Associations of IL-23 with Polypoidal Choroidal Vasculopathy

Abstract

Purpose.: To determine the relationship between the levels of intraocular inflammatory cytokines and polypoidal choroidal vasculopathy (PCV).

Methods.: Prospective cohort study. Sixty-two patients with PCV and 36 control subjects were studied. The levels of cytokines, chemokines, and growth factors in the aqueous humor samples from PCV patients and control subjects were assessed for significant associations with PCV. Logistic regression analysis was used to compute the odds ratios (ORs) and 95% confidence intervals (CIs) after the study populations were divided into quartiles.

Results.: In PCV patients, IL-4, IL-10, and IL-23 were significantly higher than in the controls. Logistic analyses showed a significantly high risk for IL-23 (OR for the highest quartile compared with the lowest quartile: 16.3; 95% CI: 3.5–75.2), VEGF (5.7; 1.2–26.1), and IL-4 (4.0; 1.3–12.7). IL-10 and IL-4, but not IL-23, were significantly correlated with the VEGF levels in PCV patients (IL-10: ρ = 0.477, IL-4: ρ = 0.281). The elevated levels of IL-5, IL-10, IL-4, IL-23, and IL-1α were found to be significantly associated with exudative lesion(s) in the fluorescein angiograms.

Conclusions.: The significant associations between elevated levels of IL-23 with PCV and its activity strongly suggest an involvement of inflammatory processes in the etiology of PCV, presumably independent of VEGF. (www.umin.ac.jp/ctr number, UMIN000003854.)

Introduction

Age-related macular degeneration (AMD), including neovascular AMD (nAMD) and polypoidal choroidal vasculopathy (PCV), is the leading cause of severe visual dysfunction in the elderly population worldwide. In the Asian and African populations, PCV is the more prevalent form of AMD. ¹ PCV has a distinct morphology, course of progression, and responsiveness to photodynamic therapy. ^1–3 PCV is characterized by abnormal vascular networks, with polypoidal lesions and serosanguineous pigment epithelial detachments (PEDs). ⁴ Most of the components of the lesions are located under the RPE, and leakage from the polypoidal lesions forms the PEDs. ⁵ The pathological processes of PCV are similar to those of nAMD, and, thus, PCV is considered to be one form of AMD. 

Genetic analyses of AMD and PCV cases have shown that they share polymorphisms in the HtrA serine peptidase 1, age-related maculopathy susceptibility 2, and complement factor H. ^{6 –8} In contrast, the elastin polymorphism is PCV-specific. ⁹  

In addition to the polygenic nature and the complexity of AMD-related diseases, the severity and characteristics of AMD can be greatly affected by environmental and epigenetic factors. ^10,11 Thus, to better understand the pathogenesis of AMD, the genetic background and the environment-responding factors must be considered. An effective method to do this is to examine the mediators or cytokines secreted by eyes with PCV. 

For eyes with AMD-related diseases, an elevation of VEGF has been found in the vitreous and aqueous humor. ^12,13 VEGF is produced by retinal pigment epithelial cells, vascular endothelial cells, and glial cells, ^14,15 and it appears to be the major stimulator of neovascular growth and vascular permeability. Because of the relationship between VEGF and neovascular growth, intravitreal anti-VEGF therapies have become standard therapy for eyes with nAMD and PCV. ^{1,16 –18} The very successful outcomes indicate that VEGF is a significant risk factor for AMD; however, we have very limited information on the real odds ratio (OR) for the risk of nAMD or PCV provided by the VEGF levels. In addition, the elevation of VEGF in AMD patients is sometimes not significant, most likely because of the limited statistical power of methodology and the contribution of other factors. ¹⁹ Thus, cooperation of VEGF and VEGF-independent factors may better explain the disease pathogenicity. 

Cytokine IL-8 was reported to be associated with nAMD by analyses of the single nucleotide polymorphisms or by its expression in choroidal neovascularizations (CNVs). ^20,21 IL-6, on the other hand, was shown to be associated with nAMD by analyses of the serum cytokines. ²² In addition, systemic elevations of C-reactive protein (CRP), homocysteine, and fibrinogen were reported to be significantly associated with the presence of nAMD. ^22,23 When the aqueous humor of eyes with nAMD was examined by multiple cytokine screening, only VEGF and platelet-derived growth factor (PDGF)-AA among the different inflammatory cytokines were significantly associated with nAMD, and the other cytokines were not significantly elevated. ¹² PDGF was also shown to be elevated in eyes with PCV. ¹³ Collectively, we have evidence that AMD has an inflammation-related aspect; however, the contributions of the cytokines to the pathogenicity of AMD remain undetermined.  

