Vascular abnormalities in the eye are the leading cause of many forms of inherited and acquired human blindness. Loss-of-function mutations in the Wnt-binding co-receptor LRP5 leads to aberrant ocular vascularization and loss of vision in genetic disorders such as osteoporosis-pseudoglioma syndrome. The canonical Wnt-β-catenin pathway is known to regulate retinal vascular development. However, it is unclear what precise role LPR5 plays in this process. Here, we show that loss of LRP5 function in mice causes retinal hypovascularization during development as well as retinal neovascularization in adulthood with disorganized and leaky vessels. Using a highly specific Flk1-CreBreier line for vascular endothelial cells, together with several genetic models, we demonstrate that loss of endothelium-derived LRP5 recapitulates the retinal vascular defects in Lrp5-/- mice. In addition, restoring LRP5 function only in endothelial cells in Lrp5-/- mice rescues their retinal vascular abnormalities. Furthermore, we show that retinal vascularization is regulated by LRP5 in a dosage dependent manner and does not depend on LRP6. Our study provides the first direct evidence that endothelium-derived LRP5 is both necessary and sufficient to mediate its critical role in the development and maintenance of retinal vasculature.
Citation: Huang W, Li Q, Amiry-Moghaddam M, Hokama M, Sardi SH, Nagao M, et al. (2016) Critical Endothelial Regulation by LRP5 during Retinal Vascular Development. PLoS ONE 11(3): e0152833. https://doi.org/10.1371/journal.pone.0152833
Editor: Aernout Luttun, Katholieke Universiteit Leuven, BELGIUM
Received: February 10, 2016; Accepted: March 2, 2016; Published: March 31, 2016
Copyright: © 2016 Huang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: This work was supported by the National Institutes of Health-National Institute of Arthritis and Musculoskeletal and Skin Diseases [http://nih.gov/, http://www.niams.nih.gov/] grant numbers AR36819 and AR36820 to BO. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Vision impairment and blindness are devastating conditions afflicting over 4% of the world population . In developed countries, vascular abnormalities are the major cause of many forms of inherited and acquired human blindness, such as Osteoporosis-Pseudoglioma Syndrome (OPPG), Norrie Disease (ND), Familial Exudative Vitreoretinopathy (FEVR) and diabetic retinopathy (DR) [2,3]. Both aberrant vascular development and pathological neovascularization can critically impair the high metabolic activities in the retina. The retinal vasculature consists of three vessel beds located in the nerve fiber layer (NFL), inner plexiform layer (IPL) and outer plexiform layer (OPL). Its heavy reliance on a well-timed and balanced orchestration of many factors involving different cell types, multiple signaling inputs and proper oxygen levels makes it susceptible to anomalies that are difficult to study . However, some of these blinding conditions have overlapping genetic causes and/or ocular manifestations, indicating that they likely have shared pathological mechanisms. Therefore, studies of human genetic ocular disorders have provided insights into biological and pathological processes that also underlie acquired diseases. Here, in the context of OPPG, we present data on the critical role of low-density lipoprotein receptor-related protein-5 (LRP5) during retinal vascular development.
OPPG is a rare autosomal recessive disorder characterized by severe childhood osteopenia and congenital or infancy-onset visual loss [5–7]. Major manifestations in the eye include retinal hypovascularization, retrolental fibrovascular tissue (pseudoglioma), microphthalmia and various vitreoretinal abnormalities. The disorder is caused by loss-of-function mutations in LRP5, a co-receptor in the canonical Wnt signaling pathway. Many of the ocular findings in OPPG patients overlap with those of FEVR and ND, caused by loss-of-function mutations in other Wnt signaling components, such as Frizzled-4 (FZD4) and Norrie disease protein (NDP) [8–12].
Seminal studies by the Nathans group and others have shown that Müller glial cells secrete Norrin that binds to FZD4 in endothelial cells (ECs) and regulates retinal vascular development through the canonical Wnt-β-catenin pathway [13–16]. Disruption of this pathway through loss of Norrin, FZD4 or LRP5 function not only leads to an overlapping spectrum of ocular problems in patients, but also results in similar retinal vascular defects in mice. Mice in which Fzd4 is conditionally knocked out (CKO mice) by using Tie2‐Cre (Tie2‐Cre;Fzd4fl/-) exhibit the retinal hypovascularization phenotype seen in Fzd4 null (Fzd4-/-) and Ndp null (Ndp-) mice, and inducing β-catenin function in ECs rescues these defects in Ndp- mice [14,17]. Based on these data, it has been proposed that the pathway functions in ECs to control retinal vascularization. However, cells that express Tie2-Cre include not only ECs but also several other cell types , indicating a possible contribution of non-EC-derived FZD4 to retinal vascular regulation. Furthermore, inducing β-catenin activity in ECs may bypass the need for Norrin-FZD4-β-catenin signaling in non-ECs. In addition, although activation of the Norrin-FZD4-β-catenin pathway requires the presence of either LRP5 or LRP6 in vitro , it is unclear what exact roles LRP5 and LRP6 play during retinal vascular development in vivo, and what cell types contribute to the incomplete vascularization in the retina of OPPG patients and Lrp5-/- mice.
In this study, we use multiple genetic animal models to address these questions. Our use of a highly endothelial-specific VEGF receptor 2-Cre line (Flk1-CreBreier) makes it possible, for the first time, to evaluate the critical function of LRP5 in ECs during retinal vascular development.
