Comprehensive expression patterns of inflammatory cytokines in aqueous humor of patients with neovascular age-related macular degeneration

Neovascular age-related macular degeneration (nAMD) is a complex and multi-factorial disease, and low-grade inflammation is associated with pathogenesis of nAMD. Aqueous humor could reflect intraocular immune environments in various eye diseases. The research so far used aqueous humor samples and revealed that inflammation is involved in pathophysiology of nAMD, although immunological roles of cytokines were evaluated inadequately with aspect to individual effects. Here we used 27 kinds of cytokines covering general immunologic reactions, examined specific expression patterns of cytokines, and assessed relationships between inflammation and pathophysiology of nAMD by multivariate analyses. In nAMD eyes, principal component analysis showed that IL-7, MCP-1, MIP-1β and VEGF had high principal component loadings of over 0.6 in the first principal component constituting 32.6% of all variability of the data. In exploratory factor analysis, IL-6, MCP-1 and MIP-1β had high factor loadings (FL) of over 0.5 in Factor 1 constituting 32.6% of all variability, while VEGF had FL of over 1.0 in Factor 3 constituting 10.7% of all variability. In hierarchical cluster analysis, MCP-1 and VEGF were located in the cluster of first proximate mutual distance to central retinal thickness. These data could suggest that low-grade inflammation is a principal contributor in nAMD.

Aqueous humor levels of cytokines in nAMD patients and controls. The profiles of aqueous humor levels of cytokines in nAMD patients before initiation of anti-VEGF agents and in controls are shown in Table 2. Among all the cytokines measured, IL-6 level was significantly lower (P = 9.34 × 10 −5 , Cohen's d = 1.44) but IP-10 level was significantly higher (P = 4.80 × 10 −5 , Cohen's d = 0.71) in total nAMD patients compared to controls, while the other cytokine levels were not significantly different between 2 groups.
When logMAR VA, CRT and cytokine levels were compared among subtypes of nAMD and controls (Table 3), CRT was significantly higher (P = 0.007) in type 2 group than that in control group. The levels of IL-6 and IP-10 in all groups of nAMD subtypes were significantly lower and higher than those in control group. Furthermore, IL-7 level was significantly lower (P = 0.036) in PCV group than that in control group. When comparing among nAMD subtypes, logMAR VA and CRT were significantly higher in type 2 group than those in PCV group (P = 0.004 and P = 0.002, respectively), and logMAR VA was significant higher in type 2 group than that in type 1 group (P = 0.005).
When gender ratio (proportion of males) was compared among subtype groups of nAMD and control group, a significant difference was observed between type 1 nAMD group and control group (Table 3 and Supplementary Tables S1). Therefore, we conducted a sub-analysis to examine whether this difference affects the other results. When age, logMAR VA, CRT and cytokine levels stratified by gender were compared between type 1 nAMD group and control group (Supplementary Tables S1 and Table S2), IL-6 level was significantly lower (P = 0.001) in type 1-male group than that in control-female group, IL-8 level was significantly higher in type 1-female group than those in type 1-male group (P = 0.026) and control-female group (P = 0.037), and IP-10 level was significantly higher (P = 0.008) in type 1-male group than that in control-female group.
Binomial logistic regression analysis was performed to examine whether inflammatory cytokines were associated with the risk of developing nAMD (Fig. 1). IP-10 showed a positive association [odds ratio (OR); 1.030, 95% confidence interval (CI); 1.002-1.062, P = 0.033], and IL-7 showed a negative association (OR; 0.683, 95% CI; 0.507-0.921, P = 0.013) with the development of nAMD. On the other hand, there was no significant association between the remaining inflammatory cytokines and the pathogenesis of nAMD. Expression patterns of aqueous humor cytokines by exploratory factor analysis in nAMD patients. EFA was performed to identify the underlying relationship between the variable and the phenomenon 25 . In nAMD patients, the eigenvalue of Factor1 (F1) was 3.26, and F1 accounted for 32.6% of all variability in the data (Fig. 3). The eigenvalue of Factor 2 (F2) was 2.05, and F2 accounted for 20.5% of all variability in the data.   Table 2. Aqueous humor levels of cytokines in nAMD patients and controls. Cytokine values are given in units of pg/mL. Cytokine levels were compared between nAMD eyes and controls by Mann-Whitney U test. CRT; central retinal thickness, PDGF-BB; platelet derived growth factor BB, IL-1ra; IL-1 receptor antagonist, bFGF; basic fibroblast growth factor, G-CSF; granulocyte colony-stimulating factor, GM-CSF; granulocytemacrophage colony-stimulating factor, IP-10; interferon gamma-induced protein 10, MCP-1; monocyte chemoattractant protein-1, MIP-1α; macrophage inflammatory protein-1α, RANTES; regulated on activation, normal T-cell expressed and secreted, TNFα; tumor necrosis factor-α, VEGF; vascular endothelial growth factor. **P < 0.01.
