Photoreceptor degeneration and choroidal neovascularization in aged BALB/c mice following systemic neonatal murine cytomegalovirus infection

Age-related macular degeneration (AMD) represents a leading cause of irreversible visual dysfunction in older individuals but its genesis is poorly understood. Human cytomegalovirus (HCMV), which infects 50 to 80% of humans, is usually acquired during early life and persists in a latent state for the life of the individual. Here we show that systemic neonatal murine cytomegalovirus (MCMV) infection of BALB/c mice resulted in dissemination of virus to the eye where it localized to choroidal endothelial cells and RPE cells. MCMV underwent ocular latency in all neonatally infected mice and latent/persistent ocular infection was associated with expression of MCMV immediate early genes, signicant upregulation of several inammatory/angiogenic factors. AMD–like pathology, including basal lamina deposits (BLamD), subretinal drusenoid deposits (SDD), severe photoreceptor degeneration, choroidal neovascularization (CNV) develop in eyes of latently-infected aged mice. Our study suggests a possible viral etiology for AMD as a result of latent/persistent ocular CMV infection.


Introduction
Age-related macular degeneration (AMD) is a complex, multifactorial, progressive disease, which is the leading cause of irreversible visual dysfunction in older individuals [1][2][3] . The disease is characterized in its early stages by lipoprotein deposits at the basal and apical aspects of the RPE and in its advanced forms by choroidal neovascularization (CNV) 1,2 or geographic atrophy (GA) of the outer retinal tissue, retinal pigment epithelium (RPE) and choriocapillaris [4][5][6] .
Human cytomegalovirus (HCMV) is a ubiquitous beta-herpesvirus, which infects 40 to 80% of individuals in the human population 7 . CMV persists for the life of the host through cycles of latency and reactivation following primary infection 8 . The establishment of CMV latency can occur in multiple sites and cell types in the host, including endothelial cells and hematopoietic cells 9 . HCMV is usually acquired during early life, with the incidence of congenital infection ranging from 0.5-2.4% of all live births [10][11][12] . More than 12% of 1 year old U.S. infants are infected with HCMV 13 . Since the innate and adaptive immune systems are not fully mature during early life 14 , acquiring HCMV infection during this period could lead to widespread virus dissemination throughout the body, resulting in viral latency in a number of sites including the eye. Indeed, the eye is one of the major target organs of congenital HCMV infections with the incidence of HCMV chorioretinitis reported to be 25% in infants with symptomatic congenital HCMV infection 10,[15][16][17] . Although only about 1% of infants who are asymptomatic and congenitally infected have CMV chorioretinitis 10,[15][16][17] , CMV could spread to and become latent in the eye in signi cantly more asymptomatic infants without chorioretinitis. Indeed, recent studies in our laboratory of ocular tissue from human cadavers, revealed that HCMV DNA was present in 4 of 24 choroid/RPE samples, suggesting that choroid/RPE might be a site of HCMV latency 18 .
HCMV infection might be a novel risk factor for the progression of AMD since there is a signi cant association between elevated anti-HCMV IgG titers and neovascular AMD compared to either dry AMD or non-AMD controls 19 . The cytomegaloviruses (CMVs) are species-speci c and murine cytomegalovirus (MCMV) infection of mice is widely used to mimic human diseases. By using MCMV resistant black mice, previous studies from Cousins et al 20 as well as from our own laboratory 18 have shown that MCMV exacerbates the development of CNV in both a laser-induced CNV model 20 as well as in a VEGFoverexpressing model 18 . In the studies presented herein, we demonstrate for the rst time that latent ocular infection is associated with the development of advanced AMD, including severe photoreceptor degeneration and CNV in aged, MCMV-infected BALB/c mice.

Results
MCMV disseminates to the eye following systemic infection of neonatal mice. To test our hypothesis that acquiring cytomegalovirus infection early in life results in widespread dissemination of virus to several sites including the eye, 50 pfu of MCMV were injected i.p. into BALB/c mice at either <3 days, 1 week, 2 weeks or 6 weeks after birth. All mice, including newborns, survived viral infection and appeared healthy. At day 14 post infection (p.i.), as shown in Figure 1A, virus recovery was higher in mice infected with MCMV soon after birth, with replicating virus recovered from eyes of all mice infected as newborns.
