Coordinating cell cycle exit and differentiation in the mammalian
retina and its dependence on Rb
By Marek Pacal
A thesis submitted in conformity with the requirements for the degree of Doctor of
Philosophy
Graduate Department of Laboratory Medicine and Pathobiology
University of Toronto
© Copyright by Marek Pacal, 2011
Abstract
Coordinating cell cycle exit and differentiation in the mammalian retina
and its dependence on Rb
Doctor of Philosophy (2011)
Marek Pacal
Department of Laboratory Medicine and Pathobiology
University of Toronto
Cell cycle exit (“birth”) of retinal progenitor cells (RPCs) is considered a watershed that is preceded by changing levels of cell cycle regulators, and followed rapidly by induction of a post M-phase differentiation cascade. Yet the actual dynamics of these events are largely unclear, thus whether mitosis separates pre- and post- birth differentiation cascades is unproven. We characterized the regulation of many division and differentiation markers relative to each other and final mitosis. Unexpectedly, classic
“cell cycle” markers were present well beyond exit (e.g. Ki67, Pcna), early embryonic
RPCs expressed “differentiation” markers that later labeled post-mitotic neurons exclusively (e.g. Brn3b, Tubb3, Ptf1a), and factors detected just after cell birth in the embryo were induced well beyond M-phase post-natally (e.g. Nrl, Crx). Thus, the dynamics of birth-associated events shift dramatically during development, even to either
1 side of mitosis. Instead of mitosis behaving as a cog that activates post-exit differentation events we suggest that a common trigger induces both the exit and differentiation programs in RPCs, precisely coordinating their startpoints, but that each subsequent cascade unfolds independently. This model explains the convergence of birth and differentiation but also their temporal maliability. This view fits with our observation that in the absence of the Rb tumor suppressor, differentiation still initiates even without cell cycle exit. Finally, neoplastic transformation in the mouse retina requires loss of Rb and its relative p107, and emerging tumor features suggest an amacrine cell-of-origin. We studied Rb/p107 null clones, and noted two striking features.
First, despite initial expansion of aberrantly dividing differentiating cells, apoptosis pruned clones precisely to wild type sizes. “Cell competition” maintains tissue size by selecting fitter over weaker progenitors; our data provide a unique example of competition among differentiating cells. Second, despite normal numbers of amacrine cells per Rb/p107 null clone, more clones contained amacrine cells and fewer had bipolar cells. Both this effect and ectopic division were E2f1-dependent. Thus, the oncogenic initiation event in mouse retinoblastoma triggers a very early fate switch, even before neoplastic transformation, broadening the possibilities for the cell-of-origin of retinoblastoma, and arguing that even very early stage tumors cannot be used to define cancer origin.
2 Acknowledgements
I would like to thank to Dr. Rod Bremner for giving me the opportunity to learn in his lab, and Drs. Van Der Kooy and Huang for being on my student committee. I thank my parents for their support and patience and my dear Rachel for her love and understanding.
Further, I am grateful for the generous financial support from the Vision Science
Research Scholarship, Sandra and David Smith Graduate Student Award and Canadian
Institute for Health Research.
3 Table of Contents
Abstract ...... 1 Acknowledgements ...... 3 Table of Contents ...... 4 List of Figures and Tables ...... 7 List of Abbreviations ...... 9
CHAPTER ONE: INTRODUCTION ...... 10 1.1 Coordinating Division and Differentiation in Retinal Development ...... 11 1.1.1 A few basics of cell cycle regulation ...... 13 1.1.1.1 Role of some core cell cycle components in retinal development ...... 16 1.1.1.2 Cyclins and Cdks in retinal development ...... 17 1.1.1.3 The Rb family in retinal development ...... 21 1.1.1.4 Ink4 CKIs and p19Arf in retinal development ...... 23 1.1.1.5 Cip/Kip CKIs in retinal development ...... 24 1.1.1.6 E2fs in retinal development ...... 26 1.2 Separating Rate, Differentiation programs and Exit ...... 27 1.2.1 Coupling INM to Cell Birth: “I need to get away” ...... 28 1.2.2 Birth and Exit: “The Cog Model” vs. The Trigger theory” ...... 30 1.2.3 Birth and exit in frogs: “You walk, I’ll jump” ...... 36 1.2.4 Coupling differentiation to cell cycle exit: “Let’s take the mystery tour” ...... 37 1.2.5 “We need a better map” ...... 39 1.3 Cell Competition ...... 41 1.4 The role of Rb in differentiation ...... 43 1.5 The Cell of Origin of Retinoblastoma ...... 45
CHAPTER TWO: TIMING OF CELL CYCLE MARKER SILENCING AND THE ONSET OF DIFFERENTIATION OF RETINAL GANGLION CELLS ...... 54 2.1 Introduction ...... 55 2.2 Results ...... 58 2.2.1 Ki67, but not Pcna or Mcm6, is confined to the NBL ...... 58 2.2.2 Ki67 labels all phases of the cell cycle in all RPCs ...... 59 2.2.3 A subset of Ki67 cells lack the pan-cell cycle markers Vsx2 and Ccnd1 ...... 60 2.2.4 Ki67 co-labels cells positive for presumed differentiation markers ...... 62 2.2.5 Ccnd1 and Vsx1 extinction followed by induction of Isl1, Pou4f2, then Tubb3 in early RPCs ...... 63 2.2.6 The length of G2/M ...... 65 2.2.7 Timing of Ki67 extinction in G0* cells ...... 66 2.2.8 Timing of expression of ganglion neuronal markers ...... 68 2.3 Discussion ...... 70 2.3.1 Coordinating Exit and Differentiation: The Trigger Theory ...... 71 2.3.2 Evidence for RPCs biased towards the ganglion cell fate ...... 73 2.3.3 Induction and roles of Isl1 and Pou4f2 ...... 75 2.3.4 Ki67 remains in ganglion RTCs for a period of time after birth ...... 76
4 CHAPTER THREE: TEMPORAL SEQUENCE OF EVENTS DURING ROD, AMACRINEAND BIPOLAR CELL BIRTHS IN THE MOUSE RETINA ...... 94 3.1 Introduction ...... 95 3.2 Results ...... 97 3.2.1 Quantification of neuronal marker appearance with respect to cell birth ...... 97 3.2.1.1 Markers that label both RPCs and post-mitotic cells...... 98 3.2.1.2 Markers that label exclusively post-mitotic cells and can be used to detect rods and/or bipolar cells ...... 102 3.2.1.3 Markers that label exclusively post-mitotic cells and can be used to detect amacrine cells ...... 110 3.3 Discussion ...... 115 3.3.1 Panel of RTC markers ...... 116 3.3.2 Flexibility in expression of Isl1 and Crx during retinal development ...... 117 3.3.3 Insights into photoreceptor development ...... 118 3.3.4 Prox1 is expressed already in S-phase RPCs ...... 120 3.3.5 Does Ptf1a label a subset of RPCs biased towards amacrine/horizontal neurons? ... 122
CHAPTER FOUR: RAPID CELL FATE SWITCH IN VIVO FOLLOWING AN INITIATING ONCOGENIC EVENT ...... 142 4.1 Introduction ...... 143 4.2 Results ...... 145 4.2.1 Sporadic Rb inactivation ...... 145 4.2.2 DKO clones expand but contract precisely to wild type sizes ...... 146 4.2.3 Ectopic division of specific differentiating cell types ...... 148 4.2.4 Rb/p107 loss generates more amacrine-containing clones ...... 149 4.2.5 All clonal Rb/p107 defects, including fate change, are E2f1-dependent ...... 152 4.3 Discussion ...... 154 4.3.1 The Origin of Retinoblastoma: a Moving Target ...... 154 4.3.2 Deregulated E2f1, not tissue-specific factors, drives fate and differentiation defects ...... 155 4.3.3 Cell competition in the retina ...... 158
CHAPTER FIVE: MATERIALS AND METHODS ...... 170 5.1 Mice ...... 171 5.2 BrdU labeling ...... 171 5.3 Immunostaining ...... 171 5.4 Retroviral constructs ...... 172 5.5 Morphological Identification of GFP+ Cell Types ...... 172 5.6 Retroviral Injections ...... 173 5.7 TUNEL staining ...... 173 5.8 Laser capture microdissection ...... 174 5.9 Statistics ...... 175 5.10 List of antibodies ...... 175
5 CHAPTER SIX: DISCUSSION AND FUTURE DIRECTIONS ...... 177 6.1 Discussion and Future Directions ...... 178 6.2 Characterizing RPCs that may be biased towards the ganglion and amacrine/horizontal cell fates ...... 178 6.3 Rb proteins do not affect retinal progenitor division but lock neurons out of the cell cycle ...... 180 6.4 What drives the fate switch in Rb;p107 DKO clones? ...... 181 6.5 What is the basis of retinal cell competition? ...... 183
REFERENCES ...... 188
6 List of Figures and Tables
Fig 1.1 Retinal histogenesis. …………………………………………………………………...48
Fig 1.2. Some Key Regulators of G1/S transition. …………………………………………….49
Fig 1.3. Cell cycle regulation in RPCs versus differentiating RTCs. ………………………….50
Fig 1.5. Notch and INM control birth/exit pathways. ……………………………………….....51
Fig 1.4. Uncoupling cell birth and cell cycle exit. ………………………………...... 52
Fig 1.6. Cell cycle Exit and Differentiation are not rigidly coupled. ………………………….53
Fig 2.1. Overview of retinal structure and development. ……………………...... 78
Fig 2.2. The expression patterns of Pcna, Mcm6 and Ki67 in the E14.5 embryonic retina. …...79
Fig 2.3. Ki67 is detectable in all phases of cell cycle in the E14.5 retina. ……………………..80
Fig 2.4. The expression pattern of Ccnd1 and Vsx2 in the embryonic retina...... 82
Fig 2.5. The expression pattern of Ki67 and Pax6, Isl1 and Neuna60 in the embryonic retina. .83
Fig 2.6. The co-expression pattern of Ki67 and Pou4f2, Tubb3, Calb2 and Pou4f1 in the embryonic retina. …………………………………………………...... 86
Fig 2.7. Pax6 is expressed in RPCs throughout the cell cycle. ………………………………..87
Fig 2.8. Isl1 and Pou4f2 are expressed in a subset of RPCs. ………………………………….88
Fig 2.9. Tubb3 is expressed in a subset of G2/M RPCs at E12.5. …………………………….89
Fig 2.10. Calb2 and Pou4f1. …………………………………………………………………....90
Fig 2.11. The length of G2/M. ………………………………………………………………....91
Fig 2.12 The length of G0*. …………………………………………………………………....93
Fig 3.1. Overview of retinal structure and development. …………………………………….124
Fig 3.2. Summary of cell cycle and neuronal markers used in this study …………………….125
Fig 3.3. Otx2 is expressed in a subset of RPCs. ……………………………………………...127
Fig 3.4. Neurod1 is expressed a in a subset of RPCs. ………………………...... 128
Fig 3.5. Prox1 is expressed in S-phase RPCS at E14.5. ……………………………………...129
Fig 3.6. Prox1 is expressed at S-phase RPCs at P3 and P5. ……………………………….…130
Fig 3.7. Ptf1a is expressed in a subset of G2/M RPCs at E14.5. ……………………………..131
Fig 3.8. Crx is expressed in a minute number of G0* RTCs at E14.5. ……………….….…..133
7 Fig 3.9. Crx is expressed in a subset of G0* RTCs in the post-natal retina. …. …………….…134
Fig 3.10. Recoverin, Nrl and Nr2e3 are expressed in post G0* RTCs. …………………….….135
Fig 3.11. Isl1 and Gnao1 are expressed in post G0* RTCs. …………………………………...136
Fig 3.12. Cabp5 and Prkca are expressed in post G0* RTCs. ………………...... 137
Fig 3.13. Tfap2a and Dcx are expressed in a minute subset of G0* RTCs. …………………...139
Fig 3.14. Elav2/3/4 is expressed in a minute number of G0* RTCs. ………………………….140
Fig 3.15. Mtap1b and Uchl1 are expressed in mature post-mitotic neurons. ………………….141
Fig 4.1. A sporadic model of Rb deletion. ……………………………………………………..161
Fig 4.2. Ectopic division increases mutant clone size, but apoptosis rapidly restores cell numbers to WT levels. …………………………………………………………….162
Fig 4.3. Normal levels of dividing and dying cells around Rb/p107 DKO clones...... 163
Fig 4.4. Rb or Rb/p107 loss triggers ectopic division of each type of RTC. …………………..164
Fig 4.5. Additional markers confirm that Rb or Rb/p107 loss triggers ectopic division of each type of RTC ………………………………………...... 165
Fig 4.6. Clone size, cell proportions and differentiation in P21 WT, RbKO,
Rb/p107 DKO and Rb/p107/E2F1 TKO clones. …………………………………...... 166
Fig 4.7. The photoreceptors in RbKO clones display normal morphology. …………………...167
Fig 4.8. Confirmation of INL cell type identity. ……………………………………………….168
Fig 4.9. Rb/p107 deletion increases the production of amacrine containing clones...... 169
Table 2.1. Ki67 is expressed in all phases of the cell cycle. ……………………………………81
Table 2.2. A small population of Ki67+ cells lack Vsx2 and/or Ccnd1. ………………………..81
Table 2.3. A small population of Vsx2+ cells lack Ccnd1. ……………………………………..81
Table 2.4. Marker analyses. ……………………………………………………………………..85
Table 2.5. Proportion of BrdU-labeled PH3+ or Ccnb1+ cells at given time points. …………...92
Table 2.6. Cell cycle length analyses. ……………………………………………………….…..92
Table 3.1. Analyses of markers that are expressed in both RPCs and RTCs/neurons. ……..….126
Table 3.2. Analyses of markers that are expressed in photoreceptor/bipolar RTCs & neurons....132
Table 3.3. Analyses of markers that are expressed in amacrine neurons. …………………..…..138
8 List of Abbreviations
Rb: retinoblastoma (gene, protein)
E2f: E2 (adenoviral) promoter binding factor
E: embryonic (day)
P: post-natal (day)
RPC: retinal progenitor cell
RTC: retinal transition cell
INM: interkinetic nuclear migration
ONBL: outer neuroblastic layer
INBL: inner neuroblastic layer
ONL: outer nuclear layer
INL: inner nuclear layer
GCL: ganglion cell layer
S (phase): DNA Synthesis phase of the cell cycle
M (phase): Mitosis
G1/2 (phase): Gap 1/2 phases of the cell cycle
Cdk: cyclin-dependent kinase
CKI: cyclin-dependent kinase inhibitor
9
Chapter One: Introduction
The work in Chapter 1 is an expanded version of a book chapter written by Dr. Rod Bremner and Marek Pacal published in the Ecyclopedia of the Eye, Academic Press, 2010, pgs. 398-407.
10 1.1 Coordinating Division and Differentiation in Retinal Development
This introduction focuses on the mechanism through which differentiation and cell cycle exit are coordinated, the consequences when they are not, and whether exit is required for initiation of differentiation programs or is simply an important parallel event. Cell cycle regulation is also critical to control tissue size, which is discussed here briefly in the context of cell competition (Hever et al. 2006; Buttitta and Edgar 2007; Lecuit and Le
Goff 2007).
In the earliest phase of eye development the emergent neuroepithelial sheet expresses transcription factors that affect division and render retinal progenitor cells
(RPCs) competent for differentiation. For example, in pre-competent RPCs (yet unable to make retinal cell types) Sox2 is essential for normal levels of division and to achieve competence (Taranova et al. 2006). In mice, competence is conferred around
~embryonic day 11 (E11) after which RPCs can generate differentiating retinal transition cells (RTCs) which terminally differentiate into one of six neurons or Müller glia. The appearance of an RTC is called cell birth, and is defined as the time when cells complete their final M phase. Once RPCs are competent to generate differentiating RTCs, they gradually shift from symmetric divisions that produce two RPCs, to asymmetric production of one RPC and one RTC, completing their role as dividing cells with symmetric production of two RTCs, which is mostly over by ~ post-natal day 8 (P8) in mice (Fig 1.1) (Cayouette et al. 2006). Cell birth (i.e. cell cycle exit) is linked to differentiation, and several markers, such as RA4 in chick ganglion cells (Waid and
McLoon 1995), Brn3b (Pou4f2) in mouse ganglion cells (Pan et al. 2008), and Ptf1a in mouse amacrine cells (Fujitani et al. 2006) are thought to be activated very soon after 11 exit. However, whether cell cycle exit always precedes and thus might be required for a new differentiation program will be disputed here.
Classic retroviral lineage studies revealed that any single RPC can generate clones with multiple cell types, implying that these cells are remarkably flexible (Turner and
Cepko 1987; Turner et al. 1990). There are constraints, in that early or late RPCs differentiate into limited cell types: ganglion, horizontal, cone and amacrine cells are
“early” while rods, bipolar cells and Müller glia are primarily “late” cell types (Fig 1.1).
Early RPCs cannot generate late born cells unless they continue to divide and change competence, upon which they become unable to generate early born cell types (Cepko et al. 1996). Numerous transcription factors combine to define competence, and it is believed that after the “watershed” of cell birth is crossed, additional transcription factors are induced that cooperate with competence factors to carve out the transcriptomes of particular differentiating retinal cell types. These factors, many of which are bHLH, forkhead, or homeobox proteins, are reviewed in detail elsewhere (Hatakeyama and
Kageyama 2004; Kumar 2008). Here, we will discuss how cell cycle regulation is coordinated with cell birth. We will begin with some general features of cell cycle control focusing on proteins known to have major roles in governing exit, then we will discuss how these factors affect retinal development, and we will end with a discussion of how the birth program, governed by the contest between Notch signaling and neurogenic transcription factors, is integrated with cell cycle regulation.
12 1.1.1 A few basics of cell cycle regulation
The typical cell cycle has four distinct phases. Chromosomes are duplicated in S (DNA
synthesis) and distributed equally between two daughter cells in M (mitosis). During G2
(gap), which separates S and M, DNA is checked for errors. During G1, which separates
M and S, cells grow and respond to mitogens or inhibitory cues that determine whether to
enter S-phase or not. Two-thirds of the way into G1 cells cross a restriction point (R)
after which they no longer require mitogens to stay in the cell cycle and will continue
even if serum is withdrawn (Fig 1.2). Cells can be driven out of G1 into reversible G0 by
serum starvation or contact inhibition, or into irreversible G0 by terminal differentiation
or senescence.
Fluctuating Cyclin levels determine Cyclin dependent kinases (Cdk) activity.
Cyclin D-Cdk4/6 complexes form in G1, Cyclin E-Cdk2 complexes act at G1-S, Cyclin
A-Cdk2 complexes drive S, and Cyclin B-Cdk1 complexes drive M. However these
complexes may act in other phases and, as discussed later, Cyclins and Cdks show
considerable functional redundancy (Malumbres and Barbacid 2005; Musgrove 2006).
Treatment of resting cells with mitogens induces Cyclin D1 transcription through Ras-
Raf-Erk mediated activation of Ap1/Ets (Albanese et al. 1995; Winston et al. 1996; Aktas
et al. 1997). Gsk3β-mediated Thr phosphorylation of Cyclin D reduces stability and
nuclear localization, mitogens also activate the Ras-Pi3k-Akt pathway which
phosphorylates and inactivates Gsk3β, doubling Cyclin D1 half life and promoting
nuclear translocation (Diehl et al. 1997; Diehl et al. 1998; Lin et al. 2006). D-Cdk4/6
complexes phosphorylate Rb proteins on multiple sites. The D-Cdk4/6 complex also
13 titrates the Cip/Kip family of Cdk2 inhibitors (CKI), discussed more below (Sherr and
Roberts 1999).
Rb inhibits division in two major ways. First, it binds the “activating” E2f
transcription factors (E2f1, 2 and 3a), which are potent inducers of genes that positively
regulate the cell cycle (e.g. Cyclins E and A) and are necessary for the nuts and bolts of
DNA replication (e.g. PCNA, RRM1) (van den Heuvel and Dyson 2008). Rb binding
quenches E2f activity, and recruits silencing cofactors to permanently shut down
“proliferation” genes in terminally differentiating or senescent cells (Burkhart and Sage
2008). Rb also binds E2f3b and E2f4, whereas its relatives p107 (Rbl1) and p130 (Rbl2)
preferentially bind E2f4 and 5, which are thought to be “repressive” E2fs that primarily
mediate inactivation of target genes (although E2f3b can mimic some functions of E2f3a
(Chong et al. 2009a)). However, p107/p130 can switch to binding activating E2fs if E2f4
is missing (Lee et al. 2002). E2fs 6-8 do not bind the Rb family of “pocket proteins” but
inhibit transcription by recruiting other corepressors (Trimarchi and Lees 2002). The
extent to which E2f target gene induction involves direct activation (mediated by
activator E2fs) versus derepression (loss of pocket proteins from repressor E2fs) is not completely resolved.
Second, Rb, but not p107/p130, binds the Cdh1 subunit of Anaphase Promoting
Complex or Cyclosome (APC/C) (Binne et al. 2007). APC/C degrades securin and
cyclins to permit passage through and escape from M phase, but in G1 it degrades Skp2,
part of another E3 ubiquitin ligase (SCFSkp2) that degrades Cip/Kip CKIs to promote
Cdk2 activity. Rb binds Skp2 (Ji et al. 2004), presenting it for destruction to APC/C
14 (Binne et al. 2007), thus preventing degradation of Cip/Kip CKIs and blocking Cdk2
action.
Rb phosphorylation weakens binding to E2f, resulting in induction of Cyclin E,
Cdk2 activation, further Rb phosphorylation, and induction of Cyclin A, a sequential
process that is necessary for cell cycle progression (Lundberg and Weinberg 1998;
Harbour et al. 1999). Cyclin D function is dispensable in Rb-deficient cells (Lukas et al.
1995), but Cyclin E is required for division even in the absence of Rb, indicating they
have other targets (Ohtsubo et al. 1995). Indeed, independent of Cdk2 activation, Cyclin
E promotes loading of MCM proteins onto origins, and E-Cdk2 phosphorylates this
complex to trigger DNA replication (Blow and Dutta 2005; Geng et al. 2007). Cyclin E over-expression can drive the cell cycle even when E2f activity is blocked (Lukas et al.
1997). Apart from unleashing E2f, Rb phosphorylation also releases it from both Skp2
and Cdh1, thus activating Cdk2 by a second route (Ji et al. 2004). This positive feedback
loop allows cells to pass R and enter S; indeed E2f1 can behave as a “bistable switch” to
drive this irreversible transition (Fig 1.2) (Yao et al. 2008).
