Inhibition of Jagged-Specific Notch Activation Reduces Luteal Angiogenesis and Causes Luteal Hemorrhaging of Hormonally Stimulated Ovaries

Abstract

Robust angiogenesis in the corpus luteum is critical for maintenance of pregnancy and thus mammalian female fertility. During angiogenesis, blood vessels sprout from pre-existing vasculature and recruit pericytes to induce maturation and vessel quiescence. Pericytes are associated with capillaries and regulate endothelial cell proliferation, vessel diameter, and vascular permeability. Endothelial induction of Notch signaling in adjacent pericytes helps recruit and maintain pericyte coverage in some but not all tissue types. We have employed a Notch decoy, N1^10–24, which blocks Notch signaling in a ligand-specific manner, and determined that pharmacological inhibition of Notch ligand Jagged blocks luteal angiogenesis after normal ovulation, resulting in reduced luteal vasculature. Conversely, after ovarian hyperstimulation, a condition which occurs during fertility treatments, Jagged inhibition causes vascular dilation and hemorrhage. These results indicate that Jagged inhibition has effects in different ovarian angiogenic conditions, promoting vascular growth in the corpus luteum and vascular stability in hyperstimulated ovaries.

Angiogenesis, the formation of new blood vessels via sprouting from existing vessels, is critical to establish the circulatory system during embryonic development. In adult mammals, angiogenesis is mainly restricted to regenerating tissue and during wound healing; however, the female reproductive tract exhibits a high level of physiological angiogenesis.¹ Ovarian and uterine angiogenesis are required for maintenance of the estrous/menstrual cycle and for establishment and maintenance of pregnancy. Abnormalities in reproductive angiogenesis can underlie devastating dysmenorrhea, infertility, or recurrent miscarriage. Ovarian hyperstimulation syndrome (OHSS), caused by chorionic gonadotrophin (hCG) exposure during fertility treatments, is associated with severe vascular hyperpermeability and can become debilitating or fatal.^1,2

The reproductive cycle of the ovary passes through several distinct stages, including production of follicles, follicular maturation, ovulation of oocytes, development of the postovulatory corpus luteum (CL), and luteal regression³ (Figure S1A). During ovulation, follicular granulosa cells are induced to proliferate and differentiate into luteal cells, producing steroid hormones estradiol and progesterone critical for pregnancy maintenence.³ Luteal angiogenesis initiates immediately following ovulation, when the basement membrane surrounding the ovulated follicle is degraded and theca vessels invade the previously avascular granulosum^4,5 (Figure S1A). Formation of the functional luteal capillary plexus is necessary for luteal function and hormone dissemination. Proliferation and migration of the theca endothelial cells is dependent on VEGF-A, which is primarily secreted by the follicular granulosa cells.^6,7 VEGF-A blockade inhibits CL angiogenesis, impairs luteal function, and blocks fertility.⁸⁻¹⁰ VEGFR-2 is an essential receptor for VEGF-A in endothelial cells during luteal angiogenesis.^11,12

Mural cells consist of pericytes and vascular smooth muscle cells (vSMCs). Single-cell sequencing shows substantive overlap in expression profiles between pericytes and vSMCs, indicating that the more general term “mural cell” is preferred unless multiple markers have been used to determine mural subtype.¹³ Pericytes surround, support, and stabilize nascent capillaries and postcapillary venules, and in most tissues do not express alpha-smooth muscle actin (αSMA). Vascular smooth muscle cells (vSMCs) support arterioles and larger vessels, have contractile functions, and express αSMA. In studies examining pericytes, loss of pericyte function can lead to loss of endothelial stabilization and overt hemorrhage.¹⁴ Unlike in most angiogenic tissue settings, where pericyte recruitment typically follows endothelial cell invasion, pericytes may be the first vascular cells to invade the ovulated follicle.^5,15 In sheep ovaries, luteal pericytes promote vascular recruitment to the CL through secretion of VEGF-A. Pericytes also fulfill their traditional supportive vascular role in the CL by providing maturation cues and support to nascent capillary vessels; thus, the blockage of pericyte recruitment through expression of soluble PDGFRβ ectodomain causes severe luteal hemorrhaging and decreased luteal vascularization in mice.^16,17 Pericytes therefore have dual functions in luteal angiogenesis, promoting both endothelial recruitment and vessel stability.

