Mesenchymal Osr1⁺ cells regulate embryonic lymphatic vessel formation

The lymphatic vasculature establishes a blind-ended hierarchical network of vessels with crucial roles in interstitial fluid homeostasis, immune cell response and lipid metabolism. In peripheral tissues, lymphatic vessels form thin-walled capillaries that drain interstitial fluid and facilitate the transport of macromolecules and cells into larger pre-collecting lymphatic vessels. Here, the lymph is moved unidirectionally by the cooperative action of lymphatic valves, the synchronic contraction of lymphatic vessel-associated mural cells and passively by the force generated in surrounding tissues such as skeletal muscles and arteries. Finally, the lymph is transported into the blood stream at the subclavian vein (Tammela and Alitalo, 2010; Schulte-Merker et al., 2011; Vaahtomeri et al., 2017).

Lymphatic vessels are formed by specialized endothelial cells, the lymphatic endothelial cells (LECs), which originate from venous and non-venous tissues depending on the vascular bed (Jafree et al., 2021; Lioux et al., 2020; Stanczuk et al., 2015). By far the best understood process is the formation of LECs via dedifferentiation of embryonic venous endothelial cells. In mice, reprogramming of venous cells towards an LEC fate occurs at embryonic day (E)9-E10, relying on the activation of key LEC transcription factors such as prospero homeobox1 (PROX1) and SRY-related HMG-box 18 (SOX18), and is associated with the expression of markers such as lymphatic vessel endothelial hyaluronan receptor 1 (LYVE1) (Tammela and Alitalo, 2010; Schulte-Merker et al., 2011; Francois et al., 2008). After this initial step in LEC commitment, which occurs at the dorsal part of the cardinal vein, LECs delaminate and migrate into the surrounding mesenchyme, forming transient primordial lymphatic vascular structures called primordial thoracic ducts, formerly known as lymph sacs, which can be observed in the mouse embryo during the stages E11.5-E14.5 (Tammela and Alitalo, 2010; Schulte-Merker et al., 2011; Alitalo et al., 2005; Oliver, 2004). LEC initial migration from the cardinal vein is dependent on the production of vascular endothelial growth factor C (VEGFC) and the activation of its primary receptor fms-related tyrosine kinase 4 (FLT4), also known as VEGFR3 (Karkkainen et al., 2004; Hagerling et al., 2013). Activation of canonical or non-canonical VEGFR3-signaling, and enzymes controlling the proteolytic activation of VEGFC, play an essential role in LEC initial migration as well as in development and maintenance of lymphatic vessels (Karkkainen et al., 2004; Wang et al., 2020; Janssen et al., 2016; Jeltsch et al., 2014; Koltowska et al., 2015). In addition, directed migration of LECs is controlled by the CXCL12/CXCR4 chemokine axis (Peng et al., 2022; Niimi et al., 2020; Cha et al., 2012). Both Vegfc and Cxcl12 are highly expressed by embryonic mesenchymal cells adjacent to the developing lymphatic vasculature (Karkkainen et al., 2004; Wang et al., 2020; Peng et al., 2022; Vallecillo-Garcia et al., 2023). Despite the knowledge gained on mechanisms supporting LEC migration in recent years (Shiiya and Hirashima, 2023), little is known about how the final pattern of lymphatic vessels in different tissues of the embryo is achieved. Several cell types have been described to influence lymphatic vessel development in zebrafish and mice, such as platelets, arterial endothelial cells, neurons, myeloid cells and mural cells (Wang et al., 2020; Peng et al., 2022; Bussmann et al., 2010; Uchida et al., 2015). However, embryonic mesenchymal cells adjacent to lymphatic vessels have remained understudied, in part due to the lack of specific markers that label mesenchymal cell subpopulations during development. Mesenchymal cells are important producers of the extracellular matrix (ECM) scaffold during embryogenesis and adult life (Walma and Yamada, 2020). Of note, LEC-ECM interactions and biophysical properties of the ECM associated with lymphatic vessels influence lymphatic vessel development and function (Frye et al., 2018; Wiig et al., 2010).

The transcription factor Osr1 is expressed in a variety of mesenchymal cells derived from the lateral plate and intermediate mesoderm (Mugford et al., 2008; Wang et al., 2005). In the limb, Osr1⁺ mesenchymal cells are present before skeletal muscle progenitors colonize the limb bud mesenchyme, and they produce guidance and a proper ECM scaffold for skeletal muscle patterning (Vallecillo-Garcia et al., 2017). Osr1 expression decreases in late fetal stages of development but is reactivated after tissue damage (Vallecillo-Garcia et al., 2023, 2017; Stumm et al., 2018; Kotsaris et al., 2023). We recently showed that Osr1 is required in mesenchymal cells to organize lymph node lymphatic vasculature assembly, and that Osr1-expressing cells cooperate with LECs in the lymph node anlage driving lymph node initiation (Vallecillo-Garcia et al., 2023). This suggests a crosstalk between mesenchymal Osr1⁺ cells and LECs as an important event for lymphatic vessel formation. Moreover, the lymphatic vasculature can be remodeled in pathogenic conditions such as inflammation, wound healing, tumor formation, hypertension or tissue transplantation (Jafree et al., 2021; Pichol-Thievend et al., 2018), where mesenchymal cell-LEC interactions might be necessary to achieve lymphangiogenesis. Nevertheless, the role of these mesenchymal Osr1-expressing cells or the ECM produced by them in the formation of murine lymphatic vasculature remains obscure.

Here, we show that mesenchymal cells expressing the transcription factor Osr1 are in close association with the developing venous-derived lymphatic endothelial cells. Functionally, the lack of Osr1 revealed a non-cell-autonomous function of embryonic Osr1⁺ mesenchymal cells that control lymphatic vessel formation by producing a beneficial ECM scaffold and signaling molecules necessary for LEC proliferation and directed migration.

