Chylomicrons Regulate Lacteal Permeability and Intestinal Lipid Absorption

Abstract

Background:

Lymphatic vessels are responsible for tissue drainage, and their malfunction is associated with chronic diseases. Lymph uptake occurs via specialized open cell-cell junctions between capillary lymphatic endothelial cells (LECs), while closed junctions in collecting LECs prevent lymph leakage. LEC junctions are known to dynamically remodel in development and disease, but how lymphatic permeability is regulated remains poorly understood.

Methods:

We employed various genetically engineered mouse models in combination with cellular, biochemical, and molecular biology approaches to elucidate the signaling pathways regulating junction morphology and function in lymphatic capillaries.

Results:

By studying the permeability of intestinal lacteal capillaries to lipoprotein particles known as chylomicrons, we show that Rho-associated kinase (ROCK)-dependent cytoskeletal contractility is a fundamental mechanism of LEC permeability regulation. We show that chylomicron-derived lipids trigger neonatal lacteal junction opening via ROCK-dependent contraction of junction-anchored stress fibers. LEC-specific ROCK deletion abolished junction opening and plasma lipid uptake. Chylomicrons additionally inhibited VEGF-A signaling. We show that VEGF-A antagonizes LEC junction opening via VEGFR2 and VEGFR3-dependent PI3K/AKT activation of the small GTPase RAC1, thereby restricting RhoA/ROCK-mediated cytoskeleton contraction.

Conclusions:

Our results reveal that antagonistic inputs into ROCK-dependent cytoskeleton contractions regulate the interconversion of lymphatic junctions in the intestine and in other tissues, providing a tunable mechanism to control the lymphatic barrier.

Graphical Abstract

INTRODUCTION

Lymphatic vessels control fluid homeostasis and immune cell trafficking in most tissues of the body, and lipid absorption in the small intestine^1–3. They form a unidirectional circulatory system with blind ended capillaries that drain back via collecting vessels and the thoracic duct into the venous circulation. To execute their functions, lymphatic endothelial cells (LECs) lining lymphatic vessels develop two types of cell-cell junctions. Capillary LECs are tied together by discontinuous button-like junctions⁴. Fluids, lipids, and immune cells enter via openings between the buttons, without disrupting junctional integrity. By contrast, collector LECs have continuous, zipper-like junctions that prevent fluid leak and ensure lymph transport towards the thoracic duct, aided by smooth muscle cells (SMC) and intraluminal valves⁵.

Both junction types contain the same proteins and can transform into each other via intermediate forms⁶. A few prior studies show that button junctions appear postnatally and revert back to zippers during inflammation, indicating dynamic remodeling^4,6. Infection-induced dermal capillary zippering inhibits dissemination of vaccinia virus⁷, identifying lymphatic junctions as critical regulators of infection resistance. Junction zippering of intestinal lymphatic capillaries termed lacteals inhibits chylomicron uptake and confers resistance to diet-induced obesity⁸, demonstrating effects of LEC junction status on whole-body metabolism. Targeting junction remodeling appears as a promising approach to control lymphatic function, but mechanisms regulating this process remain largely unknown.

We have here investigated LEC junction remodeling in the context of lipid uptake in neonatal intestine. Upon ingestion of breastmilk after birth, the newborn mammalian gastrointestinal (GI) tract becomes a lipid-rich environment that supplies energy and essential signaling constituents⁹. Milk-derived lipids are absorbed by enterocytes and repackaged into chylomicrons, large lipoprotein particles containing triglycerides, cholesterol and apolipoprotein B48 (ApoB48). Newly synthesized chylomicrons exit the abluminal side of enterocytes and are taken up by lacteals¹⁰. Lacteal growth inhibition via VEGF-C/VEGFR3 inactivation, loss of lacteal DLL4/Notch activity, or loss of calcitonin receptor–like receptor (Calcrl), resulted in reduced chylomicron absorption^11–16. Impaired lacteal extension into the villus, because of abnormal SMC coverage, shifted lipid absorption from the lymphatic system to the portal circulation, leading to hepatic lipidosis¹⁷. This indicates that the lacteals are essential for chylomicron uptake and transport.

Herein, we show that newborn lacteal junctions open in response to ingested lipids that trigger their own absorption by activating ROCK-dependent cytoskeletal pull. Chylomicrons also inhibit intestinal VEGF-A signaling by upregulating the VEGF-A decoy receptor VEGFR1 in blood vascular endothelial cells (BECs) via ApoB48^8,18. We show that VEGF-A antagonizes junction opening via VEGFR2/VEGFR3-dependent PI3K activation of the small GTPase RAC1, which inhibits RhoA/ROCK-mediated cytoskeleton contraction. Our findings thus reveal that ROCK-dependent cytoskeleton remodeling controls the permeability of lymphatic junctions in response to environmental cues.

METHODS

DATA AVAILABILITY

The RNAseq data from HUVECs and HDLECs are available in the NCBI Gene Expression Omnibus (GEO) database (accession numbers GSE209855 and GSE212743). All data are included in the manuscript and the supplementary materials. Additional information can be obtained from the corresponding authors upon reasonable request.

Mice

Vegfr2^{Y949F 19} and Akt1^{−/− 20} mice were described previously. Vegfr2^{flox/flox 21}, Vegfr3^flox/flox22, PLCγ1^{flox/flox 23}, Erk1^−/− (JAX Strain #019113), Erk2^flox/flox (JAX Strain #019112), Rock1^{flox/flox 24} and Rock2^{flox/flox 25,26} and iMB-Vegfr2 mice²⁷ mice were intercrossed with Prox1CreER^T2 mice²⁸. All mice were on a C57BL/6J background and housed in pathogen free animal facilities with a 12 hour (hr)/12 hr light-dark cycle. Mice of both sexes were used. Animal experiments were approved by the Institutional Animal Care and Use Committees (IACUCs) of Yale University and Zhongshan Ophthalmic Center, Sun Yat-sen University.