However in mice infected with MCMV at 6 weeks of age, no replicating virus could be detected. MCMV DNA was detected in eyes of all mice which were infected with MCMV at or before 2 weeks of age (26 of 26 eyes) whereas MCMV DNA was detected in only 2 of 6 eyes of mice which were infected with MCMV at 6 weeks of age (p <0.0001, Chi-square test). Not surprisingly, MCMV was recovered from salivary glands and lungs in all 4 groups of mice which were infected with MCMV during the time interval <3 days to 6 weeks after birth ( Figure 1A).
To investigate the exact ocular localization of MCMV in mice infected at <3 days after birth, eyes were immunostained with both anti-MCMV EA and anti-RPE65. We observed that the majority of MCMV infected cells were located in the choroid ( Figure 1B, Table 1). A few MCMV infected cells were also observed in the RPE layer ( Figure 1B, indicated by arrow). In addition, virus-infected cells were also occasionally observed in anterior segments including the iris and ciliary body (Table 1). No MCMV EA positive cells were observed in the inner retina ( Figure 1B). A small number of in ltrating cells were observed in the subretina/photoreceptor layer of 2 weeks-old uninfected control eyes, but signi cantly more were observed in the photoreceptor layer of MCMV-infected mice of the same age (supplemental   table 1). These cells did not stain with either anti-RPE-65 or anti-MCMV EA ( Figure 1B, indicated by arrow heads). Although some in ltrating cells in the photoreceptor layer of MCMV-infected mice stained positive for CD45 ( Figure 1C), these cells stained negative for several immune markers including Iba-1( Figure 1D), CD11b ( Figure 1C), DX-5 (not shown), and CD11C (not shown) indicating that these cells were most likely not systemic immune cells. Of note, signi cantly more CD45 positive cells were observed in the choroid of MCMV-infected mice than in the choroid of age matched control mice (Supplemental Table 1). H & E staining also showed some in ltrating cells in the photoreceptor layer ( Figure 1F).
However, no remarkable pathological changes were observed in the inner retina ( Figure 1F). To investigate more closely the localization of MCMV in the eye, infected ocular tissue was examined electron microscopically following immunogold staining with anti-MCMV EA. These experiments showed that immunogold-labeled MCMV EA was present in the nuclei of some vascular endothelial cells ( Figure  1G) and pericytes ( Figure 1H) in the choriocapillaris, and also in sporadic RPE cells ( Figure 1I).
Electron microscopical analysis was used to determine if MCMV infection in the choroid and RPE was associated with a compromised blood-retina barrier and/or pathological changes in the inner retina. As shown in Supplemental Figure 2, the outer blood-retina barrier (BRB) appeared to be intact and Bruch's membrane showed no remarkable disorganization or disruption when viewed electron microscopically. While the majority of RPE cells in infected mice were indistinguishable from RPE cells in uninfected control mice, large vesicles were noted in occasional RPE cells while increased phagocytosis of outer segments (OS) was also observed. Neither cell death nor any remarkable in ltration were found in the inner retina of MCMV-infected mice although a few in ltrating cells were observed in the photoreceptor layer.
Ocular MCMV infection becomes latent in choroid and RPE. To determine if MCMV infection undergoes latency in the eye and extraocular organs/tissues following systemic infection, 50 pfu of MCMV were injected i.p. into BALB/c mice at either <3 days, 1 week, 2 weeks or 6 weeks after birth. Mice were sacri ced at 3 months p.i. and tissue samples were collected. No replication competent MCMV could be recovered from eyes, lungs or salivary glands at this time. Nevertheless, MCMV DNA was still present in eyes of mice infected during early life and was detected in 9 of 9 eyes of mice infected at <3 days after birth and 5 of 8 eyes of mice which were infected at 1 week of age. However, when mice were infected with MCMV at 2 weeks of age, MCMV DNA was presented in only 2 of 10 eyes. No MCMV DNA was detected in eyes of mice which were infected at 6 weeks of age. In contrast, MCMV DNA was present in the lungs of all mice, irrespective of time of infection.
Signi cantly more in ltrating cells were observed in the choroid and photoreceptor layer during acute infection (Supplemental Table 1). To determine if these in ltrating cells were still present in the photoreceptor layer and choroid during latency, eyes were collected 4 months post neonatal infection and immunostained with anti-CD45. As shown in Supplemental Table 1, there were no signi cant differences in the numbers of in ltrating cells between MCMV-latently infected eyes and age matched uninfected controls, although the number of CD45+ cells was slightly elevated in choroids of latently infected mice (5.20±3.96) compared to uninfected controls (1.60±1.52).