The vascular Notch pathway components regulate both endothelial function and pericyte/endothelial interactions during angiogenesis. Notch1 and Notch4 are expressed by luteal and theca endothelial cells.^18,19 Mouse models mutant for Jag1 showed that endothelial Jagged1 activates Notch signaling in mural cells to promote mural cell maturation and coverage of the vasculature in other organs.²⁰⁻²² Mice deficient for Notch1 and Notch3 display abnormal angiogenesis and impaired pericyte function.²³ Deletion of endothelial Jag1 in the retina decreased vascular density and αSMA expression by the mural cell compartment.²² Blocking Jag1–Notch binding causes decreased tumor vessel density and loss of pericyte/endothelial cell association in the tumor vasculature.²⁴ In particular, loss of Jag1 disrupts the stability of pericyte/endothelial interaction in adult wound healing and, remarkably, in established adult heart and retinal vasculature.^25,26 The Notch ligand Jag1 is expressed in luteal endothelial cells and mural cells, suggesting that Jag1 may be the major activator for Notch signaling in the corpus luteum.¹⁹

We here demonstrate that Jagged-specific Notch signaling is required for recruitment of both pericytes and endothelium to the CL during normal ovulation in the mouse. Under the more intense luteal angiogenic conditions of ovarian hyperstimulation, which mimic human fertility treatments, Jagged–Notch signaling is critical for luteal vessel stability, and Jagged inhibition causes luteal hemorrhaging and loss of vascular pericyte association.

Results and Discussion

Luteal Pericytes Express Jag1 and Notch1

To determine if Notch signaling contributes to vascular function in the CL, we analyzed Notch receptor and ligand expression in luteal vascular cells. Previous work has shown that Jag1 is expressed in the endothelium of both luteal neovasculature and mature thecal vasculature.¹⁹ Intense Jag1 expression was also noted in a subset of mural cells, but the nature of this subset has not been elucidated.¹⁹ To assess luteal angiogenesis, ovulation was induced in sexually immature female C57BL6 mice by intraperitoneal injections of 5 IU of pregnant mare serum gonadotropin (PMSG) to induce folliculogenesis, followed by 5 IU of hCG 2 days later to induce ovulation and luteinization (Figures 1A and S1A). Ovaries were collected 2 days after hCG injection (day 4).

Figure 1.

Jagged inhibition with N1^10–24 decoy results in smaller ovaries after physiologic ovulation. (A) Timeline and dosages of stimulation of physiologic ovulation and adenoviral N1^10–24 decoy administration. (B) Diagram of N1^10–24 decoy, composed of human Notch1 EGF-like repeats 10–24 fused to human IgG Fc. (C) Subset of NG2-positive pericytes in the corpus luteum express Jagged1 (arrows). Scale bars: 50 μm. (D) H&E staining of ovary sections shows apparently normal corpus lutea after N1^10–24 decoy treatment. Scale bars: 200 μm. (E) N1^10–24-decoy-treated ovaries that have undergone physiologic ovulation have decreased mass. (F and G) No significant difference was seen between the number of corpus luteum (CL) (F) or CL development (G) after N1^10–24 decoy treatment. n = 4 mice, 2 ovaries/mouse. Error bars represent standard deviation (SD) in all figure panels.

We defined the CL mural cell populations using the markers NG2, PDGFRβ, desmin, and αSMA to distinguish between mural subtypes. NG2, a pericyte marker, strongly stained the mural cells in the CL, but the staining intensity for NG2 was weak in the theca (Figure S1B). PDGFRβ was strongly expressed by luteal and theca mural cells, consistent with its roles in both pericytes and vascular smooth muscle cells (vSMCs) (Figure S1C). The mural cells in the theca layer expressed higher levels of desmin than those in the CL mural cells, suggesting that desmin may mark a more mature mural cell population (Figure S1D). αSMA expression, a marker of contractile vSMCs, was restricted to the mural cell population of the theca layer, while the luteal mural population appeared to be devoid of contractile function (Figure S1E). These differences suggest that the theca layer contains larger and more established blood vessels requiring support from more mature and contractile αSMA-positive vSMCs, while the transient capillary plexus of the CL may require more versatile and active mural cells such as NG2-positive pericytes.

