Pericyte ALK5/TIMP3 axis contributes to endothelial morphogenesis in the developing brain

Summary

The murine embryonic blood-brain barrier (BBB) consists of endothelial cells (ECs), pericytes (PCs) and basement membrane. Although PCs are critical for inducing vascular stability, signaling pathways in PCs that regulate EC morphogenesis during BBB development remain unexplored. Herein, we find that murine embryos lacking the TGFβ receptor activin receptor-like kinase 5 (Alk5) in brain PCs (mutants) develop gross germinal matrix hemorrhage-intraventricular hemorrhage (GMH-IVH). The germinal matrix (GM) is a highly vascularized structure rich in neuronal and glial precursors. We show that GM microvessels of mutants display abnormal dilation, reduced PC coverage, EC hyperproliferation, reduced basement membrane collagen and enhanced perivascular matrix metalloproteinase activity. Furthermore, ALK5-depleted PCs down-regulate tissue inhibitor of matrix metalloproteinase 3 (TIMP3), and TIMP3 administration to mutants improves endothelial morphogenesis and attenuates GMH-IVH. Overall, our findings reveal a key role for PC ALK5 in regulating brain endothelial morphogenesis and a substantial therapeutic potential for TIMP3 during GMH-IVH.

In brief (eTOC Blurb)

Dave et al. demonstrate that Pdgfrb-Cre, Alk5^(flox/flox) embryos (mutants) have deleterious non-cell autonomous effects on endothelial cells and germinal matrix hemorrhage (GMH). Mutant vessels display abnormal dilation, reduced pericyte coverage, endothelial hyperproliferation, reduced collagen and enhanced matrix metalloproteinase activity. ALK5-depleted pericytes down-regulate TIMP3, and TIMP3 administration attenuates GMH in mutants.

Introduction

Vascularization of the developing murine brain occurs through the process of angiogenesis (Marin-Padilla 1985), and the sub-ventricular zone predominantly consists of periventricular microvessels (capillaries) that are associated with mural cells called pericytes (PCs) (Armulik et al. 2011). Tight regulation of endothelial morphogenesis in the developing brain is essential for the formation of an intact blood-brain barrier (BBB). The murine embryonic BBB is comprised of endothelial cells (ECs) and PCs embedded in the basement membrane (Daneman et al. 2010). PCs play a key role in inducing vascular stability (Benjamin et al. 1998), and PC coverage of ECs is critical for BBB regulation during development, maintenance and disease (Armulik et al. 2011, Armulik et al. 2010, Bell et al. 2012, Daneman et al. 2010). EC-derived factors, such as platelet-derived growth factor (PDGF)-B are known to regulate PC recruitment to developing endothelial neovessels (Bjarnegard et al. 2004, Hellstrom et al. 1999, Larsson et al. 2001, Leveen et al. 1994), and embryonic mice lacking brain PCs (null for Pdgfb or its receptor Pdgfrb) display cerebral microaneurysms and poor BBB integrity (Bjarnegard et al. 2004, Leveen et al. 1994, Lindahl et al. 1997, Soriano 1994). However, signaling mechanism(s) in PCs that dictate crosstalk with ECs during development or disease are unknown.

TGFβ signaling is broadly implicated in vascular development and disease (Hirschi et al. 1998, ten Dijke and Arthur 2007), but its role in PC biology and in PC-mediated non-cell autonomous regulation of ECs remains elusive. Global knockout of the transforming growth factor (TGF)β type 1 receptor activin receptor-like kinase 5 (Alk5) in mice results in impaired yolk sac vasculogenesis and lethality at E10.5 (Larsson et al. 2001). With the exception of one study of limited scope in cultured bovine retinal PCs (Van Geest et al. 2010), prior investigations have not evaluated the role of ALK5 in PCs. Our studies demonstrate that embryonic mice with deletion of Alk5 in PCs develop a markedly severe cerebral phenotype in the germinal matrix (GM), known as germinal matrix hemorrhage-intraventricular hemorrhage (GMH-IVH). These findings demonstrate that ALK5 signaling in PCs is a key regulator of EC morphogenesis during BBB development.

GMH is associated with frank disruption of the BBB in the periventricular microvessels of the GM, a fragile and highly vascularized brain region rich in neuronal and glial precursors located ventral to the lateral ventricles. During severe GMH, the bleeding extends into the ventricles. Perinatal IVH is a major neurological disorder of prematurity that occurs in ~12,000 infants annually in the USA (Ballabh 2010, Guyer et al. 1997, Mazurek et al. 2017, Whitelaw 2001) and is associated with substantial neonatal mortality and morbidity (Klebermass-Schrehof et al. 2012, Patra et al. 2006). Interestingly, a study in human fetuses demonstrated lower levels of the ligand TGFβ1 and reduced PC coverage of EC vessels in the GM during mid-gestation compared to the white matter or cortex, and the authors suggested these findings increase propensity for GMH (Braun et al. 2007).

TGFβ1 induces tissue inhibitor of metalloprotease 3 (TIMP3) in prostate stromal cells (Cross et al. 2005) and chondrocytes (Su et al. 1998), and based on a vascular endothelial growth factor (VEGF) overexpression model, neurovascular proteases, including matrix metalloproteases (MMPs), are implicated in GMH-IVH pathogenesis (Yang et al. 2013). TIMP3 is a potent inhibitor of angiogenesis, and PC-derived TIMP3 induces vascular quiescence by blocking perivascular MMP activity (Brew and Nagase 2010) and inhibiting VEGF signaling in ECs (Qi et al. 2013, Qi et al. 2003). EC-PC co-culture studies have demonstrated that PC-derived TIMP3 expression is required to prevent EC tube regression (Saunders et al. 2006). Furthermore, recent studies in a traumatic brain injury mouse model suggest that mesenchymal stem cell-derived TIMP3 significantly improves the stability of the BBB (Menge et al. 2012). These findings raise the possibility that PC-derived TIMP3 protects against GMH.

Herein, we utilize Pdgfrb-Cre, Alk5^(flox/flox) mutant embryonic mice to investigate relevant mechanisms. Our findings demonstrate that attenuating the TGFβ pathway in PCs non-cell autonomously impairs EC tube morphology in the GM of the developing brain. GM vessels of mutants display hyperproliferative ECs that form abnormally distended tubes with reduced PC coverage. Additionally, mutant GM vessels exhibit enhanced perivascular gelatinolytic activity and reduced collagen levels in the basement membrane. Complementary studies suggest that siRNA-mediated silencing of ALK5 in human brain PCs impairs network formation of co-cultured human brain ECs, and conditioned media (CM) from ALK5-silenced PCs induces a hypermigratory phenotype in ECs. TGFβ upregulates TIMP3 in cultured PCs, and Alk5-deficient PCs have reduced TIMP3 expression in culture and in vivo. Furthermore, we show that administration of a broad MMP inhibitor or recombinant TIMP3 (rTIMP3) significantly reduces GMH in mutant mice and improves PC coverage of EC vessels. Our findings provide critical insights about the role of brain PC ALK5 in regulating endothelial development during GM vessel morphogenesis. Importantly, these studies demonstrate a strong therapeutic potential of TIMP3 for GMH pathogenesis, which promises to direct further investigations geared at devising effective therapies to combat this devastating disease of prematurity.

Results

Vascular morphogenesis in the GM

During murine embryonic development, the GM is comprised of lateral and medial ganglionic eminences (GEs), which arise from the ventral telencephalon at embryonic day (E) E11.5 (Fig. S1A). With the morphogenesis of the lateral ventricles by E12.5, the GEs become more prominent (Fig. S1A). Vascularization of the GM begins with the appearance of nascent isolectin B4 (IB4)⁺ EC vessels that mostly lack mural cell coverage at E10.5 but rapidly become covered by neuron-glial antigen 2 (NG2)⁺ PCs over the next day (Figs. S1B, C). Similarly, in the GM of Pdgfrb-Cre, ROSA26R^(mTmG/+) embryos, GFP marks NG2⁺ and PDGFRβ⁺ PCs by E11.5 but is not detectable in EC vessels in the GM that lack PCs at E10.5 (Figs. S1D, E). The GM of Pdgfrb-Cre, ROSA26R^(Zs/+) embryos were further utilized to determine the specificity of the Pdgfrb-Cre transgenic line (Fig. S2). Pdgfrb-Cre does not mark ECs (Figs. S2A–D), and in the GM, the expression of Pdgfrb-Cre is limited to mural cells (i.e., PCs) of alpha-smooth muscle actin (SMA)⁻ small caliber vessels (capillaries; Fig. S2E).

