Involvement of Periostin in Regression of Hyaloidvascular System during Ocular Development

Abstract

Purpose.: A timely regression of the hyaloid vascular system (HVS) is required for the normal ocular development. Although macrophages have a critical role in this process, the exact mechanism remains undetermined. Periostin is a matricellular protein involved in tissue and vascular remodeling. The purpose of our study was to determine whether periostin is involved in the HVS regression.

Methods.: We used wild type (WT) and periostin knockout (KO) mice. Indocyanine green angiography and immunohistochemistry with isolectin B4 were used to evaluate the HVS regression. TUNEL-labeling was used to quantify the number of apoptotic hyaloid vascular endothelial cells. F4/80 and Iba-1 staining was performed to determine the number and location of macrophages in the vitreous. The location of periostin also was investigated by immunohistochemistry. To determine the functional role of periostin, the degree of adhesion of human monocytes to fibronectin was measured by an adhesion assay.

Results.: The HVS regression and peak in the number of TUNEL-positive apoptotic endothelial cells were delayed in periostin KO mice. The number of F4/80 positive cells in the vitreous was higher in periostin KO mice. Only a small number of Iba-1-positive cells near the hyaloid vessels was co-stained with periostin, and peripheral blood monocytes were not stained with periostin. Adhesion assay showed that periostin increased the degree of attachment of monocytes to fibronectin.

Conclusions.: These results suggest that periostin, which is secreted by the intraocular macrophages, enhances the HVS regression by intensifying the adhesion of macrophages to hyaloid vessels.

Introduction

The hyaloid vascular system (HVS) is a transient vascular network that is present in developing mammalian eyes. It is required for the normal growth and maturation of the crystalline lens and the primary vitreous. ¹ In normal eyes, the HVS regresses along with part of the retinal vasculature. A failure in the HVS regression causes congenital ocular pathologies, known as persistent fetal vasculature (PFV), which include persistent hyperplastic tunica vasculosa lentis (TVL), persistent hyperplastic primary vitreous, and persistent prepupillary membrane. The presence of these conditions can result in severe visual reduction. ^2,3  

The HVS does not regress in macrophage-deficient mice indicating that macrophages have a key role in this vascular remodeling. ⁴ It was demonstrated recently that the production of Wnt7b by macrophages in the vitreous was the mediator of the HVS regression. ⁵ It also was shown that the secretion of angiopoietin-2 (Ang2) from pericytes around the hyaloid vessels promoted the secretion of Wnt7b by the macrophages, and Ang2 itself induced the HVS regression by an inhibition of cell-survival signals. ⁶ In addition, it was shown that macrophages in the vitreous produced Ninjurin-1, which regulates the expression of Wnt7b and Ang2. ⁷ There are many reports of an association between the HVS regression and other bioactive molecules, for example transforming growth factor (TGF)-β, LYVE-1, and Arf. ^8–10 However, to our knowledge the details of the precise mechanisms leading to the HVS regression still have not been determined definitively. 

Periostin is a 90-kDa secreted matricellular protein and a member of the fasciclin (fas) family. It is composed of an amino-terminal EMI domain, tandem repeat of 4 fas I domains, and a carboxyl-terminal domain, including a heparin-binding site at its C-terminal end. It interacts with extracellular matrix (ECM) proteins, such as fibronectin and collagen I, through its EMI and fas I domains. In addition, periostin serves as a ligand of integrins, such as αvβ3 and αvβ5, and promotes cell adhesion and motility. ¹¹ Through these interactions, periostin is associated with the remodeling of tissues, such as teeth, heart, and kidney, during the developmental stages. ^12–14  

Periostin also is secreted under pathologic conditions, such as acute myocardial infarction, bronchial asthma, cutaneous wounds, and cancer, and it has a crucial role in tissue remodeling. ^15–18 We have shown that the expression of periostin was upregulated in the vitreous cavity of eyes with proliferative diabetic retinopathy. ¹⁹  

Previous studies on atherosclerotic and rheumatic cardiac valve degeneration showed that periostin-positive peripheral blood monocytes invaded the pathologic tissues and remodeled the tissues of the heart valves. ²⁰  

Macrophages in the vitreous originate from blood monocytes. ²¹ Although the HVS regression is recognized as a general model of vascular remodeling, the relationship between macrophages and periostin has not been determined to our knowledge. Thus, the purpose of our study was to investigate the role of periostin in the HVS regression. To accomplish this, we studied periostin knockout (KO) mice and determined whether periostin was involved in the HVS regression. 