Impaired lymphatic vasculature assembly in embryonic lymph node anlage and back lymphedema found in Osr1-deficient embryos indicates a more general role of Osr1⁺ mesenchymal cells in lymphatic vasculature formation (Vallecillo-Garcia et al., 2023; Wang et al., 2005). Therefore, we analyzed Osr1 expression and distribution of Osr1⁺ cells in association with the developing lymphatic vasculature. Using an Osr1-GFP reporter mouse line (Osr1^GCE/+) (Mugford et al., 2008), we identified Osr1⁺ cells in close association with LECs during the first stages of lymphatic vessel development. At E11.5, Osr1⁺ cells were found in the mesenchyme populated by delaminating and migrating PROX1⁺ LECs dorsal and ventral of the cardinal vein (Fig. 1A,A′; Fig. S1A). Of note, at E11.5, Osr1⁺ cells were not found in dorsal skin tissues (Fig. 1A″). At E12.5, we observed that the primordial thoracic duct was surrounded by a mesenchyme rich in Osr1⁺ cells (Fig. 1B). At E14.5, when the first lymphatic vascular structures are established in the dermis of the back, whole-mount immunofluorescence of E14.5 Osr1^GCE/+ isolated skin showed dermal lymphatic vasculature embedded by Osr1⁺ dermal fibroblasts (Fig. 1C; Fig. S1B).

In order to follow Osr1⁺ mesenchymal cell descendants and their contribution to the mesenchyme adjacent to blood and lymphatic vasculature, we performed whole-mount immunofluorescence of E14.5 Rosa26^mTmG/+ Osr1^GCE/+ back skin after tamoxifen induction at E11.5. Genetic tracing of E11.5 Osr1⁺ mesenchymal cells (schematic depiction in Fig. S1C) showed contribution to dermal fibroblasts between blood and lymph vessels (Fig. 1D), to mural smooth muscle actin-expressing cells (αSMA⁺) in arteries, and to cells associated with veins in the skin (Fig. S1C). In the back skin, Osr1⁺ descendants were found in the lymphatic avascular midline, suggesting that Osr1⁺ cells progress ahead of LECs in the back dermis (Fig. 1E), anticipating LEC migration through the forming dermis. Of note, Osr1⁺ descendants were located in close association with blood and lymphatic vasculature but, in agreement with previous results (Vallecillo-Garcia et al., 2023), mGFP signal was not found in endothelial cells (Fig. S1D).

Analyses of Osr1 expression in embryonic stages and adult tissues have revealed that Osr1 expression decreases at late stages of development (Vallecillo-Garcia et al., 2017; Stumm et al., 2018; Kotsaris et al., 2023) and Osr1⁺ cells are found in the stroma of some adult tissues (Vallecillo-Garcia et al., 2023, 2017; Kotsaris et al., 2023). We analyzed Osr1⁺ cell distribution in association with blood and lymphatic vasculature in adult ear dermis and lymph nodes of Osr1^lacZ/+ animals. Whole-mount lacZ staining revealed that Osr1⁺ cells were associated with arteries in the dermal tissue of the ear and with lymphatic vessels in the medulla of mesenteric lymph nodes of Osr1^lacZ/+ animals (Fig. 1F; Fig. S1E). Ear punch trauma led to an activation of Osr1 expression in surrounding mesenchyme (Fig. 1F, right panel; Fig. S1F) in cells closely associated with lymphatic and blood vasculature (Fig. 1G; Fig. S1G). In summary, Osr1⁺ cells accompany LEC migration during lymphatic vasculature formation in the embryo and persist as mesenchymal vessel-associated cells, including mural cells, in adult tissues.

As Osr1-expressing mesenchymal cells associated with migrating LECs and embedded the first lymphatic vasculature in the skin, we asked whether impairment of the lymphatic vasculature could be responsible for the back edema observed in Osr1^GCE/GCE (knockout; KO) embryos (Fig. S2A; Wang et al., 2005). We first analyzed whether the commitment or delamination of LECs from the cardinal vein are affected in Osr1^GCE/GCE embryos. Immunofluorescence of 100-µm thick sections from E11.5 Osr1^+/+ and Osr1^GCE/GCE embryos showed no significant changes in LEC commitment or delamination into the surrounding mesenchyme, together with normal blood vasculature structures (Fig. 2A; Fig. S2B). In line, quantification of PROX1⁺ LECs inside the cardinal vein and migrating into the mesenchyme dorsally from the cardinal vein were not significantly changed in E11.5 Osr1^GCE/GCE embryos (Fig. 2A, lower panel). At E12.5, the first impairments in lymphatic structures were identified by immunofluorescence in Osr1^GCE/GCE embryos, before a back edema was visible. Migration of LECs towards the dorsal region of the embryo was decreased and the PROX1⁺/VEGFR3⁺ vascular network in the primordial thoracic duct region showed reduced vascular branching (Fig. 2B). We found a reduction of VEGFR3⁺ lymphatic vessels after immunofluorescence of 100-µm thick sections at the back of E12.5 Osr1^GCE/GCE embryos (Fig. 2B, lower panel). In line, whole-mount immunofluorescence of isolated back skin from E14.5 embryos showed markedly impaired migration of dermal LECs in the skin of E14.5 Osr1^GCE/GCE embryos, with significantly increased distances between growing fronts in both cervical and lumbar regions compared with Osr1^+/+ controls (Fig. 2C-E; Fig. S2C,D). In addition, lymphatic vessels showed a decrease in the number of branching points (Figs 2C,F,G) and increased caliber (Fig. 2F,H) in E14.5 Osr1^GCE/GCE skin. In line with the defects found in LEC migration, LECs in the migrating front of E14.5 Osr1^GCE/GCE dermis presented a reduced number of filopodia per PROX1⁺ LEC (Figs 2I; Fig. S2E) suggesting an impaired ECM-LEC interaction. Of note, in the skin of E14.5 Osr1^GCE/GCE embryos, blood vasculature showed no significant impairment in vessel thickness or branching point numbers (Fig. S2F).

Overall, this shows that the transcription factor Osr1 expressed in mesenchymal cells is crucial for the establishment of the lymphatic vasculature; the loss of Osr1 leads to severe impairments in LEC migration and subsequent formation of lymphatic vasculature structures.

