Involvement of the Bcl2 gene family in the signaling and control of retinal ganglion cell death

Introduction

Retinal ganglion cell (RGC) death is the end-stage pathology of all glaucomas. In this review, we summarize the current level of knowledge of the molecular events associated with this process, particularly the involvement of a class of genes related to the Bcl2 gene family. Much of the research that has lead to our understanding of the molecular events associated with ganglion cell death has originated from the study of rodent models of both experimental glaucoma and chronic ocular hypertension (Weinreb and Lindsey, 2005). For this reason, the genetic nomenclature that we use here is restricted to convention for rats and mice.

Corresponding author. Tel.: 608-265-6037; Fax: 608-262-1479; E-mail: nickells@wisc.edu

More than a decade ago, several laboratories reported that RGCs died by a process similar to developmental programmed cell death, known as apoptosis. This modality of cell death was evident in models of acute optic nerve lesion, such as axotomy and crush (Berkelaar et al., 1994; GarciaValenzuela et al., 1994; Quigley et al., 1995; Li et al., 1999), and in models of experimental glaucoma (Garcia-Valenzuela et al., 1995; Quigley et al., 1995; Libby et al., 2005b), which is now believed to be a milder version of the crush lesion (Nickells, 2007). Similar observations have also been reported for ganglion cell death in human glaucoma and ischemic optic neuropathy (Levin and Louhab, 1996; Kerrigan et al., 1997).

Morphologically, dying ganglion cells exhibit many of the characteristic features of apoptosis, including dendritic tree and soma shrinkage (Misantone et al., 1984; Weber et al., 1998; Morgan

424

et al., 2000; Morgan, 2002), chromatin condensation, nuclear envelope dissolution, and the formation of apoptotic bodies (Quigley et al., 1995), and DNA fragmentation. More recent observations now suggest that ganglion cells death is subdivided into regional compartments (Whitmore et al., 2005), which can occur as independent autonomous self-destruct pathways. In this newer model, ganglion cell death begins with degeneration of the axon in the optic nerve and then spreads to the dendrites and soma in the retina. This model is consistent with the evidence that the initial site of damage in glaucoma is the lamina cribrosa (Quigley et al., 1981; Howell et al., 2007) and suggests a process that links elevated intraocular pressure (IOP) with the activation and eventual execution of the entire ganglion cell (Nickells, 2007). Technically, apoptosis is the term used to describe the autonomous selfdestruct pathway executed by cell somas.

Intrinsic apoptosis vs. extrinsic apoptosis

There are two basic pathways of apoptosis (Adams and Cory, 2007), both of which culminate in the activation of a cascade of cysteine proteases called caspases (Salvesen and Dixit, 1997; Slee et al., 1999; Adams and Cory, 2002). The caspases function to literally digest the cellular contents from within, thus allowing for the elimination of the cell without adversely imparting significant stress on the surrounding tissue of the whole organism. Caspases are present as pro-proteins in healthy cells and they become activated when the prodomain is cleaved off. There is a great deal of evidence that caspase activation is present in dying ganglion cells, supporting the early observations that these cells die by apoptosis with more molecular evidence (Kermer et al., 1998; McKinnon et al., 2002; Huang et al., 2005a). Caspases can be activated by two distinct apoptotic pathways. Extrinsic apoptosis, also referred to as the death receptor pathway, relies on the interaction of an extracellular ligand of the Tumor Necrosis Factor (TNF) family with a death receptor on the surface of the target cell. Once bound, the receptor engages adaptor proteins resulting in the direct activation of the caspase cascade through the

proteolytic cleavage of procaspase 8. Alternatively, caspases can also be activated by an internal signaling pathway called intrinsic apoptosis. The intrinsic pathway is significantly more complicated and relies heavily on mitochondrial involvement. In this pathway, signaling molecules from the Bcl2 gene family are recruited to the mitochondrial outer membrane causing disruption of mitochondrial function and the release of the electron transporter cytochrome c. Once released, cytochrome c complexes with a molecule called apoptosis inducing factor-1 (Apaf-1) and procaspase 9, creating a structure known as the apoptosome (Adams and Cory, 2002). This complex facilitates the cleavage of procaspase 9, which then is able to activate the caspase cascade. An important consideration when comparing the two pathways is that, by itself, mitochondrial dysfunction is lethal to the cell. This is principally due to disruption of the electron transport chain, resulting in the loss of ATP production. In addition, because molecular oxygen is no longer converted to water by the transfer of electron free radicals, mitochondrial dysfunction also leads to the generation of excessive reactive oxygen species. Thus, involvement of mitochondria is often considered an irreversible step in the apoptotic pathway (Chang et al., 2002; Nickells, 2004). This is experimentally evidenced by studies using caspase inhibitors, which can provide only a transient protective effect to cells undergoing intrinsic apoptosis (Chang et al., 2002). A caveat to the concept of distinct apoptotic pathways is that the two can share common elements in some cells. One process of the extrinsic pathway is to recruit the intrinsic pathway via the activation of signaling molecules that affect mitochondria. A more detailed discussion of these molecules and their function will be given in a subsequent section of this review.