www.nature.com/scientificreports www.nature.com/scientificreports/      Table 3. Comparisons of cytokine levels in aqueous humor among nAMD subtypes and controls. LogMAR VA, CRT and cytokine levels were compared among 4 groups by Kruskal-Wallis test followed by post-hoc Steel-Dwass test for each comparison. *P < 0.05, **P < 0.01. www.nature.com/scientificreports www.nature.com/scientificreports/ The eigenvalue of Factor 3 (F3) was 1.07, and F3 accounted for 10.7% of all variability in the data. The cumulative contribution rate of F1, F2 and F3 was 63.8%. In factor loading (FL) analysis of F1, IL-6, MCP-1 and MIP-1β had FLs of 0.5 or above. Especially, MCP-1 had FL of approximately 0.9, and was the cytokine with the strongest influence in F1. On the other hand, IL-12 had FL of -0.27, and was the cytokine with the weakest influence in F1. In FL analysis of F2, IL-7, IL-12, and IL-13 had FLs of 0.5 or above. In particular, IL-12 had FL of approximately 0.9, and was the cytokine with the strongest influence in F2. IL-6, IL-8, MCP-1 and VEGF had minus FLs, and MCP-1 with FL of -0.07 was the cytokine with the weakest influence in F2. In FL analysis of F3, VEGF had FL of over 1.0, and was the cytokine with extremely strong influence in F3, while IL-6 and IL-13 had low FL of approximate -0.15. www.nature.com/scientificreports www.nature.com/scientificreports/ Expression patterns of aqueous humor cytokines by hierarchical cluster analysis in nAMD patients and controls. Cluster analysis was performed to classify the inflammatory cytokines, logMAR VA and CRT into relative similar groups called clusters 26 . In nAMD patients, the factors were roughly classified into two principal groups as follows: a high cytokine expression group consisting of IP-10, MCP-1, VEGF and CRT; and a low cytokine expression group consisting of the remaining inflammatory cytokines and logMAR VA (Fig. 4). In the high cytokine expression group, MCP-1 and VEGF were included in the cluster of first proximate mutual distance to CRT, while IP-10 was located in the first proximate mutual distance to the cluster of MCP-1, VEGF and CRT. Regarding the distribution of subtypes of nAMD, type 1, type 2 and PCV groups were not separated clearly in the dendrogram (Fig. 4, vertical axis). In controls, the inflammatory cytokines were also roughly divided into two principal groups as follows: a high cytokine expression group consisting of IL-6, IP-10, MCP-1, VEGF and CRT; and a low cytokine expression group consisting of the remaining inflammatory cytokines and logMAR VA. In the high cytokine expression group, IP-10 was located in the first proximate mutual distance to CRT, while IL-6, MCP-1 and VEGF were included in the cluster of second proximate mutual distance to CRT.
Correlations among logMAR VA, central retinal thickness and levels of aqueous humor cytokines in nAMD patients. Table 4 show the correlations between the inflammatory cytokines and logMAR VA or CRT in nAMD patients. Multiple linear regression analysis was performed to examine the www.nature.com/scientificreports www.nature.com/scientificreports/ independent associations of inflammatory cytokines with logMAR VA or CRT 27 . In nAMD patients, logMAR VA correlated positively with CRT and levels of IL-13 and VEGF (P = 0.023, P = 0.048 and P = 0.017, respectively). CRT correlated positively with MCP-1 level (P = 0.021), and negatively with VEGF level (P = 0.028). When Spearman correlation analyses were conducted to examine correlation among individual cytokines, logMAR VA and CRT, logMAR VA correlated positively with IL-6 level (P = 0.037) in nAMD patients, but there was no significant correlation between CRT and the inflammatory cytokines.

