Influence of Genetic Polymorphisms on the Short-Term Response to Ranibizumab in Patients With Neovascular Age-Related Macular Degeneration

Abstract

Purpose: To determine whether genetic risk single nucleotide polymorphisms (SNPs) for age-related macular degeneration (AMD) influence short-term response to intravitreal ranibizumab treatment.

Methods: Forty-four treatment-naive AMD patients were included in a prospective observational study. They underwent three monthly injections of intravitreal ranibizumab for neovascular AMD. After an initial clinical examination (baseline measurement), a follow-up visit was performed to determine treatment response one month after the third injection (treatment evaluation). Patients were evaluated based on ophthalmoscopy, fluorescein angiography, optical coherence tomography (OCT), and OCT angiography. Peripheral venous blood was collected for DNA analysis at baseline visit. Patients were genotyped for single-nucleotide polymorphisms within AMD-relevant genes and classified on good or poor responders based on visual acuity, central retinal thickness, intraretinal fluid, and subretinal fluid.

Results: One hundred ten AMD-associated SNPs have been analyzed. Six were found to be relevant when associated to ranibizumab treatment response. The genetic variants rs890293 (CYP2J2), rs11200638 (HTRA1), rs405509 (APOE), rs9513070 (FLT1), and rs8135665 (SLC16A8) predisposed patients to a good response, whereas rs3093077 (CRP) was associated with a poor response. FTL1, SLC16A8, and APOE were the SNPs that showed significance (P < 0.05) but did not pass Bonferroni correction.

Conclusions: This is the first study that links novel polymorphisms in genes such as CRP, SCL16A8, or CYP2J2 to treatment response to ranibizumab therapy. On the other hand, HTRA1, FLT1, and APOE are linked to a good ranibizumab response. These SNPs may be good candidates for short-term treatment response biomarkers in AMD patients. However, further studies will be necessary to confirm our findings.

Age-related macular degeneration (AMD) is the leading cause of irreversible visual impairment among people older than 65 years worldwide.¹ The neovascular form (nAMD) is characterized by severe vision loss caused by the growth of neo vessels under or within the macula.² Since the approval of anti-vascular endothelial growth factor (anti-VEGF) pharmacotherapy in 2006, the prevalence of legal blindness and visual impairment caused by nAMD has been considerably reduced, removing nAMD from the list of incurable diseases.³ The pathophysiology of nAMD is not fully understood; however, many studies have shown it to be multifactorial including genetic factors, metabolic factors, and environmental factors.^4,5 The rate of AMD heritability has been estimated at somewhere between 45% to 75%.⁶ The majority of the genes identified for AMD play several parts in susceptibility to disease and pathogenesis,⁷ although it should be noted that the potential association of these genes with treatment response is not as developed.^8,9 

Ranibizumab, a recombinant fragment of a monoclonal antibody with a high affinity for VEGF-A, was approved for the treatment of nAMD by the United States and the European Union, respectively, in 2006 and 2007 after the phase III clinical trials MARINA and ANCHOR.^10–13 Multiple treatment regimens are being used in clinical practice including from monthly dosing, to pro re nata or treat-and-extend strategies.¹⁴ Regardless of the specific intervention, even though most patients experience a reduction in symptoms after the first intravitreal injection, a pool of patients shows a poor or no response to treatment.^15,16 Pharmacogenetics could help us understand the response profiles to treatment by stratifying patients based on their individual genetic profile and optimizing therapeutic decisions to obtain the best efficacy possible and avoid adverse effects of drugs. 

