The results of the combined proteomic, transcriptomic, and confirmation experiments both validate previously reported changes in retinal protein expression with diabetes and identify novel changes. These findings have implications for both our understanding of the pathogenesis of DR and as preclinical drug development biomarkers. Additionally, the combination of discovery approaches used provides comparative insight into commonly used quantitative proteomic techniques.
Biological functions of confirmed proteins
In agreement with previous reports, diabetes induced retinal expression of multiple crystallin isoforms was observed. Crystallin isoforms Cryaa (α-crystallin A), Crybb2 (β-crystallin B2), Cryab (α-crystallin B) and Cryba3 (β-crystallin A3) were confirmed to be induced after 3 months of diabetes. Interestingly, Cryba3 protein expression was only partially normalized by insulin treatment and Crybb2 protein expression was unaffected.
Crystallin proteins are associated with a variety of functions ranging from lens structure to stress response, but little is known about their potential involvement in diabetic retinopathy. Alpha crystallins, which are similar to small heat shock proteins, have been reported to have neuroprotective effects 
. Cryab is upregulated in activated astrocytes following ischemic damage 
and decreases peroxide-induced apoptosis of astrocytes through suppression of caspase-3 signaling 
. The upregulation of Cryaa and Cryab with diabetes in our work suggests a potential compensatory response in the neural retina that may act to protect neurons and glia from inflammatory insults. Beta crystallins are expressed more highly in the retina than in other non-lens tissues 
, and Cryba3 was recently reported to be restricted to astrocytes in the neural retina 
Neuroprotection of retinal ganglion cells in a rodent model of ocular hypertension has also been reported to depend on upregulation of Crybb2. Altered expression of crystallin proteins has been previously reported with diabetes, although this is the first study with targeted confirmation analysis of multiple crystallin isoforms. The findings of this work and other literature 
suggest a protective role of crystallin upregulation in the retina with diabetes. The continued induction of Cryba3 and Crybb2 after restoration of normoglycemia by insulin may be beneficial but also suggests continuing stress to the retina.
Dbi (diazepam binding inhibitor) and Anxa5 (Annexin A5) were identified as targets of interest in our iTRAQ analysis. Dbi is an insulin-regulated transport molecule with numerous functions and protein interactions ranging from regulation of lipid metabolism 
to binding components of inhibitory neurotransmission machinery 
. In the retina, Dbi is expressed by pigment epithelium cells and Müller glia 
, and secrete this protein in response to protein kinase C activation 
. Dbi binds GABA(A) receptors with high affinity 
, and it is thought that suppression of GABAergic signaling is one mechanism through which Dbi modulates retinal neurotransmission, particularly in the inner plexiform layer 
, which further increases following activity-dependent phosphorylation 
, which is a hallmark of diabetes-related retinal pathology 
. Combined with previous reports of insulin-mediated regulation of Dbi expression 
, the responsiveness of Dbi protein and mRNA to insulin therapy in the present study suggests that this target may be a promising candidate for treatments focused on modulating neurovascular function in the diabetic retina.
Although well-characterized for its utility as a marker of early apoptosis due to its high binding affinity for phosphatidylserine, Anxa5 mediates a variety of cellular functions including adhesion and calcium flux in addition to phosphatidylserine-catalyzed inflammation and apoptosis. This glycoprotein is expressed in vascular cells 
, neurons 
and in the blood 
among other locations. In the retina, Anxa5 has been localized to the ganglion cell layer 
, where it may influence neuronal function through binding of the postsynaptic scaffolding plasma membrane protein PSD95 
Our data agree with a previous report of upregulated Fgf2 protein levels after 3 months of diabetes in the Sprague Dawley rat 
. A novel finding here was that 1.5 months of insulin treatment did not restore Fgf2 protein or mRNA expression to control levels. Previously a sustained induction of Fgf2 mRNA levels in spite of insulin treatment at a one month time point has been reported 
. Increased retinal Fgf2 protein levels may serve a neuroprotective role 
and as such, continued induction after insulin therapy may not be deleterious. However, as Fgf2 is induced by a wide range of retinal insults, Fgf2 induction may indicate continued damage to the retina even after restoration of normoglycemia.
