The simplicity of the retina makes it an ideal tissue to study neurogenesis. Its development proceeds through three overlapping steps starting with retinal progenitor cell (RPC) proliferation, followed by birth of post-mitotic retinal transition cells (RTCs, also referred to as precursors), and ending with terminal differentiation of seven major cell types (A) [1
]. RPCs are multipotent and exit the cell cycle to generate different RTCs at specific time periods in development [1
]. This process of RTC “birth” requires coupling of differentiation and cell cycle exit. Once born, post-mitotic RTCs migrate and form different retinal layers. Rods and cones make up the outer nuclear layer (ONL); horizontal, bipolar, and amacrine cells, as well as Müller glia cell bodies, reside in the inner nuclear layer (INL); and ganglion and displaced amacrine cells form the ganglion cell layer (GCL) (A). The outer plexiform layer (OPL) and inner plexiform layer (IPL) house synaptic connections separating the ONL/INL and INL/GCL, respectively.
E2f1, but Not E2f2 or E2f3, Loss Rescues Ectopic Division and Cell Death in the Rb KO Retina
The retinoblastoma protein (Rb) is critical for cell cycle exit during retinal transition cell birth. Rb
knockout (KO) RTCs continue to proliferate inappropriately and some (rod, ganglion, and bipolar cells) die by apoptosis [2
]. Rb controls the cell cycle by binding and inhibiting E2f transcription factors (E2fs) (B), first defined as transcription factors that bind adenoviral E2 regulatory elements and subsequently shown to be critical cell cycle regulators [4
]. E2fs bind to DNA as heterodimers with proteins of the related Tfdp family. E2f1, E2f2, and E2f3a are “activating E2fs” that are required for fibroblast division. They are strong transcriptional activators that can drive G0 fibroblasts into cycle, and are inhibited when bound to Rb [4
]. Ectopic division in Rb
KO embryos can be rescued to various extents in different tissues by knocking out E2f1, E2f2,
], but which member(s) drive division in Rb
KO RTCs is unknown. Other members of the family, such as E2f4 and E2f5, are known as “repressive E2fs” because they are weak activators and appear to be primarily involved in gene silencing in quiescent or differentiated cells.
Activating E2fs may also promote apoptosis in the Rb
KO retina (B). Originally, E2f1 was considered the primary pro-apoptotic member of the family [10
]. However, this view was reevaluated when it was shown that either E2f1
deletion rescues apoptosis in the developing central nervous system (CNS) of Rb
KO embryos [6
]. Subsequently, CNS apoptosis was shown to be an indirect result of placental defects and probable hypoxia [12
]. Indeed, E2f3-induced apoptosis in fibroblasts has recently been shown to require E2f1 [15
]. Thus, it is controversial whether E2f3 is required for apoptosis of any Rb
KO cell type. Determining which activating E2fs promote death in distinct Rb
KO tissues requires conditional rather than germ line models of Rb
deletion to avoid secondary indirect effects (such as hypoxia).
E2f family diversity is expanded by E2f3 isoforms. Alternative promoters generate two forms (a and b) that are identical except for distinct first exons [16
]. E2f3a is a strong activator, and, like other activating E2fs, its expression is induced when quiescent cells are stimulated to divide [16
]. E2f3b, like repressive E2fs, is present in both quiescent and dividing cells, and in quiescent fibroblasts it associates primarily with Rb, suggesting that it mediates repression [16
]. Indeed, silencing the Cdkn2d (p19Arf
) locus in unstressed cells relies on E2f3b [19
]. Other E2fs may also exist in isoforms since at least two mRNA species have been detected for E2f1 and E2f2 [16
]. The roles of E2f isoforms in vivo are unknown.
E2fs are also regulated by subcellular localization. Although this feature has been best characterized for repressive E2fs [20
], it also affects activating E2fs [23
]. The distribution of E2f isoforms has never been assessed.
It has been known for many years that Rb loss perturbs neuronal differentiation [26
]. However, prior work could not exclude the possibility that differentiation defects are simply an indirect consequence of abnormal division and death. If Rb does regulate differentiation directly it is unclear whether it does so in all or a subset of neurons. Moreover, the mechanism has never been solved. In other cell types where Rb may promote differentiation directly, such as muscle and bone, it seems to do so through E2f-independent means by potentiating tissue-specific transcription factors (B) [30
]. In the retina, others have noted abnormally shaped Rb
KO rods and have suggested Rb may directly promote their morphogenesis by activating retina-specific factors [29
]. However, differentiation defects in any Rb
KO neuron could be an indirect effect of ectopic division and/or apoptosis (B). Thus, it is critical to study differentiation of Rb
KO cells in the absence of ectopic proliferation and death.
Here, we establish that Rb suppresses RTC division and death by inhibiting E2f1, not E2f2 or E2f3. When these defects were rescued, most retinal neurons, including rods, survived, differentiated, and functioned normally. Thus, unexpectedly, retina-specific differentiation factors function independently of Rb. However, comprehensive assessment of the Rb
double-null rescued retina revealed a differentiation defect in cholinergic starburst amacrine cells (SACs). Recent breakthroughs have revealed that these interneurons are critical for direction selectivity and developmentally important rhythmic bursts [34
]. However, their differentiation is poorly understood. Contrary to the prevailing view that Rb promotes differentiation through E2f-independent tissue-specific transcription factors, we show that Rb facilitates SAC development through E2f3. Defects in Rb
null SACs correlated with specific E2f3 expression in these cells, and E2f3 expression was absent in neurons that differentiated without Rb. E2f3 is also present in a specific subset of other CNS neurons, implying that this may be a general mechanism by which Rb facilitates neurogenesis. To define the mechanism in even more detail, we determined which E2f3 isoform Rb targets to control SAC differentiation. E2f3b mediates Rb function in quiescent fibroblasts [19
], yet no prior studies to our knowledge have dissected E2f3a or E2f3b functions in vivo. Using an isoform-specific null mouse we show that Rb drives SAC differentiation through E2f3a. Thus, independent of E2f1-mediated effects on division and death, Rb does regulate neuronal differentiation, but only in specific neurons and, unexpectedly, through E2f3a, not tissue-specific differentiation factors.