To learn more about pathogenesis of AMD, we focused on one representative form of AMD in Asians. Thus, the purpose of this study was to obtain a profile of the different inflammatory cytokines expressed in eyes with PCV, and to determine their association(s) and to identify new disease marker(s). Logistic regression analyses were used to show that IL-23, in conjunction with VEGF, was significantly associated with PCV. 

Methods

Subjects: Eligibility Criteria and Diagnosis

The study protocol was approved by the Ethics Committee of Tottori University, and the procedures used conformed to the tenets of the Declaration of Helsinki. An informed consent was obtained from all of the participants after an explanation of the procedures to be used. Clinical trial registration ID is UMIN 000003854. 

The diagnosis of PCV was made by the presence of definitive PCV lesions in the fluorescein angiographic, indocyanine green angiographic, and spectral domain optical coherence tomographic (3D-OCT-1000 MARK II; Topcon, Tokyo, Japan) images that met the guidelines of the Japanese Study Group. ²⁴ The diagnostic findings were the characteristic polypoidal lesions seen by indocyanine green angiography and the presence of subretinal orange nodules by fundus examination. 

The inclusion criteria were PCV eyes with macular edema or subretinal hemorrhage and/or PED involving the fovea. The exclusion criteria were PCV eyes with laser photocoagulation and photodynamic therapy within the past 6 months, and intraocular surgery including cataract surgery in the study eye within the past 6 months. The percentages of patients with a history of hypertension, diabetes mellitus, or smoking were 51.6%, 11.3%, and 11.3%, respectively. Male sex was 80.6%. 

The PCV lesions are usually seen as nonexudative lesions in the fluorescein angiograms (FAs); however, a certain percentage of PCV eyes have exudative PCV lesions. ^25,26 The components of the exudative type arise from thin vessels of the polypoidal lesions or from fibrinous exudation associated with leaking polyps. ^4,25,26 Because the exudative nature of PCV lesions can affect the visual outcome and appears to reflect the disease activity, we classified the PCVs into exudative types and nonexudative types depending on the leakage of the lesions in the FAs. Forty-two eyes (68%) were classified as the nonexudative type, and 20 eyes (32%) were the exudative type as determined by FA. 

The central retinal thickness was measured by OCT. All of the clinical parameters were assessed and statistically analyzed. The patients then received the anti-VEGF treatment, that is, an intravitreal injection of 1.25 mg bevacizumab (Avastin, Genentech and Hoffmann La Roche, Basel, Switzerland), 0.5 mg of ranibizumab (Lucentis; Novartis, Basel, Switzerland), or 0.3 mg of pegatanib (Pfizer, New York, NY). On the day of the initial treatment, 100 to 200 μL of aqueous humor was collected by paracentesis before the injection of the anti-VEGF agent and stored frozen for cytokine measurements. 

Control Group

Reference samples were obtained from 36 age-matched eyes of 36 patients who were scheduled to undergo routine cataract extraction. The exclusion criteria were any type of retinal diseases, uveitis, previous intraocular surgery, use of immunosuppressive drugs, or malignant tumors. Aqueous humor samples were collected immediately before the cataract surgery by anterior chamber paracentesis and stored frozen. The percentages of patients with a history of hypertension, diabetes mellitus, or smoking were 47.3%, 2.8%, and 8.3%, respectively. These incidences did not differ significantly from that in the PCV patients. Male sex was 47.2%. 

Measurement of Cytokines, Chemokines, and Growth Factors

Samples were analyzed using a chemiluminescence-based ELISA for high sensitivity detection. Q-Plex human cytokine-screen (Quansys Biosciences, West Logan, UT) was used to measure the concentrations of IL-1α, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-8, IL-10, IL-12, IL-13, IL-15, IL-17, IL-23, TNF-α, lymphotoxin-α, and IFN-γ in the aqueous samples. The levels of monocyte chemotactic protein 1 (MCP-1), VEGF, and eotaxin-2 were measured with ELISA kits (PeproTech, Rocky Hill, NJ; R&D Systems, Minneapolis, MN; Antigenix America, Franklin Square, NY, respectively). Aqueous humor samples were diluted 5-fold using the diluent according to the manufacturers' instructions, and incubated on the capture antibody-coated plates overnight at 4°C and processed for ELISA detection. 