Materials and Methods

Mice

We used C57BL/6 wild type (WT) mice and periostin KO mice on a C57/BL6 background. ¹⁵ All experimental procedures on the animals were performed according to the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research. 

To investigate the phenotypic alterations of the HVS regression, mice were killed by cervical dislocation from postnatal days (P) 4 to 17, and the eyes were enucleated. An ELISA kit (MOSF20; R&D Systems, Minneapolis, MN) was used to determine the genotype of the experimental mice. Heart tissues were used because periostin is expressed in the heart for up to 8 weeks after birth. ²⁰ The heart tissues were homogenized in tissue protein extraction reagent (T-PER; Pierce, Rockford, IL) containing protease inhibitors (Halt protease inhibitor cocktail kit, EDTA-Free; Pierce). 

Observations of HVS Architecture

The HVS consists of a complex of vascular networks, that is pupillary membrane, anterior/posterior TVL, iridohyaloid vessels, vasa hyaloidea propria, and hyaloid artery. The PFV expresses several phenotypes depending on which component of the fetal vasculature remains. ² Therefore, we separated the HVS into 2 groups: the anterior network consisting of the pupillary membrane, TVL, and the iridohyaloid vessels; and the posterior network consisting of the vasa hyaloidea propria and hyaloid artery. We investigated the time course of the regression of the different components. The anterior network was evaluated in two ways. The first way consisted of indocyanine green (ICG) angiography using a confocal scanning laser ophthalmoscope (Heidelberg Engineering, Heidelberg, Germany) as described in detail. ²² In brief, mice were anesthetized by an intraperitoneal injection of 15 mg/kg ketamine and 7 mg/kg xylazine, and 50 μg/g body weight of ICG was injected intraperitoneally. Pupils were dilated with 1% tropicamide and 2.5% phenylephrine. Photographs were taken within 3 minutes after the injection and used for the evaluations. 

The second method consisted of examining cross-sections to evaluate the anterior HVS network particularly the TVL. For this, mouse eyes were fixed in 4% paraformaldehyde (PFA) in PBS, embedded in paraffin, and thick sections (3 μm) were cut. After removal of the paraffin, the sections were rehydrated and stained with hematoxylin and eosin. The number of vessels around the crystalline lens was compared between 5 WT and 5 periostin KO mice. The posterior network was evaluated by immune-staining of retinal flat mounts with isolectin B4 (described below). 

Preparation of Retinal Flat Mounts and Immunohistochemistry

Retinal flat mounts containing the HVS were prepared as described in detail. ^7,23,24 In brief, the eyes were enucleated and fixed in 4% PFA for 2 hours at 4°C. After the fixation, the cornea, lens, uvea, and sclera were removed. The remaining retina containing the posterior HVS were rinsed in PBS and placed in 100% methanol for 10 minutes at room temperature. The retinas then were placed in PBS containing 2% bovine or 10% swine serum and 1% Triton X-100 for 1 hour at room temperature. The retinas were removed, and radial cuts were made for the preparation of flat mounts. The retinas were flat mounted on glass slides in mounting medium (TA-030-FM, Mountant Permafluor; Lab Vision Corporation, Fremont, CA). A fluorescent microscope (BZ-9000; KEYENCE, Osaka, Japan) was used to examine the flat mounts. 

In addition to the retinal flat mounts, cross-sections of the eyes were prepared. After removal of the paraffin, they were rehydrated and blocked. The samples then were incubated overnight in primary antibodies at 4°C and then in the secondary antibodies for 1 hour at room temperature. The periostin was made visible by a conventional avidin-biotin-peroxidase protocol with 3-amino-9-ethylcarbazole as the substrate. The slides were examined with a light microscope. 

The primary antibodies were fluorescein-labeled isolectin B4 (1:150 dilution, FL-1201; Vector Laboratories, Burlingame, CA), periostin (2 ng/μL, AF2955; R&D Systems), αSMA (1:250 dilution, F3777; Sigma, St. Louis, MO), F4/80 (1:100 dilution, MCA497R; Serotec, Raleigh, NC), Iba-1 (1:500 dilution, 019-19,741; Wako Pure Chemical Industries, Osaka, Japan), CD31 (1:500 dilution, 550,274; BD Bioscience Pharmingen, San Diego, CA), and Ninjurin-1 (1:100 dilution, 610,776; BD Bioscience Pharmingen). The secondary antibodies were Alexa Fluor 488, 546, and 647 (1:1000 dilution; Molecular Probes, Eugene, OR). 