Since MCMV infected cells was mainly observed in the choroid ( Figure 1B), posterior eyecup cultures were employed to determine if latent ocular MCMV could be reactivated in vitro. Eyes were collected at 4 months p.i. Posterior eyecups, consisting of sclera, choroid, and RPE, were separated and cultured at 37°C. Culture medium was collected bi-weekly and assayed by plaque assay for replicating virus. Posterior eye cup cultures began to produce replicating virus beginning at day 7 of culture (2/12 positive) while at day 14 of culture. 8 of 12 eye cup cultures produced replicating virus. Following 3 weeks in culture, virus was detected in 11 of 12 samples. To identify the location of the reactivated virus in vitro, eye cup cultures were stained for MCMV EA antigen. This showed that MCMV EA was present in sclera and choroid, and also co-localized with some RPE-65 positive RPE cells (Figure 2A). In addition, we separated RPE cells and choroid from posterior eyecups of MCMV latently infected mice according to a protocol described previously 21 and performed PCR to detect MCMV DNA. These experiments indicated that MCMV DNA was present in both RPE cells (4/4) and choroid (4/4).
Virus reactivation in latently infected eyes following systemic immunosuppression. Our previous studies have shown that latent virus can be reactivated from infected eyes by deeply systemic immunosuppression several months after MCMV intraocular inoculation 22 . To determine if latent MCMV could be reactivated in eyes of neonatally infected mice by immunosuppression in vivo, newborn BALB/c mice were inoculated i.p. with 50 pfu of MCMV and 4 months later, Some mice were deeply immunosuppressed with methylprednisolone plus anti-T cell antibodies, which depleted more than 99% of CD4 and CD8 T cells as previously described 22,23 . After 2 weeks, eyes and extraocular organs were collected and analysed by plaque assay. This showed that no replicating virus was present in eyes or extraocular tissues, including salivary glands and lungs, of any latently infected immunocompetent mice. In contrast, replicating MCMV was recovered from 6 of 8 eyes of immunosuppressed mice and also from the majority of extraocular tissues ( Figure 2B). Although signi cantly more replicating virus was recovered from lungs and salivary glands compared to eyes during acute infection ( Figure 1A), a surprisingly similar amount of replicating virus was recovered from eyes and salivary glands or lungs during reactivation by immunosuppression ( Figure 2B).
In order to assay for the presence of virus in leukocytes, DNA and total RNA were isolated from white blood cells of the two groups of mice. Expression of the MCMV IE1 and GB genes were analyzed by real time RT-PCR while PCR was used to test for the presence of virus DNA. Although no MCMV DNA was detectable in the blood of non-immunosuppressed, latently infected mice (0/6), MCMV DNA was detected in leukocytes in 5 of 8 immunosuppressed mice. No MCMV IE1 or GB transcripts were detectable in any blood samples from either immunosuppressed or non-immunosuppressed latently infected mice.