There are two distinct CKI families that bind and inhibit Cyclin-Cdks (Sherr and
Roberts 1999). The Ink4 family, which includes p16Ink4a, p15Ink4b, p18Ink4c and p19Ink4d
(encoded by Cdkn2a/b/c/d, respectively) inhibits Cdk4/6. Ink4 CKIs act upstream of Rb-
E2f and need Rb plus p107 or p130 (Bruce et al. 2000), and E2f4 or E2f5 (Gaubatz et al.
2000) to block division. The Cip/Kip family of CKIs, which includes p21Cip1, p27Kip1 and
p57Kip2 (encoded by Cdkn1a/b/c, respectively) inhibit Cyclin A/E-Cdk2 and CyclinB-
Cdk1 and can act downstream of Rb. In the complete absence of the Rb family, MEFs
15 lack a G1 restriction point, and progress even if mitogens are withdrawn, but these cells
arrest at G2 due to the combined action of Cip/Kip CKIs and p53 (Foijer et al. 2005).
Inhibitory mitogens, such as TGFβ, block Cyclin D induction or inhibit its
activity by inducing the expression of Ink4 CKIs (Reynisdottir et al. 1995). TGFβ also
inhibits progression by triggering nuclear translocation of an E2F4/E2f5- p107-Smad3
complexes that associate with Smad4 protein, then bind and silence the c-Myc promoter
through a Smad-E2f element (Chen et al. 2002).
In summary, Rb and Cip/Kip CKIs cooperatively inhibit division by constraining
E2f and Cdk2 mediated induction of S-phase gene transcription and replication origin
firing, respectively. Rb and p27Kip1 cross talk positively by promoting Skp2 degradation
and blocking Rb phosphorylation, respectively. Rb phosphorylation by sequential
Cyclin-Cdk action, Cip/Kip sequestration by Cyclin D-Cdk4/6, and Cip/Kip degradation by free and stabilized Skp2 activate E2f and Cdk2. E2f and Cdk2 cross talk positively by inducing Cyclins and phosphorylating Rb, respectively, This dual axis triggers the
production and/or activation of the components needed for DNA replication (Fig 1.2).
1.1.1.1 Role of some core cell cycle components in retinal development
The above scheme is based mainly on studies in cultured cells, and as much as it is
convenient to think of the G1/S progression in these terms, it should not be considered
invariable. For example, although cyclin D-CDK4/6 complexes are sufficient to drive
proliferation, they may not be the only factors to mediate the cell’s response to mitogens
since fibroblasts lacking all three cyclin D proteins or both Cdk4 and 6 can still respond
16 to serum and enter S phase albeit with lower efficiency (Kozar et al. 2004; Malumbres et
al. 2004). Also, while in vitro studies focus on exit from quiescence (G0 into G1-S-G2-
M), the reverse flow (from M to G0) is the relevant direction in terminal differentiation
(i.e. retinal development). Nevertheless, the model provides a helpful framework within which to consider the role of a few selected cell cycle regulators in RPC division and exit.
1.1.1.2 Cyclins and Cdks in retinal development
RPCs express higher levels of Cyclin D1 than any other embryonic tissue (Sicinski et al.
1995). D1 absence causes severe retinal hypocellularity (Fantl et al. 1995; Sicinski et al.
1995). Down-regulation/inhibition of D1 during differentiation is important since ectopic
expression in differentiating photoreceptors prevents normal cell cycle exit (Skapek et al.
2001), mimicking the effect of pocket protein loss (Chen et al. 2004; MacPherson et al.
2004; Ajioka et al. 2007). The large induction in p27Kip1 protein translation in newborn
retinal neurons no doubt out-competes any remaining D1 in these cells (Levine et al.
2000; Dyer and Cepko 2001a; Green et al. 2003; Lee et al. 2006). D1 loss reduces RPC
division beyond E16.5 but not earlier, suggesting that the earliest phase of RPC
expansion is D1-independent (Sicinski et al. 1995). This delayed requirement for D1
may reflect the gradual increase in Rb and p107 expression in RPCs during development
(Spencer et al. 2005; Donovan et al. 2006). Consistent with the role of D cyclins in
inactivating Rb and sequestering CKIs (see above) the D1 null retina has
hypophosphorylated Rb, no CyclinE-Cdk2 activity and hypocellularity is rescued when
p27Kip1 is also missing (Geng et al. 2001; Tong and Pollard 2001; Landis et al. 2006). A 17 Cyclin D1KE point mutant binds but fails to activate Cdk4/6 and partially rescues the D1-
null retina (Landis et al. 2006). Like normal D1-Cdk, D1KE-Cdk complexes sequester
p27Kip1, and consequently both Cdk2 activity and Rb phosphorylation are increased in
D1KE versus D1-null retinas (Landis et al. 2006).
Importantly, the roles of D1 in the retina development are likely not limited to the
cell cycle roles described above. Recent genome-wide location analysis (chromatin immunoprecipitation coupled to DNAmicroarray) showed that during the mouse development D1 binds to promoters of a large number of genes (Bienvenu et al. 2010).
Importantly in the retina D1 binds the promoter of Notch1 and controls its expression by recruiting CREB binding protein (CBP) histone acetyltransferase. In D1-null retina, CBP recruitment and histone acetylation of the Notch1 promoter are reduced, leading to decreased expression of Notch1 which likely contributes to the hypocellularity of D1-null
retinas (Bienvenu et al. 2010).
The D family has two other members, D2 and D3. The defect in D1 null mice is
rescued in mice expressing D2 from the D1 locus, so they are interchangeable in this
regard (Carthon et al. 2005). Normally however, D2 is not expressed in the retina and
while D3 is expressed in Müller glia (Dyer and Cepko 2000b), it is present at very low
levels in RPCs (Sicinski et al. 1995; Geng et al. 1999). While D3 is induced in D1 null
retina, there is no induction of D2 (Tong and Pollard 2001) and D1/D2 or D1/D3 null
retinas do not show a more severe phenotype than the D1 knockout (Ciemerych et al.
2002). The loss of p27Kip1 in D1-null retinas also does not affect D2 and D3 expression
(Geng et al. 2001). Triple null mice survive to ~E15 and, consistent with the idea that D
18 cyclins are not required for early RPC division the early triple null retina also appears
WT (Kozar et al. 2004).
The Cyclin E family consists of E1 (formerly E) and E2 (Lauper et al. 1998), and
both are expressed in retina (Lauper et al. 1998; Ochocinska and Hitchcock 2008;
Trimarchi et al. 2008). Apart from a spermatogenesis defect in E2 null males, E1 and E2
null mice are normal (Geng et al. 2003; Parisi et al. 2003). E1-/-;E2-/- mice die around
mid-gestation (~E10) due to endoreduplication (repeated S with no M-phase) defects in
trophoblasts, that perturb placenta development (Geng et al. 2003; Parisi et al. 2003). By
using tetraploid WT blastocysts, which form a normal placenta but do not contribute to
the embryo proper, E1-/-;E2-/- embryonic stem (ES) cells injected into these blastocysts
could, in about half the cases, generate a normal embryo that, like WT ES cells in this
approach, survive to birth (Geng et al. 2003). Half of E1-/-;E2-/- embryos die with cardiac
defects, all megakaryocytes (like trophoblasts) show defective endoreduplication, and
while MEFs could divide normally, they were unable to exit quiescence (Geng et al.
2003; Parisi et al. 2003). Importantly, no neural phenotypes were reported in E1-/-;E2-/-
embryos (Geng et al. 2003), suggesting that Cyclin E functions are redundant in RPCs,
perhaps due to the combined actions of Cyclins D1 and A.
Remarkably, knocking in Cyclin E1 to the D1 locus partially rescues the Cyclin
D1-null retinal phenotype with retinal thickness reaching ~75% WT (Geng et al. 1999).
Rb phosphorylation is only slightly higher than in the D1 KO suggesting that low levels of Rb phosphorylation are sufficient to permit RPC division. Indeed, Cyclin E can over- ride a cell cycle block induced by over-expressing a mutated version of Rb lacking most
19 of its phosphorylation sites (Lukas et al. 1997). The Cdk-independent role of Cyclin E in
loading replication origins may be important in this context (Geng et al. 2007).
Like Cyclins, there is also considerable redundancy among Cdks. Although
Cdk4-/- mice are 20% smaller at birth, show pancreatic islet cell hypotrophy, and null
MEFs divide more slowly due to elevated p27Kip1 (Rane et al. 1999; Tsutsui et al. 1999),
there are no obvious retinal defects (Pei et al. 2004). Apart from a mild hematopoietic
defect Cdk6-/- mice are normal, and Cdk4/6 redundancy in RPCs remains to be addressed
since double null mice die after E14.5 (Malumbres et al. 2004). Cdk3 is rarely mentioned
as it is defective in many mice and is thus dispensable (Santamaria et al. 2007). Cdk2 null mice appear normal except for a defect in meiosis, and deleting Cdk2 does not affect fibroblast division (Ortega et al. 2003). Cdk2/4/6 triple null mice die around E12.5 and retinal explant studies have not been attempted (Santamaria et al. 2007). The latter study also shows that Cdk1 (also called Cdc2), the most ancient cell cycle kinase best known for its role at G2/M in mammals, can substitute for other kinases to mediate much of the proliferation needed for embryogenesis. Moreover, it is required for the first division after fertilization (Santamaria et al. 2007). A conditional allele will be needed to study its role in later stages, including RPC expansion.
In summary, Cyclin D1 promotes RPC division both by phosphorylating Rb proteins and sequesterating p27Kip1, although early RPC expansion appears Cyclin D- independent. E cyclins, Cdk4/6, or Cdk2 each seem redundant, but further studies are required to tease apart overlapping Cyclin and Cdk roles and to determine if Cdk1 is
essential for RPC division.
20 1.1.1.3 The Rb family in retinal development
Rb is the tumor suppressor mutated in the familial cancer retinoblastoma (DiCiommo et
al. 2000; Pacal and Bremner 2006; Burkhart and Sage 2008). Rb and p107 proteins are
present in mouse and human RPCs as well as post-mitotic RTCs, whereas p130 seems to
be confined to the latter (Spencer et al. 2005; Donovan et al. 2006; Lee et al. 2006). As
expected, Rb loss triggers extra division (Chen et al. 2004; MacPherson et al. 2004;
Zhang et al. 2004). However, Rb and p107 seem to be inactive in RPCs as removing Rb or both Rb and p107 in mouse retina does not affect the number of M-phase or Chx10+
RPCs (Chen et al. 2004; MacPherson et al. 2004). Inactivation of Rb in RPCs may be due to the extremely high levels of Cyclin D1 (Sicinski et al. 1995) and, as noted above,
Rb is hypophosphorylated in D1 null retinas. The irrelevance of Rb in controlling RPCs is a hard pill to swallow in view of multiple in vitro over-expression experiments implying that Rb tempers the expansion of all dividing cells. Yet Rb null ES cells divide normally, and Rb null embryos are not enlarged despite minimal apoptosis (Wu et al.
2003), and the same is true of p107/p130 null embryos, and there is not much developmental compensation (i.e. induction of Rb relatives) in either case (Jiang et al.
1997). Moreover, Rb loss does not impact much of Xenopus embryonic development, again implying that it is already inactive (Cosgrove and Philpott 2007).
Instead of tempering progenitor expansion, Rb is employed mainly to promote permanent cell cycle exit, such as in terminally differentiating or senescing cells, or to execute arrest in DNA-damaged cells (Pacal and Bremner 2006; Burkhart and Sage
2008). Thus, in contrast to the undetectable effect of Rb loss in RPCs, there is a dramatic effect in differentiating RTCs. RTCs missing Rb, or Rb and one or more of its relatives, 21 divide ectopically (Chen et al. 2004; Ajioka et al. 2007; MacPherson et al. 2007). Some cell types use apoptosis to defend the tissue from cancer, which is E2f1-dependent (Chen et al. 2007), but p53-independent (MacPherson et al. 2004). Indeed, most defects in the
Rb null retina are rescued in the Rb/E2f1-null retina, except for a notable cell cycle and cell-death independent differentiation defect in a subset of amacrine interneurons that is caused by E2f3a (Chen et al. 2007). Intriguingly, E2f3 perturbs migration in the Rb null forebrain, also independent of its cell cycle/death functions (McClellan et al. 2007).
Amacrine cells are one of a subset that survive loss of Rb proteins, and most escape tumorigenesis not by death, but by Rb-independent cell cycle escape (Bremner et al.
2004; Pacal and Bremner 2006). However, it is thought that rare Rb/p107 null or
Rb/p130 null amacrine cells, through post-pocket protein events that over-ride this arrest,
can form sporadic retinoblastoma (Robanus-Maandag et al. 1998; Chen et al. 2004;
Dannenberg et al. 2004; MacPherson et al. 2007). Human retinoblastoma does not
require p107 or p130 loss likely reflecting broader expression of these proteins in other
species. Intriguingly, long-term ectopic division of differentiation cells in fly tissues also
requires disruption of multiple cell cycle regulators (Buttitta et al. 2007) Ectopically
dividing horizontal cells are also resistant to apoptosis, but they are better protected against transformation and require loss of Rb, p130 and one allele of p107 to form tumors
(Ajioka et al. 2007). Natural resistance to apoptosis is an attractive feature for a cancer cell-of-origin, especially one like retinoblastoma that requires fewer rate-limiting events than adult cancers (Bremner et al. 2004; Trinh et al. 2004; Pacal and Bremner 2006).
Several post-Rb events have been identified in human retinoblastoma (Corson and Gallie
22 2007) and likely facilitate progression past a benign “retinoma” state which ends in
senescence (Dimaras et al. 2008).
In summary, pocket proteins apparently do not to regulate the cell cycle in normal
RPCs, but are poised to act in differentiating RTCs where Rb is required to quench E2f1
activity. p107 and p130, although non-essential, act as backups when Rb is removed (Fig
1.3). The CKIs p19Ink4d and p27Kip1 may play an important role in activating Rb proteins in differentiating RTCs (Cunningham et al. 2002; Lee et al. 2006). Pocket protein loss creates a dangerous state where ectopically dividing RTCs risk neoplastic transformation.
This risk is countered in some cells by apoptosis and others by Rb-independent means of cell cycle exit.
1.1.1.4 Ink4 CKIs and p19Arf in retinal development
Cdkn1c encoding p18Ink4c is expressed weakly in the embryonic NBL (Trimarchi et al.
2008), but is dispenable for retinal development (Pei et al. 2004). Cdkn2b, which
encodes p15Ink4b, lies adjacent to Cdkn2a, which encodes two transcripts that have
distinct first but shared downstream exons and encode p16Ink4a and the p53-activating
protein p19Arf (p14ARF in humans) from different reading frames (Gil and Peters 2006).
p15Ink4b expression has not been reported in the retina and while deleting both Cdkn2b/2a
loci (which removes all three proteins) renders mice extremely susceptible to
tumorigenesis (Krimpenfort et al. 2007), eye defects in addition to those seen in Cdkn2a
or p19Arf null mice were not described. p19Arf is expressed in embryonic vitreal pericytes
where it represses PDGFRB expression independent of MDM2 or p53, limiting
expansion of these endothelial support cells, thus its absence triggers abnormal expansion 23 and severe defects in the adult eye (McKeller et al. 2002; Martin et al. 2004; Silva et al.
2005; Thornton et al. 2007). As expected, Cdkn2a null mice, lacking both p16Ink4a and p19Arf, also have this defect, but no obvious retinal defects, consistent with the absence of
these proteins in RPCs. The fourth Ink4 protein, p19Ink4d, is encoded by Cdkn2d. Its
expression pattern is consistent with a role in facilitating cell cycle exit and, indeed, null
mice show abnormal division followed by elevated apoptosis (Fig 1.3) (Cunningham et
al. 2002). The defects may be a milder version of those seen in the absence of Rb or
following over-expression of E2f1 or Cyclin D1 in differentiating retinal neurons (Lin et
al. 2001; Skapek et al. 2001; Chen et al. 2004).
1.1.1.5 Cip/Kip CKIs in retinal development
Some p21Cip, encoded by Cdkn1a, is expressed in the WT retina, which increases in the
absence of Rb, implying a context-specific role in retinal cell cycle control (Chen et al.
2007). p21Cip absence alone does not affect retinal development, so it would be
interesting to know its effect when combined with Rb loss.
In the embryonic retina, a few cells express p57Kip2 (e.g., 3%, E14.5) and its loss
triggers extra division, which may reflect extra RPCs and/or ectopically dividing RTCs;
the latter fits the observation that p57Kip2 is expressed in late G1 or G0 (Dyer and Cepko
2001a, 2001b). Expression ceases around P0, and is reactivated in a subset of post-
mitotic amacrine cells, consistent with a role for p57Kip2 in differentiation.
The most influential Cip/Kip CKI in the retina is p27Kip1, as suggested by its
broader expression pattern in mouse and human retina (e.g, ~50%, E14.5) (Levine et al.
2000; Dyer and Cepko 2001a; Lee et al. 2006). p27Kip1 mRNA is high in mouse RPCs 24 and low in differentiating RTCs but vice versa for protein, implying that rapid translation
and mRNA degradation parallels differentiation (Green et al. 2003). In human RTCs,
p27Kip1 expression seems in general to precede Rb protein expression, suggesting that a
dual wave of inhibitors is employed to shut division down (Lee et al. 2006). p27Kip1 is
thought to be induced at G2 in RPCs, earlier than p57Kip2, consistent with an early role in
promoting cell cycle exit in newborn RTCs (Dyer and Cepko 2001a). The Cdkn1b-null
retina has excess dividing cells at least until P10 (Levine et al. 2000; Dyer and Cepko
2001a), which could reflect extra RPCs and/or ectopically dividing differentiating RTCs
(Fig 1.3).
Some adult p27Kip1-deficient retinas contain focal hyperplastic lesions, possibly
due to reactive gliosis (Nakayama et al. 1996; Levine et al. 2000; Dyer and Cepko
2001a). These lesions are more severe when the Cdkn1b gene is replaced by an altered
protein (p27CK-) that cannot bind Cyclin-Cdk complexes (Besson et al. 2007). p27+/CK- mice do not get retinoblastoma, but do develop lung tumors (Besson et al. 2007). The increased phenotypic severity in p27+/CK- mice relative to p27Kip1-null mice reveals that
p27Kip1, when freed from Cyclin-Cdks, has a dominant disruptive function (Besson et al.
2007). This activity may relate to the cytoplasmic role of p27Kip1 in regulating the Rho-
Rock pathway which, intriguingly, is also targeted by p21Cip and p57Kip2 (Besson et al.
2008). p27Kip1 has other cytoplasmic activities such as binding the microtubule regulator
Stathmin (Baldassarre et al. 2005). p21Cip and 27Kip are distributed between the
cytoplasm and nucleus in mouse retina (Chen et al. 2007).
Cell cycle defects in the p19Ink4d or p27Kip1-deficient retinas are enhanced when
both are missing, consistent with cooperative Rb activation Cdk inactivation to promote
25 cell cycle exit (Cunningham et al. 2002). Apoptosis and dysplasia is also more severe.
As in the Rb-/- retina (MacPherson et al. 2004), deleting p53 removal does not rescue apoptosis, yet surprisingly it does reverse the dysplasia (Cunningham et al. 2002).
1.1.1.6 E2fs in retinal development
Multiple E2f mRNAs have been detected in the retina (Dagnino et al. 1997; Chen et al.
2007), and protein expression has been confirmed for E2f1 as well as both a and b isoforms of E2f3 (Chen et al. 2007). Deleting E2f1 slows RPC division ~2-fold, whereas
E2f2/E2f3 loss has no effect (Chen et al. 2007; Chen et al. 2009), suggesting redundancy.
Strikingly, E2f1-3 null retinal progenitor cells still divide although division in these retinas is reduced by ~40% at E14.5 and ~75% at P0. We attribute this effect to functional interchangeability with Mycn (Chen et al. 2009). The superior role for E2f1 in ages>4259- rs from MEFs where E2f3 is more important (Wu et al. 2001). As noted above, E2f1 drives ectopic division in differentiating Rb-/- RTCs (Chen et al. 2007) (Fig 1.3) and
transgenic E2f1 expression in photoreceptors also impairs cell cycle exit (Lin et al. 2001).
E2f3a perturbs differentiation in some Rb-/- amacrine cells, which is similar to its effect in the Rb-/- forebrain (McClellan et al. 2007). E2f4 loss affects Shh expression in the
telencephalon, but its role in the retina, as for other E2fs, awaits further study
(Ruzhynsky et al. 2007).
26 1.2 Separating Rate, Differentiation programs and Exit
The above overview summarizes how cell cycle regulators affect RPC expansion and
RTC cell cycle exit. Now we turn to how the factors that promote initiation of
differentiation programs influence the cell cycle machinery. The topic is, admittedly,
confusing because individual regulators can have multiple functions and, even worse, different functions in different contexts. To facilitate discussion, we will define (what we see as) separate activities that combine to influence cell number and cell type. Here, an
“activity” is not the same as a molecule, as the latter may have many activities engaged
together or separately at distinct times during development. Five separate activities are
particularly important to distinguish in this model:
Activity I: Rate. This activity affects the rate of cell expansion by adjusting cell
cycle length. It may alter organ size or the time it takes to reach the size limit.
Surprisingly, it seems that slowing rate is not the key to driving birth (see below).
Activity II: Competence. This activity influences the ability of RPCs to generate certain cell types. It is permissive, but not instructive in that it creates potential, but it is not sufficient to force cell birth on its own.
Activity III: Interkinetic Nuclear Migration (INM). RPCs are connected by processes to the apical (also called outer or ventricular) and basal (also called inner or vitreal) surface of the neuroepithelial sheet. Their nuclei move during the cell cycle such that they are located on the apical surface in M-phase, traverse S-phase in the basal 2/3rds of the tissue, with G1 and G2 occurring between the two (Fig 1.1). An important inbuilt feature here is polarity which, as we shall see has important effects on the timing of birth.
27 Activity IV: Differentiation program initiation. This activity instructs an RPC to switch on a new epigenetic program which facilitate differentiation. The type(s) of cell generated depends on activity II.
Activity V: Exit. This activity ensures that differentiation is coupled to cell cycle exit.
Other activities include VI: Terminal Differentiation and VII: Death. The former reutilizes the factors that conferred competence plus new factors induced by cell birth to activate the gene expression program needed to build a specific cell type. Apoptosis is engaged to prune excess cells, “weak” cells and/or aberrantly dividing neurons. Also, as discussed bellow, apoptosis can be engaged to regulate tissue size when growth is perturbed through “cell competition.” Although this process has been studied mostly in
Drosophila, the genes involved in this process are conserved in mammals and likely play roles in cancer initiation (Moreno 2008).