An Essential Role of LRP5 in Retinal Vascular Development
To examine the role of LRP5 in retinal vascular development, we first analyzed the retinal vasculature in Lrp5-/- mice at different developmental stages in detail. Retinal whole mount immunofluorescence (IF) staining of collagen IV (ColIV) showed that Lrp5-/- retinas exhibited retarded endothelial outgrowth with sparse vessel coverage in the NFL during early postnatal development (Fig 1A). In adult Lrp5-/- mice, retinal angiography with FITC-dextran perfusion showed that the intraretinal vessel layers in the IPL and OPL were mostly absent, with occasional incomplete vascular development in the IPL, and vessels penetrating from the NFL terminated in clusters without branching (Fig 1B and S1 Movie and S2 Movie). Intravitreal hemorrhage was frequently observed and hyaloid vessels persisted into adulthood, long after the time when they normally regressed in WT mice (Fig 1B). These retinal hypovascularization defects in Lrp5-/- mice confirm findings in previous studies [14,19]. In addition, the NFL exhibited chaotic vessel overgrowth (neovascularization) with arterio-venous anastomoses, microaneurysms, convoluted neovascular tufts, and leaky vessels (Fig 1B and 1C). Electron microscopy (EM) showed widespread endothelial fenestrations (EF) in adult Lrp5-/- retinas but not during early postnatal development (Fig 1D), indicating that EF is unlikely a primary defect caused by loss of Lrp5. As increased levels of VEGF can often lead to endothelial fenestration , we examined retinal VEGF amounts in the mice through ELISA assays. As shown in Fig 1E, total retinal VEGF levels in Lrp5-/- mice were only slightly increased, about 1.3-fold, at P5 and P8 compared to controls, but were 6.6-fold higher in adults. This suggests that elevated retinal VEGF is likely a secondary response to the retinal hypovascularization defect in Lrp5-/- mice during postnatal development and a contributor to EF defects when it reaches a high level in adulthood.
(A) ColIV whole mount IF staining showing retinal vessels of Lrp5-/- and control mice at P9 and P30. Quantification of vascular sprout numbers at P5 shown at right. (B) Adult Lrp5-/- retinas showing persistent hyaloid vessels (black arrows, CD31 IHC staining), aneurysms (open arrow, CD31 IHC staining), neovascular overgrowth (white arrows, fibronectin IF staining) and lack of IPL and OPL vascular development (lower panels, green: FITC-Dextran perfusion). (C) FITC-Dextran perfusion (green) showing retinal vascular leakage (white arrows) in adult Lrp5-/- mice compared to control. Scale bars = 100μm. (D) EM analysis of endothelium of Lrp5-/- and control retinas at P5, P8 and P30. Arrows point to area of endothelial fenestration. Scale bars = 500nm. (E) Total amounts of VEGF protein in retinas of Lrp5-/- and control mice at P5, P8 and P30. Each ELISA was done in duplicate and normalized to total retinal protein amount. n = 9, 5, 12 for controls (at P5, P8, P30) and 9, 8, 8 for Lrp5-/- (at P5, P8, P30). * p<0.05, ** p<0.01. Data are represented as means ± SD. Ctrl, control.
LRP5 Signaling in Tie2+ Cells but Not VE-Cadherin+ Cells is Essential for Retinal Vascular Development
In the retina, Lrp5 is expressed predominantly in Müller glia and in ECs [19,21]. To identify the primary cell population requiring Lrp5 expression for retinal vascularization, we used mice with Lrp5 floxed alleles  to conditionally knock out Lrp5 in retinal neural/glial cells using Rx-Cre  and in ECs using VE-cadherin-Cre (VE-Cad-Cre)  or Tie2-Cre mice . Loss of Lrp5 in retinal neural/glial cells had no impact on retinal vessels (S3 Movie). Surprisingly, while VE-Cad-Cre;Lrp5fl/- CKO mice (harboring one floxed and one null Lrp5 allele) also demonstrated a completely normal retinal vasculature (Fig 2A), Tie2-Cre;Lrp5fl/fl mice exhibited vascular defects that were almost identical to those of Lrp5-/- mice. Briefly, the hyaloid vessels failed to regress, intraretinal vascular beds were absent in adult mice, vessels penetrating from the NFL terminated in club-like clusters and the NFL exhibited chaotic neovascularization (Fig 2B).
(A and B) ColIV IF staining (red) and FITC-Dextran perfusion (green) showing adult retinal vasculature in VE-Cad-Cre;Lrp5fl/- and Tie2-Cre;Lrp5fl/fl CKO mice (4w) compared to Lrp5fl/fl control. Arrows: hyaloid vessels; open arrows: neovascular tufts; arrowheads: neovascular overgrowth. At right: Quantification of vascular branch points in OPL at 4w; ns not significant. (C) TdTomato signals in VE-Cad-Cre;tdTomato (8w) and Tie2-Cre;tdTomato (7w) mice showing predominant endothelial expression of VE-Cad-Cre and Tie2-Cre in the NFL, IPL and OPL of the retina. Note that VE-Cad-Cre is also widely expressed in myeloid cells around the vessels in the NFL (upper left panel, also see D). Arrows point to IB4 (green) positive vascular area with negative VE-Cad-Cre;tdTomato signals in a P5 mouse indicating incomplete VE-Cad-Cre recombination. (D) Myeloid and microglial localization of VE-Cad-Cre;tdTomato and Tie2-Cre;tdTomato signals in P5 (left and mid-left panels) and adult (right and mid-right panels) retinas. Arrows point to myeloid cells and arrowheads point to microglial cell. Green: F4/80 IF staining. Scale bars = 100nm.
As both VE-Cad-Cre and Tie2-Cre transgenes are predominantly expressed in ECs [24,25] and are widely used in EC-related genetic studies, this large discrepancy of phenotypes in VE-Cad-Cre;Lrp5fl/- and Tie2-Cre;Lrp5fl/fl CKO mice casts doubt on the conclusion that EC-dependent LRP5 signaling is critical for retinal vascular development. First, VE-Cad-Cre+ and Tie2-Cre+ cells are known to give rise to half and most adult hematopoietic cells, respectively, in addition to ECs [18,24,26]. Therefore, we considered the possibility that the essential function of LRP5 may be associated with Tie2-Cre+ hematopoietic cells rather than the Tie2-Cre+ ECs. Secondly, although we increased the Lrp5 knockout efficiency in VE-Cad-Cre;Lrp5fl/- CKO mice by incorporating one Lrp5 null allele, we also considered the possibility that VE-Cad-Cre may have a much lower recombination efficiency in retinal ECs than Tie2-Cre, and that incomplete endothelial deletion of Lrp5 in VE-Cad-Cre;Lrp5fl/- mice may explain the lack of a vascular defect. To distinguish between these possibilities, we further investigated the expression of VE-Cad-Cre and Tie2-Cre in the eye.
Endothelial and Non-endothelial Expression of VE-Cadherin-Cre and Tie2-Cre in the Retina
Using a tdTomato reporter line , we generated VE-Cad-Cre;tdTomato and Tie2-Cre;tdTomato double transgenic mice. Whole-mount retinal microscopy showed that both VE-Cad-Cre and Tie2-Cre were predominantly expressed in retinal ECs in all three vascular beds (Fig 2C). However, vascular patches with negative VE-Cad-Cre;tdTomato signals could often be observed in the developing retina (Fig 2C), indicating incomplete recombination with VE-Cad-Cre in endothelium. Tie2-Cre+ and VE-Cad-Cre+ signals were also widely present in microglia and macrophage/myeloid cells (Fig 2D), but their distribution in these cells differed developmentally.