Statistical Analyses

Data are presented as the means ± SEMs. 

Logistic regression analysis was carried out to compute the ORs and 95% confidence intervals (CIs) based on quartiles of each cytokine level. Quartiles were calculated as the three points that divide all the sampled population into four equal numbers of groups after data points were sorted in ascending order. ORs were calculated, and adjusted for age and sex. Each cytokine quartile was compared with the lowest quartile as the reference category. To evaluate the differences between groups, unpaired t-tests or Mann-Whitney U tests were used to determine their significance. Spearman correlation analysis was used to determine the coefficient of correlation (ρ) between factors. P less than 0.05 was considered significant. For multiple comparisons, Bonferroni correction was applied to the calculated P values. 

Results

Sixty-two eyes of 62 PCV patients scheduled to receive intravitreal anti-VEGF therapy were studied. Thirty-six eyes of 36 cataract cases were studied as controls after obtaining informed consent. The mean age was 74.5 ± 1.1 years in the PCV group and 74.1 ± 1.7 years in the control patients (P > 0.05). 

To screen for disease markers of PCV, we first determined the level of candidate molecules in the aqueous fluid of the PCV patients, including the inflammatory cytokines, chemokines, and growth factors. Concentrations of the different factors in the aqueous humor are shown in Table 1. In the PCV patients, IL-4, IL-10, and IL-23 were significantly elevated from that of the controls after Bonferroni correction for multiple comparisons was used (Table 1, Fig. 1). 

Table 1.  

 

Concentrations of Cytokines, Chemokines, and Growth Factors in Aqueous Humor of PCV Patients and Control Subjects

	(pg/mL)	IL-1α	IL-1β	IL-2	IL-4	IL-5	IL-6	IL-8	IL-10	IL-12	IL-13	IL-15	IL-17	IL-23	IFN-γ	TNF-α	TNF-β	MCP-1	VEGF	Eotaxin-2
PCV	Mean	14.8	13.6	4.5	1.7	4.9	26.9	5.5	8.0	6.8	8.9	10.8	5.7	76.0	1.6	9.3	10.2	384.2	55.0	5
	Median	18.0	11.7	3.5	1.0	2.4	10.4	3.7	5.0	4.1	5.3	6.7	5.0	78.0	1.1	5.0	11.1	321.3	59.1	5.1
	SEM	1.6	1.9	0.6	0.2	1.0	11.5	1.4	0.9	1.8	1.6	2.1	0.5	9.5	0.3	2.8	1.1	43.2	5.3	0.6
	IQR	1.3–25.0	0.1–19.1	0.4–7.0	0–3.1	0–8.2	4.7–14.4	2.4–5.7	1.4–15.2	0–8.1	4.1–10.2	4.0–12.6	3.8–7.2	19.4–119.4	0.4–1.5	2.5–6.2	2.9–14.5	211.0–442.3	22.8–84.2	0.7–8.3
Control	Mean	7.5	6.2	2.3	0.6	2.1	12.3	3.4	2.9	4.9	8.2	11.4	5.7	31.5	1.7	14.7	9.1	290.2	33.4	2.8
	Median	3.4	2.0	1.9	0.0	1.1	11.0	2.9	2.1	2.9	4.9	5.5	6.5	15.4	0.4	3.0	3.7	234.5	14.0	0.8
	SEM	1.6	1.5	0.4	0.1	0.5	1.8	0.5	0.5	1.0	1.6	2.1	0.5	8.8	0.5	5.3	1.8	33.8	6.7	0.9
	IQR	0–14.9	0–9.1	0.5–3.3	0–1.0	0–2.8	5.0–15.9	0.8–5.3	1.2–2.9	0–8.5	3.3–10.8	3.2–13.7	4.0–7.5	13.3–25.6	0–2.1	1.6–6.0	2.6–13.9	130.8–376.0	1.9–68.8	0.4–2.4
P value		0.003	0.007	0.009	0.0002	0.04	NS	NS	<0.0001	NS	NS	NS	NS	0.001	NS	NS	NS	NS	0.01	0.04
P value (Bonferroni correction)		NS	NS	NS	0.004	NS	NS	NS	<0.005	NS	NS	NS	NS	0.02	NS	NS	NS	NS	NS	NS

 IQR, interquartile range.