Quantification of HVS Regression of Posterior Network

To quantify the degree of the HVS regression of the posterior network, we measured the caliber of the hyaloid vessels at different postnatal ages. The flat mounted retinas were stained with isolectin B4 as described. Four main trunks in each retina were selected arbitrarily, and the caliber of 3 points from the central, intermediate, and peripheral areas were measured on each trunk. The sum of the widths of the 12 points/retina was calculated. Six samples were measured in each group, and the results were averaged. The averaged values for the WT mice were compared to that from the periostin KO mice at each time point. 

Evaluation of Number of Macrophages in Vitreous

To quantify the number of macrophages in the vitreous, we counted the number of F4/80-positive cells around the hyaloid vessels in flat mounted retinas. We counted the number of F4/80-positive cells in 4 areas (500 × 500 μm) selected randomly for each retina. The sum of the number of F4/80-positive cells in the 4 areas/retina was calculated. Six samples were counted for each group, and the results were averaged. We compared the average values between WT and the periostin KO mice at each time point. 

TUNEL Assays

To detect apoptotic hyaloid vascular endothelial cells, a TUNEL assay was used. This was done on WT and periostin KO mice (10 mice each) with the Apop Tag Plus fluorescein In Situ Apoptosis Detection Kit (S7110; Millipore, Billerica, MA) according to instructions of the manufacturer. Eyes were flat mounted as described, and co-stained with CD31. TUNEL-positive hyaloid vascular endothelial cells were counted over the entire sample. We compared the numbers in the WT to that in the periostin KO mice at each time point. 

Leukocyte Invasion Assay

To determine whether leukocytes from the peripheral blood invaded the intraocular space, 50 μL of 1 mg/mL acridine orange (AO, 012-08942; Wako Pure Chemical Industries) was injected into the heart of P5 day WT mice. The eyes were enucleated 2 hours later, and the retinas containing the posterior HVS were flat mounted. The retinas were stained for periostin, CD31, Iba-1, or F4/80. 

Laser-Capture Microdissection

Laser-capture microdissection was used to purify the macrophages in the vitreous. Eyes of P5 WT and periostin KO mice (6 eyes each) were enucleated and embedded in OCT compound (4583; Sakura Finetech, Tokyo, Japan) and kept at −80°C until sectioning. Then, 15 μm thick sections were cut with a cryostat and placed on glass slides designed for laser microdissection microscopy (LMD6500; Leica Microsystems, Wetzlar, Germany). Macrophages were stained with toluidine blue and collected after laser microdissection. 

Peripheral Blood Monocyte Purification

Peripheral blood monocytes were isolated by positive selection using magnetic-activated cell separation (MACS) selection with anti-FITC beads, MACS columns, and MACS reagents (130-042-201 and 130-048-701; Miltenyi Biotech, Auburn, CA), according to the manufacturer's instructions. Spleens of P5 WT mice (10 mice) were removed and homogenized. Cells were stained with FITC-conjugated F4/80 antibody (11-4801-81; eBioscience, San Diego, CA), and the monocytes were isolated. 

mRNA Isolation and Quantitative Real-Time RT-PCR

Total RNA was isolated using the PureLink RNA Micro Kit (12183-016; Invitrogen, Carlsbad, CA), and real-time quantitative RT-PCR was performed with the OneStep RT–PCR kit (210,210; Qiagen, Valencia, CA) and the Light Cycler instrument (Roche, Basel, Switzerland). Primer nucleotide sequences were as follows: 

Periostin.

forward: 5′-TGAATGCCTTACACAGCCACA-3′ 

reverse: 5′-CGAGCACAGTTCACAGTGACAA-3′ 

Ninjurin-1.

forward: 5′-TCATCGTCGTGGTCAACATCTTC-3′ 

reverse: 5′-GCAGGTCCGGTACCCTTAAAGTC-3′ 

Beta-Actin (Control Gene).

forward: 5′-GATGACCCAGATCATGTTTGA-3′ 

reverse: 5′-GGAGAGCATAGCCCTCGTAG-3′. 