1.2.1 Coupling INM to Cell Birth: “I need to get away”
To place cell cycle in the context of cell birth we will first discuss a recently elucidated model of retinal differentiation that emphasizes the role of INM in deciding fate. By tracking fluorescent RPCs undergoing INM, it was discovered that deeper basal migration correlates with cell birth (Baye and Link 2007). Cytochalasin B (CCB) treatment, which disrupts actin filaments and thus INM, causes premature neurogenesis in chick retina, consistent with a link between INM and birth (Murciano et al. 2002).
Deep migration was dependent on atypical protein kinase λ and ζ, key components of the
28 polarity complex that are associated with adherens junctions at the apical side (Baye and
Link 2007). How they influence INM distance is unknown. However, there is an explanation as to why deeper migration promotes cell birth. There is a higher concentration of Notch at the apical surface, and nuclei that migrate basally appear to escape its influence, achieving a status that allows them to resist Notch when they revisit the apical side (Del Bene et al. 2008). These data fit observations in other species.
Conditional deletion of Notch1 in mice or inhibition of the pathway in chicks triggers premature differentiation (Jadhav et al. 2006a; Yaron et al. 2006; Nelson et al. 2007).
Conversely, activating the Notch pathway maintains cells in a progenitor like state (Bao and Cepko 1997; Furukawa et al. 2000; Scheer et al. 2001; Jadhav et al. 2006b). Deleting
Hes1, a transcriptional repressor that mediates Notch signaling, has a similar effect
(Tomita et al. 1996). Hes proteins repress transcription of neurogenic bHLH factors (e.g.
Math5, Mash1, NeuroD etc) and over-expression of the latter is alters fate (Perron and
Harris 2000; Kageyama et al. 2008). Thus, the key to cell birth may be physical distance from the inhibitor Notch. Leaving Notch behind seems to permit induction of neurogenic factors to the levels necessary to initiate the cell birth transcriptional program after the
RPC traverses M for the last time.
A further compelling aspect of this model is that cells in the basal portion of the retina are in S-phase (Fig 1.4) and there is evidence that this is where critical decisions regarding fate occur. For example, when ferret cortical progenitor cells in their final S- phase are transplanted to an older brain they switch fate (McConnell and Kaznowski
1991), and cells in the retina can switch fate right up to M-phase (Belliveau and Cepko
1999). These results show that at least some fate decisions are made late in S/G2,
29 although they do not specifically address whether this is when the general cell birth/anti-
Notch signal accumulate. However, other studies have shown that pro-neurogenic bHLH factors are up-regulated in S-phase (Matter-Sadzinski et al. 2005). Nevertheless, it should be emphasized that when, exactly, individual retinal cells choose their final fate is moot. An important starting point is to carefully characterize the timing of events around cell birth, which is the subject of chapters 2 and 3 in this thesis.
It is not clear how the apparent requirement for sampling distinct concentrations of Notch is recapitulated (or even necessary) in dissociated embryonic rat retinal cultures, where clones arise with the same numbers and proportions as seen in vivo (Cayouette et al. 2003). However, a two cell clone could establish gradients of membrane bound proteins even in vitro, and INM by definition creates an intracellular organelle gradient
whereby the nucleus escapes from the influence of Notch signaling at the opposite
surface. But the occurrence and relevance of such a gradient in vitro has yet to be
addressed. For the purposes of discussion, we will assume that the distance of nuclear
migration distance is the key to escaping Notch signaling and provoking cell birth.
1.2.2 Birth and Exit: “The Cog Model” vs. The Trigger theory”
The phrase “Cell cycle exit and differentiation are tightly coupled” is common, but what
does it mean? Are birth and differentiation functionally connected like two interlocked
cogs driving interdependent machines, or are they simply two events running in parallel?
The cellular wiring may be distinct in higher versus lower vertebrates. First, we consider
mammals.
30 Some evidence is consistent with the idea that cell cycle inhibition may be a
prerequisite for differentiation in the mouse retina. First, they are temporally coupled:
activation of the “differentiation transcriptome” is always accompanied by exit,
irrespective of cell type. Second, cell cycle lengthens as the retina matures and correlates
with the birth of later born cells (Alexiades and Cepko 1996). Third, the CKI p27Kip1 is up-regulated in G2 of the last cell cycle prior to cell birth (Levine et al. 2000; Dyer and
Cepko 2001a). Fourth, over-expression of p27Kip1 in single RPCs at P0 drives cell cycle
exit, generating smaller clones of cells with increased proportions of earlier born rods and
fewer later born bipolar and Müller glia (Dyer and Cepko 2001a). Fifth, olomoucine, a
Cdk inhibitory drug, lengthened the cell cycle and promoted early differentiation in the mouse E9.5 telencephalic neuroepithelium (Calegari and Huttner 2003). And fifth, in the
zebrafish studies discussed above there was a tentative correlation between lengthening of the cell cycle and cell birth; however, the average cell cycle length across a population of RPCs was not predictive, rather length only correlated with birth when two siblings derived from the same parent RPC were compared.
The temporal proximity of exit and differentiation and correlations between birth and cell cycle lengthening or the appearance of cell cycle inhibitors do not distinguish whether exit is required for differentiation, or whether they are simply parallel events.
Also, while CKI over-expression or pharmaceutical Cdk inhibition indicate that exit can drive differentiation, they do not show that this is the physiologically relevant means of triggering differentiation in vivo. Finally, the correlation between cell cycle timing and birth might be a side-effect of deeper basal migration to escape Notch; i.e., the critical parameter may be distance, not time. Strikingly, there was no correlation between time
31 nuclei lingered in the maximal basal position and cell birth, rather it was the depth of
migration that mattered (Baye and Link 2007). It is also noteworthy that several
mutations that lengthen cell cycle time, including loss of E2f1, Cyclin D1, or Chx10, do not drive the birth of early born neurons (Burmeister et al. 1996; Dhomen et al. 2006;
Livne-Bar et al. 2006). In the latter two cases retarded RPC division rate (activity I above) is due to elevated p27Kip1 but this physiological increase was not enough to drive
early birth since the proportion of dividing and differentiating cells remains unchanged
(Green et al. 2003). Thus, it would appear that division can be slowed considerably
without triggering birth. Conceivably, artificial high CKI over-expression pushes the cell
cycle below this low threshold, forcing birth (we speculate how this might occur later).
Notably, mutations that do force early cell birth and retard the cell cycle inhibit the Notch
pathway (Tomita et al. 1996; Bao and Cepko 1997; Furukawa et al. 2000; Jadhav et al.
2006b; Jadhav et al. 2006a; Yaron et al. 2006; Nelson et al. 2007; Wall et al. 2009),
which brings us full circle to the problem of deciding if cell cycle inhibition is the
physiological mechanism that overcomes Notch (exit is required for differentiation) or
whether Notch inhibition triggers differentiation and cell cycle inhibition in parallel pathways.
A key issue, not addressed by over-expression assays, is whether exit is required for birth, and there is genetic evidence that it is not. Knocking out Rb impairs cell cycle exit in differentiating neurons, which in both the retina and forebrain is due to unleashed
E2f1 activity (Chen et al. 2007; McClellan et al. 2007). Ectopic division is exasperated if both Rb and its relatives are missing (Chen et al. 2004; MacPherson et al. 2004; Ajioka et al. 2007). Critically, Rb loss does not prevent differentiation (Fig 1.5). For example, in
32 the Rb-/- forebrain, ectopic division does not block the appearance of migrating βIII-
tubulin (Tubb3)+ neurons (Ferguson et al. 2002). In the retina, loss of Rb or even Rb and
p107 does not block production of any early cell type (e.g. Trβ2+ cones, Crx+
photoreceptors etc) and differentiated horizontal cells divide in mice lacking 5/6 of the
Rb/p107/p130 alleles (Chen et al. 2004; Ajioka et al. 2007). Similar results have been
observed in the Rb-null inner ear hair cells (Sage et al. 2005). Over-expression of Cyclin
D1 or E2f1 in newborn photoreceptors also disrupts exit, but not the photoreceptor gene
expression program (Lin et al. 2001; Skapek et al. 2001). There is also ectopic division
of differentiating cells in the p19Cdk4d-/-;p27Kip1-/- retina, but again all the usual cell types
are generated and there is no obvious shift to late born cells (Cunningham et al. 2002).
In addition to the genetic evidence discussed above, other observations support
the notion that differentiation programs might not strictly depend on the cell cycle status.
First, transcriptional profiling of RPCs revealed that mRNA transcripts of many factors
involved in the differentiation of retinal neurons are already present in RPCs. This
suggests that many RPCs are perhaps poised to became neurons come the end of the last
M-phase, and that subsets of RPCs may even be biased to producing specific retinal cell
types (Trimarchi et al. 2008). Second, and in agreement with the first observation, we detected the expression of proteins that were previously associated only with post-mitotic ganglion neurons already in RPCs in the late S and G2/M phases of the cell cycle
(Chapter 2).
Together, these data suggest that differentiation can be uncoupled from exit. As discussed earlier, there are associated costs of failing to couple differentiation to exit (e.g. apoptosis, tumorigenesis), but irrespective, initiating the “differentiation program” (i.e.
33 inducing genes unique to differentiating neurons) does not depend on cell cycle exit.
Coupled with the INM-Notch birthing model discussed above, these data suggest that
induction of neurogenic factors is the major requirement for differentiation, and that cell
cycle exit is a parallel to rather than necessary for this event. In this scenario (“The
Trigger theory”) birth and differentiation are like two runners in a race. One trigger (the starting gun) induces both events (running towards the finish line), while numerous variables affect the end result (fitness, injuries, equipment, weather etc). In an RPC, one signal (e.g., Notch) could set off two separate cascades that lead to exit and differentiation, while multiple variables might influence the timing of both events
(extrinscic and intrinsic activators and inhibitors). In this way the coupling of exit and differentiation would be more plastic such that across development the same differentiation event could occur at significanlty distinct times, even either side of M phase (Fig 1.6).
But if exit is not used to drive differentiation, why does artificial over-expression of CKIs trigger birth and what is the physiological purpose of this (presumed) connection from the cell cycle to Notch? Perhaps the cell cycle machinery is coupled to expression/activity of Notch pathway components, which is logical given that their shared objective is to ensure RPC expansion. It is known that Notch drives the expression of factors required for division, such as Cyclin D and Myc (Ronchini and
Capobianco 2001; Weng et al. 2006), so could there also be positive feedback? Indeed, the Hes family members Hes1 and Hey1 are E2f target genes (Vanderluit et al. 2007;
Hulleman et al. 2009). If the connection between the cell cycle machinery and Notch signaling components was highly redundant (e.g. any one E2f is sufficient to facilitate
34 expression) this may explain why complete arrest (e.g. total inhibition of E2f) promotes birth, but partially impairing cell cycle progression does not.
A notable exception to the general rule that blocking division promotes birth is the now classic experiment in which the S-phase inhibitors hydroxyurea and aphidicolin were used to block division in Xenopus embryos but CNS differentiation was virtually normal (Harris and Hartenstein 1991). These drugs do not act on regulators like E2f and so may not perturb feedback to the Notch pathway. It would be interesting to compare the effect of these drugs and CKI over-expression on Notch pathway activity.
Another explanation as to why CKIs may influence birth is their Cdk-independent functions. In addition to it’s ability to regulate Rho-ROCK activity, p27Kip1 also binds and stabilizes Neurog2 (Ngn2/Math4a/Atoh4) to facilitate neuronal differentiation in the cortex (Nguyen et al. 2006). Conceivably, excess NeuroG2 could favor early cell birth in the retina. Indeed, over-expressing the N-terminal portion of p27Ck- which binds
Neurog2 but not Cdks drove early neurogenesis in the cortex (Nguyen et al. 2006).
In summary, while in reality RPCs use distance to escape Notch (Baye and Link
2007; Del Bene et al. 2008), CKI over-expression may override Notch artificially, possibly through non cell cycle activities or because complete cell cycle blockade artificially down-regulates Notch pathway components. Notably, partial inhibition of the cell cycle (as in E2f1, D1 and Chx10 null RPCs) does not drive early birth perhaps because Notch components have multiple redundant links to the cell cycle (e.g. the redundant E2f family). This putative link fits the synergy between division and anti- differentiation signals. Links between lengthened cell cycle and earlier birth may be a
35 consequence of deeper migration to escape Notch, rather than a mechanism for driving
birth per se.
1.2.3 Birth and exit in frogs: “You walk, I’ll jump”
The above model may not apply in Xenopus. While over-expressing the neurogenic
bHLH factor Xath5 did not affect the timing of cell cycle exit, it redirected later born
bipolar cells to become ganglion neurons, an early cell type. p27Xic1 , unlike mammalian
p27Kip1 , has a cell differentiation activity that favours Müller glia genesis, but the N- terminal portion of p27 has been used to study cell cycle exit in isolation (although it is not yet clear if p27Xic1 like mammalian p27Kip1 also binds Neurog2) (Ohnuma et al.
1999). This version of p27Xic1 drove early exit and increased the ability of Xath5 to
promote ganglion cell genesis, whereas cyclin E delayed exit and prevented the Xath5-
induced switch to early born cells (Ohnuma et al. 2002). Cyclin E did not perturb
survival in this system. These data suggest an “exit-then-differentation” model,
contrasting the mouse model above where exit seems dispensable for birth. However, not
all the data fitted the exit-then-differentiation model since Xath5 + p27 did not increase
early born horizontal or amacrine cell proportions. Moreover, the Cyclin E vector was
not RPC-specific, and would likely drive ectopic division in RTCs, which may impair
marker expression in some neurons (e.g. ganglion cells?) and increase other cell numbers
by driving expansion of already born neurons (e.g. bipolar cells?). It would be useful to
assess ectopic division of differentiating cells in this system and to determine whether
removal of cell cycle inhibitors (e.g. Rb) mimics the Cyclin E effect. Another study
reported that Cyclin A-Cdk2 over-expression caused a shift to late-born cells (Casarosa et 36 al. 2003). Again though, the shift was not uniform: early born cones and ganglion cells
were decreased and late-born bipolar cells were increased, but early born amacrine and
horizontal cells and late-born Müller glia were unaffected. Conceivably, all cells were
born at the correct time, but ectopic division affected survival and or marker expression
differentially in distinct RTCs. If exit really is required for differentiation in frogs, then
perhaps it reflects other aspects of RPC wiring that are distinct in this species, e.g., while
Notch inhibits differentiation in both mice and frogs, it drives cell cycle exit in the latter,
but not the former (Dorsky et al. 1995; Ohnuma et al. 2002).
1.2.4 Coupling differentiation to cell cycle exit: “Let’s take the mystery tour”
Although differentation does not need exit (in mice), the two are paired, and blocking
Notch-Hes signaling triggers both early birth and differentiation, so how is this achieved?
In general, cell birth coincides with down-regulation of Cyclins and Cdks, contrasting the
high levels of neurogenic factors and CKIs (Dyer and Cepko 2000a; Levine et al. 2000;
Dyer and Cepko 2001a; Ohnuma et al. 2002; Barton and Levine 2008). An attractive model might be that fate determining transcription factors regulate CKI expression, e.g.
Six6 (XOptx2 in frogs (Zuber et al. 1999)) promotes RPC proliferation, represses
Ink4d Kip1 Kip2 Kip1 p19 , p27 P57 mRNA and binds directly the p27 promoter in vivo (Li et al.
2002). Pax6 also directly regulates CKI expression in the optic primordium, but unlike
Six6 it induces CKIs (Duparc et al. 2007). Other early determinants of eye/retinal
development, such as Rx and Lhx2 also regulate CKI expression directly or indirectly
(Hardcastle and Papalopulu 2000; Andreazzoli et al. 2003; Duparc et al. 2007; Tetreault
et al. 2008). These studies did not assess CKI protein levels, and while this could be a 37 mechanism for controlling division in pre-competent RPCs (before E11 in mouse), the
induction of p27Kip1 protein at cell birth in competent RPCs (beyond E11) is a
translational, not transcriptional, event (Green et al. 2003). Hes1 can also repress p27Kip1 transcription (Murata et al. 2005) but this appealing mechanism cannot explain how
Notch down-regulation is connected to p27Kip1 protein induction during the RPC-RTC
transition. Conceivably, CKI regulation switches from transcriptional to post-
transcriptional mechanisms after RPC competence is induced at ~E11. The idea that
RPC wiring is altered at this time is further supported by the observation that Pax6
inhibits division before but promotes expansion after RPCs become competent to
generate neurons (Marquardt et al. 2001; Duparc et al. 2007).
Multiple mechanisms control p27Kip1 translation (e.g. Elav proteins or miRNAs)
and protein stability (Chu et al. 2008), yet how these are deployed during retinal cell birth
is unclear. Intriguingly, active Notch1 induces Skp2 just as it does Hes1, and Skp2
promotes p27Kip1 protein degradation (Sarmento et al. 2005). Forkhead proteins can
increases p27Kip1 stability which, in the case of Foxo4 is through repression of a
proteosome subunit (Yang et al. 2005), but whether this is employed by factors like
Foxn4, which promotes amacrine and horizontal cell births (Li et al. 2004), is unknown.
Absence of Chx10 or Tlx4 triggers cyclin D1 down-regulation and p27Kip1 protein induction, but the mechanism is also unknown (Green et al. 2003). Notch inhibition causes rapid down-regulation of Chx10 and Tlx4, providing indirect links to p27Kip1 protein levels (Nelson et al. 2007). Regulation of p27Kip1 levels by Vsx2 must be context
specific as Vsx2 loss affects early, but not late stage RPC division (Burmeister et al.
1996; Livne-Bar et al. 2006).
38 The molecular links between birth and the p19Ink4d-Cyclin D-Cdk4/6-Rb-E2f pathway are also unclear. p19Ink4d levels are regulated both at transcriptional and post-
transcriptional levels (Matsuzaki et al. 2002; Yokota et al. 2004; Forget et al. 2008;
Katayama et al. 2008), but the mechanisms employed in the retina are unknown. E2fs are
expressed in neurons (Dagnino et al. 1997), and E2f1 activity in RTCs is blocked by Rb
(Chen et al. 2007). The dramatic drop in Cyclin D1 mRNA and protein levels in
differentiating cells (Barton and Levine 2008; Trimarchi et al. 2008) likely activates Rb
(and would also facilitate p27kip1 action), but although numerous pathways regulate D1 transcription (as well as translation and stability) (Musgrove 2006), the retinal mechanism is not known. The intracellular cleaved portion of Notch can activate the D1 promoter (Ronchini and Capobianco 2001), but surprisingly, pharmaceutical inhibition of
Notch for 8 hr in the retina induces Cyclin D1 mRNA (Nelson et al. 2007). D1 down- regulation may require intermediate repressors accumulating at that point.
In summary, while the link between Notch/Hes down-regulation and derepression of neurogenic bHLH genes is well established, the dots connecting Notch to induction of
CKIs/loss of Cyclins/Cdks have not been joined.
1.2.5 “We need a better map”
RPCs exhibit diverse transcriptomes (Trimarchi et al. 2008), making it difficult to decipher how individual cells decide to transit to a newborn RTC. In coupling birth to cell cycle exit Notch is likely the principal player, however other pathways such as Shh or Wnt signalling may also be involved. It has dual connections that block neurogenic factors needed for the birth program, and control cell cycle regulators that coordinate 39 birth with exit. Nuclear distance from Notch, achieved through INM and dependent on
polarity signals, is an intriguing mechanism through which the RPC may succumb to
birth and exit at the next mitosis. How this choice to migrate more basally is made, and
how it interferes with Notch signaling, is unclear. The cell cycle molecules that are
activated or down-regulated/repressed are the usual suspects (CKIs, Rb proteins, Cyclins,
Cdks, E2fs etc). However, we do not understand, in detail, how Notch down-regulation
connects to these factors. Rb protein, poised yet inactive in RPCs (perhaps through
Cyclin D1/Cdk4/6 mediated phosphorylation?), is likely activated through
dephosphorylation in newborn neurons following D1 down-regulation and CKI
induction. p27Kip1 activation is post-transcriptional, but how this occurs in the retina is
unknown. Despite temporal connectivity and CKI over-expression data suggesting that
birth might require exit, genetic studies counter this notion. Reducing the rate of division
does not force birth even down to quite a low threshold, and although exit (driven by
over-expressed CKIs) can force birth, it is not necessary for birth. Exit, however, is more than a convenient afterthought, since interfering with it risks tumorigenesis. Even then,
the vast majority of retinal cells evade that route, either by dying or by exploiting
considerable redundancy to exit the cell cycle.
Despite striking advances in the last decade towards understanding cell birth,
numerous gaps remain. And we have not even mentioned the debates around extrinsic
versus intrinsic regulation and the role of asymmetric cytokinesis. The results in the
following three chapters will revisit and reiterate the points raised in this section
regarding the loose connection between the cell cycle and cell birth. In Chapter 2, we
discuss the characterization of neuronal markers that allowed us to track the initial events
40 in the development of the first retinal neurons, the ganglion cells. We observed that markers once thought to be exclusively associated with post-mitotic ganglion cell neurons can in fact be observed prior to cell birth in a small subset of RPCs. These observations, in addition to the other evidence discussed above, helped us formulate “the
Trigger Theory”. One corollary of this theory is that across development the same differentiation event could occur at significanlty distinct times, even either side of M phase (Fig 1.6). In Chapter 2 we show that the onset of appearance of ganglion neuornal markers with respect to cell birth varies markedly within the course of several days. In
Chapter 3, we extend this analysis to markers recognizing other retinal neuorns. Using this knowledge we build a panel of markers that enable us to unequivovally distinguish post-mititoc neurons from proliferating RPCs. We use this panel of markers in Chapter 4 to quantify the extent of ectopic division of Rb and Rb/p107 null retinal RTCs in which differentiation and cell cycle progression have been uncoupled.
1.3 Cell Competition
The ultimate purpose of coordinating cell birth and cell cycle exit processes is to build tissue with the correct number of cells and correct proportion of cell types. In
Drosophila, one way in which tissue size is regulated is through the “cell competition” that can occur when growth is perturbed. In the imaginal discs, which give rise to much of the external tissues of the adult fly, cell competition coordinates growth and apoptosis and is required for consistent size regulation (de la Cova et al. 2004).
