Postnatal lymphatic partitioning from the blood vasculature in the small intestine requires fasting-induced adipose factor

We found that newborn mice, homozygous for a null allele of Fiaf (9) and with a mixed 129/SvJ:C57BL/6J (B6) genetic background, are born at a normal Mendelian ratio (26.5%). Although they appear grossly normal at birth, the vast majority (41/44; 93%) dies by 3 weeks of age. In contrast, their heterozygous littermates live through adulthood. Interestingly, Fiaf−/− mice can be rescued from the lethal phenotype by backcrossing them one generation onto a B6 background: only 4.1% (5/121) of these B6-enriched Fiaf−/− animals die by postnatal day 15 (P15), with the remainder surviving through adulthood.

Fiaf−/− mice with the original hybrid 129/Sv:B6 background develop shoulder girdle wasting and protuberant abdomens during the first postnatal week of life (Fig. 1A). The small intestine, but not the colon, assumes a bright red color beginning at P3 (Fig. 1B). By P7, small intestinal villi are filled with blood (Fig. 1C). The protuberant abdomen reflects a marked increase in the small intestine-to-body weight ratio (3.5-fold compared with wild-type and heterozygous littermates at P6–P10; P < 0.0001; n = 11–20 animals per group) (Fig. 1D).

P7 Fiaf−/− animals are not anemic (blood hemoglobin, 16.0 ± 0.7 g/dl versus 15.3 ± 0.6 g/dl in wild-type and Fiaf+/− littermates; P = 0.32, n = 6–14 mice per group). In addition, their intestinal contents are not bloody. However, histologic analysis of P7 Fiaf−/− intestines revealed blood in dilated, thin-walled mesenchymal vessels that stain positive for lymphatic vessel endothelial hyaluronan receptor 1 (Lyve-1), a lymph-specific biomarker that is not expressed in blood vascular endothelial cells (13) (Fig. 2A and B). This dilatation is not accompanied by significant differences in levels of intestinal expression of Lyve-1 and another well established lymphatic marker, podoplanin, in knockout compared with wild-type animals [see supporting information (SI) Figs. 6 and 7].

Blood-filled lymphatic vessels could arise from aberrant communications between lymphatic and blood vascular systems or, less likely, generalized extravasation from the blood vascular system and subsequent uptake of blood cells by neighboring lymphatics. Three observations argue against extravasation. First, P7 Fiaf−/− mice and their +/− and +/+ littermates were given a single i.v. injection of FITC-conjugated Bandeiraea simplicifolia 1 lectin (BS-1) (n = 4 animals per group). BS-1 binds to α-d-galactosyl- and N-acetyl-α-d-galactosaminyl-containing glycans expressed by both blood and lymphatic vascular endothelial cells (11, 14). In normal mice, i.v. administered FITC-BS-1 labeled endothelial cells in the small intestine's subepithelial blood microvasculature, but not the endothelium that lines its Lyve-1-positive lymphatic system, consistent with the fact that venous contents do not normally flow into lymphatics. In contrast, FITC-BS-1 labeled both lymphatic and blood endothelial cells in the intestines of Fiaf−/− animals (Fig. 2, compare C–E with F–H). Importantly, we did not observe BS-1 staining outside of vessels in the Fiaf−/− intestinal mesenchyme, even though this lectin binds to epitopes commonly found both within and outside of blood vessels. Second, transmission electron microscopic studies of blood vascular endothelial cell morphology revealed that, despite the somewhat roughened appearance of their apical surfaces, tight junction morphology is similar in Fiaf−/− and wild-type littermates (see SI Fig. 8). Third, immunohistochemical assessments of the envelopment of the intestinal blood vasculature by smooth muscle cells (see SI Fig. 9), and TUNEL assays of endothelial apoptosis (see SI Fig. 10) in P7 Fiaf−/− and wild-type littermates, failed to reveal any differences.