Hemorrhage in the GM of Pdgfrb-Cre, Alk5^(flox/flox) mutant embryos

Because TGFβ is critical for smooth muscle cell (SMC) development and homeostasis (Carvalho et al. 2007, Li et al. 2014), and the role of the TGFβ receptor ALK5 in PC biology has not been analyzed in vivo, we studied the morphogenesis of Pdgfrb-Cre, Alk5^(flox/flox) mutant embryonic mice. During murine development, Alk5 is broadly expressed in the GM at E11.5 and E13.5 (Figs. S2F, G), consistent with previous work, (http://www.informatics.jax.org/assay/MGI:4450893#Tgfbr1_E11_5_id). Unless noted otherwise, in all the mouse studies described herein, the mutant genotype is Pdgfrb-Cre, Alk5^(flox/flox) and the control genotype is Pdgfrb-Cre, Alk5^(flox/+). These mutants develop intracerebral hemorrhage (ICH) which is first grossly visible at E12.5 (Fig. 1A), increases in severity by E13.5 (Figs. 1A–C) and results in perinatal lethality (data not shown). To microscopically analyze this phenotype, coronal sections of controls and mutants at E12.5 and E13.5 were stained by H&E (Figs. 1D, E). At E12.5, ICH was detected predominantly in the medial GEs of mutants, and by E13.5, severe hemorrhage encompasses the lateral and medial GEs and extends to the surrounding brain parenchyma and lateral ventricles. Since Pdgfrb-Cre marks both PCs and SMA⁺ SMCs (Fig. S2E) (Foo et al. 2006), we analyzed SMA expression in the GM of Pdgfrb-Cre, ROSA26R^(Zs/+) embryos. Our data suggest that the GM vessels consist of Zs⁺ (Cre-reporter) mural cells that lack SMA expression (i.e., PCs) (Fig. S2E). Furthermore, we generated Acta2-CreER, Alk5^(flox/flox) embryos to determine the contribution of SMA⁺ cells (e.g. SMCs) in the GMH phenotype of Pdgfrb-Cre, Alk5^(flox/flox) mutants. Our results indicate that tamoxifen-induced Acta2-CreER, Alk5^(flox/flox) mutants do not develop GMH (Figs. S2H–J). Thus, PC-ALK5 is critical in regulating vascular development in the brain, and Pdgfrb-Cre, Alk5^(flox/flox) mutant embryos develop gross GMH-IVH phenotype.

Figure 1. Brain hemorrhage of Pdgfrb-Cre, Alk5^(flox/flox) mutant embryonic mice.

(A, B, D and E) Mutant and Pdgfrb-Cre, Alk5^(flox/+) control embryos are shown at E11.5–13.5 as indicated. In (A), whole mount views of embryos demonstrate gross ICH initially visible at E12.5 (black arrows). In (B) Hemorrhage area is quantified as the ratio of hemorrhage area (white dotted-line) to head area (red dotted-line). In (D and E), H&E stained coronal brain sections demonstrate GMH (yellow asterisks) and IVH (white +’s) in mutants. Boxed regions are shown as close-ups on the right. L and M, lateral and medial ganglionic eminences; LV, lateral ventricle. Scale bars, 1 mm (A, B) and 200 μm (D, E). (C) Quantification of gross hemorrhage area in mutants at E11.5–13.5, calculated as described in (B). Each symbol represents an individual embryo. Bars indicate averages ± SD. One way ANOVA with Tukey’s post hoc test, ***p<0.001, ****p<0.0001. See also Figures S1 and S2.

We next examined the time course of blood vessel pathology in the developing GM of mutant Pdgfrb-Cre, Alk5^(flox/flox) embryos by staining for markers of ECs (Cadherin5, CDH5) and PCs (NG2). At E10.5, similar to control embryos, most of the nascent GM vessels in mutants are CDH5⁺ NG2⁻ (Fig. S3A). As noted above, the normal GM vasculature at E11.5 consists of EC vessels largely covered by PCs whereas in mutants at this time point, many CDH5⁺ EC vessels are deficient in coverage by NG2⁺ PCs (Figs. 2A, C). At E12.5 and E13.5, when extensive hemorrhage is grossly evident in the mutants, the CDH5⁺ EC vessels in the GM have become dilated and tortuous (Figs. 2B, S3A). As observed at E11.5, the mutant GM vessels at E13.5 show incomplete coverage by NG2⁺ PCs (Figs. 2B, C). Additionally, mutant GM vessels at E13.5 display increased lumen area, vessel area and vessel diameter (Fig. 2D) and fail to develop a functional BBB at E15.5 (Fig. S3B) (Ben-Zvi et al. 2014). In contrast to vessels of the GM, those of the cortex lack detectable abnormalities (Fig. S4). Analysis of PC coverage of EC vessels and EC vessel morohology revealed no difference in the cortex of controls and mutants (Figs. S4A–C). To understand the anatomical specificity of the observed phenotype in the brain, we assessed the pSMAD3 levels in PCs of the GM and cortex (Fig. S4D). Our results indicate that the percent of PCs that are pSMAD3⁺ is ~2.5 fold higher in GM vessels compared to PCs in cortical vessels (Fig. S4E).

Figure 2. Abnormal EC and PC phenotype in GM of Pdgfrb-Cre, Alk5^(flox/flox) mutants.

(A and B) GM sections of mutant and Pdgfrb-Cre, Alk5^(flox/+) control embryos at E11.5 (A) and E13.5 (B) stained for CDH5 (ECs) and NG2 (PCs). Boxed regions are shown as close-ups in columns on the right. Arrowheads indicate areas of mutant EC vessels lacking PC coverage. Scale bars, 25 μm. (C and D) Quantification of PC coverage of EC vessels (at indicated ages) and of EC vessel area, diameter and lumen area at E13.5. At least 150 blood vessels were scored per mouse (n=3 mice for each condition). Data are averages ± SD. Student’s t test, **p<0.01, ***p<0.001, ****p<0.0001. See also Figures S3 and S4.

Enhanced EC proliferation in the GM of mutant embryos

Embryonic mice lacking brain PCs have enhanced EC proliferation (Hellstrom et al. 2001). Given the marked EC vessel distention with reduced PC coverage in Pdgfrb-Cre, Alk5^(flox/flox) embryos, we next assessed vascular cell proliferation in the GM of control and mutant embryos. Pregnant dams were injected with bromodeoxyuridine (BrdU). Six hours later, embryos were harvested, and sections were stained for BrdU, isolectin B4 (IB4, EC marker) and the EC-specific nuclei marker ERG (Figs. 3A, S3C). In comparison to controls, mutants have a ~2 fold increase in BrdU⁺ ERG⁺ EC nuclei at E11.5–13.5 (Figs. 3B, S3D). A similar increase in EC proliferation was confirmed by an independent approach of co-staining for the proliferation marker Ki67, IB4 and DAPI (Figs. 3C, D). In contrast to these results in GM vessels, proliferation of cortical vessel ECs in controls and mutants does not differ (Figs. S4F, G). Because of species overlap between antibodies directed against PCs and proliferation markers, we utilized mutant and control mice also carrying the ROSA26R^(mTmG/+) Cre reporter and identified PC nuclei as ERG⁻ DAPI⁺ nuclei in GFP⁺ cells (Fig. 3E). BrdU incorporation studies revealed no difference in PC proliferation between mutant and control vessels at E11.5 and E13.5 (Fig. 3F).

Figure 3. EC hyperproliferation in Pdgfrb-Cre, Alk5^(flox/flox) mutant embryos.

(A, C and E) GM sections of mutant and Pdgfrb-Cre, Alk5^(flox/+) control embryos in (A and C) also bearing ROSAR26^(mTmG/+) in (E) at E11.5 and E13.5 were stained for: ERG (EC nuclei), isolectin B4 (IB4, EC) and BrdU in (A); Ki67 (proliferation marker), IB4 and propidium iodide (PI, nuclei) in (C); and ERG, GFP (Cre-reporter), BrdU and DAPI in (E). BrdU was injected in pregnant dams 6 hours prior to dissection in (A and E). Yellow arrowheads indicate proliferative ECs as marked by BrdU⁺ ERG⁺ nuclei of IB4⁺ cells in (A) and Ki67⁺ nuclei within IB4⁺ vessels in (C). Yellow arrows indicate proliferative PCs as marked by BrdU⁺ DAPI⁺ and ERG– nuclei in GFP⁺ cells in (E). Scale bars, 25 μm. (B, D and F) Histograms representing the percentage of ECs or PCs that are proliferative in the GM vessels for control and mutant embryos at E11.5 and E13.5. Quantification of the percentage of EC nuclei that are BrdU⁺ in (B) or Ki67⁺ in (D) and PC nuclei that are BrdU⁺ in (F). At least 150 blood vessels were scored per mouse (n=3 mice for each condition and age). Data are averages ± SD. Two-way ANOVA with Sidak’s multiple comparison test, **p<0.01, ***p<0.001, versus control.

