GATA2 regulates blood/lymph separation in a platelet-dependent and lymphovenous valve-independent manner

Abstract

Lymphatic vessels collect interstitial fluid, immune cells and digested lipids and return these bodily fluids to blood through two pairs of lymphovenous valves (LVVs). Like other cardiovascular valves LVVs prevent the backflow of blood into the lymphatic vessels. In addition to LVVs platelets are necessary to prevent the entry of blood into the lymphatic vessels. Platelet thrombi are observed at LVVs suggesting that LVVs and platelets function in synergy to regulate blood/lymphatic separation. However, whether platelets can regulate blood/lymph separation independently of LVVs is not known. We have determined that the lymphatic vasculature-specific deletion of the lymphedema-associated transcription factor GATA2 results in the absence of LVVs without compromising blood/lymph separation. In contrast, deletion of GATA2 from both lymphatic vasculature and hematopoietic stem cells results in the absence of LVVs, reduced number of platelets and blood-filled lymphatic vasculature. These observations suggest that GATA2 promotes blood/lymph separation through platelets. Furthermore, LVVs are the only known sites of interaction between blood and lymphatic vessels. The fact that blood is able to enter the lymphatic vessels of mice lacking LVVs and platelets indicates that under these circumstances the lymphatic and blood vessels are connected at yet to be identified sites.

Introduction

Lymphatic vasculature is important for maintaining interstitial fluid homeostasis, absorbing digested lipids, and for immune response. Defects in the lymphatic vasculature results in lymphedema or by other lymphatic disorders such as lymphatic malformation. These lymphatic disorders are characterized by symptoms such as the swelling of limbs, chylothorax and chylous ascites. Lymphatic endothelial cells (LECs) originate predominantly from the embryonic veins ¹. At embryonic day E10 a sub-population of venous endothelial cells express the homeobox transcription factor PROX1 ^1,2. These are the LEC progenitors that give rise to most of the lymphatic vasculature, although additional sources make minor contribution ^3,4. Although LECs originate from the veins, the lymphatic vasculature is devoid of RBCs. This process, known as blood/lymph separation, is regulated by two-pairs of bilaterally located lymphovenous valves (LVVs) and platelets ^5–9. Like all valves LVVs function as a mechanical gate to regulate the unidirectional flow of fluid. LVVs permit lymph return to blood circulation while preventing the backflow of blood into the lymphatic vessels. In addition, the transmembrane glycoprotein podoplanin that is expressed in LECs interacts with and activates C-type lectin-like receptor 2 (CLEC2) in platelets to form transient thrombi at LVVs to prevent the entry of RBCs into the lymphatic vasculature ⁷. Mice lacking podoplanin or CLEC2 develop blood-filled lymphatic vessels although they have LVVs ^6,7,10,11. Therefore, platelets are thought to function at LVVs to regulate blood/lymph separation. However, it is not known if platelet activity is required at additional sites to prevent blood/lymph mixing.

The zinc finger transcription factor GATA2 is an important regulator of hematopoietic cells and endothelial cells ¹². Heterozygous loss-of-function mutations in GATA2 are associated with congenital lymphedema ¹³. Gata2−/− mice die at embryonic day E10 just as LECs are starting to be specified ¹⁴. Conditional deletion of Gata2 from all endothelial cells results in severely edematous embryos with small blood-filled lymph sacs ^8,15–18. Conditional deletion of Gata2 from LECs using Lyve1-Cre or Prox1-CreERT2 results in the absence of LVVs, and mispatterned lymphatic vessels that lack lymphatic valves (LVs) ^15,16,18. Intriguingly, Lyve1-Cre;Gata2^f/f, but not Prox1-CreERT2;Gata2^f/f embryos develop blood-filled lymphatic vessels ¹⁸. We addressed the reason for this important difference in phenotype by using additional endothelial and hematopoietic cell specific Cre lines. Our findings reveal that platelets can regulate blood/lymph separation in an LVV-independent manner.