We co-stained ovary sections with NG2 and either Notch1, Notch3, or Jag1 to determine the Notch expression profile of luteal pericytes. A small subset of luteal pericytes express Notch1, while Notch3 expression is restricted to the mural cells of the theca layer (Figure S2A,B). Luteal pericytes express varying levels of the Notch ligand Jag1 with a large subset that express very high levels of Jag1 (Figure 1C, arrows). Pericytes interact closely with endothelial cells; thus, to ensure accurate identification of ovarian pericytes, all images were taken at a resolution where endothelial and pericyte markers (i.e., CD31 and NG2) showed clear separation. Additionally, we assessed Jag1-positive cells that were matched with NG2 positive cells and which shared cellular morphology. We therefore hypothesized that disrupting Jagged/Notch interactions could alter pericyte coverage and function of the CL capillary plexus.

Jag1 Inhibition Permits Ovulation but Blocks Luteal Angiogenesis

To inhibit Jagged-specific activation of Notch, we employed a construct comprised of the EGF-like repeats 10–24 of the human NOTCH1 extracellular domain fused to a human Fc domain, called the N1^10–24 decoy (Figure 1B). The N1^10–24 decoy specifically and nonproductively interferes with the activity of Jagged-class Notch ligands to inhibit Jagged–Notch signaling and has been shown to disrupt pericyte/endothelial interactions in tumor angiogenesis.^24,27 We administered either N1^10–24 decoy or human Fc control by infecting mice with adenoviral expression constructs which infect the murine liver, causing hepatocytes to produce and secrete protein into systemic circulation. Serum human Fc and N1^10–24 decoy can be detected the day immediately following adenovirus injection, with increasing levels though day 4 (Figure S3A,B). To block Jagged signaling during luteinization while permitting normal folliculogenesis, we induced physiologic ovulation as described above and injected the mice with adenovirus on the day between PMSG and hCG treatment (Figure 1A).

Ovaries isolated from N1^10–24 decoy-treated mice were smaller by weight, but sections of ovaries showed no significant difference in the number of CL or distribution between normal and stalled CL (Figure 1D,G). Decreased ovarian weight without changes in CL number can be caused by inhibiting luteal angiogenesis.¹¹ We therefore examined CL endothelial content and determined that the CL of N1^10–24 decoy treated mice display a 30% reduction in vessel density (p = 0.002, Figure 2A–C), demonstrating reduced angiogenesis.

Figure 2.

Jagged inhibition with N1^10–24 decoy reduces vascular recruitment to the corpus luteum after physiologic ovulation. Ovary sections from human Fc-treated (A–A′′′) and N1^10–24 decoy-treated mice (B–B′′′) stained for endothelial marker CD31 (red), pericyte marker NG2 (green), and DAPI (blue). (C–E) Quantification of CD31 and NG2 staining within the CL normalized to total luteal cell content (DAPI) shows that N1^10–24 decoy treatment reduces endothelial content (C) and pericyte content (D) but that luteal vessels have normal pericyte coverage (E). n = 4, 2 ovaries/mouse, 2 sections/ovary. Scale bars: 100 μm.

To determine whether Jagged inhibition disrupts pericytes in the CL, we stained ovarian sections for the pericyte marker NG2.^15,17 N1^10–24 decoy-treated mice showed a 30% reduction in CL NG2-positive pericyte cell content (p = 0.003, Figure 2D). However, no difference was observed in the relative vascular pericyte coverage in the CL capillary plexus under physiologic conditions, suggesting that the angiogenic block does not derive from pericyte loss (Figure 2E).