In the first postnatal week, compared to VE-Cad-Cre, the Tie2-Cre;tdTomato signal was seen in much more microglial cells both ahead of the vascular leading edge and around the developing vessels (Fig 2D). Whole mount IF staining with F4/80, a macrophage marker, showed that almost all F4/80+ macrophage/myeloid cells in the NFL had a Tie2-Cre+ origin while only about half were VE-Cad-Cre;tdTomato+ (Fig 2D). In adult mice, a very large number of myeloid cells around the NFL vessels showed VE-Cad-Cre;tdTomato signals with only a few showing signs of being derived from Tie2+ cells (Fig 2D). However, intraretinal microglial cells still presented strong tdTomato signals from a Tie2+ but not VE-Cad+ origin (Fig 2D).
These data indicate that compared to Tie2-Cre, VE-Cad-Cre is less efficient in recombining targets in retinal ECs, and this may explain the lack of any vascular defects in VE-Cad-Cre;Lrp5fl/- CKO mice. At the same time, Tie2+ cells also give rise to myeloid and microglial cells in the retina, thus not excluding a possible contribution of these cells to the vascular phenotype in Tie2-Cre;Lrp5fl/fl CKO mice.
LRP5 Signaling in Endothelial but Not Myeloid Cells is Essential for Retinal Vascular Development
To further examine the potential contributions of Tie2+ myeloid/microglial cells to the vascular phenotype in Tie2-Cre;Lrp5fl/fl CKO mice, we used two in vivo strategies. First, we conditionally deleted Lrp5 in myeloid/glial cells with M lysozyme-Cre (LysM-Cre)  and CD11b-Cre . Both LysM-Cre and CD11b-Cre have been shown to have a high degree of recombination in bone marrow macrophages and granulocytes as well as peritoneal macrophages [28,29]. While CD11b-Cre also demonstrates recombination activity in brain microglial cells and LysM-Cre is less effective in tissue macrophages, the expression of these Cre’s in the retina is unclear. We generated double transgenic mice of tdTomato with these two Cre lines and showed that LysM-Cre was primarily expressed in myeloid and microglial cells in developing retinas and mostly in myeloid cells around adult NFL (Fig 3A and 3B). CD11b-Cre was abundantly expressed in microglia and Müller glia both in developing and mature retinas (Fig 3D and 3E). Nonetheless, both LysM-Cre;Lrp5fl/- and CD11b-Cre;Lrp5fl/- CKO mice developed a three-tier retinal vasculature similar to controls (Fig 3C and 3F), indicating that myeloid/microglial LRP5 is dispensable for retinal vascularization.
(A) Whole mount IF staining with IB4 (blue) or ColIV antibody (right panel, green) showing distribution of LysM-Cre;tdTomato+ (red) myeloid and microglial cells in retinas at P5, P8 and 6w. Most LysM-Cre;tdTomato+ myeloid cells were also F4/80+ (green) (white rectangle in left panel indicates area shown at higher magnification in the upper right corner inset). (B) Localization of LysM-Cre;tdTomato+ myeloid cells in three adult (6w) retinal vascular beds. Green: ColIV. (C) Retinal vasculature of LysM-Cre;Lrp5fl/- CKO mice (12w). Red: IB4. (D) Distribution of CD11b-Cre;tdTomato+ (red) myeloid and microglial cells in retinas at P6 (left and mid panels, green: IB4) and 4w (right panel, cross section). Blue: Hoechst. (E) Localization of CD11b-Cre;tdTomato+ signals (red) in myeloid, microglial, Müller glial and perivascular macrophage cells in adult (4w) retina. Blue: IB4. White rectangle in right panel outlines cell shown at higher magnification in upper right corner inset. (F) Retinal vasculature of CD11b-Cre;Lrp5fl/- CKO mice (4w). Green: FITC-Dextran perfusion. Scale bars = 100nm.
Next, we generated another endothelial Lrp5 CKO with VEGF receptor 2-driven Cre (Flk1-CreBreier) transgenic mice . Unlike Tie2-Cre and VE-Cad-Cre, Flk1-CreBreier;tdTomato mice showed that Flk1-CreBreier was specifically expressed in ECs with few, if any, signs of myeloid expression in the retina as well as in bone marrow, bone and skeletal muscle (Fig 4A and 4B). This remarkable endothelial specificity is likely due to the limited set of Flk1 regulatory promoter regions used to generate the Flk1-CreBreier line, although Flk1 expressing progenitor cells give rise to endothelial, hematopoietic and muscle lineages . In Flk1-CreBreier;Lrp5fl/- CKO mice 43.8% of the retinal area exhibited similar vascular abnormalities as Tie2-Cre;Lrp5fl/fl CKO mice, including absent intraretinal vessel layers, persistent hyaloid vessels and neovascularization in the NFL (Fig 4C and S4 Movie). In the retinal OPL of these CKO mice, areas with normal capillaries were often close to areas lacking any vascular development (Fig 4C). This suggests that horizontal endothelial growth in the OPL alone is insufficient to form and extend a large capillary network to areas where no vertical branches developed from the NFL. In addition, when one tdTomato allele was incorporated in these CKO mice, the vessels in areas with vascular abnormalities were always tdTomato+ while normal areas often included many tdTomato negative vessel branches (Fig 4D). This indicates that LRP5 functions autonomously in retinal ECs. Collectively, our data provide the first direct evidence that ECs are the primary cells that cause retinal hypovascularization during development and chaotic neovascularization in adulthood when LRP5 signaling is lost. In addition, based on the unique endothelial specificity of Flk1-CreBreier, combined with other studies showing endothelial requirement for Norrin-FZD4-β-catenin signaling in retinal vascularization, our data further suggest that the Norrin-induced canonical Wnt pathway functions through endothelial LRP5.