To induce postnatal gene recombination, we injected mice intraperitoneally (IP) or intragastrically (IG) with tamoxifen (TAM, Sigma; #T5648; 20mg/ml in corn oil). Postnatal mice received 75μg/g TAM daily for three days between P0-P2, P5-P7 or P16-P18 (IP or IG). Adult mice received 2mg TAM on days 1, 2, 3, 5 and 7, starting at age week 5 (IP). All Rock1/2iLKO mice were analyzed within two weeks after the first TAM injection. TAM-injected Cre negative littermates were used as controls.

Statistics

For statistical analysis we used Prism 9-GraphPad software. For comparisons of mouse phenotypes, multiple measurements per mouse were averaged prior to statistical analysis. We analyzed at least 4 mice per condition and mouse numbers per experiment are indicated in Figure legends. Animals were included based on genotypes, proper age, and sex. In some experiments, mice were randomly divided into groups to receive different treatments. For western blot quantifications, protein expression was normalized to endogenous control expression, to correct for loading variation between samples. All timepoints/conditions were then normalized to the average value of baseline/control measurements. Changes were compared using the appropriate statistical test, as indicated in Figure legends for each experiment.

For two group comparisons data were subjected to the non-parametric Mann-Whitney U test. For multiple group comparisons data were analyzed either by One-Way ANOVA (data with equal variances) followed by the Tukey post-hoc test, or by Welch’s ANOVA (data with unequal variances) followed by the Dunnett T3 post-hoc test. For ANOVA, normal distribution was determined using the Shapiro-Wilk test. p < 0.05 was considered statistically significant. Data are expressed as mean ± SEM (standard error of the mean) along with scatter plots of individual data points. Each dot represents one mouse or one experiment, as described in the Figure legends. No outliers were excluded.

RESULTS

Chylomicron-induced cytoskeletal contraction opens lacteal junctions via ROCK

To assess the status of perinatal lacteal junctions, intestinal villi in mouse jejunum were stained for the junctional marker VE-Cadherin (Fig. 1A) and the junctions were categorized as “buttons” or “zippers” based on their morphology (Fig. 1B, see Methods). Lacteals from non-fed prenatal embryonic day (E) 20.5 mice displayed predominantly (>85%) continuous, zipper-like junctions (Fig. 1A–C). By contrast, milk-fed postnatal day (P) 0.5 mice exhibited significantly fewer (<50%) zipper junctions and more discontinuous, button-like junctions (Fig. 1A–C). As the appearance of discontinuous, presumably open LEC junctions, coincided with lacteal exposure to chylomicrons, we asked if the chylomicrons affected junction opening. In vitro, confluent LECs cultured in complete media displayed continuous, straight VE-Cadherin⁺ adherens junctions, likely due to the presence of VEGF-A in fetal bovine serum⁸. Upon chylomicron exposure, the junctions transformed into discontinuous, parallel linear segments of VE-cadherin, connected to radial actin stress fibers via focal vinculin-positive patches (Fig. 1D). Treatment with arachidonic acid (AA), the most common long-chain polyunsaturated fatty acid in human milk²⁹, mimicked the effects of chylomicrons on LECs (Fig. 1D). The effect was observed at 30 minutes (min) and disappeared at 120 min (Fig. 1D). The appearance of junction-anchored actin stress fibers suggested increased cytoskeletal contraction, as attested by increased phosphorylation of myosin light chain 2 (MLC2) in response to chylomicrons or AA treatment, compared to untreated cells (Fig. 1E–H). In vivo, wildtype lacteal VE-Cadherin⁺ discontinuous junctions colocalized with the tension sensor vinculin (Fig. 1I), leading us to hypothesize that chylomicron-derived lipids promote junction opening by triggering cytoskeletal tension changes in lacteal LECs.

Figure 1. Chylomicrons and lipids promote lacteal junction opening via cytoskeletal contractions:

(A) VE-Cadherin and LYVE1 staining of jejunum lacteals of C-sectioned E20.5 wildtype mouse embryos and breastfed postnatal (P) 0.5 neonatal pups.

(B) Example of junction morphology quantification from Figure 1A. The length of zipper and button-like junctions is annotated in green and red respectively and measured using Fiji software (see online Methods).

(C) Quantification of % zipper-like junctions out of total junction length in lacteals shown in A. Each symbol represents one mouse. N=6 mice per group. Error bars, SEM. Mann-Whitney U test.

(D) VE-Cadherin, F-actin and Vinculin staining of confluent HDLECs in complete media with or without treatment with 50μg/mL chylomicrons (CM) or 30μM arachidonic acid (AA) at the indicated timepoints. Arrowheads indicate colocalization of VE-Cadherin, F-actin and Vinculin staining at junctional sites.

(E-H) Western blot and quantification of pMLC2 (Thr18/Ser19), normalized to GAPDH, in HDLECs treated with 50μg/ml CM (E, G) or 30μM AA (F, H) at indicated timepoints. Data are expressed as fold changes compared to the average value of time (T)=0’. N=4 experiments. Error bars, SEM. ns, not significant, One-Way ANOVA with post-hoc Tukey test.

(I) VE-Cadherin, Vinculin and LYVE1 staining of jejunum lacteals of wildtype mice. Arrowheads point to VE-Cadherin and Vinculin colocalization on button-like structures.