Cell Culture and Adhesion Assays

Human CD14+ monocytes (Lonza, Walkersville, MD) were incubated at 37°C with 5% CO₂ and 95% air. The culture medium was Dulbecco's modified Eagle's medium (DMEM, D6046; Sigma) supplemented with 10% fetal bovine serum (SH30396.03; HyClone, Logan, UT), 20 U/mL DNase 1 (D4513; Sigma), 100 U/mL penicillin, and 10 g/mL streptomycin. 

Monocytes (3 × 10⁴/well) were added to fibronectin-coated 96 well plates (354,409; BD Bioscience) with human recombinant periostin (3548-F2; R&D systems, 0, 300, and 1000 ng/mL) in DMEM (100 μL/well). After 6 hours of incubation, the cells were washed with PBS, and DMEM (150 μL/well) was added to each well with 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H tetrazolium bromide (MTT, 5 mg/mL, 20 μL/well, M2128; Sigma). After 5 hours of incubation, the supernatants were decanted, and the formazan precipitates were solubilized by the addition of 150 μL/well of 100% dimethyl sulfoxide (D8418; Sigma) and shaken for 10 minutes. The living cell number was proportional to the absorbance of MTT at 550 nm using Immuno Mini NJ-2300 microplate reader. 

Statistical Analyses

All data are expressed as the means ± SD. The significance of the differences in the measurements was determined by Student's t-tests. Differences were considered statistically significant at P < 0.05. 

Results

Delay of Regression of Posterior HVS Network in Periostin KO Mice

During normal ocular development, almost all of the hyaloid vessels regress within the first two postnatal weeks. ^4,25 ICG angiography was performed to evaluate the regression of the anterior HVS network (Fig. 1A). The active anterior HVS network was detected in WT and periostin KO mice on P7. Thereafter, the anterior network began to regress at P12 and was barely detected on P17 (data not shown). It was reported that the TVL gradually regresses and completely disappears at P16. ^25,26 For the quantification of the TVL regression, we also compared the number of vessels around the lens between WT and periostin KO mice in paraffin sections stained with HE. We confirmed that the time course of the TVL regression was the same in both types of mice (Fig. 1B). These results indicated there was no obvious difference in the time course of regression of the anterior HVS network. 

Figure 1. 

 

HVS regression delayed in periostin knockout (KO) mice. (A) ICG angiographic images from WT and periostin KO mice at P7 and P12. Anterior (upper) and posterior (lower) indicate the anterior and posterior network of HVS. The active anterior HVS network is detected around the crystallin lens similarly in WT and periostin KO mice. The anterior network gradually regressed in WT and periostin KO mice at P12. (B) Cross sections with HE staining of WT (upper left) and periostin KO (upper right) mice at P7. Arrowheads: TVL. Types of mice have the same time course in the TVL regression (lower). (C) Isolectin B4 staining of the posterior HVS network of WT (left) and periostin KO (right) mice at P6, 12, and 17. Arrowheads: components of the posterior network. There is a significant difference in the vascular density between WT and KO mouse at P12. (D) Isolectin B4 staining (green) of retinal flat mounts of WT and periostin KO mice at P4-P15. Arrowheads: hyaloid vessels. HVS regression is delayed in the periostin KO mice. The vasa hyaloidea propria persists in the peripheral area (surrounded by dot-line). (E) Isolectin B4 staining of the posterior HVS network of WT (upper) and periostin KO (lower) mice at P7. The caliber of a hyaloid vessel trunk was measured in the center, intermediate, and peripheral regions (wedged between arrowheads). Four trunks per retina were selected randomly, and the sum of the 12 points was calculated. (F) The sum of the hyaloid vessel calibers was larger in the periostin KO mice than WT mice at P7 to P12. (G) Double immunofluorescent stainings for CD31 (red) and αSMA (green) of cross sections from WT and periostin KO mice at P6. ONH, optic nerve head. Scale bar: 100 μm (D, E) and 40 μm (G). *P < 0.05, **P < 0.01.

Next, the regression of the posterior network was investigated using retinal flat mounts with isolectin B4 staining. For a better observation of the HVS network, retinal flat mounts were prepared without cutting the HVS as described. ⁷ However under these conditions, complete petaloid flat mounts were difficult to achieve because of the traction of the residual vitreous and HVS. Nevertheless, the overall posterior HVS architecture was easier to see. The posterior HVS network was more persistent in periostin KO mice than WT mice (Fig. 1C). To identify the residual HVS networks in detail, complete petaloid retinal flat mounts also were prepared (Fig. 1D). In periostin KO mice, there was a pronounced delay in the HVS regression. 

