Multiple roles of lymphatic vessels in peripheral lymph node development

Lymphatic capillaries take up interstitial fluid, antigens, and antigen-presenting cells, and collecting vessels transport lymph to LNs. Afferent collecting lymphatics deliver lymph to the LN subcapsular sinus (SCS) through a fibrous capsule surrounding LNs. From the SCS, lymph reaches trabecular and medullary lymphatic sinuses and exits via the efferent collecting vessels. The LN and peripheral lymphatic endothelial cells (LECs) are molecularly distinct (Petrova and Koh, 2018). Such specialization is important for intranodal migration of dendritic cells (DCs; Ulvmar et al., 2014), filtration of small versus large molecular weight components (Rantakari et al., 2015), and immune tolerance (Cohen et al., 2010; Lund et al., 2012; Tewalt et al., 2012).

Mechanisms of how such a highly organized system arises during development are not fully understood. Pioneering studies have identified key roles for hematopoietic lymphoid tissue inducer (LTi) and stromal lymphoid tissue organizer (LTo) cells, proposing a model where LNs are initiated after CXCR5⁺ pre-LTi cell egress to specific locations from blood vessels in response to LTo cell-derived CXCL13 (Ansel et al., 2000; Mebius et al., 2001; Yoshida et al., 2001; Luther et al., 2003; Ohl et al., 2003; van de Pavert et al., 2009; van de Pavert and Mebius, 2010; Brendolan and Caamaño, 2012). Clustering and crosstalk of LTαβ⁺ LTi and LTβR⁺ (lymphotoxin-β receptor) LTo cells result in further signal amplification, leading to LTi cell maturation and accumulation. Analysis of Prox1^−/− embryos, which lack lymphatic vessels, revealed that most LNs are initiated in the absence of lymphatics, but their expansion is compromised (Vondenhoff et al., 2009b). Accordingly, Vegfc^+/− embryos with defective lymphangiogenesis have small LNs (Lee and Koh, 2016). Recently, Onder et al. (2017) demonstrated that LEC-specific loss of LTβR and RANK impaired LN formation and proposed an alternative model, in which lymphatics initiate LNs by bringing disseminated LTi cells to the LN development site. The mechanism of LN capsule and SCS formation is also enigmatic. The SCS, the site of arrival of lymph and DCs, hosts resident phagocytes filtering pathogens and particles from the afferent lymph and is continuous with afferent collecting lymphatics. Currently, it is not known whether collecting lymphatic vessel development and maturation is linked to LN capsule and SCS formation.

We propose that early LN development comprises both LTi cell egress from the veins, occurring independently of lymphatic vessels, and lymphatic transport of disseminated LTi cells to the LN site. We further show that efficient LTi cell transport, formation of the LN capsule, and specialization of the SCS at later embryonic stages require collecting lymphatic vessels. Perinodal lymphatics also promote local interstitial flow, which amplifies CXCL13 expression by LTo cells in cooperation with lymphotoxin-β signaling and thereby promotes LTi cell retention. Our data reveal a new role of mechanical forces in LN development and provide a unified model in which lymphatic vessels intervene at multiple points to assist LN expansion.

Mouse embryonic lymphatic vessels first form a capillary-like plexus (Escobedo and Oliver, 2016), and at E15.5–E16.5, in response to increased lymph flow, part of this plexus matures into collecting vessels, characterized by low levels of LYVE1 and VEGFR3 and presence of intraluminal valves (Mäkinen et al., 2005; Norrmén et al., 2009; Sabine et al., 2012, 2015). We visualized remodeling of LN-associated lymphatic vessels using whole mount imaging of inguinal LNs (iLNs) at different stages of embryogenesis. At E15.5, as described earlier (Eberl et al., 2004; Vondenhoff et al., 2009b), single extravascular CD4⁺ LTi cells and aggregates accumulated at bifurcations of the inguinal blood vessels, beneath the subepigastric vein/artery and above lymphatic vessels (Fig. 1, A and B). At E16.5, the lymphatic vessel beneath the LN anlage expanded and formed a shallow disk (Fig. 1 B). The disk progressively enlarged, forming a cup-like structure around growing CD4⁺ LTi cell area, and almost fully engulfed it by E20–E20.5 (Fig. 1 B). While both peripheral and LN LECs express VEGFR3, NRP2 important for LEC sprouting (Xu et al., 2010) was detected only on peripheral LECs. The NRP2 was low in the lymphatic cup around LN anlage and was expressed only in few sprouting LECs at the cup edge (Fig. 1, C and D). 20% of LECs proliferated in the lower part of the cup, and this proportion increased to 30% in the upper part (Fig. 1, E, E′, and F). Lymphatic vessel expansion around the LN anlage is thus a process distinct from lymphatic vessel growth occurring in skin. In the former case, we observed a coordinated extension of continuous double LEC layers, which eventually engulfs the LN anlage. In contrast, peripheral lymphatic vessels display multiple sprouts and form a mesh-like network (Fig. 1 G).