Cell competition was initially observed in the imaginal discs of D. melanogaster several decades ago as a phenotype whereby mutant slowly dividing, but viable, cells 41 were ousted from a population of more rapidly dividing cells (Morata and Ripoll 1975;
Simpson 1979; Simpson and Morata 1981). Current definition of cell competition posits
that the “loser” cells should survive when surrounded by cells of the same genotype.
Thus, this process is non-cell autonomous and depends on the presence of WT cells
(Moreno et al. 2002).
Cell competition in Drosophila is triggered by juxtaposition of slowly and fast
growing populations. In the original studies, the slowly growing cells had defects in cell metabolism due to the Minute ribosomal protein mutations and the hypomorphic dmyc allele mutants, and the winners were the healthy WT cells (Lambertsson 1998; Marygold et al. 2007). However, the slowly growing cells are not always the “losers” and the wild type cells don’t always win. Cells with inactivating mutations in the tumor suppressor scribbled (scrib) lose polarity and over proliferate, but this is curbed by the surrounding normal cells which kill the mutants via JNK-mediated apoptosis (Brumby and
Richardson 2003). In this way WT cells guard the tissue against ectopic growth.
However, the WT cells can also lose the battle against mutant cells. For example, mutants overexpressing dmyc turn into “super-competitors,” expand and kill the wild type neighbors (de la Cova et al. 2004; Moreno and Basler 2004). Other super- competitors are cells with inactivating mutations in the Salvador–Warts–Hippo pathway
(Tyler et al. 2007).
Since both Myc family proteins and Hippo pathway components are deregulated in human cancers (Vita and Henriksson 2006; Harvey and Tapon 2007; Saucedo and
Edgar 2007), cell (super)competition might also be involved in cancer initiation. Cell competition might serve as a protective mechanism against early lesions. For example,
42 cells with activated Ras survive poorly when surrounded by WT tissue (Land et al. 1983).
Conversely, super competitors might gain control over tissue by actively killing the
normal neighbors. Thus, cancer initiation might involve not only the gain of super-
competitive oncogenic properties by rogue cells, but also the failure of WT cells to
eliminate the mutant cells. In Chapter 4, we consider the potential role of cell
competition in constraining the effects of Rb loss in the retina.
1.4 The role of Rb in differentiation
Rb binds >100 proteins (Morris and Dyson 2001) and interacts with transcription factors
that promote differentiation of various tissues (Skapek et al. 2006). However, the role of
Rb in differentiation is controversial. Rb-null embryos do not survive past embryonic
day (E) 13.5–14.5 (Clarke et al. 1992; Jacks et al. 1992; Lee et al. 1992). Gross pathology
in the developing central and peripheral nervous systems and in erythrocytes were
thought to cause the embryonic lethality, and suggested key roles for Rb in mouse
development. However, in Rb-null mice which developed with WT placenta, lethality, neuronal cell death and the excess nucleated erythrocytes are rescued, suggesting that
these defects were mediated by Rb in placenta (Wu et al. 2003; de Bruin et al. 2003).
Despite this, some developmental abnormalities remain such as excess proliferation and
apoptosis in the lens and defects in skeletal myogenesis. Moreover, the rescued Rb-null
mice die shortly after birth due to lung failure that appears to be caused by skeletal
muscle defects. Rb interacts with MyoD in cultured cells and this is believed to be
crucial for the process of myogenesis (Novitch et al. 1996; Novitch et al. 1999).
However, whether Rb has a direct role in myogenesis in vivo or whether the muscle 43 defects in Rb-null mice are driven by deregulated activity of E2fs has not been fully resolved.
E2f1, 2, 3a and 3b are all expressed in differentiating myocytes (Asp et al. 2009), but defects in myogenesis are not rescued in Rb;E2f1-/- and Rb;E2f3-/- mice (Tsai et al.
1998; Ziebold et al. 2001). However, E2f1 functions might be redundant since the loss of
E2f1 does not affect myogenesis in vitro, and E2f3b appears to have Rb-independent roles in myogenic differentiation (Asp et al. 2009). Thus, only compound knock-out mice lacking multiple E2f family members will address the roles of the different E2fs in muscle development.
In addition to myogenesis, Rb has also been implicated in another aspect of mesenchymal development. Rb loss biases mesenchymal progenitors towards the adipogenic fate, which results in reduction of bone formation and increases the levels of brown fat (Calo et al. 2010). Further, the loss of Rb also converts differentiating osteoblasts into adipocytes. These functions of Rb appear to be E2f-dependent, as for example both Rb and E2f4 bind promoters of the master activator of adipogeneiss
(PPAR-γ) (Calo et al. 2010).
In the retina, the loss of Rb alone or together with p107 results in ectopic differentiation of neurons and cell death of several retinal neuronal cell types. The mature mutant retinas lack several neuronal subtypes and many of the surviving cells display differentiation defects (Chen et al. 2004; MacPherson et al. 2004) (also, see
Chapter 4). Strikingly, the loss of E2f1 rescues the proliferation and death, and virtually all proliferation defects ((Chen et al. 2007) and Chapter 4). A subset of Rb-null amacrine cells exhibit defects in the level and transport of synaptic proteins. Although
44 this defect was not rescued by the loss E2f1, it was suppressed by deleting E2f3 (Chen et
al. 2007). Together, our data raise the possibility that the only critical Rb activity in
retinal development is to quench E2f family protein function in RTCs. E2f1 might
trigger differentiation defects either because it directly regulates expression of differentiation transcription factors (e.g., E2f1 and E2f4 bind the promoter of the adipogenesis regulator PPAR-c (Fajas et al. 2002; Calo et al. 2010) and/or because of the secondary consequences of ectopic proliferation and/or the non autonomous effects related to death of neighbors. Irrespective, our data reveals that retina-specific factors do not require Rb/p107 to promote differentiation of retinal cells.
1.5 The Cell of Origin of Retinoblastoma
In Section 1.1.2.2, we discussed the critical roles of Rb family genes (Rb, p107, p130) in guarding retinal tissue against retinoblastoma. Briefly, the loss of Rb leads to retinoblastoma in humans, whereas mouse retinal tumors are triggered by the loss of Rb and p107 or p130 (Dyer and Bremner 2005). The information about tumor cell of origin is commonly based on cellular markers expressed in the given tumor. For example, previous immunostaining of human retinoblastomas indicated that many but not all samples contained photoreceptor (e.g., IRBP, rod and cone opsins) and glial (e.g., GFAP) cell markers. However markers that stain other retinal cell types or are not exclusive to one cell type were also present, e.g., Vsx2 (Chx10) (RPCs, bipolar cells and Muller glia),
Map2 (ganglion, horizontal, amacrine cells), Nestin (RPCs, Muller glia) (Gonzalez-
Fernandez et al. 1992; Nork et al. 1995; Sakata and Yanagi 2008). More recently, Xu et al., reported that a large number if not all human retinoblastoma samples express cone 45 markers, and cone-specific signaling circuitry is employed for tumor growth (Xu et al.
2009b). This study also showed that markers that label other retinal cell types were Rb+ and thus infiltrating WT cells (Xu et al. 2009b). In contrast, mouse retinal tumors stain positive for markers that are specific to amacrine or horizontal cells but not to photoreceptors (Robanus-Maandag et al. 1998; Chen et al. 2004; Dannenberg et al.
2004). This discrepancy might indicate that human and mouse retinal tumors have two different cells of origin. However, cancer by definition is a transforming event. Thus the presence of a certain cell type in a tumor may not indicate the tumor cell of origin but instead it may reflect the consequence of the transforming events and re-programming of cell fate.
The concept of dedifferentiation and cell fate reprogramming in cancer is not new, but how relevant it is in tumor initiation and progression is unknown. Oncogenes
(e.g., Myc, Sox2) drive reprogramming somatic cells into induced pluripotent stem cells
(iPSCs) while the tumor suppressor p53 inhibits this process (Takahashi and Yamanaka
2006; Takahashi et al. 2007; Yu et al. 2007; Kawamura et al. 2009). Importantly, several studies have also implicated Rb in cell fate control. First, the loss of Rb biases adipocytes towards becoming brown fat cell fate (Hansen et al. 2004). In agreement with this, a recent study showed that the loss of Rb can induce fate switch in differentiating osteoblasts towards becoming brown fat adipocytes. Critically, on p53-null background,
Rb loss favors the formation of hibernomas (fat tissue tumors) over osteosarcomas arising from p53-null cells (Calo et al. 2010). Second, C. elegans, mutations in Rb pathway revert the gene expression pattern of somatic cells to germ cell status (Calo et al. 2010).
Despite the unique roles of Rb proteins in the retina, the role of Rb in retinal cell fate
46 choices has never been reported. Strikingly, in Chapter 4 we present evidence for fate switch occurring in Rb/p107-null retinal cells and discuss the important implications for the cell-of-origin of retinoblastoma.
47 Peak of cell birth:
Cell types:
Ganglion Horizontal Cone Amacrine Rod Bipolar Müller neuron neuron neuron neuron glia Time E11.5 E13.5 E15.5 E17.5 P0 P3 P8
Cell cycle length: ~15rhs ~30rhs ~60rhs
Type of Division:
RPC RTCs
Fig 1.1 Retinal histogenesis. The upper schematic shows approximate periods of genesis for each of the seven major murine retinal cell types. The lower half depicts when symmetric production of two RPCs gradually switches to symmetric production of two differentiating RTCs. This switch is matched by an increase in cell cycle length. Dividing RPCs are depicted with white cytoplasm and green nuclei. Post mitotic differentiating RTCs are depicted with red nuclei and colored cytoplasm.
48 p15Ink4b p16Ink4a p18Ink4c Mitogens p19Ink4d
p107 CycD1/2/3 p130 E2f4 Cdk4/6 E2f5 p27Kip1 cdh1 p107 Rb APC/C Skp2 p21Cip p130 p57Kip2
E2f1 E2f2 CycE/A E2f3 Cdk2
DNA replication proteins Replication origin firing
S-phase
Fig 1.2. Some Key Regulators of G1/S transition. The two major axes of G1/S inhibition are the Rb-E2f axis and the Cip/Kip-Cdk2 axis. Mitogens stimulate division by inhibiting INK4a CKIs (e.g. p19Ink4d) and increasing CycD levels. Activated Cdk4/6 phosphorylates and inhibits the Rb family. D-Cdk4/6 complexes also sequester Cip/Kip CKIs (e.g. p27). Active Rb binds and inhibits gene transactivation by E2f1-3. p107/130 form repressor complexes with E2f4/5 that target the same genes as E2f1-3. On the other axis, Cip/Kip CKIs bind and inhibit Cdk2 complexes. P107 and p130, unlike Rb, can bind and inhibit Cdk2 complexes. There are also feed forward and feed back links between the Rb-E2f and Cip/Kip-Cdk2 axes. Feed forward effects include blockade of Skp2-mediated degradation of p27 by Rb-Cdh1, and E2f-mediated induction of CycE/A. Feedback effects include inhibition of Rb by Cdk2 (thus further activating both E2fs and Skp2). Positive regulators of G1/S inhibition are in red, negative in green. The figure does not, by any means, include all regulators and links.
49 Dividing RPCs p19Ink4d Mitogens
p107 CycD1 E2f4 Cdk4/6
E2f5 Kip1 Rb APC/Ccdh1 Skp2 p27 p107
E2f1 E2f2 CycE/A E2f3 Cdk2
DNA replication proteins Replication origin firing
S-phase
Fig 1.3. Cell cycle regulation in RPCs versus differentiating RTCs. A) In dividing RPCs extremely high levels of CycD1 activate Cdk4/6 and maintain Rb and p107 in an inactive phosphorylated state and sequester p27Kip1. The latter is also efficiently degraded, presumably due to high levels of Skp2. Inactive pocket proteins leave the activating E2fs free to induce transcription of genes required for DNA replication and CycE/A that activate Cdk2. p27Kip1 absence leaves CycE/-Cdk2 free to fire origins and further inhibit negative cell cycle regulators (e.g. Rb, p27Kip1).
Differentiating RTCs p19Ink4d Mitogens
p107 CycD1 p130 E2f4 Cdk4/6
E2f5 Kip1 Rb APC/Ccdh1 Skp2 p27 p107 p130
E2f1 E2f2 CycE/A E2f3 Cdk2
DNA replication proteins Replication origin firing
S-phase
B) In differentiating RTCs both p19Ink4d and p27Kip1 proteins are induced and inhibit Cyc-Cdk complex activity. Cyclin expression is downregulated. Pocket proteins are activated by dephosphorylation an quench E2f activity. p107/p130 act in other cells to form repressor complexes with E2f4/5, although whether this is the case in the retina has not been shown explicitly. Rb may facilitate p27Kip1 stability by bringing Skp2 into contact with APC/C. However, in human retina, there is only brief overlap in Rb and p27Kip1 expression in differentiating neurons, so the picture is more complex than depicted.
50
Apical surface
High Low
Nicd
Hes1 Anti-birth/exit Skp2/CycD1
p27Kip1
Pro-birth txn factors
Pro-birth/exit
Basal surface
Fig 1.4. Notch and INM control birth/exit pathways. This Nicd gradient model is based on data from zebrafish studies, but is still controversial (see text for details). Also, the link between Nicd and Skp2/Cyclin D1 induction has been shown in other cell types but not, as yet, RPCs. In this speculative model, the RPC nucleus on the left migrates less basally than the RPC nucleus on the right. As a result the former is under control of the apical gradient of Notch signaling mediated by the cleaved diffusable Notch intracellular domain (Nicd), which drives transcription of Hes1 (green) as well as Skp2 and Cyclin D1 (blue) which counter expression of neurogenic bHLH factors and inhibit p27Kip1, respectively. On the contrary, the nucleus that migrates deeper is less affected by Nicd because it has to diffuse further (dotted grey arrows), and the reduced levels of Hes1 and Skp2/CycD1 permit induction of pro-birth bHLH factors (red) and stabilization of p27Kip1 (orange).
51 Apical surface M M G1 G2 G0* WT
S G0 T.D.
Basal surface
Rb-/-
RPC division RTC exit unaffected delayed
Fig 1.5. Uncoupling cell birth and cell cycle exit.
A) In the WT retina the RPC nucleus/cell body changes position as the cell progresses through the cell cycle. This process is termed interkinetic nuclear migration (INM). Following cell birth, the apical process begins to retract as the newborn RTC moves to its final destination and undergoes terminal differentiation (T.D.) Green and red nuclei depict whether the cell is dividing or post-mitotic, respectively. Yellow cytoplasm indicates induction of a distinct transcriptome in the RTC.
B) In the absence of Rb, RPC division is unaffected, but differentiating RTCs divide ectopically (note green instead of red nucleus). However, the birth transcriptome (yellow cytoplasm) is activated. The response of different RTCs to Rb loss is cell type specific. Some (as depicted here) survive, terminally differentiate and exit division via Rb-independent means. During ectopic division these cells are at risk for neoplastic transformation (not shown). Other cell types (e.g. ganglion cells) escape transformation by undergoing apoptosis (not shown).
52 A.
Exit
Pre-exit cascade Gene X
Cog
B. Cell 1 Cell 2
Exit Exit
Pre-gene Gene X Pre-gene Gene X X events X events Trigger
Pre-exit Pre-exit cascade cascade
Fig 1.6. Cell cycle Exit and Differentiation are not rigidly coupled. A. Cell cycle exit (“birth”) of retinal progenitor cells (RPCs) is considered a watershed that is preceded by changing levels of cell cycle regulators, and followed rapidly by induction of a post M-phase differentiation cascade. In this way, initiation of differentiation is rigidly coupled to cell cycle exit as though through a cog wheel (“The Cog Model”). B. We characterized the regulation of many division and differentiation markers relative to each other and final mitosis. Unexpectedly, early embryonic RPCs expressed “differentiation” markers that later labeled post-mitotic neurons exclusively (e.g. Brn3b, Tubb3, Ptf1a), and factors detected just after cell birth in the embryo were induced well beyond M-phase post- natally (e.g. Nrl, Crx). Thus, the dynamics of birth-associated events shift dramatically during development, even to either side of mitosis (compare Cell 1 and Cell 2). In Cell 1, differentiation cascade “runs” slower than exit cascade and thus differentiation begins before the cell reaches the M phase. In Cell 2, exit cascade is faster, and thus exit occurs before the onset of differentiation. Therefore, instead of cell birth behaving as a cog that activates post-exit differentiation we suggest that a common trigger induces both the exit and differentiation programs in RPCs, precisely coordinating their start-points, but that each subsequent cascade unfolds independently (“The Trigger Theory”). This model explains the convergence of birth and differentiation but also why their connection is malleable.
53
Chapter Two: Timing of Cell Cycle Marker Silencing and the Onset of Differentiation of Retinal Ganglion Cells
The work in Chapter 2 was written by Marek Pacal and Dr. Rod Bremner as a manuscript in preparation.
54 2.1 Introduction
Cell cycle exit and differentiation have to be carefully coordinated during tissue development. Inappropriate activation of the cell cycle machinery in cells that should be post-mitotic can trigger profound differentiation defects and/or cause cell death, and is associated with neurodegenerative diseases and cancers (Bremner et al. 2004; Herrup and
Yang 2007; Chong et al. 2009b). The coordination of exit and differentiation is best studied in the nervous system. The retina is a particularly useful CNS tissue in this regard as it is easily accessible and can be manipulated without compromising viability.
For these reasons, there is a considerable literature and numerous reagents available to study retinal development. Despite these advantages, several fundamental questions have not been addressed. The expression “tightly coupled” is used commonly to describe the link between cell cycle exit and differentiation, but its exact meaning is not well defined.
In a classic study, Waid and McLoon showed that the chick ganglion cell marker, RA4, is activated within ~15 minutes of cell cycle exit (Waid and McLoon 1995). Apart from suggesting that the program to achieve this event was set up in the dividing progenitor, this data also suggests that cell cycle exit is a watershed which, once crossed, is accompanied by striking changes in the gene expression program.
Strictly, the term “cell birth” refers to the end of the last mitosis in a progenitor that is producing one or two post-mitotic daughter cells, but it is common to assume that birth coincides with the activation of specific gene products. But how maliable or rigid is this relationship? For example, are differentiation regulators/markers that appear around the time of cell birth always induced at the same interval before or after M phase?
Can induction vary during development? If so, is this maleability confined to one side of 55 M phase such that an event can take more or less time but always occurs on one side of
M phase? Or is the process so variable that induction can occur either side of birth? If the latter was the case, it would argue against the notion that completion of M phase is linked to gene induction in the newborn cell. These questions are as poorly answered for the timing with which progenitor markers and positive cell cycle regulators are silenced.
It is unclear which markers are extinguished prior to exit, whether some linger on beyond
M phase, and if so for how long? Apart from understanding the biology of cell birth, these issues have important practical implications. The field uses several markers to define dividing or post-mitotic cells yet in most cases it has not been rigorously defined whether they trully do delineate these separate populations.
One way to think of birth and differentiation is that they are controlled by two
interlocked cogs. The downstream events are distinct, yet inexorably linked such that
their timing is precisely coordinated. If, for example, events at M phase directly regulate
immediate differentiation events in the newborn cell, there would be no possibility that
the latter could occur either side of M phase in different progenitors. An alternative
model is that birth and differentiation are akin to two runners in a race. One trigger (the
starting gun in a race) induces both events (the runners moving to the finish line), but
numerous variables affect the end result (fitness, injuries, equipment, weather etc). In a
progenitor approaching cell birth, one signal could set off two separate cascades that lead
to exit and differentiation, but beyond the trigger multiple variables would influence the
timing of both events (such as the concentration of countless activators and inhibitors)
and thus the coupling of exit and differentiation would be more plastic such that across
56 development the same differentiation event could occur at significanlty distinct times, even either side of M phase.
To study this problem we examined the birth of mouse retinal ganglion cells.
Ganglion cells are the first retinal neurons generated in all animal models studied. In humans, the loss of these neurons is associated with glaucoma, a leading cause of blindness, and diabetic retinopathy, a major cause of morbidity in diabetic patients (Kern and Barber 2008; Weber et al. 2008). The life and death of these neurons has therefore been much scrutinized. In the mouse, ganglion cells are genereated between the embryonic day (E) E11 and aproximtaly the post-natal day P0, although the majority of ganglion cells are born between the days E12-16 (Fig 2.1 ) (Hinds and Hinds 1974;
Drager 1985; Young 1985a). Hierarchical model of ganglion cell development has been proposed in which the bHLH transcription factor Atoh7 (Math5) promotes the ganglion cell fate specification and supresses the non-ganglion cell fate; followed by the expresion of the POU-homeodomain (POU-HD) factor Pou4f2 (Brn3b) and the LIM-homeodomain
(LIMHD) factor Isl1 (Islet1) which activate further targets (Mu et al. 2005). However,
Atoh7 is neither sufficient nor necessary to specify all ganglion cells; not all
Atoh7/Math5 expressing cells differentiate into ganglion cells and Atoh7 -null mice still generate a small (~5%) population of ganglion cells (Brown et al. 2001). Ganglion cells seem to be born normally in the absence of Pou4f2 and Isl1, but the loss of these transcription factors results in defective ganglion cell differeantiation and massive cell death (Pan et al. 2005; Pan et al. 2008). Thus these two factors appeared dispensable in the innitial steps of ganglion cell differentiation. However, recent and more careful analysis of Pou4f2-null retinas revelaled that many Pou4f2-null cells thought to be
57 specified as ganglion cells are in fact other cell types and that Pou4f2 does confer ganglion cell fate (Qiu et al. 2008). Importantly, the expression of Pou4f2, albeit reduced, occurs even in the absence of Atoh7, indicating that the expression of Pouf42 in some cells is not downstream of Atoh7 (Brown et al. 2001; Pan et al. 2008). Further, the presence of Isl1 in a subset of retinal progenitors is intriguing and suggests that Isl1 may have additional roles in ganglion cell development prior to its later roles (Pan et al. 2008).
Thus, while the previosuly described Atoh7/Pou4f2/Isl1 hiearchy may underlie the development of most ganglion cells, it may not account for differentiation of all ganglion cells. Indeed, recent trasncriptional profiling reveald a great heterogeniety in the retinal progenitors and ganglion cells (Trimarchi et al. 2007; Trimarchi et al. 2008).
Thus, the innitial steps in ganglion cell differentiation are likely not identical across the whole population. This study attempts to adress these issues.
2.2 Results
2.2.1 Ki67, but not Pcna or Mcm6, is confined to the NBL
Our overall goal was to define the timing of induction of ganglion cell markers relative to
cell birth, to determine if it stays constant during development and, if it varies, to define
whether it at least remains constant with respect to the position of M phase. We first
examined the expression of common cell cycle markers with some well established
ganglion cell markers. Ki67 and Pcna are used widely to mark all phases of the cell
cycle. A more recent study has also highlighted the utility of the DNA replication
licensing factor Mcm6 as a pan-cell cycle marker in the retina (Barton and Levine 2008).