To quantify the degree of the HVS regression, especially the posterior network, we measured the caliber of the trunks of the hyaloid vessel in the retinal flat mounts (Fig. 1E). The average total caliber of the hyaloid vessels was thicker by approximately 30% in the periostin KO mice than in the WT mice in P7-12 mice. These results demonstrated a delay of the regression of the posterior HVS networks in periostin KO mice compared to that of WT mice (Fig. 1F). 

The localization of pericytes, which secrete Ang2, ⁶ around the HVS was confirmed by the immune-staining with CD31 and alpha-smooth muscle actin. There was no difference in the distribution of pericytes between periostin KO mice and WT mice (Fig. 1G). 

To detect apoptosis of the hyaloid vascular endothelial cells, double staining with TUNEL and CD31 was performed in retinal flat mounts (Fig. 2A). The number of apoptotic cells reached a peak at P7 in WT mice and at P12 in periostin KO mice (Fig. 2B). This indicated a delay in the peak of the HVS regression in periostin KO mice. 

Figure 2. 

 

Apoptosis of endothelial cells of hyaloid vessels is delayed in periostin KO mice. (A) TUNEL (green) and CD31 (red) stainings show the apoptotic hyaloid vascular endothelial cells. (B) Time course of the total number of apoptotic endothelial cells at P 4 to P15. The peak in the number of apoptotic cells was at P7 in WT and at P12 in the periostin KO mice. Scale bar: 40 μm.

Several persistent HVS models have been studied, and the persistence resulted from the absence or delay of retinal physiologic angiogenesis. ²⁷ However, the results of our study showed that retinal angiogenesis and remodeling appeared almost normal (Supplemental Figs. A–F). Although a temporal delay of the posterior HVS regression was detected, HE staining showed no obvious abnormality in the ocular development, for example the formation of the pupil and lens, retinal layered structure (data not shown). These results implicate periostin as being directly associated with the HVS regression. 

Increase of Macrophages in Vitreous in Periostin KO Mice

Isolectin B4 staining of retinal flat mounts showed a significant increase in the number of round cells around the posterior HVS networks in periostin KO mice (Fig. 3A). We confirmed that these round cells were macrophages by their staining by F4/80 (Fig. 3B). The number of F4/80-positive cells was significantly higher in periostin KO mice than in WT mice at all times except at P4 (Fig. 3C). 

Figure 3. 

 

Increase in number of macrophages in vitreous of periostin KO mice. (A) Flat mounts of retinas stained with isolectin B4 (green) of P7 WT mouse (upper left) and periostin KO mouse (upper right). A marked increase in the number of round cells can be seen in the periostin KO mice in the magnified image in the white-outlined boxes. (B) Immunostaining with F4/80 (red) shows that the round cells are macrophages. (C) Number of F4/80-positive cells was determined at ages P4–P15. The number of F4/80-positive cells was significantly higher in the periostin KO than WT mice at every time point except at P4. Scale bar: 40 μm. *P < 0.05, **P < 0.01.

Periostin Expression during HVS Regression

Periostin was expressed in the round cells, that is macrophages, around the HVS and the inner and outer nuclear layers of the retina (Fig. 4A). To determine the location of the expression of periostin relative to the macrophages and hyaloid vessels, we performed triple staining with periostin, Iba-1, and CD31 antibodies. Only a small number of Iba-1-positive cells that were located near the hyaloid vessels was co-stained with periostin (Fig. 4B). 

Figure 4. 

 

Periostin expressed in macrophages near hyaloid vessels and inner nuclear layer but not in peripheral blood monocytes. (A) The expression of periostin is made visible by the conventional avidin-biotin-peroxidase protocol with 3-amino-9-ethylcarbazole as the substrate. Cross-section of whole eye from P6 WT mouse (left), and the magnification of the vitreous (upper right) and retina (lower right). Periostin is expressed in round cells in the vitreous and inner nuclear layer. (B) Triple staining with periostin (red), Iba-1 (green), and CD31 (blue) antibodies of P5 WT mice. A small number of Iba-1-positive cells is co-stained with periostin, which are located near hyaloid vessels. No periostin-positive cells are detected in any CD31-positive cells. (C) Periostin (red) is co-localized with only F4/80 (blue), but not with AO (green). This indicates that only macrophages that had infiltrated into vitreous cavity can secrete periostin. (D) Peripheral blood monocytes were obtained from spleens of P5 WT mice using MACS. Triple staining with periostin (red) and F4/80 (green) antibodies, and Hoechst 33258 was examined. An expression of periostin could not be detected in F4/80-positive cells. INL, inner nuclear layer; ONL, outer nuclear layer. Scale bar: 50 μm (A), 100 μm (B), 40 μm (C), and 20 μm (D).