Lipids such as AA can activate ROCK, the major regulator of cytoskeletal contraction, in a RhoA-dependent or independent manner^30–34, and pharmacological ROCK inhibition zippered lacteal junctions and prevented chylomicron uptake in postnatal mice⁸. This suggested that lipid-activated ROCK signaling could regulate lacteal permeability. In cultured LECs, serum starvation induced the formation of discontinuous VE-Cadherin⁺ adherens junctions. Silencing of both mammalian ROCK isoforms, ROCK1 and 2³⁵, reverted these junctions back into straighter ones, confirming a LEC-autonomous role for ROCK in zippering lymphatic junctions (Fig. S1A). To determine this effect in vivo, we generated LEC-specific ROCK conditional mutant mice by intercrossing Rock1^flox/flox, Rock2^flox/flox and Prox1CreER^T2 mice (Prox1CreER^T2;Rock1^flox/flox;Rock2^flox/flox: Rock1/2iLKO)^24–26,28. Tamoxifen (TAM) administration to these mice reduced ROCK1 and 2 protein expression significantly in initial lymphatics, confirming LEC-specific gene deletion (Fig. S1B–C). Analysis of P6 Rock1/2iLKO mice that had been administered TAM at P0-P2 revealed straight zipper-like junctions in jejunal lacteals, whereas Cre-negative littermate controls (Rock1^flox/flox;Rock2^flox/flox: Rock1/2f/f) had mainly discontinuous button-like junctions (Fig. 2A–C). Lacteal length or width were not significantly altered in the mutants (Fig. 2D). Functionally, the P6 Rock1/2iLKO pups had significantly less chyle in their mesenteric lymphatic vessels than their littermate controls (Fig. 2E–F), attesting to functional defects in chyle transport. To investigate whether ROCK1/2 loss of function (LOF) affects button maintenance, we analyzed P19–21 Rock1/2iLKO mice after TAM administration between P5-P7 (Fig. 2G). These mice exhibited more lacteal zipper junctions than control littermates (Fig. 2H–I), plus increased length and unaltered width of lacteals (Fig. 2J), and lethality by P35 due to as yet undetermined causes (Fig. 2K). The increase of blood triglyceride levels was also suppressed in adult mutant mice that received olive oil gavage 10 days after completion of TAM administration (Fig. 2L–M). Taken together, our results show that ROCK activity in LECs is required for the formation and maintenance of lacteal button junctions, and for fat uptake in the GI tract.

Figure 2. ROCK is required for lacteal junction opening and lipid uptake:

(A) Experimental strategy to induce Rock1/2 deletion in B-F. Arrowheads represent 3X75μg/g TAM injections, followed by analysis at P6.

(B) VE-Cadherin and LYVE1 staining of jejunum lacteals from P6 Rock1/2f/f and Rock1/2iLKO mice.

(C-D) Quantification of % lacteal zipper-like junctions normalized to total junction length, and lacteal length and width of P6 Rock1/2f/f and Rock1/2iLKO mice. Each dot represents one mouse. N=4 or 5 mice per group. Error bars, SEM. ns, not significant. Mann-Whitney U test.

(E-F) Images of mesenteries and quantification of chyle-filled lymphatic collectors of P6 Rock1/2f/f and Rock1/2iLKO mice. Each dot represents one mouse. N=4 or 5 mice per group. Mann-Whitney U test.

(G) Experimental strategy to induce Rock1/2 deletion in H-J. Arrowheads represent 3X75μg/g TAM injections, followed by analysis at P19–21.

(H) VE-Cadherin and LYVE1 staining of jejunum lacteals from P19–21 Rock1/2f/f and Rock1/2iLKO mice.

(I-J) Quantification of % lacteal zipper-like junctions, normalized to total junction length, and lacteal length and width from P19–21 Rock1/2f/f and Rock1/2iLKO mice. Each dot represents one mouse. N=4 or 5 mice per group. Error bars, SEM. ns, not significant. Mann-Whitney U test.

(K) Survival curves of Rock1/2f/f and Rock1/2iLKO mice after 3X75μg/g TAM injections at P5, P6 and P7. N=7 or 14 mice per group.

(L-M) Timeline of TAM administration (2mg/day for 5 days) and plasma triglyceride measurements of adult Rock1/2f/f and Rock1/2iLKO mice. Mice were fasted for 6 hours, gavaged with 200μl of olive oil, and plasma triglycerides were measured. N=9 or 15 mice per group. Error bars, SEM. Mann-Whitney U test was performed to compare Rock1/2f/f and Rock1/2iLKO groups at the same timepoints.

LEC VEGFR2 activation zippers lymphatic capillary junctions

An increase of VEGF-A bioavailability or exogenous VEGF-A administration can zipper initial lymphatics, and chylomicrons are known to suppress VEGF-A signaling by upregulating the VEGF-A decoy receptor VEGFR1 in BECs via ApoB48^8,18. In line with this, VEGFR2 phosphorylation was significantly reduced in jejunum from milk-fed P0.5 mice, compared to non-fed prenatal E20.5 mice (Fig. 3A–C). To determine if this process required LEC-autonomous VEGFR2 signaling, we produced LEC-specific inducible Vegfr2 knock-out mice (Prox1CreER^T2;Vegfr2^flox/flox: Vegfr2iLKO)²¹. TAM administration between P16–18 abolished VEGFR2 protein expression in Vegfr2iLKO lacteals (Fig. S2A–C). Loss of Vegfr2 from the lymphatics did not affect lacteal junction morphology, or lacteal length or width at baseline conditions (Fig. 3D–F). However, intravenous (IV) VEGF-A injection increased the ratio of lacteal zipper- vs. button-like junctions in control mice, but not in Vegfr2iLKO mice (Fig. 3E, 3G).

Figure 3. VEGFR2 activation zippers lymphatic capillary junctions:

(A-C) Western blot and quantifications of phospho-VEGFR2 (pY949 and pY1173), normalized to total VEGFR2, in jejunum tissue lysates from non-fed prenatal E20.5 and milk-fed P0.5 mice. Data are expressed as fold changes compared to the average value of E20.5. Each dot represents one mouse. N=8 mice per group. Error bars, SEM. Mann-Whitney U test.

(D) Experimental strategy to induce Vegfr2 deletion or overexpression, shown in E-K. Arrowheads represent 3X75μg/g TAM injections, followed by analysis at P21.

(E) VE-Cadherin and LYVE1 staining of jejunum lacteals from P21 Vegfr2f/f and Vegfr2iLKO mice 30 min after IV injection of PBS or 250ng/g VEGF-A.

(F) Quantifications of jejunum lacteal length and width in P21 Vegfr2f/f and Vegfr2iLKO mice. Error bars, SEM. Each dot represents one mouse. N=4 mice per group. ns, not significant. Mann-Whitney U test.