Discussion
The major findings of the present study are as follows: (1) binomial logistic regression analysis, PCA and EFA revealed that low-grade inflammation is a principal contributor in the development of nAMD, and (2) in hierarchical cluster analysis and multiple regression analysis, MCP-1 was a prominent cytokine associated with CRT which is an essential clinical index of diagnosis and treatment in nAMD.
In this study, CRT and aqueous humor VEGF level did not differ significantly between nAMD patients and controls. Aqueous humor level of VEGF in untreated nAMD eyes varied among previous reports 9, 19,28,29 . Furthermore, previous studies have shown no correlation between CRT and aqueous humor VEGF level in untreated nAMD eyes 18,29,30 . Therefore, we suppose that VEGF-driven pathways regulating CNV may be a part of the complex processes in the development of nAMD, and other inflammatory mediators may play a crucial role in deviant angiogenesis 9,18,31 .
In this study, binomial logistic regression analysis identified IL-7 and IP-10 as significant factors for the development of nAMD. IP-10 is a chemokine that attracts type 1 T helper (Th1) cells, and activates Th1 cell-mediated immune response by binding C-X-C motif chemokine receptor 3 (CXCR3) 32,33 . Previous in vivo studies found that CC-chemokine receptor 3 and CXCR3 were associated with the development of nAMD 34,35 . In addition, IP-10 functions as an antiangiogenic and antifibrotic factor 36,37 . Bodnar et al. 38 reported that IP-10 inhibited VEGF-induced endothelial cell motility and tube formation in vitro. In postmortem eyes with early AMD, IP-10 was strongly expressed in neovascular endothelial cells and connective tissue matrix associated with CNV 39 . Therefore, it is conceivable that IP-10 primarily acts as an antiangiogenic factor to inhibit the development of CNV, and may secondarily induce Th1 cell-mediated immune response to produce low-grade inflammation in nAMD eyes. On the other hand, IL-7 plays a critical role in T cell development and peripheral homeostasis, especially, IL-7 is required for development and survival of regulatory T (Treg) cells 40,41 . Several papers have indicated significant involvement of para-inflammation in the pathophysiology of nAMD, participated by immune cells such as eosinophils, mast cells, macrophages and T cells 9,18,42,43 . Therefore, it is conceivable that IL-7 is a potent mediator of low-grade inflammation, a pathogenetic mechanism of nAMD, via both innate and acquired immune responses.
In this study, IL-6 level was markedly suppressed in nAMD patients compared with controls. IL-6 is both a pro-and an anti-inflammatory cytokine, and induces inflammatory and angiogenic cytokines such as VEGF 44,45 . Agawa et al. 9 reported that aqueous humor IL-6 level in nAMD patients tended to be low compared with that in cataract patients. Therefore, we suppose that down-expression of IL-6 in nAMD eyes may be a part of homeostatic response to suppress CNV development and vascular permeability, although some contradictory results have been reported 46,47 .
PCA is a statistical procedure that uses an orthogonal transformation to convert a set of many correlated variables into a set of fewer uncorrelated variables called principal components 24 . PC1 is the component with primary comprehensive influence on the subject. In this study, the PC1 of nAMD patients was composed of many inflammatory cytokines, and all of those cytokines had positive PCLs. Therefore, these results indicate that inflammatory cytokines have dominant influences on the pathophysiology of nAMD. Especially, IL-7, MCP-1, MIP-1β and VEGF showed high PCLs, suggesting that these inflammatory cytokines and VEGF are responsible www.nature.com/scientificreports www.nature.com/scientificreports/ factors involved in the pathophysiology of nAMD. MCP-1 is one of the key chemokines regulating migration and infiltration of monocytes/macrophages 48 . Monocytes/macrophages are innate immune cells that digest antigens and present them to other immune cells, enabling them to interact with the adaptive immune system 49 . Previous studies using human aqueous humor reported that MCP-1 is implicated in the pathogenesis of nAMD 9,19,46 . In addition, Lechner et al. 14 have suggested that plasma level of MCP-1 is elevated in nAMD, and monocytes may  www.nature.com/scientificreports www.nature.com/scientificreports/ contribute to the development of nAMD. On the other hand, expression of MIP-1β is increased in the blood of elderly individuals 50,51 . MIP-1β, MCP-1 and IP-10 as well as VEGF in the aqueous humor of nAMD eyes were significantly higher compared with controls 9 . Therefore, based on previous reports and our results, we could assume that inflammation is a principal contributor in the development of nAMD.