At E18.5, SMCs were associated with the exterior LN LECs and closely followed the expanding LEC sheet (Fig. 2, A and A′), indicating coordinated formation of the LN capsule. Staining for collagen IV revealed increased extracellular matrix deposition around the LN capsule (Fig. 2 B). PDGFβ, a growth factor essential for vascular SMC recruitment (Wang et al., 2017), was expressed by the outer LEC layer (Fig. 2 C). Taken together, our data demonstrate that the growing LN induces a coordinated remodeling and expansion of adjacent lymphatic vessels. As a result, the newly formed lymphatic cup envelops the developing LN while maintaining overall vessel integrity and function (Fig. 2 D).

Earlier models proposed that LNs are formed by RORγt⁺ CD4⁻ CXCR5⁺ pre-LTi cells exiting from blood vessels in response to CXCL13 produced by fibroblastic LTo cells (van de Pavert et al., 2009, 2014). Onder et al. (2017) suggested that lymphatic vessels initiate LN formation by transporting LTi cells to the LN site and by increasing LTi cell retention through CCL21⁺ perinodal lymphatics. To characterize the LN initiation process in more detail we first visualized the localization of CD4⁺ LTi cells and blood vessels at different development stages by staining for αSMA (α-smooth muscle actin), the venous marker endomucin, and the basement membrane component laminin α5. We observed significantly higher SMC coverage of the subepigastric vein at the iLN site at E16.5 in comparison to E15.0 (Fig. 3, A–C). At E15.0, only arteries produced laminin α5, which inhibits the T cell egress from blood vessels (Wu et al., 2009; Fig. 3 B). We also observed a strikingly different SMC coverage of the subepigastric vein along the anterior–posterior axis: the vein in the axillary region had a continuous SMC layer, but SMCs at the venous bifurcation in the iLN area were so sparse that it was difficult to detect the vein contour (Fig. 3, C–E). In contrast, we observed high laminin α5 deposition and complete venous SMC coverage in both locations after E16.5 (Fig. 3, B and C). Importantly, at E15, we observed LTi cell trafficking in SMC-low gaps in the venous EC layer (Fig. 3, F and F′). Significantly, we did not observe CD4⁺ LTi cells in the venous lumen. Because CD4 marks mature LTi cells (van de Pavert et al., 2009, 2014), we analyzed E15.0 Rorc(γt)^EGFP/+ embryos, in which both RORγt⁺ CD4⁻ pre-LTi and mature RORγt⁺ CD4⁺ LTi cells can be observed. Indeed, only a proportion of extravascular iLN RORγt⁺ cells was CD4⁺ (Fig. S1, A and B). Most importantly, we observed intravenous RORγt⁺ cells in the iLN area in close contact with ECs (Fig. S1, C and C′). These results are consistent with the model in which immature RORγt⁺ CD4⁻ LTi cells egress from the vein and undergo further extravascular maturation.

To study if LTi cell transmigration alters venous SMC coverage, we analyzed Id2-deficient embryos, which lack LTi cells (Fig. S1, D–F; Yokota et al., 1999; Rawlins et al., 2009). Venous SMC coverage was still reduced in the iLN regions in these mice (Fig. S1, D–F); therefore, low SMC coverage is an intrinsic characteristic of the vein at this location at E15. Finally, we analyzed Pdgfb^ret/ret embryos, which have low perivascular Pdgfβ and reduced recruitment of mural cells to blood vessels (Lindblom et al., 2003). The iLNs were significantly bigger in Pdgfb^ret/ret embryos in comparison to the controls, indicating that SMC coverage of blood vessels is an important factor regulating LN development (Fig. 3, G and H). These results support the notion that during LN initiation, pre-LTi cells egress at LN development sites using gaps in venous mural coverage and then undergo further local maturation.

To study whether lymphatic vasculature is necessary for pre-LTi cell egress from the vein, we blocked lymphangiogenesis by administering VEGFR3-blocking antibodies. In agreement with Zheng et al. (2014), the treatment severely diminished dermal lymphatics (Fig. 3 I). However, prominent, albeit reduced LTi cell accumulation was still observed in LN anlagen in the absence of associated lymphatic vessels (Fig. 3 I; LN volume, 10³ µm³; control IgG, 319.2 ± 115.9; VEGFR3-blocking IgG, 79.2 ± 25.5; P = 0.0272). Furthermore, we readily visualized LTi cells in SMC-low gaps in the vein in both control and VEGFR3 IgG-treated embryos (Fig. 3 I′).