The majority of experimental evidence points to RGCs executing the intrinsic apoptotic pathway. This evidence is partially based on studies of mitochondrial changes in dying ganglion cells (Mittag et al., 2000; Tatton et al., 2001), and the activation of caspase 9 in these cells (Kermer et al., 2000). In addition, as will be discussed below, genetic studies using mice mutated for different

members of the Bcl2 gene family, show dramatic effects in preventing ganglion cell soma death. Lastly, even though intrinsic apoptosis is likely the dominant mechanism of ganglion cell death in glaucoma, there is also compelling evidence that downstream activation of extrinsic apoptosis may play a role in the overall pathology of this disease.

The Bcl2 family of proteins

Both the extrinsic and intrinsic apoptotic pathways involve proteins with structurally similar amino acid domains. Originally, these domains were identified in a protein called BCL-2, the product of a gene found in a t(14;18) chromosomal translocation common to human B-cell lymphomas (Bakhshi et al., 1985). Members of the Bcl2 gene family share homology with at least one of four conserved amino acid domains, hence the name Bcl Homology (BH) domains (Fig. 1). The BH1, BH2, and BH3 domains of a single protein can organize into a hydrophobic groove, which may allow family members containing all three to form pore structures in lipid bilayers. More importantly, however, proteins containing the BH3-only domain are able to interact with the groove by virtue of an amphipathic a-helical domain (Adams and Cory, 1998). Additionally, most members of the Bcl2 gene family have the ability to anchor, or insert, themselves into membranes, which has led to the speculation that they function to create membranous pore structures or destabilize lipid bilayers.

Within the Bcl2 gene family, there are three subfamilies of proteins. The first group contains related proteins that are anti-apoptotic and promote cell survival. The major representatives of this subfamily are Bcl2 and BclX, although at least 10 other similar proteins have been identified. The second group contains related proteins that are pro-apoptotic and promote cell death. This subfamily is smaller, containing three known members, with the most well characterized being Bax and Bak. Finally, a third group has also been classified in the Bcl2 family by virtue of containing a single BH3 domain. Whereas genes within the first two subfamilies tend to be homologs of each

425

other, the BH3-only proteins are generally unrelated except for the BH3 domain. There are at least eight known members of this subfamily and they also play a role in activating apoptosis.

Gene expression studies of RGCs indicate that members of all three subfamilies are transcribed in these cells. Quantitative mRNA analysis and localization experiments indicate that BclX is the most prevalent member of the anti-apoptotic genes in ganglion cells, exhibiting at least a 16-fold greater abundance over Bcl2 transcripts in the whole retina (Levin et al., 1997). This does not preclude that Bcl2 is also expressed in the ganglion cells, since sensitive PCR-based methods of mRNA analysis are also able to detect transcripts from this gene in whole retinal extracts (Chaudhary et al., 1999). Overexpression of anti-apoptotic proteins has a dramatic impact on ganglion cell survival after optic nerve trauma. Transgenic mice that selectively upregulate Bcl2 in neurons exhibit prolonged RGC survival after optic nerve lesions (Bonfanti et al., 1995; Bonfanti et al., 1996; Cenni et al., 1996), while introduction of exogenous BCL-X by gene therapy or permeable fusion proteins prevents ganglion cell loss in rat axotomy models (Liu et al., 2001; Kretz et al., 2004; Malik et al., 2005).

Of the pro-apoptotic subfamily, Bax appears to be the most relevant molecule for ganglion cell death. This observation comes from genetic experiments using mice with engineered mutations in the Bax gene (Mosinger Ogilvie et al., 1998). Mice homozygous for the Bax mutation exhibit complete abrogation of ganglion cell soma death in models of acute and chronic optic nerve damage (Li et al., 2000; Libby et al., 2005b). This is generally true for many neuronal cell-types, while non-neuronal cells appear to express comparable amounts of Bax and Bak and both genes must be disabled to block cell death (Wei et al., 2001). There is also some controversy regarding the transcriptional regulation of Bax in damaged ganglion cells. Some studies have suggested that Bax levels increase early in the apoptotic pathway (Isenmann et al., 1997; Na¨pa¨nkangas et al., 2003), while others have not been able to confirm this finding (McKinnon et al., 2002) (S. Semaan and R. Nickells, unpublished data). Historically, Bax

mRNA and proteins are thought to exist as latent molecules in neurons (Putcha et al., 2003; Adams and Cory, 2007). A comprehensive discussion of Bax function in RGCs is presented in a later section of this review.