Effects of PC ALK5 silencing on EC proliferation, migration, sprouting and capillary network formation

Because of the dramatic EC phenotype in Pdgfrb-Cre, Alk5^(flox/flox) mutants, we sought to query the role of ALK5 in PCs using a reductionist approach. To this end, we evaluated the effects of ALK5 silencing in cultured PCs on PC and EC function. siRNA-mediated knockdown of ALK5 in human brain PCs down-regulated downstream effectors phosphorylated SMAD2/3 in PCs (Fig. S5A), reduced PC attachment (Fig. S5B) and decreased PC migration (Figs. S5C, D). In addition to these direct effects on PCs, conditioned media (CM) from ALK5-silenced PCs induces non-cell autonomous effects on human brain microvascular ECs. ECs cultured in CM from ALK5-silenced PCs have enhanced EC migration (Figs. 4A, B) and proliferation (Fig. 4C), as well as sprouting and invasion in three dimensional (3D) collagen matrices (Figs. S5E, F). Furthermore, co-culture of ALK5-silenced PCs with ECs on 3D Matrigel disrupts EC capillary network and cord formation (Figs. 4D–G, S5G).

Figure 4. Effects of human PC-ALK5 silencing on EC migration, proliferation and endothelial capillary network formation.

(A) ECs were cultured in conditioned media (CM) collected from PCs pretreated with scrambled (Scr) or ALK5-specific siRNA (siALK5) for 48 hours. Pretreated ECs were placed in a Boyden apparatus top chamber to assess migration towards VEGF-A (100 ng/ml) for 10 hours, and migrated cells (i.e., on the membrane’s bottom surface) were stained with Crystal Violet. (B) Quantification of migrated ECs per field from images as shown in (A) (n=3 experiments, 6 fields per experiment). (C) ECs were cultured in CM collected from PCs pretreated with scrambled (Scr) or ALK5-specific siRNA (siALK5) for 48 hours or 72 hours. BrdU was added 6 hours prior to fixing ECs and percent of BrdU⁺ ECs were quantified (n=3 experiments, 9 fields per experiment). (D) PCs pretreated with Scr siRNA or siALK5 were co-cultured with ECs on Matrigel for 6 hours, and Brightfield images are shown. (E to G) Histograms generated from images as in (D) show EC branches (E), branch points (F) and branch length (G) for siALK5 PC + EC co-culture normalized to Scr PC + EC control (n=3 experiments, 6 fields per experiment). Scale bars, 100 μm. Data are averages ± SD. Student’s t test, *p<0.05, **p<0.01, ***p<0.001. See also Figure S5.

Abnormal basement membrane and enhanced protease activity in mutant GM vessels

Thus, reduced ALK5 in PCs induces non-cell autonomous deleterious effects on ECs both in vitro and in vivo. We next examined deposition of basement membrane constituents around GM vessels of control and mutant embryos. To avoid antibody species overlap, we utilized Pdgfrb-Cre, ROSA26R^(mTmG/+) mice (to label PC sleeves) and co-stained with collagen type IV (Fig. 5A) or collagen type I (Fig. 5B). Immunohistochemistry revealed markedly reduced levels of collagen type IV and type I around PC sleeves in the mutant GM (Fig. 5C) but not in the cortex (Figs. S4H, I).

Figure 5. Abnormal basement membrane and enhanced MMP activity in Pdgfrb-Cre, Alk5^(flox/flox) mutants.

(A, B, F and G) GM sections of Pdgfrb-Cre, Alk5^(flox/+) control and mutant embryos also carrying ROSA26R^(mTmG/+) at E13.5 stained for GFP (PCs) and either collagen type IV (Col-IV) (A), collagen type I (Col-I) (B), MMP9 (F) or MMP2 (G). In mutants, white arrowheads indicate GFP⁺ PC sleeves lacking basement membrane Col-IV (A) and Col-I (B), and arrows indicate up-regulated perivascular MMP9 (F) and MMP2 (G) levels in mutants. (C) Quantification of Col-IV and Col-I coverage of GFP⁺ PC sleeves at E13.5 from images in (A and B). (H) Quantification of MMP2 and MMP9 staining intensity around GFP⁺ PC sleeves at E13.5 from images in (F and G). In (C and H), at least 100 blood vessels were scored per mouse (n=3 mice for each condition). Data are averages ± SD. Two-way ANOVA with Sidak’s multiple comparison test, *p<0.05, **p<0.01. (D and E) In situ zymography with fluorescein-conjugated DQ-gelatin substrate and staining for IB4 of GM cryosections from mutant and control embryos at E13.5. Yellow arrowheads indicate gelatinolytic activity around IB4⁺ vessels in mutants. Treatment of cryosections with a MMP inhibitor (BB94 or rTIMP3) in (E) attenuates the enhanced gelatinolytic activity of mutants. Scale bars, 25 μm. See also Figures S4 and S5.

In addition to the cellular changes, neurovascular proteases, including MMPs, are implicated in GMH pathogenesis (Yang et al. 2013), and thus, we next evaluated protease activity in the GM of mutants. In situ zymography with the DQ-gelatin substrate demonstrated a marked increase in gelatinolytic activity around IB4⁺ GM vessels of mutant mice (Fig. 5D), but not cortical vessels (Fig. S4J), and treatment of mutant cryosections with the MMP inhibitor batimastat (BB94) substantially blocked this enhanced perivascular gelatinolytic activity (Fig. 5E). Furthermore, immunohistochemistry revealed up-regulation of MMP9 and MMP2 in the GM around PC sleeves in mutant embryos carrying the ROSA26R^(mTmG/+) Cre reporter (Figs. 5F–H). In addition, PCs and ECs isolated from the brains of mutant embryos also carrying the ROSA26R^(Zs/+) Cre reporter (Figs. S5H–J, L) have increased transcript levels of Mmp2 and Mmp14 in ECs (Fig. S5K) and Mmp3, Mmp9, Mmp13 and Mmp14 in PCs (Fig. S5M). Hence, up-regulated MMPs in mutants may contribute to GMH pathogenesis.

To test the effects of MMP inhibition, we added BB94 to co-cultured human brain ALK5-silenced PCs and microvascular ECs and observed that BB94 attenuates the abnormal EC capillary network phenotype (Figs. S6A–D). Furthermore, BB94 injection of pregnant dams bearing mutants markedly reduced GMH in mutant embryos at E12.5 (Figs. S6E, F) and E13.5 (Figs. S6G–I). Immunostaining for markers of ECs (CDH5) and PCs (NG2) reveal that GM vessels of BB94-treated mutants show ~25% increase in coverage by NG2⁺ PCs as compared to mutants without treatment (Figs. S6J, K).

Reduced TIMP3 with Alk5 down-regulation and beneficial effects of rTIMP3

TIMP3 is a major MMP inhibitor that induces EC tube stabilization and vascular quiescence (Qi et al. 2013, Qi et al. 2003, Saunders et al. 2006), and TGFβ1 stimulates TIMP3 expression in chondrocytes and prostate stromal cells (Cross et al. 2005, Su et al. 1998). To examine the expression pattern of Timp3 in the GM of developing murine brain, we performed in situ hybridization. At E11.5, the expression pattern of Timp3 in the GM (Fig. S7A) is reminiscent of the developing vascular network. We hypothesized that depletion of PC ALK5 results in TIMP3 deficiency, leading to up-regulation of MMPs, basement membrane degradation, EC tube destabilization and ultimately GMH. Indeed, knockdown of ALK5 in cultured human PCs down-regulates TIMP3 transcript levels by ~65% (Fig. 6A), and conversely, TGFβ1 treatment induces TIMP3 mRNA expression (Fig. 6B). EC-PC co-culture has been shown to induce TIMP3 protein levels (Saunders et al. 2006), but TIMP3 expression is relatively reduced in co-cultures of ALK5-silenced PCs with ECs (Fig. 6C). Furthermore, we isolated brain PCs from E13.5 embryonic mutant mice and controls also carrying ROSA26R^(Zs/+) (Figs. S5H–J, L) and similarly found that Timp3 transcript levels are reduced by more than 50% in PCs from mutants (Fig. 6D). In contrast, ECs isolated from mutants and controls displayed no difference in Timp3 transcript levels (Fig. S7B).