Based on these results, we concluded that the presence of blood within the small intestinal lymphatic microvasculature of Fiaf−/− mice was due to a functional lymphatico-venous partitioning abnormality. However, we were not able to definitively detect anastamoses between Lyve-1-positive lymphatic and Lyve-1-negative/CD31-positive blood vessels, despite surveying >500 sections prepared from the small intestines of P2–P17 knockout mice.

We did not observe blood in lymphatics until P2–P3 in Fiaf−/− pups: embryonic day 18 (E18) and P0 Fiaf−/− animals exhibited identical histology as wild type (n = 5 per group) (see Fig. 3), indicating that embryonic lymphatico-venous partitioning is intact at birth. Consistent with these findings, we documented a marked increase of intestinal Fiaf expression on the first postnatal day in wild-type mice, peaking at P2 (17.9 ± 1.5-fold relative to E18) (Fig. 4A). These results indicate that functional separation of the lymphatic and blood vasculature is actively preserved during postnatal life through a process that requires Fiaf.

The “conjoined” lymphatico-venous abnormality extended to the small intestinal mesentery. Lyve-1 staining of paraffin-embedded sections of mesentery revealed blood-filled and dilated lymphatics in P7–P12 Fiaf−/− animals but not in +/− or +/+ littermates (n = 5 per group) (SI Fig. 11). Gastric gavage with Nile red, a neutral lipid-binding fluorescent dye that is absorbed into the lymphatic system from the gut lumen, disclosed ectatic lymphatic vessels that resembled “strings of sausages” as well as microscopic leakage of the dye, but not erythrocytes, into the surrounding mesentery. Neither of these mesenteric phenotypes was observed in age-matched wild-type littermates (see SI Fig. 11). We interpret the Nile red results as indicating that there are low levels of leakiness in the lymphatics of Fiaf-deficient mice: this leakiness is not of sufficient magnitude to produce chylous ascites but is sufficient to be detected with this sensitive fluorescent dye.

Intravenous injection of FITC-BS-1, intradermal injection of Evans blue dye (a commonly used lymphatic tracer that binds albumin), plus Lyve-1 immunohistochemical studies of E18–P12 Fiaf−/− mice and their age-matched wild-type and Fiaf+/− littermates disclosed that the lymphatic vascular defects associated with Fiaf deficiency did not extend to involve central lymphatics (see SI Fig. 12) or the lymphatic microvasculature of the skin (see SI Fig. 13). The liver, which also expresses high levels of Fiaf, developed normal vascular features (data not shown). We did note dilated blood-filled lymphatic vessels in the colons of the small subset of Fiaf−/− mice that survived until P17 (see SI Fig. 14), emphasizing the evolving nature of this defect within the gut. Fiaf-deficient mice do not have thoracic effusions or edematous and hemorrhagic skin (n = 41), providing additional evidence that the abnormality is gut-specific.

We have shown previously that the intestinal microbiota is important for postnatal development of the intestine's blood vascular system (15) and that Fiaf expression is higher in the villus epithelium of germ-free compared with conventionally raised wild-type mice (9, 11). Therefore, we rederived 129/Sv:B6 Fiaf−/− mice as germ-free to examine whether expression of this abnormal lymphatic phenotype was elicited by the microbial colonization of the postnatal gut. We found that the lymphatico-venous abnormality and postnatal lethality were fully penetrant in these germ-free animals and indistinguishable from what was noted in their syngeneic, conventionally raised Fiaf−/− counterparts. Thus, the impact of Fiaf deficiency on postnatal development of the intestinal lymphovascular system occurs independent of the gut microbiota.