At E13.5, Osr1 is transcribed mainly in mesenchymal cell populations and is not detected in LECs (Vallecillo-Garcia et al., 2023, 2017). To assess LEC defects in E13.5 Osr1-deficient embryos, we first isolated E13.5 LECs (PDPN⁺ PECAM1⁺) via fluorescence activated cell sorting (FACS) (Fig. 3A). Enrichment of LECs was confirmed via real-time (RT)-qPCR analysis of key LEC markers Vegfr3, Prox1 and Ccl21 compared with FACS-isolated E13.5 blood endothelial cells (BECs) and GFP⁺ mesenchymal cells from Osr1^GCE/+ embryos (Fig. 3B). Of note, relative percentages of BECs and LECs were not impaired in E13.5 Osr1^GCE/GCE embryos (Fig. S3A), suggesting that there was no general defect in LEC pool expansion and maintenance. Next, we performed transcriptome analysis of E13.5 LECs from Osr1^controls and Osr1^GCE/GCE embryos. Consistent with a high expression of LEC markers observed via RT-qPCR (Fig. 3B), RNA-seq analysis from E13.5 LECs showed higher expression of LEC markers compared with BEC (Jurisic and Detmar, 2009) and mesenchymal markers (Table S1). This analysis revealed 1386 differentially expressed genes (DEGs) with 877 down-, and 509 upregulated genes (Fig. 3C). We further characterized DEGs by performing gene ontology (GO) analysis. GO analysis for biological processes on all DEGs revealed an enrichment of terms associated with ECM organization and cell migration (Fig. 3D; Fig. S3B). ECM-associated terms were also identified by GO analysis using the Jensen compartment database (Fig. 3E), and the term ECM-receptor interaction was enriched after GO analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database (Fig. 3F; Fig. S3C). Further GO analysis for biological processes by separating down- and upregulated genes showed an enrichment of ECM terms in genes downregulated in LECs of Osr1^GCE/GCE embryos (Fig. 3G). In line, GO analysis against the Jensen Compartment database by separating down- and upregulated genes revealed genes associated with ECM-related terms as top ranked within downregulated genes (Fig. S3D). Among genes upregulated in LECs of E13.5 Osr1^GCE/GCE embryos, GO analysis for biological processes revealed terms associated with vascular development ranking top (Fig. 3G). In line with this, selected genes known to have a positive effect in lymphatic vasculature formation (Schulte-Merker et al., 2011; Alitalo et al., 2005; Gauvrit et al., 2018; Liu et al., 2018; Xiang et al., 2020) were found to be upregulated in LECs of E13.5 Osr1^GCE/GCE embryos (Fig. 3H).

In summary, lack of Osr1 in mesenchymal cells leads to a series of transcriptional adaptations in embryonic LECs. Mesenchymal cell impairment leads to decreased expression of ECM and ECM-receptor interaction genes and downregulation of genes involved in cell migration in LECs of Osr1^GCE/GCE embryos. Finally, E13.5 Osr1^GCE/GCE LECs activate genes positively involved in lymphatic vessel formation, suggesting a compensatory mechanism.

RNA-seq analysis revealed that LECs in Osr1^GCE/GCE embryos have a deregulated transcriptional signature of ECM organization and ECM-receptor interaction, with many genes related to these terms being downregulated (Fig. 3; Fig. S3). To address this from the perspective of Osr1⁺ mesenchymal cells, we used our previously published RNA-seq dataset of Osr1⁺ cells from E13.5 Osr1^GCE/+ versus Osr1^GCE/GCE embryos (Vallecillo-Garcia et al., 2017). To specifically select for dermal fibroblast-expressed genes, we intersected the 511 E13.5 Osr1^GCE/GCE DEGs with the 976 genes found to be highly abundant in the dermal cluster of E13.5 Osr1-expressing cells extracted from a published single-cell (sc)RNA-seq dataset (Vallecillo-Garcia et al., 2023), given that this cluster was enriched for key embryonic skin dermal fibroblast population markers (Ge et al., 2020; Gupta et al., 2019) (Fig. 4A). Among the genes deregulated in E13.5 Osr1^GCE/GCE cells, we identified 50 deregulated ‘dermal’ genes (Fig. 4B; Table S6). We subjected this set of genes to GO analysis and found the terms ‘small leucine-rich proteoglycan (SLRP) molecules’, ‘NCAM1 interactions’ and ‘TGF-beta regulation of extracellular matrix’ as significant terms downregulated in this intersection (Fig. 4C). Further GO analysis of the molecular functions of upregulated genes in this intersection revealed that ‘metalloendopeptidase activity’ and ‘metallopeptidase activity’ were the only significantly enriched terms (Fig. 4D). Consistent with the observations of ECM defects seen in embryos lacking Osr1 in other tissues (Vallecillo-Garcia et al., 2017; Kotsaris et al., 2023), this analysis suggested that in dermal E13.5 Osr1-deficient cells, ECM production and organization mediated by mesenchymal cells may be also impaired.

Transcriptome analyses of E13.5 Osr1⁺ cells and LECs suggested defective ECM organization in the developing lymph vasculature of Osr1^GCE/GCE embryos. We therefore characterized the ECM in direct contact with the dermal vasculature. Dermal fibroblasts are the primary source of ECM scaffold components such as Col12a1, Col1a1, Fn1, Col3a1, or small leucine-rich proteoglycans (SLRPs) such as Dcn or Lum; of note, Osr1 expression is enriched in dermal fibroblasts (Rezza et al., 2016) (Fig. S4A). Whole-mount immunofluorescence of E14.5 skin using antibodies against the ECM proteins COL12A1, COL6 and TNC revealed an impaired ECM scaffold in direct contact with lymphatic vasculature showing frayed fiber organization and increased TNC expression in E14.5 Osr1^GCE/GCE embryos (Fig. 5A-C). The representative images shown in Fig. 5A-C display lymphatic vessels and the surrounding ECM of lymphatic vessels at the front of the migrating zone; of note, ECM-organization defects were observed in ventral regions as well as in the region of the dorsal migrating front of LECs (Fig. S4B). Despite normal formation of dermal blood vessels observed in E14.5 Osr1^GCE/GCE embryos (Fig. S2F), closer appreciation of the basal lamina in dermal blood capillaries showed an irregular, ruffled COL1 staining along the PECAM1⁺ endothelial layer (Fig. 5D). In contrast to blood capillaries, lymphatic vessels at this stage showed almost an absent COL1 staining along the VEGFR3^high endothelial layer in both controls and Osr1^GCE/GCE embryos (Fig. S4C).

