Background
Age-related macular degeneration (AMD) is a neurodegenerative disease that strikes the macula, causing irreversible blindness to people over the age of 50 in industrialized countries. The global prevalence of AMD has created substantial economic and social burden with a projected estimate of 196 million people living with AMD in 2020 [
1]. Clearly, better eye care strategies need to be designed to provide health care services to those in need. Clinically, AMD can be predicted by a severity scale that is a function of drusen deposition and pigment abnormalities [
2]. The accumulation of drusen both in size and in number has become a hallmark of AMD progression. As it progresses, AMD transitions from early benign stages into advanced vision-threatening stages, presenting with either choroidal neovascularization (CNV, wet form) and/or geographic atrophy (GA, dry form). Although CNV is the severe subtype of the advanced AMD, it is clinically managed using the anti-vascular endothelial growth factor (VEGF) therapy [
3]. However, there is no effective treatment to slow down the more prevalent dry form, which makes up approximately 90% of all AMD cases. Retinal pigment epithelium (RPE) cell death and secondary photoreceptor degeneration are two signature changes that lead to central vision loss in GA, the advanced stage of dry AMD. Hence, it is paramount to understand the fundamental mechanisms underlying these devastating impacts to the retina, especially the ones that undermine RPE health.
Despite the lack of consensus on the exact cell death pathway(s) involved, there have been three candidate cell death mechanisms proposed to underlie RPE atrophy in GA, including necrosis, apoptosis, and pyroptosis [
4‐
6]. Necrosis is a classic form of RPE cell death in GA, which was reported in earlier clinical-pathological studies by Sarks et al. [
4], and also in basic research projects with ultrastructural and histochemical data [
7]. Apoptotic RPE cell death, on the other hand, has gained substantial support from the literature in recent years. Using postmortem human eyes, Kaneko and colleagues identified the activation of caspase-3 in the RPE layer of GA eyes, but not in normal control eyes [
5]. Moreover, Dunaief et al. demonstrated statistically significant differences in the number of terminal deoxynucleotidyl transferase dUTP nick end-labeling (TUNEL) positive retinal cells in postmortem retinas with AMD, compared to normal controls [
8]. The third proposed mechanism of RPE death is pyroptosis, which is an inflammatory form of programmed cell death [
9]. The cornerstone of pyroptosis is the activation of an intracellular multi-protein complex named the inflammasome. NLR family pyrin domain containing 3 (NLRP3) inflammasome is the most widely studied machinery and consists of NLRP3, active caspase-1, and a bridging adaptor, apoptosis-associated speck-like protein containing a carboxy-terminal CARD (ASC). For the NLRP3 inflammasome to function, it requires sequential treatment of two types of pathological stimuli, (1) a priming signal to activate nuclear factor kappa B (NF-κB) and upregulate the transcription of NLRP3 and interleukin (IL)-1β precursor protein; (2) an activation signal to trigger the inflammasome assembly for the production of two pro-inflammatory cytokines, IL-1β and IL-18. The relationship between NLRP3 inflammasome and AMD pathology has been an attractive subject in the field, and current knowledge on this subject is reviewed elsewhere in detail [
10]. Although all three cell death pathways described above seems to govern RPE cell fate in GA to some extent, it is still unclear whether these mechanisms act independently or in synergy.
Earlier work, including ours, demonstrated that components found in drusen, (e.g.
, amyloid beta, complement cascade products) increase in the aged retina [
11,
12]. Previously, we have established a rat intraocular injection model to mimic the increasing amyloid beta (Aβ) load associated with drusen in human eyes [
13]. We demonstrated that drusen component, Aβ, triggers a short lasting pro-inflammatory response in RPE via the activation of NF-κB and NLRP3 inflammasome [
13,
14], which can be specifically abolished by vinpocetine (an NF-κB inhibitor) [
15]. Intriguingly, RPE cell loss is not a feature of the short lasting pro-inflammation associated with the acute Aβ intravitreal model [
13]. In the present study, we extended the duration of outer retina pro-inflammation by making sequential Aβ injections, in order to better model the chronic pro-inflammatory events and associated cell death underlying the pathogenesis of the dry form of AMD.