(A) Specific endothelial location of Flk1-CreBreier;tdTomato signals in developing retinas (P5, upper panels) overlapping with IB4 IF signals (blue). F4/80 IF staining (green) showing that most macrophages were tdTomato negative. In adult retina, Flk1-CreBreier was specifically expressed in all three layers of the retinal vascular beds (lower panels) with no signs of any myeloid cell expression. Penetrance of Flk1-CreBreier expression in ECs could also be incomplete as shown in (D). (B) Specific endothelial location of Flk1-CreBreier;tdTomato signals in adult (4w) bone marrow, cortical bone and skeletal muscle. M: muscle; CB: cortical bone; BM: bone marrow. Blue: Hoechst. (C) Whole mount IF staining of ColIV (red) and IB4 (green) showing retinal vasculature in adult (4w) Flk1-CreBreier;Lrp5fl/- CKO mice with disorganized NFL vessels, vertical vessel branches terminating in ball-like structures in the IPL and lack of OPL vascular bed. Arrow points to persistent hyaloid vessels. Note that regions with vascular abnormalities often had patchy normal-looking areas located in the neighborhood. In the selected OPL image, an area with well-developed vessels is next to another with no vascular development. (D) Whole mount IF staining of IB4 (green) showing that normally developed vascular areas in Flk1-Cre;Lrp5fl/-;tdTomato retinas (8w) included many Flk1-Cre;tdTomato negative vessels (arrows), whereas abnormal vessel structures were all tdTomato positive (open arrows). Scale bars = 100μm.
Conditional Restoration of LRP5 Signaling in Endothelial Cells Rescues the Vascular Defects in Lrp5-/- Mice
If loss of LRP5 in ECs alone is sufficient to cause the vascular defects observed in Lrp5-/- mice, re-expressing LRP5 only in ECs in the Lrp5-/- background should rescue the defective retinal vasculature in these mice. To test this hypothesis, we took advantage of an Lrp5 knock-in hypomorph allele, Lrp5 a214v(n), where Lrp5 is transcribed at a very low level . Following Cre-mediated removal of a neo cassette, this allele is converted into a high bone mass-causing allele (Lrp5 a214v) that has been shown to be comparable to WT LRP5 in its ability to transduce canonical Wnt signaling . Retinal angiography with FITC-dextran perfusion showed that the Lrp5 hypomorphic mice displayed retinal vascular defects similar to those of Lrp5-/- mice. Adult Lrp5a214v(neo)/- mice had retinas lacking OPL vascular development with their NFL and IPL vessels varying from normal to defective as in Lrp5-/- mice, albeit to a lesser degree (Fig 5A and S5 Movie). Neovascularization in the NFL was also less prominent in Lrp5a214v(n)/- than in Lrp5-/- mice. Conditional restoration of LRP5 expression in ECs using Flk1-CreBreier or VE-Cad-Cre (Fig 5B and S6 Movie) in Lrp5a214v(n)/- mice, restored normal retinal vascular development, while conditional restoration of Lrp5 in myeloid cells with LyzM-Cre had no such effect (Fig 5B). These data strongly indicate that EC-derived LRP5 alone is sufficient for its control in normal retinal vascular development.
(A) FITC-Dextran perfusion showing distorted NFL and IPL vessels and lack of OPL vascular development in Lrp5a214v(n)/- hypomorph mice (lower panels) compared to controls (upper panels) both in retinal whole mount (4w) and cross section (2m) images. (B) VE-Cad-Cre;Lrp5a214v(n)/+ (upper left panels, FITC-Dextran perfusion), VE-Cad-Cre;Lrp5a214v(n)/- (lower left panels, FITC-Dextran perfusion) and Flk1-Cre;Lrp5a214v(n)/- (upper right panels, red: ColIV, green: IB4) mice all developed a normalized three-tier retinal vascular structure, while LysM-Cre;Lrp5a214v(n)/- (lower right panels, red: ColIV) mice displayed similar retinal vascular abnormalities compared to control Lrp5a214v(n)/- mice (A, lower panels). Quantification of vascular branch points in OPL is shown in graph at right; **P<0.01, ns not significant. All mice are between 4 to 5 weeks of age except otherwise labeled. Scale bars = 100nm.
Endothelial Cell-Derived LRP5 Does Not Share Redundant Roles with LRP6 in Its Regulation of Retinal Vascular Development
LRP5 and LRP6 often serve as interchangeable co-receptors for the canonical Wnt pathway and share redundant roles in many developmental or pathological processes . Activation of Norrin-FZD4-β-catenin signaling requires the presence of either LRP5 or LRP6 in vitro . Endothelial deletion of Lrp6 in the background of Lrp5-/- mice resulted in severely attenuated brain vascularization and bleeding, but had little impact on the retinal vascular defects seen in Lrp5-/- mice . To further examine whether LRP5 and LRP6 have similar redundant functions in ECs during retinal vascular development, we deleted different copies of Lrp5 and Lrp6 with the most robust endothelial Cre line–Tie2-Cre–among the three that were used in this study. As shown in Fig 6A and 6B, when a single copy of Lrp5 is deleted in ECs, additional removal of either one or both copies of Lrp6 had no impact on retinal vascular development. Similarly, when both copies of Lrp5 were deleted in ECs, adding a deletion of one Lrp6 allele had no impact on the observed vascular defects in Tie2-Cre;Lrp5fl/fl CKO mice (Fig 6C and 6D). Our data are consistent with the Zhou et al. study  and further suggest that while LRP5 is essential for retinal vascularization, LRP6 has a dispensable role in this process in vivo. This further indicates that the Norrin-FZD4-β-catenin pathway regulates retinal vascular development in vivo exclusively through LRP5 and not LRP6.
Whole mount IF staining of ColIV (red) and FITC-Dextran perfusion showing (A) Retinal vasculature of Tie2-Cre;Lrp5fl/+;Lrp6fl/+ CKO mice. (B) Retinal vasculature of Tie2-Cre;Lrp5fl/+;Lrp6fl/fl CKO mice. (C) Retinal vasculature of Tie2-Cre;Lrp5fl/fl CKO mice. (D) Retinal vasculature of Tie2-Cre;Lrp5fl/fl;Lrp6fl/+ CKO mice. All mice are 8 weeks of age. Scale bars = 100nm.