PharmGKB, a platform that disseminates knowledge about clinically actionable gene-drug associations,¹⁷ has clinical annotations on five variants possibly associated with anti-VEGF treatment for AMD. Some of these genetic variants in genes such as CFB, VEGFA, and VEGFR1 suggest an inclination toward a proper response.⁹ On the other hand, variants of NRP1 and HTRA1/ARMS genes have been associated with poor or no response to anti-VEGF treatment,^18,19 but the evidence either is substandard or does not support this association because other studies failed to replicate it.^20,21 There is a lack of agreement in the pharmacogenetic studies of nAMD response to anti-VEGF treatment, because different periods have been studied such as one, three,^22,23 four,²⁴ five²⁵ or six²⁶ months when it comes to short-term response and 12 to 24 months to long-term response.^27,28 This makes it difficult for results comparison, particularly, in short-term analysis of the response. Short-term response is the optimal point to measure morphological and functional changes as maximum response is usually achieved at this point^29,30 and could help in better clinical management. 

Given the high prevalence of AMD and the associated high socioeconomic burden of the anti-VEGF therapies,³¹ it is of the utmost relevance to be able to predict treatment outcome. Therefore a key element for the implementation of individualized therapies is the early identification of poor and non-responders, aiming to maximize visual outcomes while minimizing the number of injections and their associated risks. The purpose of this study is to evaluate the pharmacogenetic relationship between the different genotypes found in the scientific literature of single nucleotide polymorphisms (SNPs) known to be associated with AMD, and response to treatment with ranibizumab for nAMD with the intention to find new SNPs that could predict short-term treatment response. 

Methods and Design

Characteristics of the Study Group

This was an observational, prospective study of naïve patients with neovascular AMD eligible to start treatment with ranibizumab according to clinical practice. Patients were recruited and evaluated by ophthalmologists from University Clinical Hospital of Santiago de Compostela, Galicia, Spain, from May 2019 to August 2021. After an initial clinical examination (baseline measurement), patients underwent a loading phase that consisted of three-monthly injections with ranibizumab (Lucentis; Novartis, Basel, Switzerland), and a follow-up visit was performed to determine treatment response one month after the third injection (treatment evaluation). Patients were evaluated based on ophthalmoscopy, fluorescein angiography, optical coherence tomography (OCT) (Dri-Triton; Topcon Corp Inc, Tokyo, Japan) and OCT angiography (SS-DRI-Triton-OCTA device; Topcon Corp Inc). This instrument has an A-scan rate of 100,000 scans per second, using a light source with a wavelength of 1 µm for a deeper penetration into tissue. All patients in this study were imaged using the 3 × 3 mm, 6 × 6 mm, and 9 mm scan patterns centered on the fovea. OCT angiography images were processed, and measures were analyzed using IMAGEnet 6. 

The study and the data collection strictly adhered to the declaration of Helsinki principles. Ethical approval was given by the local Institutional Review Board and the autonomous region of Galicia (2017/614). 

Patient's Criteria

Patients had to meet all the inclusion criteria, no exclusion criteria and had to give their consent for the collection and storage of blood samples. Inclusion criteria were defined as neovascularization secondary to AMD, age over 55 years, and visual acuity (VA) greater than 25 Early Treatment Diabetic Retinopathy Study (ETDRS) letters at baseline. Exclusion criteria included any retinal disease other than nAMD. In addition, patients who had previously undergone any ophthalmologic surgery other than cataract surgery or had signs of intraocular inflammation were also excluded. None of the subjects had received previous anti-VEGF treatment for nAMD. 

Measures of Response to Treatment

Treatment response analysis to ranibizumab therapy was performed four months after the initiation of ranibizumab treatment. The classification of this response is based on Amoaku et al.³² Patients with intraretinal fluid (IRF) or subretinal fluid (SRF) resolution, reduced central retinal thickening (CRT), and improvement of at least five ETDRS letters were considered good responders. Patients with less than 25% decrease from baseline on OCT-CRT, no change in VA (<5 letters), and continued or new IRF, SRF after anti-VEGF therapy were classified as poor or nonresponders. 

Genotype Determination

Peripheral venous blood was collected in two EDTA tubes (4 mL) at the baseline visit. The separation of leukocytes in blood was done within 24 hours. Five milliliters of blood were spun in a centrifuge at 2280g for 15 minutes at room temperature. After the leukocytes were separated from the other components of blood, they were washed twice with saline solution, and they were resuspended in 500 µL of saline solution. Afterward, the leukocytes samples were stored at −80°C until the DNA isolation. 