58 However, the same study reported that, based on flow cytometry cell cycle analysis, both
Pcna and Mcm6 are also found in a significant proportion of G0/G1 retina cells (Barton and Levine 2008). Similarly, based on another study it appears that the expression of mRNA of both Pcna and Mcm6 extends beyond RPCs into the ganglion and amacrine neuronal populations (Trimarchi et al. 2008). Thus, Pcna and Mcm6 may not distinguish
RPC and RTC cell populations. Indeed, we found many Mcm6 + cells, and to a lesser extent Pcna + cells, in the ONBL and the forming GCL where post-mitotic ganglion neurons are found in the early embryonic retina ( Fig 2.2). In comparison, Ki67 + cells did not appear to extend as far into the ONBL as Mcm6 or Pcna, although there were occasional faintly labeled cells that barely encroached into the forming GCL ( Fig 2.2 ).
These data indicate that Mcm6 and Pcna are retained by some G0 neurons, and suggest
that more careful analysis is warranted to define whether some of the Ki67 + cells are post mitotic RTCs.
2.2.2 Ki67 labels all phases of the cell cycle in all RPCs
To more comprehensively define Ki67 expression we first examined the assumption that it co-labels all phases of the cell cycle in retinal progenitors. A short pulse of BrdU is commonly used to label S-phase RPCs. The expression of Ccna2 (Cyclin a2) and Ccnb1
(Cyclin b1) can be used to visualize RPCs in the G2 - M-phases (Barton and Levine
2008). Both cyclins can be detected in RPC processes and display perinuclear and nuclear staining in cells near the apical surface (Barton and Levine 2008). Finally, the phosphorylation of histone H3 on Ser 10 accompanies chromosome condensation at the beginning of mitosis (Hendzel et al. 1997). In cultured cells, anti-PH3 antibodies detect 59 strong nuclear staining from the beginning of prophase to telophase (Prigent and
Dimitrov 2003; Li et al. 2005) and are widely used to detect mitotic RPCs.
Ki67 co-localized with all of four of these markers (short BrdU pulse, Ccna2,
Ccnb1 and pH3) at multiple time points during retinal development ( Fig 2.3A, Table
2.1). These data confirm the notion that Ki67 marks all RPCs throughout the cell cycle.
To address whether it is exclusive to RPCs we examined additional markers, as outlined below.
2.2.3 A subset of Ki67 cells lack the pan-cell cycle markers Vsx2 and Ccnd1
If Ki67 only labels RPCs and is absent in RTCs, then it should overlap with other pan- cell cycle markers. Ccnd1 is best known for its effects in G1 but in the retina it is expressed in all phases of the RPC cell cycle (Barton and Levine 2008). Indeed, we observed that Ccnd1 + cells were also positive for BrdU, Ccna2, Ccnb2 and PH3; staining was strongest in S-phase nuclei labeled with BrdU (0.5h) in the inner half of the NBL, and was nuclear or perinuclear in apically located G2/M cells ( Fig 2.4A, B ). Vsx2 is also expressed in RPCs and is required for proliferation of early but not late progenitors
(Burmeister et al. 1996; Livne-Bar et al. 2006). In the post-natal retina, Vsx2 is also expressed in bipolar neurons and Müller glia. Thus, this marker can be used as an unambiguous RPC marker only in the embryonic retina (Burmeister et al. 1996; Rowan and Cepko 2004; Livne-Bar et al. 2006). Indeed, Vsx2 was found together with BrdU,
Ccna2, Ccnb2 and PH3 in the embryonic retina; rich nuclear Vsx2 staining was observed in the inner NBL, while a mixture of nuclear and perinuclear staining was observed on the apical surface ( Fig 2.4A, B ). 60 Having confirmed that Vsx2 and Ccnd1 label all phases of the cell cycle we analyzed the fraction of Ki67 cells that were Vsx2 + and Ccnd1 +. Surprisingly, a relatively large fraction of Ki67 cells was Vsx2 - (e.g., E12.5: 247/3300; 7 ± 0.5%) and a
slightly greater fraction were Ccnd1 - (e.g., E12.5: 327/3300; 9.3 ± 0.2 %) ( Table 2.2, Fig
2.3B ). Consistent with this finding, some Vsx2 + cells were Ccnd1 -(e.g., E12.5:
127/3000; 4.3 ± 0.2%), but all Ccnd1 + cells were Vsx2 + (T able 2.3, Fig 2.4A ). This data
supports the possibility that some Ki67 + cells in the NBL are post-mitotic RTCs rather than RPCs. Another explanation, however, is that a significant fraction of Ki67 + RPCs
lack Vsx2and more so Ccnd1.
Towards resolving these alternative possibilities, we first assessed the fraction of
RPCs in S-phase that express Vsx2 or Ccnd1. As noted earlier, all BrdU + RPCs co-label with Ki67 ( Table 2.1). If the entire ~7-9% of Ki67 + cells that lack Vsx1 and Ccnd1 are
RPCs, and Vsx/Ccnd1 are expressed throughout the cell cycle, one would expect a similar fraction of BrdU + RPCs to lack Vsx1, whereas if these cells are all/mainly RTCs
then no/few BrdU +/Vsx - or BrdU +/Ccnd1 - cells should exist. The latter applied, since only a very small population (<1%) of BrdU + RPCs lacked (or displayed barely detectable levels) of Vsx2 (E12.5: 24/3000; 0.8 ± 0.1%), and a slightly larger but still very small fraction lacked Ccnd1 (E12.5: 46/3000; 1.5 ± 0.1%). All G2/M phase marked cells were always Vsx2 + and Ccnd1 + ( Fig 2.4B and data not shown), but because there are few cells in these phases it is difficult to exlude the possibility that a tiny fraction lacked Vsx2 and Ccnd1, as for the more abundant S-phase cells. In summary, while some Ki67 +/Vsx2 - and Ki67 +/Ccnd1 - RPCs, were S-phase RPCs, the fraction (1-1.5%)
was too small to explain the entire ~7-9 % of Ki67 + cells that were Ccnd1 - and Vsx2 - ,
61 arguing that many or most of these cells are post-mitotic RTCs. Nevertheless, it remained formally possible that many of the Ki67 +/Vsx2 - or Ki67 +/Ccnd1 - cells were not
RTCs, but rather RPCs in G1, for which there is no specific marker.
2.2.4 Ki67 co-labels cells positive for presumed differentiation markers
To further test the notion that Ki67 might be maintained in RTCs we co-labeled embryonic retinas for Ki67 and several well established markers that are expressed exclusively in ganglion cells (Pou4f1 (Brn3a), Pouf4f2 (Brn3b)), or in ganglion cells plus other retinal neurons (Tubb3 (Tuj1), Calb2 (Calretinin), Neuna60 (NeuN)) (Quina et al.
2005; Lee et al. 2003; Sharma et al. 2003; Pan et al. 2005; Philips et al. 2005; Sharma and
Netland 2007; Pasteels et al. 1990; Haverkamp and Wassle 2000; Raymond et al. 2008).
We also analyzed Isl1 (Islet1) which is considered primarily to mark post-mitotic cells, and Pax6 which labels RPCs and both ganglion as well as other neurons, and thus served as a control for double labeling with Ki67 (Marquardt et al. 2001; Pan et al. 2008).
As expected, many Pax6 + cells co-labeled with Ki67 (e.g., E12.5: 45.3 ± 4.9%
Pax6 + cells, Fig 2.5A, Table 2.4). Isl1 also co-labeled with Ki67 (e.g., E12.5: 12.4 ±
0.7% Isl1 + cells, Fig 2.5B, Table 2.4). In contrast, Neuna60 was found only in the forming GCL and never co-localized with Ki67 ( Fig 2.5C, Table 2.4), indicating that it marks G0 ganglion neurons exclusively. Strikingly, all of the remaining five differentiation markers stained Ki67 + cells to varying extents ( Fig 2.6, Table 2.4). For
example, at E12.5 Ki67 was detected in 15.4 ± 2.5% of Pou4f2(Brn3b) + cells, 12.4 ±
0.7% Tubb3 + cells, 6.2 ± 1.0% Calb2 + and 0.1 ± 0% Pou4f1(Brn3a) + cells ( Fig 2.6 A, B,
C, D ; Table 2.4). These results further support the possibility that Ki67 expression is 62 maintained in post-mitotic neurons. However, it is also possible that some of these differentiation markers are induced in RPCs. To define this issue more precisely, we next asked whether: 1. Any of the presumed post-mitotic ganglion cell markers co-localize with cell cycle markers other than Ki67, and 2. The timing of Ki67 extinction precedes or follows cell birth.
2.2.5 Ccnd1 and Vsx1 extinction followed by induction of Isl1, Pou4f2, then Tubb3 in early RPCs
We assessed marker expression relative to BrdU (0.5h), Ccna2, Ccnb1, PH3, Ccnd1 and
Vsx2. As expected, a fraction of Pax6 + cells stained for all six RPC markers, ( Fig 2.7,
Table 2.4). Previously, it was suggested that Isl1 may be expressed in some RPCs, but
the extent to which this is the case and the temporal pattern has not been quantified
(Marquardt et al. 2001; Pan et al. 2008). At E12.5, a tiny fraction of Isl1 + cells were
BrdU + (0.5 hr; 1.6 ± 0%) and Ccnb1 + (0.4 ± 0.1%). However, Isl1 was either absent or
extremely weak in PH3 + RPCs implying it is temporarily extinguished during M phase
(Fig 2.8A , Table 2.4). At E14.5 and E16.5, Isl1 + cells never co-localized with BrdU,
Ccnb1 or PH3, suggesting that Isl1 expression at these later time points occurs after cell birth.
The expression pattern of Isl1 overlaps with that of Pou4f2 in both migrating and mature ganglion cells from E11 to adulthood (Pan et al. 2008). However, Pouf4f2 is assumed to label only post-mitotic cells. Surprisingly, and similar to Isl1, we found a tiny fraction of Pou4f2 + cells at E12.5 that were BrdU + (0.5 hr; 1.2 ± 0.3%) and Ccnb1 +
(0.2 ± 0.0%). Unlike Isl1, which was absent in M-phase, a minute number of
63 Pou4f2 +/PH3 + cells was also observed (0.2 ± 0.0%) ( Fig 2.8B, Table 2.4). At E14.5, the fraction of Pou4f2 + RPCs was dramatically reduced: a small subset of Pou4f2 + cells (0.2
± 0%) were BrdU + after a 2h pulse. No Pou4f2 +/Ccnb1 + or Pou4f2 +/PH3 + cells were
observed at E14.5, ( Table 2.4), which could mean it is temporarily silenced in G2/M
specifically at E14.5, or as seen at E12.5 there may be some Pou4f2 + G2/M cells and larger counts would be required to examine this possibility. In the past, Pou4f2 has been used to define RTCs, but this new analysis shows that although this is true beyond E14.5, at E12.5, and marginally at E14.5, it is induced in S-phase RPCs.
Upon closer inspection it was noticable that all Pou4f2 +/BrdU + and Isll +/BrdU + cells showed punctuate BrdU staining in the lower/apical half of the NBL. Previous studies established that this pattern marks late S-phase RPCs, as opposed to the more uniform staining seen in early S-phase RPCs (Humbert and Usson 1992; O'Keefe et al.
1992; Dimitrova and Gilbert 1999). Thus, Isl1 and Pou4f2 are expressed at the end of the
S-phase, perhaps in RPCs about to give birth to RTCs. As discussed above, around 1% of RPCs were BrdU +/Vsx -/Ccnd1 - ( Fig 2.4A ), thus we wondered if this subset is also the group that induces Isl1 and Pou4f2. Indeed, a high percentage of BrdU +/Vsx2 - and
BrdU +/Ccnd1 - RPCs were positive for Isl1 or Pou4f2 ( Fig 2.8A,B ). Moreover, 100% of
BrdU +/Isl1 + and BrdU +/Pou4f2 + RPCs lacked Vsx2 and Ccnd1 ( Fig 2.8A,B ). Altogether, the above data indicate that a fraction of E12.5 RPCs in late S-phase extinguish Ccnd1 then Vsx1 expression, and then they induce Isl1 and Pou4f2. A reasonable hypothesis is that these Ccnd1 -/Vsx2 -/Isl1 +/Pou4f2 + RPCs are about to give birth to post-mitotic
differentiating ganglion cells.
64 Unlike Pou4f2 and Isl1, Tubb3 was not detected in BrdU + cells, nor was it co- detected with Ccnd1 or Vsx2 ( Fig 2.9A, Table 2.4 and data not shown). However, specifically at E12.5, a minute fraction of Tubb3 + cells co-localized with Ccnb1 (0.05 ±
0.04%) and PH3 (0.1 ± 0.04%) ( Fig 2.9B , Table 2.4). At later developmental time points
Tubb3 was never co-localized with PH3 or Ccnb1 ( Table 2.4 and data not shown). We
also detected a population of PH3 +/Tubb3 +/Pou4f2 + cells ( Fig 2.9C ); consistent with the idea that Tubb3 induction follows the Isl1 and Pou4f2 induction revealed above.
Finally, in contrast to Pou4f2, neither the related transcription factor Pou4f1 nor the calcium binding protein Calb2 co-localized with the cell cycle markers BrdU (0.5h),
Ccna2, Ccnb1, PH3, Ccnd1 and Vsx2 ( Fig 2.10A, B and data not shown). As noted earlier, Neuna60 was induced after Ki67 induction (Fig 2.5C ), so in summary, of six markers expressed in ganglion cells, Isl1, Pou4f2 and Tubb3 are induced just prior to cell birth at E12.5 (in that order), whereas Calb2, Pou4f1, Nuena60 are induced after cell birth.
2.2.6 The length of G2/M
To complement the above analyses we wanted to define when markers are induced relative to the end of S-phase and G2/M. Prior work defined the average length of S- phase and the entire cell cycle at various times during rat retinal development (Alexiades and Cepko 1996), but the length of G2/M has not been measured. If the length of G2/M were known, BrdU labeling could be used to define when markers are extinguished or induced relative to the end of G2/M and hence cell birth.
65 G2/M length is determined by measuring the time necessary to label all mitotic cells with BrdU (Takahashi et al. 1995) ( Fig 2.11A); when the BrdU pulse is too short no
S-phase cells reach the end of M phase and thus some PH3 + cells remain BrdU -, but when
the BrdU pulse is as long as G2/M all PH3 + cells become BrdU +. We performed BrdU
labeling for 1, 2, 4 and 6 h at seven distinct embryonic and post-natal time points (E12.5,
E14.5, E16.5, P0, P3 and P5), then determined the fraction of BrdU +/PH3 + cells. After 2
h of labeling, > 90% of PH3 + cells were BrdU + in the embryonic retina, which fell to 82.6
± 0.6%, 41.1 ± 3.7% and 0 ± 0% by P0, P3 and P5, respectively ( Fig 2.11B, Table 2.5).
Thus, G2/M spans slightly over 2 hrs in the embryonic retina and expands gradually to 4 hrs by P5 ( Table 2.6). In agreement, a greater fraction of perinuclear Ccnb1 + (G2) cells were BrdU + for each pulse length at every stage of development ( Fig 2.11 C, Table 2.5).
This increase in G2/M during development is consistent with the concomitant increase in total cell cycle length (Alexiades and Cepko 1996).
2.2.7 Timing of Ki67 extinction in G0* cells
Characterizing G2/M length ( Table 2.6) allowed us to apply BrdU pulse-chase analysis to define when Ki67 is extinguished relative to cell birth. Retinas were exposed to a single pulse of BrdU at five developmental stages (E12.5, 14.5, 16.5, P0, 3), and the duration necessary to detect the first Ki67 -/BrdU + cells was determined. If Ki67 was extinguished before or after cell birth, a period less or more than the length of G2/M would generate Ki67 -/BrdU + cells, respectively. Strikingly, even after 12 hours of BrdU labeling no Ki67 -/BrdU + cells were detected at any of the five developmental time points.
However, by 24 hours, Ki67 -/BrdU + cells were detected at all developmental time points 66 (Fig 2.12A). We confirmed that the Ki67 -/BrdU + cells were differentiating neurons because they co-labeled with Tubb3, which only labels post-mitotic cells from E14.5 and beyond, or Calb2 and Pou4f1, which only label post-mitotic neurons ( Fig 2.12B and data not shown). Moreover, at long labeling periods BrdU is diluted out of continually cycling RPCs and only RTCs born immediately after M phase are labeled intensely. For any single chase period (e.g. 48 hrs) the fraction of Ki67 -/BrdU + cells decreased as
development proceeded ( Fig 2.12A), and at any single developmental time point, the
fraction of Ki67 -/BrdU + cells increased with the length of the chase ( Fig 2.12A). Thus,
the minimal time to extinguish Ki67 after S phase is between 12 and 24 hrs, and since
G2/M is ~2 hr or 4 hr at E12.5 or P3 ( Table 2.6), Ki67 is retained for a minimum of 10 hr
and 22 hr after cell birth, respectively. For the purposes of discussion, we refer to this
period of Ki67 retention in post-mitotic cells as G0*.
As noted earlier, Ki67 is detected in the embryonic NBL but disappears as cells
begin to emerge into the forming GCL ( Fig 2.2). Thus, as a second approach to
approximate the length of G0* pregnant dams were pulse-labeled with BrdU at E12.5,
E14.5 and E16.5, chased for 8, 12, 24 and 48 hr, and BrdU +/Tubb3 + neurons in the GCL
were quantified. The chase time for newborn cells to migrate to the GCL ( Fig 2.12C),
where Ki67 disappears, was ~ 12 hrs at E12.5 and E14.5 and between 12 and 24 hrs at
E16.5, which is similar to the chase times required to observe the first BrdU+/Ki67 - cells
(Fig 2.12A).
67 2.2.8 Timing of expression of ganglion neuronal markers
In the previous sections, we determined the extent of expression of neuronal markers in
RPCs based on their extent or lack of co-expression with Ki67 and several other cell cycle markers. We determined that Isl1, Pou4f2 and Tubb3 appear to be expressed in subsets of RPCs, since they co-localize with Ki67 and (critically) other cell cycle markers. On the other hand, Pou4f1 and Calb2 appear to be expressed after cell birth, as they co-localize with Ki67 but not with any of six other cell cycle markers. Finally,
Neuna60 stains G0 neurons in the forming GCL as its does not co-localize with Ki67.
Here, we applied BrdU pulse-chase labeling to determine more precisely the onset of expression of these markers relative to the cell birth.
As described above Tubb3 signal appears in a small subset of G2/M RPCs at
E12.5 but is restricted to post-mitotic neurons at later times. At E12.5, when G2/M length is approximately 2h, a 1h BrdU pulse did not label Tubb3 + cells. After a 2h BrdU pulse, a minute number (0.1 ± 0.04%) of Tubb3 +/BrdU + double-labeled cells were observed ( Fig 2.9D , Table2.4). As reported above, the same fraction of Tubb3 cells colabeled with the G2/M markers Ccnb1 and PH3 ( Fig 2.9B, Table 2.4) suggesting that
Tubb3 protein is translated just before cell birth in E12.5 RPCs. Interestingly, these
Tubb3 +/BrdU +(2h) cells expressed Pou4f2, supporting the notion that they are biased towards the ganglion cell fate ( Fig 2.9D ).
Notably, at later developmental time points Tubb3 +/BrdU + cells were observed
after increasingly longer BrdU labeling, corresponding to the increase in the G2/M
length. At E14.5 and onwards the length of G2/M is ~ 4hrs. The first Tubb3 +/BrdU + cells born at E14.5 were observed after 4h of BrdU labeling (0.06 ± 0.0%), but cells born 68 at E16.5 were BrdU-labeled after 6h (0.07 ± 0.06%). In the post natal retina, 10 hrs of labeling was required to detect the first Tubb3 +/BrdU + cells at P0 (0.42 ± 0.1%) and P3
(0.26 ± 0.05%). This fraction rose at longer labeling times, consistent with gradual accumulation of Tubb3 + post-mitotic neurons ( Table 2.4). Importantly, at E16.5 and
beyond Tubb3 did not co localize with G2/M markers ( Table 2.4 and data not shown),
confirming that Tubb3 is induced after cell birth.
Since Calb2 and Pou4f1 never co-localized with any cell cycle markers apart from
Ki67, these neuronal markers seem to label G0* RTCs that retain Ki67. Indeed, in the
embryonic retina, 2h or 4h of labeling was insufficient and 6h, well beyond G2/M length,
was required to observe the first Calb2 +/BrdU + cells (0.25 ± 0%; E14.5: 0.1 ± 0.1%). At
E16.5 the first Calb2 +/BrdU +cells appeared only after 12h of labeling (0.4 ± 0.1%). Since
at all time points examined, the length of BrdU labeling exceeded the length of G2/M, it
is clear that Calb2 is detectable in RTCs after the last G2/M. For example, at E12.5 and
E16.5 Calb2 is detectable 4h and 8h after the last M-phase, respectively ( Fig 2.10C,
Table 2.4).
Pou4f1 was detectable in a minute number of Ki67 + cells only at E12.5, suggesting it might be induced around the end of G0* ( Table 2.4). Indeed, at this time
point, Pou4f1+/BrdU + cells (2.5 ± 0.3%) only appeared after 12h of BrdU labeling ( Fig
2.10D, Table 2.4). Thus Pou4f1 appears in RTCs 10h after cell birth at E12.5 and 6h
later than Calb2.
Neuna60 never co-localized with Ki67 and was detected in the GCL only after
~E13, arguing that it is induced beyond G0*. In agreement, only a few Neuna60 +/BrdU +
neurons were detected in retinas pulsed with BrdU at E12.5 and chased for 24h; which
69 approximately doubled after another 24h of labeling ( Fig 2.5D; Table 2.4). Since, the length of G2/M at E12.5 is 2h, this marker is expressed by the GCL neurons approximately 22h after the last M-phase.
In summary, from the above results we infer the order of appearance of these markers during the development of ganglion cells: Pax6 (many if not all RPCs, G0* and
G0 neurons) > Isl1 and Pou4f2 (a subset of late S/G2/M RPCs at E12.5, G0* RTCs and
G0 neurons) > Tubb3 (G2/M-RPCs at E12.5 and G0* RTC and G0 neurons) > Calb2
(mid-stage G0* RTCs, G0 neurons) > Pou4f1 (late stage G0* RTCs, G0 neurons) >
Neuna60 (G0 neurons).
