1. Association of Genetic Polymorphisms with Age-Related Macular Degeneration Clinical and Visual Phenotypes and Response to Treatment
AMD Pharmacogenetics and Metabolomics: Moving towards Personalized Medicine
Milam Brantley, MD, PhD, Vanderbilt Eye Institute
American Geriatrics Society Annual Meeting - 4 May 2013

2. Disclosures I have no financial disclosures.
I will discuss the off-label use of bevacizumab.

3. Age-related macular degeneration
Disease of central retina
Blind spots, metamorphopsia
Slow progression or rapid change
Reading, TV, driving,
recognizing faces

6. Wet
Dry

7. Can dry AMD become wet?
Dry
Advanced
Early
Intermediate
Wet

8. Treatment - intravitreal injections
Anti-Vascular Endothelial Growth Factor (anti-VEGF)
Multiple monthly treatments
Can slow vision loss and even improve vision
bevacizumab
ranibizumab
aflibercept

9. Jahnigen supported research

10. AMD Studies Genotype-phenotype associations Pharmacogenetics
Oxidative stress quantification
Metabolomics

11. Science March 2005 PNAS May 2005 Science 2005;308:419-21
Proc Natl Acad Sci 2005;102:

12. Complement Factor H Y402H Chromosome 1q32
Nucleotide change (1277): T to C
Controls AMD TT TC CC
OR for AMD TT
TC OR
CC OR
Edwards, Science 2005;308:421-4
Amino acid change:
Tyrosine to Histidine
Y402H

13. AMD-associated Genes Replicated in multiple studies
CFH
ARMS2/HTRA1
C2/BF C3
Developing evidence
CFI
TIMP3
LIPC
Variable evidence
VEGFA
SERPING1 13

14. AMD genetics & genotype-phenotype associations

15. 153 patients with NVAMD, looked at CFH Y402H
185 eyes with gradable lesions
Classified lesions as Predominantly Classic or Occult
Examined the % of PC lesions among genotypes
CC – 47% PC
TC – 37% PC
TT – 17% PC OR=2.01 for each C allele
Concluded C allele associated with PC in this cohort
Am J Ophthalmol 2007;144:

16. Address conflicting literature
556 NVAMD cases, 256 controls
3 SERPING1 tSNPs
Address conflicting literature
Examined both allele frequency and genotypes for SNPs
Br J Ophthalmol 2010;94:

17. SERPING1 Conclusions rs2511989 negatively associated with AMD
rs positively associated with AMD

18. AMD Pharmacogenetics

19. 86 pts rec’d bevacizumab monotherapy
Treated as needed to inactivate lesion
Followed for an average of >9 months
Ophthalmology 2007;114: 19

20. Clinical phenotypes of AMD patients by CFH genotype
Risk alleles 1 2
Characteristic TT TC CC
p value
Pre-treatment VA
20/248
20/206
0.86
Post-treatment VA
20/166
20/170
20/341
0.016
Eyes with improved VA (%)
5 (50.0)
31 (54.4)
2 (10.5)
0.004 * *
* Patients with CFH CC genotype did significantly worse with bevacizumab treatment
Brantley et al., Ophthalmology 2007;114:

21. Conclusion – Bevacizumab study
Patients with CFH CC genotype faired worse with intravitreal bevacizumab

22. 156 pts rec’d ranibizumab monotherapy
Treated as needed to inactivate lesion
Followed at least 9 months
Br J Ophthalmol 2009;93: 22

23. VA outcomes and number of ranibizumab injections by CFH Y402H genotypes
Risk alleles 1 2 TT TC CC
p value
Initial VA, mean logMAR (95% CI)
0.85
( )
0.89
( )
0.83
( )
Post-treatment VA at 6 months, mean logMAR (95% CI)
0.71
( )
0.73
( )
0.62
( )
0.38
Post-treatment VA at 9 months, mean logMAR (95% CI)
0.75
( )
0.78
( )
0.70
Mean no of injections in 9 months (95% CI)
3.3
( )
3.8
( )
3.9
( )
0.09
Lee et al., Br J Ophthalmol 2009;93:

24. Recurrent event analysis of interval between injections by CFH Y402H variant genotypesTC CC
Hazard ratio (compared with TT genotype)
1.25
1.37
95% CI
0.94 to 1.67
1.01 to 1.87
p value
0.12
0.04
Lee et al., Br J Ophthalmol 2009;93:

25. Conclusions – ranibizumab study
CFH CC genotype more likely to need ranibizumab injections

26. Oxidative Stress and AMD

27. Quest for AMD biomarkers
Working hypothesis
Certain biochemicals could identify those at risk for AMD
Thiol redox
Lipid peroxidation