We next monitored the global distribution of CD4⁺ LTi cells in the embryonic skin. We observed CD4⁺ cells associated with the upper dermis blood vascular plexus (Fig. 3, J and J′). Isolated CD4⁺ cells and clusters were frequently present in the lymphatic plexus at E15.5 and later in lymphatic capillaries and in afferent and efferent LN collecting vessels (Fig. 3, K and L; and Fig. S2, A–D). A similar distribution of GFP⁺ cells was observed in Rorc(γt)^EGFP/+ embryos (Fig. S2, E–G). Our data corroborate lymphatic vessel–independent LN initiation model also for the iLN (Vondenhoff et al., 2009b). However, in agreement with Onder et al. (2017), they show that a reservoir of disseminated LTi cells critically contributes to LN expansion. These results unify the previous models and further suggest that the SMC coverage in part defines the site and the time of pre-LTi cell egress from veins.

Following VEGFR3 blockade, we sometimes observed isolated LEC spheroids (Fig. 4 A), likely formed by nonvenous dermal LEC progenitors (Martinez-Corral et al., 2015). The LEC spheroids contained CD4⁺ cells (Fig. 4 B), indicating that even disconnected LECs attract LTi cells. In adults, CCR7 drives migration of DCs and T cells toward and inside of CCL21⁺ initial lymphatics (Ohl et al., 2004; Bromley et al., 2005; Debes et al., 2005; Russo et al., 2016). Because LTi cells express functional CCR7 (Honda et al., 2001), we analyzed iLN development in Ccr7^−/− mice. While the lymphatic vascular development was not affected (Fig. S2, H and I), Ccr7^−/− iLNs were smaller, and we observed almost no intralymphatic LTi cell clusters (Fig. 4, C–E). Meanwhile, the extravascular CD4⁺ LTi cell fraction was increased (Fig. 4 F). Therefore, in addition to the role of CCL21 in LTi cell retention at LN site (Onder et al., 2017), CCL21⁺ initial lymphatics attract and collect disseminated LTi cells.

In addition to CD4⁺ cells in lymphatic capillaries and afferent LN lymphatics, we observed a low but significant number of CD4⁺ cells in iLN efferent vessels (Fig. S2, A and D). The dynamic equilibrium between arriving, retained, and departing LTi cells thus defines the iLN size. To study whether defective retention of LTi cells shifts the balance toward LTi cell accumulation in lymphatic vessels, we analyzed Cxcr5^−/− embryos. Loss of CXCR5⁺ LTi cell attraction by CXCL13⁺ LTo cells prevents formation of most peripheral LNs, including iLN (Förster et al., 1996; Ansel et al., 2000; Ohl et al., 2003; Fig. 4, G, G′, and H). Lymphatic vessels at the presumptive Cxcr5^−/− iLN site did not develop the characteristic lymphatic cup found in WT mice but maintained a collecting vessel phenotype (Fig. 4, G and G′). As in Cxcr5^−/− animals, lymphatic vessel remodeling into a LN capsule was abolished in LTb^−/− embryos (Koni et al., 1997), which do not form peripheral LNs due to lack of surface LTβ presentation (Fig. S2, J and K), further confirming the driving role of LN anlagen in LN capsule and SCS development. Strikingly, lymphatic vessels of Cxcr5^−/− mice were filled with intralymphatic CD4⁺ clusters (Fig. 4 I and Fig. S2 L). Such clusters were biggest at the presumptive iLN site, but smaller clusters were also disseminated throughout the lymphatic vascular network (Fig. 4 I). Furthermore, the number of isolated CD4⁺ cell in the skin interstitium was increased (Fig. 4 J). Our results thus show that extravascular LTi cell accumulation is a prerequisite for the induction of lymphatic vessel remodeling and LN capsule formation.

Acquisition of a capillary or collecting phenotype is a key step of lymphatic vascular development. In skin, this process starts at E15.5–E16.5 and thus coincides with initiation of iLN capsule formation. We studied whether all LECs of the LN lymphatic cup belong to one or another vessel type by staining for collecting or capillary markers FOXC2 or LYVE1 (Schulte-Merker et al., 2011). The LECs of the outer layer expressed high levels of FOXC2, whereas LECs in the inner layer were FOXC2⁻ but produced higher levels of LYVE1 (Fig. 5 A), indicating a hybrid collecting–capillary identity of the prenatal SCS.

Such organization is reminiscent of adult LN SCS, which contains distinct populations of ceiling LECs (cLECs) and floor LECs (fLECs). Only cLECs produce the chemokine receptor CCRL1, while fLECs express LYVE1, ITGA2B, and MadCAM1 (Cohen et al., 2010; Ulvmar et al., 2014; Cordeiro et al., 2016). In adult LNs of Ccrl1-EGFP mice, FOXC2 was coexpressed with CCRL1 in cLECs (Fig. S3, A and A′), confirming conservation of asymmetric FOXC2 localization within the SCS. In embryonic iLNs, the inner LECs expressed fLEC-specific ITGA2B and MadCAM1, whereas a subset of outer LECs produced CCRL1 (Fig. 5, B–D). We observed continuous CCRL1 expression in cLECs in axillary LN, more advanced in its development (Fig. 5 D; Rennert et al., 1996). The VEGFR3⁺ collagen IV⁺ LECs interconnected the cLEC and fLEC layers (Fig. 5 E). Such pillars were LN specific and were absent in all other lymphatic vessels examined. They likely represent sinus-traversing strands of adult LN SCS (Ohtani et al., 2003; Rantakari et al., 2015) and may be preventing the SCS lumen collapse, in addition to their role in filtering the afferent lymph. Finally, the CD169⁺ macrophages inserted into the adult SCS fLEC layer prevent dissemination of pathogens and deliver antigens to adjacent B cells (Carrasco and Batista, 2007; Junt et al., 2007; Phan et al., 2007). Surprisingly, we observed CD169⁺ macrophages within the fLEC layer already at E18.5 (Fig. 5 F). Taken together, our data show that specialization of SCS LECs is established prenatally, and this timing coincides with initiation of collecting lymphatic vessel maturation.

