Critical roles for murine Reck in the regulation of vascular patterning and stabilization

Extracellular matrix (ECM) is known to play several important roles in vascular development, although the molecular mechanisms behind these remain largely unknown. RECK, a tumor suppressor downregulated in a wide variety of cancers, encodes a membrane-anchored matrix-metalloproteinase-regulator. Mice lacking functional Reck die in utero, demonstrating its importance for mammalian embryogenesis; however, the underlying causes of mid-gestation lethality remain unclear. Using Reck conditional knockout mice, we have now demonstrated that the lack of Reck in vascular mural cells is largely responsible for mid-gestation lethality. Experiments using cultured aortic explants further revealed that Reck is essential for at least two events in sprouting angiogenesis; (1) correct association of mural and endothelial tip cells to the microvessels and (2) maintenance of fibronectin matrix surrounding the vessels. These findings demonstrate the importance of appropriate cell-cell interactions and ECM maintenance for angiogenesis and the involvement of Reck as a critical regulator of these events.

both vascular endothelial cells and mural cells 27 . Dilated vessels with abnormal luminal shapes can be observed in these tissues in mice with reduced Reck expression. Abundant Reck-expression has also been found in fibroblastic cells associated with bifurcating vessels, leading to the speculation that Reck may play a role in non-sprouting angiogenesis (e.g., intussusception and pruning) 27 .
In the present study, we dissected the roles for Reck in different vascular cell types during angiogenesis by using multiple lines of newly developed Reck mutant mice. We also employed aortic ring assay (ARA) 28,29 to assess the ability of aortic tissue explants to form small vessels (microvessels) in vitro. We found that selective inactivation of Reck in vascular mural cells caused embryonic death around E10.5 with vascular defects, suggesting that the mid-gestation lethality of Reck-null mice can be attributed to the absence of Reck in mural cells. In addition, we unexpectedly found that impaired Reck function leads to excessive sprouting of unstable microvessels in vitro, raising the possibility that the abnormal, dilated vessels found in Reck-deficient mice may arise by lateral fusion of unstable vessels rather than, or in addition to, abortive intussusception.
Effects of Reck-deficiency on microvessel formation. To understand the roles of Reck at the cellular level, ARAs 28,29 were utilized, which allowed assessment of the ability of dorsal aorta tissue pieces (aortic rings) to form microvessels in vitro. Aortae from 5-weeks old, tamoxifen-induced Reck knockout (Reck cKO) and control (Cont) mice carrying the mTmG reporter were used. Under optimized conditions (Supplementary Figs S2 and S3a), control aortic rings showed a slow but steady increase in microvessel number over the time course observed (Fig. 2a, blue line), whereas Reck cKO samples showed an initial rapid increase (up to day 10) followed by a decline (from day 12) in the number of microvessels (Fig. 2a, red line). This decline was accompanied by aggregation and thickening of microvessels ( Fig. 2b-8, 9, arrowheads; Supplementary Fig. S3b,c). Morphometry of fluorescent images ( Fig. 2c-f and Supplementary Fig. S4) indicated that the Reck cKO microvessels were wider (Fig. 2c) and covered a broader area (Fig. 2d) with lower lacunarity (Fig. 2e). In addition, Reck cKO aortic rings were often accompanied by peri-aortic halos, indicating increased local lysis of collagen gel 33 (Fig. 2b-10, arrow; Fig. 2f). Thus, Reckdeficiency leads to the formation of an excessive number of unstable microvessels in this assay.

Identity of Reck-mG cells.
To determine how Reck-deficiency leads to such microvessel phenotypes, the nature of Reck-mG cells in ARA was examined. In control cultures, some Reck-mG cells localize at microvessel tips ( Fig. 3a-1, arrowhead), whilst others are associated with microvessel stalks (Fig. 3a-1, arrow). The former look similar to the Tie2-mG or CD31-positive endothelial cells ( Fig. 3a  . Taken together, these findings suggest that Reck is required to achieve adequate compositions of, and interactions between, vessel-forming cells. Functional relationship between Reck and FN in microvessel formation. FN and its receptor are known to be protected by RECK 18,34,35 . In ARA with Reck-positive cells, abundant FN fibrils could be visualized by immunofluorescent staining (Fig. 5a, panel 3 and 7), and prominent signals were found in the areas where mural (Sm22-mG) cells tightly associated with microvessels (Fig. 5a, panels 1-4, arrows). Near the tips of these microvessels, localized loss of FN fibrils were found near microvessel tips (Fig. 5a, panel 5-8, arrow). In Reck cKO cultures, signals for both FN (Fig. 5a-11) and laminin α 5 (Lama5), a component of vBM ( Fig. 5b-5), were dampened and diffuse. When Reck cKO aortic rings were embedded in collagen gel supplemented with purified FN (Fig. 5c), some Reck cKO phenotypes, such as increased number, width, and aggregation of microvessels, increased area covered by microvessels, and decreased lacunarity of vascular network, were significantly suppressed ( Fig. 5c-g) with increased perivascular FN-and Lama5-immunoreactivity (Supplementary Fig. S7a and b). Other Reck cKO phenotypes, such as peri-aortic halo and poor association of Reck-mG cells with microvessels, were not fully suppressed by FN supplementation (Fig. 5h, i). Hence, some Reck cKO phenotypes in ARA, such as excessive sprouting and destabilized microvessels, may be attributable to the reduced ambient FN in the absence of Reck.

