Generation of mice with a floxed Angpt1 allele. A BAC recombineering approach was used to generate a floxed Angpt1 allele (Figure 1A), with loxP sites inserted around exon 1. Cre-mediated excision of the floxed allele is predicted to generate a null Angpt1 allele through frameshift. Correctly targeted ES cell clones were identified by Southern blot analysis with probes outside the region of homology (Figure 1B). The targeting frequency was 16 out of 700 clones (2.3%). After aggregation and breeding of chimeras, heterozygous floxed Angpt1 mice were bred to mice expressing flp recombinase to remove the Neo cassette.

To confirm that excision of the floxed allele results in a null Angpt1 allele, Angpt1^lox/+ mice were bred to the pCaggs Cre-driver line that expresses Cre recombinase at the 1-cell embryo stage (11), creating an Angpt1^del allele. Homozygous deletion (Angpt1^del/del) resulted in embryonic lethality at E9.5 to E12.5 that was identical to that of the conventional knockout and confirmed that the Angpt1^del allele is functionally a null allele (Figure 1C). Angpt1^del/del embryos were studied at E9.5 and E10.5. At E9.5, Angpt1^del/del embryos showed heart defects and loss of heart trabeculations (Figure 1D) identical to those of the conventional knockout (3). Heterozygous embryos showed no phenotype, and a deleted allele was used in breeding to increase Cre-dependent excision.

Angpt1 is critical for cardiac development. The earliest phenotype in the Angpt1^del/del mice was observed in the heart at E9.5. Specifically, there was marked simplification of the cardiac trabeculation pattern evident at E9.5 and E10.5 (Figure 2, A and B). Optical projection tomography (OPT) was used to study the vasculature of the whole embryos in detail. At E9.5, OPT also showed lack of cardiac trabeculations, but, surprisingly, no other vascular defects were observed. By E10.5, the Angpt1^del/del embryos were markedly growth restricted (Figure 2C and Figure 3) and exhibited widespread vascular defects, with fusion of vessels with larger diameters and absence of patterning (Figure 3). Despite the vascular defects, Angpt1^del/del vessels showed coverage of pericytes similar to that of controls (Supplemental Figure 1, A and B; supplemental material available online with this article; doi: 10.1172/JCI46322DS1).

Given that the cardiac defects preceded the major vascular defects by an entire day, we wondered whether the extensive vascular defects observed at E10.5, which are reported in the original knockout paper (3), might be a direct consequence of abnormal heart function. To test this, we deleted Angpt1 specifically in cardiomyocytes using the Nkx2.5–Cre-driver line, Angpt1^{del/^(heart)} mice. We found that cardiac-specific deletion of Angpt1 causes embryonic lethality and a phenotype that is virtually identical to that caused by global and conventional Angpt1 deletion. Quantification of the vascular area of the hindbrain showed a significant increase in vasculature in both Angpt1^del/del and Angpt1^{del/^(heart)} embryos compared with that of controls (Figure 3P). Thus, the profound vascular remodeling defects at E10.5 were secondary to the phenotype of cardiac abnormalities and are perhaps due to abnormal blood flow and hemodynamics (Figure 3 and Supplemental Videos 1 and 2). Future studies that allow deletion of Angpt1 in specific vascular beds but leave cardiac levels intact are required to determine whether there are additional roles for Angpt1 at this early time point.

Angpt1 is required to shape the developing vasculature. To determine whether Angpt1 plays a direct role in vascular development, we used an inducible system to knock out the Angpt1 allele at each embryonic day. Upon administration of doxycycline (DOX) to the pregnant dam, the ROSA-rtTA/tetO-Cre bitransgenic system was activated to excise the floxed Angpt1 allele from the entire embryo (Figure 4A). In most cases, an Angpt1^del allele was used to enhance the degree of excision. PCR and Southern blot analysis of tail genomic DNA confirmed excellent excision of the Angpt1 gene, i.e., mice only have a deletion band and no floxed band after Cre recombination (Supplemental Figure 2, A and B).

