Joe Corbo (Postdoctoral Fellow)
Bo Chen (Postdoctoral Fellow)
Timothy Cherry (Graduate Student)
Ashu Jadhav (Graduate Student)
Doug Kim (Postdoctoral Fellow)
Jeff Trimarchi (Postdoctoral Fellow)
Takahiko Matsuda (Postdoctoral Fellow)

The vertebrate retina comprises six major classes of neurons and one class of glia (Figure 1). These cell types are produced in an orderly manner that is generally conserved (reviewed in Altshuler et al., 1991). Lineage analyses have indicated that mitotic progenitors are multipotent (See Turner and Cepko, 1987; Turner et al., 1990). This multipotency persists throughout development, as revealed most clearly by the smaller retrovirally marked clones initiated by infection of the postnatal rat retina. Many of these clones had only two cells, and these could comprise two very different cell types (Figure 2A).  How can such clones arise? Two extreme models can be proposed. In Model 1 (Figure 2B), the environment is completely responsible for dictating the choice of cell fate. One type of progenitor is invoked and is responsive to an array of extrinsic cues such that several cell types can be produced. In Model 2 (Figure 2C), all information concerning cell fate derives from intrinsic information. However, as stated above, lineage analyses have shown that the retina does not have committed progenitors restricted to producing only one cell type, with the possible exception of a rod progenitor. Thus, in Model 2, distinct progenitors that each make more than one cell type (with the exception of a rod progenitor) are predicted.

         We have addressed the question of whether extrinsic or intrinsic cues are important for cell fate determination. We found that both types of information feed into the cell fate determination process. As detailed in “Single cell microarray analysis”, we are examining single progenitor cells for differences. We are also looking at the roles of extrinsic cues, in part using microarrays to look at the response of cells to exogenous cues added to cultures. In addition, retinal cells are transduced in vivo or in vitro using electroporation of plasmids, or infection with a retrovirus, to assay gene function. For example, transduction of a dominant negative or constitutively active alles of a receptor can be used to probe the activities that underlie the response to an extrinsic cue. Animals with knock-our mutations in some of the genes that are key for cell fate determination have also been generated or acquired. These are assayed by microarrays or SAGE to determine the genes whose levels are dependent upon a gene, particularly transcription factors (Livesey et al. Current Biology; Blackshaw et al. Cell, 2001). The focus is currently on basic helix-loop-helix genes and paired type homeobox genes, though several other classes of genes are also being investigated.

         One particular gene that can be classified as playing a role intrinsically in the interpretation of extrinsic information is the Notch receptor. This receptor is present on retinal progenitor cells from different ages. Expression of an activated allele of Notch results in formation of cells that express features of glia as well as features of late progenitor cells. In addition, expression of this activated allele in late retinal progenitor cells results in hyperproliferation. We are now examining the retina for the effects of removal of Notch, using a conditional allele, as well as expression of the activated allele at different ages.

We are also studying how a cell differentiates, that is, turns on the genes required to make it into a particular cell type. Our studies of rod photoreceptor development are the most advanced to date (Figure 4). There is an unusually long period between withdrawing from the cell cycle and beginning synthesis of the outer segment, the specialized membranous process of photoreceptors where phototransduction takes place. During this time, commitment to the rod fate occurs. Cells in this stage can also be influenced by environmental cues to become bipolar cells. Taurine, a small molecule derived from cysteine, and members of the CNTF family of cytokines can influence this decision. The Crx homeobox gene is not necessary for rod fate determination, but is required for the growth of the outer segment (Figure 5). It is most likely also required for maintenance of the photoreceptor as mutations in Crx in humans have been found in several types of blinding illnesses. The mechanisms at work during the period shown in Figure 4 are currently under study.

FIGURE 1. The cell types in the vertebrate retina are shown here in a cross sectional diagram taken from “Eye, Brain, and Vision, by David Hubel. Photoreceptors, rods and cones, are found in the outer layer of the retina. Their outer segments are membranous structures that capture light and carry out phototransduction. They form synapses with bipolar and horizontal cells, found in the inner nuclear layer. Also in the inner nuclear layer are amacrine cells, which synapse with bipolar cells and the output cell type, ganglion cells. Ganglion cells then send the result of all of this processing to the brain via the optic nerve.