Activation of the VEGFC/VEGFR3 signaling pathway is essential for lymphatic vessel formation driving LEC delamination from the cardinal vein and LEC proliferation (Karkkainen et al., 2004; Zhang et al., 2010; Wang et al., 2001; Makinen et al., 2001; Flister et al., 2010). Available scRNA-seq data for the stages E9.5-E13.5 (Cao et al., 2019) showed that mesenchymal cells and endothelial cells were the major source of Vegfc expression (Fig. S5A). We confirmed that E13.5 Osr1⁺ cells showed high expression of Vegfc, together with endothelial cells, by comparing Vegfc transcriptional expression in Osr1⁺ cells, BECs and LECs separated by FACS (Fig. 6A). At E11.5, Vegfc expression in BECs was higher compared with mesenchymal Osr1⁺ cells; however, at E13.5 BECs and Osr1⁺ cells express comparable Vegfc levels (Fig. S5B). Transcriptome analysis of E13.5 Osr1^GCE/+ and E13.5 Osr1^GCE/GCE mesenchymal cells revealed a decrease in Vegfc transcripts in cells lacking Osr1 (Vallecillo-Garcia et al., 2017). We confirmed Vegfc transcript downregulation in E13.5 Osr1^GCE/GCE mesenchymal cells, whereas at E11.5, Osr1^GCE/GCE mesenchymal cells produced similar Vegfc transcript levels as controls (Fig. 6B). Contradictory to the notion that reduced Vegfc in mesenchymal cells may cause a downregulation of Vegfr3 in LECs, we observed an increase in Vegfr3 expression in LECs of E13.5 Osr1^GCE/GCE embryos, together with other genes involved in the VEGFR3-signaling axis such as Hhex, Maf or Egr1 (Koltowska et al., 2015; Gauvrit et al., 2018; Shin et al., 2008; Dieterich et al., 2017) (Fig. 6C). As BECs are also a source of Vegfc, we analyzed Vegfc expression in E13.5 Osr1^GCE/GCE embryos via RT-qPCR analysis of FACS-isolated BECs and did not observe a change in Vegfc expression (Fig. S5C). Of note, available transcriptome analysis of E14.5 skin populations confirms fibroblasts and melanocytes as the main cell types expressing Vegfc (Sennett et al., 2015) (Fig. S5D).

E13.5 LECs isolated by FACS from E13.5 whole Osr1^+/+ and Osr1^GCE/GCE embryos and cultured in vitro did not show defects in cell proliferation (Fig. 6D). Interestingly, using whole-mount immunofluorescence we detected a heterogeneous proliferation of LECs in wild type embryos, depending on their distribution in the skin. LECs in the migrating front show higher proliferation quantified by PROX1/Ki67 co-staining, compared to LECs on the ventral side (Fig. 6E). Quantification of E14.5 LECs proliferation showed a reduced proliferation in E14.5 Osr1^GCE/GCE embryos specifically in the migrating front, while LECs in the ventral lymphatic vasculature (Fig. 6E) showed similar proliferation in Osr1-deficient and control embryos.

In order to untangle the mechanism used by Osr1⁺ mesenchymal cells to control LEC proliferation, we aimed to separate the effects coming from signaling molecules secreted by Osr1⁺ cells and the effects of the ECM scaffold produced by Osr1⁺ cells. For this purpose, we isolated E13.5 Osr1^GCE/+ and Osr1^GCE/GCE cells via FACS and let them produce either an ECM scaffold or a conditioned medium (Fig. 6F). Next, we isolated dermal LECs from the tail (tdLECs) of adult wild-type mice (Fig. S5E,F) and quantified the effects of decellularized ECM (dECM) or cell-conditioned media on tdLECs proliferation (Fig. 6F). LECs were cultured for 48 h on dECM produced by Osr1⁺ cells (Fig. S5G). LECs cultured on dECM produced by E13.5 Osr1^GCE/GCE cells showed reduced proliferation, quantified by PROX1⁺/Ki67⁺ co-staining, compared with LECs cultured on dECM produced by E13.5 Osr1^GCE/+ cells (Fig. 6G). Conversely, conditioned medium produced by E13.5 Osr1^GCE/GCE cells did not significantly reduce LEC proliferation compared with E13.5 Osr1^GCE/+ conditioned medium (Fig. 6H).

We conclude that, although mesenchymal Osr1⁺ cells are an important source of Vegfc in the mouse, decreased Vegfc in Osr1^GCE/GCE cells did not lead to decreased VEGFR3 downstream target expression in LECs. Instead, a defective ECM secreted by Osr1^GCE/GCE cells affects LEC proliferation in vitro in line with decreased LEC proliferation in E14.5 Osr1^GCE/GCE embryos at the migrating front of the growing lymph vasculature.