Results

GATA2 regulates blood/lymph separation in an LVV-independent manner

We used 6 Cre lines to investigate the tissue and time-specific roles of GATA2 (Table 1). First, we used a novel BAC transgenic Tg(Prox1-Cre) to delete Gata2 ¹⁹. We compared the phenotypes of Tg(Prox1-Cre); Gata2^f/f mice with that of Lyve1-Cre;Gata2^f/f and Prox1-CreERT2;Gata2^f/f animals. Consistent with our previous report, at E16.5 both Lyve1-Cre;Gata2^f/f and Prox1-CreERT2;Gata2^f/f (tamoxifen at E10.5) embryos developed edema and lacked LVVs, but only Lyve1-Cre;Gata2^f/f embryos developed blood-filled lymph sacs and dermal lymphatic vessels (Figure 1A-C, E-G). E16.5 Tg(Prox1-Cre); Gata2^f/f embryos were edematous and lacked LVVs, but were devoid of blood cells in the lymph sacs (Figure 1D, H). Unlike Lyve1-Cre;Gata2^f/f embryos that died at E16.5, Tg(Prox1-Cre); Gata2^f/f mice were born alive, able to suckle and devoid of chylous ascites or blood-filled dermal lymphatic vasculature phenotypes (Figure 2A, B). Instead, the dermal lymphatic vessels of Tg(Prox1-Cre); Gata2^f/f mice were filled with chyle. This phenotype was likely caused by the backflow of chyle into lymphatic vessels that lacked LV (Figures 1H, 2E-F).

Table 1:

Tissues targeted by the various Cre lines used in this study.

	Arteries	Veins	Sinusoidal ECs	LECs	HSCs
Lyve1-Cre	Subset	Subset	Yes	Yes	Subset
Tg(Prox1-Cre)	Subset	Subset	No	Yes	No
Prox1+/GFPCre	Subset	Subset	No	Yes	Did not analyze
Prox1-CreERT2	No	No	No	Yes	No
Vav-Cre <E12.5	No	No	Did not analyze	No	Yes
Vav-Cre >E12.5	Subset	Subset	Did not analyze	Yes	Yes
Cdh5-CreERT2 TM<E10.5	Yes	Yes	Yes	Yes	Yes
Cdh5-CreERT2 TM>E12.5	Yes	Yes	Yes	Yes	No

Figure 1:

Deletion of Gata2 with Lyve1-Cre, but not Prox1-CreERT2 or Tg(Prox1-Cre) results in blood-filled lymphatic vasculature.

(A-D) E16.5 embryos in which Gata2 was deleted using the indicated Cre lines. While using Prox1-CreERT2 tamoxifen was intraperitoneally administered to pregnant dams at 3 mg/40 gm bodyweight at E10.5. Arrows indicate edema in the dermis and arrowhead points to blood-filled dermal lymphatic vessels.

(E-H) The embryos were frontally sectioned and the LVVs (arrows) were analyzed using the indicated antibodies. All the conditional homozygous embryos lacked LVVs and had dilated lymph sacs (LS). However, only the Lyve1-Cre;Gata2^f/f embryos had blood-filled LS.

Statistics: n=5 embryos per genotype.

Figure 2:

Tg(Prox1-Cre);Gata2^f/f mice lack lymphatic valves and develop milk-filled lymphatic vasculature phenotype.

(A-D) The dermal lymphatic vessels of newborn Tg(Prox1-Cre);Gata2^f/f pups was filled with milk as seen near the naval (B) and the eye (D).

(E, F) Lymphatic valves (E, arrows) were seen in the dermis of E16.5 control mice, but not in Tg(Prox1-Cre);Gata2^f/f littermates (F).

Statistics: n=3 embryos per genotype per stage.

Based on the above observations we conclude that the blood/lymph separation defect in Lyve1-Cre;Gata2^f/f mice is not caused by the absence of LVVs.

Platelet development is defective in Lyve1-Cre;Gata2^f/f embryos

As mentioned in the Introduction platelets and podoplanin are critical for blood/lymph separation. Hence, we tested whether platelets or podoplanin is defective in Lyve1-Cre;Gata2^f/f embryos. In E11.5 wild type embryos podoplanin is expressed in LECs lining the developing lymph sacs. A few blood cells were observed within the lymph sacs of wild type embryos as reported previously ^8,20. The lymph sacs of E11.5 Lyve1-Cre;Gata2^f/f embryos were dilated due to increased number of blood cells. Podoplanin expression was largely intact in the mutant LECs although chimerism was occasionally noted (Figure 3B, arrow).