To identify the ocular location of MCMV reactivation, we performed antibody staining with both anti-MCMV EA and anti-RPE-65. As previously observed during acute infection (Table 1), MCMV-infected cells were present in the choroid of all eyes (6/6) and in RPE cells of some eyes (3/6) (Table 1, Figure 2C). No virus was detected in the inner retina, although virus-infected cells were observed in anterior segments including the iris and ciliary bodies of 6 of 6 mice ( Figure 2D), compared to only 1 of 7 mice during acute infection (Table 1). This suggests that reactivated MCMV might spread from the choroid to the ciliary body and iris via the uveal tract. H & E staining showed no remarkable pathological changes in the inner retina ( Figure 2E, 2F, 2G). AMD-like pathology by Spectral-Domain Optical coherence tomography (SD-OCT). SD-OCT is a noninvasive imaging technique, providing high-resolution, cross-sectional images of the retinal microstructure in vivo 24,25 . By using the Envisu R2210 system of Leica Microsystems, SD-OCT examinations were performed and the retinal thickness was measured in eyes of MCMV latently infected mice and eyes of age matched controls at 4, 8 and 18 months p.i. ( Figure 3A). Whereas no remarkable pathological changes were detected by SD-OCT in eyes of MCMV latently infected mice at 4 months p.i., the mean retinal thickness was signi cantly lower in eyes of latently infected mice, compared to agematched control eyes, beginning at 8 months p.i., as shown in Figure 3B. In control mice, mean retinal thickness deteriorated with age and was lower in eyes of control BALB/c mice at 18 months of age compared to control mice at 8 months of age. However, apart from changes in retinal thickness, no other remarkable pathological changes were detected by SD-OCT in eyes of MCMV latently infected mice at 8 months p.i. or in control eyes at all ages. In contrast, besides signi cantly lower retinal thickness in all 40 eyes examined, as shown in Figure 3A, severe photoreceptor degeneration, including disappearance of the entire outer nuclear layer (ONL) in some areas, was observed in 21 eyes at 18 months p.i. Other pathological changes observed in these 21 eyes included CNV-like lesions (average 2 per eye) ( Figure 2C) and retinal detachment (not shown), which were noted in 6 and 4 eyes respectively. Eyes with CNV-like lesions were removed, sectioned and stained with anti-CD31 and isolectin. The results con rmed the presence of CNV lesions in the photoreceptor layer ( Figure 3D, 3E). Speci cally, at 4 months p.i., many large lipid vesicles were observed inside RPE cells as shown in Supplemental Figure 3D. Lipid vesicles were also noted in Bruch's membrane (BM) (Supplemental Figure   3B, 3C), which appeared to be thickened (Supplemental Figure 3B) and darkly stained in some areas (Supplemental Figure 3D, 3E). Although platelets were occasionally observed in the choriocapillaris of age-matched uninfected control eyes, more were observed in eyes of MCMV latently infected mice at 4 months p.i. Platelets were attached to vascular endothelia in the choroid and appeared activated (Supplemental Figure 3F, 3G, 3H), while cell death was also noted in some areas of the choroid (Supplemental Figure 3E). However, the de ning pathologies which are the hallmark of AMD, such as basal lamina deposits (BlamD) and photoreceptor degeneration, were not observed at this time point. At 8 months p.i., many large lipid vesicles were still present inside some RPE cells ( Figure 4B, arrows), while only occasional lipid vesicles were noted in RPE cells of control eyes ( Figure 4A). Although lipid vesicles were not noted in BM, BlamD deposits were now present in some areas between the plasma membrane and basal lamina of the RPE ( Figure 4B, 4C, 4E, 4H) in 3 of 3 eyes from MCMV latently infected mice at this time point. The majority of BlamD deposits were stained lightly (4B, 4C, 4E, 4H), although a few were dark-stained (not shown). As shown in Figures 4C and 4D, there was a loss of tight junctions between some RPE cells located below the BlamD, with evacuated areas created between these RPE cells. In addition, some RPE cells appeared atypical and exhibited marked vacuolization ( Figure 4F, 4G). In the subretinal space near these atypical RPE cells, we observed large structures, which were separated by an intact septum or membrane ( Figure 4F, indicated by star) and which contained heterogeneous material, including dark-staining OS-like fragments ( Figure 4F, indicated by arrow heads).
The morphology of nearby photoreceptors was disturbed, with shortening and loss of outer segments ( Figure 4F, indicated by arrows). Although the identity of this structure is uncertain, it may be a large phagolysosome or a subretinal drusenoid deposit (SDD) separated by an intact septum. SDD was recently recognized as another type of extracellular lesion located in the subretinal space of AMD eyes [26][27][28] . A few in ltrating cells were observed in the subretinal space of infected ( Figure 4H, 4I, indicated by arrows) and control mice (not shown) at 8 months p.i. while, dark-staining OS-like fragments were observed inside some in ltrating cells of infected mice. As shown in Figure 4J, discs were occasionally noted in these fragments.
At 18 months p.i., 5 eyes from infected mice, including 1 eye with severe photoreceptor degeneration and CNV-like lesions identi ed by SD-OCT, as well as 3 eyes from age matched, uninfected control mice, were removed and processed for electron micrsocopy. Large lipid vesicles were present in some RPE cells of eyes of 18 month old uninfected mice ( Figure 5A) and a few in ltrating cells were also observed in the subretinal space of these mice ( Figure 5A, indicated by arrow head). In addition, loss or shortening of OS was sometimes noted ( Figure 5B). However, no other pathologies, such as BlamD or SDD were observed in any of the 3 control eyes examined.
