Transcriptome changes in age-related macular degeneration

Standard differential expression analysis produced only a small number of statistically significant genes associated with AMD classes. The authors then relaxed their filtering criteria and computationally identified modules of genes within and among AMD classes, tissue types, and anatomical regions (macula and periphery) based on coordinate expression and putative shared biological function. This elegant clustering method, designed to detect patterns confounded by background variation, revealed that the RPE/choroid in eyes with CNV showed increased expression of genes associated with angiogenic processes (such as VCAM1 and WNT4), whereas samples with geographic atrophy (GA) were enriched for apoptosis-related genes (such as caspases).

In both the RPE/choroid and retina, genes elevated across all AMD grades were enriched for regulators of cell-mediated immunity. In the RPE/choroid, these included immunoglobulin genes, a number of cytokines, and several CD antigens associated with T cell activation. These findings provide strong support for the concept that the microenvironment in AMD eyes is pro-inflammatory, with increased numbers and/or activities of leukocytes that may be responsible for injury to resident RPE or choroidal cells. Moreover, the elevation of these transcripts at the earliest stages of AMD, even before disease may be clinically identified, may suggest early opportunities for therapeutic management, which would be a major advance.

A hallmark of many systems-level studies is the integration of large datasets with molecular networks. These models become frameworks for further analyses, such as the identification of pathologically altered pathways or the discovery of central modulators of network function, which may serve as drug targets. Newman et al. used genes from statistically significant disease modules to retrieve protein association relationships from online interaction repositories. From this information, they constructed interactomes for RPE/choroid and retina. Data regarding macular versus extramacular expression and disease progression was then overlaid on the network. Both interactomes were significantly enriched for genes previously associated with AMD. In the macular retina, phototransduction elements decreased expression with increasing AMD severity. This measured decrease likely results from the loss of rod and cone photoreceptor cells observed in advanced AMD (for example, Figure 1D, E).

While the study by Newman et al. is notable as the first transcriptomics study to systematically analyze AMD cases and to so do with a very large sample cohort, several issues remain to be addressed by the field. Presumably, most of the mRNA changes detected will result in altered levels of their gene products. However, the relationship between changes in mRNA expression and protein production are not always linear. For example, a recent study analyzed gene expression in porcine tissues using three platforms (two that interrogated RNA levels and one that quantified proteins) [15]. By comparing the ratios of positively correlated genes common to the three technologies, the authors found good correspondence between transcript ratios and protein ratios for about 75% of these genes, with approximately 25% showing poor correspondence. Thus, integration of transcriptome profiling, as presented by Newman et al., with new and previously published proteomics surveys of AMD tissue [16‐18] may resolve which transcripts are translated directly or are under post-transcriptional regulation.

In addition, methods are now available to assess not only the total abundance of mRNA molecules encoded by a gene, but also their specific transcripts and novel isoforms [19, 20]. Distinct splicing isoforms of several genes are associated with both inherited and acquired diseases [21] and identifying the specific isoforms of AMD-associated molecules may reveal novel regulatory mechanisms.

There are several potential sources for transcriptional signals that vary between affected and unaffected tissues, including: altered expression of the gene within resident cells; degeneration of a cell type in disease that contributes a specific transcript in the healthy organ; and increased migration/proliferation of a cell type such that a new synthetic source for the mRNA has been introduced. The elevation of cellular immunity genes in all stages of AMD is very interesting, and may represent altered behavior of resident RPE/choroidal cells and/or the introduction of increased numbers of leukocytes in the choroid of eyes with AMD, as described previously with immunohistochemistry [22, 23]. As Newman and colleagues point out, identifying the cellular source(s) of the AMD-associated transcripts will be important extensions of this work.

Finally, a clearer picture has emerged over the last several years of how variants in genes associated with the complement pathways influence risk of AMD [24]. An extension of the excellent studies by Newman and colleagues would be comparing the transcriptome of anatomically healthy eyes at genetically low- or high-risk for developing AMD. Identifying altered patterns of complement proteins has been explored in human tissues with known AMD-associated genotypes [25, 26]. Systems based analysis of tissues with different genotypes may uncover the initial molecular events in eyes predisposed to disease, even before the genesis of any overt pathology.