Deletion of Angpt1 at E10.5, Angpt1^{del/^(E10.5)}, resulted in embryonic lethality between E17.5 and P0 (Figure 4B). Vascular abnormalities were observed in several organs and were particularly apparent in the liver and kidney, in which vessels were dilated, with veins being more affected than arteries (Figure 4, E–H). In addition, dilated atria were observed in Angpt1^{del/^(E10.5)} mice (Figure 4, C and D). Quantification of vessels in the liver showed that there were more vessels and greater vessel area in Angpt1^{del/^(E10.5)} mice (Figure 4, V and X). Nonetheless, pericytes were present in equivalent numbers compared to those of controls, as shown by stainings for smooth muscle actin and Pdgfrb (Figure 4, I–N, and Supplemental Figure 1C), and lymphatic vessels appeared normal (Supplemental Figure 3).

In the kidney, dramatic defects were observed in the specialized microvascular beds of the glomeruli, the site of formation of primary urinary filtrate. Many glomerular capillaries exhibited dilated capillary loops and, in some cases, only a single large, open loop (Figure 4, O–T). There are also segments of the glomerular basement membrane (GBM) that were grossly disrupted with numerous folds (Figure 4U). Although mesangial cells were present in Angpt1^{del/^(10.5)} embryos, the number appeared reduced in some glomeruli. While the GBM directly beneath the podocytes appeared to be intact, the subendothelial GBM was markedly disorganized and endothelial cell attachment was disrupted (Figure 4U). Structure of podocytes was intact, and they expressed markers of differentiation, such as podocin, similar to wild-type embryos, confirming that this specialized pericyte was also intact (Figure 4, Q and R). Thus, elimination of Angpt1 at E10.5 results in a primary abnormality of the glomerular endothelium and associated matrix production.

Angpt1 is dispensable in the adult vasculature. Deletion of Angpt1 at any point up to E12.5 had profound consequences, causing death in the perinatal period and the range of structural abnormalities described above. On the other hand, deletion at any time point after E13.5 [Angpt1^{del/^(E13.5)}] did not affect survival and produced viable, fertile Angpt1-deficient mice, with no overt phenotype (Figure 4B). To ensure that the lack of phenotype was not the result of poor excision, we determined the degree of genomic rearrangement of the allele from multiple tissues using PCR (data not shown). Southern blot analysis showed only a deletion band and no floxed band, confirming close to 100% degree of excision (Supplemental Figure 2A). Real-time PCR demonstrated absence of Ang1 (Supplemental Figure 2, C and D). To determine whether expression of other Tek ligands might be upregulated to compensate for deletion, we used real-time PCR. No difference was observed in levels of Angpt2 or Angpt3 (Supplemental Figure 2, C and D). In addition, we found no difference in the expression of Tek, Tgfb1, Vegfa, Pdgfb, and Pdgfrb (Supplemental Figure 2, C and D).

Since retinal vascular development occurs from P0 to P5, we examined the retinas of Angpt1^{del/^(E13.5)} mice and compared them to those of control littermates. While the retinal neural layers were reduced in number (Supplemental Figure 4A), the retinal vasculature was indistinguishable from that of wild-type mice. Although the number of branch points was more variable in Angpt1^{del/^(E13.5)} mutants, the overall number of branch points was not significantly different between the groups (Supplemental Figure 4, B and C). Hyaloid vessels regress by P9 — and again there was no difference between mutant and wild-type mice (n = 2–3; Supplemental Figure 4D). In addition, pericytes were present in equivalent numbers in control and Angpt1^{del/^(E13.5)} retinas (Supplemental Figure 1, D and E).

Role of Angpt1 in adult tissues in wound healing and microvascular injury. The viability and absence of a phenotype in mice with Angpt1 deleted after E13.5 demonstrates that Angpt1 is dispensable in quiescent vasculature. Moreover, these animals provide a useful model for testing the role of Angpt1 in other settings. Previous studies have indicated that Angpt1 may act as a negative regulator of angiogenesis when levels of proangiogenic factors, such as Vegfa, are elevated (12). To test this in vivo, we used ear punch as a simple model of tissue injury and wound healing in which angiogenesis is stimulated. Using mice with intact capacity for Angpt1 production, we confirmed local upregulation of Vegfa in keratinocytes at the site of wound healing using a Vegfa-lacZ mouse (data not shown), consistent with previous reports using real-time PCR (13). We then compared wound healing responses in wild-type mice and mice with induced deletion of Angpt1 starting at day E13.5–E16.5 [Angpt1^{del/^(E13.5)} mice]. The Angpt1^{del/^(E13.5)} mice showed an accelerated wound healing response, with almost complete closure by day 60, compared with the minimal response seen in littermate controls. After 60 days, the mean wound area in controls was 3.44 ± 0.30 mm², compared with only 0.82 ± 0.14 mm² in Angpt1^{del/^(E13.5)} mice (Figure 5, A–C). The cartilage and hair follicles were not regenerated in either group, demonstrating that the mechanism of closure is through fibrosis rather than regeneration (Figure 5B). There was a substantial enhancement of angiogenesis and fibrosis in the closed ear tissue in Angpt1^{del/^(E13.5)} mice compared with that of controls, as shown by endothelial staining and Masson Trichrome staining of collagen fibers (Figure 5, B and D). Vessels had normal coverage of pericytes (Figure 5E). Thus, after tissue injury, Angpt1 acts to constrain angiogenesis and fibrosis, and, in its absence, these features of wound healing are exaggerated.