In eyes of MCMV-infected 18 month old mice, in ltrating cells were sometimes observed in the subretina while dark-staining OS-like fragments were also observed inside some of these cells ( Figure 5C, 5D). As shown in Figure 5G, BlamD deposits were observed in 4 of 5 eyes examined. Some SDD-like structures were noted in the subretinal space of all 5 eyes of MCMV-infected mice. (Figure 5F, 5G, 5H) and although some were well formed, these structures are apparently not cellular vesicles since they were not separated by an intact septum or membrane. Some deposits contained dark-staining OS-like fragments ( Figure 5F, 5H, indicated by arrows), while others appeared drusen-like and were lightly stained. The morphology of nearby photoreceptors was disturbed, with shortening and loss of OS and IS ( Figure 5F, indicated by arrow heads), while in some areas, OS were completely absent with SDD deposits extending into the IS layer ( Figure 5F, indicated by star). Some photoreceptors appeared apoptotic, with nuclear shrinkage and strong chromatin condensation. ( Figure 5E). In addition, a few ectopic photoreceptor nuclei were noted in the IS and OS layers ( Figure 5I) while large lipid vesicles were occasionally observed inside RPE cells of MCMV-infected mice ( Figure 5C). Thinned and disrupted RPE structure was observed adjacent to areas of SDD ( Figure 5H) and some RPE cells exhibited marked vacuolization ( Figure 5E). Extensive choroidal platelet in ltration both within choroidal capillaries and in perivascular tissue ( Figure 5J) was observed in 5 of 5 eyes examined. One particularly striking abnormality was the presence of relatively large arteries in some areas of the choroid. An example is shown in Figure 5K, with a large artery located between the sclera and Bruch's membrane although no capillaries were observed in this region. .
The presence of CNV lesions and severe retinal degeneration was observed in one eye and con rmed by electron microcopy (Figure 5M, 5N). As shown in Figure 5M, new blood vessels were noted in the sub-RPE space, suggesting that their origin was choroidal rather than retinal although the RPE in this area was located adjacent to the innernuclear layer (INL) due to the loss of the entire ONL ( Figure 5M). Severe retinal degeneration was not con ned to areas of CNV and was also observed in non-CNV areas of the retina. As shown in Figure 5L, no ONL is present while the INL is reduced in size and composed of only 3 to 4 layers of cells.
Expression of MCMV immediate early genes and host in ammatory/angiogenic factors. Cytomegalovirus latent infection could result in expression of a number of virus encoded proteins with the potential to signi cantly alter homeostasis of the latently infected cell and the surrounding cellular environment 29 . Therefore, expression of the MCMV IE1, IE3 (immediate early genes) and gB (a late gene) genes was analyzed by real-time RT-PCR using RNA isolated from eyes of latently infected BALB/c mice at 4, 8 and 18 months p.i. As shown in Table 2, expression of MCMV IE gene transcripts was detected in the majority of eyes from latently infected BALB/c mice at all 3 time points. In contrast, expression of transcripts from the gB late gene was detected only rarely in latently infected eyes. No replicating virus was detected by plaque assay in any eyes of latently infected mice at 4, 8 or 18 months p.i.
CMVs have been shown to induce angiogenesis via the production of various angiogenic factors 30 including cytokines/chemokines such as IL6 31,32 and CCL5 33 and growth factors such as angiopoietin 34 , VEGF 35 , TGF-β 34 , and M-CSF 34 . Although IE proteins alone are not su cient to drive viral genome synthesis or production of infectious viral progeny in infected cells 36 , they can in uence the cellular environment 37 and previous studies have suggested that IE1 alone can trigger a proin ammatory host transcriptional response via a STAT1-dependent mechanism 38,39 . Since latent ocular infection results in expression of viral IE genes, we hypothesized that expression of some in ammatory factors and growth factors might also be stimulated by latent ocular MCMV infection. Therefore, real time RT-PCR was used to test for expression of these genes in latently infected and control eyes.