Figure 6. TGFβ pathway regulates PC TIMP3 levels, and rTIMP3 mitigates effects of ALK5-silenced PCs on EC network.

(A) ALK5 and TIMP3 mRNA levels as assessed by qRT-PCR relative to GAPDH in Scr- and siALK5-treated human PCs and normalized to Scr levels (n=3). Student’s t test, ***p<0.001, ****p<0.0001, versus Scr. (B) PCs were treated with TGFβ1 for indicated times and RNA was isolated. Histogram shows TIMP3 mRNA levels relative to 18S rRNA as assessed by qRT-PCR and normalized to 0 hour (n=3). One-way ANOVA, *p<0.05 versus 0 hour. (C) Western blot of TIMP3 and GAPDH from co-cultures of PCs pretreated with Scr or siAlk5 and ECs (n=3). (D) Zs⁺ NG2⁺ CD31⁻ cells (PCs) were isolated by FACS from the brains of E13.5 Pdgfrb-Cre, ROSA26R^(Zs/+) embryos also carrying either Alk5^(flox/+) (control) or Alk5^(flox/flox) (mutant) as described in Figure S5. RNA levels of Alk5 and Timp3 relative to 18S rRNA were measured by qRT-PCR and normalized to control (n=3). Student’s t test, **p<0.01, ***p<0.001, versus control. (E) ECs co-cultured with Scr or siALK5 pretreated PCs on Matrigel in the presence of vehicle or rTIMP3. Representative Brightfield images are shown. Scale bar, 100 μm. (F to H) Histograms showing EC branches (F), branch points (G) and branch length (H) for rTIMP3- or vehicle-treated co-culture of ECs and Scr or siALK5-pretreated PCs normalized to vehicle-treated Scr PC + EC control (n=3 experiments, 6 fields per condition for each experiment). One-way ANOVA with Tukey’s post hoc test, **p<0.01, ***p<0.001. All data are averages ± SD. See also Figures S5–S7.

We further confirmed PC-specific down-regulation of Alk5 transcript levels by isolating PCs and ECs from the embryonic brains of controls and mutants. Our results indicate that in mutants, PCs have ~85% reduced Alk5 transcript levels (Fig. 6D) while ECs from the brain as well as other organs (lung and liver) do not have altered Alk5 RNA levels (Figs. S7B, C). Moreover, to rule out non-specific effects on Alk5 signaling in ECs, we assessed the percent of ECs that are pSMAD3⁺ in the GM of controls and mutants (Fig. S7D). Our data indicate that ECs of mutant GM vessels do not display any difference in pSMAD3 expression (Fig. S7E). We also assessed Alk1 transcript levels in PCs and ECs of mutants. In agreement with a previous study (Van Geest et al. 2010), our data confirm that Alk1 is expressed in ECs but not PCs (Fig. S7F). Moreover, Alk1 transcript levels in ECs of controls and mutants does not differ (Fig. S7F).

Given increased protease activity and reduced PC TIMP3 within the mutant GM, we next explored the effects of rTIMP3 on EC-PC co-cultures and in vivo. Addition of rTIMP3 to co-cultures of ECs and ALK5-silenced PCs improved EC cord formation and branching (Figs. 6E–H). As with BB94, rTIMP3 treatment abrogated the enhanced gelatinolytic activity of the mutant GM (Fig. 5E). To evaluate whether rTIMP3 could alleviate GMH in mice, pregnant dams were injected daily with rTIMP3 or vehicle from E10.5–12.5, and embryos were harvested at E13.5. rTIMP3 treatment significantly attenuated gross brain hemorrhage in whole mount embryos and GMH/IVH in sections of mutants (Figs. 7A–C without affecting controls S7G, H). To microscopically examine blood vessel morphology, GM sections of vehicle- or rTIMP3-treated mutants were stained with CDH5 (ECs) and NG2 (PCs) (Figs. 7D, S7I). Mutant GM vessels with rTIMP3 treatment revealed significant improvement in NG2⁺ PC coverage of CDH5⁺ EC vessels (Fig. 7E) and reduced vessel area, lumen area and vessel diameter (Fig. 7F). In addition, quantification of Ki67⁺ nuclei within IB4⁺ vessels showed reduced EC proliferation in GM vessels of mutants treated with rTIMP3 (Figs. 7G, H). Effective delivery of rTIMP3 or BB94 to developing embryonic brains at E10.5–12.5 is consistent with the finding that the embryonic BBB is not functional until E15.5 (Fig. S3B) (Ben-Zvi et al. 2014). Taken together, our data suggest that deletion of Alk5 in PCs reduces TIMP3 levels and increases perivascular MMP activity resulting in basement membrane degradation, excess EC proliferation and migration and culminating in abnormal vessel morphogenesis, BBB breakdown and ultimately GMH/IVH (Fig. S7J).

Figure 7. rTIMP3 attenuates GMH in Pdgfrb-Cre, Alk5^(flox/flox) mutant embryonic mice.

Pregnant dams were injected daily from E10.5–12.5 with either vehicle (PBS) or rTIMP, and mutant embryos were harvested at E13.5. (A) Brightfield images of whole mutant embryos are shown with black arrows indicating hemorrhage. (B) Quantification of gross hemorrhage with each symbol representing an individual embryonic mouse at E13.5. Data are averages ± SD. Student’s t-test, ***p<0.001. (C) H&E stained coronal brain sections. GMH and IVH are denoted by yellow asterisks and white +’s, respectively. (D) GM coronal sections stained for CDH5 (EC) and NG2 (PC). Boxed regions are shown in the right columns as close-ups with white arrowheads indicating areas on EC vessels lacking PC coverage. (E and F) Histograms of PC coverage of EC vessels normalized to vehicle controls and EC vessel area, diameter and lumen area in controls or mutants treated with vehicle (PBS) or rTIMP3. At least 150 blood vessels were scored per mouse (n=3 mice for each condition). Data are averages ± SD. One way ANOVA with Tukey’s multiple comparision test, ****p<0.0001, **p<0.01 as compared to mutant with rTIMP3, #p<0.0001 as compared to control with vehicle. (G) GM sections stained for Ki67 (proliferation marker), IB4, and propidium iodide (PI, nuclei). Arrowheads indicate Ki67⁺ nuclei within IB4⁺ vessels. (H) Percentage of EC nuclei that are Ki67⁺ per treatment group. At least 200 blood vessels were scored per mouse (n=3 mice for each condition). Data are averages ± SD. Student’s t-test, *p<0.05. Scale bars, 1 mm (A), 200 μm (C) and 25 μm (D, G). See also Figure S7.

Discussion

Our present study establishes that ALK5-mediated signaling in PCs is critical for GM vessel development and especially for EC tube morphogenesis. This work has four main results: (1) Pdgfrb-Cre, Alk5^(flox/flox) mutant embryos develop gross GMH-IVH; (2) deletion of Alk5 in PCs has deleterious non-cell autonomous effects on EC morphology and induces perivascular protease activity; (3) TGFβ-ALK5 signaling in PCs regulates TIMP3 expression; and (4) rTIMP3 treatment markedly attenuates GMH in mutants. These findings are key steps towards understanding the role of PC-Alk5 signaling during BBB development and pathophysiology of GMH-IVH.

PCs are known to play a critical role in vascular homeostasis by inducing EC maturation and quiescence (Armulik et al. 2011, Benjamin et al. 1998, Hirschi and D’Amore 1996) and also regulate the BBB in development, postnatally and during aging and disease (Armulik et al. 2010, Bell et al. 2010, Daneman et al. 2010, Winkler et al. 2011); however, there is limited understanding of any signaling pathway(s) in PCs that modulates EC morphogenesis, BBB formation and/or ICH. Hence, studies of the effects of the TGFβ pathway, and specifically ALK5-mediated signaling, in PCs in vivo are critical. Our investigations establish that ALK5 in PCs is required to form an intact brain vasculature. During normal GM development, nascent EC vessels appear at ~E10.5 and by E11.5, are largely covered by PCs (Fig. S1). Pdgfrb-Cre, Alk5^(flox/flox) mutants are not distinguishable from controls at E10.5 (Fig. S3), but beginning at E11.5, mutants display hyperproliferative ECs and abnormally dilated EC vessels that have deficient PC coverage (Figs. 2, 3). By E12.5, mutants develop gross ICH (Fig. 1). The increased proliferation of ECs (but not PCs) in response to depletion of Alk5 in PCs (Fig. 3) most likely contributes to the reduced PC coverage destabilizing mutant EC vessels and resulting in abnormal distension and tortuosity which are detrimental to BBB morphogenesis.