Jagged Inhibition Causes Luteal Hemorrhaging, Vessel Dilation, and Pericyte Dysfunction in a Mouse Model of Ovarian Hyperstimulation Syndrome

In human diseases, the lack of adequate pericyte coverage can lead to vessel leakage, aneurysm, and hemorrhaging when vascular function is stressed. OHSS patients undergoing gonadotropin-based fertility treatment experience excessive ovarian angiogenesis and luteal vessel leakiness, resulting in nausea, swelling, ovarian rupture, or even death.²⁸ We therefore investigated whether Jagged–Notch signaling affects hyperstimulated CL angiogenesis.

To induce hyperstimulation, we increased the dose of gonadotropins to 20 IU PMSG and 20 IU hCG and again administered human Fc or N1^10–24 decoy via adenovirus injection (Figure 3A), which resulted in high serum expression levels on experimental day 4 (Figure 3B). Hyperstimulated ovaries treated with N1^10–24 decoy displayed severe hemorrhaging, numerous blood-filled corpuscles, and a greater increase in ovarian weight than control hyperstimulated ovaries (Figure 3C,D). H&E staining of N1^10–24 decoy-treated ovaries showed a significant increase in blood-filled hemorrhagic CL per ovary, with no significant increase in the number of histologically normal CL (Figure 3E,F).

Figure 3.

Jagged inhibition causes expansion and hemorrhage in hyperstimulated ovaries. (A) Timeline and dosages of ovulatory hyperstimulation and adenoviral N1^10–24 decoy administration. (B) N1^10–24 decoy and control Fc are robustly expressed in circulating serum at the time of ovary collection. An anti-Fc antibody was used to detect both proteins. (C) Intact hyperstimulated N1^10–24 decoy-treated ovaries show overt hemorrhaging. (D) N1^10–24 decoy-treated hyperstimulated ovaries have increased mass. (E) H&E staining of N1^10–24 decoy-treated hyperstimulated ovaries show blood-filled CL and a significant increase in hemorrhaging CL (F). n = 4 mice, 2 ovaries/mouse. Scale bars: 100 μm (C), 200 μm (E).

In contrast to physiologic ovulation, where Jagged inhibition causes reduced angiogenesis in the CL, the N1^10–24 decoy-treated hyperstimulated ovaries showed no significant change in endothelial content, pericyte content, or endothelial/pericyte ratio compared to control hyperstimulated ovaries (Figure 4A–D). Overt vessel lumens, as defined by openings in the tissue bordered by cells positive for CD31 and/or NG2, were rare in the capillary plexus of the Fc-treated CL (Figures 4A and 5A). Vessels in the N1^10–24 decoy-treated nonhemorrhagic CL plexus showed multiple visible lumens, suggestive of vessel dilation (Figures 4A,B and 5A,B). The capillary plexus in N1^10–24 decoy-treated hemorrhaging CL showed abundant visible vessel lumens dramatically larger than those in Fc controls, suggesting severe vessel dilation (Figure 5D). The vessels with enlarged lumens often showed discontinuous endothelial lining, where the NG2-positive pericytes were directly lining the capillary lumen (Figure 5D). This phenotype was not observed in the rare hemorrhaging CLs of the control hyperstimulated ovaries (Figure 5C).

Figure 4.

Jagged inhibition does not reduce vascular recruitment in hyperstimulated corpus lutea. Sections from human-Fc-treated (A-A′′′) and N1^10–24 decoy-treated (B–B′′′) hyperstimulated ovaries stained for endothelial marker CD31 (red), pericyte marker NG2 (green), and DAPI. (C–D) Quantification of CD31 and NG2 staining within the CL shows no effect of N1^10–24 decoy treatment on endothelial content or pericyte content in hyperstimulated ovaries. n = 4 mice, 1 ovary/mouse, 2 sections/ovary. White areas = vascular lumens. Scale bars: 50 μm.

Figure 5.

Jagged inhibition causes vessel dilation, abnormal vessel morphology, and disorganized pericytes in hyperstimulated corpus lutea. (A–D) Confocal maximum projection stacks of control (A, C) and N1^10–24 decoy-treated (B, D) hyperstimulated nonhemorrhaged (A, B) and hemorrhaged (C, D) CL stained for CD31 (red), NG2 (green), and DAPI. Boxes indicate higher magnification panels below (A′–D′′). H = hemorrhagic CL core, dotted line = periphery of CL, white areas = vessel lumens, arrowheads = pericytes cellular processes not in contact with endothelium. Scale bars: 50 μm.