To identify additional molecular correlates of the lymphatico-venous abnormality, we performed quantitative RT-PCR (qRT-PCR) studies of total small intestinal RNA isolated from P7 Fiaf−/− mice and their wild-type littermates (n = 4 animals per group, each analyzed individually). There were modest (≤2-fold) reductions in expression of known biomarkers of lymphatic vascular development (Vegf-C and Neuropilin 2, P < 0.05) whereas, as noted above, levels of mRNAs encoding other lymphatic markers (Lyve-1, podoplanin, and Vegr 3) did not change (see SI Fig. 6). These results, together with our immuohistochemical analyses, indicate that the observed lymphatico-venous malformations are manifestations of a specific rather than a global change in the nature of lymphatic vessels.

Prox1 is a homeobox gene essential for normal embryonic development (16–18): Prox1−/− mice die at approximately E14.5 without any lymphatics (1, 16). Heterozygotes do not survive unless they are on an NMRI-enriched genetic background (16), and only a small proportion of these animals live past the postnatal period (19). Surviving Prox1+/− mice have dilated and leaky intestinal lymphatics but no aberrant lymphatico-venous connections. This lymphatic phenotype is most pronounced in the small bowel (19). These observations prompted us to assess the relationship between Fiaf and Prox1 expression. qRT-PCR studies revealed that the perinatal induction of Prox1 parallels that of Fiaf in wild-type mice (Fig. 4B). Fiaf deficiency is associated with a 3-fold diminution in small intestinal Prox1 expression at P7 (P < 0.05) but no significant change in its expression before that time or in the skin at P7 (P = 0.92) (Fig. 4B and SI Fig. 13). Follow-up immunohistochemical studies disclosed that, in contrast to wild-type mice, the Lyve-1-positive lymphatic endothelium of P7 Fiaf−/− animals rarely contain Prox1-positive nuclei (Fig. 4 C–F).

We cannot yet provide a definitive mechanistic explanation about how Fiaf regulates Prox1, or why a very small subset of Lyve-1-positive cells retains the ability to express normal levels of Prox1 in P7 Fiaf−/− animals. Any effects of Fiaf on lymphatic endothelial Prox1 expression are likely to occur via paracrine signaling. Fiaf, which encodes a secreted protein, is prominently expressed in the epithelium but not in the underlying mesenchymal compartment of the small intestine where lymphatic endothelial cells reside, whether judged by qRT-PCR of laser capture microdissected cell populations or β-galactosidase immunohistochemistry of small intestinal sections prepared from Fiaf+/− mice where the disrupted locus contains an inserted lacZ reporter (ref. 9 and data not shown).

Although these findings demonstrate that Prox1 expression in the postnatal intestinal lymphatic endothelium is modulated by Fiaf, the difference in phenotypes between Prox1-deficient and Fiaf-deficient mice suggests that Fiaf can also effect maintenance of lymphatico-venous partitioning through Prox1-independent pathways. The Prox1-deficient phenotype consists of absent or severely abnormal vessels that are leaky: hence the chylous ascites. The phenotype in Fiaf−/− mice is qualitatively different: the aberrant postnatal lymphatico-venous communication in these animals does not lead to chylous ascites; however, such communication could produce retrograde flow within lymphatics causing the small bowel wall edema that we observe (see Fig. 1 and SI Fig. 8B Inset).

To further confirm a central role for Fiaf in postnatal intestinal lymphovascular development, we harvested small intestines from E15 or E18 Fiaf−/− mice and their +/+ littermates (all on a mixed 129/Sv:B6 background) and implanted them into the dorsal s.c. tissue of syngeneic young adult Fiaf+/+ recipients. The intestinal isografts, which had been harvested from donors and placed in recipients under sterile conditions, were recovered 14 days after implantation and characterized by using histochemical and immunohistochemical methods. Our previous studies using wild-type donors and recipients had shown that an “extrinsic” vasculature from the adult recipient animal rapidly envelops the isograft and supports subsequent growth, villus morphogenesis, and epithelial cytodifferentiation of the intestinal implant in its s.c. habitat (20, 21).

Remarkably, intestinal isografts obtained from embryonic wild-type and Fiaf-deficient donors both exhibited normal lymphovascular development after implantation into Fiaf-expressing wild-type recipients, as judged by Lyve-1 and Prox1 staining of serially sectioned, well oriented, nascent crypt villus units (four isografts per group; n ≥ 200 villi per genotype) (Fig. 5).