It generally is recognized that macrophages in the vitreous originate from peripheral blood monocytes. ²¹ To investigate whether these macrophages secreted periostin only in the intraocular space or also in peripheral blood, we triple stained with AO, F4/80, and periostin antibodies. The infiltrated AO-positive leukocytes were co-localized with F4/80, but not with periostin. Periostin co-localized with F4/80, but not with AO-positive leukocytes (Fig. 4C). None of the cells was triple stained. These findings suggested that periostin was secreted by only macrophages in the intraocular space and not by monocytes in the peripheral blood. 

To confirm further that monocytes in peripheral blood were not the source of periostin, we collected peripheral monocytes from WT mice at age P5 by MACS. After purification, the cells were triple stained with F4/80, periostin antibodies, and Hoechst 33528. None of the cells was double stained with F4/80 and periostin (Fig. 4D). This supported our findings that macrophages secreted the periostin in the intraocular space. 

Association of Periostin and Expression of Ninjurin-1

Ninjurin-1 is a regulator molecule for the secretion of Wnt7b and Ang2 that are required for the HVS regression. ⁷ Therefore, we investigated the relationship between periostin and Ninjurin-1 during the HVS regression. Macrophages in the vitreous were collected using LCM from WT and periostin KO mice at age P5 (Fig. 5A). After the isolation of the mRNAs, the expression of periostin and Ninjurin-1 was determined by real time RT-PCR. As expected, the amount of the mRNA of periostin was below the detection level in periostin KO mice (Fig. 5B). There was no significant difference between WT and periostin KO mice in the production of the mRNA of Ninjurin-1 (Fig. 5C). Immunohistochemistry of isolectin B4 and Ninjurin-1 showed that there was no significant difference in the distribution of Ninjurin-1 between WT and periostin KO mice (Fig. 5D). 

Figure 5. 

 

Periostin does not affect expression of Ninjurin-1. (A) Laser capture microdissection of vitreous macrophages was done. Cryosections of 15 μm were prepared from P5 WT mice and stained with toluidine blue to detect vitreous macrophages. Immunostaining before (left) and after (right) laser microdissection. (B) PCR products of periostin for vitreous macrophages of P5 WT and periostin KO mice. (C) Real time RT-PCR of Ninjurin-1 for vitreous macrophages of P5 WT and periostin KO mice. There was no significant difference in the levels. (D) Ninjurin-1 is expressed in round isolectin B4-positive cells. These cells were shown to be vitreous macrophages in Figure 3B. There is no difference in the distribution of Ninjurin-1 expression between WT and periostin KO mice. Scale bar: 100 μm (A) and 40 μm (B).

Functional Role of Periostin in HVS Regression

A study of eosinophilia showed that periostin increased eosinophil adhesion to fibronectin, a major extracellular components. ²⁸ Therefore, we examined the role of periostin in the attachment of macrophages to fibronectin using the MTT assay. Periostin enhanced the adhesion between monocytes and fibronectin (Fig. 6). 

Figure 6. 

 

Periostin enhances monocyte attachment to fibronectin. The number of attached human monocytes is proportional to the absorbance at 550 nm. Periostin (1000 ng/mL) led to a statistically significant increase in the number of cells attached to fibronectin compared to the control. *P < 0.01.

Discussion

A PubMed search with the words “periostin” and “hyaloid vascular system” did not extract any publications examining whether periostin is involved with the HVS regression. It generally is recognized that macrophages have a critical role in HVS regression. ^4–7 We found a temporal delay of the regression of the posterior HVS network, the vasa hyaloidea propria and hyaloid artery, in the periostin KO mice. In addition, the number of F4/80-positive cells in the vitreous was higher in periostin KO mice than in WT mice. No abnormality of retinal vascularization and remodeling was detected in the periostin KO mice. Therefore, we suggest that periostin could increase the macrophage-induced hyaloid vascular endothelial cell-death signaling, and this increase was a compensatory mechanism for the HVS regression. 