Figure 1. 

 

Distinct profiles of inflammatory cytokines, chemokines, and growth factors in PCV. Fold increases in the concentrations in control subjects are shown. IL-4, IL-10, and IL-23 were significantly elevated. *P < 0.05, **P < 0.005 after Bonferroni correction for multiple comparison.

To determine the significance of the association of these cytokines, we used logistic regression analyses to evaluate their disease-associated risk after adjustment for age and sex. 

We determined the association of significantly elevated cytokines (i.e., IL-4, IL-10, and IL-23) with the disease (Table 1, Fig. 1). Of these, IL-23 showed a significant risk with a high OR of 16.3 for the highest quartile (P = 0.0003; Table 2). The highest induced cytokine was IL-4 (Fig. 1), which had a significant risk with an OR of 4.0 for the highest quartile (P = 0.018). IL-10 was not statistically associated (P = 0.05). 

Table 2.  

 

Association of IL-23, IL-4, and VEGF in Aqueous Humor with Polypoidal Choroidal Vasculopathy by Logistic Regression Analysis

	Quartile of Cytokines (pg/mL)	Age- and Sex-Adjusted Odds Ratio	95% Confidence Interval	P Value
IL-23	1 (≤15.1)	1.0	Reference
	2 (>15.1–24.0 )	2.5	1.5–4.2
	3 (>24.0–108.9)	6.4	2.3–17.8
	4 (>108.9)	16.3	3.5–75.2	0.0003
IL-4	1 (ND)	1.0	Reference
	2 (−0.8)	1.6	1.1–2.3
	3 (>0.8–2.5)	2.5	1.2–5.4
	4 (>2.5)	4.0	1.3–12.7	0.018
VEGF	1 (ND)	1.0	Reference
	2 (−41.1)	1.8	1.1–3.0
	3 (>41.1–71.6)	3.2	1.2–8.8
	4 (>71.6)	5.7	1.2–26.1	0.025

 ND, not detected.

We next assessed the association of VEGF because it is a well-recognized target molecule for the treatment of nAMD and PCV. ^1,16–18 VEGF was elevated in PCV patients over that of the controls (Table 1). For reference, we assessed whether VEGF may indeed be a disease-associated marker of PCV. Our results showed that the VEGF level was significantly associated with PCV (Table 2). The OR for VEGF as risk for disease was 5.7 for the highest quartile after adjustments for age and sex. 

Because VEGF is the canonical disease-associated factor, we next assessed whether the induced cytokines or other cytokines reflected the VEGF-related pathological arm. To classify each cytokine into independent disease-associated or secondary to VEGF induction, we assessed whether the levels of the significantly elevated cytokines, including IL-4, IL-10, and IL-23 (Fig. 1, Table 2), were correlated with that of VEGF. We found significant correlations between VEGF and IL-10 and IL-4 (correlation coefficient by Spearman correlation analysis: 0.477, P < 0.0005; 0.281, P < 0.05, respectively). In addition, levels of IL-10 and IL-4 were significantly correlated with a high correlation coefficient of 0.80 (P < 0.0001). Thus, IL-10 and IL-4 may have relationship with the VEGF-related pathological process of PCV. IL-23 was not significantly correlated with VEGF (correlation coefficient: 0.24, P = 0.16); however, it was correlated with IL-10 (correlation coefficient: 0.53, P < 0.001). This suggests that IL-23 may well be placed in indirect relationship with the VEGF-dependent pathological process. 

Next, we evaluated how each cytokine is associated with the different characteristics of PCV. First, PCV patients were categorized by the disease activity. PCV patients were classified based on the presence of leakage from the polypoidal lesions in the FAs. ^4,25,26 PCV types with positive fluorescence leakage were categorized as exudative, and PCV types without leakage were categorized as nonexudative. Twenty eyes (32%) were classified as exudative and 42 eyes (68%) were classified as nonexudative. The mean visual acuity was 0.56 ± 0.11 logMAR units (logarithm of the minimum angle of resolution) in the exudative type of eyes and 0.48 ± 0.06 logMAR units in the nonexudative type eyes (P > 0.05). A representative exudative type PCV is shown in Figure 2. 