(G) Quantification of % zipper-like junctions out of total junction length in lacteals shown in E. Each symbol represents one mouse. N=4 or 5 mice per group. Error bars, SEM. One-Way ANOVA with post-hoc Tukey test.

(H) pERK and LYVE1 staining of VEGFR2^CA-tdTomato (+) and (−) ear dermal lymphatic capillaries of P21 iMB-Vegfr2;Prox1CreER^T2 mice. CA: constitutive active. LV: lymphatic vessel.

(I) Quantification of pERK intensity in VEGFR2^CA-tdTomato (+) and (−) lymphatic vessels outlined in H. Each dot represents one mouse. N=4 mice per group. Error bars, SEM. Mann-Whitney U test.

(J) VE-Cadherin, tdTomato and LYVE1 staining of ear dermal lymphatic capillaries from P21 iMB-Vegfr2;Prox1CreER^T2 mice.

(K) Quantification of zipper-like lymphatic junctions out of total lymphatic junction length in VEGFR2^CA-tdTomato (+) and (−) lymphatic vessels in J. Each dot represents one mouse. N=6 mice per group. Error bars, SEM. Mann-Whitney U test.

In a complementary VEGFR2 gain-of-function (GOF) approach, we investigated the effects of VEGFR2 activation on junction morphology by using Prox1CreER^T2-iMb-Vegfr2. After TAM treatment, these mice express either a tdTomato-tagged VEGFR2 with constitutive ligand-independent kinase activity, or a YFP fused-dominant negative VEGFR2, or a far-red fluorescent mKate2²⁷ in stochastically recombined LECs. Since tissue preparation diminished strong tdTomato signals in the lacteals, we analyzed initial lymphatics in the dermis. As predicted, TAM administration led to LEC-specific expression of red tdTomato fluorescence in a subset of dermal initial lymphatics, along with significantly increased ERK phosphorylation, confirming VEGFR2 overactivation (Fig. 3H–I). We found more zippered junctions in tdTomato-positive than tdTomato-negative LECs (Fig. 3J–K). Taken together, our data show that LEC VEGFR2 signaling is dispensable for junction opening at baseline, whereas its activation is sufficient to zipper junctions.

VEGFR2 Y949 phosphorylation is dispensable for modulation of lacteal junctions

Next, we addressed the mechanisms of VEGF-A/VEGFR2 induced LEC junction zippering. VEGF-A-induced permeability in BECs is regulated by phosphorylation of VEGFR2 at tyrosine (Y) 949, which mediates the binding of VEGFR2 to the Src homology 2-domain of T cell–specific adaptor (TSAd), leading to c-Src activation, and increased VE-Cadherin phosphorylation. This mechanism disrupts adherens junctions and increases vascular permeability^19,36. We tested whether Y949 is involved in LEC junction remodeling by analyzing junction morphology in Vegfr2^Y949F/Y949F mice that carry a Y to F (Phenylalanine) substitution at position 949, which blocks VEGF-A induced permeability¹⁹. Blood vascular leakage and extravasation of fluorescent dextran after intravenous injection of VEGF-A was prevented in the Vegfr2^Y949F/Y949F mutant intestinal vasculature (Fig. 4A–B), confirming BEC permeability regulation via Y949. However, LEC junctions were not affected, as VEGF-A injection increased the number of zipper-like junctions to the same extent in control and Vegfr2^Y949F/Y949F mice (Fig. 4C–D). Thus, Y949 signaling does not regulate LEC junction morphology downstream of VEGFR2.

Figure 4. VEGFR2 Y949 regulates permeability in blood but not in lymphatic vessels:

(A) Dextran leakage of jejunum villus blood vessels in P11-P13 Vegfr2^Y949F/Y949F mice and wildtype littermates, 10 min after retro-orbital injection of fluorescent labeled IsoB4/dextran (70kDa) with 250ng/g VEGF-A or PBS.

(B) Quantification of dextran leakage shown in A. Each symbol represents one mouse. N=4 mice per group. Error bars, SEM. ns, not significant. One-Way ANOVA with post-hoc Tukey test.

(C) VE-Cadherin staining of jejunum lacteals from P21 wildtype and Vegfr2^Y949F/Y949F mice 30 min after IV injection of PBS or 250ng/g VEGF-A.

(D) Quantification of % zipper-like junctions out of total junction length in lacteals shown in C. Each symbol represents one mouse. N=4 or 6 mice per group. Error bars, SEM. ns, not significant. One-Way ANOVA with post-hoc Tukey test.

(E) RNAseq analysis of SH2D2A (encoding TSAd), C-SRC and LYN differential expression in cultured HDLECs vs. HUVECs. Shown are transcript per million (TPM) values of the indicated genes. N=4 experiments. Error bars, SEM. Benjamini-Hochberg test.

(F) Western blot analysis for the indicated proteins in HDLECs vs. HUVECs treated with 50ng/ml VEGF-A for 0, 5, 15 and 30 min.

(G-L) Quantifications of the protein blots, normalized to GAPDH, shown in F. Data are expressed as fold changes compared to the average value of T=0’ of HUVEC group. Error bars, SEM. N=5 experiments. ns, not significant. Mann-Whitney U test was performed to compare HDLECs vs. HUVECs at the same timepoints.

Bulk RNA sequencing revealed very low SH2D2A (encoding TSAd) expression in HDLECs vs. HUVECs (Fig. 4E). Interestingly, we observed a significant decrease in C-SRC transcripts and protein in HDLECs vs. HUVECs, while another Src family kinase (SFK) LYN was significantly increased (Fig. 4E–H). The YES1 and FYN SFK protein levels were not changed despite mild to moderate increase of their transcripts in HDLECs vs. HUVECs (Fig. S3A–D). In BECs in vivo, phosphorylation of VE-Cadherin at Y658 and Y685, mediates junction opening downstream of SFKs^37,38. Signaling studies revealed that the VEGF-A-induced phosphorylation of VE-Cadherin at Y685 and Y658 was profoundly reduced in HDLECs vs. HUVECs, despite robust phosphorylation of VEGFR2 Y951 (the human Y949 homologue) (Fig. 4F–L). Thus, intrinsic differences in TSAd/SFKs expression between BECs and LECs could account for the differential responses to VEGF-A activation and VE-Cadherin phosphorylation in BECs vs. LECs, and explain why LECs do not remodel their junctions in response to Y949 phosphorylation.