Although the role of TGFβ signaling in PCs has not been previously examined in BBB development or GMH-IVH, a number of studies suggest that a key pathway in GMH pathogenesis involves inhibition of TGFβ-mediated signaling in ECs (Arnold et al. 2014, Li et al. 2011, Nguyen et al. 2011). Indeed, EC deletion of TGFβ receptors Alk5 or Tgfbr2 (with Alk1-Cre) or the downstream effector Smad4 (with SP-A-Cre) results in aberrant vascular sprouting and ICH (Li et al. 2011, Nguyen et al. 2011). In addition, mice null for Itgav (McCarty et al. 2002) or with neuroepithelial deletion (with Nestin-Cre) of Itgb8 have impaired TGFβ signaling in ECs and develop glomeruloid-like EC collections and GMH (Arnold et al. 2014). Thus, in the present study we have meticulously analyzed and validated that PCs, but not ECs, in the GM vessels of Pdgfrb-Cre, ROSAR26^(Zs/+) embryos are marked by the Cre-reporter Zs (Fig. S2). Furthermore, ECs from controls and mutants do not differ in Alk5 transcript levels or percentage of pSMAD3⁺ ECs (Fig. S7). Pdgfrb-Cre also targets SMCs, however, our data suggest that the mural cells of GM vessels express NG2 and PDGFR-β but not SMA (Fig. S2), which is the molecular signature of brain PCs (Hill et al. 2015). Taken together, these findings demonstrate that ALK5-mediated signaling in PCs is indispensable for BBB development of GM vessels.

In addition to cell-type specificity, detailed spatial analysis revealed that the hemorrhage and aberrant vessel morphology in mutants is restricted to the GM. Mutant cortical vessels were indistinguishable with regard to EC proliferation, PC coverage, vessel caliber, perivascular proteolytic activity and reduced basement membrane collagen (Fig. S4). The anatomical specificity of the brain hemorrhage in mutants is intriguing and is likely attributed in part to the following factors: (1) high adundance of PCs in the CNS compared to other organs (Armulik et al. 2011); (2) rapid development of periventricular vessels in the GM (Vasudevan and Bhide 2008); and (3) fragility of the GM vasculature (Ballabh 2010). Importantly, we find that ALK5-mediated signaling is ~2.5-fold higher in PCs of periventricular GM vessels in comparison to PCs of cortical vessels during normal development (Fig. S4). This regional heterogeneity of ALK5 signaling in PCs may reflect developmental differences in PC sub-populations or compartment-specific gradients of TGFβ ligand within the embryonic brain and warrants further investigations.

Enhanced MMPs are widely implicated in disease pathogenesis following traumatic brain injury or acute stroke (Lakhan et al. 2013, Sifringer et al. 2007). Notably, over-expression of VEGF in the embryonic murine GM induces neurovascular proteases and results in GMH-IVH (Yang et al. 2013), and PC expression of MMP9 is an important factor in APOE4-mediated breakdown of the BBB (Bell et al. 2010). In agreement with these findings, Pdgfrb-Cre, Alk5^(flox/flox) embryos have increased perivascular MMP activity and expression of MMP2, 3, 9 and 14 (Figs. 5, S5). In physiological conditions, the capillary basement membrane modulates cell-cell and cell-matrix interactions and inhibits EC proliferation and migration. However, in disease, MMP-dependent basement membrane degradation induces ECs from intact vessels to migrate and proliferate (Kalluri 2003). Specifically, MMP9-induced basement membrane degradation has been shown to liberate and mobilize extracellular-bound pro-angiogenic growth factors initiating EC activation, proliferation and migration (Bergers et al. 2000). Thus, factors liberated from MMP-mediated extracellular matrix breakdown may induce EC hyperproliferation in Pdgfrb-Cre, Alk5^(flox/flox) mutants.

TIMP3 is a PC-derived MMP inhibitor that induces vessel quiescence and vascular stabilization (Saunders et al. 2006, Schrimpf et al. 2012). TIMP3 is expressed in perivascular mesenchyme of a developing murine embryo, expression of PC TIMP3 is strongly induced by ECs in vitro, and silencing of PC-TIMP3 induces EC tube regression (Saunders et al. 2006). In situ hybridization suggests that Timp3 is expressed in the GM at E11.5, and the pattern of TIMP3 expression is similar to the developing GM vasculature (Fig. S7A). Furthermore, TGFβ induces TIMP3 expression in PCs, and ALK5-depleted PCs express lower levels of TIMP3 both in mice and cultured cells (Fig. 6). Addition of rTIMP3 to co-cultures of brain microvascular ECs and ALK5-silenced brain PCs or to dams pregnant with Pdgfrb-Cre, Alk5^(flox/flox) embryos significantly rescues EC capillary morphogenesis and attenuates GMH-IVH (Figs. 6, 7). The beneficial effects of TIMP3 may be largely due to inhibition of MMPs; however, other mechanisms are likely to contribute as well. For instance, up-regulation of the VEGF pathway is implicated in GMH pathogenesis, and VEGF inhibitors reduce GMH in a rabbit hyperosmolality model (Ballabh et al. 2007, Yang et al. 2013). TIMP3 binds to VEGFR2, inhibiting VEGF-mediated signaling and angiogenesis (Qi et al. 2013, Qi et al. 2003). In addition, TIMP3 can restore EC barrier dysfunction after traumatic brain injury (Menge et al. 2012) and has direct neuroprotective functions that enhance neuronal survival and dendritic outgrowth (Gibb et al. 2015). Finally, anti-inflammatory effects of TIMP3 (Mohammed et al. 2004) may be central.

In contrast to the gross ICH in embryos with disrupted ALK5 signaling in PCs (Fig. 1), mutants with reduced PCs (Pdgfb or Pdgfrb hypomorphs or nulls) have developmental brain microaneurysms (Hellstrom et al. 2001, Leveen et al. 1994, Lindahl et al. 1997, Soriano 1994) and increased BBB permeability due to aberrant vesicle trafficking in ECs (Armulik et al. 2010, Daneman et al. 2010) but lack frank ICH. This marked difference and the fact that rTIMP3 induces only a partial rescue of the phenotype (Fig. 7) suggests that other mechanisms contribute to the GMH phenotype as well. Our cell culture experiments demonstrate that conditioned media from Alk5-silenced PCs is sufficient to induce EC hyperproliferation, higher EC migration and increased angiogenesis (Figs. 4, S5). We hypothesize that ALK5-silenced PCs up-regulate expression and secretion of certain pro-angiogenic factors which promote abnormal EC morphogenesis and GMH. In the future, it will be critical to identify these pro-angiogenic factors to better understand the molecular mechanism(s) of GMH pathogenesis.

Overall, our studies yield essential findings characterizing the process of neurovascular morphogenesis and the pathogenesis of GMH. The results emphasize an essential role of signaling in PCs for non-cell autonomous EC vessel morphogenesis and elucidate a compelling mechanism involving PC ALK5-TIMP3 and crosstalk with microvascular ECs in the developing brain. Thus, augmenting TIMP3 activity may be an attractive therapeutic strategy for perinatal GMH-IVH, a common and devastating complication of prematurity that urgently needs efficacious therapies.

The STAR Methods

Contact for reagent and resource sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Daniel Greif (daniel.greif@yale.edu). Requests will be handled according to Yale University policy regarding MTA and related matters.

Experimental model and subject details

Mice

All mouse experiments were approved by the Institutional Animal Care and Use Committee at Yale University. C57BL/6 wild type, ROSA26R^(Zs/Zs) (Madisen et al. 2010) and ROSA26R^(mTmG/mTmG) (Muzumdar et al. 2007) mice were purchased from Jackson Laboratory. Alk5^(flox/flox) (Larsson et al. 2001), Pdgfrb-Cre (Foo et al. 2006) and Acta2-CreER (Wendling et al. 2009) mice were previously described. Both male and female embryonic mice were used in these experiments.

Cell culture

PCs and microvascular ECs from human brain (ScienCell) were cultured up to passage 6 in low glucose (LG) DMEM supplemented with 10% FBS or complete EC medium (ScienCell), respectively.