We asked next whether disruption of collecting vessel specialization affects LN development. In Foxc2^−/− mice, primary lymphatic capillary plexus fails to remodel into collecting vessels (Petrova et al., 2004; Norrmén et al., 2009). We analyzed LN development in Foxc2^f/f; Prox1-CreERT2 (Foxc2^lecKO) mice with LEC-specific loss of FOXC2. As described earlier (Sabine et al., 2015), Foxc2^lecKO embryos had a complete arrest of collecting vessel development, as determined by the absence of valves and expression of capillary markers LYVE1 and CCL21 in all LECs (Fig. 6 A and Fig. S3 B). In parallel with the failed collecting vessel maturation, formation of the LN capsule was severely disrupted: instead of a continuous lymphatic cup, Foxc2^lecKO iLN anlage was surrounded by a mesh-like vasculature, only partially enclosing the LN (Fig. 6 B).

The LN LECs in Foxc2^lecKO mice failed to express SCS LEC markers ITGA2B or CCRL1, whereas LYVE1 and MAdCAM1 were uniformly high. Moreover, mislocalized SMCs encircled the LN lymphatic vessels (Fig. 6, C–F′). Furthermore, while CD169⁺ macrophages were aligned with fLECs in the WT iLN, in Foxc2^lecKO mice the number of macrophages was reduced, and they were dispersed throughout the LN (Fig. 6 G). These results demonstrate that FOXC2 ensures both collecting vessel maturation and prenatal LN LEC capsule specialization.

From E18.5 onward, the Foxc2^lecKO iLNs were significantly smaller than controls (Fig. 7 A). Organization of axillary and mesenteric LNs was also perturbed, but the formation of nonencapsulated Peyer’s patches was not affected (Fig. 7, B–F). We next investigated the reason for the reduced Foxc2^lecKO LN size. The LTi cells in both E18.5 control and Foxc2^lecKO animals were mostly quiescent, and staining for activated caspase 3 did not reveal differences in cell death (Fig. S3, C–E). Instead, we observed accumulation of intra- and extravascular CD4⁺ cells and large intralymphatic clusters both in the vicinity of LN anlagen and throughout the dermal and mesenteric lymphatic vessels of Foxc2^lecKO mice (Fig. 7, G–K; and Fig. S3 F). These results demonstrate that collecting vessel maturation is essential for LN capsule formation, SCS specialization, and further LN expansion.

Because of intralymphatic accumulation of LTi cell clusters in both Cxcr5^−/− and Foxc2^lecKO mice, we compared the expression of CXCR5 ligand Cxcl13 as well as of Ccl21 and Ccl19 chemokines in Foxc2^lecKO or WT iLNs. WT iLNs produced more Cxcl13, Ccl21, and Ccl19 mRNA in comparison to the skin, and only Cxcl13 mRNA was significantly decreased in the absence of FOXC2 (Fig. 8, A and B). On the protein level, CXCL13 but not VCAM1, ICAM1, and RANKL was reduced in Foxc2^lecKO iLNs (Fig. 8, C–J). Because lower CXCL13 could be a result of reduced LTi cell number in Foxc2^lecKO iLNs, we quantified CXCL13 and VCAM1 levels in E17.5 WT and E18.5 Foxc2^lecKO iLNs, which have similar sizes (Fig. 8 K). Only CXCL13 but not VCAM1 was decreased in E17.5 Foxc2^lecKO iLNs (Fig. 8, L–O), demonstrating that LTi cell number is not a critical parameter at this stage. These results show that lymphatic vessel specialization is important for stromal CXCL13 expression and LN expansion.

LTβR signaling promotes CXCL13 expression by fibroblastic LTo cells (Rennert et al., 1998; Dejardin et al., 2002; Müller and Siebenlist, 2003). However, LTβR, LTβ, or LTα expression was not affected in Foxc2^lecKO iLNs (Fig. 8 B). We then reasoned that LTβR signaling, supplied by LTαβ⁺ LTi cells (Vondenhoff et al., 2009a), may be necessary but not sufficient to fully activate Cxcl13 expression in LTo cells.