In line with a reduced LEC migration observed in the skin of E14.5 Osr1^GCE/GCE embryos, cell migration was one of the most enriched terms in the GO analysis for biological processes in E13.5 LEC RNA-seq data, in which the majority of deregulated genes were downregulated (Fig. 7A). We therefore aimed to clarify whether Osr1 controls the expression of signaling molecules in mesenchymal cells that could act on LECs to promote their migration. We assessed LEC migration in vitro by performing scratch assays in cultures of tdLECs isolated from the dermis of adult tail tissues supplemented with conditioned media produced by E13.5 Osr1^GCE/+ and Osr1^GCE/GCE cells. tdLECs cultured for 24 h in conditioned medium from E13.5 Osr1^GCE/GCE cells migrated slower into the acellular space compared with LECs cultured in conditioned medium from E13.5 Osr1^GCE/+ control cells (Fig. 7B). To evaluate putative interactions between E13.5 Osr1⁺ cells and LECs, we analyzed ligand-receptor interactions using all deregulated genes found in E13.5 Osr1^GCE/GCE cells (511 genes) and in E13.5 LECs (1386 genes) that matched with the interacting pairs defined in Skelly et al. (2018) (Fig. 7C; Table S7). Within the interactions of deregulated ligands in Osr1⁺ cells and deregulated LEC receptors, we found that the Osr1-LEC ligand-receptor pairs TNC-EGFR/ITGA9/ITGAV represented interactions of genes found upregulated in both cell types in line with increased TNC protein abundance in Osr1^GCE/GCE skin (Fig. 5B). By contrast, the ligand-receptor pairs COL3A1-DDR1/2 and CXCL12-CXCR4 depicted ligands highly expressed by Osr1⁺ cells (Col3a1 and Cxcl12), which were downregulated in E13.5 Osr1^GCE/GCE cells. Conversely, their interaction partners (Ddr1/2 and Cxcr4) were downregulated in LECs of Osr1^GCE/GCE embryos. The chemokine CXCL12 controls LEC migration in zebrafish development and in newborn mice via its receptor CXCR4 (Peng et al., 2022; Cha et al., 2012). Therefore, we asked whether CXCL12 might have a similar function in embryonic dermal lymphatic vasculature formation in mouse embryos. First, we assessed Cxcl12 expression in E13.5 Osr1⁺ cells, BECs and LECs and observed that mesenchymal Osr1⁺ cells appear as the main source of Cxcl12 expression (Fig. 7D). Next, we used the Cxcr4^KO/KO line (Zou et al., 1998) to assess the importance of the CXCL12/CXCR4 axis and performed whole-mount immunofluorescence for LEC markers. At E14.5 LECs showed impaired migration to the dorsal midline in E14.5 Cxcr4^KO/KO embryos, with a concomitant increase in lymphatic vessel caliber and reduced arborization (Fig. 7E; Fig. S6A), similar to E14.5 Osr1^GCE/GCE embryos. In line with the pro-migratory function assigned to cxcl12 in zebrafish (Peng et al., 2022), LECs in Cxcr4^KO/KO embryos did not show a reduced proliferation at the tip of the migrating front (Fig. 7F). This suggests that, in the mouse, CXCL12/CXCR4 signaling is required for LEC dorsal migration downstream of mesenchymal Osr1.

The formation of lymphatic vasculature during development is crucial for body fluid homeostasis and, therefore, for animal survival. Although several steps in lymphatic vessel development have been well characterized, the role of mesenchymal cells in lymphatic vessel development has remained understudied.

We have shown that mesenchymal Osr1⁺ cells accompany the early migration path of LECs from the cardinal vein to peripheral tissues, including the mesenchyme surrounding the primordial thoracic duct, and Osr1⁺ cells remain in close association with lymphatic vasculature in several vascular beds. Lineage tracing experiments suggest that mesenchymal Osr1⁺ cells appear in the mesenchyme preceding the LEC migration front in the dermis of the embryo. Osr1⁺ descendants also contribute to mural cells of arteries and veins; it remains to be elucidated whether Osr1⁺ cells constitute a population of mural cells for mature lymphatic vessels. Interestingly, in the medulla of lymph nodes, Osr1 expression remains active in close association to blood and lymphatic vessels, and Osr1 and Vegfc/Vegfa expression are both found in lymph node mesenchymal stromal subpopulations (data not shown, Cyster Lab Shyni Server).

Reminiscent of Osr1 expression induction in skeletal muscle mesenchymal fibro-adipogenic progenitors triggered by acute muscle injury (Stumm et al., 2018), we observed transcriptional activation of Osr1 in the dermis of the ear after trauma. This indicates that Osr1 activation may represent a general mechanism employ by mesenchymal cells as a trauma response, possibly to induce a pro-remodeling phenotype.

Lack of Osr1 in mesenchymal cells in Osr1^GCE/GCE embryos led to impairment of lymphatic vasculature formation, indicating that Osr1⁺ mesenchymal cells are a new important player controlling LEC behavior in a non-cell-autonomous fashion. In line with the function assigned to mural cells in zebrafish controlling LEC migration and survival (Peng et al., 2022), we observed LEC migration defects in Osr1-deficient embryos and a reduced dermal LEC proliferation specifically at the tips of the migrating front. Conversely, blood vessel morphology and pattern were not significantly impaired in Osr1-deficient embryos. Of note, the collagen-rich basal lamina of capillaries in the skin of E14.5 Osr1^GCE/GCE embryos shows defects in COL1 abundance, hinting at a possible function of Osr1⁺ cells in blood vessel stability, which remains to be explored. The differential impact of Osr1 deficiency on blood and lymphatic vasculature formation is in agreement with the earlier formation of the blood vascular plexus in the embryo (Walls et al., 2008) and a later appearance of Osr1⁺ mesenchymal cells in close proximity to blood vasculature (Mugford et al., 2008; Stricker et al., 2006).