Method details

Animal breeding and treatments

Mice were bred and embryos harvested at different ages with E0.5 considered the time of vaginal plug. All agents were injected intraperitoneally. Embryos were harvested the day following the last treatment of pregnant dams with BB94 (Sigma, 30 mg/kg/dose), which was injected daily either from E10.5–11.5 or E10.5–12.5. Similarly, vehicle (PBS) or rTIMP3 (R&D, 80 μg/kg/dose) was administered daily from E10.5–12.5, and embryos were analyzed at E13.5. For Acta2-CreER animals, tamoxifen (Sigma, 1 mg/day) was injected daily from E8.5–12.5, and embryos were harvested at E13.5. For proliferation studies, pregnant dams were injected intraperitoneally with BrdU (Sigma, 1 mg/kg body weight), and then embryos were harvested six hours later.

Immunohistochemistry

Embryos were fixed in 4% paraformaldehyde, transferred to 30% sucrose solution, embedded in OCT compound (Tissue Tek), frozen and stored at −80^oC. Coronal cryosections of murine embryonic brain (10 μm thick) were incubated with blocking solution (5% goat serum in 0.1% Triton X-100 in PBS) and then with primary antibodies diluted in blocking solution overnight at 4°C. For all BrdU studies, cryosections were incubated in 2N HCl for 45 minutes at 37°C and washed with 0.1M boric acid prior to incubation in blocking solution. On the next day, sections were washed with 0.5% Tween 20 in PBS and incubated with secondary antibodies for 1 hour. Primary antibodies used were rat anti-Cadherin5 (BD Pharmingen, 555289), rabbit anti-NG2 (Millipore, ab5320), rabbit anti-Ki67 (Thermo Fisher Scientific, PA5-19462), rat anti-BrdU (BioRad, MCA2060), rabbit anti-ERG (Abcam, ab92513), rabbit anti-pSMAD3 (Abcam ab52903), chicken anti-GFP (Abcam, ab13970), rabbit anti-MMP9 (Abcam ab38898), rabbit anti-MMP2 (Abcam ab37150), rabbit anti-Collagen I (Abcam ab34710), rabbit anti-Collagen IV (AbD Serotec, 21501470), Dylight 594 directly conjugated Isolectin B4 (Vector Laboratories, DL-1207) and biotinylated goat anti-PDGFR-β (R&D, BAF1042). Secondary antibodies were conjugated to either Alexa 488, 555, 647 (Molecular Probes) or Dylight 555, 649 (Jackson ImmunoResearch) fluorophores. Elite ABC reagents (Vector Laboratories) and fluorescein tyramide system (PerkinElmer) were used to amplify the biotinylated PDGFR-β staining as described previously (Greif et al. 2012). DAPI (Sigma, D9542) or PI (Thermo Fisher Scientific, P3566) were used for nuclear staining.

In situ zymography

Gelatinolytic activity was assessed in unfixed cryosections (10 μm thick) of control and mutant embryos at E13.5 using DQ-gelatin, fluorescein conjugate as a substrate (Molecular Probes). The fluorescein is quenched by gelatin, and gelatin digestion yields highly fluorescent peptides. DQ-gelatin was dissolved at 10 mg/ml in water and then diluted to 1 mg/ml in PBS in the presence or absence of an inhibitor (BB94, 25 μg/ml or rTIMP3, 0.5 μg/ml) immediately prior to adding to cryosections. After 30 minutes at room temperature, slides were rinsed with PBS, fixed in 4% paraformaldehyde for 20 minutes, rinsed with 0.5% Tween 20 in PBS and then immunostained.

Ribo-probe generation and in situ hybridization

cDNA from E11.5 WT embryonic heads was used as a template to amplify Alk5 (5′-AAAGCAGTCAGCTGGCCTTG-3′ and 5′-TGCGTCCATGTCCCATTGTC-3′; 480 bp) and Timp3 (5′-CTCGGACTGTAGCATCAGCG-3′ and 5′-GGTAGCGGTAATTGAGGCCC-3′; 500 bp). PCR products were cloned into pCR II-TOPO vector using TOPO TA Cloning Kit (Invitrogen) as per manufacturer’s instructions, and positive clones were confirmed by sequencing. To generate anti-sense and sense riboprobes, pCR II-TOPO vector containing Alk5 or Timp3 inserts was linearized by digesting with XhoI (anti-sense probe) or HindIII (sense probe). Linearized pCR II-TOPO vector was used as a template for in vitro transcription (Roche) to generate DIG-labelled anti-sense and sense ribo-probes. In situ hybridizations with Alk5 and Timp3 sense and antisense probes were performed on coronal sections (10 μm) of WT embryos using standard protocols (Richard Behringer 2002). Signal was developed using NBT-BCIP (Roche) in a 5% polyvinyl alcohol (MW 146–186 kDa, Sigma) solution, as previously described (De Block and Debrouwer 1993).

Fluorescent activated cell sorting (FACS)

PCs and ECs were isolated from dissected heads of Pdgfrb-Cre, ROSA26R^(Zs/+) embryos at E13.5 also carrying Alk5^(flox/+) (control) or Alk5^(flox/flox) (mutant) (Fig. S5H–J, L). Heads were digested with freshly prepared 0.5% collagenase B (Worthington) for 30 minutes and then 0.5% collagenase 4 for 15 minutes. All digestions were incubated at 37ºC. Resulting tissue lysates were filtered through a 70 μm cell strainer (Falcon) to generate single cell suspensions and collected in ice-cold stain buffer (BD Pharmingen). The filtered cells were centrifuged for 5 minutes, and the pellet was washed twice with ice-cold PBS and resuspended in 100 μl stain buffer. This single cell suspension was incubated with Mouse Fc Block (BD Pharmigen) for 10 minutes and then with Cy5 directly conjugated NG2 antibody (Bioss) and BUV395 directly conjugated to CD31 antibody (BD Pharmigen) for 30 minutes prior to washing in stain buffer. The cell pellet was resuspended in pre-sort buffer (BD Biosciences) and sorted on a BD FACSAria II cell sorter. Zs was excited at 488 nm, and emission was collected through a 525 nm/50 bandpass filter, BUV395 was excited at 348nm and collected through 395nm/50 bandpass filter, Cy5 was excited at 640 nm and collected through a 670 nm/30 bandpass filter. Cells isolated from ROSA26R^(Zs/+), Alk5^(flox/+) (Cre negative) embryos without antibody incubation were used as a negative control. The results were analyzed using the FACSDiva and FlowJo software programs. CD31⁺ (ECs) and Zs⁺ NG2⁺ CD31⁻ (PCs) cells from controls and mutants were subjected to qRT-PCR.

EC isolation using anti-CD31-coated magnetic beads

Embryonic brains were dissected and digested as described above for FACS. The cell pellet was resuspended in cold PBS to generate a single cell suspension and ECs were isolated as previously described with minor modifications (Sheikh et al. 2015). Breifly, sheep anti-rat-IgG Dynal magnetic beads (Invitrogen) were resuspended in PBS containing 0.1% FBS and incubated with mouse anti-rat CD31 (1:250, BD Biosciences) monoclonal antibody overnight at 4°C. The single cell suspension was then incubated with these anti-CD31-coated beads for 20 minutes at room temperature and washed four times with sterile PBS. A magnet was used to separate cells bound to beads from unbound cells. In order to remove beads and exogenenous cell debri, cells were cultured in EC culture medium (DMEM containing 20% FBS supplemented with 2% endothelial growth supplement, 1% penicillin-streptomysin, 0.1% gentamycin and 0.05% heparan) for 4 weeks and passaged every week. Cells were then utilized for qRT-PCR, FACS analysis or fluorescence imaging.

Tracer injection to assess embryonic BBB formation

Five microliters of 10-KDa dextran tetramethylrhodamine, lysine fixable (Invitrogen, 5 mg/ml) was injected into the embryonic liver as described previously (Ben-Zvi et al. 2014). After 4 minutes of tracer uptake, embryos were fixed in 4% paraformaldehyde for 1 hour, washed with PBS and transferred to 30% sucrose solution overnight. The embryos were embedded in OCT compound, frozen and stored at −80ºC. Coronal cryosections of murine embryonic brain (10 μm thick) were utilized to assess tracer uptake in the GM vessels.