In the current study, we report that loss of LRP5 causes a profound change in the retinal vasculature including both hypovascularization and neovascularization. Through the use of multiple genetic models we demonstrate that specific endothelial deletion of Lrp5 recapitulates the retinal vascular defects observed in Lrp5-/- mice, while specific endothelial restoration of LRP5 in Lrp5-/- mice rescues those defects. This suggests that LRP5 regulates an essential retinal vascularization program solely through ECs. In addition, LRP5 likely exerts a dosage dependent effect as the severity of the retinal vascular defects in Lrp5 hypomorphic mice was between that of WT and Lrp5-/- mice. Finally, LRP6 is completely dispensable in this process.
The overlapping ocular manifestations and genetic causes in OPPG, FEVR and ND patients strongly suggest a common genetic pathway involved in the pathogenesis of the disorders. Our findings that the retinal vascular defects in Lrp5-/- mice also highly resemble those in Ndp- and Fzd4-/- mice [13,14,16] further confirm the critical regulation by each component of the ligand (Norrin)—receptor (FZD4)—co-receptor (LRP5) trio in the canonical Wnt pathway. Although Norrin-FZD4 has been shown to control retinal vascularization in ECs through β-catenin, how exactly LRP5 functions in this process remained to be determined. In addition, the use of Tie2-Cre, a Cre line that shows robust recombination in both endothelial and hematopoietic cells, in conditional knockout studies of Fzd4 has left open the possibility of non-EC contributions to the essential regulation of Norrin-FZD4-mediated retinal vascular development. In the present study, we exclude any essential contribution of retinal myeloid/microglial cells to LRP5-mediated retinal vascularization. Using the highly endothelial specific Flk1-CreBreier line and Flk1-CreBreier;Lrp5fl/- and Flk1-CreBreier;Lrp5a214v(n)/- mice, we directly demonstrate an endothelial autonomous function of LRP5, as well as the Norrin-FZD4-LRP5-β-catenin pathway, during retinal vascular development.
Interestingly, although both LRP5 and LRP6 function as interchangeable essential co-receptors for activation of β-catenin signaling by Norrin-FZD4 in vitro  and have redundant functions in canonical Wnt signaling-mediated brain vascular formation in vivo , our study shows that LRP5, but not LRP6, is the critical co-receptor in Norrin-FZD4-β-catenin-mediated retinal vascularization, consistent with a previous report . The differential requirement for LRP5 and LRP6 in the retina and brain could reflect different availability of the Wnt ligands in the two tissues during vascular development or different adaptation to additional regulation from other local factors, such as macrophages in the eye and oxygen levels in the brain.
The lack of a phenotype in VE-Cad-Cre;Lrp5fl/- CKO mice is likely associated with incomplete Cre recombination and a wild type EC rescue effect. However, additional factors may also contribute, because in Flk1-CreBreier;Lrp5fl/- CKO mice, we often observed retinal areas with fully developed OPL vessels next to areas with no vessels. The fact that horizontal branching of capillaries alone is not sufficient to vascularize a large area in the OPL indicates that different cues may guide ECs to grow vertically into the OPL and horizontally throughout the OPL. Angiopoietin 2 (ANG2) has been shown to be essential for hyaloid vessel regression and OPL vascular development as Ang2-/- mice lack OPL vasculature and have persistent hyaloid vessels [33,34], two major abnormalities observed in Lrp5-/- retinas as well. This suggests that ANG2 may have an important interaction with Norrin-FZD4-LRP5 signaling in regulating retinal vascular development. Several lines of evidence support this enticing possibility. During hyaloid vessel regression, ANG2 critically stimulates WNT7B expression in macrophages, which is essential for inducing hyaloid EC death (also see the next paragraph); ANG2 is highly expressed in endothelial tip cells  and the leading edge of proliferating vessels . Meanwhile, it is also significantly upregulated in retinal horizontal cells located at the outer border of the inner nuclear layer right at the time when OPL vessels start to develop . Whether cooperation of ANG2 and WNT in macrophages also takes place in ECs during retinal endothelial outgrowth, with ANG2 either being a major endothelial target of Norrin-FZD4-LRP5 signaling or serving as a major guidance cue that requires the endothelial Norrin-FZD4-LRP5 pathway, remains an open question.
Defective regression of hyaloid vessels is a feature of Ndp-, Fzd4-/- and Lrp5-/- retinas [14,38,39]. This indicates that signaling from Norrin through FZD4-LRP5 is critical for regression of the hyaloid vasculature. Macrophage-derived WNT7B, also essential for apoptosis of hyaloid ECs, stimulates cell cycle entry through FZD4-LRP5, a crucial step in the subsequent ANG2-induced cell death . Whether Norrin and WNT7B regulate hyaloid apoptosis through exactly the same canonical Wnt pathway remains to be determined. However, it is possible that the same Norrin-FZD4-LRP5 pathway regulates both retinal vascular outgrowth and hyaloid vessel cell death to ensure coordinated control of retinal vessels and appropriate local oxygen and VEGF levels. How could Norrin, a sticky matrix-associated protein primarily secreted by Muller glia cells, directly regulate apoptosis in hyaloid ECs? The possible answer to this question is provided by the findings that Norrin can be expressed by macrophages  and a large number of macrophages are directly in contact with hyaloid vessels during the regression phase. Therefore, it is enticing to hypothesize that macrophages are the major source of Norrin at hyaloid vessels. Perhaps the additional control through macrophage WNT7B ensures that macrophages are available for the aftermath of endothelial cell death, and does so through an efficient utilization of the same FZD4-LRP5 complex. Whether ANG2 also serves as a regulator of retinal EC growth and hyaloid EC apoptosis in coordination with Norrin-FZD4-LRP5 signaling awaits further investigation.
The rescue results using Lrp5a214v(n) hypomorphic mice have important therapeutic implications for OPPG patients. Expression of a single Lrp5 a214v allele in osteocytes is sufficient to cause high bone mass in mice , while expression of the same Lrp5a214v allele in ECs of both Lrp5 a214v/- and Lrp5 a214v/+ mice results in normalized retinal vascular development (Fig 5B). This suggests that the retinal vasculature is less sensitive to the effects of the Lrp5 a214v allele compared to bone. Therefore, therapeutic strategies, based on manipulation of LRP5 or canonical Wnt signaling, may be effective in targeting both osteoporosis and retinal vascular defects in OPPG patients, with the retinal vasculature being more resistant to potential side effects.