We observed that transcript levels of several cytokines and growth factors, particularly CCL5, were strongly upregulated in latently infected eyes compared to age matched control eyes (Table 3). At 4 months p.i., the relative levels of CCL5, angiopoietin I and TGF-β1 transcripts were more than 10 times higher than in control, uninfected eyes while overall 9 of 11 gene transcripts studied were upregulated in latently infected eyes. By 8 months p.i., levels of CCL5 gene transcripts remained elevated, while mRNA levels of the other in ammatory factors and growth factors had returned to levels similar to those of controls. By 18 months p.i., CCL5 levels were still elevated, while IL6 transcripts were approximately 27fold higher relative to controls.
To determine if elevated transcript levels were associated with increased protein expression, relative protein levels of CCL5, angiopoietin I, TGF-β1 and its receptors, as well as IE6 and VEGF were measured by ELISA or western blot. At 4 months p.i, relative protein levels of angiopoietin I ( Figure 6C, 6D) and CCL5 ( Figure 7H) were signi cantly increased in latently infected eyes compared to age-matched controls. Although relative levels of the 45 KD inactive form of TGF-β1 ( Figure 6A, 6B) and TGF-β receptors including TGF-β R1 and TGF-β R2 ( Figure 6C, 6D) were signi cantly increased in eyes of latently infected mice at 4 months p.i., surprisingly, protein levels of the active forms of TGF-β1 including both the 25 KD dimer and 12KD monomer ( Figure 6A, 6B) were signi cantly lower in latently infected eyes compared to age-matched control eyes. By 8 months p.i., protein level of CCL5 remained signi cantly elevated in eyes of latently infected mice ( Figure 7H), while protein levels of other in ammatory and growth factors had returned to levels similar to those found in control eyes (Figure 6). At 18 months p.i., protein level of CCL5 were signi cantly elevated in eyes of latently infected mice, compared to age matched control eyes or compared to eyes from latently infected mice at earlier time points (4 or 8 months p.i.) ( Figure 7A). Interestingly, protein levels of CCL5 were signi cantly elevated in 18 month old aged control eyes, compared to 4 or 8 month old control eyes. Although relative levels of the 45 KD inactive form of TGF-β1 ( Figure 6A, 6B) was signi cantly increased in eyes of latently infected mice at 18 months p.i compared to controls, protein levels of the active forms of TGF-β1 including both the 25 KD dimer and 12KD monomer ( Figure 6A, 6B) as well as TGF-β receptors including TGF-β R1 and TGF-β R2 ( Figure 6C, 6D) were signi cantly lower in latently infected eyes compared to age-matched control eyes. In contrast, IL6 protein levels were signi cantly elevated in 18 month old latently infected eyes compared to age-matched control eyes ( Figure 7B).

Discussion
Several studies have shown that MCMV can initiate latent ocular infection following intraocular Virus-induced disruption of an immature outer blood-retina barrier (BRB) and weak anti-virus immunity 47 in neonatal mice could contribute to the establishment of MCMV latency in the choroid/RPE of almost all mice which were infected early in life. Systemic MCMV disseminated not only to the choroidal endothelia but also passed through the outer blood-retina barrier and infected pericytes and RPE cells in the majority of neonatally-infected mice, although the barrier appeared intact when viewed electron microscopically.
The eye is an immune privileged site in which in ammatory responses are limited in order to minimize the risk to vision integrity [48][49][50][51][52] . RPE cells play an important role in this phenomenon by producing immunosuppressive factors such as TGF-β, alpha-melanocyte-stimulating hormone (α-MSH) and vasoactive intestinal peptide [48][49][50][51] as well as by inhibiting immune T cells and converting T cells to regulatory (Treg) cells 52 . Therefore, the immunosuppressive ocular microenvironment around the choroid and RPE could also facilitate the establishment of MCMV latency there.
Many patholgies and abnormalities observed following prolonged MCMV ocular latency are similar to those observed in AMD, a progressive degenerative disease and the leading cause of severe, permanent vision loss in people over age 60 [1][2][3] . Although the exact events which contribute to the development of AMD remain uncertain, studies have implicated immunological and in ammatory mechanisms [53][54][55] with various clinical and genetic data supporting a tight association between chronic low-grade in ammation and the pathogenesis of AMD 56,57 . Although AMD models in mice, rats, rabbits, pigs and non-human primates have recreated many of the histological features of the disease and provided much insight into its underlying pathological mechanisms; there is currently no single animal model in which all of the characteristics of AMD develop in a progressive manner 53,54,58 .