Quantitative real-time RT-PCR

RNA was isolated with the PureLink RNA Kit (Invitrogen) and reverse transcribed with the iScript cDNA Synthesis Kit (Biorad), and transcript levels were determined by qRT-PCR. As indicated in figure legends, normalized transcript levels are relative to the levels of either GAPDH or 18S rRNA. Forward and reverse primer pairs are listed in Table S1. For the TGFβ1 time course, RNA was isolated from human brain PCs treated with recombinant TGFβ1 (R&D, 5 ng/ml) for the indicated time.

siRNA-mediated Alk5 knockdown

Human brain PCs were transfected with Lipofectamine 2000 (Life Technologies) containing siRNA (Dharmacon) targeted against ALK5 (50 nM) or scrambled RNA for 6 hours. Cells were then washed in DMEM and cultured for 72 hours. To analyze the effect of knockdown on downstream signaling, cells were treated with recombinant TGFβ1 (5 ng/ml) for 30 minutes prior to cell lysis and Western blotting.

Migration assay

Conditioned media from PCs pretreated with ALK5 siRNA or scrambled siRNA was collected at 72 hours post-transfection and applied to human brain microvascular ECs for 48 hours. ECs were trypsinized and immediately added to the top of Boyden chamber polycarbonate membranes (Corning Costar, 5 μm pores) precoated with 0.1% gelatin. The lower compartment of the Boyden chamber contained serum-free M199 with VEGF-A (100 ng/ml). ECs were allowed to migrate for 10 hours from the upper chamber towards the VEGF-A gradient (lower compartment). The membrane was immersed in 4% paraformaldehyde for 20 minutes, and then its upper surface was scraped with a cotton swab to remove non-migrated cells. Cells on the bottom surface (i.e., migrated cells) were imaged and counted after staining with 0.1% Crystal Violet. For PC migration, following treatment with ALK5 siRNA or scrambled siRNA, human brain PCs were trypsinized, resuspended in LG DMEM with TGFβ1 (5 ng/ml) and immediately added to the top of Boyden chamber polycarbonate membranes (Corning Costar, 8 μm pores) precoated with 0.1% gelatin. The lower compartment of the Boyden chamber contained serum-free LG DMEM with TGFβ1 (5 ng/ml) and recombinant PDGF-BB (R&D, 10 ng/ml). PCs were allowed to migrate for 6 hours. Membranes with migrated PCs were fixed, stained, imaged and quantified as described above.

Human brain PC and microvascular EC co-culture assay

For co-culture studies, Matrigel (Basement Membrane Matrix, Corning) was added to A/2 96 well plate (Corning, 25 μl/well) and incubated at 37°C for 30 minutes to allow for Matrigel matrices to polymerize. Meanwhile, human brain PCs pretreated with scrambled or ALK5 siRNA and human brain microvascular ECs were trypsinized and collected. PCs and ECs were labeled with fluorescent membrane dyes PKH67 (green) and PKH26 (red), respectively, according to the protocol of the manufacturer (Sigma). Labeled cells were mixed in an equal ratio of ECs to PCs and added to Matrigel matrices (5,000 cells/well) in the absence or presence of either: i) rTIMP3 (0.5 μg/ml) or vehicle (PBS); or ii) BB94 (25 μg/ml) or vehicle (DMSO). Co-cultures were incubated at 37°C for 8 hours and fixed in 4% paraformaldehyde. For each experiment, the branch points, branch length and number of branches were assessed by manually scoring EC cords from three individual wells per treatment group.

Three-dimensional invasion assay

Conditioned media from human brain PCs pretreated with ALK5 siRNA or scrambled siRNA was collected at 72 hours post-transfection and applied to human brain microvascular ECs. Three-dimensional invasion experiments were established using collagen matrices (2.5 mg/ml) containing 1 μM sphingosine-1-phosphate (Avanti Polar Lipids), 40 ng/ml VEGF and basic fibroblast growth factor as described previously (Dave et al. 2016, Kwak et al. 2012). EC invasion cultures were incubated for 20 hours, fixed in 3% gulteraldehyde (Sigma) and stained with toluidine blue (Sigma). Invasion density was quantified manually from 6 fields per treatment as number of invading cells per field.

Cultured EC proliferation assay

Conditioned media from PCs pretreated with ALK5 siRNA or scrambled siRNA was collected at 72 hours post-transfection and applied to human brain microvascular ECs cultured on glass coverslips for 48 and 72 hours. BrdU (0.1 mg/ml) was added to the cultured ECs 6 hours prior to 48 or 72 hour timepoints. Coverslips were fixed in 4% paraformaldehyde, rinsed in Tris-glycine buffer (0.3% Tris, 1.5% glycine) for 15 minutes, and permeabilized with 0.5% Triton X-100 for 30 minutes with gentle agitation before blocking in buffer containing 0.5% Triton X-100 and 1% BSA, FBS for 1 hour. ECs were stained with rat anti-BrdU (BioRad) primary antibody for 1 hour, washed three times in 0.5% Tween 20 in PBS for 5 minutes and then incubated with secondary antibody conjugated to Alexa 488 (Molecular Probes) and DAPI for 1 hour. Coverslips were finally washed three times in 0.5% Tween 20 in PBS for 5 minutes and mounted on slides using fluorescence mounting medium (Dako). For each experiment, 18 fields were imaged for each treatment and each timepoint. Proliferating ECs were calculated as percentage ratio of BrdU⁺ ECs (BrdU⁺ DAPI⁺) to total ECs (DAPI⁺).

Cell attachment assay

Cell attachment assay was performed as described previously with minor modifications (Dave et al. 2013). Briefly, high binding EIA/RIA 96-well plates (Corning-Costar) were coated with either 0.1% gelatin, 10 μg/ml fibronectin or Matrigel (1:100 in PBS) for 1 hour, washed in PBS and blocked with 10 mg/ml BSA. PCs pretreated with scrambled or ALK5 siRNA were trypsinized and resuspended in LG DMEM with TGFβ1 (5 ng/ml). PCs were immediately added to precoated wells (20,000 cells/well) and incubated at 37°C for 1 hour. Plates were washed in saline to remove unbound cells and fixed in 3% formalin for 3 hours before staining with 0.1% Amido Black in 10% acetic acid and 30% methanol for 15 minutes. Plates were washed with saline and dried before adding 50 μl of 2N NaOH/well. Raw absorbance (595 nm) was measured using a Synergy 2 plate reader (BioTek). Higher absorbace (595 nm) corresponds to increased cell attachment.

Western blot

Total cell lysates were prepared by solubilizing cells in boiling 1.5X Laemmli sample buffer at 95°C for 10 minutes. Protein samples were resolved by 7–15% SDS-PAGE, transferred to Immobilon PVDF membranes (Millipore), blocked with 5% nonfat dry milk or BSA, washed in tris-buffered saline with 0.1% Tween 20 (TBS-T) and probed with primary antibodies overnight at 4 °C. Membranes were incubated with HRP-conjugated secondary antibodies (Dako), washed in TBS-T, developed with Supersignal West Pico Maximum Sensitivity Substrate (Pierce) and analyzed with the G:BOX imaging system (Syngene). Primary antibodies used for Western blot analysis were rabbit anti-ALK5 (Abcam, ab31013), rabbit anti-phospho-SMAD2 (Cell Signaling, 3108), rabbit anti-phospho-SMAD3 (Abcam, ab52903), rabbit anti-SMAD2/3 (Cell Signaling, 3102), rabbit anti-GAPDH (Cell Signaling, 2118) and mouse anti-TIMP3 (Millipore, MAB3318).

Imaging

Fluorescent images of coronal brain sections and cultured cells were acquired on a confocal microscope (PerkinElmer UltraView Vox Spinning Disc). Brightfield images of H&E staining, Boyden chamber migration assays and Matrigel co-culture assays were captured using inverted microscopes (Eclipse 80i and Eclipse TS100, Nikon). Volocity software (PerkinElmer) and Adobe Photoshop were used to process images.

Quantification and statistical analysis

Quantification of PC coverage of EC vessels and EC vessel morphology

Confocal images of vessels in the embryonic GM were stained for CDH5 and NG2. All quantifications were done using ImageJ software. PC coverage of EC vessels was quantified and expressed as percentage ratio of CDH5⁺ EC vessel length covered by NG2⁺ PCs to total CDH5⁺ EC vessel length. Vessel area was quantified as the total area of CDH5⁺ EC vessels, lumen area was calculated as the area of cross-section of CDH5⁺ EC vessels and vessel diameter was quantified as the thickness CDH5⁺ EC vessels. Vessel area, lumen area and vessel diameter were normalized to controls.