Methods
Preparation of oligomeric Aβ
The lyophilized, synthetic Aβ1–40 peptide (hereafter referred to as “Aβ”) in the HCl salt form was purchased from American Peptide (Sunnyvale, CA). We chose Aβ1–40 peptide over its structurally similar but more toxic, Alzheimer’s disease (AD)-specific, form of Aβ1–42 peptide based on earlier studies that demonstrated the presence of Aβ1–40 in drusen deposits in postmortem human eyes [
13]. Oligomeric Aβ was prepared according to published protocols [
16,
17]. Briefly, the synthetic Aβ peptide was first reconstituted in 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP, Sigma Aldrich, St. Louis, MO) and was evaporated by speed vacuum, resulting in thin transparent Aβ peptide film. The Aβ peptide film was then reconstituted in 100% dimethyl sulfoxide (DMSO, Sigma Aldrich) to a concentration of 1 mM, and further diluted in pre-warmed phosphate buffered saline (PBS, pH 7.4) to produce the Aβ injection solution of 323 μM, equivalent to 1.4 μg/μL. The injection solution was subsequently incubated at 37 °C for 48 h to form the oligomeric Aβ. Confirmation of successful oligomeric Aβ formation was achieved by western blot (WB) using our published protocols [
11]. Reverse Aβ40–1 peptide (hereafter, referred to as “control”) served as a sequence control for Aβ and was prepared in the same fashion. The absence of protein bands recognized by a mouse monoclonal anti-Aβ1–16 antibody (clone 6E10, Table
1) in the control solution indicated proper preparation (Additional file
1: Figure S1).
Table 1
List of primary antibodies
Amyloid-beta amino acid 1-16 (Aβ1–16) | Mouse monoclonal anti-Aβ1–16 (clone 6E10) | 1:2000 1:500 | BioLegend, Dedham, MA | Western blot Immunohistochemistry |
Caspase-1 | Rabbit monoclonal | 1:300 | Abcam, Cambridge, UK | Immunohistochemistry |
Phosphorylated NF-κB p65 (Ser 276) | Rabbit polyclonal | 1:75 | Santa Cruz Biotechnology, Dallas, TX | Immunohistochemistry |
Interleukin-18 (IL-18) | Rabbit polyclonal | 1:100 1:1000 | Santa Cruz Biotechnology, Dallas, TX | Immunohistochemistry Western blot |
Active caspase-3 (aCasp-3) | Rabbit monoclonal | 1:1000 | Cell Signaling Technology, Beverly, MA | Immunohistochemistry |
X-linked inhibitor of apoptosis (XIAP) | Mouse monoclonal | 1:1000 | BD Transduction Laboratories, San Jose, CA | Western blot |
Gasdermin D (GSDMD) | Mouse monoclonal | 1:100 | Santa Cruz Biotechnology, Dallas, TX | Western blot |
GAPDH | Mouse monoclonal | 1:10,000 | EMD Millipore, Billerica, MA | Western blot |
Animals
Adult female Long-Evans rats at the age of 4.5 month (Charles River, Wilmington, MA) were randomly divided into two groups. Group 1 (
N = 16) comprised rats receiving intravitreal injections of Aβ (5 μL at 1.4 μg/μL as previously published [
13]) once every 4 days for a total of three injections. Group 2 (
N = 16) rats received intravitreal injections of the control solution (reverse Aβ40–1 peptide) the same way as described for group 1. All rats were sacrificed on the 14th day after initial injection (day 14). Eyes were immediately enucleated and frozen for WB, polymerase chain reaction (PCR), and enzyme-linked immunosorbent assay (ELISA) or fixed in 4% paraformaldehyde diluted in Dulbecco’s PBS (Invitrogen, Carlsbad, CA) for 48–72 h prior to paraffin embedding.