With increasing prevalence of diabetes, DR has become a leading cause of blindness in adults in the western world. A key feature of DR is retinal vascular abnormalities including capillary leakage, microaneurysms, hemorrhages and fibrovascular tissue formation . These vascular problems in turn lead to disruption or loss of vision. Interestingly, many of these neovascularization features are also presented in Lrp5-/- mice. In addition, loss of LRP5, Norrin, or FZD4 causes decreased coverage of mural cells to retinal ECs [14,17], which is also considered a key attribute to the early phase of DR leading to loss of blood-retinal barrier and increased permeability . What role does Wnt signaling play in developmental neovascularization caused by loss of gene function as in Lrp5-/- retinas compared with pathological neovascularization as in DR? This study shows that loss of LRP5 leads to formation of retinal neovascular tufts and overgrowth during development. In contrast, pathological neovascularization in oxygen-induced retinopathy (OIR) mice is associated with upregulation of FZD4 and LRP5, and loss of Lrp5 function reduces the neovasculature in the mice . On the other hand, intravitreal injection of Norrin reduces neovascular tufts in the OIR model . Further work is clearly needed to resolve these contradictory findings, but genetic models such as the mice used in this study offer tools for mechanistic investigations and therapeutic studies of retinal neovascularization that could benefit patients with DR.
With genetic tools of great specificity, we demonstrate that mice lacking LRP5 only in endothelial cells recapitulate the retinal vascular defects seen in mice that have LRP5 totally knocked out. Additionally, when LRP5 function is reinstalled only in these endothelial cells in LRP5 knockout mice, the retinal vascular abnormalities are completely rescued. Thus, an endothelial autonomous function of LRP5 is a key component of the cellular mechanisms underlying OPPG and likely diabetic retinopathy as well.
Materials and Methods
All animal experiments were performed in compliance with NIH's Guide for the Care and Use of Laboratory Animals and Guidelines from the Harvard University Institutional Animal Care and Use Committee. The study was approved by the Harvard Medical School Institutional Animal Care and Use Committee (Protocol number 02074). Mice were deeply anesthetized with ketamine (150 mg/kg body weight) and xylazine (15mg/kg body weight). For euthanasia CO2 asphyxiation (longer exposure) was followed by decapitation.
Lrp5-/-  and Lrp5a214v(neo)/+  mice were generated as described. Lrp5fl/fl and Lrp6fl/fl , Flk1Breier-Cre , Rx-Cre  and CD11b-Cre  mice were kindly provided by Drs. Bart O. William, Kevin P. Campbell, Eric Swindell and Roland Baron, respectively. VE-Cad-Cre , Tie2-Cre , LysM-Cre  and tdTomato  mice were purchased from the Jackson Laboratory. Flk1Breier-Cre, Rx-Cre, CD11b-Cre, VE-Cad-Cre and Tie2-Cre mice were all transgenic lines generated by fusing the Cre gene to a fragment of the promoter sequence of Flk1, Rx, CD11b, VE-Cad and Tie2, respectively. LysM-Cre mice were knock-ins, generated by targeted insertion of the Cre cDNA into the endogenous M lysozyme locus. When animals from different genetic backgrounds were crossed, littermate controls were used to avoid confounding effects.
The following antibodies were used for immuno-staining: ColIV (EMD Millipore), CD31 (BD Pharmingen), fibronectin (Sigma), F4/80 (Invitrogen), Alexa Fluor 488 (or 555) conjugated anti-rabbit secondary antibody (Invitrogen).
For whole-mount retina IF staining, eyes were fixed in 4% PFA in PBS for 30min at room temperature (RT) or overnight at 4°C. Retinas were dissected, fixed for an additional 30min at RT, permeabilized with PBS containing 1% Triton-100 for 1 hr at RT, blocked for 1–4 hrs in PBST (0.2% Triton-100 in PBS) containing 2% BSA at RT and incubated with primary antibodies or IB4-biotin (Sigma) overnight at 4°C in PBST containing 1% BSA. After thorough washes in PBST, retinas were incubated with fluorescent secondary antibodies or streptavidin-Alexa Fluor dye conjugates together with Hoechst 33342 (Molecular Probes) overnight at 4°C, washed with PBST and flat mounted with mounting medium (Vector Lab). Images were acquired with either MetaMorph software on a Nikon 80i microscope or Nikon Elements software on a Nikon A1R laser scanning confocal microscope. Z stack images were acquired with a step size of 0.5 microns using a Nikon Ti-E motorized A1R microscope with Perfect Focus System. Z series were reconstructed as a video with Nikon Elements.
For quantification of vascular sprouting in the retina of 5-day old mice, the numbers of vascular sprouts at the edge of the growing vascularized area were counted in 20X-magnified images of IB4-biotin-stained whole mounts from 6 control and 6 Lrp5-/- mice as described . For quantification of vascular density within the OPL, the number of vascular branch-points were counted in 20X-magnified images of FITC-Dextran-perfused retinas from four Lrp5fl/fl, VE-Cad-Cre;Lrp5fl/-, Lrp5 a214v(n)/+, Lrp5 a214v(n)/-, and VE-Cad-Cre;Lrp5 a214v(n)/- mice.
2 ml of PBS containing 500 units of heparin were perfused through the heart of a deeply anesthetized mouse, followed by 1 ml PBS containing 15mg/ml FITC conjugated Dextran (Sigma) and 4% paraformaldehyde (PFA), prepared immediately prior to use. Eyes were post-fixed in 4% PFA in PBS overnight at 4°C and retinas were dissected, post-fixed in 4% PFA in PBS for 30min at RT and flat-mounted in Mounting medium (Vector Lab).
Mice were anesthetized and perfused through the heart with a solution containing 2% PFA, 2.5% glutaraldehyde and 0.1M sodium cacodylate (pH 7.4). Eyes were enucleated and fixed in the same solution overnight at 4°C. Retinas were dissected out, post-fixed for 2hrs at RT, cut in half across the optical nerve center, treated with 1% osmiumtetroxide and 1% potassium ferrocyanide, en bloc stained with 1% uranyl acetate, dehydrated and embedded in Epon (Marivac Inc., Canada). 80nm thin sections were cut and stained with uranylacetate and lead citrate. The sections were analyzed and micrographs were acquired with a JEOL 1200EX electron microscope.
