Age-related macular degeneration (AMD) is a leading cause of blindness in people over 50 years of age in many developed countries. Drusen are yellowish extracellular deposits beneath retinal pigment epithelium (RPE) found in aging eyes and considered as a biomarker of AMD. However, the biogenesis of drusen has not been elucidated. We reported previously that multicellular spheroids of human RPE cells constructed a well-differentiated monolayer of RPE with a Bruch's membrane. We determined that RPE spheroids exhibited drusen formation between the RPE and Bruch's membrane with expression of many drusen-associated proteins, such as amyloid β and complement components, the expression of which was altered by a challenge with oxidative stress. Artificial lipofuscin-loaded RPE spheroids yielded drusen more frequently. In the current study, we showed that drusen originates from the RPE. This culture system is an attractive tool for use as an in vitro drusen model, which might help elucidate the biogenesis of drusen and the pathogenesis of related diseases, such as AMD.
Age-related macular degeneration (AMD) is one of the most common irreversible causes of severe visual loss in elderly individuals (Fine et al., 2000; Klein et al., 2003). Drusen, a hallmark of AMD, is a local basal deposit located between the basement membrane of the retinal pigment epithelium (RPE) and the inner collagenous layer of Bruch's membrane, a basal membrane complex of RPE (Abdelsalam et al., 1999). Despite a significant correlation between the presence of drusen and the incidence of AMD, the origin and biogenesis of drusen and its role in the development of AMD are not understood fully (Klein et al., 2004; Johnson et al., 2003). We reported recently that multicellular spheroids of RPE cells formed a well-differentiated monolayer of RPE with a Bruch's membrane (Sato et al., 2013). The current study showed that this three-dimensional culture system yielded drusen-like deposits with similar protein localization to that in aging human eyes, suggesting that the origin of hard drusen is the RPE cells. This spheroid culture system is useful as a novel in vitro drusen model to elucidate drusen biogenesis and the pathogenesis of AMD.
RESULTS AND DISCUSSION
Drusen formation on the surface of RPE spheroids
RPE cells, which are conventionally cultured on a flat plate, were suspended with F10-containing methylcellulose and 10% fetal calf serum (FCS) in a 96-well culture plate with a U-shaped bottom. The next day, a spheroid was constructed with the RPE cells aggregated in each well. Such a spheroid comprises an inner volume of apoptotic cells and an outer monolayer of RPE cells accompanying a Bruch's membrane on the outermost surface (Sato et al., 2013). In our previous study, the spheroids formed tight junctions and gap junctions, and showed most of the components of Bruch's membrane (the basement membrane of RPE, the elastic layer, and the inner and outer collagenous layers) on the outermost surface, and showed phagocytosis of inner apoptotic cellular components and outward deposition of lipoproteins, which are consistent with physiological functions of RPE (Sato et al., 2013). Therefore, it is assumed that RPE cells in this culture construct polarity at least to perform physiological functions. Thus, because the basal side of RPE cells had no plastic scaffolds like a Transwell membrane, this culture system may be advantageous to clarify the basal functions of the RPE as it allows the visualization of the behavior on the basal side. In fact, this system showed directly that RPE produced lipoproteins with apolipoprotein (apo) B-100, as suggested by Curcio and colleagues (Li et al., 2006). On the spheroidal surface, hyaline-like homogeneous deposits and clustered lipoproteins were observed between the RPE and the elastic layer of Bruch's membrane (Fig. 1). The deposits could be distinguished from apoptotic cells upon pretreatment of the RPE cells with fluoresceinated microspheres. In addition, our results showed that microspheres phagocytosed by RPE cells did not move into the drusen-like deposits, and this might support the evidence of previous reports showing differences in spectra lipofuscin-attributable autofluorescence from that of drusen and paucity of lipofuscin in macular drusen (Tong et al., 2016; Rudolf et al., 2008). The deposits involved substructural vesicles with a structure similar to human drusen. The live images suggested that drusen is formed by cytoplasmic budding. In contrast to human drusen, most of the deposits on the spheroidal surface were protruded outward, making the demarcated elastic layer curved. That may be because the spheroids possess an immature Bruch's membrane with insufficient support of collagenous layers, while RPE cells aggregate tightly into a spherical shape.