Immunohistochemistry (caspase-1, IL-18, NF-κB, active caspase-3)
Paraffin-embedded rat eye tissues were processed following established protocols [
13]. Sections from both the Aβ and the control groups were processed simultaneously in an effort to minimize variability in immunoreactivity conditions (
N = 3 per group). Primary antibodies recognizing total caspase-1, IL-18, NF-κB, and active caspase-3 are described in Table
1. Non-specific isotype IgGs (Sigma Aldrich) matching the species of primary antibodies are used on negative control tissue sections. For visualization, the slides were developed using the Vector® AEC substrate kit (Vector Laboratories, Burlington ON, Canada) and were counterstained with Mayer’s Hematoxylin (Sigma Aldrich) for the nuclei. Sections processed simultaneously were analyzed and scored so as to avoid difference in immunostaining due to conditions such as temperature and stock of antibodies. Caspase-1, IL-18, and active caspase-3 immunoreactivity was scored in a masked fashion and semi-quantitatively based on a 0–3 point scale. A score of 0 indicates no detectable staining above the background level as compared to the negative control sections, whereas a score of 1, 2, or 3 suggests weak, intermediate, and robust intensity of the immunoreactivity, respectively. The immunoreactivity scores of caspase-1, IL-18, and active caspase-3 were averaged and normalized to the control group.
To detect NF-κB translocalization, an antibody recognizing the phosphorylated Ser 276 locus on NF-κB p65 subunit was used (Table
1). Immunoreactivity was measured quantitatively, in a masked fashion, using a × 60 objective lens and × 10 eyepieces. Positive RPE nuclei were identified as containing both the red AEC chromogen and blue hematoxylin counterstain, thus resulting in a purple appearance distinct from the unlabeled RPE nuclei that were blue in color due to the hematoxylin counterstain alone. The number of NF-κB positive nuclei was converted to percentage of all RPE nuclei in the sample area and was normalized to the control group.
Suspension array for vitreal cytokines
An ELISA-based cytokine assay for vitreal cytokines was carried out (Bio-Plex 200 System, Bio-Rad Laboratories, Hercules, CA). The assay targeted the following cytokines: erythropoietin (EPO), granulocyte colony stimulating factor (G-CSF), granulocyte macrophage colony stimulating factor (GM-CSF), chemokine (C-X-C motif) ligand 1 (GRO/KC), interferon-gamma (IFN-γ), IL-1α, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-10, IL-12p70, IL-13, IL-17, IL-18, macrophage colony stimulating factor (M-CSF), macrophage inflammatory protein 1alpha (MIP-1α), MIP-3α, regulated on activation, normal T cell expressed and secreted (RANTES), TNF-α, and vascular endothelial growth factor (VEGF). Vitreous from rat eyes in the same group (either Aβ or control) were pooled (
N = 7). Experiments were carried out following methods in our earlier publication [
15].
Western blot
To determine the level of X-linked inhibitor of apoptosis (XIAP), whole retina tissues (including neuroretina, RPE, Bruch’s membrane, and choroid) from each of the two injection groups were used (
N = 5). To investigate IL-18 secretion and gasdermin D (GSDMD) cleavage, RPE-choroid tissues from each of the two injection groups were dissected out and pooled to use (
N ≥ 3). Tissues were homogenized in ice-cold RIPA buffer (Thermo Fisher Scientific, Waltham, MA) containing protease inhibitor cocktail (Roche Diagnostics, Indianapolis, IN). Protein lysates were run under reducing conditions and established blotting procedures were followed. Detailed information on the primary antibodies used in western blotting can be found in Table
1 [
11]. As an internal protein loading control, GAPDH was detected either on the stripped membrane or using freshly thawed protein lysates (Table
1). The protein band intensity of XIAP (57 kDa), IL-18 (18 kDa), pro-GSDMD (53 kDa), N-GSDMD (30 kDa), and GAPDH (36 kDa) was individually measured using ImageJ (NIH, Bethesda, MD) and was converted into ratios relative to GAPDH. The final relative intensity of XIAP, IL-18, pro-GSDMD, or N-GSDMD was normalized to the control group.