Figure 3:

Platelets are reduced in Lyve1-Cre;Gata2^f/f embryos.

(A, B) At E11.5 control embryos had some blood cells within the developing lymph sacs (A, LS). (B) The LS of Lyve1-Cre;Gata2^f/f embryos had substantially higher amount of blood cells. Podoplanin was expressed in both control and Lyve1-Cre;Gata2^f/f LECs although chimerism in podoplanin expression was occasionally observed in Lyve1-Cre;Gata2^f/f embryos (B, arrow).

(C, D) CD41⁺ platelets were observed in the internal jugular vein (IJV), LS and at the opening of the developing LVVs of E12.0 control and Lyve1-Cre;Gata2^f/f embryos. However, the platelet numbers appear to be reduced in Lyve1-Cre;Gata2^f/f embryos.

(E, F) CD41⁺ megakaryocytes and platelets were reduced in the liver of E12.0 Lyve1-Cre;Gata2^f/f embryos.

(G) CD41⁺ megakaryocytes were quantified from representative sections collected from E12.0 Lyve1-Cre;Gata2^f/f embryos. Each dot represents an individual embryo.

(H) The livers of E14.5 Lyve1-Cre;Gata2^f/f embryos (arrow) were substantially smaller than controls.

Statistics: n=3 embryos per genotype.

LVVs start forming at E12.0 as indicated by the increased expression of PROX1 in a subset of venous endothelial cells known as LVV-forming endothelial cells (LVV-ECs) ^8,9. LVV-ECs were observed in both E12.0 wild type and Lyve1-Cre;Gata2^f/f embryos (Figure 3C, D, arrow). Furthermore, CD41⁺ platelets could be seen aggregating close to LVV-ECs of wild type embryos. Platelets could also be seen inside the lymph sacs and veins. In contrast, CD41⁺ cells appeared to be reduced in the vicinity of LVV-ECs, and inside the lymph sacs and veins of Lyve1-Cre;Gata2^f/f embryos. Hematopoietic stem cells (HSCs) develop from the hemogenic endothelium of the aorta-gonad-mesonephros (AGM) region of embryos at ~E9.5 ²¹. From this location HSCs migrate to the liver where they proliferate and differentiate further until late embryonic stage. GATA2 is a critical regulator of HSC development ²¹, and Lyve1-Cre is active in a subset of hematopoietic cells ^22,23. Therefore, we analyzed the liver of E12.0 control and Lyve1-Cre;Gata2^f/f embryos and determined that platelets and megakaryocytes were strikingly reduced in the Gata2-conditional homozygous embryos (Figure 3E-G). Furthermore, the livers of E14.5 Lyve1-Cre;Gata2^f/f embryos were strikingly smaller when compared to their control littermates, suggesting a reduction in the hematopoietic cells (Figure 3H). No obvious differences in platelets was observed in the livers of Prox1-CreERT2;Gata2^f/f and Tg(Prox1-Cre); Gata2^f/f embryos (Supplementary Figures 1 and 2).

We performed lineage tracing using Lyve1-Cre, Prox1-CreERT2 and Tg(Prox1-Cre). All three Cre lines efficiently labelled LVVs and lymph sacs of E16.5 embryos (Figure 4A-C). Hepatocytes were labelled by Prox1-CreERT2 and Tg(Prox1-Cre) and sinusoidal endothelial cells were labeled by Lyve1-Cre. Importantly, platelets in the liver were labeled only by Lyve1-Cre (Figure 4D-F). Thus, we conclude that the platelet count is reduced in Lyve1-Cre;Gata2^f/f embryos due to Lyve1-Cre activity in platelet progenitors or sinusoidal endothelial cells. We used additional approaches to distinguish the role of GATA2 in platelet progenitors and sinusoidal endothelial cells.

Figure 4:

Lyve1-Cre, but not Prox1-CreERT2 or Tg(Prox1-Cre), is active in megakaryocytes.

Lineage tracing was performed by breeding the Cre lines with R26^+/tdTomato reporter. Prox1-CreERT2; R26^+/tdTomato embryos were exposed to tamoxifen at E10.5.