DNA was isolated following the manufacturer's instructions with the Wizard Genomic DNA Purification Kit (Promega Corporation, Madison, WI, USA). DNA concentration and purity were measured with Qubit 2.0 (Thermo Fisher Scientific, Waltham, MA, USA) and NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific). For the panel design, genotyping was performed using a custom made OpenArray technology using the QuantStudio 12K Flex System (Thermo Fisher Scientific), which is a real-time PCR system that uses Taqman technology. The results were analyzed using two software, the QuantStudio 12K Flex Real-Time PCR Software and Taqman Genotyper Software, which allows a cluster analysis by biomarker. 

A selection of DNA biomarkers has been carried out through the development of a bibliographic search based on scientific evidence so far on articles related to AMD and pharmacogenetics. Fifty-seven articles between the years 2004 to 2021 have been reviewed.³³ A total of 110 of AMD-associated single nucleotide polymorphisms have been analyzed (SNP selected in Supplementary Table S1). Biomarkers that emerged as significant in prior research and any biomarkers that leaned toward significance have been selected, despite not being particularly significant in the study. The chosen biomarkers belonged to both treatment response biomarkers and AMD susceptibility and pathogenesis biomarkers. 

Statistical Analysis

Pharmacogenetic data were analyzed with PASW Statistics 20 (IBM SPSS Statistics; IBM, Armonk, NY, USA) and R for Windows. Differences between good responder versus poor responder groups were explored using statistical tests; nonparametric Mann-Whitney test in the case of quantitative variables, and chi-square test or Fisher's exact test in the case of categorical variables. A logistic regression analysis was performed to detect the genetic association for each SNP and response to ranibizumab treatment using codominant, dominant, recessive, and log-additive genetic models, with sex used as covariate. Correction (for multiple tests) was performed via the Bonferroni method. 

Results

Cohort Description

This was an observational study where 44 naïve patients have been included. The demographic and response parameters of the study population are summarized in Table 1. The median age was 80 years old with 75% of the patients being over the age of 75 at the start of treatment with ranibizumab. 59% of the patients were women with a worse outcome than men in the short-term response to ranibizumab treatment (P = 0.014). 

Table 1.

 

Demographic and Clinical Parameters

 

Regarding the analysis of the response at the end of the loading phase, 52.3% of our series presented a complete response to treatment (gain ≥5 ETDRS letters and disappearance of fluid), 43.2% presented a partial response (gain <5 ETDRS letters or persistence of any type of fluid), and 4.5% did not present any type of response. Nonresponders and poor responders were pooled in a single group for analysis purposes. 

The median visual acuity was 60 ETDRS letters (interquartile range [IQR], 45–70) and the median CRT was 298 µ (IQR, 260–354) at the initial visit. In the OCT parameters: 61.7% of the eyes had IRF at diagnosis, 75% SRF. None of the patients presented areas of geographic atrophy as defined by the Classification of Atrophy Meetings at baseline.³⁴ 

At the first visit after completing the loading phase, the median VA was 66 ETDRS letters (IQR, 55–75) and the median CRT was 214 µm (IQR, 194–250). We found IRF in just 18.1 % of the cases, and SRF in 20.5%. No intraretinal hemorrhages were found in any of our patients. Regarding the improvement of VA, an overall improvement of 7 ± 13 letters has been quantified after the loading phase (Figs. 1, 2). 

Figure 1.

 

Macular neovascularization (MNV) without response to treatment. (A) Baseline image SRF can be observed over MNV. (B) Image after loading phase were SRF can be seen, along with an increase in the central subfield of the thickness of the ETDRS grid.

Figure 2.

 

Macular neovascularization (MNV) showing response to ranibizumab treatment. (A) Baseline image were IRF and SRF can be observed over MNV. (B) Image after loading phase without any type of fluid have can be seen, along with a recuperation of the normal thickness in all of the subfields of the ETDRS grid.