To clarify the mechanism of Cxcl13 down-regulation in the Foxc2^lecKO iLNs, we first asked whether LECs enhance CXCL13 production by LTo cells. We cultured mouse embryonic fibroblasts (MEFs) with LN LECs or LN LEC–conditioned medium in the presence or in the absence of LTβR-activating antibody (Fig. S4, A–C). As expected (Dejardin et al., 2002; Müller and Siebenlist, 2003), LTβR activation induced Cxcl13 and Vcam1 in MEFs (Fig. S4 C). However, co-culture with LECs or with LEC-conditioned medium suppressed the induction of Cxcl13 (Fig. S4 C). Thus, direct contact with LECs is unlikely to induce Cxcl13 in LTo cells.

By absorbing plasma leaking from blood capillaries, lymphatic vessels generate IFF, which has important morphogenetic functions (Starling, 1896; Wiig and Swartz, 2012). We reasoned that engulfment of LN anlagen by growing LN capsule creates high IFF, from the moment when embryonic maturing lymphatics initiate directional lymph transport. To analyze whether IFF affects Cxcl13 expression in vitro, we cultured MEFs in three dimensions in the presence or absence of LTβR-activating antibody in static or IFF conditions (Fig. 9 A). As expected (Chambliss et al., 2013), IFF induced formation of actin stress fibers indicating cytoskeleton reorganization (Fig. 9 B). Most importantly, IFF potentiated the production of Cxcl13 mRNA in response to LTβR activation (Fig. 9 C). To study IFF in vivo, we microinjected FITC-dextran in the forelimb interstitium of E18.5 control or Foxc2^lecKO embryos. WT lymphatic vessels readily drained the dye, but transport into lymphatic vessels of Foxc2^lecKO mice was strongly reduced (Fig. 9, D and E). We suggest that in addition to their role in LTi cell transport, perinodal lymphatics promote IFF, which cooperates with LTβR signaling to induce CXCL13 in fibroblastic LTo cells and thereby enhances LTi cell retention in the LN anlage (Fig. 10).

We report a high-resolution analysis of coordinated changes occurring during lymphatic vessel and LN development. Our results clarify blood and lymphatic vessel contribution to LN initiation and expansion, describe the process of LN capsule and SCS formation, and reveal the key role of collecting lymphatic vessels and their transport function in LN development.

Together with the results of Vondenhoff et al. (2009b), our data show that initiation of LNs occurs independently of lymphatic vessels. We propose that similar to leukocyte extravasation in inflammation, CD4⁻ pre-LTi cells initially egress in specific venous locations with low SMC coverage, thus defining positioning of the future LN, and undergo local maturation into CD4⁺ LTi cells. Embryos with reduced SMC coverage develop larger LNs, indicating that immature blood vessels are a prerequisite for LTi cell entry into the future LN area. In agreement with Onder et al. (2017), we also show that lymphatic vessels transport the disseminated LTi cells and thus play a critical role in the expansion of early LNs. Our results unify the previous views and further suggest that venous SMC coverage status in part defines the site and the time of pre-LTi cell egress. Our static imaging approach has limitations: it remains possible that the observed venous trafficking of LTi cells in iLN anlage represents the reverse migration into blood vessels. However, we found only CD4⁻ RORγt⁺ intravenous LTi cells in the iLN area, which is in agreement with previous work (Mebius et al., 1997) that postulated the existence of extravascular maturation of LTi cells. Live imaging and characterization of extravascular versus intravascular LTi cells will be invaluable for further unraveling this fascinating process. Because development of different LNs is guided by distinct kinetics, similar high-resolution studies should also be extended to other LNs.

Lymphatic vessels undergo striking remodeling during LN development. Similar to other lymphatic vascular beds, this process is VEGFC/VEGFR3-dependent (Lee and Koh, 2016; this study), but it differs from sprouting lymphangiogenesis observed in peripheral lymphatics. The canonical lymphangiogenic response includes selection of tip cells, sprouting in response to a VEGFC gradient, followed by proliferation of stalk cells, which establish a vascular lumen and partially anastomose to form a mesh-like lymphatic vasculature (Ulvmar and Mäkinen, 2016). In contrast, collective migration and proliferation of LN LECs lead to formation of a double-walled cup-like structure, engulfing the developing LN, with only rare sprouting LECs observed at the top of converging EC sheets (Fig. 1 G). Why VEGFC induces LN engulfment rather than a sprouting response remains to be investigated. This could be due to the uniformly high production of VEGFC in LN cells, resulting in the absence of VEGFC gradient. Indeed, LTo cells express high VEGFC levels (Okuda et al., 2007), while LN LECs produce only low levels of the guidance molecule NRP2 that promotes VEGFC-driven LEC sprouting (Xu et al., 2010). Our results reveal the diversity of organ-specific mechanisms of lymphatic vessel expansion and highlight the need to further investigate the underlying mechanisms.