2.3 Discussion
The conversion from dividing RPCs to post-mitotic RTCs involves the inactivation and induction of a large number of cell cycle and differentiation promoting factors, respectively, but the sequence and timing of these events is poorly defined. To dissect the mechanisms underlying cell birth it is critical to first understand the flow of events and to distinguish dividing and post-mitotic cell populations. A useful initial parameter is the spatial expression pattern, but for proteins detected in the NBL additional analyses is required. To address this issue, we utilized Vsx2, Ccnd1, Ccna2, Ccnb1, BrdU and PH3 to label dividing RPCs. To further define when markers are activated in RTCs we determined the length of G2/M, then performed BrdU pulse chase experiments to establish when the first marker +/BrdU + cells appear in the retina. Using these tools, we
performed an in depth analysis of events surrounding the birth and maturation of
ganglion cells. 70 Many patterns of expression were defined including: 1. Extinction in RPCs (e.g.
Vsx1 & Ccnd1), 2. Induction in many RPCs + RTCs + Neurons (e.g. Pax6); 3.
Induction in late stage RPCs + RTCs + Neurons (e.g.: Tubb3, Isl1 & Pou4f2 at E12.5);
4. Induction in RTCs and Neurons (e.g. Calb2 & Pou4f1 at all times, Tubb3, Isl1 and
Pou4f2 at E14.5 onwards); 5. Extinction in late stage RTCs (Ki67, defining “G0*” ); and
6. Induction beyond G0* in the GCL (e.g. Neuna60). Unexpectedly, several markers used widely to “specifically” mark dividing RPCs (e.g. Ki67, Pcna) or post-mitotic ganglion cells (e.g. Isl1, Pou4f2, Tubb3) remained on longer or were induced earlier than expected, respectively. Most strikingly, there were not strict rules regarding whether a differentiation marker was induced before or after birth, but this could change as the embryonic retina aged. Apart from raising concerns about the use of markers to definitively mark specific states, our analyses suggest a new model for how cell cycle exit and ganglion cell differentiation are coordinated, discussed below.
2.3.1 Coordinating Exit and Differentiation: The Trigger Theory
It is obvious that cell cycle exit and retinal differentiation are linked. Usually it is noted that they are “tightly connected”, much like two interlocked cogs driving interdependent machines. This notion is perhaps best encapsulated by the now classic observation that
RA4 is induced 15 minutes after the birth of chick ganglion cells (Waid and McLoon
1995). However, the degree to which the timing of cell cycle exit and retinal differentiation are connected over time has not been fully examined. Our data now show that the timing of ganglion cell marker induction relative to exit is not inflexible, but rather that it varies considerably across development. For example, Isl1, Pou4f2 and 71 Tubb3 were detectable already in G2/M RPCs at E12.5, while after this time point, these proteins were detected solely in the post-mitotic RTCs. These variations in the timing of induction are easier to detect in the mammalian retina than the evolutionary older avian, fish or amphibian retina, because development is more protracted. One of the predictions of such variability is that in early RPCs fate may be decided sooner than in late RPCs, just before rather than after cell cycle exit. To test this notion it would be necessary to live image cells that induce Brn3b before or after exit and determine whether they always become ganglion cells.
Why does the timing of cell cycle exit and differentiation factor induction vary?
We propose that these two events are set in motion by one switch, but that subsequently
their journeys are independent. Thus, rather than two inseparable cogs, cell cycle exit
and differentiation factor induction can be viewed as two runners in a race; one trigger
(the starting gun) initiates both, but beyond that their passage to the finish line is
independent. An excellent candidate for the trigger is the loss of Notch signaling because
Notch can inhibit differentiation and drive division, e.g. by inducing Hes1 and Ccnd1,
respectively (Wu et al. 2002; Liu et al. 2003). Indeed, we found that Ccnd1 is down-
regulated in S-phase RPCs that induce differentiation factors. Following the “trigger”,
numerous variables could affect when exit or differentiation-factor-induction occur, such
as the abundance of regulators, the epigenetic state of key genes, etc, thus explaining why
the timing of these two coordinated events can vary. In summary, we propose that
differentiation and exit are not rigidly connected (The Cog Model), but that they always
occur at roughly the same time because they rely on a common upstream switch (The
Trigger Theory).
72 2.3.2 Evidence for RPCs biased towards the ganglion cell fate
Previously, qualitative analysis of Isl1 suggested that it is expressed mainly in post- mitotic ganglion cells (Pan et al. 2008), while Pou4f2 was thought to be expressed exclusively in post-mitotic ganglion cells (Pan et al. 2005). We performed a thorough quantitative analysis of several thousands of cells and found that both markers were found in ~1% of late S-phase RPCs at E12.5. Pou4f2was also readily detectable in G2/M
RPCs, although Isl1 temporarily disappeared until cell birth. Intriguingly, these
Isl1 +/Pou4f2 + RPCs lacked Ccnd1 and Vsx2 and thus potentially the coordinated down-
regulation of the RPC/cell cycle markers and induction of Isl1 and Pou4f2 marks a
unique subset of RPCs biased towards, or perhaps even fully committed to, the ganglion
cell fate.
This notion is supported by the fact that Pou4f2 co-labeled with another
differentiation marker, Tubb3, in M-phase RPCs. Tubb3 is often used as a marker of
post-mitotic neurons, but the dynamics of Tubb3 expression are not uniform throughout
the CNS. For example, while Tubb3 +/BrdU + neurons only appear after at least 24 hrs
BrdU labeling in post-mitotic mouse embryonic cortical neurons (Menezes and Luskin
1994), the olfactory bulb interneurons and their progenitors in the anterior part of the subventricular zone (SVZa) express Tubb3 and other neuronal markers while still in the cell cycle (Memberg and Hall 1995; Menezes et al. 1995; Luskin et al. 1997; Brock et al.
1998). This suggests that in the SVZa, the expression of Tubb3 protein can be initiated as early as S and M phases. In addition to SVZa progenitors, a small number of cells with mitotic spindles expressing Tubb3 are found in the rat ventricular zone and
DRG/spinal cord at E11.5, but not later (not quantified) (Memberg and Hall 1995). In 73 agreement with the latter studies we found that Tubb3 was present in a fraction of G2/M
RPCs at E12.5. Interestingly, these cells also express Pou4f, suggesting that they are
RPCs that will likely become ganglion neurons at the end of mitosis. It appears that a subset of ganglion neurons at this early stage undergo rapid fate specification preceding cell birth, as manifested by the expression of both Isl1 and Pou4f2 in a subset of S-phase
RPCs, followed by the expression of Tubb3 as early as G2/M. Indeed, this accelerated cell fate specification/differentiation has been predicted previously based on the expression of the ganglion cell marker RA4 in neurons at the time of the last M-phase
(Waid and McLoon 1995). Olfactory neuronal precursors are committed to producing neuronal precursors and olfactory neurons (Luskin 1998; Coskun and Luskin 2002).
Similarly, Tubb3 + G2/M retinal RPCs may be fully committed towards becoming
neurons. Expression of Tubb3 in dividing progenitors may be a signature of neuronal
commitment.
Recent data indicate that mRNAs of large number developmental transcription
factors for many retinal neuronal subtypes, including ganglion cells, are expressed
already in RPCs (Trimarchi et al. 2008). The analysis focused on RNA, but conceivably
many of the cognate proteins that guide ganglion cell differentiation are already on prior
to cell cycle exit. Our data indicate that this is indeed the case for Pou4f2 and Isl1, but
only at the earliest period of retinal cell differentiation. RPCs biased to rod
photoreceptors have been suggested for many years (Cayouette et al. 2006) and recent
data shows that there are also RPCs that produce unique repeated patterns of daughter
cell progeny (Cohen et al. 2010). In view of these results, RPCs biased to specific
retinal subtypes may be a normal feature of the RPC pool.
74 2.3.3 Induction and roles of Isl1 and Pou4f2
The bHLH protein Math5 is required to induce Isl1 and Pou4f2 in ~95% of ganglion cells
(Wang et al. 2001; Yang et al. 2003). It has been stated that Math5 is expressed in post- mitotic cells, contradicting our claim that Isl1 and Pou4f2 are already present in late S phase RPCs at E12.5. However, the notion that Math5 protein is induced in post-mitotic cells has never been proven, due mainly to the lack of anti-Math5 antibody that can be used for immunostaining. However, a study using Math5-LacZ knock-in (KI) reporter mouse, described a tiny fraction of E11.5-E15.5 βgal +/Ccnd1 + RPCs and also a small subset of βgal +/PH3 + cells (Le et al. 2006). Also, Math5 mRNA was found in 18% of
E16.5 RPCs labeled by a 1h pulse of 3H-thymidine (Trimarchi et al. 2008). Further, a study using a Math5-HA knockin mouse line, which reproduces the expression of the endogenous Math5 protein, showed that while all Pou4f2 + cells in the NBL were Math5-
HA +, approximately 50% of Math5-HA + cells lacked the Pou4f2 signal (Fu et al. 2009).
Further, in agreement with Le et al (06) this study clearly showed Math5-HA + cells
located at the ventricular edge of the retina, where mitotic cells are located. Taken
together these data indicate that it is likely that some RPCs express Math5 protein,
consistent with our data that Isl1 and Pou4f2 are induced in early RPCs. It would be
informative to co-stain the Math5-HA mouse retina for HA, Isl1 or Pou4f2, and markers
of late S-phase or G2/M. If Math5 was not found in Isl1 + or Pou4f2 + RPCs this would suggest that the latter reflect the small proportion of cells that give rise to ganglion neurons independent of Math5 (Brown et al. 2001).
75 Isl1 and Pou4f2 affect survival and axon growth, however cell fate specification
has not been ascribed to these transcription factors because A) they are presumed to be
induced in post-mitotic ganglion RTCs, and B) because ganglion cells are specified even
in the absence of both of these transcription factors (Pan et al. 2008).
2.3.4 Ki67 remains in ganglion RTCs for a period of time after birth
The initial goal of defining when ganglion cells markers are induced was built on a
number of commonly held assumptions. As mentioned above, Pou4f2 was believed to
label post-mitotic ganglion RTCs. Further, markers such as Pcna and Ki67 are
commonly used to distinguish dividing and post-mitotic cell populations. We found that
although Ki67 labels all S, G2 and M RPCs, it also remains detectable in post-mitotic
RTCs expressing Calb2 and Pou4f1, markers that are never colocalized with any of the
six other dividing RPC markers we studied. Moreover, BrdU pulse chase revealed that
the onset of Calb2 and Pou4f1 detection occurred several hours after cell birth. Thus, it
was clear that Ki67 remained in Calb2 + and Pouf1 + RTCs for some time after cell birth.
Pulse chase analysis revealed, surprisingly, that this period lasts at least 12h and increases
with development. The ability to label newborn RTCs within the first 12-24h (e.g.,
Calb2 +/Ki67 +) could be exploited to isolate these cells and define other differences from
RPCs and G0 neurons. The DNA replication licensing factor Ctd1 is also expressed in post-mitotic CNS neurons, but unlike Ki67 it remains on and thus cannot distinguish G0* and G0 (Sakaue-Sawano et al. 2008). The disappearance of Ki67 typically occurred at the boundary of the NBL and GCL, although a few Ki67-/Pou4f1 + cells were observed migrating across the NBL. Whether the down-regulation of Ki67 and induction of 76 Pou4f1 is connected, regulated by coincident signals, or unrelated awaits future analysis, but the coincidence is intriguing.
In conclusion, we observed that markers believed to label post-mitotic ganglion neurons such as Pou4f2 or Tubb3 were found in a subset of progenitors. Thus, unlike the current dogma, the timing of ganglion cell development might be shifted back in time into the RPC stage. Conversely, we observed that Ki67, a marker commonly used to label diving cells is expressed for ~ 12-24h after cell birth, and thus can serve when used in conjunction with other neuronal markers such as Calb2 to conveniently label newborn
RTCs.
77 Developing Mature RPE RPE
M
G0* ONL G1 RPC OPL G2 INBL
RTC INL S G0
IPL
Ganglion
cell GCL GCL/ONBL
Figure 2.1. Overview of retinal structure and development. A. In the embryonic retina, RPCs pass through the cell cycle in a repetitive topological pattern: S-phase in the apical INBL, M-phase at the ventricular side, and G2 or G1 in between. Ganglion cell genesis starts at E11, peaks between E12.5-14.5 and diminishes after E16.5. Note that other early cell types that begin to populate the embryonic retina after E12 (amacrine and horizontal cells and photoreceptors) were omitted for simplicity. B. The structure of mature retina at P21 and onwards. Ganglion neurons make up half of the GCL along with amacrine neurons. Note that the number of ganglion cells is reduced approximately by half during the first post-natal week. E- embryonic day; P- post-natal day; INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer; RPC: retinal progenitor cell; RTC- retinal transition cell.
78 Merge/DAPI Ki67/Pcna/Mcm6 Ki67 Pcna Mcm6 INBL
GCL/ONBL E14.5
Figure 2.2. The expression patterns of Pcna, Mcm6 and Ki67 in the E14.5 WT embryonic retina. Mcm6 (grey) and to a lesser extent Pcna (red) labelled cells in both INBL and ONBL. In contrast, Ki67 (green) was more confined to the INBL. DAPI (blue) stains cell nuclei. Yellow arrows point to cells co-labelled for all three markers, pink arrows show cells co-labelling for Pcna and Mcm6 but not Ki67, while gray arrows depict cells positive only for Mcm6. INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bars are 10μm.
79 A.
Merge/DAPI Ki67 Merge/DAPI Ki67 Merge/DAPI Ki67 INBL
BrdU(0.5h) Ccna2/b1 PH3
E14.5 GCL/ONBL
B. Merge/DAPI Ki67/Ccnd1 Ki67 Ccnd1 INBL
E14.5 GCL/ONBL
Figure 2.3. Ki67 is detectable in all phases of cell cycle in the WT E14.5 retina. A. Ki67 (green) co- localized (yellow arrows highlight examples) with a short pulse of BrdU (red, S-phase), Ccna2/b1 (red, G2/M phase) or PH3 (red, M-phase). B. All Ccnd1+ cells (red) co-localized with Ki67 but some Ki67+ cells did not stain positive for Ccnd1 (white arrows). Nuclei in A and B are also stained with DAPI (blue). The lower panels are blow ups of the boxed region in the upper panel. INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bars are 10μm.
80 Table 2.1. Ki67 is expressed in all phases of the cell cycle. Ki67 co-localized with a short pulse of BrdU (S- phase), Ccna2/b1 (G2/M phase) and PH3 (M-phase). The numbers show the percentage of marker+ cells out of the total marker+ cell population. For example all Ccnd1+ cells were Ki67+.
BrdU(0.5h) Ccnb1 PH3 Vsx2 Ccnd1 E12.5 3000/3000; 100 ± 0% 3000/3000; 100 ± 0% 1500/1500; 100 ± 0% 3000/3000; 100 ± 0% 3000/3000; 100 ± 0% E14.5 3000/3000; 100 ± 0% 3000/3000; 100 ± 0% 1500/1500; 100 ± 0% 3000/3000; 100 ± 0% 3000/3000; 100 ± 0% P0 3000/3000; 100 ± 0% 3000/3000; 100 ± 0% 1500/1500; 100 ± 0% 3000/3000; 100 ± 0% 3000/3000; 100 ± 0%
Table 2. 2. A small population of Ki67+ cells lack Vsx2 and/or Ccnd1. The numbers show the percentage of Ki67+ /Vsx2- or Ki67+/Ccnd1- cells out of the total Ki67+ cell population. Thus while all Ccnd1+ cells were also Ki67+ (Table 2.1) , not all Ki67+ were Ccnd1+.
VSX2 CCND1 E12.5 247/3300; 7 ± 0.5% 327/3300; 9.3 ± 0.2 % E14.5 187/2950; 6.3 ± 0.1% 187/2950; 10.3 ± 0.2% P0 215/3600; 5.9 ± 0.2% 337/3600; 9.3 ± 0.2%
Table 2.3. A small population of Vsx2+ cells lack Ccnd1. While there was a small population of Vsx2+ cells that lacked Ccnd1, all Ccnd1+ cells contained Vsx2. The numbers show the percentage of Vsx2+ /Ccnd1- out of total Vsx2+ cell population and Ccnd1+/Vsx2- cells out of the total Ccnd1+ cell population.
(Vsx2+/Ccnd1-)/Vsx2+ (Ccnd1+/Vsx2-)/Ccnd1+ E12.5 127/3000; 4.3 ± 0.2% 0/3000; 0 ± 0 % E14.5 129/3000; 4.3 ± 0.1% 0/3000; 0 ± 0 % P0 132/3000; 4.4 ± 0.1% 0/3000; 0 ± 0 %
81 A. Merge/DAPI Vsx2/Ccnd1 Vsx2 Ccnd1 BrdU(0.5h) INBL
E12.5 ONBL
B. Merge/DAPI Ccnd1 Vsx2 Ccna2/b1 INBL
ONBL E14.5
Merge/DAPI Ccnd1 Vsx2 PH3 INBL
ONBL E14.5 Figure 2.4. The expression pattern of Ccnd1 and Vsx2 in the WT embryonic retina. A. At E12.5, both Vsx2 (red) and Ccnd1 (green) were expressed throughout the INBL. The lower panels are blow ups of the boxed region in the upper panel. Most cells labelled by a short 0.5h BrdU pulse (white) were also Ccnd1+ or Vsx2+ (green arrows), but a fraction of S-phase cells lacked these markers (white arrows). Further, some Vsx2+ cells lacked Ccnd1 (purple arrows). B. Both Ccnd1 (red) and Vsx2 (green) co-localized with Ccna2/b1 (a2 and b1 antibodies were combined) (white, G2/M phase) or PH3 (white, M-phase), at the apical surface. The white arrows depict perinuclear Ccna2/b1staining in G2/M cells, or PH3+ M-phase cells, whereas purple arrows point to non G2 Ccna2/b1+ and PH3+ cells. INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bars are 10μm. 82 A. C. E12.5: Pax6/Ki67 E13 E14.5 E16.5
Neuna60/Ki67 INBL INBL GCL/ONBL GCL/ONBL
D. B. E12.5: Isl1/Ki67 E12.5 + a 48h BrdU chase ( = E14.5) Neuna60/BrdU INBL INBL GCL/ONBL GCL GCL/ONBL
Figure 2.5. The expression pattern of Ki67 and Pax6, Isl1 and Neuna60 in the embryonic retina. A. At E12.5, many Pax6+ cells (green) co-localized with Ki67 (red, white arrows), and many resided in the GCL/ONBL. B. At E12.5, a small subset of Isl1+ cells (green) co-localized with Ki67 in the INBL (white arrows), while many Isl1+ cells resided in the GCL/ONBL. C. Neuna60 (green) did not co-localize with Ki67 at any time point and was always expressed in the GCL/ONBL. A tiny number Neuna60+/BrdU+ cells were detected after 24h BrdU chase (red), while this number increased after another 24h, shown here. This suggests that Nuena60 appears in GCL cells ~24h after cell birth. INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bar is 10μm.
83 Table 2.4. Marker analyses. The numbers represent percentage of cells positive for the given cell cycle marker at different time points (e.g., (Isl1+ and Ki67+ cells / all Isl1+ cells). The orange and gray panels represent co-localization or the lack of co-localization of neuronal markers with proliferation markers, respectively.