We found that Osr1⁺ cells are a source of Vegfc in embryonic tissues. Interestingly, Vegfc expression at early stages of development was higher in endothelial cells (Cao et al., 2019), whereas at E13.5 Osr1⁺ mesenchymal cells and BECs expressed similar amounts of Vegfc. Despite Vegfc downregulation observed only in E13.5 Osr1^GCE/GCE mesenchymal cells, the LEC transcriptome data suggested that embryonic E13.5 LECs show increased expression of genes involved in the VEGFC-VEGFR3 cascade, at least when assessed at the transcriptional level. This might represent a compensatory mechanism in line with the transcriptional upregulation of ECM components, ECM-interaction genes and other genes positively involved in vascular development, as observed in GO analysis of E13.5 LECs from Osr1^GCE/GCE embryos. Alternatively, two important further aspects may separately or in combination play a role in this transcriptional upregulation. One aspect is the ECM and its mechanical properties sensed by LECs in Osr1^GCE/GCE embryos. We have shown that Osr1⁺ cells are important producers of ECM proteins and that lack of Osr1 leads to impaired ECM in muscle connective tissue (Vallecillo-Garcia et al., 2017), and in the skin of E14.5 embryos embedding lymphatic vasculature and capillaries (Fig. 5; Fig. S4). It has been demonstrated that the VEGFR3 tyrosine kinase function is highly dependent on mechanical forces (Galvagni et al., 2010; Planas-Paz et al., 2012; Baeyens et al., 2015), representing a crucial aspect in signaling pathway activation downstream of VEGFR3, even independent of VEGFC action. It was shown that reduced tissue stiffness leads to the upregulation of genes positively implicated in lymphatic vascular formation, e.g. Vegfr3, Itga9 or Hey1 (Frye et al., 2018), which we found upregulated in LECs from E13.5 Osr1^GCE/GCE embryos. Intriguingly, we have observed reduced tissue stiffness upon loss of Osr1 in a model of skeletal muscle injury (Kotsaris et al., 2023). Thus, the defective EMC we found in Osr1^GCE/GCE embryos may lead to the transcriptional upregulation of genes involved in lymphangiogenesis in LECs. Another important aspect also coupled to the ECM component is fluid shear stress sensed by LECs. LECs are very sensitive to changes in shear stress and VEGFR3 builds a mechanosensory complex modulating different aspects of LEC behavior (Baeyens et al., 2015; Angeli and Lim, 2023). Osr1^GCE/GCE embryos show malformation in the septum primum (Wang et al., 2005) and we here found lymphatic vasculature defects; both aspects very likely affect the shear stress experienced by LECs in E13.5 Osr1^GCE/GCE embryos. Thus, the combination of changes in fluid shear stress experienced by LECs on the luminal side and an impaired ECM and defective support of signaling molecules at the basal side could, in combination, explain upregulation of genes involved in lymphangiogenesis in LECs of E13.5 Osr1^GCE/GCE embryos. Furthermore, Vegfc^+/− heterozygous embryos showed only very subtle defects in embryonic lymphangiogenesis (Hagerling et al., 2013; Srinivasan et al., 2014). Altogether, this argues against reduced Vegfc expression in Osr1^GCE/GCE cells as the primary explanation for the lymphatic vasculature defects observed in Osr1 KO embryos.

In the embryonic dermis, fibroblasts are the main producers of ECM components and therefore key players in the formation of the ECM scaffold embedding dermal blood and lymphatic vasculature (Sennett et al., 2015). Lack of Osr1 leads to a severely disorganized ECM scaffold in the dermis of E14.5 embryos, paralleling our previous observations in skeletal muscle (Vallecillo-Garcia et al., 2017). In addition, the most prominent terms after GO analysis of deregulated genes in LECs of E13.5 Osr1^GCE/GCE embryos are related to ECM and ECM-interaction genes, suggesting that LECs themselves react to the altered ECM produced by Osr1-deficient mesenchymal cells. Decellularized ECM, but not conditioned medium, produced by E13.5 Osr1^GCE/GCE cells in vitro failed to properly sustain LEC proliferation.

Altogether, this suggests that the reduced proliferation of LECs at the tips of the migrating zone in the dorsal skin of Osr1^GCE/GCE embryos was caused by aberrant ECM deposition from Osr1-expressing mesenchymal cells rather than by deregulated expression of signaling molecules such as VEGFC. This agrees with previous reports showing that ECM composition and stiffness can modulate BEC (Yeh et al., 2012; Sack et al., 2016; LaValley et al., 2017) and LEC (Frye et al., 2018) behavior.

In addition to producing the bulk of ECM, mesenchymal cells express signaling molecules that control LEC migration (Peng et al., 2022). Conditioned media experiments suggested that Osr1⁺ mesenchymal cells produce signaling molecules necessary for LEC migration, and transcriptome-based Osr1 cell–LEC interaction analysis highlighted the CXCL12/CXCR4 axis as a possible mechanism. Of note, Cxcl12 expression is directly regulated by Osr1 and Cxcl12 is highly expressed by Osr1⁺ cells (Vallecillo-Garcia et al., 2017). In support of this idea, CXCR4^KO/KO embryos displayed similar defects in lymphatic vessel formation as Osr1^GCE/GCE embryos. However, in the dermis of E14.5 CXCR4^KO/KO embryos, LEC proliferation at the tip of the migrating front was not affected. In agreement, cxcr4 inhibition in zebrafish mainly affected LEC migration and not their proliferation (Peng et al., 2022). This suggests that the CXCL12/CXCR4 signaling pathway may contribute to LEC migration during lymphatics development in the mouse and in zebrafish, whereas LEC proliferation, specifically at the migration front, is primarily controlled by the ECM. It is noteworthy that the defects in arterial orientation and impairments in smooth muscle cell coverage observed in CXCR4 KO mouse embryos (Ara et al., 2005; Li et al., 2013, 2021) could affect the source of CXCL12 signals required for LEC migration, making it difficult to distinguish between primary and secondary effects.

However, the ECM and factors involved in endothelial-ECM interaction also play a role in controlling endothelial migration and stability (Zhang et al., 2021; Levchenko et al., 2003; Aase et al., 2007). Interestingly, Amot and genes encoding direct AMOT interactors such as Kank1, Kank3 and Flnc (Angeli and Lim, 2023) are downregulated in LECs of E13.5 Osr1^GCE/GCE embryos (Fig. S6B). We also observed a reduced formation of filopodia at the tips of migrating LECs in E14.5 Osr1^GCE/GCE embryos, altogether suggesting a defective LEC-ECM interaction. We therefore cannot exclude that the altered ECM in Osr1^GCE/GCE embryos may also in part contribute to impaired LEC migration.

In summary, our data show that mesenchymal Osr1⁺ cells play a fundamental role in LEC migration, proliferation and lymphatic vasculature assembly. Hereby, Osr1 is a key player controlling the production of critical ECM scaffold components and signaling ligands in a bimodal manner to create a microenvironment necessary for lymphatic vessel formation. This parallels the mode of action applied by Osr1⁺ cells in the developmental formation and adult regeneration of skeletal muscle (Vallecillo-Garcia et al., 2017; Kotsaris et al., 2023) and the developmental formation of lymph nodes (Vallecillo-Garcia et al., 2023), and thus suggests an overarching mechanism by which mesenchymal cells control organ formation with the transcription factor Osr1 at a key nexus.