Good responders did not differ in VA from poor responders at baseline (P = 0.715), but we found significant differences in VA between groups after the loading phase (P = 0.029). These differences were not found between responders and poor responders in the baseline CRT, nor after treatment. 

Median CRT decreased significantly over time (P < 0.0001) in both responders (292 µm; IQR, 251–343 to 212 µm; IQR, 190–242) and poor responders (328 µm; IQR, 273–395 to 218 µm; IQR, 198–274). VA surged over time in both groups, but it was found to have statistical significance only in the responders’ group (P < 0.0001). 

Genetic Effectiveness Outcomes

A total of 110 of AMD-associated single nucleotide polymorphisms have been analyzed (Supplementary Table S1). No polymorphism was excluded because of noncompliance with the established quality control. All polymorphisms fulfilled the Hardy-Weinberg equilibrium. The genetic association between the selected SNPs with the different types of treatment response was explored (Supplementary Table S2). The SNPs with a significant association in the logistic regression analysis are shown in Table 2. The odds ratio (OR) reports the effect of the rare allele in the study (the frequent homozygote is always taken as reference). The rest of the SNP evaluated in this study showed no association with ranibizumab treatment (Supplementary Table S2). 

Table 2.

 

SNPs That Showed Significant Association With Response in Any of the Tested Models

 

Six SNPs were found to be relevant in their association to treatment response with ranibizumab. The SNPs that had better response to ranibizumab treatment were rs890293 of CYP2J2 (cytochrome P450 2J2) gene, rs11200638 of HTRA1 (human high temperature requirement serine protease A1) and rs8135665 of SLC16A8 (solute carrier family 16 member 8) gene. On the other hand, risk alleles of rs3093077 of C-reactive protein (CRP) and rs9513070 of FTL1 suggested worse response. However, these SNPs did not reach significance after the Bonferroni correction. 

The allele T (minority allele) of rs8135665 (SLC16A8) was associated with a risk of better response to treatment with ranibizumab after the loading dose in a dominant model. This SNP not only preserved significance, but it also increased once the model was adjusted for sex. 

No homozygotes for the rare allele (A) of rs890293 (CYP2J2) were found, and heterozygotes only appeared in the group of non-responders, so the C allele may be beneficial when patients were treated with ranibizumab. The allele G of rs405509 (APOE) was associated with improved ranibizumab treatment outcomes after the loading dose in a dominant model when the model was adjusted for sex. In the case of rs9513070 of FLT1 gene, the G allele is the risk allele and tends to be frequent in patients with a substandard response to treatment with ranibizumab therapy in a recessive model. 

The SNP that appeared to have worse response was rs3093077 of CRP gene where the C allele had higher frequency of heterozygotes in poor responders, but this SNP failed to display any association once the analysis was corrected for sex (P = 0.0856). It should be noted that the count was excessively low in the response group of this SNP. 

The frequent allele (G) of rs11200638; HTRA1, had the worst response to treatment because it appeared more frequently in the group of nonresponders (P = 0.0404; OR = 0.25; 95% CI, 0.06–0.98), but it also did not signify any association once the analysis was corrected for sex (P = 0.1318). 

Discussion

The main purpose of pharmacogenetics is the possibility of stratifying patients based on their individual genetic profile and to provide them with the most appropriate therapeutic decision to avoid lack of efficacy and adverse effects of drugs. Therefore early identification of nonresponders or poor responders is essential for the implementation of personalized therapies in the hope of maximizing visual outcomes and minimizing the number of treatments and associated risks. 

We have analyzed biomarkers related both to response and to the development of macular neovascularization. Although the former is more scarce than the latter in the literature, both provide new insights of treatment response. 