Angpt1 protects the glomerular microvasculature in diabetic nephropathy. To examine the role of Angpt1 in a setting of more complex microvascular injury, we used a model of diabetic kidney injury in the mouse. The early phase of diabetic kidney disease is characterized by neoangiogenesis and increased vascular permeability, with leakage of proteins into the urine (albuminuria and proteinuria). Serum levels of both VEGFA and ANGPT2 are increased in diabetics, and there is an elevation of the ANGPT2/ANGPT1 ratio, which is associated with worse cardiovascular and kidney outcomes (9). We confirmed the upregulation of Vegfa, Angpt2, and Tgfb1 in glomerular cell fractions in the streptozotocin (STZ) model of diabetes by using a kinase insert domain protein receptor–GFP (Kdr-GFP) (Kdr is also known as Flk1) transgenic mouse and FACS (Figure 6A).

To examine the role of Angpt1 in diabetes, controls and mice with Angpt1 deletion starting between E16.5 and P0 [Angpt1^{del/^(E16.5)} mice] were used. At the age of 4–6 weeks, they were given injections of the β cell toxin STZ to induce diabetes and were then monitored for up to 20 weeks. This regimen typically causes robust hyperglycemia but has only modest effects on kidney structure and function in wild-type mice (14), which was the case for control mice in our study. By contrast, 20% of the diabetic Angpt1^{del/^(E16.5)} mice died before the end of the study, whereas all control mice survived (P < 0.05, Figure 6B). The surviving diabetic Angpt1^{del/^(E16.5)} mice have impaired function of the glomerular filtration barrier (GFB), manifested by significant albuminuria, with urinary albumin/creatinine ratios of 0.25 (+0.08, –0.06) mg/mg in diabetic Angpt1^{del/^(E16.5)} mice compared with ratios of 0.06 ± 0.01 mg/mg in diabetic controls (Figure 6C). On histopathological examination of the kidney, diabetic controls had minimal abnormalities confined to mild mesangial expansion. However, there were marked changes in glomerular histology in the diabetic Angpt1^{del/^(E16.5)} mice, with dramatic mesangial matrix expansion and glomerulosclerosis (Figure 6D); similar pathological features can be seen in humans with advanced diabetic nephropathy. Such changes were never seen in diabetic controls or nondiabetic groups and have rarely been reported in other studies of diabetic kidney injury in mice (14, 15). Despite the differences in glomerular pathology, the extent of hyperglycemia achieved was not different between the groups, as reflected by the similar levels of glycosylated hemoglobin (HbA1c) at the end of the study [0.088% ± 0.003% vs. 0.090% ± 0.003% for diabetic controls and diabetic Angpt1^{del/^(E16.5)} mice, respectively] (Figure 6E). F4/80-positive macrophage infiltration was increased in kidneys from diabetic compared with nondiabetic mice; however, there was no difference between controls and Angpt1^{del/^(E16.5)} mice in diabetic or nondiabetic conditions (data not shown).

To define the key cellular sources of Angpt1 that protect the glomerulus in diabetes, we generated separate lines of mice with Angpt1 deleted specifically from podocytes or mesangial cells [Angpt1^{del/^(glom)} mice] using podocyte- and mesangial-expressed Cre drivers (16, 17). After STZ treatment, only compound mutants (i.e., both Nphs1-Cre and Pdgfrb-Cre mice) showed a similar degree of accelerated glomerular damage as the global diabetic Angpt1^{del/^(E16.5)} mice, indicating that production of Angpt1 by each of these glomerular cell populations provides protection against microvascular injury in diabetes (Figure 6D). Thus, local production of Angpt1 by glomerular cell populations, including the podocyte and mesangial cell, provides significant protection against diabetic kidney injury.