Mice were maintained in an enclosed, pathogen-free facility, and experiments were performed in accordance with European Union regulations and under permission from the Landesamt für Gesundheit und Soziales (LaGeSo) Berlin, Germany (permission numbers ZH120, G0346/13, G0240/11, G0268-16). Mouse lines have been described previously; Osr1^GCE (Mugford et al., 2008), R26R^mTmG (Muzumdar et al., 2007), Osr1^lacZ (Stumm et al., 2018), Cxcr4^+/− (Zou et al., 1998).

As we described previously (Vallecillo-Garcia et al., 2023, 2017), tamoxifen (Sigma Aldrich) was dissolved in a 1:10 ethanol/sunflower oil mixture. For lineage tracing experiments, we bred R26R^mTmG/mTmG females to Osr1^GCE/+ males. Pregnant females were injected with 150 µl of a 20 mg ml⁻¹ tamoxifen stock. Tissues were collected at E14.5.

Embryonic tissues were fixed in 4% paraformaldehyde (PFA) for 2 h on ice. Tissues were dehydrated in two steps using 15% and 30% (w/v) sucrose (Roth) solutions before O.C.T. (Sakura) cryo-embedding in a chilled ethanol bath. Embryonic tissue was sectioned at 12 or 100 μm thickness.

Cryosections were warmed up for at least 30 min at room temperature (RT). Sections and E14.5 isolate skin tissue were blocked with 5% (v/v) horse serum (Vector Laboratories) in 0.1% (v/v) Triton X-100 (Sigma-Aldrich) PBS for 1 h at RT. Primary antibodies in blocking solution were incubated at 4°C overnight, followed by secondary antibody staining of 1 h at RT. Antibodies used are listed in Tables S2 and S4. Specimens were counterstained with 5 µg µl⁻¹ 4′, 6-diamidino-2-phenylindole (DAPI; Invitrogen) and mounted with FluoromountG (SouthernBiotech).

Isolation of E13.5 Osr1⁺ cells has been described before (Vallecillo-Garcia et al., 2023, 2017). Briefly, inner organs, tail, limbs and cranial tissue above the tongue were removed from E13.5 Osr1^GCE/+ and Osr1^GCE/GCE embryos. Next, embryonic tissue was minced using a small scissor in 1 ml high-glucose Dulbecco's modified eagle medium (DMEM, Pan Biotech) containing 10% fetal bovine serum (FBS, Pan Biotech) and 1% penicillin/streptomycin (P/S) solution. Further enzymatic digestion was performed using 0.7 mg ml⁻¹ of Collagenase (Collagenase A, Roche) in DMEM medium at 37°C for 45 min. For the isolation of E13.5 LECs and BECs, embryonic tissue was dissected from E13.5 Osr1^controls and Osr1^GCE/GCE embryos and treated as described above for E13.5 Osr1⁺ mesenchymal cells. Antibody labeling (for antibodies see Table S3) was performed for 20 min on ice.

Isolation of primary tail-derived LECs (tdLECs) has been described before (Hagerling et al., 2018). Briefly, tails of at least seven wild-type adult animals (8-30 weeks) were used. Tails were cut at the attachment site and washed twice with Hank's Balanced Salt Solution (HBSS) containing 1% P/S. Next, epidermis and dermis were isolated mechanically from the underlying musculoskeletal system. Isolated tissue was cut in pieces of 2 cm and digested for 1 h at 37°C in a HBSS solution containing 1% P/S and 2 U ml⁻¹ Dispase II (Roche). The epidermal layer was separated from the dermis using two tweezers. Collected dermal tissues were further digested enzymatically in 30 ml DMEM medium containing 10% FBS, 1% P/S and 1 mg ml⁻¹ collagenase A (Roche) for 90 min. Digested tissue was filtered through a 100 µm cell strainer and cells were collected by centrifugation at 300 g for 10 min. Cell suspensions were cultured on 0.4% gelatine-coated dishes using LEC medium that contained high-glucose DMEM, 20% FBS, 1% P/S solution, 10 µg/ml endothelial cell growth supplement (ECGS) (Thermo Fisher Scientific), 50 µM 2-mercaptoethanol, 1% non-essential amino acids solution (Thermo Fisher Scientific) and 50 µM for 4-7 days. Before FACS purification, cells were detached using Accutase (Thermo Fisher Scientific) for epitope conservation and washed once with HBSS supplemented with 1% P/S and 0.4% FBS. Antibody labeling was performed for 20 min on ice.

Before flow cytometry, cell suspensions were washed using a solution containing PBS, 0.4% FBS and 2 mM EDTA, collected by centrifugation at 300 g for 5 min and passed through a 35-µm cell strainer filter (BD Biosciences). To assess viability, cells were stained with Propidium Iodide (2 μg ml⁻¹, eBioscience) immediately before sorting or analysis.

Sorts and analyses were performed on a FACS Aria II and FACS Aria fusion (BD Biosciences). Data were collected using FACSDIVA software. Further analyses were performed using FlowJo 10 (FlowJo LLC) software. Sorting gates were defined based on unstained and fluorescence negative controls. Cells were collected into 400 µl high-glucose DMEM containing 10% FBS, and 1% P/S solution for Osr1⁺ mesenchymal cells or LEC medium.

For the production of conditioned media, 80,000 E13.5 Osr1^GCE/+ and Osr1^GCE/GCE cells isolated by FACS were plated in 24-well plates. After 100% confluence was reached, Osr1⁺ cells were cultured in 300 µl high-glucose DMEM containing only 1% P/S solution. After 24 h, media containing secreted molecules from E13.5 Osr1^GCE/+ and Osr1^GCE/GCE cells were collected and used as a conditioned medium. Conditioned media from several time points was collected.

For wound healing assays, 24-well plates were coated with 0.4% gelatine and 20,000 tdLECs (P3-P5) were seeded into two-wells culture-inserts (Ibidi^®). After 45 min two-wells culture-inserts were removed and gap closure was monitored every 2 h using a Leica DMi8 microscope.