We used cells from passages 3 to 6 because human RPE cells (hRPECs) could be re-differentiated in this culture system independently of the passage number (Sato et al., 2013), and hRPECs from primary culture have varying amounts of native lipofuscin granules, which might unintentionally affect the experiment. However, as an alternative, stem cell-derived RPE cells, which potentially could possess a differentiated profile for RPE (Miyagishima et al., 2016), can be used to form spheroids (data not shown). Especially, spheroids of induced pluripotent stem cell-derived RPE cells prepared from patients with a specific disease may be useful to elucidate its pathogenesis.
Drusen formation induced by artificial lipofuscin loading
Lipofuscin accumulates in RPE cells after birth as indigestible byproducts derived from daily, physiological phagocytosis of photochemically oxidized outer segments of photoreceptors (Okubo et al., 1999). Thereafter, drusen formation is observed in some eyes of patients over 40 years of age and is associated with the future development of AMD (Abdelsalam et al., 1999). Therefore, the effects of lipofuscin loading were investigated. Native lipofuscin was eliminated promptly by repeated cellular division in the cellular culture (data not shown). Alternatively, glycoxidized nanospheres were fabricated as artificial lipofuscin. RPE cells were pretreated by adding glycoxidized nanospheres in the culture medium, simulating lipofuscin accumulation in the cytoplasm of RPE cells through phagocytosis. Spheroids made of artificial lipofuscin-loading RPE cells produced drusen and a basal laminar deposit-like material more frequently than spheroids without artificial lipofuscin loading (Fig. 2). In theory, basal laminar deposits and basal linear deposits with different localization should be identified by transmission electron microscopy. However, it is hard to preserve Bruch's membrane and deposits on the outermost surface of spheroids during the processing for transmission electron microscopy (e.g. wash-out, dehydration and embedding). Glycoxidized nanospheres have components that are different from native lipofuscin. Nevertheless, in our previous study, a rabbit AMD model that simulated lipofuscin accumulation by use of glycoxidized nanospheres produced not only choroidal neovascularization (CNV) and geographic atrophy but also drusen and abnormal fundus autofluorescence in specific patterns that were similar to human aging eyes with a high risk for AMD (Yasukawa et al., 2007; Hirata et al., 2009). In contrast, plastic beads could not induce AMD-associated markers except for RPE atrophy. We assume that glycoxidized nanospheres may be partially degradable, similar to lipofuscin and different from plastic beads, and that this system is useful at least to evaluate mechanical impacts of lipofuscin. Thus, lipofuscin accumulation likely is associated with drusen formation.
Expression of drusen-associated proteins on RPE spheroids
Recent studies have suggested that several molecules might be implicated in the pathogenesis of AMD including complement components, complement factors, amyloid β (Aβ), vitronectin and apoE. Polymorphisms in the genes coding complement factor H (CFH) and B (CFB) predict the risk of developing AMD (Scholl et al., 2007). Aβ, produced by the sequential proteolytic processing of amyloid precursor protein (APP), has been observed in human drusen in the senile plaques of Alzheimer disease (Anderson et al., 2004). It has also been reported that Aβ is bound to complement factor I (CFI) and blocks its ability to inactivate C3b, resulting in upregulated complement activation (Wang et al., 2008). Vitronectin, a fluid-phase regulator that binds the terminal component complex to inhibit cytolytic activity and the ability of the complex to bind to the cellular membrane, is also a marker of drusen (Hageman et al., 1999). Some studies have reported that the ε2 allele of the apoE gene increases the risk of AMD, while the ε4 allele is supposed to be protective (Edwards and Malek, 2007). Drusen and thickened Bruch's membranes have been reported to contain lipoproteins with apoproteins A-I, B-100, C-I, C-II and E (Li et al., 2006).
RPE spheroids expressed APP, CFH, vitronectin and apoE with localizations similar to those in human aging eyes (Fig. 3). APP was expressed in the cytoplasm of RPE cells. CFH was localized in the basolateral extracellular space. Vitronectin and apoE were distributed diffusely in deposits, drusen and Bruch's membrane.