(A-C) At E16.5 all the Cre lines had efficiently labelled the lymph sacs (LS) and the lymphovenous valves (arrows). While Lyve1-Cre and Tg(Prox1-Cre) had also labelled the superior vena cava (SCV) and the venous valves (arrowheads), Prox1-CreERT2 (tamoxifen at E10.5) did not. This is since the venous valves develop only at E14.5 (Geng et al., 2016).

(D-F) CD41⁺ megakaryocytes (D, arrowheads) and sinusoidal endothelial cells of the liver were labelled by Lyve1-Cre. In contrast, hepatocytes, but not megakaryocytes or sinusoidal endothelial cells were labelled by Prox1-CreERT2 and Tg(Prox1-Cre).

Statistics: n=3 per genotype.

Deletion of Gata2 from HSCs results in blood-filled lymphatic phenotype

We used additional Cre lines to dissect the role of GATA2 in platelet progenitors and sinusoidal endothelial cells. Vav-iCre is commonly used to delete genes from hematopoietic cells ^24,25. We performed lineage tracing experiments with Vav-iCre and determined that LVV-ECs were not labeled by Vav-iCre at E12.5 (Figure 5A, arrow). However, most endothelial cells, including LECs and LVV-ECs were labeled by Vav-iCre at E16.5 (Figure 5B, arrows). We generated E12.5 Vav-iCre;Gata2^f/f embryos to study the HSC-specific role of GATA2. Platelets are reduced in the vicinity of LVV-ECs, veins and liver as anticipated (Figure 5C-F). Importantly, the lymph sacs of these mice were filled with blood.

Figure 5:

HSC-specific function of GATA2 is required for blood/lymph separation.

(A, B) Lineage tracing was performed by breeding Vav-iCre with R26^+/tdTomato reporter. The resulting embryos were analyzed at E12.5 (A) or E16.5 (B). The developing LVVs were not labelled by Vav-iCre at E12.5 (A, arrow). However, LVVs were strongly labelled at E16.5 (B, arrows).

(C-F) The LVVs and liver of Vav-iCre;Gata2^f/f embryos were analyzed at E12.5. CD41⁺ platelets were reduced in the veins, LVV (arrow) and liver of Vav-iCre;Gata2^f/f embryos.

Abbreviations: LS, lymph sac; IJV, internal jugular vein; SVC, superior vena cava.

Statistics: n=3 per genotype.

As HSCs originate from the CDH5⁺ hemogenic endothelial cells of dorsal aorta we additionally used Cdh5(PAC)-CreERT2 to investigate the HSC-specific activity of GATA2. We generated E16.5 Cdh5(PAC)-CreERT2;Gata2^f/f embryos in which Gata2 was deleted in a time-specific manner at E10.5, E11.5 or E12.5. HSC development from dorsal aorta is gradually reduced during this time window. However, Cdh5(PAC)-CreERT2 activity remains stable in endothelial cells. Deletion of Gata2 at any of these stages resulted in edema (Figure 6A-D) and in the absence of LVVs at E16.5 (Figure 6E-H). However, deletion of Gata2 at E10.5 or E11.5, but not at E12.5, resulted in blood-filled lymph sacs (Figure 6A-D, I-L). Importantly, platelets were substantially reduced in the livers of E16.5 Cdh5(PAC)-CreERT2;Gata2^f/f embryos that were treated with tamoxifen at E10.5 or E11.5 when compared to those that were treated with tamoxifen at E12.5 (Figure 6M-P and Supplementary Figures 3).

Figure 6:

Blood-filled lymphatics phenotype correlates with reduced number of platelets in Cdh5(PAC)-CreERT2;Gata2^f/f embryos.