VEGFR3 is required for VEGF-A induced lacteal junction zippering

VEGFR3 is the main receptor involved in lymphangiogenesis and lymphatic maintenance³⁹. VEGFR3 binds VEGF-C and VEGF-D, but not VEGF-A⁴⁰. To determine if VEGFR3 regulates LEC junctions, we generated lymphatic specific Vegfr3 mutants by crossbreeding Vegfr3 floxed mice²² with Prox1CreER^T2 mice (Prox1CreER^T2;Vegfr3^flox/flox: Vegfr3iLKO). Deletion of Vegfr3 dramatically decreased lacteal length (Fig. S4A–B), in line with previous reports⁴¹. We therefore opted for a short deletion window that did not affect the ratio of lacteal zipper/button junctions at baseline conditions or lacteal length, although it minimally decreased lacteal width (Fig. 5A–E). Intriguingly, while VEGF-A injection increased the number of lacteal zipper-like junctions in control mice, Vegf3iLKO mice maintained mostly buttoned junctions (Fig. 5B–C). As VEGFR3 does not bind VEGF-A directly, we hypothesized that VEGFR3 either modulates VEGFR2 expression levels, or VEGFR2 signaling output via the formation of VEGFR2/VEGFR3 heterodimers^42,43. We did not observe any changes in VEGFR2 protein levels in lacteals of Vegfr3iLKO mice (Fig. 5F–G), or any changes to VEGFR2 mRNA levels in cultured LECs after silencing VEGFR3 (Fig. 5H). On the other hand, silencing of VEGFR3 in HDLECs reduced AKT phosphorylation in response to VEGF-A, while phosphorylation of Y951, Y1175, Y1214 and ERK remained unchanged (Fig. 5I–J). These results suggested that VEGFR3 contributes to LEC junction zippering by reinforcing PI3K/AKT signaling downstream of VEGF-A/VEGFR2, presumably via the formation of VEGFR2/VEGFR3 heterodimers.

Figure 5. VEGFR3 is required for VEGF-A/VEGFR2 induced lymphatic junction zippering:

(A) Timeline of experiments shown in B-G. Arrowheads represent 3X75 μg/g TAM injections and analysis at P21.

(B) VE-Cadherin staining of jejunum lacteals from P21 Vegfr3f/f and Vegfr3iLKO littermates, 30 min after PBS or 250ng/g VEGF-A IV injection.

(C) Quantification of % zipper-like junctions out of total junction length in lacteals shown in B. Each symbol represents one mouse. N=4 or 6 mice per group. Error bars, SEM. One-Way ANOVA with post-hoc Tukey test.

(D) LYVE1 staining of jejunum lacteals of P21 Vegfr3f/f and Vegfr3iLKO mice.

(E) Quantifications of lacteal width and length from D. Each dot represents one mouse. N=4 or 5 mice per group. Error bars, SEM. ns, not significant. Mann-Whitney U test.

(F) VEGFR2 and LYVE1 staining of jejunum villi from P21 Vegfr3f/f and Vegfr3iLKO mice. LV: lymphatic vessel. BV: blood vessel.

(G) Quantification of VEGFR2 intensity in the lacteals outlined in F. Data were normalized to the average VEGFR2 intensity of Vegfr3f/f lacteals. Each dot represents one mouse. N=4 or 5 mice per group. Error bars, SEM. ns, not significant. Mann-Whitney U test.

(H) qPCR analysis of VEGFR2 and VEGFR3 expression in HDLECs 2 days after transfection with control (CTRL) or VEGFR3 siRNA. N=4 experiments. Error bars, SEM. ns, not significant. Mann-Whitney U test.

(I-J) Western blots and quantifications of VEGFR2, ERK and AKT phosphorylation, normalized to the indicated loading controls, in HDLECs transfected with siCTRL or siVEGFR3 and treated with 50ng/ml VEGF-A for 0, 5, 15 and 30 min. Data are expressed as fold changes compared to the average value of T=0’ of siCTRL group. N=5 experiments. Error bars, SEM. ns, not significant. Mann-Whitney U test was performed to compare siCTRL vs. siVEGFR3-treated groups at the same timepoints.

VEGF-A zippers lacteal junctions via PI3K/AKT signaling

To determine PI3K/AKT signaling requirement for lacteal junction zippering in vivo, we analyzed Akt1 mouse mutants (Akt1^−/−: Akt1KO)²⁰. Global loss of Akt1 did not affect junction morphology or lacteal length or width (Fig. 6A–D). Intravenous VEGF-A injection increased the number of zipper-like junctions in control mice, whereas Akt1KO mice were resistant to VEGF-A induced zippering (Fig. 6A–B). Similar results were observed when we treated the mice with the PI3K inhibitor Wortmannin that blocks VEGF-A induced phosphorylation of AKT (Fig. 6A–B, S5A–B). In parallel, we investigated the role of PLCγ-ERK pathway downstream of VEGFR2 Y1175 phosphorylation⁴⁴. We analyzed lymphatic specific mutants of Plcγ1 (Prox1CreER^T2;Plcγ1^flox/flox: Plcγ1iLKO)²³, as well as compound mutants of Erk1/2 (Erk1^-/;Prox1CreER^T2;Erk2^flox/flox: Erk1KO;Erk2iLKO). Plcγ1iLKO and Erk1KO;Erk2iLKO mice showed unaltered junction morphology at baseline, and zippered their junctions in response to intravenous VEGF-A (Fig. 6E–G), demonstrating that PI3K/AKT but not PLCγ1/ERK1/2 signaling mediates lacteal junction zippering in response to VEGF-A.