Reverse transcription PCR (RT-PCR)
Total RNA of RPE-choroid was isolated from pooled rat eye tissues (
N ≥ 3) using ultRNA Column Purification kit (Applied Biological Materials, Richmond BC, Canada). 200 ng total RNA from each injection group was reverse transcribed into cDNA using the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, Carlsbad, CA). RT-PCR was carried out on the 7500 Fast Real-time PCR System (Applied Biosystems) using the following cycling conditions: 95 °C for 30 s, 50 °C for 30 s, 72 °C for 30 s, 40 cycles. RT-PCR primer sequences can be found in Table
2. Melting curve analysis was automatically performed right after the cycles’ completion. The results were expressed as mRNA fold-change relative to the control group after normalization to the reference gene, GAPDH, using the 2
−ΔΔCT method.
Table 2
List of primer sequences
X-linked inhibitor of apoptosis (XIAP) | CACACAGTCTACATCTCCT | CTACAACCTGTCCAGTTCT |
Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) | CTCTTGTGACAAAGTGGAC | CCATTTGATGTTAGCGGGA |
RPE morphological assessment
To evaluate the morphological changes of RPE cells due to induced inflammasome activity, we developed a custom Photoshop algorithm to measure the area of a set length of RPE monolayer based on RPE pigmentation. For each animal group, a total of nine sections from each animal were used for the analysis. All RPE micrographs were taken under × 60 magnification and subsequently cropped into 12 cm × 2 cm rectangular areas for further processing. Next, choroidal pigments were manually removed and the RPE-only areas were selected by applying “fuzziness”. Then, the pixel count for the cropped area was obtained. The RPE area measurement was expressed as number of pixels. This area measure was equivalent to, but more accurate than measuring the thickness of RPE monolayer.
RPE cell nuclei count and retinal thickness measurement
Following our established protocol, retinal cross sections within 200 μm distance from the optic disc were chosen for the analysis because of their uniform retinal thickness regardless of embedding orientation [
13,
18]. Briefly, RPE nuclei were counted by scanning the whole retinal section under × 20 magnification in 10
3 μm increments. Retinal thickness was measured from the inner limiting membrane to the photoreceptor outer segments/RPE junction. The mean values of RPE nuclei count per increment, retinal thickness as well as ONL thickness were averaged over a minimum of 4–6 retinal sections per animal at each time point.
Statistical analyses
Data are presented as mean ± SD. Non-parametric tests were used throughout the study except the vitreal cytokine levels and RT-PCR were analyzed by one-tailed Student’s t test. For the non-parametric comparisons between the two groups (Aβ vs control), one-tailed Mann-Whitney U tests were used. All analyses were conducted with GraphPad Prism version 6 (GraphPad Software, La Jolla, CA). Statistical significance was set at p ≤ 0.05.
Discussion
In late stage AMD, RPE cells die, yet the cell death pathways responsible remain a mystery. Based on pathological specimens of AMD donor eyes, atrophic RPE cells were thought to die from a necrotic cell death pathway, in which hypopigmented RPE cells filled with membrane-bound melanolipofuscin were eliminated, resulting in the increased pigmentation and cell body enlargement in adjacent RPE cells [
4]. More recently, reports from Kaneko et al. [
5] and Tarallo et al. [
6] highlight other cell death mechanisms (such as apoptosis and pyroptosis) that also likely contribute to atrophic RPE demise. However, the question remains as to what factors initiate the cell death cascades in AMD. In this study, using an animal model in which we mimicked pro-inflammation due to prolonged (14 days) exposure to a drusen component, Aβ, we assessed the level of involvement of pyroptotic and apoptotic pathways.
RPE inflammasome activation is a feature of the model
NLRP3 inflammasome activation, one of the fundamental innate immune defense mechanisms, has recently been studied for its role in the development of AMD [
22]. Our earlier study indicated that a single Aβ injection resulted in a peak in pro-inflammation at 4 days post injection, but then dramatically subsided [
13]. In the current study, we extended the duration of pro-inflammation from 4 to 14 days by making sequential injections of Aβ every 4 days to achieve better retinal penetration and mimic a chronic inflammatory microenvironment in the outer retina (Additional file
3: Figure S3). Our results demonstrated that longer exposure to pro-inflammation triggered robust NF-κB p65 subunit translocalization in RPE nuclei, elevated levels of total caspase-1 immunoreactivity, and enhanced secretion of IL-1β in vitreous and increased IL-18 presence in retina. Collectively, these results support inflammasome activity in the RPE (Fig.