Up- and down-regulation of drusen-associated proteins under stressed conditions
The pathogenic mechanisms whereby environmental factors contribute to the development of AMD remain elusive. However, epidemiologic studies indicate that cigarette smoke is the greatest environmental risk factor for dry and wet AMD (Smith et al., 1996). A previous study found that nicotine enhanced the size and severity of experimental CNV (Suner et al., 2004). Growing evidence indicates that oxidative stress may be responsible for AMD (Liang and Godley, 2004).
Oxidative stress in human eyes increases with age, and is derived from accumulated glycoxidized lipofuscin granules in RPE cells, accumulated oxidized lipoproteins beneath RPE cells, and exposure to high levels of oxygen from choroidal circulation and light exposure. However, it could be difficult to simulate these conditions experimentally. Nevertheless, production of reactive oxygen species would be expected to be a shared outcome of all of these conditions. On the basis of these reasons, spheroids were stressed with nicotine, cobalt chloride (CoCl2) and hydrogen peroxide (H2O2) to evaluate the effect of oxidative stress and nicotine. Quantitative RT-PCR analysis showed that C3, C5 and APP were upregulated, while apoE, vitronectin and CFH were downregulated in these conditions (Fig. 4). These results suggest that oxidative stress might disrupt the physiological status of the RPE and activate an inflammatory reaction with upregulation of complement components and APP.
RPE spheroids as an in vitro drusen model
Based on a complicated pathogenesis potentially involving aging, genetics, smoking, diet and hyperlipidemia, among others reasons, the morphological changes in Bruch's membrane and/or the morphological or metabolic changes in the RPE likely are observed prior to formation of drusen (Li et al., 2006; Okubo et al., 1999). These preconditions may lead to formation of drusen and diffuse basal deposits, which are biomarkers of AMD, although the mechanisms remain unclear (Abdelsalam et al., 1999). One major difficulty in elucidating the mechanism of AMD is the lack of optimal animal or culture models of AMD. To understand the mechanism of RPE-related diseases, such as AMD, it will be essential to clarify unknown basal functions of RPE involving possible transmembranous deposition of lipoproteins and maintenance of Bruch's membrane. We showed previously that this three-dimensional spheroid culture allowed visualization of the behavior of Bruch's membrane formation and lipoprotein deposition (Sato et al., 2013). In the current study, spheroids showed colocalization of drusen deposits and key drusen biomarker proteins, in which proinflammatory complement components and Aβ were upregulated and homeostatic components were downregulated with exposure to oxidative stress. The spheroid culture may be useful to elucidate the biogenesis of drusen, the pathogenesis of AMD, and the physiologic basal functions of the RPE.
Recently, a culture system using Transwell inserts has been reported to produce sub-RPE deposits that are comparable to those in human aging eyes (Pilgrim et al., 2017; Galloway et al., 2017). This Transwell culture system is advantageous for the construction a highly polarized monolayer of RPE cells with tight junctions (Sonoda et al., 2009), and basal deposits can also be observed in this system. Therefore, the Transwell culture system is ideal to investigate barrier functions of RPE as it allows measurement of transepithelial resistance, cellular polarity (by measuring cytokines secreted to the apical and basal sides) and basal deposits, by means of immunohistochemistry and electron microscopy. On the other hand, the 3D spheroid culture system, as used here, could be advantageous for the live imaging for basal deposits and relatively larger deposition including drusen formation, to assess the chain of functions from phagocytosis to lipoprotein deposition, and elucidate the morphogenesis of Bruch's membrane. However, it may be inadequate to assess polarity and barrier functions because it is difficult to measure transepithelial resistance and because the spheroids are usually used for the first 4 weeks, during which active phagocytosis of inner apoptotic cellular components is observable (Sato et al., 2013), while the Transwell culture requires 6 weeks for preconditioning to enhance polarity and tight junctions (Pilgrim et al., 2017; Galloway et al., 2017). Nevertheless, in the spheroid culture, increased expression of cytokeratin and reduced expression of smooth muscle actin shows that the RPE differentiate. Our previous study showed that the spheroids as used in this study formed tight junctions and gap junctions, suggesting that the cells established polarity (Sato et al., 2013). In addition, the spheroids show the majority of the components of the Bruch's membrane on the outermost surface, phagocytosis of inner apoptotic cellular components and outward deposition of lipoproteins, which are consistent with physiological functions of RPE (Sato et al., 2013). Spheroids of vascular endothelial cells have been reported to be highly differentiated, with expression of intercellular junction markers (Heiss et al., 2015). It is known that spheroids of cardiomyocytes start to show synchronized beating and improve their survival after implantation into the mouse heart (Hattori et al., 2010). Thus, spheroid culture be a highly useful method to more directly reproduce physiological conditions.