MCMV latency in the choroid and RPE cells of neonatally-infected BALB/c mice is associated with expression of virus immediate early genes including IE1 and IE3 in the absence of de novo viral protein synthesis. Major immediate-early enhancer (MIE) activity alone does not guarantee full virus reactivation and production of progeny virus 37 , since ectopic expression of the IE proteins is not su cient to drive viral genome synthesis or infectious progeny production in infected cells 36 . Nevertheless IE expression does modulate the cellular environment as well as the transactivation of early virus genes 37 . For instance, it has been reported that, HCMV IE1 can trigger a proin ammatory host transcriptional response via a STAT1-dependent mechanism 38,39 . In our experiments, expression of IE genes in latently infected eyes was associated with increased transcription and protein production of several in ammatory/angiogenic factors, including CCL5, at all 3 time points studied. CCL5 is produced by platelets and other cells including macrophages, eosinophils, broblasts, endothelium, epithelium and endometrial cells 59 . This chemokine participates in multiple biological processes, from pathogen control to enhancement of in ammation 59 and angiogenesis 60 . Previous studies have suggested that CCL5, which is produced by human RPE cells following chronic in ammatory stimulation, could play an important role in the development of AMD via interactions with CCR3 61 . A recent studies demonstrated that patients with GA have a higher plasma levels of CCL-5 and a higher expression of CCR5 in peripheral blood mononuclear cells than healthy controls 62 . Therefore CCL5 production during MCMV ocular latency might contribute to the development of AMD-like pathology.
In summary, the results presented herein demonstrate for the rst time that systemic MCMV infection of BALB/c mice can spread to the eye with subsequent establishment of latency at the choroid and RPE when infection is acquired early in life. MCMV latency in the choroid/RPE of BALB/c mice was associated with the upregulation of in ammatory/angiogenic factors and most importantly, the development of AMD-like pathology in a progressive manner, including deposits in both basal and apical aspects of the RPE, severe photoreceptor degeneration and eventually, CNV in later life. Therefore, we propose that systemic neonatal MCMV infection of BALB/c mice may represent a useful model in which to study these pathologies which are typical of AMD, particularly since the time course of virus-induced pathologies is relatively prolonged, mimicking the situation in human AMD development. This should allow for a more precise dissection of early and late events which eventually culminate in CNV. We have recently demonstrated the presence of HCMV DNA in the choroid/RPE, of approximately 17% of a group of human cadavers 18 , consistent with latent infection. It is possible that CMV latency could therefore be a risk factor for the occurrence and development of AMD.

Methods
Cells and virus. MCMV strain K181 was originally provided by Dr. Edward Morcarski, Emory University, Atlanta, GA. Virus was prepared from the salivary glands of MCMV-infected immunosuppressed BALB/c mice and virus stocks were titered on monolayers of mouse embryo broblast (MEF) cells as described previously 63 . A fresh aliquot of virus stocks was thawed and diluted to the appropriate concentration for each experiment.
Mice. Breeding pairs of BALB/c mice were purchased from Jackson Laboratory (Bar Harbor, ME). All mice were given unrestricted access to food and water and were maintained on a 12-hour light cycle alternating with a 12-hour dark cycle. Anesthesia protocols have been described previously 64 . The breeding and treatment of animals in this study adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and was approved by the Institutional Animal Care and Use Committee of Augusta University. The rd8 mutation was excluded by genotyping.
Antibodies. Anti-MCMV early antigen (EA) 65 was labeled with FITC (Sigma-Aldrich, St. Louis, MO) as previously described 66 . Anti-RPE65 was kindly provided by Dr. Michael Redmond (National Eye Institute, National Institutes of Health, Bethesda, MD). Other antibodies used in this study were obtained from the following sources: Rabbit anti-mouse TGF-β antibody (#3711), rabbit anti-mouse TGF-β-R2 antibody (#79424) and mouse anti-mouse β-actin antibody (#3700) were all from Cell Signaling Technology, Inc. Eyes, salivary glands and lungs were collected for analysis by plaque assay, RT-PCR, immuno uorescence staining and hematoxylin and eosin (H&E) staining.