Dissected retinas were rinsed with ice-cold PBS and residual PBS was removed by quickly tapping the retina to the dry area of a petri dish several times. Retinas were immediately homogenized in ice-cold M-PER lysis buffer (Thermo Scientific) containing proteinase and phosphatase inhibitors (Thermo Scientific) and stored at -80°C. After a freeze-thaw cycle, homogenates were centrifuged and the supernatants were used to quantify VEGF levels using the mouse VEGF Quantikine ELISA kit (R&D Systems). VEGF values were normalized to the total retinal protein amount analyzed by Pierce Coomassie Protein Assay Kit (Thermo Scientific). All samples were measured in duplicates.
S1 Movie. Retinal vasculature of Lrp5-/- mice.
S2 Movie. Retinal vasculature of control mice.
S3 Movie. Retinal vasculature of Rx-Cre;Lrp5fl/- CKO mice.
S4 Movie. Retinal vasculature of Flk1-Cre;Lrp5fl/- CKO mice.
IB4 whole mount IF staining.
S5 Movie. Retinal vasculature of Lrp5a214v(n)/- mice.
We would like to thank Drs. Georg Breier (Technische Universität, Dresden) and Kevin P. Campbell (University of Iowa) for kindly providing the Flk1-CreBreier mice, Drs. Milan Jamrich (Baylor College of Medicine) and Eric Swindell (the University of Texas Medical School at Houston) for the Rx-Cre mice and Drs. Jean Vacher (Institut de Recherches Cliniquies de Montréal) and Roland Baron (Harvard School of Dental Medicine) for the CD11b-Cre mice. We are also grateful to Dr. Bart Williams (Van Andel Research Institute, MI) for providing Lrp5fl/fl and Lrp6fl/fl mice. We thank Dr. Jennifer Waters and Lauren Alvarenga at the Nikon Imaging Center at Harvard Medical School for advice and assistance with microscopy. We thank Dr. Naomi Fukai for intellectual and technical help, Sofiya Plotkina and Steven Hann for assistance with genotyping and Mark Cohen for help with the manuscript.
Conceived and designed the experiments: WH BO MW. Performed the experiments: WH QL MA MH SS. Analyzed the data: WH QL MN BO. Contributed reagents/materials/analysis tools: MW. Wrote the paper: WH MN BO.
- 1. Stevens GA, White RA, Flaxman SR, Price H, Jonas JB, Keeffe J, et al. Global prevalence of vision impairment and blindness: magnitude and temporal trends, 1990–2010. Ophthalmology. 2013;120: 2377–2384. pmid:23850093
- 2. Gariano RF, Gardner TW. Retinal angiogenesis in development and disease. Nature. 2004;438: 960–966.
- 3. Gilbert C. Retinopathy of prematurity: a global perspective of the epidemics, population of babies at risk and implications for control. Early Hum Dev. 2008;84: 77–82. pmid:18234457
- 4. Grammas P, Riden M. Retinal endothelial cells are more susceptible to oxidative stress and increased permeability than brain-derived endothelial cells. Microvasc Res. 2003;65: 18–23. pmid:12535867
- 5. Frontali M, Stomeo C, Dallapiccola B, Opitz JM, Reynolds JF. Osteoporosis‐pseudoglioma syndrome: Report of three affected sibs and an overview. Am J Med Genet. 1985;22: 35–47. pmid:3931475
- 6. Gong Y, Vikkula M, Boon L, Liu J, Beighton P, Ramesar R, et al. Osteoporosis-pseudoglioma syndrome, a disorder affecting skeletal strength and vision, is assigned to chromosome region 11q12-13. Am J Hum Genet. 1996;59: 146–151. pmid:8659519
- 7. Gong Y, Slee RB, Fukai N, Rawadi G, Roman-Roman S, Reginato AM, et al. LDL receptor-related protein 5 (LRP5) affects bone accrual and eye development. Cell. 2001;107: 513–523. pmid:11719191
- 8. Nikopoulos K, Venselaar H, Collin RW, Riveiro‐Alvarez R, Boonstra FN, Hooymans JM, et al. Overview of the mutation spectrum in familial exudative vitreoretinopathy and Norrie disease with identification of 21 novel variants in FZD4, LRP5, and NDP. Hum Mutat. 2010;31: 656–666. pmid:20340138
- 9. Chen ZY, Battinelli EM, Fielder A, Bundey S, Sims K, Breakefield XO, et al. A mutation in the Norrie disease gene (NDP) associated with X-linked familial exudative vitreoretinopathy. Nat Genet. 1993;5: 180–183. pmid:8252044
- 10. Robitaille J, MacDonald ML, Kaykas A, Sheldahl LC, Zeisler J, Dubé M, et al. Mutant frizzled-4 disrupts retinal angiogenesis in familial exudative vitreoretinopathy. Nat Genet. 2002;32: 326–330. pmid:12172548
- 11. Toomes C, Bottomley HM, Jackson RM, Towns KV, Scott S, Mackey DA, et al. Mutations in LRP5 or FZD4 underlie the common familial exudative vitreoretinopathy locus on chromosome 11q. 2004;74: 721–730. pmid:15024691
- 12. Warburg M. Norrie's disease. A congenital progressive oculo-acoustico-cerebral degeneration. Acta Ophthalmol (Copenh). 1966: Suppl 89:1–47.