Figure 6. PI3K/AKT signaling mediates VEGF-A-induced lymphatic junction zippering:

(A) VE-Cadherin staining of jejunum lacteals from P21 wildtype, Akt1KO and Wortmannin-treated mice, 30 min after IV administration of PBS or 250ng/g VEGF-A.

(B) Quantification of % zipper-like junctions out of total junction length in lacteals shown in A. Each symbol represents one mouse. N=4 or 6 mice per group. Error bars, SEM. ns, not significant. One-Way ANOVA with post-hoc Tukey test.

(C) LYVE-1 staining and (D) quantifications of length and width of jejunum lacteals from P21 wildtype and Akt1KO mice. Each dot represents one mouse. N=4 mice per group. Error bars, SEM. ns, not significant. Mann-Whitney U test.

(E) Experimental strategy to induce Plcγ or Erk2 deletion in F-G. Arrowheads indicate 3X75μg/g TAM injections and analysis at P21.

(F) VE-Cadherin staining of jejunum lacteals from P21 PlcγiLKO and Erk1KO;Erk2iLKO mice 30 min after IV injection of PBS or 250ng/g VEGF-A.

(G) Quantification of % zipper-like junctions out of total junction length in lacteals shown in F. Each symbol represents one mouse. N=4 to 5 mice per group. Error bars, SEM. Mann-Whitney U test.

VEGF-A/VEGFR2 signaling suppresses RhoA in LECs via RAC1 activation

Next, we investigated how VEGF-A/VEGFR2 interacts with RhoA/ROCK during LEC junction remodeling. Stimulation of serum-starved HDLEC monolayers with VEGF-A for 30 min transformed the discontinuous adherens junctions into straighter zipper junctions and reduced cytoskeletal stress fibers^8,45 (Fig. 7A). Consistently, phosphorylated MLC2 was significantly decreased in HDLECs 30 min after VEGF-A stimulation (Fig. 7B). In contrast, serum-starved HUVECs showed increased stress fiber formation and no difference in MLC2 phosphorylation 30 min after VEGF-A stimulation (Fig. 7A–B). In line with the reduced pMLC2 levels in HDLECs, we found a decrease of activated, GTP-bound RhoA following VEGF-A treatment (Fig. 7C). This depended on VEGFR2, since silencing of VEGFR2 sustained RhoA activity (Fig. 7C). VEGF-A stimulation of HDLECs induced GTP-bound RAC1, another small GTPase of the Rho family, in a VEGFR2-dependent manner (Fig. 7D). As RAC1 inactivates RhoA/ROCK, this could explain the reduced RhoA activity and actin depolymerization observed in VEGF-A stimulated HDLECs⁴⁶. VEGFR2-mediated RAC1 activation depended on PI3K, as pretreatment with the PI3K inhibitor Wortmannin completely abolished the elevation in RAC1-GTP levels after VEGF-A stimulation (Fig. 7E). Thus, VEGF-A/VEGFR2 signaling in LECs reduces stress fiber formation by inhibiting RhoA/ROCK/MLC2 via activation of RAC1, in a PI3K-dependent manner.

Figure 7. VEGF-A/VEGFR2 inhibits stress fiber formation in HDLECs:

(A) VE-Cadherin and phalloidin staining of confluent HUVECs and HDLECs starved in 0.2% serum for 6 hours and treated with 50ng/ml VEGF-A or PBS for 30 min.

(B) Western blot and quantification of pMLC2 (Thr18/Ser19), normalized to TUBULIN, in confluent HUVECs and HDLECs starved for 6 hours and stimulated with 50ng/ml VEGF-A for 0, 5, 15 and 30 min. Data are expressed as fold changes compared to the average value of T=0’ of each cell group. N=6 experiments. Error bars, SEM. ns, not significant. One-Way ANOVA with post-hoc Tukey test (for HDLEC), or Welch’s ANOVA with post-hoc Dunnett T3 test (for HUVEC).

(C-D) RhoA and RAC1 pulldown assays in HDLECs transfected with CTRL or VEGFR2 siRNA and treated with 50ng/ml VEGF-A for 30 min. Top panel: Western blots of GTP-bound RhoA/RAC1 from the pulldown assay and total RhoA/RAC1 from total cell lysates. Bottom panel: Quantifications of the ratios of GTP bound RhoA/RAC1 to total RhoA/RAC1. Data are expressed as fold changes compared to the average value of T=0’ of each siRNA-treated group. N=4–5 experiments. Error bars, SEM. ns, not significant. Mann-Whitney U test.

(E) RAC1 pulldown assay in HDLECs after 5 hours Wortmannin (1μM) or DMSO treatment and 30 min VEGF-A (50ng/ml) stimulation. Shown are western blots (top) and quantification (bottom) of the ratios of GTP bound RAC1 to total RAC1 from cell lysates. Data are expressed as fold changes compared to the average value of T=0’ of each treatment group. N=5 experiments. Error bars, SEM. ns not significant. Mann-Whitney U test.

ROCK regulates lymphatic junction morphology in dermal initial lymphatics

To address whether VEGF-A and ROCK signaling control junction remodeling outside the GI tract, we analyzed junction morphology in dermal lymphatic capillaries. Intradermal injection of VEGF-A induced junction zippering in the dermal initial lymphatics of the ear⁸ (Fig. 8A–B), while other vascular permeability inducers, such as histamine or bradykinin, did not induce junction zippering (Fig. 8A–B). ROCK signaling was also required for the maintenance of button junctions in dermal initial lymphatics, as Rock1/2iLKO mutants displayed predominantly zipper junctions at baseline (Fig. 8C–E). To examine whether button morphology is necessary for fluid uptake, we injected Evans Blue into the footpads of adult mice. The dye was quickly absorbed by lymphatic capillaries and transported to the draining lymph node in control mice, while VEGF-A injection resulted in significantly reduced dye uptake (Fig. 8F–G). We also observed a significant reduction in the intensity of Evans Blue-filled lymphatics connected to popliteal lymph nodes in the Rock1/2iLKOs (Fig. 8H–J). These results show that ROCK regulates lymphatic junction transformation and solute uptake function across different lymphatic capillary beds.