(A-D) Cdh5(PAC)-CreERT2;Gata2^f/f embryos were exposed to tamoxifen (TM) at E10.5, E11.5 or E12.5 and analyzed at E16.5. Gata2-conditional homozygous embryos developed edema (B-D) irrespective of the time of Gata2 deletion. A subset of Cdh5(PAC)-CreERT2;Gata2^f/f embryos that were exposed to tamoxifen at E10.5 or E11.5 developed blood-filled dermal lymphatic vessels (B, C, arrows)

(E-P) Control and Gata2-conditional homozygous embryos were frontally sectioned and the region where LVVs normally form (E-L) and the livers (M-P) were analyzed using the indicated antibodies. LVVs were seen in control embryos (E, I, arrows). In contrast, Gata2-conditional homozygous embryos uniformly lacked LVVs (F-H, J-L). A subset of Cdh5(PAC)-CreERT2;Gata2^f/f embryos that were exposed to tamoxifen at E10.5 or E11.5 developed blood-filled lymph sacs (F, G, J, K) and had reduced number of platelets and megakaryocytes in the liver (N, O). However, Cdh5(PAC)-CreERT2;Gata2^f/f embryos were exposed to TM at E12.5 did not develop blood-filled lymph sacs (H, L) and appeared to have normal number of platelets in the liver (P).

Abbreviations: LS, lymph sac; IJV, internal jugular vein; SVC, superior vena cava.

Statistics: n=7 embryos per group.

Finally, we used Prox1^+/GFPCre to delete Gata2 from LECs. Prox1^+/GFPCre embryos are heterozygous for Prox1 and are devoid of LVVs ^8,9,26. In contrast to mice lacking Gata2, LVV-EC differentiation does not happen in Prox1^+/GFPCre embryos. We anticipated that the Prox1^+/GFPCre;Gata2^f/f embryos will recapitulate the phenotype of Prox1-CreERT2;Gata2^f/f and Tg(Prox1-Cre); Gata2^f/f animals. Surprisingly, although both E14.5 Prox1^+/GFPCre;Gata2^+/f and Prox1^+/GFPCre;Gata2^f/f embryos lacked LVVs as anticipated, the lymph sacs of Prox1^+/GFPCre;Gata2^f/f embryos were filled with blood (Figure 7A, B). PROX1 was identified as a regulator of HSC development ²⁷. Consistent with that report platelets appeared to be substantially reduced in the veins and liver of Prox1^+/GFPCre;Gata2^f/f embryos, indicating that the blood-filled lymphatic phenotype is likely caused by a defect in platelet development.

Figure 7:

Blood-filled lymphatics phenotype correlates with reduced number of platelets in Prox1^+/GFPCre;Gata2^f/f embryos.

LVV-forming region (A, B) and the liver (C, D) of E14.5 Prox1^+/GFPCre;Gata2^+/f and Prox1^+/GFPCre;Gata2^f/f littermates were analyzed. LVVs were absent in both models (A, B, arrows). However, Blood-filled lymph sacs was observed only Prox1^+/GFPCre;Gata2^f/f embryos (B). Additionally, CD41⁺ platelets and megakaryocytes appeared to be reduced in the livers of Prox1^+/GFPCre;Gata2^f/f embryos (D).

Abbreviations: LS, lymph sac; IJV, internal jugular vein; SVC, superior vena cava.

Statistics: n=3 embryos per genotype.

In summary, deletion of Gata2 by Lyve1-Cre, Vav-iCre, Cdh5(Pac)-CreERT2 (tamoxifen at E10.5 or E11.5) or Prox1^+/GFPCre results in blood-filled lymphatics phenotype. In these mutants the blood-filled lymphatics phenotype correlates with a reduction in the number of platelets in the liver and circulation. In contrast, deletion of Gata2 by Prox1-CreERT2 or Tg(Prox1-Cre) abolishes LVVs without causing blood-filled lymphatics phenotype. Previous reports have shown that platelet aggregates at LVVs likely play a critical role in regulating blood/lymph separation ⁵. Together, these data strongly indicate that Gata2 regulates blood/lymph separation in a platelet-dependent and LVV-independent manner.

Discussion

LVVs are the gatekeepers of the lymphatic vasculature. Proper functioning of LVVs is critical to prevent the entry of blood into the lymphatic vasculature. However, our data strongly supports the idea that blood/lymph separation can take place in the absence of LVVs. This possibility is also supported by the phenotype of Prox1^+/− embryos. Most Prox1^+/− embryos that are devoid of LVVs undergo normal blood/lymph separation ^8,9,26. One could argue that both lymph drainage and blood backflow are prevented in Prox1^+/− embryos due to the absence of LVVs. Surprisingly, Prox1^+/−;Vegfr3^+/− embryos that also lack LVVs develop blood-filled lymph sacs suggesting that blood could enter the lymph sacs through yet to be identified sites ⁹. Prox1^+/−;Vegfr3^+/− embryos have reduced expression of podoplanin in LECs. Therefore, LEC/platelet interaction is likely compromised in Prox1^+/−;Vegfr3^+/− embryos resulting in blood/lymph mixing.