In the present study, it is notable that IL-6, IL-12 and VEGF were identified as significant variables with strong influences in the PC1 of nAMD patients, while these cytokines had PCLs of almost 0 in the PC1 of controls. IL-12 activates T cells and NK cells, resulting in the induction of Th1-related immune response [52][53][54] . In the clinical setting, IL-12 and IL-23 are the major etiologies of psoriasis, psoriatic arthritis and Crohn's disease 55 . Besides, recent studies indicate that IL-12 has anti-angiogenic effect, and acts as an angiogenic inhibitor in various stages of angiogenesis in vitro and in vivo 52,56 . Therefore, we presume that IL-12 may primarily play the role of anti-angiogenic factor to inhibit CNV, and secondarily induce Th1-related inflammation associated with the pathophysiology of nAMD. In controls, PCA identified IL-7, IL-8, IL-13, IP-10, MCP-1 and MIP-1β as the principal etiological elements of cataract. However, the cataract patients (controls) were otherwise normal elderly individuals. If the PC1 represents an intraocular immunological environment of the elderly people, the above-mentioned inflammatory cytokines could also be candidates of age-associated inflammatory elements in the elderly eyes.
In PCA, variables with positive PCLs are in conflict with those which have negative PCLs. In the PC2 of nAMD patients, the cytokine group consisting of IL-12 and IL-13 was located on the opposite side of another cytokine group consisting of IL-6, IL-8, IP-10 and MCP-1. In a large scale-cohort study, Ristau et al. 57 suggested that allergy has a protective effect on the development of AMD, although the mechanism remains unclear. Therefore, we would hypothesize that the innate immune responses are in conflict with the adaptive immune responses in nAMD eyes.
EFA is a statistical technique that reduces a large data to a smaller set of summary variables, and identifies underlying relationships between variables and the respondent 25 . In EFA, a factor implies the systematic summary component which is a part of the whole phenomenon, and the influence of the factor depends on the size of the eigenvalue, the same as the contribution rate of the factor. In this study, the F1 was principally composed of IL-6, MCP-1 and MIP-1β, but VEGF had almost no FL in the F1. Besides, the FL of IL-12 was a negative value. In terms of cytokine functions, the F1 may represent the summary component of innate immune response in the pathophysiology of nAMD. The F2 was composed of IL-12, IL-13 and IL-7, but the FL of MCP-1 was below -0.1. Therefore, the F2 may represent the summary component of adaptive immune response. The F3 was composed predominantly of VEGF. Therefore, the F3 would represent the summary component of angiogenic response. Based on the results of EFA, we propose that rather than angiogenic response, inflammatory immune response contributes predominantly in the pathophysiology of nAMD, because the cumulative contribution rate of the F1 and the F2 was 53.1% in the whole pathophysiology of nAMD, while the F3 accounted for only 10.7%.
Cluster analysis is an exploratory method used to classify objects or cases into relative similar groups called clusters 26 . In this study, the cluster related to BCVA was different from that related to CRT. The cluster of cytokines with high expression levels composed of IP-10, MCP-1, VEGF and CRT was common to both nAMD patients and controls. VEGF and MCP-1 are critical cytokines related to vascular permeability 58 . Therefore, VEGF and MCP-1 would be candidates of factors associated with macula edema in nAMD patients as well as elderly individuals. On the other hand, nAMD patients could not be clearly classified into the subtypes of nAMD on the basis of expression levels of inflammatory cytokines (Fig. 4, vertical axis). Therefore, we suppose that there is no significant difference in the specific expression patterns of inflammatory cytokines among the subtypes of nAMD.