Quantification of collagen coverage of PC sleeves and MMP staining

Confocal images of GM vessels of Pdgfrb-Cre, Alk5^(flox/+) controls and Pdgfrb-Cre, Alk5^(flox/flox) mutants also carrying ROSAR26^(mTmG/+) were stained for GFP and either Col-I, Col-IV, MMP2 or MMP9. All quantifications were done using ImageJ software. Collagen coverage of GFP⁺ PC sleeves was quantified and expressed as percentage ratio of GFP⁺ PC sleeve length covered by Col-I or Col-IV to total GFP⁺ PC sleeve length. MMP staining was quantified as the intensity of MMP2 or MMP9 staining around the GFP⁺ PC sleeves and normalized to controls.

Quantification of EC and PC proliferation

Proliferation in the GM vasculature was analyzed through confocal images of immunostained coronal cryosections. In EC proliferation studies, the total number of ECs was determined by counting ERG⁺ nuclei of IB4⁺ cells, and BrdU⁺ERG⁺ nuclei of IB4⁺ cells were scored as proliferating. The percentage of ECs that were proliferative is reported. Alternatively, in Ki67 experiments, EC proliferation was quantified as the percent of DAPI⁺ nuclei of IB4⁺ cells that were also Ki67⁺. For PC proliferation studies, as there is species overlap between PC and proliferation marker antibodies, we used Pdgfrb-Cre, ROSA26R^(mTmG/+) mice that were also either Alk5^(flox/+) (controls) or Alk5^(flox/flox) (mutants). We focused on ERG− nuclei of GFP⁺ cells to exclude ECs and determined the percent of total PCs (DAPI⁺ ERG− nuclei in GFP⁺ cells) that were also BrdU⁺.

Statistics

Student’s t test and multifactor ANOVA with Tukey’s and Sidak’s post hoc tests were used to analyze the data (Graphpad Prism version 6.03 software). Statistical significance threshold was set at p≤0.05. All data are presented as means ± SD.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
rat anti-Cadherin5	BD Biosciences	Cat# 555289; RRID:AB_395707
rabbit anti-NG2	Millipore	Cat# ab5320; RRID:AB_91789
rabbit anti-Ki67	Thermo Fisher Scientific	Cat# PA5-19462; RRID:AB_10981523
rat anti-BrdU	BioRad	Cat# MCA2060; RRID:AB_323427
rabbit anti-ERG	Abcam	Cat# ab92513; RRID:AB_2630401
rabbit anti-pSMAD3	Abcam	Cat# ab52903; RRID:AB_882596
chicken anti-GFP	Abcam	Cat# ab13970; RRID:AB_300798
rabbit anti-MMP9	Abcam	Cat# ab38898; RRID:AB_776512
rabbit anti-MMP2	Abcam	Cat# ab37150; RRID:AB_881512
rabbit anti-Collagen I	Abcam	Cat# ab34710; RRID:AB_731684
rabbit anti-Collagen IV	AbD Serotec	Cat# 21501470; RRID:AB_2082660
Dylight 594 conjugated Isolectin B4	Vector Laboratories	Cat# DL-1207; RRID:AB_2336415
Alexa 647 conjugated Isolectin B4	Invitrogen	Cat# I32450
biotinylated goat anti-PDGFR-β	R&D Systems	Cat# BAF1042; RRID:AB_2162632
anti-rat Alexa 488	Invitrogen	Cat# A11006; RRID:AB_141373
anti-rat Alexa 546	Invitrogen	Cat# A11081; RRID:AB_141738
anti-rat Alexa 647	Invitrogen	Cat# A21247; RRID:AB_141778
anti-rabbit Dylight 488	Vector Labs	Cat# DI-1488; RRID:AB_2336402
anti-rabbit Dylight 555	Vector Labs	Cat# DI-1549; RRID:AB_2336407
anti-rabbit Dylight 649	Jackson ImmunoResearch	Cat# 211-492-171; RRID:AB_2339164
anti-chicken Alexa-488	Abcam	Cat# ab150169
DAPI	Sigma	Cat# D9542; RRID:AB_2636803
Propidium Iodide (PI)	Thermo Fisher Scientific	Cat# P3566
anti-DIG conjugated alkaline phosphatase	Roche	Cat# 11093274910; RRID:AB_514497
Mouse Fc Block	BD Pharmingen	Cat# 553141; RRID:AB_394656
Cy5 conjugated NG2	Bioss	Cat# BS-11192R-Cy5
BUV395 conjugated CD31	BD Biosciences	Cat# 740231
rabbit anti-ALK5	Abcam	Cat# ab31013; RRID:AB_778352
rabbit anti-phospho-SMAD2	Cell Signaling	Cat# 3108; RRID:AB_490941
rabbit anti-SMAD2/3	Cell Signaling	Cat# 3102; RRID:AB_10698742
rabbit anti-GAPDH	Cell Signaling	Cat# 2118; RRID:AB_561053
mouse anti-TIMP3	Millipore	Cat# MAB3318; RRID:AB_94813
HRP-conjugated secondary antibodies	Dako	Cat# P0448; RRID:AB_2617138 Cat# P0447; RRID:AB_2617137
Bacterial and Virus Strains
One Shot DH5α-T1 competent cells	Thermo Fisher Scientific	Cat# K4520-01
Chemicals, Peptides, and Recombinant Proteins
rTIMP3	R&D System	Cat# 973-TM
Batimastat	Sigma	Cat# SML0041
Tamoxifen	Sigma	Cat# T5648
10-KDa dextran tetramethylrhodamine	Invitrogen	Cat# D1817
BrdU	Sigma	Cat# B5002
TGFβ1	R&D Systems	Cat# 240-B-002
VEGF	R&D Systems	Cat# 293-VE-050
bFGF	Peprotech	Cat# 100-18B
PDGF-BB	Sigma	Cat# P4056
S1P	Avanti Polar Lipids	Cat# 860492
Lipofectamine 2000	Invitrogen	Cat# 1815581
Matrigel	Corning	Cat# 354234
Collagen Type 1	Corning	Cat# 354249
Collagenase B	Worthington	Cat# LS004147
Collagenase 4	Worthington	Cat# LS004188
DQ-gelatin	Invitrogen	Cat# D12054
Critical Commercial Assays
Vectastain Elite ABC kit	Vector Laboratories	Cat# PK-6100
TSA Plus Fluorescein	PerkinElmer	Cat# NEL701A001KT
TOPO TA Cloning kit	Thermo Fisher Scientific	Cat# K457501
GoTaq Green mastermix	Promega	Cat# M7122
anti-rat-IgG Dynal magnetic beads	Invitrogen	Cat# 11035
Purelink RNA kit	Invitrogen	Cat# 12183018A
iScript cDNA Synthesis kit	BioRad	Cat# 170-8891
Sso Fast Evagreen mastermix	BioRad	Cat# 172-5201
Experimental Models: Cell Lines
Human brain vascular pericytes	ScienCell	Cat# 1200
Human brain microvascular endothelial cells	ScienCell	Cat# 1000
Experimental Models: Organisms/Strains
C57BL/6 wild type	The Jackson Laboratory	JAX: 000664; RRID:IMSR_JAX:000664
ROSA26R(mTmG/mTmG)	The Jackson Laboratory	JAX: 007576; RRID:IMSR_JAX:007576
ROSA26R(Zs/Zs)	The Jackson Laboratory	JAX: 007906; RRID:IMSR_JAX:007906
Alk5(flox/flox)	Stefan Karlsson	MGI:2680164 (Larsson et al. 2001)
Pdgfrb-Cre	Ralf Adams	MGI:3763193 (Foo et al. 2006)
Acta2-CreER	Pierre Chambon	MGI:5305721 (Wendling et al. 2009)
Oligonucleotides
ON-TARGET plus human TGFBR1 siRNA SMARTpool	Dharmacon	Cat# L-003929-00-0005
ON-TARGET plus non-targeting pool	Dharmacon	Cat# D-001810-10-05
In situ hybridization Alk5: F- aaagcagtcagctggccttg, R- tgcgtccatgtcccattgtc	This paper	N/A
In situ hybridization Timp3: F- ctcggactgtagcatcagcg, R- ggtagcggtaattgaggccc	This paper	N/A
Primers for qRT-PCR, see Table S1	This paper	N/A

Highlights.

• Pdgfrb-Cre, Alk5^(flox/flox) mutant embryos develop germinal matrix hemorrhage (GMH)

• Alk5-depleted pericytes induce deleterious effects on endothelial morphogenesis

• ALK5 signaling in pericytes regulates TIMP3 expression

• TIMP3 treatment markedly attenuates GMH in Pdgfrb-Cre, Alk5^(flox/flox) mutants