Vessel dilation and hemorrhage is a common phenotype associated with vasculature that lacks adequate mural cell coverage, as displayed by mice with impaired endothelial Jagged1 expression.²⁰ In the control CL, we observed pericytes tightly lining the capillary vessels (Figure 5A,C). However, in both hemorrhagic and nonhemorrhagic N1^10–24 decoy-treated CL capillaries, the pericytes appeared to be disordered with abundant cellular processes that failed to make contact with the endothelium (Figure 5B,D). The observed luteal hemorrhage suggested that Jagged inhibition caused pericyte-endothelial disassociation, possibly underlying the vascular instability in hyperstimulated ovaries. N1^10–24 decoy-treated hyperstimulated ovaries showed no major changes in αSMA expression in the vSMC of the ovarian theca, but fluorescent intensity was mildly reduced, suggesting that the thecal vessels may also have reduced mural coverage (Figure S4).

Discussion

Our data provides a new role for Notch signaling in the ovarian vasculature and provides evidence for Jagged function in reproductive angiogenesis. Luteal angiogenesis is a tightly regulated process that includes CL capillary plexus formation, which ensures systemic delivery of luteal hormones. Our data demonstrate that during normal ovulation Jagged–Notch signaling acts to promote angiogenesis in the CL and permits normal recruitment and/or growth of both endothelium and pericytes into the luteum.

Luteal angiogenesis assures delivery of ovarian hormones to sustain pregnancy. In humans, implantation occurs 7–9 days after ovulation and requires a functional CL. Analysis of the effects of N1^10–24 decoy on fertility was beyond the scope of this study, but our results suggest that Jagged inhibition may prevent successful implantation and pregnancy. N1^10–24 decoy may therefore represent a novel mechanism for emergency contraception.

Conversely, under conditions of ovarian hyperstimulation that mimic conditions designed to promote fertility for patients undergoing in vitro fertilization, Jagged inhibition caused capillary dilation, hemorrhage, and pericyte dysfunction but did not alter overall endothelial or pericyte content. Ovarian hyperstimulation occurs in a subset of patients undergoing clinical gonadotropin administration as part of fertility treatments, and while some risk factors are known, patient response is currently unpredictable. These results suggest that underlying variations in Notch signaling may influence extreme response to gonadotropin and dictate whether OHSS progresses to ovarian effusion and hemorrhage.

It is well-established that endothelial Jagged1 is required during embryonic development to promote vSMC coverage of large vessels.^20,29 The role of Notch signaling in pericyte/endothelial cell interactions is less clear. Jagged1 expressed by endothelial cells and mural cells activates Notch signaling in mural cells to regulate their maturation and coverage of the vasculature in multiple tissues.^20,23,30,31 Our lab has previously showed that Jagged-specific Notch decoy inhibits vascular density during tumor angiogenesis and that it impairs pericyte association with the tumor vasculature.²⁴ Recent work has shown that loss of Notch cofactor Rbpj profoundly disrupts pericyte function, but this effect may be independent of Notch signaling.¹⁴ Here we demonstrate that Jagged-specific activation of Notch does not alter the quantity of pericyte coverage of luteal vessels, but it instead affects the ability of pericytes to properly associate with the endothelium during luteal angiogenesis.

It is not clear why N1^10–24-decoy-mediated Jagged inhibition causes antiangiogenic effects in physiologic ovulation while causing vascular dilation in hyperstimulatory contexts. It is possible that physiological angiogenesis can compensate for altered pericyte function and achieve a functional, despite a sparser luteal capillary network. However, in the pathological context of OHSS characterized by very high levels of endothelial sprouting and VEGF expression, the effects of altered pericyte function cannot be overcome and result in a dysfunctional luteal capillary plexus. Additionally, it is unclear which cell type is most critical for Jagged signaling. N1^10–24 decoy is secreted systemically and should inhibit ligand presentation by all Jag1 expressing cells in the CL, which may include the luteal endothelium, thecal endothelium, and several subsets of pericytes and vSMC, as well as cells expressing Jag2, such as granulosa cells.³² Future experiments with tissue-specific manipulation of Jag1 and Jag2 may determine the mechanistic and cellular origin of Jagged-specific reduced angiogenesis.