Marker/Age Ki67 BrdU Ccnb1 PH3 Vsx2 Pax6 E12.5 351/780; 45.3 ± 4.9% (0.5h) 200/840; 23.8 ± 1.2% 26/800; 3.25 ± 0.6% 9/800; 1.1 ± 0.6% 950/2450; 38.7 ± 1.2% Isl1 (Islet1) E12.5 128/1024; 12.4 ± 0.7% (0.5h) 20/1220; 1.6 ± 0% 5/1010; 0.4 ± 0.1% 0/1020; 0 ± 0% 0/1010; 0 ± 0% E14.5 26/1020; 2.5 ± 0.5% (2h) 0/1000; 0 ± 0% 0/1010; 0 ± 0% 0/1020; 0 ± 0% 0/1010; 0 ± 0% (4h) 2/300; 0.6 ± 0.5% (8h) 29/300; 9.6 ± 1.5% Pou4f2 (Brn3b) E12.5 165/1085; 15.4 ± 2.5% (0.5h) 19/1230; 1.5 ± 0% 3/1500; 0.2 ± 0.0% 3/1500; 0.2 ± 0.0% 0/900; 0 ±0% E14.5 23/1030; 2.2 ± 0.1% (2h) 5/1800; 0.2 ± 0% 0/900; 0 ± 0% 0/1000; 0 ± 0% 0/900; 0 ±0% (4h) 25/1800; 1.3 ± 0.2% (6h) 64/1800; 3.5 ± 0.2% (8h) 90/1800; 5.0 ± 0.3% (10h) 117/1800; 6.5 ± 0.1% Calb2 (Calretinin) E12.5 36/580; 6.2 ± 1.0% (1,2,4h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (6h) 3/1200; 0.25 ± 0% (12h) 18/1200; 1.5 ± 0.2% E14.5 36/600; 6.0 ± 1.0% (2,4h) 0/2100; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (6h) 1/900; 0.1 ± 0.1% (12h) 3/900; 0.3 ± 0% (24h) 8/900; 0.8 ± 0.1% E16.5 49/600; 8.1 ± 1.0% (2,4,6h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (12h) 5/1200; 0.4 ± 0.1% (24h) 40/1200; 2.6 ± 0.6% E18.5 30/788; 3.8 ± 0.3% (2h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% P0 20/900; 2.2 ± 0.3% (2h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0%
84 Table 2.4. Marker analyses continued
Marker/Age Ki67 BrdU Ccnb1 PH3 Vsx2 Tubb3 (Tuj1) E12.5 95/700; 13.5 ± 0.4% (0.5,1h) 0/3600; 0 ± 0% 2/3600; 0.05 ± 0.04% 4/3600; 0.1 ± 0.04% 0/3600; 0 ± 0% (2h) 4/3600; 0.1 ± 0.04% (4h) 8/3600; 0.2 ± 0.04% (6h) 17/3600; 0.4 ± 0.04% (12h) 33/2250; 1.4 ± 0.3% (GCL) (24h) 377/2400; 15.7 ± 0.6% (GCL) (48h) 560/2700; 20.7 ± 0.6% (GCL) E14.5 106/800; 13.2 ± 0.5% (0.5,1,2h) 0/2100; 0 ± 0% 0/3600; 0 ± 0% 0/3600; 0 ± 0% 0/3600; 0 ± 0% (4h) 3/4500; 0.06 ± 0.0% (6h) 5/1800; 0.2 ± 0.1% (8h) 7/2100; 0.3 ± 0.0% (10h) 10/2100; 0.4 ± 0.0% (12h) 8/2700; 0.29 ± 0.06% (GCL) (24h) 52/4500; 1.15 ± 0.3% (GCL) (48h) 450/6000; 7.5 ± 0.5% (GCL) E16.5 95/700; 12.9 ± 0.7% (1,2,4h) 0/2700; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (6h) 2/2700; 0.07 ± 0.0% (8h) 5/2700; 0.1 ± 0.0% (12h) 12/6000; 0.2 ± 0.05% (GCL) (24h) 41/15000; 0.27 ± 0.3% (GCL) (48h) 82/24000; 0.3 ± 0.03% (GCL) P0 89/800; 11.1 ± 1% (1,2,4,6,8h) 0/2700; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (10h) 13/3130; 0.4 ± 0.0% (24h) 30/4200; 0.7 ± 0.% (48h) 45/4200; 1.07 ± 0.07% (72h) 63/4200; 1.5 ± 0.07% P3 95/700; 11.3 ± 1.4% (1,2,4,6,8h) 0/2700; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (10h) 8/3000; 0.2 ± 0.0% (24h) 19/4200; 0.4 ± 0.04% (48h) 44/4200; 1.04 ± 0.1% (72h) 63/4200; 1.5 ± 0.07% Pou4f1 (Brn3a) E12.5 3/1650; 0.1 ± 0% (1,2,4, 8h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (12h) 31/1200; 2.5 ± 0.3% (24h) 28/580; 4.8 ± 0.2% (48h) 149/1130; 13.2 ± 0.9% E14.5 0/900; 0 ± 0% E16.5 0/900; 0 ± 0% Neuna60 (NeuN) E12.5 0/600; 0 ± 0% (1,2,4,6,8h) 0/2700; 0 ± 0% (24h) 31/580; 5.3 ± 0.7% (48h) 150/1130; 13.2 ± 1.2% E14.5 0/900; 0 ± 0% E16.5 0/900; 0 ± 0% E18.5 0/900; 0 ± 0%
85 A. B. E12.5: Pou4f2/Ki67 E12.5: Tubb3/Ki67 INBL INBL GCL/ONBL GCL/ONBL
C. D. E12.5: Calb2/Ki67 E12.5: Pou4f1/Ki67 INBL INBL GCL/ONBL GCL/ONBL
Figure 2.6. The co-expression pattern of Ki67 and Pou4f2, Tubb3, Calb2 and Pou4f1 in the embryonic retina. A. Pou4f2 (green) labels both migrating ganglion cells and cells located in the ONBL/GCL. A small subset of Pou4f2+ cells co-localized with Ki67 (red) in the INBL (white arrows) B. Tubb3 (green) labels most if not all neurons born during the retinal development and many Tubb3+ cells co-labelled with Ki67 in the INBL (white arrows). C. Calb2 (green) labelled a subset of migrating ganglion cells, which often contained Ki67 (whit arrow), as well as a subset of neurons in the ONBL/GCL. D. The expression of Pou4f1 was largely contained to the ONBL/GCL, although a minute fraction of Pou4f1+ cells co-labelled with Ki67 (white arrow). The arrow-heads point to cells co-expressing a neuronal marker and Ki67. INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bars are 10μm.
86 E12.5
Pax6/BrdU(0.5h) Pax6/Ccnd1/Vsx2 INBL GCL/ONBL
Pax6/Ccnb1 Pax6/PH3
INBL INBL
Figure 2.7. Pax6 is expressed in RPCs throughout the cell cycle. Pax6+ cells (green) were labelled by a short pulse BrdU and also co-labelled with Ccnb1 and PH3 (all red), indicating that Pax6 was expressed in S-phase and G2/M RPCs, respectively. This notion was further confirmed by Pax6 co-localizing with the pan-cell cycle markers Vsx2 and Ccnd1 (red). White arrows depict examples Pax6+ cells co-localized with the given markers. INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bar is 10μm.
87 A. B. E12.5 E12.5
Isl1/BrdU(0.5h)/Ccnd1 Pou4f2/BrdU(0.5h)/Ccnd1
NBL NBL INBL INBL
GCL GCL GCL/ONBL
Isl1/Ccnb1 Isl1/PH3 Pou4f2/Ccnb1 Pou4f2/PH3
NBL NBL NBL
Figure 2.8. Isl1 and Pou4f2 are expressed in a subset of RPCs. Top: Both Isl1 and Pou4f2 (green) were expressed in a subset of cells labelled by a short pulse of BrdU (red), indicating that they were expressed in a subset of S-phase RPCs (white arrows). However these Isl1+/BrdU+ and Pou4f2+/BrdU+ RPCs did not stain positive for Ccnd1 (white) or Vsx2 as indicated by purple arrows (A, B top and not shown). In the blow up part, white arrows depict Ccnd1+/BrdU+ cells; note that these cells display more uniform BrdU signal than that observed in the Pou4f2+/BrdU+ cells which have a more punctate BrdU staining pattern (purple arrows). Bottom: Both Isl1 and Pou4f2 were detected in a tiny number of Ccnb1+ G2 cells (red, white arrows). Isl1 was not found to co-localize with PH3 and Pou4f2 was found to co-localize very rarely with PH3 (red, white arrows). INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bars are 10μm.
88 A. B. E12.5: Tubb3/BrdU(0.5h)/Ccnd1 Tubb3/Ccnb1/DAPI Tubb3/PH3/DAPI
E12.5 E12.5 INBL INBL
Tubb3 Tubb3 ONBL C. E12.5: Tubb3/PH3/Pou4f2 INBL INBL
DAPI DAPI
D. E12.5: Tubb3/BrdU(2h)/Pou4f2 INBL INBL
Figure 2.9. Tubb3 is expressed in a subset of G2/M RPCs at E12.5. A. At E12.5, Tubb3 (green) was not detected in BrdU+ cells (red), however, a minute fraction of Tubb3+ cells co-localized with Ccnb1 and PH3 (red, purple arrows); DAPI stains nuclei, B. C. We also detected a population of PH3+/Tubb3+/Pou4f2+ cells (red, green, white, respectively, purple arrows), consistent with the idea that Tubb3 induction follows the Isl1 and Pou4f2 induction. D. The Tubb3+/BrdU+ cells appeared after a 2h BrdU chase (red, purple arrows). These cells were also Pou4f2+ (white, purple arrows). INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bars are 10μm.
89 A. C. E12.5: Calb2 Calb2
BrdU(0.5h) Ccnd1 Ccnb1 PH3 BrdU(6h) BrdU(12h) INBL INBL GCL/ONBL
B. D. E12.5: Pou2f1 Pou4f1/BrdU(12h)
BrdU(0.5h) Ccnd1 Ccnb1 PH3 INBL INBL
Figure 2.10. Calb2 and Pou4f1. A, B, Neither Calb2 nor Pou4f1 (green) co-localized with BrdU(0.5h), Ccnb1, PH3, Ccnd1 or Vsx2 (all red), indicating that they were not expressed in RPCs. C. Calb2+ cells became BrdU+ (red) after a 6h BrdU chase, while Pou4f1+ cells became BrdU+ after a 12h BrdU chase (purple arrows), D. INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. Scale bars are 10μm.
90 A.
BrdU chase (hr)
Ccnb1+ PH3+ PH3+ G1 S G2 M
C. B. PH3 Ccnb1 BrdU chase: 1hr 2hr 4hr 6hr BrdU chase: 1hr 2hr 4hr 100 100
80 80
60 60
40 40
20 20 PH3/BrdU:pH3 (%)
0 CCNB1/BrdU: CCNB1 (%) 0 E12.5 E14.5 E16.5 P0 P3 P5 E12.5 E14.5 E16.5 P0 P3 P5
E12: PH3/BrdU(2h)/DAPI E12: Ccnb1/BrdU(2h)/DAPI
Figure 2.11. The length of G2/M. A. G2/M length is determined by measuring the time necessary to label all mitotic cells with BrdU; when the BrdU pulse is too short no S-phase cells reach the end of M phase and thus some PH3+ cells remain BrdU-, but when the BrdU pulse is as long as G2/M all PH3+ cells become BrdU+ B. BrdU labeling was performed for 1, 2, 4 and 6 h at seven distinct embryonic and post-natal time points (E12.5, E14.5, E16.5, P0, P3 and P5), then determined the fraction of BrdU+/PH3+ cells. After 2 h of labeling, > 90% of PH3+ cells were BrdU+ in the embryonic retina. Thus, G2/M spans slightly over 2 hrs in the embryonic retina and expands gradually to 4 hrs by P5. C. In agreement, a greater fraction of perinuclear Ccnb1+ (G2) cells were BrdU+ for each pulse length at every stage of development. The values from which charts were constructed are provided in Table 5. Scale bars are 10μm.
91 Table 2.5. Proportion of BrdU-labelled PH3+ or Ccnb1+ cells at given time points.
Numbers represent percentage of BrdU+/PH3+ or BrdU+/Ccnb1+ cells out of the total PH3+ and Ccnb1+ cell population, respectively.
Table 2.6. Cell cycle length analyses
92 A. C. 30 50 45 25 40
(%) 20 35 + 30 15 25 /BrdU (#) /section cells - + 20 10
Ki67 15
5 ;Tubb3 10 + 5 0 0 0 0 0 0 0 0 0 12 24 48 12 24 48 12 24 48 12 24 48 72 12 24 48 72 8 12 24 48 8 12 24 48 12 24 48
E12.5 E14.5 E16.5 P0 P3 GCL BfrdU E12.5 E14.5 E16.5
B. Tubb3/Ki67/BrdU Ki67 Tubb3 BrdU
E14.5 INBL INBL 12h GCL/ONBL INBL INBL 24h GCL/ONBL INBL INBL
48h GCL/ONBL
Figure 2.12 The length of G0*. A. Retinas were exposed to a single pulse of BrdU at five developmental stages (E12.4, 14.5, 16.5, P0, 3), and the duration necessary to detect the first Ki67-/BrdU+ cells was determined. By 24 h, Ki67-/BrdU+ cells were detected at all developmental time points. B. We confirmed that the Ki67-/BrdU+ cells were differentiating neurons because they co-labeled with Tubb3. C. The chase time for newborn cells to migrate to the GCL, where Ki67 disappears, was ~ 12 hrs at E12.5 and E14.5 and between 12 and 24 hrs at E16.5, which is similar to the chase times required to observe the first BrdU+/Ki67- cells as seen in A. INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer. 93
Chapter Three: Temporal Sequence of Events during Rod, Amacrineand Bipolar Cell Births in the Mouse Retina
The work Chapter 3 was written by Marek Pacal and Dr. Rod Bremner as a manuscript in preparation.
94 3.1 Introduction
We have shown in Chapter 2 that during ganglion cell genesis induction of differentiation markers can vary considerably during development. This argues against the existence of strict pre and post mitotic gene induction programs. Instead, we proposed a model in which a common trigger induces both the exit and differentiation programs in RPCs. This trigger would precisely coordinate the beginning of exit and differentiation but both of these processes would proceed independently. Surprisingly, we also observed that even mitosis is not a watershed as several markers (e.g., Pou4f2,
Tubb3) are activated before this time point and shift to post-mitotic induction later in development. Here, we asked whether this was true for other cell types, and if so, for which markers/determinants. Further, in Chapter 2 , we defined markers that distinguish newborn ganglion retina transition cells (RTCs) from retinal progenitors cells (RPCs) and define different stages of RTC maturation. Here we defined a toolbox of such markers for rods and amacrine and bipolar interneuorns.
Dividing RPCs expand within the neuroblastic layer (NBL), and undergo interkinetic nuclear migration (INM) that involves stereotypic, although not uniform, movements along processes extending between the apical and basal surfaces of the NBL
(Norden et al. 2009). The M-phase occurs at the apical (ventricular/outer) edge of the
NBL, S-phase occurs in the basal (vitreal/inner) half of the NBL, and G1 and G2 phases occur in between ( Fig 3.1A ). Thus, nuclear position provides clues to cell cycle phase.
After the apical process retracts as the RTC migrates to its final destination and undergoes terminal differentiation.
95 Rod and cone photoreceptors are located in the outer nuclear layer (ONL), bipolar cell bodies are located in the apical (outer) part of the inner nuclear layer (INL) interspersed with the rare horizontal cells while amacrine cell bodies are located in the basal (inner) part of the INL and also make up 50% of the ganglion cell layer (GCL) cell population ( Fig 3.1A). Amacrine cells and photoreceptors cell births commence around
the same time in the early embryonic retina, but while amacrine cell births peak around
the embryonic day (E) 16, rod births peaks around the postnatal day (P) 0. Bipolar cells
are born mainly post-natally, peaking around P3. Division in the central retina ceases
around P8 and terminates in the far periphery by P10/11 ( Fig 3.1B ).
Here we determined the timing of appearance of neuronal markers with respect to cell birth. Many of these markers are transcription factors expressed throughout the retinal development. This allowed us to compare the onset of appearance at multiple developmental time points. For example, transcription factors that control photoreceptor differentiation (e.g. Crx) are excellent for this purpose since photoreceptors are generated throughout retinal development from ~E12 - ~ P7 (Carter-Dawson and LaVail 1979;
Young 1985a). Moreover, it was suggested that pre- and post natal rods constitute two distinct populations with intrinsically distinct timing of differentiation programs (Morrow et al. 1998). Our results provide molecular explanation for different timing of gene induction in early and late born photoreceptors.
96 3.2 Results
3.2.1 Quantification of neuronal marker appearance with respect to cell birth
As described in detail in Chapter 2, co-labeling of a neuronal marker with Ki67 cannot
discriminate between expression in RPCs or post-mitotic G0* cells. This issue can be
resolved by co-labeling with other cell cycle specific markers including BrdU (0.5-1h
pulse; S-phase), Ccna2 or Ccnb1 (G2/M phase), PH3 (M-phase) and the pan cell cycle
markers Vsx2 (before ~P2) and Ccnd1 ( Fig 3.2). Below, analysis was performed at
multiple developmental time points for each cell type to cover the entire period of their
birth. To determine the onset of marker expression relative to cell birth, BrdU pulse
chase analysis was used. The onset of marker expression after cell birth was calculated
as the number of hours of BrdU chase minus the G2/M length. As presented in Chapter
2, the length of G2/M is ~2h at E12.5 and ~4h at later time points.
The ensuing sections are divided into two main parts. Part (A) consists of
analyses of cell markers that are present in RPCs and are therefore unsuitable for
specifically labeling RTCs. Part (B) presents the analysis of markers that are only
expressed in RTC or later neurons but never in RPCs. These markers can be used to
different extents to distinguish differentiating rods, amacrine and bipolar cells.
Importantly, many of the markers label more than one cell type. In some cases, position
can be used to distinguish between different cell types (e.g., photoreceptors vs bipolar
cells). Also, a combination of other markers might be used to double-check cell type
specificity (e.g., rod vs cone). Each section dealing with a particular marker ends with a
conclusion as to whether this marker is suitable for RTC detection.
97 In the accompanying figures, the title box for each panel of stained sections indicates which cell cycle marker (BrdU, Ccna2/b1, PH3, Ccnd1, Vsx2) was used, and is filled yellow when that marker co-labeled cells expressing the differentiation marker, and is filled white if it never did. This information is also summarized in Figure 3.2.
Quantification is provided in Table 3.1 and 3.2. Note that negative staining could reflect absence of protein or epitope masking, but irrespective of which is actually the case, the critical issue is whether a marker is detected or not.
3.2.1.1 Markers that label both RPCs and post-mitotic cells
3.2.1.1.1 Otx2 and Neurod1 display an identical expression pattern in early and late
RPCs
Otx2 labels: RPCs, rod, cone, & bipolar cells;
Neurod1 labels: RPCs, rod, cone, amacrine, horizontal, & bipolar cells
Orthodenticle homeobox 2 (Otx2) is essential and sufficient for the determination of photoreceptors and is also involved in the late development and maturation of photoreceptor and bipolar cells (Nishida et al. 2003; Koike et al. 2007). The latter function appears to be performed in co-operation with the cone-rod homeobox transcription factor (Crx) (Chen et al. 1997; Furukawa et al. 1997; Furukawa et al. 1999;
Koike et al. 2007). Otx2 is expressed in developing and mature cones and rods, and later in bipolar cells in mice and humans (Baas et al. 2000; Nishida et al. 2003); (Fossat et al.
2007; Koike et al. 2007); (Glubrecht et al. 2009). We found that at least 25% of Otx2 +
cells expressed Ki67 and approximately 6% of Otx2 + cells were BrdU + after a 2hr pulse
at multiple time points between E14.5 and P3 ( Fig 3.3A, B, C; Table 3.1). In addition, at 98 E14.5 some Otx2 + cells were Ccnb1 + (0.25 ± 0.25% Otx2 +), and PH3 + (0.25 ± 0.0%
Otx2 +). At later time points, the signal in Ccnb1 + and PH3 + cells was indistinguishable from background and thus considered negative ( Fig 3.3B, C; Table 3.1).
Neurod1 (formerly NeuroD), is a bHLH transcription factor involved in cell fate determination and differentiation of photoreceptor and amacrine cells as well as survival of photoreceptors (Morrow et al. 1999; Inoue et al. 2002; Pennesi et al. 2003). In the embryonic mouse retina, Neurod1 +/Islet1 + cells mark developing amacrine cells
(Elshatory et al. 2007a). In the adult mouse retina, Neurod1 is expressed in
photoreceptors, amacrine cells, horizontal cells and bipolar cells (Morrow et al. 1999;
(Pennesi et al. 2003; Cho et al. 2007). We found that approximately half of Neurod1 +
cells in the E14.5 retina express Ki67. Also, the small proportions of Neurod1 +/BrdU +
(17.0 ± 2.2% Neurod1 +), Neurod1 +/Ccnb1 + (0.5 ± 0.2% Neurod1 +) and
Neurod1 +/PH3 +(0.6 ± 0.1% Neurod1 +) cells indicate that Neurod1 is expressed only in a
small subset of S, G2 and M-phase RPCs. Further, as seen in adult amacrine cells with
the same antibody (Cho et al. 2007) we found faint Neurod1 staining in the E14.5 INBL
(Fig 3.4A ) where the G0 amacrine cells would be located, suggesting that Neurod1 is
down regulated (or the epitope becomes less accessible) soon after amacrine cells
maturation.
Neurod1 is also expressed in a large number of bipolar cells in the adult mouse
retina (Cho et al. 2007). We investigated the expression pattern of Neurod1 in the early
post-natal retina when bipolar cells are generated. A large population of Neurod1 + cells
(75 ± 1.6%) expressed Ki67 at P3 and small population of Neurod1 + cells (10 ± 1.4%)
expressed Ki67 at P5. As in the E14.5 retina, a small subset of P3 and P5 Neurod1 + cells
99 was labeled by a 1hr BrdU-pulse ( Fig 3.4B, Table 3.1). This number increased after a 2 hr pulse-labeling ( Table 3.1). As reported for Otx2 above, often a very faint but irreproducible Neurod1 signal was observed in Ccnb1+ and PH3 + cells, but excluded from the counts.
Both Otx2 and Neurod1 are detected in subsets of RPCs throughout the retinal development. Thus, although both of these markers are also expressed in post-mitotic neurons they cannot be used to conclusively distinguish mitotic and post-mitotic retinal cell populations.
3.2.1.1.2 Prox 1 + RPCs increase during retinal development
Labels RPCs, amacrine, horizontal & bipolar cells
Prox1, the vertebrate homolog of Drosophila prospero is expressed in several places of the murine CNS, including the eye (Oliver et al. 1993; Torii et al. 1999; Duncan et al.
2002). In the mouse retina, prospero homeobox 1 (Prox1) is expressed in developing and mature horizontal cells, AII amacrine cells and bipolar cells (Dyer et al. 2003a; Elshatory et al. 2007a; Elshatory et al. 2007b) and is both sufficient and necessary for horizontal cell development (Dyer et al. 2003a). The expression of Prox1 coincides with neuronal commitment and differentiation in the CNS (Steiner et al. 2008). However, in several places in the CNS Prox1 is already detected in S-phase progenitors (Misra et al. 2008;
Steiner et al. 2008). In contrast, in the retina Prox1 expression was suggested to commence at G2 or after (Dyer et al. 2003a).
Consistent with the previous study (Dyer et al. 2003a), we observed both brightly
(Prox1 hi ) and weakly (Prox1 lo ) stained cells in the E14.5 and P0, P3 and P5 retina ( Fig
100 3.5, 3.6). All the Prox1 hi cells were Ki67 -, suggesting they are G0 neurons ( Fig 3.5, 3.6).
For example, at E14.5 all Prox1 hi cells were located primarily in the basal part of the
NBL, labeling G0 amacrine cells. However, some Prox1 hi Ki67 - cells were also located in the NBL ( Fig 3.5 ., 3.6). At E14.5, approximately 1/5 th of all Prox1 + cells were Ki67 + and the proportion of Prox1 +/Ki67 + cells was much increased (~60% Prox1 + cells) at P0 and
P3 ( Fig 3.5, 3.6; Table 3.1). Importantly, this double positive fraction was always
Prox1 lo (Fig 3.5, 3.6). The Prox1 lo cells were BrdU-labeled after 1hr pulse, and this proportion also increased with time (E14.5: ~ 3%, P0: ~20%; P3: ~10% Prox1 + cells; Fig
3.5, 3.6; Table 3.1). We also detected Prox1 lo staining in small populations of Ccnd1 +,
Vsx1 +, Ccnb1 +, and PH3 + RPCs at all time points ( Fig 3.5, 3.6; Table 3.1). In agreement
with the previous study (Dyer et al. 2003a), we observed that Prox1 staining intensity
rose slightly after 4 and 6 hrs of BrdU chase in the E14.5 retina ( Fig 3.5). This time coincides with the end of G2/M and cell cycle exit and suggests that Prox1 levels increase during G2/M and cell birth, as previously suggested (Dyer et al. 2003a).
However, contrary to the previous study which mapped the onset of Prox1 expression to
G2 and later, our results indicate that Prox1 is already present in S-phase RPCs.