Multiple linear regression analysis is a statistical approach used to describe the simultaneous associations of several variables with one continuous outcome 27 . In nAMD eyes, BCVA had a negative correlation with CRT, and MCP-1 was an exacerbation factor of CRT in multiple linear regression analysis. Furthermore, BCVA had a negative correlation with IL-6 level, and IL-6 level showed a positive correlation with MCP-1 level in Spearman's rank correlation. Jonas et al. 30 reported a positive correlation of MCP-1 level in the aqueous humor with macular thickness in nAMD eyes. On the other hand, VEGF is a key cytokine related to retinal vascular permeability, but previous papers reported no correlation between CRT and aqueous humor VEGF level 18,29,30 . In clinical research, the enrolled nAMD patients had various clinical stages, ranging from early to later clinical stages with macular atrophy 18,59 . Therefore, we suppose that intraocular VEGF levels will fluctuate depending on the clinical stages of nAMD eyes, if strict eligibility criteria for BCVA and macula lesions are not set in the enrollment of nAMD patients.
In exploratory cluster analysis and multiple linear regression analysis, MCP-1 was identified as a critical cytokine associated with CRT of nAMD eyes. This finding suggests that low-grade inflammation affecting macula thickness may be induced by innate immune response via macrophage/monocyte activities in nAMD eyes. Furthermore, BCVA could be improved indirectly depending on the improvement of CRT in nAMD eyes. In the future, combination therapy with anti-VEGF agents and steroids, immunosuppressive drugs and/or anti-TNFα antibodies would be useful to suppress pathological inflammation in nAMD eyes [60][61][62][63] .
The present study has several limitations. First, all of the nAMD patients and controls enrolled in this study were Japanese, although it is known that AMD has complex and multi-factorial etiologies including genetic factors, race, high-fat diet and color of the iris 4 . In the future, a large-scale international multicenter study could reveal various immunological features of nAMD depending on the differences in phenotype, genotype, race, as well as living environment and lifestyle. Second, the examination items were limited to 27 inflammatory cytokines, BCVA and CRT as clinical indices. Previous studies reported various nAMD-associated etiological factors other than those examined in our study, including lesion types and lesion size in the macula, insulin-like growth factor-1, angiogenin, monokine induced by interferon γ, soluble intercellular adhesion molecule 1, soluble vascular cell adhesion molecule 1, epithelial growth factor, human growth factor, intercellular adhesion molecule-1, IL-1a2, matrix metalloproteinase 9, plasminogen activator inhibitor 1, and C-reactive protein 9,19,46 . In the future, further multivariate analysis using more variables associated with AMD is needed to provide better understanding (2019) 9:19447 | https://doi.org/10.1038/s41598-019-55191-x www.nature.com/scientificreports www.nature.com/scientificreports/ of the pathophysiology in AMD. Third, there is a possibility that cytokine levels in aqueous humor are partially influences by cytokine levels in plasma, and are also affected by some unspecified inflammatory substances in the anterior chamber. Fourth, the cytokine levels in aqueous humor are not in full accordance with those in vitreous fluid that is in contact with the impaired retina in nAMD eye. Further studies using various biological samples such as plasma and vitreous fluid together with aqueous humor are warranted to confirm our preliminary results. Fifth, cataract patients were not perfect controls for nAMD patients in this study, because intraocular inflammation in the anterior chamber could contribute to cataract formation. However, performing limbal paracentesis on healthy individuals is not feasible from ethical viewpoint. Therefore, in interpreting the control data, potential increase of inflammatory substances that may influence the levels of aqueous inflammatory cytokines should be considered. Sixth, in analysis of nAMD subtypes, gender ratio was significantly higher in type 1 nAMD group than in control group. However, all multivariate analyses were performed using data of total nAMD patients, and there was no difference in gender ratio between total nAMD patients and controls. Hence, the difference in gender ratio of type 1 nAMD group should not affect the interpretation of results of multivariate analyses.
The strength of the present study was considered that the study population was relatively homogeneous. Fifteen (24.2%) of the enrolled nAMD patients had undergone cataract surgery, and the surgery was performed more than 6 months before this study. Furthermore, the enrolled nAMD patients had not been treated with regular anti-inflammatory eye drops including steroids for postoperative management after cataract surgery. Therefore, we suppose that the aqueous humor samples of nAMD eyes were collected under relatively unbiased conditions, and thus allowed precise evaluation of the specific expression patterns of inflammatory cytokines in nAMD eyes.