MATERIALS AND METHODS
The Ethics Committee of Leipzig University approved the use of human material. The study was performed according to the tenets of the Declaration of Helsinki. Human RPE cells (hRPECs) were obtained from several donors within 48 h of death and prepared as described previously (Enzmann et al., 1999). hRPECs were obtained from an eye bank. All tissue culture components and solutions were purchased from Gibco BRL (Paisley, UK). The cells were suspended in Ham F-10 medium containing 10% fetal calf serum (FCS), diagnostic medium (Glutamax II, Invitrogen, Karlsruhe, Germany) and 1% gentamicin, and cultured in tissue culture flasks (Greiner, Nürtingen, Germany) in 95% air/5% carbon dioxide at 37°C. The epithelial nature of the RPE cells was identified routinely by immunocytochemistry using the monoclonal antibodies AE1 (recognizing most of the acidic type I keratins) and AE3 (recognizing most of the basic type II keratins), both from Chemicon (Hofheim, Germany).
Spheroids of hRPECs were generated with minor modification according to the protocol described previously (Sato et al., 2013). Briefly, 4500 hRPECs were suspended in 150 µl of culture medium containing methylcellulose (2.4 mg/ml) (Sigma-Aldrich, Munich, Germany) in non-adherent 96-well culture plates with round bottoms (Nunc, Denmark). After 7 days of culture, the spheroids were harvested and, if necessary, cultured continuously with 15 ml of the same culture medium used with a normal RPE culture in non-adherent 10-cm culture dishes.
Preparation and loading of artificial lipofuscin granules into RPE cells
Glycoxidized nanospheres, to act as an artificial lipofuscin, were produced as described previously (Yasukawa et al., 2007). Briefly, 500 mg of albumin was dissolved in 1.8 ml of distilled water, and 0.2 ml of glycolaldehyde (123 mg/ml) was added. This mixture was added immediately to 10 ml of toluene with 2.8% ethylcellulose. The mixture then was stirred by a vortex mixer and ultrasonicated for 2 min. The resultant emulsions were added drop-wise to 10 ml of toluene with 2.8% ethylcellulose and mixed continuously using a magnetic stirrer for 4 days at room temperature, followed by counteraction for 1 h with 1 ml of glycine solution (1 M). The resultant glycoxidized nanospheres were washed three times each with toluene, isopropanol and distilled water. Fluorescein isothiocyanate (FITC) was conjugated to some glycox-nanospheres (NS). Briefly, after 250 mg of microspheres were rehydrated in 10 ml of carbonate-bicarbonated solution (pH 9.5), 10 mg of FITC dissolved in 10 ml of the same buffer was added to this suspension and the mixture kept at room temperature for 3 h. The resultant suspension was washed with a mixture of the same volumes of dimethylsulfoxide and distilled water four times and with distilled water three times. Finally, samples were washed with phosphate-buffered saline (PBS) three times, concentrated and stored at −80°C. Particles were confirmed to be ∼1 µm in diameter, similar to that of lipofuscin in human aging eyes.
Transmission electron microscopy
For transmission electron microscopy, the spheroids were harvested and fixed for 60 min in 0.2 M PBS with 4.0% paraformaldehyde and 2.5% glutaraldehyde. After washing, the spheroids were inserted into LUMiTainer (KG Lerche, Berlin, Germany). Then spheroids were postfixed in 1% osmium tetroxide in 0.1 M PBS for 1 h, dehydrated in graded ethanol series, embedded in epoxy resin (Nishin EM Co. Ltd., Tokyo, Japan), trimmed, and cut into ultrathin sections. The sections were stained with 2% uranyl acetate and lead citrate and were imaged on a JEM-2000EX TEM (JEOL Co. Ltd., Tokyo, Japan).