It already has been shown that some subtypes of macrophages are associated with the HVS regression. ^26,29 These macrophages sometimes are referred to as hyalocytes, and they originate from blood monocytes. ²¹ We found that periostin was co-localized with F4/80, but not with AO in the vitreous, and also that the peripheral blood monocytes did not express periostin. These findings suggested that macrophages secreted periostin only in the intraocular space, but not in peripheral blood. Although it is not known whether cytokines and/or chemokines are present in the vitreous cavity during development, it is possible that some factors in the vitreous cavity induce the macrophages to secrete periostin. The expression of periostin is increased by TGF-β, IL-4, and IL-13 in the periosteum and also during subepithelial fibrosis in bronchial athsma. ^30,31 Thus, these factors also could be inducers of periostin in the developing vitreous. Periostin also has been detected in the inner nuclear layer of the retina. Because it is possible that periostin secreted from the retina may be associated with the HVS regression, further studies are required to investigate how periostin expression is regulated in the retina and vitreous during ocular development. 

The macrophages that were located near hyaloid vessels were co-stained with periostin. Wnt7b and Ang2 have a crucial role in HVS regression. ^5,6 Ninjurin-1 is recognized not only as a regulator of Wnt7b and Ang2, but also as an enhancer of adhesion between macrophages and vascular endothelial cells. ⁷ Our results showed that periostin had no effect on the expression of Ninjurin-1. Although a detail examination for Wnt7b and Ang2 was not performed, the distribution of pericytes around hyaloid vascular endothelial cells appeared normal in periostin KO mice. It is possible that periostin expression is a downstream event of the Ninjurin-1 signaling pathway, and periostin affects the secretion of Wnt7b and Ang2. Therefore, further studies on the relationship between periostin and Ninjurin-1-Wnt7b-Ang2 signaling pathway are needed. 

Because only the macrophages near the hyaloid vessels express periostin, we assumed that periostin can affect the macrophage-induced hyaloid vascular endothelial cell-death signaling. This was because periostin is required for strong adhesions between macrophages and hyaloid vascular endothelial cells during the HVS regression due to the insolubility of Wnt7b. ⁵ During the developmental stage of several tissues, periostin is required for the proper mobilization of ECM by interacting with collagen I and V, and fibronectin. ^12–14 Periostin is involved in cell adhesion and motility by binding to the integrin families. ^32–34 We investigated whether periostin could affect human monocyte adhesion to fibronectin by the adhesion assay. Our results demonstrated that periostin is able to strengthen the adhesion of macrophages to the ECM components by acting as a “paste.” This then would enhance the effectiveness of the Ninjurin-1-Wnt7b signaling pathway. 

To the best of our knowledge, this is the first study to demonstrate that periostin is secreted from macrophages in the vitreous and is involved in HVS regression. PFV consists of many phenotypes depending on which component of the HVS remains. ² Consistent with this, in spite of our finding that periostin KO mice have a delay in the regression of the posterior HVS network, the regression of the anterior HVS network was similar in WT and periostin KO mice. It generally is accepted that macrophages infiltrate the vitreous from surrounding areas of the lens at specific times during the HVS regression. ^26,29 Although there is no doubt that macrophages have the main role in the HVS regression, it is possible that the molecules that affect the functions of macrophages are completely different depending on their localization in the anterior chamber or in the vitreous. It was assumed that some molecules other than periostin modified the adhesion between macrophages and endothelial cells in the anterior chamber. 

In conclusion, the results of our study indicated that periostin, which is secreted by the intraocular macrophages, enhances HVS regression by strengthening the adhesion between macrophages and hyaloid vessels. We suggest that periostin KO mice are a useful model to study the mechanism involved in HVS regression. 

Supplementary Materials

pdf S1 

Acknowledgments

Masayo Eto (Kyushu University, Fukuoka, Japan) and Hirotsugu Yamaguchi (Kyodo Byori, Kobe, Japan) provided excellent technical assistance. 

Footnotes

 Supported in part by grants from the Ministry of Education, Science, Sports and Culture, Japan (TI and SY), and Takeda Science Foundation (SY).

Footnotes

 Disclosure: M. Arima, None; S. Yoshida, None; T. Nakama, None; K. Ishikawa, None; S. Nakao, None; T. Yoshimura, None; R. Asato, None; Y. Sassa, None; T. Kita, None; H. Enaida, None; Y. Oshima, None; A. Matsuda, None; A. Kudo, None; T. Ishibashi, None