Discussion
In this study, the importance of Reck in both mural and endothelial cells was documented in vivo and in vitro. Although selective inactivation of Reck in mural cells in vivo resulted in mid-gestation lethality and vascular defects reminiscent of Reck-null mice (Fig. 1c-e and ref. 2), Reck-inactivation in Tie2-positive cells also resulted in embryonic death, albeit at later stages with major impacts in the brain (Fig. 1c, f,g). In vitro, Reck-positive cells contribute to both mural and endothelial lineages (Fig. 3) whilst Reck-inactivation results in increased sprouting, decreased microvessel stability (Fig. 2) and altered composition (Fig. 3d-f) alongside a change in behavior that includes defective localization and reduced association (Figs 3b,4; Supplementary Fig. S5d) of vascular cells whose normal counterparts express Reck. Senger, Stratman, and Davis have proposed that mural-endothelial interaction triggers perivascular FN-deposition and subsequent vBM-deposition required for vascular stability in vivo 8,9 . Our data in vitro suggest that Reck plays a key role in this process by protecting FN from degradation ( Supplementary Fig. S7c). The failure of FN to fully normalize the association of Reck-mG cells to the microvessels (Fig. 5i) fits with this model, (Supplementary Fig.S7c) which places the mural-endothelial interaction upstream of FN-deposition; raising a new question as to how Reck promotes this upstream event. Attenuated PDGF-receptor immunoreactivity found in Reck-deficient cells ( Supplementary Fig. S8a and b) may be suggestive of its involvement in this failure, since PDGF is known to play a key role in pericyte-recruitment by endothelial cells 36 (Supplementary Fig. S8d-1).
The mechanism underlying the contribution of Reck-deficiency to the mislocalization of Reck-mG tip cells is presently unclear but several hypotheses can be envisaged. In Reck cKO culture, Flk1 signals were dampened and scarce in cells at the microvessel tips ( Supplementary Fig. S5d, magenta), suggesting abortive tip-stalk specification. Reck-deficient mice exhibit precocious neuronal differentiation, and this has been explained by attenuated Notch-signaling in neural precursor cells, due to de-regulated Adam10 that clips off Notch-ligands from adjacent cells 16 . In the vascular system, Notch signaling is known to suppress the tip cell phenotype and to promote acquisition of the stalk cell phenotype 6 . Hence, an obvious model is that attenuated Notch-signaling in this system leads to ectopic expression of tip cell phenotype, which can explain the excessive sprouting found in Reck cKO rings in ARA. This model, however, predicts upregulation and/or ectopic expression of tip-cell markers (such as Flk1), but  Fig. S5d). An alternative model involves direct Flk1-downregulation, for instance, by proteolytic cleavage that may scramble tip-stalk specification ( Supplementary Fig. S8d-2). Another potential model is binding between VEGF and FN 37 that somehow contributes to tip-stalk association or specification ( Supplementary Fig. S8d-3), which fits with our finding that FN may suppress the mislocalization of tip cells found in Reck cKO samples to some extent (Supplementary Fig. S8c; Fig. 5i).
Likewise, multiple models can be proposed to explain the increased cell proliferation found in Reck cKO cultures. Since Notch-signaling is known to confer quiescence to the vessels 6 , reduced Notch-ligands may promote cell proliferation. Alternatively, as observed in colon cancer cells 38 , Reck-deficiency may promote cell cycle progression by upregulating Skp2, thereby downregulating the Cdk inhibitor p27. These issues need to be addressed in future studies.
We have previously speculated that dilated vessels found in Reck-deficient mice may result from impaired non-sprouting angiogenesis (i.e., intussusception) 27 . The present data, however, support an alternative model that dilated vessels could result from aggregation and lateral fusion of excessive, unstable microvessels (Fig. 5j-4). Our data also raise the interesting possibility that Reck may serve as a key regulator of vascular morphogenesis  by regulating the number and size of the vessels, the area covered by the vessels, as well as the site and timing of anastomoses (Fig. 5j). We also speculate that RECK-dysfunction may underlie various conditions that give rise to fragile, leaky blood vessels, such as cancers 39 and RASopathies 40 , since several activated oncogenes, including mutated RAS, strongly suppress RECK expression 12,41 . Compounds capable of upregulating endogenous RECK 42 may be useful in ameliorating such conditions.