For immunohistochemistry, the spheroids were harvested at 2 weeks and fixed for 5 min in 0.2 M PBS (pH 7.8) containing 4% paraformaldehyde and for 15 min in PBS containing 4% paraformaldehyde. After fixation, the spheroids were washed five times with PBS and transferred back into PBS with 0.3% Triton-X and 5% normal goat serum at room temperature. After 60 min, the spheroids were incubated with the primary antibodies at 4°C overnight. The primary antibodies used were monoclonal mouse anti-APP antibody (1:100; cat. no. MAB348, Chemicon, Temecula, CA), polyclonal rabbit anti-elastin (1:200; cat. no. 324756, Calbiochem, Darmstadt, Germany), monoclonal mouse anti-vitronectin (1:200; cat. no. V7881, Sigma, Saint Louis, MO), monoclonal mouse anti-apoE (1:100; cat. no. A8599, Sigma), and monoclonal mouse anti-CFH (1:200, cat. no. A229, Quidel, San Diego, CA). As the second antibody, Cy2 (green channel)- or Cy3 (red channel)-conjugated IgG (Dianova, Hamburg, Germany) was applied for 2 h at room temperature. Nuclei were stained with 4′6-diamidino-2-phenylindole (DAPI) (blue channel). In addition, the sections were stained with aminoethyl-carbazole (AEC) for light microcopy, using the Pathostain ABC-POD (M) kit (Wako, Osaka, Japan). The lack of unspecific staining was confirmed by negative controls by omitting the primary antibody (data not shown). Images were recorded by confocal laser scanning microscopy (LMS 510 Meta, Zeiss, Jena, Germany).
Semi-quantitative and real-time RT-PCR
Spheroids incubated for 2 weeks were stimulated with 5 µM H2O2, 5 µM nicotine or 150 µM CoCl2 in Ham F-10 medium containing 0.5% FCS for 1 h. As a control, the spheroids incubated with Ham F-10 medium containing 0.5% FCS for 1 h were used. After stimulation, normal medium (F-10 medium with 10% FCS) was applied and spheroids were incubated for 24 h.
Using an RNeasy Mini Kit (Qiagen, Hilden, Germany), total RNA was extracted from the spheroids and treated with DNase I (Life Technologies, Darmstadt, Germany). cDNA was synthesized from 1 µg total RNA in a 20 µl reaction (200 U Superscript II RT, Life Technologies), 500 µM dNTP, and 0.5 µg oligo(dT)15. Real-time PCR using 2 µl cDNA was performed with a PCR detection system (Single-Color Real-Time PCR Detection System, BioRad, Munich, Germany) with the primer pairs described in Table S1. The PCR mix was denatured at 95°C for 6 min, followed by 40 cycles of melting at 95°C for 10 s, annealing at 58.5°C for 25 s, and elongation at 72°C for 25 s. Fluorescence changes were monitored after each cycle. mRNA expression was normalized to the levels of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) mRNA, and the changes were calculated as described previously (Pfaffl, 2001). To calculate the relative expression changes, we used the fluorescence signal of the lower cycles during the log phase of the product formation. Real-time PCR efficiency (E) was calculated according to the equation: E=10[−1/slope].
The authors thank Lynda Charter for professional medical English editing, and Tomomi Atsumi for technical assistance.
Conceptualization: A.N., T.Y.; Methodology: A.N., L.L., T.Y.; Validation: J.K., W.E., R.W., A. Kato, Y.Y., A.B., J.S., T.Y.; Formal analysis: A.N., L.L., T.Y.; Investigation: H.U., A.N., L.L., R.W., A. Kato, t.n., S.K., K.O., Y.Y., A. Kubota; Supervision: J.K., W.E., A.B., Y.O., J.S., P.W., T.Y.; Project administration: Y.O., P.W., T.Y.; Funding acquisition: Y.O., P.W., T.Y.
This investigation was supported by the Deutsche Forschungsgemeinschaft (grant WI880/9-1), and by a 2016 Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research (no. 16K11293) to T.Y.
The authors declare no competing or financial interests.