For ECM deposition, 80,000 embryonic E13.5 Osr1^GCE/+ and Osr1^GCE/GCE cells isolated by FACS were plated on a 10 mm coverslip coated with 1 mg/ml fibronectin and cultured in DMEM containing 10% FBS and 1% P/S solution until 100% confluence. Next, cells were cultured for 3 weeks in DMEM containing 10% FBS, 1% P/S solution and 2 µ/ml ascorbic acid. ECM produced by Osr1⁺ cells was decellularized using a freeze/thaw method as described previously (Vallecillo-Garcia et al., 2017). Briefly, culture medium was aspirated and cultures were washed with PBS. Next, PBS was removed and exchanged with 300 µl distilled water. Cultures were frozen in a −80°C freezer and thawed in a 37°C water bath. Remaining water was careful aspirated by pipetting and tdLECs were seeded on the dECM. To assess tdLECs proliferation cultures on dECM, they were cultured for 48 h in LEC medium.

X-Gal staining of whole lymph nodes or the dermis of the ear was documented using a Zeiss SteREO Discovery V12 stereomicroscope. Confocal images of immunolabeled sections were taken using the confocal laser scanning microscope systems LSM710, LSM810 (Zeiss) or Leica DMi8 microscopes. Images were captured using Zen 2010 (Zeiss) and LAS Life System (Leica). For quantifications shown in Fig. 2A, at least three consecutive 100 µm sections were quantified per embryo. In Fig. 2B, four to five regions along the rostro-caudal axis of the back from four to five consecutive 100 µm sagittal sections were quantified per embryo. VEGFR3 expression was quantified using the ImageJ algorithm ‘integrated density’ as the sum of pixels per region of VEGFR3 staining. For normalization, each value was normalized to the average value obtained from wild-type sections. In Fig. 2D,E, the lumbar and cervical regions were quantified separately, assessing the distance of the tips to the center of the avascular midline in at least 20-30 lymphatic vessels per embryo. In Fig. 2G,H and Fig. S2F, branching points, vessel thickness and vessel density were quantified in three different regions of 1 mm² skin per embryo. In Fig. 2I, the number of filopodia per PROX1⁺ LEC was quantified in six different regions at the tip of the migrating zone per embryo. In Figs 6E and 7F, PROX1⁺/KI67⁺ double positive cells were quantified in three different regions at the tip and three different regions at the ventral side per embryo. In Fig. 6G,H, Prox1^High cells were quantified per region, choosing three to four regions of 1 mm² per biological replicate (cells or medium derived from one embryo). In each region 20-500 cells per region were quantified. In Fig. 2A,B,D,E,G-I, Fig. S2F, Fig. 6E,G,H and Fig. 7F, measurements from a single embryo are depicted in the same color.

Total RNA extraction from FACS isolated cells was performed using Direct-zol^TM RNA MicroPrep (Zymo Research) following the manufacturer's protocol. Reverse transcription was conducted using the M MuLV Reverse Transcriptase Kit (Biozym). Relative gene expression analyses were performed using GoTaq^® qPCR kit (Promega) or Blue S'Green qPCR kit (Byozim) on a 7900HT Real Time PCR system or QuantStudio 7 Flex Real-Time-PCR-System (Applied Biosystems). Primer sequence information is provided in Table S5. Data were acquired and analyzed using SDS 2.0 and QuantStudio^TM Real-Time PCR software (Applied Biosystems).

To obtain the total RNA amount necessary for RNA-seq, we pooled LECs from eight to ten E13.5 Osr1^controls embryos and eight to ten Osr1^GCE/GCE embryos per sample. E13.5 LECs were isolated by FACS as described above. Total RNA was isolated using Direct-zol RNA Microprep (ZYMO RESEARCH) following the manufacturer's protocol. Total RNA was measured using a Qubit 4 Fluorometer and RNA quality was assessed using an Agilent RNA 6000 Nano kit before library preparation. Library preparation was conducted according to Illumina instructions TruSeq Library Preparation Kit V2. Next, libraries were subjected to high-throughput sequencing using an HiSeq 2500 device. Obtained fastq data were further analyzed using the platform Galaxy Europe (https://usegalaxy.eu). We obtained and mapped 53-81 million reads using STAR (Dobin et al., 2013) against the genome of Mus musculus version mm10. Quantification of aligned reads at the gene level was conducted using featureCounts (Liao et al., 2014). Differential gene expression analysis was performed using DESeq2 (Love et al., 2014). Transcript per million (TPM) abundances were calculated using normalized gene counts from DESeq2 analysis of E13.5 Osr1^GCE/GCE LEC samples. Genes with an absolute log2 fold change of ≥0.3, a Benjamini-Hochberg adjusted P-value (padj)<0.05, and TPM value>1 were considered as being differentially expressed between E13.5 Osr1^cantrols and Osr1^GCE/GCE LECs. GO analysis was performed using Enrichr (Chen et al., 2013). The ligand-receptor interaction network was drawn using graph-tool v2.45_5 (https://graph-tool.skewed.de/) with a hierarchical edge bundling. Raw fastq and count data were uploaded to the Gene Expression Omnibus (GEO) database under the accession number GSE269397.

Unpaired, two-tailed Student's t-test and one-way ANOVA with Dunnett's post-hoc comparison were performed using Prism 8 (GraphPad) software. Error bars in all figures, including supplementary information, represent the mean±standard error of the mean (s.e.m.). In Fig. 6G, a paired, two-tailed Student's t-test was used due to the differential proliferation showed by tdLECs isolated from different experiments.

We gratefully acknowledge Andrew P. McMahon (Keck School of Medicine of USC, USA) and Andreas Kispert (Hannover Medical School, Germany) for providing Osr1^GCE and R26^mTmG mouse lines. We thank Carmen Birchmeier and Ines Lahmann (Max Delbrück Center for Molecular Medicine, Berlin, Germany) for providing Cxcr4^+/− mice. We thank Uta Marchfelder and Erwin Weiß (Max Planck Institute for Molecular Genetics, Germany) for flow cytometry support. We thank Stefan Mundlos, Nobert Brieske and Thorsten Mielke for their support (Max Planck Institute for Molecular Genetics, Berlin, Germany). We are grateful to Olaf Penack (Charité, Berlin, Germany) for his support.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202747.reviewer-comments.pdf