3, Additional file
3: Figure S3). When compared with other inflammasome studies on RPE cells including ours, the current study demonstrated significant increase of IL-18 in neuroretina and RPE-choroid, but not in the vitreous where IL-1β was found upregulated by more than 50% (Figs.
1 and
3, Additional file
2: Figure S2). This secretion pattern is unique in that it diversifies the downstream biological events of these inflammasome cytokines: in our previous acute model of single Aβ injection, vitreal IL-18 increased by more than 3-fold at day 14 whereas vitreal IL-1β elevated by less than 50% compared to the controls [
13]. Such discrepancy in IL-1β/18 vitreal levels can be partially explained by the “distinct licensing requirements” for processing these two cytokines by NLRP3 inflammasomes, which might involve further regulation of caspase-11 and/or reactive oxygen species (ROS), as evident in immune cells [
23]. This mechanism is also very likely to be true in RPE cells since Aβ has recently been shown to induce NLRP3 inflammasome activation via NADPH oxidase- and mitochondria-dependent ROS pathway in vitro [
24]. Other studies suggest that a high level of retinal IL-18 (including RPE-choroid) is more related to the cell death events in GA [
6,
25], which support our findings that higher IL-18 immunoreactivity levels in Aβ-injected retina are correlated with the activation of pyroptotic pathway.
Sequential injections of Aβ also provide a good model to study Aβ’s role in inflammasome activation in the eye. Considered as one of the pathological hallmarks in AD, the deposition of Aβ in the AD brain is associated with elevated NLRP3 inflammasome activity, particularly the elevation of IL-1β production [
26‐
28]. Using microglia culture models, Halle et al. demonstrated the importance of NLRP3 inflammasome activation for the recruitment of microglia to Aβ deposits in the AD brain [
29]. Heneka et al. further discovered that in the absence of NLRP3 or caspase-1, mice carrying mutations associated with familial AD were largely protected from spatial memory loss, and demonstrated reduced IL-1β secretion and enhanced Aβ clearance [
27]. However, the involvement of these pathways has not been well established in the eye tissue. As a continuation from our previous studies [
13], our current Aβ multi-injection model recapitulates seminal features associated with each step of the NLRP3 inflammasome activation cascades. Therefore, by giving the animals sequential Aβ injections, we were able to sustain Aβ-induced pro-inflammatory responses in the neuroretina and RPE at a comparably high level until day 14, when the presence of Aβ was much diminished in the single Aβ injection model [
13].
Pyroptosis and apoptosis may contribute to RPE cell death in this model
The involvement of NLRP3 inflammasome activation in RPE has been studied in many different AMD models [
30‐
33]. However, once the inflammasome is activated, little is known of the exact biological events that occur and whether these events lead to cell death. Using rodent and non-human primate models, Doyle and colleagues demonstrate the efficacy of IL-18 treatment as a potential alternative, adjuvant therapy for CNV [
34‐
36]. On the other side, Ambati and collaborators showed evidence that IL-18 drives RPE degeneration after NLRP3 inflammasome activation in murine models [
6,
37]. In the current study, we also assessed the changes after activation of the NLRP3 inflammasome by Aβ. Intriguingly, after prolonged pro-inflammation in outer retina, we found enlarged or swollen RPE cells and significant increases in the proteolytic cleavage of full-length GSDMD in the RPE-choroid tissues. As reported by earlier studies, the cytolytic effects of pyroptosis are mediated by the oligomerization of the GSDMD’s N-terminal fragments (N-GSDMD) in cellular membrane, resulting in the cell-burst pore formation [
38]. Hence, it confirms the activation of pyroptotic pathway in the Aβ-injected animals [
19]. Therefore, we have demonstrated the co-existence of two events that occur after inflammasome activation: the secretion of mature pro-inflammatory cytokines, including the inflammasome products (IL-18 and IL-1β) and morphological and western blot evidence that supports the GSDMD-mediated pyroptotic pathway activation in RPE cells. Such an orchestrated response has also been seen in non-ocular cell types [
39].