28. Methods
Blood drawn from 152 participants
77 AMD pts
75 controls

29. Plasma biomarker levels in AMD pts v. controls
Cases
(n =)
Con.
(mean ± SD)
Controls
Unadjusted p-value
Adjusted p- value
T-test
Wilcoxon Rank Sum
Cys (μM) 69 68
4.19 ± 1.45
4.11 ± 1.39
0.770
0.835
0.632
CySS (μM)
55.27 ± 11.21
50.66 ± 10.27
0.013
0.025
0.108
GSH (μM) 67
1.77 ± 0.65
1.89 ± 0.97
0.396
0.492
0.167
IsoP (ng/ml) 72
0.051 ± 0.034
0.059 ± 0.064
0.372
0.850
0.345
IsoF (ng/ml)
0.22 ± 0.38
0.15 ± 0.18
0.115
0.087
0.194
Brantley et al., Am J Ophthalmol 2012;153:

30. Conclusions
CySS higher in AMD pts compared to controls prior to age adjustment

31. AMD Metabolomics

32. Moving toward Personalized Medicine in Ophthalmology
Long-term goal: To develop personalized treatment regimens for patients with degenerative retinal diseases
Approach: Determine metabolic and genetic profiles associated with disease and treatment response

33. Metabolic Profiling
Metabolome – snapshot of individual’s physiological state
Downstream of the genome and the proteome
Liquid Chromatography-Mass Spectrometry (LC-MS)
unique
metabolites

34. MWAS
In a metabolome-wide association study (MWAS) of the 26 NVAMD patients and 19 controls, 94 unique metabolic features were identified that significantly differed between the two groups using FDR (q=0.05) This is a Manhattan plot of –log p for each metabolite expressed as a function of the m/z, with the significantly different features appearing above the broken line.
94 individual metabolites significantly different between AMD patients and controls (FDR = 0.05)

35. We performed correlation analysis on the 40 most consistently discriminant features. Seventeen features formed four correlation clusters, indicating that these metabolites move together and suggesting that these features may have a biological basis for association. We searched the Metlin metabolomics database for potential matches to these features. Interesting matches included a series di- and tripeptides which were elevated in NVAMD and bile acids and vitamin D metabolites which had lower levels in NVAMD.

36. Metabolites Associate with ARMS2 Genotypes
T - risk allele
To investigate metabologenetic relationships, we combined the metabolomic analysis with genotype data for the ARMS2 rs SNP. Using metabolomic data for the 26 NVAMD patients and 19 controls, we used OPLS-DA to identify features that contribute to separation of the ARMS2 GG genotype (no risk alleles) from the ARMS2 GT+TT genotype (1-2 risk alleles). We then overlaid the 94 FDR significant features that discriminated NVAMD from controls. This figure shows the loading plot showing the 26 metabolites that were associated with either ARMS2 GG (n=18, left) or ARMS2 GT+TT (n=8, right).
Ovals represent metabolites that discriminate between the ARMS2 GG and GT+TT genotypes

37. Hierarchical Clustering Analysis (HCA) 44 AMD patients and 29 Controls
16 m/z features
We then used Hierarchical Clustering Analysis (HCA) to show the relationship between participants and the 16 discriminatory metabolites. The analysis generated 14 clusters of individuals distributed into two major groupings: Group A, made up of 58% controls, and Group B, consisting of 78% AMD patients. Note Cluster 14 in Group A and Cluster 13 in Group B, both consist entirely of AMD patients, but are well separated by the cluster analysis.

38. Principal Component Analysis
All 73 patients
AMD Clusters 13 and 14
We performed principal component analysis of all 73 subjects using the 16 significant m/z features. This showed partial separation of controls and AMD patients. Clusters 13 and 14 from the HCA, all AMD patients, were largely separated from each other. These clusters were selected to investigate whether the metabolic profiles could be used to subclassify AMD patients. PCA of subjects present in Clusters 13 (n=8) and 14 (n=6) showed complete separation.

39. Manhattan plot for 2708 m/z features discriminating Clusters 13 and 14
We then used FDR to evaluate 2708 features for these 14 individuals in Clusters 13 & 14 to test for differences between these two clusters of AMD patients. A total of 335 features differed at a stringent q= This Manhattan plot shows that many m/z features had very significant p values.
335 m/z features differed at q=0.01

40. AMD Metabolomics
Metabolites and metabolic pathways differ between AMD and control.
Clinically indistinguishable AMD patients may be subcategorized based on their metabolic phenotype.
These studies may improve understanding of AMD pathophysiology and may help reveal causes of variable disease progression and treatment response.

41. Personalized Ophthalmology
DNA profile for all AMD patients at the VEI
This information will help us:
Choose specific treatment
Establish frequency of follow-up
Determine risk for relatives

42. Acknowledgments Vanderbilt Melissa Osborn Megan Parks Goodwin Burgess
Emily Wade
Samantha Williamson
Christopher Estopinal
Paul Sternberg
Emory
Dean Jones
Youngja Park
Karan Uppal
Sky Lee
Funding
Jahnigen Award from AGS
Reeves Foundation
American Health Assistance Foundation
International Retina Research Foundation
Macula Society
Vanderbilt Eye Institute
RPB Core Grants