Methods

Animals

For ovary studies, female wildtype C57BL6/J mice, aged 3 weeks (± 3 days), were obtained from Jackson Laboratories. Pregnant mare gonadotropin (Sigma) and human chorionic gonadotropin (Sigma) were used for stimulation of folliculogenesis and ovulation, respectively. Adenovirus was administered via IP injection at 2.5 × 10⁸ pfu. All procedures were carried out according to approved protocols and guidelines established by the Columbia University Institutional Animal Care and Use Committee.

Ovary Preparation and Staining

Following sacrifice, ovaries were removed and weighed after the fallopian tube was removed. Ovaries were fixed with 4% PFA for 2 h at 4 °C and then dehydrated in 30% sucrose overnight at 4 °C and embedded in OCT for sectioning. Ovary sections of 7 μm were fixed and permeabilized with acetone. Additionally, 30 μm ovary sections were fixed with 4% PFA and permeabilized with 0.5% Triton-X. Hematoxylin and eosin staining was performed via standard protocols. For immunohistochemistry, sections were incubated for 1 h in blocking solution (3% bovine serum albumin and 2% donkey serum), then incubated overnight at 4 °C in blocking solution with primary antibody. Primary antibodies used include anti-CD31 1:500 (BD Biosciences, 553370), anti-NG2 1:750 (Millipore, AB5320), anti-PDGFR-β 1:500 (Cell Signaling, 3169), and anti-αSMA 1:1000 (Sigma, C6198). After washing in 1× PBS, ovary sections were then incubated for 1 h at room temperature in blocking solution containing Alexa Fluor-conjugated secondary antibodies (Invitrogen) diluted at 1:1000. Vectashield mounting medium with DAPI (Vector Laboratories) was used to visualize nuclei.

Validation of Serum Decoy Levels

For analysis of serum decoy levels, blood was isolated from the tail vein or cardiac puncture. Blood was centrifuged at maximum speed for 5 min to separate plasma and frozen at −20 °C. Aliquots (2 μL) serum were run on SDS-PAGE by standard protocol. N1^10–24 decoy and Fc control were detected with HRP-conjugated antibody antihuman Fc (Sigma; A0170) at 1:5000 in 2% milk, 2% BSA, and 0.1% Tween in PBS.

Image Acquisition

A Nikon Eclipse 200 fluorescent microscope and a Nikon A1R confocal microscope were used to obtain images of fluorescently stained images of ovary sections. ImageJ and Adobe Photoshop were used for image processing.

Vessel Density and Pericyte Coverage Quantification

ImageJ was used for all quantification. For ovarian analysis, merged 10× stacks were obtained of 30 μm sections to cover the entire ovary section and to account for changes in the focal plane. Each corpus luteum was outlined by hand. For each corpus luteum, thresholding was used to quantify corresponding total cell content by DAPI staining, endothelial cell content by CD31 staining, and pericyte content by NG2 staining.

Statistical Analysis

Statistical significance was assessed using the two-tailed Student’s t test. Values of p < 0.05 were considered significant. All data represents three independent experiments unless otherwise noted. Error bars represent SD in all figure panels.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsptsci.9b00050.

• Analysis of CL mural cell markers; Notch receptors in CL pericytes; validation of decoy serum levels; αSMA expression in hyperstimulated ovaries (PDF)

Author Contributions

N.M.K. designed and conducted experiments, collected and interpreted data, and drafted the manuscript. L.A.N. provided critical preparation and revisions of the manuscript. L.D.Z. conducted experiments. J.K.K. guided conception of the experiments and contributed to study design and analysis, critical manuscript revision, and final approval.

The authors declare no competing financial interest.