Figure 8. VEGF-A and ROCK signaling regulate dermal initial lymphatic drainage:

(A) VE-Cadherin staining of ear dermal initial lymphatics from P21 wildtype mice 30 min after intradermal injection of VEGF-A (50ng in 20μl PBS), histamine (500ng in 20μl PBS), bradykinin (20μg in 20μl PBS) or PBS alone.

(B) Quantification of % zipper-like LEC junctions out of total lymphatic junction length in dermal initial lymphatics in A. Each symbol represents one mouse. N=6 or 7 mice per group. Error bars, SEM. ns, not significant. One-Way ANOVA with post-hoc Tukey test.

(C) Experiment strategy to induce Rock1/2 deletion shown in panels D-E. Arrowheads represent 3X75μg/g TAM injections and analysis at P19–21.

(D-E) VE-Cadherin staining and quantification of % zipper-like junctions out of total junction length in ear dermal initial lymphatics from P19–21 Rock1/2f/f and Rock1/2iLKO mice. Each symbol represents one mouse. N=4 mice per group. Error bars, SEM. Mann-Whitney U test.

(F) Evans Blue absorption in lower leg lymphatic vessels of adult wildtype mice at the indicated time points after footpad injection of VEGF-A (20ng in 5μl PBS) or PBS alone, and 5μl of 1.5% Evans Blue solution.

(G) Quantification of dye intensity in Evans Blue-filled lymphatics in F. N=7 or 9 mice per group. Error bars, SEM. Mann-Whitney U test was performed to compare PBS and VEGF-A-treated groups at the same timepoints.

(H) Timeline of TAM administration (2mg per day for 5 days) of adult Rock1/2f/f and Rock1/2iLKO mice. All mice were analyzed 10 days after the last TAM injection.

(I) Evans Blue uptake in lower leg lymphatic vessels of Rock1/2f/f and Rock1/2iLKO mice at the indicated time points after intradermal footpad injection of 9μl 1.5% Evans Blue solution.

(J) Quantification of dye intensity in Evans Blue-filled lymphatics in I. N=7–8 mice per group. Error bars, SEM. ns, not significant. Mann-Whitney U test was performed to compare Rock1/2f/f and Rock1/2iLKO mouse groups at the same timepoints.

(K) Summary model of LEC junction remodeling in intestinal lacteals. i. High lipid uptake: Chylomicrons derived from ingested lipids activate ROCK (1), which phosphorylates MLC2 and induces actomyosin stress fiber assembly and junction opening, thereby enabling lacteal chylomicron uptake. VEGFR2 is inhibited by chylomicron upregulation of VEGFR1 and NRP1 in BECs, which reduce VEGF-A signals to LEC VEGFR2 (2). ii. Low lipid uptake: VEGF-A signals via VEGFR2/VEGFR3 heterodimers to antagonize ROCK activation and junction opening. VEGFR2/VEGFR3 activate RAC1 via PI3K/AKT, thereby reducing RhoA and ROCK-dependent MLC2 phosphorylation, resulting in actin stress fiber relaxation, cortical actin formation and junction zippering, which prevents chylomicron uptake by lacteals.

DISCUSSION

Our work identifies a two-pronged homeostatic mechanism that triggers lacteal junction opening in intestinal lacteals. Chylomicrons rapidly activate ROCK in LECs, which initiates the formation of junction-anchored stress fibers and generates the tension required to pull junctions apart (Fig. 8Ki). A similar mechanism has been described in BECs during the formation of focal adherens junctions (FAJ), which contain vinculin and partially unzip to increase blood vascular permeability. FAJ formation requires actomyosin contractility and depends on RhoA GTPase activity⁴⁷. Thus, RhoA and ROCK are key regulators of junctional morphology in both LECs and BECs.

We speculate that lipid activation of ROCK is a driving force that triggers cytoskeletal contraction and junction opening in the intestinal lacteal microenvironment. Lipids, such as lysophospholipids^48,49, sphingolipids⁵⁰ and fatty acids^31–34 can activate guanine nucleotide exchange factors (GEFs) for RhoA through G-protein coupled receptors and p38/MAPK. Arachidonic acids can also directly activate ROCK by binding to the autoinhibitory region of ROCK⁵¹. The mechanisms of ROCK activation in response to chylomicrons remains to be determined.

Following postnatal fat ingestion, VEGFR2 signaling is inhibited in the intestine. This is likely due to the chylomicron apolipoprotein ApoB48 that could induce postnatal transcriptional upregulation of the VEGF-A decoy receptor VEGFR1 in BECs of the intestinal villi^8,18. VEGFR1 sequesters peri-lacteal VEGF-A, and together with Neuropilin 1 (NRP1) inhibits VEGFR2 signaling on LECs and junction zippering¹². Therefore, chylomicrons take dual action to ensure perinatal lymphatic junction opening and their own intestinal absorption, which in mammals coincides with milk consumption immediately after birth³¹.

VEGF-A expression is tightly regulated in the microenvironment of the intestinal villi to maintain the fenestrated blood vasculature⁵². Increased VEGF-A bioavailability leads to vascular leakage in the intestinal villus⁸, which could indirectly affect LEC junctions. Here we show that the effects of VEGF-A on LEC junctions are neither a byproduct of concurrent BEC junction opening, nor via effects on other cell types, such as mural cells⁵³. Using VEGFR2 LEC-specific LOF and GOF mice, we show that VEGF-A antagonizes junction opening in a LEC-autonomous manner. VEGFR2 LOF does not alter baseline junction morphology, in contrast to ROCK deletion, revealing antagonistic effects by VEGF-A and ROCK that are both controlled by chylomicrons.