Posterior eye-cup culture. The method of posterior eye-cup culture has been previously described by our laboratory 67 . Eyes were collected from MCMV latently infected mice and the posterior eye cup, consisting of sclera, choroid, and a monolayer of RPE was isolated.. The posterior eyecup was attached to a sterile membrane lter (Schleicher & Schuell, Dassel, Germany) with the sclera side in contact with the lter and mounted on a coverslip (Nalge Nunc international, Rochester, NY) with Matrigel (BD Biosciences, Bedford, MA). The coverslip with attached eye-cup was inserted in a culture tube in 1 ml of culture medium (DMEM, 10% FBS) and cultured in a roller incubator at 37°C with a rotation rate of 10-15 rpm. Culture medium was collected after 1 day and twice weekly thereafter and examined by plaque assay for replicating virus. Cultures were also harvested and stained for MCMV EA as described below.
SD-OCT examinations and measurement of retinal thickness. Mice were anesthetized and SD-OCT was performed using the Bioptigen Spectral-Domain Ophthalmic Imaging System (Envisu R2200; Bioptigen, Morrisville, NC, USA) as described previously 18,64 . Brie y, pupils were dilated with 1% tropicamide and systane ultra lubricant eye drop was applied liberally to keep the eye moist during imaging. Images, including averaged single B scan and volume intensity scans (VIP) were acquired. A total of 3 scanning images were acquired from the center of the optic nerve head, the left sphere of the eye (from optic nerve head to left iris), and the right sphere of the eye (from optic nerve to right iris) in each mouse. The highly re ective CNV lesions located above the RPE layer were quantitated followed by measurement of and total retinal thickness was measured in the scanning image taken from the center of the optic nerve head by using InVivoVue™ Diver 2.4 software (Bioptigen).
Electron Microscopy and Immunogold Staining. As previously described by our laboratory 67 Following washing, grids were stained with 2% uranyl acetate in 70% EtoH and lead citrate and visualized in a JEM 1230 transmission electron microscope.
Nucleic acid puri cation. DNA was extracted from eyes, salivary glands, lungs and peripheral blood leukocytes by overnight digestion with Proteinase K at 56°C with continuous vigorous mixing, followed by centrifugation. The supernatant, was removed and DNA precipitated with an equal volume of isopropanol prior to resuspension in water. Genomic DNA was diluted to 50 ng/µl prior to use. Total RNA was extracted from eyes, salivary glands, lungs and peripheral blood leukocytes using the RNeasy mini Kit Mix (Bio-Rad), 0.2 µl of 20 pmol/µl primer mixture and 1 µl of DNA or cDNA prepared by reverse transcription (RT) of mRNA using 500 ng of total RNA. Reaction conditions were 5 min at 95°C followed by 40 cycles of 95°C ,15 s, 60°C, 20 s and 72°C, 20 s. Ampli cations were analyzed by CFX Maestro™ Software (Bio-rad) and all data were normalized to β-actin using the method of 2 -ΔΔCT .
Western Blot. Eyes were harvested and lenses removed with the remaining eye tissues homogenized in a lysis buffer containing protease inhibitors (Complete™ Lysis-M, Roche, Germany). Proteins were extracted as previously described 69,70 Equal amounts of protein were separated by 10% or 12% SDS-PAGE, followed by electroblotting onto polyvinylidene di uoride membranes (Amersham Biosciences, Amersham, UK). Following blocking with 5% nonfat dry milk for 1 hour at room temperature, membranes were incubated overnight at 4 0 C with primary antibody.
Binding of HRP-conjugated secondary antibody was performed for 1 hour at room temperature and bands were visualized using chemiluminescence (ECL; GE Healthcare, Chicago, IL). To verify equal loading, membranes were stained with anti-β-actin and band density was analyzed using Image J software (NIH, USA).
ELISA. Eyes were collected and homogenized in NP40 lysis buffer (ThermoFisher, Pittsburgh, PA) with 1 mM PMSF and protease inhibitor cocktail (Roche, Germany). The levels of VEGF-A, CCL5 and IL-6 in the tissue lysate were measured by using an ELISA assay kit (R&D Systems) according to the manufacturer's instructions.
Statistical Analysis. All data were expressed as means ± SEM, with n representing the number of mice used in each of the experimental groups. Statistical analyses were used to determine the signi cance of observed differences between treatment groups in all experiments. Statistical signi cance was calculated by means of two-tailed, unpaired, noparametric, Mann-Whitney test using GraphPad Prism 8 software. P values < 0.05 were considered to be signi cant.