Ganglion cells also express several BH3-only proteins, including genes called Bim, Bad, and Bid. The interplay of the proteins from these genes is a critical control over the apoptotic program in most cells, including ganglion cells. While BH3-only proteins are unable to activate cell death by themselves, they play a role in modulating the interaction between pro-apoptotic and anti-apop- totic members. In our current understanding of these proteins, they appear to act as both activators and sensors of the apoptotic process within cells, such that as cell death is initiated, different BH3-only proteins are recruited to

augment the process and ensure successful execution of the pathway. In a separate section of this review, we present a model of how BH3-only proteins may interact in dying cells of a glaucomatous retina.

The requirement of BAX for RGC soma death

In healthy cells, BAX resides in an inactive state, principally localized in the cytosol, or loosely associated with intracellular membranes. BAX molecules that are resident in or near membranes may interact specifically with anti-apoptotic proteins, thereby preventing ‘‘accidental’’ permealization of these organelles. Upon activation of apoptosis, evidence shows that BAX is translocated primarily to the mitochondrial outer

membrane (Wolter et al., 1997; Putcha et al., 1999) (Fig. 2). Once there, BAX is thought to undergo a conformational change, allowing it to insert into the membrane bilayer, resulting in the release of cytochrome c. Although the exact mechanism of this release is currently not known (Danial and Korsmeyer, 2004), a popular model is that BAX can oligomerize and form pores in the mitochondrial outer membrane, enabling the release of cytochrome c. A caveat to this model is that proteinaceous pores have never been identified either in vivo or in vitro. An alternate model suggests that BAX proteins interact specifically with mitochondrial lipids, causing a decrease in the stability of the lipid bilayer and increasing permeability (Polster and Fiskum, 2004). Of note in this whole process is the suspected antagonistic function of the anti-apoptotic proteins. BCL-X is typically already localized to mitochondrial membranes by virtue of a specific mitochondrial targeting sequence at its carboxy terminus (Kaufmann et al., 2003) (Fig. 2). Cells expressing BCL-X are able to block apoptosis even with active BAX molecules present, suggesting that there is a direct antagonism between the two. One model suggests that these two proteins bind to each other and that this prevents BAX from affecting membrane permeability, but not necessarily its ability to insert into membranes (Putcha et al., 1999). This model is challenged, however, by some evidence that anti-apoptotic BCL proteins

Fig. 2. Localization of BCL-X ad BAX in HEK 293 tissue culture cells. Proteins are visualized by immunofluoresence. In unstressed cells, BCL-X staining is punctate, typical of its association with cellular organelles, principally mitochondria (left panel). BAX labeling in unstressed cells (center panel) is more diffuse, but becomes punctate within 24 h after treatment with staurosporine (STS), which activates apoptosis in these cells (right panel). The punctate labeling pattern is associated with the translocation of BAX from the cytoplasm to the mitochondrial outer membrane.

427

can exert a protective effect without associating with BAX (Liu et al., 2006). The true mechanism (or mechanisms) of action of BCL-X is still an unanswered question.

Much of the knowledge of Bax involvement in neuronal death comes from studies using mice with an engineered mutation in the Bax gene. Mice carrying mutant Bax genes exhibit a marked decrease in developmental neuronal death in a variety of sites such as the brainstem, cerebellum, and hippocampus (White et al., 1998). This is also true of the retina, which exhibits supernumerary neurons in both the inner nuclear layer and ganglion cell layer (Mosinger Ogilvie et al., 1998). Using this mouse model, researchers found that ganglion cell death after acute optic nerve trauma (crush lesion), was absolutely dependent on Bax function, but other stimuli of ganglion cell death, such as high levels of excitotoxins, elicited a Bax-independent program of apoptosis (Li et al., 2000) (Fig. 3). The reason underlying the alternative cell death pathways is still not clear, but probably lies in which BH3-only sentinel molecules are activated under varying circumstances.

The mutant Bax allele was also used to establish Bax function in ganglion cell death in glaucoma. In these experiments, the mutant allele was congenically cross-bred onto the genetic background of the DBA/2J line of inbred mouse. During the last decade, studies on this line have shown that they exhibit abnormalities in melanocyte formation and melanin synthesis. These defects cause a breakdown in the iris stroma and dispersion of pigment debris that becomes trapped in the mouse trabecular meshwork leading to pathology of the angle tissues and outflow pathway (John et al., 1998; Chang et al., 1999; Anderson et al., 2002; Mo et al., 2003). These anterior chamber defects are clinically evident in mice at 6 months of age and very prominent in mice older than 8 months. Concomitant with these defects, the DBA/2J mouse develops elevated IOP followed shortly after by degeneration of the optic nerve and the subsequent loss of RGCs (Libby et al., 2005a; Schlamp et al., 2006). DBA/2J mice congenic for the Bax mutant allele revealed three important features of RGC loss in response to elevated IOP (Libby et al., 2005b). First,