An advantage of LN engulfment versus lymphangiogenic sprouting is the maintenance of overall vessel integrity and transport function throughout the remodeling process, which is critical for bringing in the disseminated LTi cells. We also show that CCR7⁺ LTi cells use CCL21⁺ lymphatic capillaries as a direct route toward the LNs. However, albeit reduced in number, most LNs are present in mice lacking CCR7 or its ligands (Förster et al., 1999; Luther et al., 2003), indicating that additional mechanisms controlling lymphatic trafficking of LTi cells should exist.

The adult LN SCS is lined by specialized LECs, which ensures optimal immune responses. The specification of SCS LECs occurs before birth and is characterized by acquisition of polarized expression of cLEC and fLEC markers CCRL1, ITGA2B, MADCAM1, and LYVE1; formation of transversal SCS LEC strands; recruitment of CD169⁺ macrophages; and LN capsule SMCs (Fig. 5; Cordeiro et al., 2016). Thus, all cues for the differentiation of cLECs/fLECs and for positioning of specialized LN cells are present before birth. Importantly, we found that SCS specialization coincides with formation of collecting lymphatic vessels. By analyzing Foxc2^lecKO embryos with defective collecting vessel morphogenesis we show that LN and collecting vessel development are highly coordinated, and disruption of collecting vessel function prevents formation of the LN capsule and SCS LEC specialization.

CXCL13/CXCR5 signaling is indispensable for LTi cell retention within the LN anlagen and LN development (van de Pavert and Mebius, 2010). We show now that in Cxcr5^−/− embryos, LTi cells are widely disseminated in the tissues and in the lymphatic vasculature. Foxc2^lecKO embryos display a similar phenotype indicating that functional collecting vessels participate in the regulation of CXCL13. Indeed, CXCL13 is selectively reduced in LTo cells of Foxc2^lecKO mice. Surprisingly, we found that the interstitial flow potentiates Cxcl13 expression by LTβR-activated MEFs, and such flow is impaired in Foxc2^lecKO mice, indicating that lymphatic vessel transport function and IFF contribute to LN expansion. The expression of CCL21, induced by IFF in adult LN (Tomei et al., 2009), was not affected, suggesting a differential responsiveness of embryonic versus adult stromal cells. Primary cilia sense IFF (Nag and Resnick, 2017); however, loss of cilia did not affect IFF-dependent Cxcl13 production by MEFs, indicating that alternative mechanisms are involved (unpublished data). How LTo cells sense IFF and how this process potentiates LTβR signaling at the molecular level will require more work. Most importantly, these results identify a new role of mechanical forces in LN development and suggest that IFF, generated by perinodal lymphatic vessels, enhances the efficiency of LTβR signaling between LTi and LTo cells and thus contributes to LN expansion.

In summary, our work reveals multiple roles of lymphatic vessels in LN development. While the initial LTi cell egress occurs from blood vessels, lymphatic vessels also participate in the LTi cell transport to LN anlagen. Around the LN anlage, collecting vessels reorganize into a capsule to further increase the efficiency of LTi cell delivery and retention. We further propose a novel role of IFF generated by perinodal lymphatics for potentiating CXCL13 expression by fibroblastic LTo cells in response to simultaneous LTβR signaling.

Animal experiments were approved by the Animal Ethics Committee of Canton of Vaud, Switzerland; by the Animal Experimentation Committee of the county of Münster, Germany; and by the Animal Experimentation Committee of Uppsala, Sweden. Foxc2^fl/fl (Sabine et al., 2015), Prox1-Cre^ERT2 (Bazigou et al., 2011), Cxcr5^−/− (Förster et al., 1996), Ltb^−/− (Koni et al., 1997), Ccr7^−/− (Förster et al., 1999), Pdgfb^ret (Lindblom et al., 2003), Id2-Cre^ERT2 (Rawlins et al., 2009), Rorc(γt)^EGFP/+ (Eberl et al., 2004), Ccrl1-eGFP (Heinzel et al., 2007), and Pdgfβ-CreERT2-IRES-EGFP (Claxton et al., 2008) were previously described. All mice were bred to at least the sixth generation on C57BL6/J background. For embryo experiments, matings were set up in the evening, and plugs were verified in the morning. Embryonic age (E) was determined according to the day of the vaginal plug (E0.5). For conditional gene inactivation, pregnant mice were injected with 5 mg of tamoxifen and 5 mg of progesterone in 100 µl sunflower seed oil between E13.5 and E14.5. When indicated, 5 µg/g EdU (Life Technologies) was injected i.p. 1 h before sacrifice. To block lymphangiogenesis, C57BL6/J pregnant mice were injected i.p. at E11.5 and E13.5 with 30 mg/kg rat anti-VEGFR3 (Pytowski et al., 2005) or control rat anti-HRPN IgG antibodies (BioXCell).