Figure 2. 

 

Fundus and OCT findings of representative PCV case with classic lesions. (A) A 69-year-old man with exudative-type PCV lesion with reddish-orange nodule (arrow) with serous pigment epithelial detachment. (B) Indocyanine green angiography shows polypoidal lesions (arrowhead). (C) The tomographic image of the lesion is shown in a horizontal scan (green line) by OCT. Visual acuity was 1.0 logMAR units. Aqueous levels of IL-1α, IL-4, IL-5, IL-10, and IL-23 were elevated (29 pg/mL, 4 pg/mL, 8 pg/mL, 20 pg/mL, and 208 pg/mL, respectively).

In the exudative type of PCV, IL-5, IL-10, IL-4, IL-23, and IL-1α were significantly elevated in a descendent order of induction ratio to the control (Bonferroni corrected P value for multiple comparison, calculated from Fig. 3). Thus, the Th2-type cytokines, including IL-5, IL-10, and IL-4, were associated with the disease activity and may be markers for a risk for prognosis. 

Figure 3. 

 

Profile of inflammatory cytokines, chemokines, and growth factors in PCV relative to the activity of the lesions. In the exudative type of PCV determined by fluorescence angiography, IL-1α, IL-4, IL-5, IL-10, and IL-23 were significantly elevated. *P < 0.05, **P < 0.005 after Bonferroni correction for multiple comparison.

We then analyzed the clinical parameters, including the height of the PED and macular edema. The mean PED height was higher in the exudative-type group (291 ± 87 μm) than in the nonexudative-type group (144 ± 35 μm), but the difference was not significant (P = 0.068). Macula edema was found in 27% of the patients with PCV. When cytokines were assessed for association with the absence or presence of PED or macula edema, none of the cytokines were significantly elevated after the Bonferroni correction was used. 

Discussion

Currently, AMD is considered to be an inflammation-related condition with a complex balance of inflammatory cytokines that are associated with its pathogenesis. In addition, atherosclerosis-related vascular diseases, such as retinal vein occlusion, also have inflammatory components that may significantly determine its prognosis. ²⁷ In agreement with these findings, we found significantly higher elevations of IL-23, IL-4, and IL-10 in the aqueous humor of PCV patients. 

Of these, our results provided strong evidence that there was a significant association between levels of IL-23 in patients with PCV. IL-23 is essential for recruitment and activation of inflammatory cells that are required for the induction of chronic inflammation or granuloma formation. It was previously unrecognized for its significant association with AMD or PCV. IL-23 is also a pro-angiogenic cytokine and increases angiogenesis or promotes inflammatory responses, such as an upregulation of the matrix metalloproteinase MMP9. ²⁸ Very recently, the genes for the IL-23 receptor and IL-10 were identified by genome-wide screening analysis (GWA) as the susceptible gene for Behcet's disease. ^29,30  

The involvement of the immune-mediated factors in AMD was initially revealed by genetic analyses. GWA has been successfully applied to AMD, and the results showed that complement factor H was a significant risk factor for AMD. ³¹ This outcome is recognized as the most successful of numerous GWA studies applied to common diseases. ¹¹ Subsequent GWA studies showed that polymorphisms of factor B and complement component 2 genes were also risk factors for AMD. ³² These genes are key components of the complement-mediated inflammatory cascade, which is supposed to be operative in lesions of the choroid, retinal pigment epithelium, and drusen. 

There has been accumulating evidence that the complement activation cascade positively or negatively regulates inflammatory functions. Recently, significant association between the complement system and IL-23 has been reported. In the airway hyperresponsiveness model, C3a was shown to promote IL-23 production by the dendritic cells and Th17 responses, which led to more severe disease, ³³ whereas complement activation product C5a limits frequency of Th17 cells. In the endotoxic shock model, C5a was shown to exert anti-inflammatory effects via induction of IL-10, which inhibits production of IL-23 and IL-17A. ³⁴ In experimental autoimmune encephalomyelitis, complement activation stimulates Th17 cell differentiation via C5a receptor and reciprocally reduces IL-23. ³⁵ Thus, IL-23 was positioned for intricate regulation by the C3a and C5a system, and appears especially stimulated by C3a. 