Despite our best efforts we were unable to identify direct connection between blood vessels and lymph sacs in Lyve1-Cre;Gata2^f/f embryos. The integrity of the jugular vein also appeared to be normal. Nevertheless, LECs migrate from veins and capillaries at multiple locations ^1,28–30. Additionally, blood vessels pass through the growing lymph sacs at various locations. It is conceivable that thrombi fail to form in the absence of platelets or podoplanin at sites where blood and lymphatic vessels interact. This could lead to bleeding into lymphatic vessels. Advanced histological approaches such as serial block-face scanning electron microscopy might be able to reveal these sites in the future ³¹.

Finally, our work highlights the limitations of Cre lines that are in use to study lymphatic vascular development and maintenance (Table 1). HSCs are targeted by Lyve1-Cre and Prox1^+/GFPCre; HSCs and blood vasculature are targeted by Cdh5(PAC)-CreERT2; sinusoidal endothelial cells are targeted by Lyve1-Cre; hepatocytes are targeted by Tg(Prox1-Cre), Prox1-CreERT2 and Prox1^+/GFPCre. Thus, Cre lines that specifically target the lymphatic vasculature are missing. Intersectional genetic models, in which tissue-specific Dre and Cre drivers are sequentially activated, may be able to at least partially fulfill this need ^32,33.

Materials and Methods

Mice

Tg(Prox1-Cre) ¹⁹, Prox1-CreERT2 ¹, Lyve1-Cre ²², Vav-iCre ²⁵, Cdh5(PAC)-CreERT ³⁴, Gata2^{f/f 35}, R26+/tdTomato ³⁶ mice were described previously. Mice were maintained in C57BL6 or C57BL6/NMRI mixed backgrounds. All mice were housed and handled according to the institutional IACUC protocols. Tamoxifen dissolved in peanut oil was intraperitoneally administered to pregnant dams at a concentration of 3 mg/40 gm body weight.

Immunohistochemistry of tissues

Immunohistochemistry on paraffin embedded sections was performed according to our previously published protocols ¹⁸. Briefly, freshly collected embryos were washed in 1× PBS and then fixed the embryos in 4% paraformaldehyde (PFA) overnight at 4°C. Subsequently, the embryos were washed three times (10 min each) in ice cold PBS at 4°C, followed by incubation in 15% sucrose overnight at 4°C and then again in 30% sucrose at 4° C until fully submerged in the solution. The embryos were cryo- embedded in OCT solution (Sakura). Then 12 μm thick sections were prepared using a cryotome (Thermo Fisher Scientific, model: HM525 NX) and immunohistochemistry was performed using the indicated antibodies. E11.5 embryos were sectioned in a transverse orientation and E12.0-E16.5 embryos were sectioned frontally. Around eight consecutive sections were analyzed to determine the presence or absence of LVVs and VVs and to observe CD41 positive signal in the liver.

Antibodies

Primary antibodies for immunohistochemistry were: rabbit anti-PROX1 (11–002, Angiobio), goat anti-human PROX1 (AF2727, R&D Systems), goat anti-mouse VEGFR3 (AF743, R&D Systems), rat anti-mouse CD31 (553370, BD Pharmingen), hamster anti-mouse podoplanin (127401, BioLegend), rat anti-mouse TER-119 (116201, BioLegend), goat anti-mouse VEGFR2 (AF644, R&D Systems), and rat anti-mouse CD41 (133901, BioLegend).

Secondary antibodies for immunohistochemistry were: Cy3-conjugated donkey anti-rabbit, and Cy5-conjugated donkey anti-rat antibodies (Jackson ImmunoResearch Laboratories), and Alexa 488-conjugated donkey anti-goat, Alexa 488-conjugated goat anti- chicken and Alexa 488-conjugated donkey anti-rat (Life Technologies).

Data Availability

No datasets were generated for this work.