Stat3 activation by diabetes and subsequent blood retinal barrier damage can be prevented by simvastatin treatment 
. The angiotensin converting enzyme inhibitor enalapril has been demonstrated to prevent the diabetes-induced up-regulation of Stat3 protein 
. Stat3 can exert a large number of effects as a signaling regulator to a number of downstream endpoints. As such, modulation of Stat3 expression likely impacts a variety of pathways and processes in the retina and may prove to be a useful target in drug development efforts for diabetic retinopathy.
Lgals3 (lectin, galactoside-binding, soluble 3; galectin-3) is an abundant protein expressed in the nucleus, cytoplasm, mitochondria, and plasma membrane of multiple cell types including inflammatory macrophages 
, retinal Müller glia 
and endothelial cells 
. This protein binds advanced glycation end products (AGEs) with high affinity 
and, as such, may play an important role in the development of diabetes complications. Lgals3 expression has previously been reported to increase in Muller cells of diabetic rats in association with increased acute phase genes 
. Lgals3 has also been shown to suppress retinal angiogenesis 
and promote inner blood-retina barrier dysfunction and breakdown 
. It is likely that increased Lgals3 expression with diabetes is mediated by increased AGE formation 
and contributes to inflammatory 
and vascular components of retinal complications.
Likewise, ceruloplasmin (Cp), an iron binding and metabolizing α2 globulin, was upregulated with diabetes in this work. This protein is a component of the acute phase response, and is induced by inflammation 
as well as interleukin-1 stimulation 
, although Cp upregulation appears to constitute an anti-inflammatory cytoprotective response 
. Cp protein is increased in sera from hyperglycemic patients 
and after optic nerve crush 
. Cp transcript is also increased in retinal Müller cells of diabetic rats 
which may secrete this protein and ultimately impact retinal microvasculature. It has been proposed that upregulation of Cp protects against oxidative damage by decreasing ferritin levels and reducing AGE-mediated damage 
. Given the potential protective roles of these proteins, further study is required before a definitive determination of the benefits of normalizing their expression with insulin can be made.
Previously, in focused studies examining specific proteins changes in expression of para-inflammatory and microvascular-related proteins (e.g., Vegf, Icam1, TNFα, and IL-1β) have been reported in rodent models of DR. Some of these proteins, such as Icam1, were not observed with any of the three proteomic methods used in this study. For other proteins, no changes in expression were observed (e.g., Vegf, TNFα, and IL-1β). Increased Vegf protein expression has been reported in the Sprague Dawley STZ model at after 1 and 6 months of diabetes 
but not after 3 months 
. Similarly varying results have been reported for the induction of both Il-1β and TNFα depending on the duration of diabetes 
. Clearly, there are important temporal characteristics to the expression of these proteins with diabetes. The DR field would greatly benefit from a thorough characterization of the expression and localization of these proteins across different durations and in multiple rodent models of diabetes, expanding on work previously reported by Kirwin and colleagues 
Comparison of proteomic technologies
The overarching finding from the comparison of the methods used in this study is that the different techniques implemented are complementary rather than redundant. Comparison of the biophysical properties of proteins identified by the different methods demonstrated no preferential discrimination of protein molecular weights between iTRAQ and DIGE. However, the distribution of identified iTRAQ proteins was skewed toward hydrophilic proteins while those identified by the DIGE analysis were more hydrophobic. Proteins detected by the Luminex analysis were evenly distributed across molecular weights and hydropathicity. Since Luminex assays are predesigned to select specific proteins, the technical capabilities of this assay are highly dependent on the specific antibodies used. The inclusion of transcriptomic analysis provided novel targets not observed in any of the proteomic analyses. Interestingly, all targets chosen from the transcriptomic analysis were confirmed at both the mRNA and protein levels. In total, the comparison of mRNA and protein expression levels demonstrates that while mRNA changes are not determinative of protein expression changes in many cases they are highly correlated.
This study identified a number of retinal protein expression changes with diabetes. Of particular interest for future studies are those proteins that are not normalized (Fgf2 and Crybb2) or only partially normalized (Cp, Cryba3, Lgals3, Stat3) by insulin treatment, as these persistent changes reflect the limitations of insulin to restore the retina to a nondiabetic state. The lack of complete normalization by insulin treatment offers the potential that these proteins may play a role in the ‘metabolic memory’ 
observed clinically in retinopathy pathogenesis 
. Future studies will address when these protein changes first occur as well as the cellular origin of these proteins and the mechanisms underlying their continued dysregulation despite insulin replacement.