Embryos were injected with 0.5–1 µl of 10 mg/ml 2,000-kD FITC-dextran (Sigma-Aldrich) into the interstitium of forelimb foot pads using a 10-µl Hamilton syringe. Lymphatic drainage was analyzed using Leica M205 FA stereomicroscope, Leica DFC 3000 G camera, and LAS X software.

For whole mount staining, tissues were fixed with 4% paraformaldehyde (PFA), washed with PBS, permeabilized with 0.5% Triton X-100, and blocked with 5% donkey serum. Tissues were incubated with primary followed by secondary antibodies for 3 d on a rotating platform at 4°C and cleared using 3DISCO protocol (Ertürk et al., 2012) or alternatively incubated overnight with Histodenz (Sigma) at room temperature. 8-µm cryosections were 4% PFA–fixed, washed with PBS, permeabilized with 0.5% Triton-X-100, and blocked with 5% donkey serum. 5-µm deparaffinized tissue sections were subjected to heat-induced epitope retrieval (High pH Retrieval solution; DAKO). Click-it EdU Alexa Fluor 555 kit (Life Technologies) was used for EdU detection. For staining of cultured MEFs, cells were fixed with 4% PFA, washed with PBS, permeabilized with 0.1% Triton X-100, and blocked with 5% donkey serum. Alexa 488–conjugated phalloidin (Life Technologies) and DAPI (Sigma) were added to the chips overnight. Images were taken using confocal Zeiss LSM 880 or Leica SP5 Tandem microscopes and analyzed using Imaris and Adobe Photoshop software. Frontal stacks of whole mount images were on average 10 µm thick, and transverse stacks were 5–10 µm thick. Large image tiles were acquired using a motorized Zeiss Axio Observer.Z1 with AxioVision Rel.4.7 software. All frontal images of the embryonic iLN are shown with the epidermis facing downward. A list of antibodies used in this study is provided in Table S1.

MEFs from E14.5 embryos were prepared as described previously (González-Loyola et al., 2015) and cultured in complete DMEM. MEFs were used for experiments at passages 1–5. Mouse LN LECs were prepared as described (Dubrot et al., 2014), and LEC purity was confirmed by staining for LEC-specific PROX1 and EC-specific ERG. For co-culture experiments, MEFs and LN LECs were mixed at 50% in complete αMEM medium, seeded on a 6-well plate coated with mixed bovine type I collagen (10 µg/ml; Corning) and human plasma fibronectin (10 µg/ml; Sigma), and cultured for 24 h. To study paracrine signaling, complete endothelial cell medium or conditioned medium from LECs was added to confluent MEFs. To activate lymphotoxin signaling, 2 µg/ml of LTβR (clone 4H8; Adipogen) or control antibodies (MOPC21 muIgG2a) were used.

Microfluidic devices were described in and cell seeding protocol was adapted from Kim et al. (2016). MEFs were resuspended in complete DMEM and fibrinogen solution (2.5 mg/ml) at a concentration of 10 million cells/ml and mixed with thrombin (1 U/ml; Sigma). The cell mix was injected in the left central channel, and all remaining channels were filled with fibrin (2.5 mg/ml). After fibrin gel polymerization, complete DMEM was loaded in the channels and the reservoirs. Devices were kept at 37°C in the incubator overnight. The following day, medium was replaced with complete DMEM supplemented with 2 µg/ml of LTβR (clones 4H8 or 3C8; Adipogen) or control antibodies (MOPC21 muIgG2a). To create a 2–3 µm/s interstitial flow across the central channels, a pressure gradient was applied by transferring the medium between the two channel compartments every 12 h. After 48 h, cells were used for RNA extraction or staining. Images were taken using inverted confocal Zeiss LSM 880 microscope.

iLNs or skins were dissected from E18.5 embryos and RNA isolated using the Qiagen RNeasy Plus Mini Kit (Qiagen). RNA was amplified using Ovation Pico WTA System V2 (Nugen). MEFs from chips and cells from co-culture experiments were collected using 0.05% trypsin-EDTA (Gibco) and RNA isolated using the Qiagen RNeasy Plus Micro or Mini Kit (Qiagen) and reverse transcribed using Transcriptor First Strand cDNA Synthesis Kit (Roche Diagnostics). We used StepOnePlus (Applied Biosystems) and SYBR Green PCR Master Mix (Applied Biosystems) or SensiFAST Sybr Hi-Rox Mix (Bioline) for qPCR analyses. Data were analyzed using the comparative Ct (ΔΔCt) method as described by the manufacturer. A list of primers used in this study is provided in Table S2.