These experiments indicate that development of the aberrant intestinal lymphatic phenotype in Fiaf−/− intestinal isografts can be prevented if the late gestation/perinatal gut is exposed to an environment where there is circulating Fiaf. Although the rescued Fiaf−/− isografts do not contain Fiaf locally produced from their lining epithelium, they are exposed to a milieu in which circulating host intestine-derived Fiaf is present (9). Similarly, maternal Fiaf may cross the placental barrier, which would explain why Fiaf−/− pups are normal at birth but develop a severe phenotype soon thereafter. Given the rapid neonatal induction of Fiaf that occurs in the intact normal small intestine, our findings are consistent with the notion that sustained lymphovascular partitioning requires locally produced Fiaf to signal through an as yet undefined intestinal receptor.

The molecular mechanisms that regulate normal blood vasculature development have been the subject of intensive investigation, and a detailed picture is emerging about the underlying signaling pathways (22). Despite its importance to health and disease, considerably less is known about the mechanisms that support proper lymphatic development. In this article we present evidence that separation of the intestinal mucosal lymphatic and blood microvasculature is a process whose regulation continues well beyond fetal life and demonstrate the essential, organ-specific role of Fiaf in postnatal lymphatic development.

The phenotype exhibited by Fiaf−/− mice is unique. Mice homozygous for a null allele of Slp-76, an SH2 domain-harboring adapter molecule, or the tyrosine kinase Syk, exhibit aberrant communications between their blood and lymphatic vessels (23). qRT-PCR assays of total small intestinal RNA disclose that Fiaf deficiency is associated with a 2.8 ± 0.3 increase in Syk expression (P = 0.03) (SI Fig. 6), although this may simply reflect the fact that Syk is expressed in blood cells that accumulate in the ectatic lymphatic vessels of these Fiaf knockout animals (23). Unlike Fiaf−/− mice, the defects in Syk−/− and Slp76−/− animals appear during embryogenesis and are not gut-specific. Mice deficient in Angpt2, a member of the same superfamily as Fiaf, also show postnatal intestinal lymphatic dilatation (24), but lymphatico-venous malformations are not present, and the phenotype is not intestine-specific. Fiaf is the first described tissue-specific regulator of postnatal lymphangiogenesis: as such, this observation raises the question of whether there are additional selective modulators of this process that operate in other organs.

Specified pathogen-free Fiaf−/−, +/−, and +/+ littermates on a mixed C57BL/6J:129/Sv background were housed in microisolater cages in a barrier facility and maintained under a strict 12-h light cycle. To test the effects of genetic background on phenotypes associated with Fiaf deficiency, male Fiaf+/− mice with this hybrid B6:129/Sv background were bred to wild-type C57BL/6J females (The Jackson Laboratory, Bar Harbor, ME). The resulting Fiaf+/− progeny were intercrossed to generate Fiaf−/− animals (defined in this article as “backcrossed one generation onto a C57BL/6J background”). Fiaf−/− animals with both genetic backgrounds were rederived as germ-free according to previously described protocols and reared in gnotobiotic isolators (9).

All animals were fed a standard chow diet (Lab Diet 5058; Purina Mills) (autoclaved in the case of germ-free mice). All experiments involving mice were conducted by using protocols approved by the Animal Studies Committee of Washington University.

Small intestines were removed en bloc and fixed overnight at 4°C in freshly prepared 4% paraformaldehyde (in PBS). Fixed specimens used for histochemistry were washed in 70% ethanol, mounted in 2% agar, and embedded in paraffin. Sections (7 μm thick) were cut for subsequent staining with hematoxylin and eosin. For immunohistochemical studies, sections were deparaffinized in xylene, treated with isopropanol, blocked in 1% BSA/0.1% Triton X-100/PBS, and incubated with rabbit anti-mouse Lyve-1 (1:1,000; Upstate Cell Signaling Solutions). Antigen–antibody complexes were visualized by using Vectastain Elite and Nova Red substrate (Vector Laboratories). Sections were counterstained with hematoxylin.