Unlike pyroptosis, which rapidly lyses the cell, apoptosis is considered as a non-inflammatory and non-cytolytic form of programmed cell death. From a canonical point of view, pyroptosis is biochemically characterized as caspase-1 dependent and caspase-3 independent, whereas apoptosis is often caspase-3 dependent and caspase-1 independent. Interestingly, in the current study, we observed parallel cleavage of both caspase-1 and caspase-3 in the RPE tissue, challenging the dogma of their mutual exclusiveness. Although it is biologically impossible for one single cell to undergo both distinctive cell death pathways, it is still likely for one type of tissue, such as the RPE monolayer in the retina, to accommodate these pathways in a spatially discrete and stimulus dose-dependent manner. One study using murine bone marrow-derived macrophages exploited the potential of crosstalk between pyroptosis and apoptosis. The authors exhibited a DNA-dose dependent integral model for cell death, with apoptosis seen at a lower-dose of DNA stimulation and pyroptosis at higher doses. They further concluded that such an explicit response is regulated through caspase-8 activation [
40].
Despite the fact that our data supported the activation of both pyroptotic and apoptotic pathways in the RPE-choroid tissue of multiple Aβ-injected eyes, there were no significant anatomical changes as indicated by RPE nuclei counts and retinal thickness measurements (Fig.
8). All the biochemical markers used in this study (IL-18, IL-1β, NF-κB, caspase-1, caspase-3, GSDMD, XIAP) proved changes leading towards RPE atrophy via both cell death pathways. Combined with the anatomical data, our results suggest a transition between early and middle-to-late stage of dry AMD, where RPE cells begin to show early biochemical signs of cell death and before any observable retinal structural changes.
RPE-specific vs choroidal macrophage-mediated responses in vivo
The ability to dissect RPE-choroid from neuroretina was demonstrated in our earlier work [
11,
13] and those of others [
41]. However, it is possible that in the RPE-choroid sample, the choroidal component with macrophages, may also contribute to the inflammasome activity in our in vivo work. Testing for activated macrophages by immunohistochemistry in paraffin sections or by microdissection of choroidal tissue alone [
42] may help us better understand the migration of immune cells from choroid in our Aβ-injected animal groups. It is possible that the choroidal macrophages, important immune cells associated with AMD [
43,
44] in combination with RPE may both work to exacerbate the chronic inflammatory milieu in the AMD eye. Therefore, future in vitro work on cultured RPE cells and choroidal macrophages will allow us to define the role of each cell type without the confounding effects from the other.
Conclusions
In conclusion, we have demonstrated the activation of two distinct cell death pathways in RPE following prolonged pro-inflammation induced by drusen component, Aβ. For the first time, we show that GSDMD cleavage is associated with inflammasome activation in RPE, providing the molecular basis for pyroptosis participation in this model. Consistent with other studies using postmortem human AMD donor eye tissues [
6], our model recapitulates the key events in the NLRP3 inflammasome cascade. Furthermore, the presence of swollen, enlarged RPE cells in this model represents another prominent feature of RPE morphological changes during AMD progression [
10,
45]. Even though the retinal morphological analysis found no evidence of ultimate RPE cell loss or retinal thinning in this model, the biochemical events revealed here point towards a critical transitional stage that may further lead to the occurrence of RPE cell death. Understanding the detailed molecular mechanisms associated with this stage will benefit future endeavors in the search of therapeutic agents to slow down, or even prevent, RPE cell death in AMD.
Acknowledgements
The authors acknowledge the technical assistance provided by Alison Fong, Elliott To, Cyrus Thomas, Aikun Wang, and Matthew Wong.