Whole embryonic LNs were imaged using confocal Leica SP5 Tandem microscope. LN size was quantified using Imaris software. Cell number in iLN area was quantified using Adobe Photoshop and ImageJ software. Skin tissue tiles were imaged with a motorized Zeiss Axio Observer.Z1. Skin intralymphatic clusters and isolated CD4⁺ cells were manually analyzed and quantified using Adobe Photoshop and ImageJ software. To quantify the size and number of Peyer’s patches, images were taken with a Leica DMI 3000 B microscope and Leica DFC 3000 G camera using LAS X software and analyzed using ImageJ software. For proliferation analyses, cells were double-stained for PROX1 and EdU. For every LN, whole LNs were analyzed in one transverse thick section (19.6 µm thick) at the top LN area and three thick sections homogenously distributed from the center of the LN (4.5 µm thick). Double PROX1 and EdU masks were applied using Imaris software, and cells were quantified using Adobe Photoshop and ImageJ software, counting as positive only cells in which all PROX1⁺ area was EdU⁺. False-positive cells due to partial overlap of PROX1⁺ cells with adjacent proliferating EdU⁺ cells were excluded. Activated caspase-3 intensity in CD4⁺ area was quantified using Imaris software. For SMC coverage analysis, samples were stained for αSMA and endomucin. Double αSMA/endomucin masks were applied using Imaris software, and total endomucin and αSMA/endomucin area were quantified using Adobe Photoshop and ImageJ software. For all VCAM1 and CXCL13 protein-level analyses at E18.5, five Imaris-generated 4.5-µm-thick transverse sections per LN were analyzed using Adobe Photoshop and ImageJ. For CXCL13 in E17.5 WT and E18.5 Foxc2^lecKO iLNs, ICAM1, and RANKL protein-level analyses, one cryosection per samples was analyzed using Adobe Photoshop and ImageJ. Staining intensity was quantified and normalized by the area analyzed. For FITC-draining experiments, FITC-filled lymphatic vessel area was quantified using Adobe Photoshop and ImageJ.

Number of embryos analyzed is indicated for each image in the figure legends. For in vivo experiments comparing animals, we used a two-tailed unpaired Student’s t test to determine statistical significance by calculating the probability of difference between two means. For iLN volume comparing E17.5 WT, E18.5 WT, and E18.5 Foxc2^lecKO and for SMC coverage comparing Id2-Cre^{ERT2 tg/tg} and Id2-Cre^{ERT2 tg/+}, we used one-way ANOVA with Tukey post hoc test. For in vitro experiments in microfluidic devices, we used a one-way ANOVA with Tukey post hoc test. Differences were considered statistically significant at P < 0.05. Data are shown as mean ± SD.

Fig. S1 provides further characterization of LTi cells and blood vascular maturation status during early iLN development. Fig. S2 provides characterization of intralymphatic CD4⁺ LTi cells. Fig. S3 provides further characterization of lymphatic vessels in Foxc2^lecKO mice. Fig. S4 shows that LECs reduce MEF expression of CXCL13 in response to LTβR activation. Tables S1 and S2 provide information on antibodies and primers, respectively.

We thank L. Sorokin (University of Münster, Münster, Germany) for laminin α5 antibodies; A. Gonzalez Loyola for help with MEF isolation; T. Mäkinen (Uppsala University, Uppsala, Sweden), S. Hugues and J. Dubrot Armendariz (University of Geneva, Geneva, Switzerland), M. Fruttiger (University College London, London, UK), and P. Romero (University of Lausanne, Lausanne, Switzerland) for Prox1-CreERT2, Rorc(γt)^EGFP/+, Pdgfβ-CreERT2-IRES-EGFP, and Id2-Cre^ERT2 mice, respectively; C. Beauverd for mouse genotyping and colony maintenance; A. Brendolan, J. Caamano, M. Dickinson, E. Lammert, J. Bernier-Latmani, and S. van de Pavert for useful discussions; and Animal, Cellular Imaging, Mouse Pathology and Genomic Technology Facilities of the University of Lausanne.

This work was supported by the Swiss National Science Foundation (31003A-156266 and 31ER30_160674 to T.V. Petrova), the Leenaards Foundation (to T.V. Petrova), Theodor and Gabriela Kummer and Société Académique Vaudoise fellowships (to E. Bovay), the Swedish Cancer Foundation (CAN2015/771 to C. Betsholtz), Vetenskapsrådet (VR2015-00550 to C. Betsholtz), Knut och Alice Wallenbergs Stiftelse (2012.0272 to C. Betsholtz), Fondation Leducq (14-CVD-02 to C. Olsson), the Deutsche Forschungsgemeinschaft (CRC656 and CRC1348 to F. Kiefer), and the National Research Foundation funded by the Ministry of Education of Korea (NRF-2018R1A2A1A05019550 to N.L. Jeon).

The authors declare no competing financial interests.

Author contributions: E. Bovay and T.V. Petrova conceived and designed research, interpreted the data, and drafted the manuscript; E. Bovay and B. Prat-Luri conducted experiments; E. Bovay acquired and analyzed data; A. Sabine, S. Luther, A.-H. Willrodt, C. Olsson, C. Halin, F. Kiefer, C. Betsholtz, S. Kim, K. Son, and N.L. Jeon provided reagents and intellectual input and interpreted the data; and all authors read, commented on, and approved the manuscript.