2A. Infection of the postnatal rat with a retrovirus vector results in multicellular clones of varying compositions (Turner and Cepko, 1987). The major clone types comprising only 2 cells are shown here. Although two cell clones can result from the pruning of larger clones, many two cell clones must result from a terminal division that produces two postmitotic daughter cells. These clonal compositions show that two different fates can be chosen by the two postmitotic daughter cells.

2B. A model in which the environment is responsible for the choice of the four postnatal cell fates. In this model, there is only one type of mitotic progenitor, and one type of newly postmitotic daughter cell. This single type of progenitor is responsive to different environmental cues that dictate the choice of cell fate. By having these cues present in different amounts, different ratios of the four types of postnatally-generated cells can be produced.

2C. A model in which intrinsic differences among progenitor cells drives the production of the four different postnatally-generated cell types. In this model, retinal progenitors are a mixture of distinct progenitor cell types, with each type programmed to produce different types of progeny. No reliance upon environmental cues for the cell fate decisions is proposed, as these decisions are passed on to each postmitotic daughter cell by its progenitor. The distinct types of progenitors can be present in various ratios to account for the production of different cell types in different ratios.

FIGURE 3. Retinal progenitors comprise a dynamic mixture of distinct mitotic cell types that interact with the environment to make the different postmitotic cell types. Each progenitor cell type is thought to be controlled by a complex of transcription factors that define its competence to make a particular retinal cell type, or small set of cell types. Retinal progenitors are modeled to progress from one state of competence another in only one direction. Early progenitor cells appear to be unable to jump ahead to later stages of competence. There are most likely many more states of competence than shown here, and there may be many branch points along this progression. The environment is shown to be changing over time, in part due to the production of postmitotic, differentiating cells. The postmitotic cell types can regulate, via feedback inhibition, the production of more neurons of the same type, and possibly regulate cell fate choices through other types of interactions (from 56).

FIGURE 4. A model of rod photoreceptor development. Lineage analyses in multiple species have indicated that rods are generated by a multipotent progenitor (for review see Cepko et al., 1996). The lag time between becoming postmitotic and synthesizing opsin is 5-6 days (Morrow et al. 1998; see Figure  ). Several transcription factors have been shown to be expressed in the pool of mitotic progenitors that produce rods, including three genes which encode paired type homeodomains, Rax, Pax6, and Chx10; a negatively acting bHLH, hes1; and also the transmembrane receptor, Notch 1 (see text for further details). Some time between becoming postmitotic and expressing opsin, all of these genes will be extinguished in cells that will become rods. Crx, another paired type homeodomain gene, is expressed soon after a cell fated to become a rod is born. If an activated Notch1 allele is introduced into the mitotic progenitor cells for rods in vivo, differentiation into mature rods is blocked (Bao and Cepko, 1997) and in mice heterozygous for a hes1 null allele,  rhodopsin expression was shown to be precocious for a fraction of rods (Tomita et al., 1996). If CNTF is added to mitotic cells, or to postmitotic cells that are fated to express opsin, but have not yet begun to do so, the expression of opsin is blocked. In rats, even these postmitotic cells fated to be rods are not committed to the rod fate as a fraction of these cells go on to express at least three markers of the bipolar cell fate (Ezzeddine et al., 1997). Once a cell has expressed detectable levels of opsin, it is resistant to CNTF. We speculate that the specification to the rod fate occurs over several days, perhaps beginning with information passed from the mitotic multipotent cell to its postmitotic daughter. This information might endow a cell with the competence to become a rod in response to environmental cues. These cues may be part of the process that extinguishes the repertoire of transcription factors expressed in the multipotent progenitor, and that results in the synthesis of Crx. As CNTF resistance is concomitant with opsin synthesis, the committment step, whereby a cell is no longer plastic and able to become a bipolar neuron, may occur just before opsin is detectable. Introduction of a presumed dominant-negative allele of Crx, in which the DNA binding domain of Crx is fused to the repressor domain of the Drosophila engrailed gene, leads to lack of formation of outer segments and rod termini (Furukawa et al., 1997).

FIGURE 5. Histology of the Crx knock-out mouse. A crx null  mouse was created (Furukawa et al. 1999). This animal has no detectable electroretinogram, indicating an absence of activity in response to light. The reason why there is no response is most likely due to the lack of the outer segments. At 2 months of age, no outer segments are seen in the knock-out. At 6 months of age, the knock out also exhibits a loss of rod cells. This animal is a good model for humans with Leber’s Congential Amaurosis, in which babies are born with no, or almost no, vision. Some people with Crx mutations have Leber’s.