The SLC16A8 gene encodes a monocarboxylate transporter protein (MCT3) that is specifically expressed by the RPE and is involved in the transport of pyruvate, with lactate and ketones³⁵ mediating the level of acidity in the outer retinal segments and preventing the accumulation of lactate in the subretinal space. The rs8135665 variant is of interest because it is predicted to disrupt the processing of the encoded transcript³⁶ leading to defective lactate transport that may ultimately cause retinal acidification and photoreceptor dysfunction. Interestingly, SLC16A8 gene knock-out mice have changes in visual function,³⁷ but the overall retinal histology remains normal. 

SLC16A8 has been linked to susceptibility to AMD,^36,38,39 but no data have been found in the literature linking it to response to treatment. Ours is the first study to find an association between SLC16A8 and response to ranibizumab treatment, and the prevalence and trend of the data shown allows us to establish the importance of increasing the sample size to establish conclusively whether it is a protective factor of good response. 

CYP2J2 belongs to the family of cytochromes that are common in the heart and vascular endothelium.⁴⁰ There is little published on the role of CYP2J2 in nAMD. It has been observed that products from CYP2J2 metabolism, induced angiogenesis in vivo and ex vivo in mice, and inhibition of CYP2J2 may suppress pathological vessel growth in choroidal neovascularization.⁴¹ Liutkeviciene et al. aimed to determine the relationship between early neovascular AMD and the rs890293 polymorphism of the CYP2J2 gene in a Lithuanian population. The CYP2J2 TT genotype was statistically less frequent in patients >65 years with AMD than in controls, suggesting that polymorphism rs890293 TT could be a protective genotype.⁴² 

In our case, we found no homozygotes for the rare allele (A) and heterozygotes were only found in the group of non-responders, so the C allele could be beneficial when patients are treated with ranibizumab. However, it is true that the frequency of the TT genotype is very low according to the scientific literature and it is similar to that obtained in our population (4.7%). This rarity has also been observed in other pathologies such as Alzheimer's disease⁴⁰ and atherosclerosis.⁴³ Ours is also the first study to find association between CYP2J2 and response to ranibizumab treatment. 

FLT1 is a gene which encodes a member of the vascular endothelial growth factor receptor (VEGFR) family. This protein binds to VEGFR-A, VEGFR-B, and placental growth factor and plays an important role in angiogenesis.⁴⁴ There are several studies that relate different SNPs of this gene (rs7993418, rs9582036) with improved response in AMD.^9,44 In our study, these SNPs did not reach significance. The rs9513070 is the variant that holds significance in our case, where the G allele is the risk allele and is more common in patients with a substandard response to treatment with ranibizumab. However, the small sample size does not allow for a perfect interpretation of risk quantification because of the width of the confidence interval (OR = 17.35; 95% CI, 1.3-231.33). Nonetheless, these results are consistent with a retrospective study of a Belgian population treated with anti-VEGF where the A allele was also associated with improved response to short-term treatment (as in our case), but it failed to maintain association when evaluated again a year later.⁴⁴ 

The APOE polymorphic protein acts by revamping the homeostasis of cholesterol and other lipids. It is the major apolipoprotein of the central nervous system, and it has been shown to be a component of drusen, and can accumulate in the cytoplasm of RPE cells.⁴⁵ The rs405509 variant that was significant in our study was previously related to AMD risk, but it did not identify additional risks than the other SNPs related to this gene.⁴⁶ No other studies have been discovered in relation to AMD treatment response regarding this SNP. Alternatively, two SNPs, rs7412 and rs4229358, which did not reach statistical significance in our study, were found to confer significantly better visual outcomes after anti-VEGF at short-term response.^47,48 

Genetic variants of CRP have been shown to determine serum levels of this protein, particularly in response to inflammatory stimuli.⁴⁹ Several SNPs of CRP gene have been studied in relation to treatment response but related to photodynamic therapy (PDT).⁵⁰ Two variants related to this gene, rs2808635 and rs876538, did find an association between with response to PDT, but rs3093077 was not significative. No other studies have been found that relates to anti-VEGF treatment, being this the first one that explores this possible association with the rs3093077 variant. 