- 13. Xu Q, Wang Y, Dabdoub A, Smallwood PM, Williams J, Woods C, et al. Vascular development in the retina and inner ear: control by Norrin and Frizzled-4, a high-affinity ligand-receptor pair. Cell. 2004;116: 883–895. pmid:15035989
- 14. Ye X, Wang Y, Cahill H, Yu M, Badea TC, Smallwood PM, et al. Norrin, frizzled-4, and Lrp5 signaling in endothelial cells controls a genetic program for retinal vascularization. Cell. 2009;139: 285–298. pmid:19837032
- 15. Wang Y, Rattner A, Zhou Y, Williams J, Smallwood PM, Nathans J. Norrin/Frizzled4 signaling in retinal vascular development and blood brain barrier plasticity. Cell. 2012;151: 1332–1344. pmid:23217714
- 16. Junge HJ, Yang S, Burton JB, Paes K, Shu X, French DM, et al. TSPAN12 regulates retinal vascular development by promoting Norrin-but not Wnt-induced FZD4/β-catenin signaling. Cell. 2009;139: 299–311. pmid:19837033
- 17. Zhou Y, Wang Y, Tischfield M, Williams J, Smallwood PM, Rattner A, et al. Canonical WNT signaling components in vascular development and barrier formation. J Clin Invest. 2014;124: 3825–3846. pmid:25083995
- 18. Tang Y, Harrington A, Yang X, Friesel RE, Liaw L. The contribution of the Tie2 lineage to primitive and definitive hematopoietic cells. Genesis. 2010;48: 563–567. pmid:20645309
- 19. Xia C, Yablonka-Reuveni Z, Gong X. LRP5 is required for vascular development in deeper layers of the retina. 2010;5: e11676. pmid:20652025
- 20. Roberts WG, Palade GE. Increased microvascular permeability and endothelial fenestration induced by vascular endothelial growth factor. J Cell Sci. 1995;108 (Pt 6): 2369–2379. pmid:7673356
- 21. Cui Y, Niziolek PJ, MacDonald BT, Zylstra CR, Alenina N, Robinson DR, et al. Lrp5 functions in bone to regulate bone mass. Nat Med. 2011;17: 684–691. pmid:21602802
- 22. Zhong Z, Baker JJ, Zylstra‐Diegel CR, Williams BO. Lrp5 and Lrp6 play compensatory roles in mouse intestinal development. J Cell Biochem. 2012;113: 31–38. pmid:21866564
- 23. Swindell EC, Bailey TJ, Loosli F, Liu C, Amaya‐Manzanares F, Mahon KA, et al. Rx‐Cre, a tool for inactivation of gene expression in the developing retina. Genesis. 2006;44: 361–363. pmid:16850473
- 24. Alva JA, Zovein AC, Monvoisin A, Murphy T, Salazar A, Harvey NL, et al. VE‐Cadherin‐Cre‐recombinase transgenic mouse: A tool for lineage analysis and gene deletion in endothelial cells. 2006;235: 759–767. pmid:16450386
- 25. Kisanuki YY, Hammer RE, Miyazaki J, Williams SC, Richardson JA, Yanagisawa M. Tie2-Cre transgenic mice: a new model for endothelial cell-lineage analysis in vivo. Dev Biol. 2001;230: 230–242. pmid:11161575
- 26. Zovein AC, Hofmann JJ, Lynch M, French WJ, Turlo KA, Yang Y, et al. Fate tracing reveals the endothelial origin of hematopoietic stem cells. 2008;3: 625–636.
- 27. Madisen L, Zwingman TA, Sunkin SM, Oh SW, Zariwala HA, Gu H, et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat Neurosci. 2010;13: 133–140. pmid:20023653
- 28. Clausen B, Burkhardt C, Reith W, Renkawitz R, Förster I. Conditional gene targeting in macrophages and granulocytes using LysMcre mice. Transgenic Res. 1999;8: 265–277. pmid:10621974
- 29. Ferron M, Vacher J. Targeted expression of Cre recombinase in macrophages and osteoclasts in transgenic mice. Genesis. 2005;41: 138–145. pmid:15754380
- 30. Licht AH, Raab S, Hofmann U, Breier G. Endothelium-specific Cre recombinase activity in flk-1-Cre transgenic mice. Dev Dyn. 2004;229: 312–318. pmid:14745955
- 31. Motoike T, Markham DW, Rossant J, Sato TN. Evidence for novel fate of Flk1 progenitor: contribution to muscle lineage. Genesis. 2003;35: 153–159. pmid:12640619
- 32. Ai M, Holmen SL, Van Hul W, Williams BO, Warman ML. Reduced affinity to and inhibition by DKK1 form a common mechanism by which high bone mass-associated missense mutations in LRP5 affect canonical Wnt signaling. Mol Cell Biol. 2005;25: 4946–4955. pmid:15923613
- 33. Hackett SF, Wiegand S, Yancopoulos G, Campochiaro PA. Angiopoietin‐2 plays an important role in retinal angiogenesis. J Cell Physiol. 2002;192: 182–187. pmid:12115724
- 34. Rao S, Lobov IB, Vallance JE, Tsujikawa K, Shiojima I, Akunuru S, et al. Obligatory participation of macrophages in an angiopoietin 2-mediated cell death switch. Development. 2007;134: 4449–4458. pmid:18039971
- 35. Felcht M, Luck R, Schering A, Seidel P, Srivastava K, Hu J, et al. Angiopoietin-2 differentially regulates angiogenesis through TIE2 and integrin signaling. J Clin Invest. 2012;122: 1991–2005. pmid:22585576
- 36. Maisonpierre PC, Suri C, Jones PF, Bartunkova S, Wiegand SJ, Radziejewski C, et al. Angiopoietin-2, a natural antagonist for Tie2 that disrupts in vivo angiogenesis. Science. 1997;277: 55–60. pmid:9204896
- 37. Armulik A, Genové G, Betsholtz C. Pericytes: developmental, physiological, and pathological perspectives, problems, and promises. 2011;21: 193–215. pmid:21839917
- 38. Ohlmann AV, Adamek E, Ohlmann A, Lütjen-Drecoll E. Norrie gene product is necessary for regression of hyaloid vessels. Invest Ophthalmol Vis Sci. 2004;45: 2384–2390. pmid:15223821
- 39. Kato M, Patel MS, Levasseur R, Lobov I, Chang BH, Glass DA II, et al. Cbfa1-independent decrease in osteoblast proliferation, osteopenia, and persistent embryonic eye vascularization in mice deficient in Lrp5, a Wnt coreceptor. J Cell Biol. 2002;157: 303–314. pmid:11956231
- 40. Chen J, Stahl A, Krah NM, Seaward MR, Dennison RJ, Sapieha P, et al. Wnt signaling mediates pathological vascular growth in proliferative retinopathy. Circulation. 2011;124: 1871–1881. pmid:21969016
- 41. Tokunaga CC, Chen Y, Dailey W, Cheng M, Drenser KA. Retinal Vascular Rescue of Oxygen-Induced Retinopathy in Mice by NorrinRetinal Vascular Rescue of OIR in Mice by Norrin. Invest Ophthalmol Vis Sci. 2013;54: 222–229. pmid:23188723
- 42. Clement-Lacroix P, Ai M, Morvan F, Roman-Roman S, Vayssiere B, Belleville C, et al. Lrp5-independent activation of Wnt signaling by lithium chloride increases bone formation and bone mass in mice. Proc Natl Acad Sci U S A. 2005;102: 17406–17411. pmid:16293698