For immunofluorescence studies, fixed tissue was cryoprotected in 10% sucrose in PBS (3 h at 4°C), immersed in 20% sucrose/10% glycerol in PBS (overnight at 4°C), and then frozen in OCT TissueTek compound (VWR Scientific) before preparing 8-μm-thick cryosections. A rat monoclonal antibody to mouse Lyve-1 (R & D Systems), rabbit anti-mouse/human Prox1 (Abcam), Syrian hamster anti-mouse podoplanin (AngioBio), and mouse anti-mouse/human smooth muscle actin conjugated to Cy3 (Sigma, St. Louis, MO) were diluted 1:100 in blocking buffer before use. Secondary Alexa Fluor 594- or Alexa Fluor 647-conjugated antibodies to rat, rabbit, or hamster Igs (1:100; Molecular Probes) were used to visualize antigen–antibody complexes. Nuclei were stained with DAPI (40 ng/ml blocking buffer). An Axiomat inverted microscope and AxioVision 4 software (Zeiss) were used to acquire and process immunofluorescent images. To visualize the microvasculature underlying the small intestinal epithelium, FITC-labeled BS-1 (Sigma) was injected retroorbitally (25 mg/kg of body weight) 30 min before animals were killed (11).

TUNEL assays were performed on fresh-frozen intestinal cryosections by using the TUNEL kit from Roche and the manufacturer's suggested protocols. Sections were incubated with rat anti-Lyve-1 and Alexa Fluor 647-conjugated goat anti-rat Ig before the TUNEL assay (11).

Nile red fluoresces intensely in the presence of both neutral and polar lipids (25). P12 pups were gavaged with 400 μl of a Nile red solution (12.5 μg/ml PBS; Molecular Probes) 60 min before they were killed. The gut was removed en bloc, and lipids in the mesentery were visualized with an Olympus SZX12 stereomicroscope.

P7 pups were injected with Evans blue dye (Sigma) in the hind paws 15 min before they were killed. Central lymphatics were visualized with an Olympus SZX12 microscope.

RNA was isolated from the small intestines of E18, P2, P7, and P24 wild-type and E18, PO, P2, and P7 Fiaf−/− mice (n = 4 per group; RNeasy; Qiagen). Each RNA preparation from each individual animal was reverse-transcribed by using SuperScript II and dT₁₅ primers (Invitrogen). qRT-PCR assays were performed in triplicate with gene-specific primers (SI Table 1) and SYBR green (Abgene), as described previously (9, 11). Data were normalized to L32 mRNA (ΔΔC_T analysis).

Pregnant mice were killed on day 15 or day 18 of gestation. The gastrointestinal tracts of their fetuses were removed en bloc after transection at the gastroesophageal junction and the rectum and placed into RPMI medium 1640 (Gibco/BRL), maintained at 4°C. The stomach was removed by severing the pyloric–duodenal junction, and the small intestine was isolated by cutting the ileal–cecal junction. The small intestinal specimen was then implanted into a young adult isogenic recipient mouse (20, 21). To do so, each recipient animal was anesthetized with rodent anesthesia mixture [0.15 ml of a solution of ketamine (100 mg/ml) and xylazine (20 mg/ml)], and a midline incision was made in their paravertebral region. A pocket was created in the s.c. fascia with a blunt-ended probe, and the fetal isograft was placed into this space. The ends of the isograft were fastened and sealed with 7-0 proline sutures, and the incision was closed with surgical clips. The isograft was removed by careful dissection 14 days after implantation and fixed in 4% paraformaldehyde/PBS as described above.

Data were analyzed by using Student's t test or ANOVA with Tukey's post hoc analysis.