Finally, the HTRA1 gene encodes a heat shock serine protease that regulates TGF-β.⁵¹ The HTRA1 gene is one of the few genes that have been linked to a poor response to anti-VEGF treatment and is included in PharmGKB database,¹⁷ although still with a very low level of evidence (level 4) or association. Some studies show that the AA risk genotype is more frequent among nonresponders or poor responders.^22,52 However, in our case, the AA allele of rs11200638 is protective, with the most frequent allele (GG) having the worst response to treatment because it appears more frequently in the group of nonresponders. Our results are consistent with two studies, one with short-term response¹⁹ to ranibizumab treatment, and another in long-term response to bevacizumab.⁵³ Also, homozygotes (GG or AA) seem to have poorer response than individuals who were heterozygous for the risk A allele heterozygotes (GA).⁵⁴ We found no relation to other biomarkers known to affect treatment response such as VEGF, CFH or ARMS as detailed in PharmGKB probably because of the small sample size that was recruited for this study. 

The primary limitation of this study is undoubtedly the sample size. However, this limitation originates from the strict inclusion and exclusion criteria, along with the use of a sole center for recruitment. Likewise, this work only shows the early response associated to loading phase. Short-term response was chosen because previous studies revealed that the optimal point to measure morphological and functional response is following the loading phase where maximum response is seen.^11,29 Analyzing short-term response allows for an improved understanding of the visual potential, helping to manage patients' expectations and providing guidance on clinical management.³² On the other hand, with long-term follow-up and an individualized injection schedule, the genetic profile associated with the visual outcome may change, as has been seen in other studies.^9,25,55,44 

Overall, we have only analyzed the response to treatment of nAMD with ranibizumab as anti-VEGF agent as our goal was to be able to identify the patients who could benefit the most from this treatment. This grants an opportunity to further expand this area of research as other approved agents for nAMD also exert their effect in the VEGF pathway, such as brolucizumab, aflibercept or faricimab. Surveying the genetic profiles associated to response to varying agents could offer the opportunity to customize the treatment of patients. On the other hand, due to the multifactorial pathophysiology of the disease, the treatment response may be further be explored with studies that bridge the gap between genetic markers and clinical imaging biomarkers^56,57 more readily available to daily practice such as choroidal features such as choroidal thickness and perfusion index⁵⁸ or new OCT angiography biomarkers such as three-dimensional characterization⁵⁹ that might help in future clinical decisions. 

Conclusions

Our study provides further insight into the pharmacogenetics of the clinical response of nAMD to ranibizumab therapy, by identifying novel polymorphism in genes such as SCL16A8 or CYP2J2 that are associated with better outcome to treatment response. This is the first study that links CRP genotype, which previously was only studied in PDT, to anti-VEGF therapy obtaining a good response in homozygotes. FTL1 confirms prior published data regarding positive response, on the other hand, in this study HTRA1 risk allele was associated with a good ranibizumab response. Based on our results, SCL16A8, CYP2J2, FLT1, APOE, and CRP may be good candidates for treatment response biomarkers in AMD patients. However, further studies will be necessary to confirm our findings. 

Acknowledgments

Partially funded and supported by Instituto de Salud Carlos III-ISCIII (PI17/00940) and Xunta de Galicia (IN607D 2021/001). A.F.-F, C.M-G, L.G.-Q. are grateful to the Carlos III Health Institute for financing their personnel contracts: JR18/00014, JR20/00026, CM20/00024 personnel contracts. 

Disclosure: L. García-Quintanilla, None; P. Almuiña-Varela, None; O. Maroñas, None; A. Gil-Rodriguez, None; M.J. Rodríguez-Cid, None; M. Gil-Martinez, None; M.J. Abraldes, None; F. Gómez-Ulla de Irazazabal, None; M. González-Barcia, None; C. Mondelo-Garcia, None; R. Cruz, None; A. Estany-Gestal, None; M. Fernández-Rodríguez, None; A. Fernández-Ferreiro, None