In conclusion, the present study indicates that low-grade inflammation via both innate and adaptive immune responses is a principal contributor in the development of nAMD. In the future, we will further examine predictive markers of response to anti-VEGF treatment in nAMD eyes, using comprehensive statistical analyses.

Methods
Subjects. This prospective observational study enrolled 62 eyes of 62 nAMD patients and 20 eyes of 20 cataract patients as controls. This study was performed at National Defense Medical College Hospital and Enoki Eye Clinic in Japan between September 1, 2013 and August 1, 2016. The study protocol was approved by the Ethics Committee of National Defense Medical College, and the procedures conformed to the tenets of the Declaration of Helsinki. Informed consent was obtained from all patients before enrolling in this study.
Inclusion criteria for nAMD patients in this study were as follows: (1) previously untreated choroidal neovascularization (CNV); (2) detection of intraretinal edema, subretinal fluid or pigment epithelial detachment by spectral-domain optical coherence tomography (SD-OCT); (3) absence of concurrent ocular diseases in the affected eye, which had compromised or could have compromised vision and ocular conditions. Exclusion criteria were as follows: (1) clinical features suggesting that CNV was secondary to other causes such as pathologic myopia, trauma and hereditary diseases; (2) myopia greater than -6 diopter, or axial length longer than 26 mm; (3) a history of treatments for nAMD including intravitreal drug injection, photodynamic therapy and systemic or topical steroids including sub-tenon injection; (4) previous intraocular surgery, except cataract surgery performed more than 6 months before the date of enrollment. Inclusion criteria for controls in this study were as follows: (1) no current or history of CNV, ocular trauma, diabetic retinopathy, retinal artery occlusion, retinal vein occlusion, ocular tumor, uveitis and other intraocular inflammations; (2) no remarkable drusen and geographic atrophy, (3) myopia less than -6 diopter, and axial length shorter than 26 mm; (4) any previous intraocular surgeries including scleral buckling; (5) previous external ocular surgeries performed within 6 months before the date of enrollment.
The nAMD patients enrolled in this study were different from the cataract patients enrolled as controls. No patient was recruited for both groups.
We performed a priori sample size calculation using the data of our previous study 18 . We found that for a statistical power of 0.80 64 , the sample size for each group required to detect significant differences in levels of aqueous humor cytokines was approximately 20. Therefore, we attempted to recruit around 20 cases each of type 1 nAMD, type 2 nAMD, PCV, RAP and control in this study.

Diagnostics and treatments. Diagnosis of nAMD was based on a full ophthalmological examination
including BCVA test using a decimal chart, intraocular pressure measurement, slit-lamp biomicroscopy, dilated fundus examination, color fundus photography, fundus fluorescein and indocyanine green angiographies, and SD-OCT (Cirrus HD-OCT; Carl Zeiss Meditec, Dublin, CA, USA). Neovascular AMD was classified according to the classification and diagnostic criteria of AMD into type 1 nAMD, type 2 nAMD, PCV, and RAP 2,22 . BCVA was converted to logMAR units (logMAR VA) for statistical analysis. For measurement of CRT, the ETDRS grid was employed 23 . In this study, CRT was defined as the mean retinal thicknesses of a central 1-mm circle on the ETDRS grid in the macula 65 . Aqueous humor sample collection and cytokine measurements. In nAMD group, before the first intravitreal injection of anti-VEGF agents, approximately 0.1 mL of undiluted aqueous humor was collected by performing an anterior chamber limbal paracentesis. In control group, undiluted aqueous humor samples were obtained at the beginning of cataract surgery. The aqueous humor samples were transferred into sterile tubes and stored at −80 °C until processing. No complication associated with the sampling of aqueous humor occurred. Twenty-seven cytokines in the aqueous humor samples were measured by a Bio-Plex multiplex assay (Bio-Plex Human Cytokine 27-plex panel; Bio-Rad, Hercules, CA, USA) and a multiplex bead analysis system (Bio-Plex Suspension Array System; Bio-Rad) according to manufacturers' instructions. All standards and samples were assayed in duplicate. Levels of aqueous humor cytokines below detectable levels were treated as 0 for statistical analysis 9,18 .