Methods
Mice. Animal experiments were approved by Animal Experimentation Committee, Kyoto University and conducted in accordance with its regulation. Generation and genotyping of mice carrying the Reck CrER (also known as Reck-CreER T2 or KI), Reck E1fx (also known as R1), Reck ∆ , or Reck Low (also known as R2neo) allele has been described elsewhere 21  Histology. Mouse embryos were fixed, sliced (10 μ m-thick), and stained with Hematoxylin and Eosin 21 , by immunohistochemistry with anti-CD31, anti-α SMA, or anti-laminin 27 , or by Kluver-Barrera method 43 as described previously. Tissues from mice carrying the mTmG reporter were fixed, incubated overnight in 30% sucrose, embedded in O.C.T, frozen at -80 o C, sliced (5μ m-thick), and observed with a fluorescent microscope.
Aortic ring assay. ARA was performed following the protocol of Baker et al. 29 with some modifications. The following conditions, optimised as shown in Supplementary Fig. S2, were used unless otherwise stated. Thoracic aortae were dissected from 5-week old transgenic mice after overnight starvation. The peri-aortic fibroadipose tissue and blood were carefully removed with fine microdissecting forceps, and aorta tunics were preserved without damage. After cleaning, aortae were cut into 1 mm-thick rings, rinsed 3 times with PBS (− ), and incubated overnight in Opti-MEM at 37 °C. The rings were then embedded in a mound of gel prepared on ice by mixing 5 volumes of 3 mg/ml collagen type I-A (Nitta Gelatin, Osaka, Japan), 3 volumes of 3 mg/ml collagen type IV (Nitta Gelatin, Osaka, Japan), 1 volume of 10× DMEM, and 1 volume of reconstitution buffer (2.2 g NaHCO 3 in 100 ml of 0.05 N NaOH and 200 mM HEPES). For embedding, 50 μ l of the collagen mixture was placed into wells of 96-well plates and incubated at 37 °C for 10 minutes to form a basal gel. Aortic rings were placed on the basal gel and covered with 10 μ l of cold collagen mixture. After 20-min incubation at 37 °C, each well was fed with 200 μ l Opti-MEM (Invitrogen) supplemented with 20 ng/ml VEGFa (R&D systems), 5% fetal horse serum (Cell Culture Laboratories, Ohio, USA), 100 U/ml penicillin and 100 μ g/ml streptomycin. The dishes were tightly sealed with parafilm, kept at 37 °C for two weeks, and examined every day under an inverted microscope. Micrographs were taken every day up to day 14 for morphometric analyses. In some experiments (Fig. 5c-i; Supplementary Fig. S7, S8c), bovine plasma fibronectin (Sigma, F1141) was added to the gel at 1 μ g/ml.
Morphometry. Microscopic images were recorded with a digital camera at various time points and magnification, depending on the properties to be assessed: x40 on day 3 to 7 for microvessel growth (number and length), x100 on day 8 and 9 for microvessel width, and x200 on day 10 to 14 for the cells associated with microvessels. The images were analysed using ImageJ to determine the length, width, and anastomosis of microvessels and the area covered by them. The numbers of microvessels were counted manually, following the criteria described by Aplin et al. 33 , where outgrowth constituted by singular cells such as fibroblasts, pericytes, and macrophages or segmented sprout structures were not included. AngioTool 44 was also employed to quantify microvessel area, branching points (total junctions), total length, end points, and lacunarity. Immunofluorescent staining. Aortic rings extending microvessels were fixed with 4% formalin, permeabilised with 0.25% Triton-X for 15 min and incubated in PBS (+ ) containing 10% goat serum for 1 h at room temperature to block non-specific binding.After being rinsed with PBS (+ ) for 5 min they were then incubated overnight with primary antibody diluted in PBLEC 29 . After rinsing with PBS (+ ) (3 × 10 min), samples were then incubated for 30 min at 37 °C in a cocktail of Alexa Fluor (488, 555, or Cy5) conjugated with appropriate secondary antibodies in PBLEC. Samples were rinsed then incubated (1 min) with Hoechst33342 and mounted using Fluoromount (Diagnostic BioSystems). Images were recorded using a fluorescence microscope (OlympusX70 or Keyence BZ-9000). The following primary antibodies were used: laminin (Progen 10765), CD31 (BD Pharmingen 550274), Reck [RECK-F, polyclonal rabbit antibodies (Matsuzaki et al., in preparation) and 5B11D12 12 ]; α -Sma (DAKO MO85), fibronectin (BD Biosciences, 610078), PDGFR-beta (Santa Cruz Biotechnology sc-432), Flk-1 (Santa Cruz Biotechnology sc-625), LAMA5 (Sigma-Aldrich SAB4501720) and Ki67 (Leica Biosystems NCL-Ki67p).