In the complement cascade system, acute-phase proteins were also shown to be involved in the pathogenesis of AMD and atherosclerotic diseases. For example, CRP is a systemic clinical marker for inflammatory diseases, including atherosclerosis, ^36,37 and also a marker of PCV and AMD. ^22,23 CRP is also known to be involved in the alternative complement cascade. ³⁸ Other acute-phase proteins, including IL-6 and amyloid A, are known to be readily elevated systemically under inflammatory conditions. In addition, the drusen in aged retinas contains a wide spectrum of amyloid proteins, including soluble amyloid proteins. ³⁹ Importantly, the soluble forms of the amyloids are able to strongly stimulate IL-23 production rapidly and continuously even at very low concentrations. ⁴⁰ This may occur at the sites of chronic inflammation, and this will activate the innate immune system. IL-23 is a heterodimer of p19 and p40 subunits, and p40 is shared by IL-12. We observed a specific elevation of IL-23 but not IL-12 (Fig. 1). This distinct pattern of elevation is consistent with the amyloid-induced IL-23 responses, which also does not induce IL-12. ⁴⁰  

The highest amount of cytokine elevation in PCV patients was found for MCP-1; however, it was not significantly associated with PCV by logistic regression. Studies of animal models indicate a disease-exacerbating role of MCP-1. In the murine laser-induced CNV model, MCP-1 and VEGF were shown be significantly associated with CNV. ⁴¹ MCP-1 appears to play some role in disease processes; however, it may not serve as a useful disease marker. 

The recent finding of an involvement of the CCR3 cascade in AMD has revealed important contributions of the VEGF-independent arm in the pathology of AMD. ⁴² This arm was shown to be more potent than VEGF in the murine laser-induced CNV model. The CCR3 ligands, eotaxins, are signature chemokines of the Th2-type responses. However, eotaxin-2 elevation was not significant after Bonferroni correction. Instead, we found that IL-4 had significant ORs, and IL-5 level was significantly elevated in the active exudative type of PCV (Fig. 3). Notably, IL-4 and IL-5 are the canonical Th2-type cytokines. They may represent the Th2 arm of this disease. Considering this possibility, CCR3 antagonism may still have a potential therapeutic efficacy, as suggested by Takeda et al. ⁴² This would be achieved by CCR3 antagonists, as we reported for ocular allergy. ^43,44  

Another Th2 lymphocyte-related cytokine, IL-10 generally has an anti-inflammatory role. IL-10 generally exhibits anti-angiogenic activity for microvascular endothelial cells ⁴⁵ ; however, an opposite role of IL-10 is also known. When IL-10–deficient mice were analyzed for CNVs, the CNVs were significantly reduced. In the murine CNV model, IL-10 was shown to regulate anti-angiogenic macrophage activity and promote angiogenesis. ⁴⁶ Although we observed a high correlation of VEGF with IL-10, the actual roles of IL-10 need further investigation. 

Another obvious potential therapeutic target is IL-23. We identified IL-23 as a disease marker and found a significant OR of 16.3, which exceeded the OR for VEGF. Whether an elevation of IL-23 is the cause or counter-regulator for the disease remains unsolved, and the potential role of IL-23 as a therapeutic target for AMD is undetermined. Targeting strategies for IL-23 are a real potential and are under clinical trials for the treatment of psoriasis and Crohn's disease. ^47,48  

To summarize, our data clearly show that IL-23 is significantly associated with PCV. This association is not directly recognized for VEGF. Our findings on the association of VEGF independent arm(s) may shed new light on the etiology of AMD. Considering the urgent need for better therapy to complement or substitute for anti-VEGF and photodynamic therapy, our new findings may help open new therapeutic avenues for treating this potentially blinding eye disease. 

Footnotes

 Supported by Grant-in-Aid 20592076 and 23791984 for Scientific Research from the Japanese Ministry of Education, Science, and Culture.

Footnotes

 Disclosure: S. Sasaki, None; D. Miyazaki, None; K. Miyake, None; Y. Terasaka, None; S. Kaneda, None; Y. Ikeda, None; T. Funakoshi, None; T. Baba, None; A. Yamasaki, None; Y. Inoue, None