We also show that VEGF-A-mediated inhibition of lacteal permeability is independent of Y949 but abolished by VEGFR3 deletion. Since VEGFR3 does not bind VEGF-A directly, we conclude that VEGFR3 modulates VEGFR2 signaling output via VEGFR2/VEGFR3 heterodimers. Such heterodimers are formed in cultured LECs in response to VEGF-C⁴², as well as in BECs in angiogenic sprouts in response to VEGF-A and VEGF-C⁴³. As VEGF-A-induced VEGFR2 Y951, Y1175 and Y1214 phosphorylation is unaffected by VEGFR3 knockdown, the formation of such heterodimers may affect the sites of VEGFR3 tyrosine phosphorylation^42,43.

VEGF-C-156S, a mutated form of VEGF-C that only binds VEGFR3⁵⁴, was unable to zipper the lacteal junctions of control mice⁸, and VEGFR3 inhibition with monoclonal antibodies failed to reverse zippered junctions in the inflamed mouse trachea⁶. On the other hand, VEGF-C produced by lacteal macrophages in response to microbial stimuli is required for button maintenance¹². We suggest that these effects are mediated by VEGFR2, or via VEGFR2/VEGFR3 heterodimers, since proteolytically processed VEGF-C also binds to and activates VEGFR2⁵⁵. VEGFR3 knockdown abolished AKT activation downstream of VEGF-A in HDLECs, and AKT deletion prevented VEGF-A induced lacteal junction zippering in vivo. Conversely, VEGFR2 knockdown in LECs prevents VEGF-C induced AKT phosphorylation, but not ERK activation⁵⁶. Taken together, the data are consistent with a model where VEGF-A or VEGF-C induced VEGFR2/VEGFR3 heterodimer signaling to AKT accounts for LEC permeability inhibition.

VEGF-A-mediated activation of VEGFR2 zippered lacteal junctions by suppressing RhoA via increased RAC1 activation (Fig. 8Kii). RAC1 antagonizes RhoA to remodel apical junctions as described in a variety of cellular contexts⁵⁷. In this case, RAC1 inhibits RhoA, while RhoA can in turn restrain RAC1 activity. Interestingly, in BECs, signaling via VEGFR2 leads to activation of both RAC1^58–60 and RhoA⁶¹, suggesting that the dynamics of pathway activation are different in BECs vs. LECs. PI3K/AKT signaling can activate RAC1 in many cell types^62–64, but how exactly PI3K activation downstream of VEGF-A/VEGFR2 results in activation of RAC1 in LECs remains to be determined.

ROCK deletion resulted ultimately in lethality in the Rock1/2iLKO mice. Determining the cause of death requires further investigation. Prox1CreERT² is expressed also in other cell types, such as hepatocytes, cardiomyocytes and some neuronal cells, and therefore we cannot exclude off-target effects of ROCK LOF that could underlie the increased mortality of the mutants.

Finally, we show that ROCK activity is essential for button formation also outside of the GI tract. The upstream signaling pathways activating ROCK in skin lymphatics remain to be determined. Dexamethasone activation of LEC glucocorticoid receptors, which increase RhoA-mediated contractility in various settings^65,66 promotes button formation and decreases edema in inflamed tracheal lymphatics⁶. It is therefore likely that signaling through glucocorticoid receptor or other surface receptors such as Calcrl, transforming growth factor β-receptor (TGFβR), platelet derived growth factor receptor (PDFGR) and epidermal growth factor receptor (EGFR) that also activate RhoA, may contribute to button formation in different tissues^67–69. Other environmental inputs, including enzymes such as thrombin⁷⁰, flow shear⁷¹ and mechanical forces⁷² may also be at play. Understanding the inputs into the cytoskeletal mechanism that we here uncover could allow manipulation of lymphatic permeability to control tissue drainage and immune cell trafficking in preclinical disease models.

Novelty and Significance.

What is known?

• Initial lymphatics at the intestine (termed lacteals) are responsible for absorption of chylomicrons, lipoprotein particles produced by enterocytes.

• Chylomicrons enter lacteals through open, ‘button’-like junctions between lymphatic endothelial cells.

• Closing (or ‘zippering’) of lymphatic endothelial cell junctions, prevents dietary lipid absorption in mice.

What new information does this article contribute?

• In newborn mice, chylomicrons trigger lacteal junction opening, following milk injection right after birth.

• Chylomicrons activate ROCK signaling in lymphatic endothelial cells, remodeling junctions and allowing for lipid absorption.

• VEGF-A signals via VEGFR2/VEGFR3 to antagonize junction opening by suppressing ROCK.

• ROCK suppression is mediated via PI3K-dependent activation or RAC1, which inhibits RhoA.

Lymphatic vessels are essential for intestinal lipid absorption, yet the mechanisms regulating lymphatic permeability are poorly understood. Here we show that chylomicrons, lipoprotein particles produced by enterocytes, regulate their own absorption by triggering the formation of ‘button’-like junctions in the intestinal villus, via activation of ROCK signaling and cytoskeletal remodeling. At the same time chylomicrons suppress VEGFR2 signaling, which limits ROCK activity by inhibiting RhoA. We show that these effects are mediated by VEGFR2/VEGFR3 heterodimer formation in response to VEGF-A, which activate the RhoA suppressor RAC1, via PI3K/AKT. This two-pronged homeostatic mechanism ensures fat ingestion after postnatal milk consumption. In addition, we show that ROCK-mediated cytoskeletal contractions regulate lymphatic junction morphology and permeability also outside the intestine, in the mouse dermis. This study identifies lymphatic junctions as an important regulator of lymphatic vessel function, suggesting a novel target to control lipid absorption and tissue drainage in preclinical models of disease.

Non-standard abbreviations and acronyms

LECs

lymphatic endothelial cells

BECs

blood vascular endothelial cells

ROCK

Rho-associated kinase

ApoB48

apolipoprotein B48

VEGF

vascular endothelial growth factor

VEGFR

vascular endothelial growth factor receptor

SMC

smooth muscle cell

TAM

tamoxifen

HDLECs

human dermal lymphatic endothelial cells

HUVECs

human umbilical vein endothelial cells

AA

arachidonic acid

MLC2

myosin light chain 2

SFK

Src family kinase