3.2.1.1.3 Ptf1a marks a subset of early but not late RPCs in G2/M
Labels E14.5 G2/M RPCS, amacrine and horizontal neurons
Pancreas transcription factor 1a (Ptf1a) is transiently expressed in differentiating
amacrine and horizontal cells between E12.5 and P2 (Kawaguchi et al. 2002; Fujitani et
al. 2006). It was suggested that Ptf1a may mark post-mitotic differentiating amacrine and
horizontal cells (Fujitani et al. 2006). We found a significant proportion of Ptf1a +/Ki67 +
101 cells in the embryonic retina (e.g., E14.5: 19.0 ± 1.0% Ptf1a +; E16.5: 33.0 ± 1.0% Ptf1a +;
Fig 3.7; Table 3.1) and a smaller proportion in the perinatal retina (P0: 6.0 ± 1.1%
Ptf1a +). In agreement with the prior study (Fujitani et al. 2006), Ptf1a + cells were never
BrdU + after 1hr- or 2hr-long pulses. Further, Ptf1a + cells were always Ccnd1 - and Vsx2 -
(Fig 3.7 ; Table 3.1 ). Importantly, however, a small subset of cells were Ccnb1 + (0.33 ±
0.0% Ptf1a +) and PH3 + (0.33 ± 0.0% Ptf1a +) at E14.5 but were negative for these G2/M markers at earlier or later time points ( Fig 3.7; Table 3.1). Conceivably, these RPCs are biased towards or committed to the amacrine and/or horizontal cell fates. Irrespective, this marker adds to the examples of proteins highlighted in Chapter 2 (Brn3b, Isl1,
Tubb3) that are induced prior to the end of mitosis in the early retina, but then switch to post-mitotic induction.
3.2.1.2 Markers that label exclusively post-mitotic cells and can be used to detect rods and/or bipolar cells
Rod photoreceptors are born as early as E12.5, and continue to be born until as late as P7
(Carter-Dawson and LaVail 1979; Young 1985a). Here we analyzed the expression patterns of four markers that can be used to detect differentiating rod photoreceptors.
There are additional markers that label mature post-mitotic rods in the post-natal retina such as rhodopsin (Rho) or S-antigen (Arr1), although the latter also labels cones
(Treisman et al. 1988; Morrow et al. 1998); (Watanabe and Raff 1990) (Zhu et al. 2002)).
These markers never co-localized with Ki67, confirming the previous observations (data not shown). Bipolar cells are born post-natally, but many photoreceptor transcription
102 factors are also involved in bipolar cell development, therefore here we discuss rod and bipolar cells together.
3.2.1.2.1 The onset of Crx and Rcvrn appearance lengthens with development
Both label rod, cone and bipolar cells
Crx is expressed in most if not all developing and mature mouse and human cones and rods, and later in bipolar cells (Baas et al. 2000; Nishida et al. 2003; Fossat et al. 2007;
Koike et al. 2007; Glubrecht et al. 2009). Based on in situ mRNA hybridization and reporter mice, Crx is expressed in the retina as early as E12.5 (Chen et al. 1997;
Furukawa et al. 1997; Samson et al. 2009). We found that in the E14.5 retina 2.8 ± 0.8%
Crx + cells expressed low levels of Ki67 at the far peripheral retina, but Crx + cells in the
central retina were Ki67 - ( Fig 3.8; Table 3.2). At P0 and P3 ~10% and ~5% of Crx + cells were Ki67 +, respectively, regardless of the location ( Fig 3.9; Table 3.2). At all timepoints, Crx + cells were Ccnd1 -, Vsx2 -, BrdU -(2hrs), Ccnb1 - and PH3 - ( Figs 3.8, 9;
Table 3.2). Thus, Crx is induced after cell birth and during G0*. In pulse chase assays
starting at E14.5 or E16.5, Crx + cells first became BrdU + after 6-8 h or 8-12h of chase,
respectively ( Table 3.2). However, in the retinas BrdU injected at P0 and P3, the first
Crx +/BrdU + cells were detectable after 12-24h. These striking results reveal that while
E14.5 RTCs switch on Crx ~2-4h after birth, Crx does not appear in P0 and P3 RTCs
until 8-20h after birth.
Crx appeared in the forming INL around P5, presumably in differentiating bipolar cells (although some might be displaced photoreceptors (Gunhan et al. 2003)). At this time point and onwards Crx staining became fainter and mostly perinuclear ( Fig 3.9B and
103 data not shown). This is in agreement with previous reports, and appears to be characteristic for other photoreceptor markers which display a strong nuclear signal in the developing cells and become progressively diffuse and perinuclear, such as Otx2 or
Nr2e3 (see below; (Baas et al. 2000; Peng et al. 2005)).
Recoverin (Rcvrn) is a calcium-binding protein present in a large number of rods, cones, and three subtypes of cone bipolar cells in the mouse retina (Dizhoor et al. 1991;
Milam et al. 1993; Euler and Wassle 1995; Haverkamp and Wassle 2000). We detected a few Rcvrn + cells at E18.5, and this number increased afterwards ( Fig 3.10A ). These cells
were always located in the apical NBL and the forming ONL displaying typical
photoreceptor morphology, co-staining with other photoreceptor markers as previously
described (data not shown). Importantly, Rcvrn + cells did not appear in the INL until the second post-natal week ((Sharma et al. 2003); data not shown). Thus, in the period between E18 and approximately P10, Rcvrn stains photoreceptors. Moreover, after the
INL and ONL separate Rcvrn + photoreceptors and bipolar cells can be distinguished
based on their distinct locations in these layers, respectively. Importantly, Rcvrn never
co-localized with Ki67 at three time points examined (E18.5, P0, P3), suggesting that
Rcvrn is expressed in post-mitotic G0 but not G0* photoreceptors ( Fig 3.10A; Table
3.2). Indeed after BrdU pulse at E16.5 no Rcvrn +/BrdU + photoreceptors were detected
after 72h, and only after 96 h (4 days) the first Rcvrn +/BrdU + photoreceptors were detected (4.1 ± 0.3% Rcvrn +). At P0, it took 5-7 days ( Table 3.2). These long periods are reminiscent of rhodopsin induction in maturing rods (Morrow et al. 1998).
104 3.2.1.2.2 Nrl appears later than Crx in post-natal rod RTCs
Labels: rods
To define the timing of induction of a specific rod marker we studied the neural retina leucine zipper protein (Nrl), a Maf-family transcription factor. It is expressed specifically in rod photoreceptors and in the pineal gland (Akimoto et al. 2006). Nrl is essential and sufficient for rod differentiation (Mears et al. 2001; Daniele et al. 2005; Oh et al. 2007).
Nrl regulates the expression of rod-specific genes via interaction with transcription factors such as Crx and photoreceptor-specific nuclear receptor subfamily 2, group E, member 3 (Nr2e3) (Mitton et al. 2000; Cheng et al. 2004; Yoshida et al. 2004; Akimoto et al. 2006). Nrl antibodies do not reliably detect the mouse protein, but in Nrl-GFP transgenic mice, where GFP is expressed under the control of the endogenous Nrl promoter in an integrated transgene, a few GFP + newborn rods can be detected as early as
E12, with more GFP-labeled rods generated each day, peaking around P0 (Akimoto et al.
2006), reflecting the time course of rod genesis (Carter-Dawson and LaVail 1979; Young
1985a). At E16, Nrl-GFP + cells do not co-localize with the mitotic marker PH3. Further,
BrdU pulse chase performed at E16 reveals that the first rare Nrl-GFP +/BrdU + cells
appear after 4h, more so after 6h, but not after 2h of chase (Akimoto et al. 2006). Since,
the length of G2/M at this timepoint is ~4h, the onset of Nrl expression at E16 coincides
with cell birth. At P3, Nrl-GPF + rods do not co-localize with Ccnd1 or Ki67 (Akimoto et al. 2006). Here we extended this analysis to show that Nrl-GFP+ cells do not express
Ki67 or any other cell cycle maker at P0 ( Fig 3.10B, Table3.2). Although, our analysis
of Nrl is incomplete, the onset of detection of Nrl, as for Crx, is clearly delayed with
development. At E16.6, Nrl and Crx appear at cell birth and 2-4h after birth,
105 respectively. In striking contrast, at P0 or P3 it takes 8-20 hr after BrdU pulse for Crx to become detectable Crx, and because some Crx+ cells but no Nrl-GFP+ cells are Ki67+
Nrl must be induced slightly later than Crx. In agreement, all Nrl-GFP + cells were Crx +
but there were Crx + cells that lacked Nrl-GFP (data not shown).
3.2.1.2.3 Nr2e3 is expressed later than Nrl in rod RTCs
Labels: rods
Nr2e3 is a target of Nrl and its main role is to inhibit the cone-specification of photoreceptor RTCs (Cheng et al. 2004; Peng et al. 2005) (Chen et al. 2005; Oh et al.
2008)). Based on co-staining of cone markers and a series of Nr2e3 rabbit polyclonal antibodies on mouse, human and Macaque retinal sections, several studies reported that
Nr2e3 is expressed exclusively in rods (Bumsted O'Brien et al. 2004; Chen et al. 2005;
Peng et al. 2005). One study, using a different Nr2e3 rabbit polyclonal antibody recognizing mouse Nr2e3 (amino acids 80 – 395) reported the expression of Nr2e3 in a small number of mouse cones (Haider et al. 2006). The latter study also reported that ~
50% of Nr2e3 + cells express Ki67 (Haider et al. 2006). However, we have shown that
Ki67 is maintained in RTCs and (Corbo and Cepko 2005) reported (as unpublished data)
that the Nr2e3 transcript is first detected in post-mitotic cells. We used a mouse
monoclonal antibody (H7223; R&D Systems) recognizing human and rat Nr2e3 (amino
acids 80 – 395) to stain retinal sections and observed a weak nuclear signal at E16.5 in
rare cells in the ONBL ( Fig 3.10C ). The number of Nr2e3 + cells and the intensity of the signal increased dramatically after E18.5 ( Fig 3.10C ). This is in agreement with the previous study describing that Nr2e3 mRNA is first detectable by in situ hybridization
106 around E18 (Corbo and Cepko 2005) Notably, at all time points (E16.5, 18.5, P0, P3;
2,800 Nr2e3 + cells in total) never co-localized with Ki67 ( Fig 3.10C ; Table 3.2). These
data suggest that this antibody detects the cognate Nr2e3 epitope in post-G0* rods.
Indeed, BrdU pulse chase analysis indicated that rods born at E16.5 or P0 did not become
Nr2e3 +/BrdU + until 24-48h later, 20-44h after birth ( Fig 3.10C ; Table 3.2). These results suggest that although the onset of Nrl expression in RTCs varies during retinal development, the onset of Nr2e3 does not.
3.2.1.2.4 Isl1 appears in bipolar RTCs much later than in embryonic RTCs
Labels: ganglion, amacrine, bipolar cells
In the embryonic retina, Isl1 is expressed in a subset of RPCs during S-phase at E12.5,
but subsequently is only induced in post-mitotic differentiating ganglion cells and a
subset of amacrine cells (Chapter 3). At P5 and onwards Isl1 is also induced in a large
number of ON rod and cone bipolar cells (Galli-Resta et al. 1997; Haverkamp et al. 2003;
Elshatory et al. 2007a; Elshatory et al. 2007b; Pan et al. 2008). These INL Isl1 + bipolar
cells can be clearly distinguished from Isl1 + amacrine cells based on their morphology and location in the apical versus basal INL, respectively ( Fig 3.11A ). At P5, approximately 5% of Isl1 cells expressed Ki67, mostly in the peripheral retina; however they were Ccnd1 -, Vsx2 -, BrdU +(2h) -,Ccnb1 - and PH3 - ( Fig 3.11A, Table 3.2). BrdU-
pulse chase experiments revealed that Isl1 becomes detectable in bipolar cells born at P3
after 12-24h, which is ~8-20h after cell birth ( Fig 3.11A, Table 3.2). Thus, although, Ils1
is expressed in a subset of RPCs in the early embryonic retina, in the post-natal retina Isl1
appears to be expressed in a subset of G0* and G0 bipolar cells.
107 3.2.1.2.5 Gnao1, Cabp5 and Prkca label post-G0* bipolar cells
Gnao1 labels GCL cells, photoreceptors & bipolar cells
Cabp5 labels cones & bipolar cells
Prkca labels amacrine & bipolar cells
The gene product of Gnao1, Go α, is a heterotrimeric G-protein involved in the signaling cascade downstream of the metabotropic glutamate receptor 6 (mGluR6) of ON-bipolar cells (Nomura et al. 1994; Dhingra et al. 2000; Dhingra et al. 2002). Gnao1 is expressed in ON bipolar cells (Vardi et al. 1993; Vardi 1998; Haverkamp and Wassle 2000;
Haverkamp et al. 2003), but is also expressed in GCL cells from E12.5 onwards ( Fig
3.11B). Importantly, cells that express detectable levels of Gnao1 never co-localize with
Ki67 at any time-point in the embryonic or post-natal retina ( Fig 3.11B; Table 3.2). The
Gnao1 + bipolar cells appear in the INL after P5, and are thus easily distinguishable from the GCL Gnao1 + cells. In agreement, about half of the Gnao1 + became BrdU+ 48-72h after a pulse of at P3 (Fig 3.11B; Table 3.2).
Cabp5 is expressed in cones, rod bipolar cells and some ON and OFF cone bipolar cells (Haeseleer et al. 2000; Haverkamp et al. 2003; Ghosh et al. 2004; Corbo et al. 2007). The expression of Cabp5 in the forming INL appears around P5, coinciding with bipolar cell genesis ( Fig 3.12A ). Importantly, while Cabp5 signal remains high in bipolar cells it is diminished post-natally in cones. Thus, Cabp5 + bipolar cells are easily
distinguishable from Cabp5 + cones based on their location in INL as opposed to ONL,
morphology and higher Cabp5 levels. Importantly, Cabp5 never co-localized with Ki67
and retinas BrdU-labeled at P3 displayed approximately 50% Cabp5 +/BrdU + INL cells after 48h (Fig 3.12A; Table 3.2). These data indicate that a large number of Cabp5+
108 bipolar cells switch on this marker within two days after cell birth. Thus, Cabp5 is a useful marker of differentiating bipolar cells.
Prkca is a member of the PKC family of second messengers (Hug and Sarre
1993), expressed in rod bipolar cells and a subset of amacrine cells (Greferath et al.
1990); (Haverkamp and Wassle 2000). Some anti-Prkca antibodies also detect this protein in the segments of blue cones (Wikler et al. 1998); (Haverkamp and Wassle
2000). Although others reported the expression of Prkca as early as P0 or P3 in the mouse retina (de Melo et al. 2003; Sharma et al. 2003), we did not detect a reliable Prkca signal until ~P7, becoming more prominent after P8 (Fig 3.12B and data not shown) (Fig
3.12 B). Importantly , Prkca + cells never colocalized with Ki67 at P7 or P8 as Prkca + cells only occurred at these times in the central retina where Ki67 + cells no longer occurred, indicating that Prkca is a marker of post-mitotic bipolar cells (Table 3.2 and data not shown). After a BrdU pulse chase at P3…the few Prkca + cells that appeared at P7 (4d of
BrdU pulse chase) were BrdU + (data not shown), while after a 5d BrdU pulse chase ~
75% of Prkca + cells were BrdU labeled ( Fig 3.12B, Table 3.2). This indicates that
majority of Prkca + bipolar cells started expressing this marker approximately 4d after cell
birth coinciding with the appearance of this marker in the retina. Since bipolar cell
bodies are located at the apical part of the INL, Prkca + bipolar cells can be easily distinguished from the scarce Prkca + amacrine cells located in the basal INL ( Fig 3.12B ).
109 3.2.1.3 Markers that label exclusively post-mitotic cells and can be used to detect amacrine cells
Amacrine cells are born mainly pre-natally but a small number of them are born perinatally. Thus we followed amacrine development from E12.5 to P0.
Four of the six ganglion cell markers described in Chapter 2 also label amacrine cells: Pax6, Isl1, Calb2 and Neuna60. As described in Chapter 2, in the embryonic retina only Calb2 and Neuna60 can exclusively discriminate post-mitotic amacrine neurons from dividing RPCs. In the early embryonic retina, Calb2 labels migrating ganglion and amacrine cells, and both Calb2 and Neuna60 label a mixture of ganglion and amacrine population in the INBL/GCL ( Chapter 2 and data not shown). However,
Calb2 is an excellent marker of migrating perinatal amacrine RTCs, as at this time virtually no ganglion cells are born. Further, both Calb2 and Neuna60 distinctly label amacrine cells in the basal INL and are thus excellent amacrine markers in the post-natal retina (Haverkamp and Wassle 2000) (Raymond et al. 2008).
Further, two bipolar cell markers, Gnao1 and Prkca (see above) also label rare unknown subtypes of amacrine cells (Greferath et al. 1990; Vardi et al. 1993; Vardi 1998;
Haverkamp and Wassle 2000; Haverkamp et al. 2003). Neither of these markers co- localized with Ki67 at any stage analyzed (Go α: E12.5-P8; Prkca: P5-P8). Here we describe the expression pattern of seven additional markers that can be used to label subsets of amacrine neurons.
110 3.2.1.3.1 Tfap2a is induced at the end of G0* in differentiating amacrine cells
Tfap2a labels: amacrine cells
The transcription factor Tfap2a (AP-2α), is expressed in a large number of amacrine cells
(Bassett et al. 2007). Very few Tfap2a + cells were detected as early as E12.5 in some migrating neurons in the NBL but mostly in the basal NBL. These cells did not co-label with Ki67, and in pulse chase assay were not labeled by a 12h chase and the first
Tfap2a +/BrdU + cells appeared after a 24h chase ( Table 3.3). The number of Tfap2a +
cells increase dramatically after E14.5 (Bassett et al. 2007). A minute number of Tfap2a +
cells co-labeled with Ki67 at E14.5 and E16.5 (0.2 ± 0% Tfap2a +) but not at P0; and
Tfap2a never co-localized with any of the five other cell cycle markers (BrdU(1-2h),
Ccnb1, PH3, Ccnd1, Vsx2) ( Table 3.3 and data not shown). In retinas pulsed with BrdU
at E14.5 and E16.5, the first Tfap2a +/BrdU + cells were detected already after 8-12h.
Since these BrdU-labeled cells were also Ki67 +, this suggests that the earliest Tfap2a is induced is ~4-8h after cell birth in G0* ( Fig 3.13A, Table 3.3). Importantly, in the vast majority of cells between E12.5-P0, Tfap2a is only detectable in the post G0* amacrine neurons.
3.2.1.3.2 Dcx does not label proliferating cells in the retina
Dcx labels: ganglion, amacrine and horizontal cells
Dcx (doublecortin) is a microtubule associated protein expressed widely in differentiating
CNS neurons (Gleeson et al. 1999; Brown et al. 2003). Dcx levels decline in the adult brain except for the areas where new neurons are still produced, the subventricular zone and the dentate gyrus. In that context, Dcx was detected in proliferating progenitors that
111 were also labeled with Ki67 or a 2hr pulse of BrdU (Brown et al. 2003). In the retina,
Dcx expression has been reported in developing ganglion, amacrine and horizontal cells
(Lee et al. 2003; Trimarchi et al. 2007; Lee et al. 2008; Wakabayashi et al. 2008), but accurate temporal analysis of the staining pattern is lacking. We found a strong Dcx signal in the GCL as early as E12.5, and in presumed migrating neurons in the NBL ( Fig
3.13B ). At E14.5 and onwards, a strong Dcx signal appeared in the cell bodies of some cells in the middle of the NBL ( Fig 3.13B ) which, based on a previous study are likely to be differentiating horizontal cells (Lee et al. 2003). Dcx disappeared from the cell bodies of amacrine and ganglion cells around the second post-natal week, and from the cell bodies of horizontal cells after the third post-natal week, while it remained detectable in the plexiform layers (data not shown). In the embryonic retina, a few Dcx + cells in the
NBL co-localized with Ki67 (E12.5: 13.9 ± 1% Dcx +; E14.5: 3.4 ± 0.3% Dcx +; E16.5: 0.3
± 0.1% Dcx +; Fig 3.13B; Table 3.3). However, Dcx + cells were not labeled by a short pulse of BrdU (1-2h) and did not co-localize with Ccnd1, Vsx2, Ccna2, Ccnb1 or PH3.
This suggests that Dcx +/Ki67 +cells were not RPCs but G0* RTCs. The first Dcx +/BrdU + cells appeared after 4-6h and 8-12h of BrdU chase at E12.5 and E14.5/16.5 respectively
(Fig 3.13B, Table 3.3).. Thus, Dcx is detectable in G0* RTCs 2-4h and 4-8h after cell birth at these time points, respectively. Dcx did not co-localize with Ki67 at P0, indicating that either Dcx is either not expressed in later born amacrine cells, or it is expressed after G0* ( Table 3.3 and data not shown). Importantly, in contrast to the cortex, Dcx was detected in dividing RPCs, suggesting alternative regulation of this protein in these two tissues.
112 3.2.1.3.3 Elavl 2/3/4 are induced at the end of G0*
Elavl 2/3/ 4: ganglion, amacrine and horizontal cells
Human antigens B, C and D (Elavl 2, 3 and 4) are closely related RNA binding proteins of the embryonic lethal abnormal visual (Elav) family (Levine et al. 1993; Liu et al.
1995). The HuC/D 16A11 antibody that recognizes all three of these proteins labels rat retinal ganglion and amacrine cells and transiently labels horizontal cells (Ekstrom and
Johansson 2003). In the early embryonic retina, Elavl 2/3/4 labeled a few migrating neurons in the NBL and many cells in the forming GCL. Only a minute percentage of
Elavl 2/3/4 + cells were also Ki67 + (E12.5 and E14.5: 0.2 ± 0% Elavl 2/3/4 +) and none co- stained with the cell cycle markers BrdU (0.5h), Ccnd1, Vsx2, Ccna2/b1 or PH3 ( Fig
3.14A, Table 3.3), suggesting that this marker is induced after Dlx. Indeed, BrdU pulse chase showed that at E12.5, the first Elavl 2/3/4 +/BrdU + cells were detected 6- 8h after
the pulse (0.3 ± 0.1% Elavl 2/3/4 +; i.e., ~4-6h after birth), while at E14.5 these cells were detectable after 8-12h, after the pulse (1 ± 0.1% Elavl 2/3/4 +; i.e., ~4-8h after birth)
(Figure 3.14A, Table 3.3 ). Thus, in the embryonic retina Elavl 2/3/4 labels G0* RTCs
later than Dcx by approximately 2h. Importantly, at E16.6 and P0, virtually no migrating
Elavl 2/3/4 + cells can be detected in the NBL, and no Elavl 2/3/4 + cells co-labeled with
Ki67 ( Figure 3.14A, Table 3.3, and data not shown , (Ekstrom and Johansson 2003)).
This suggests that later born amacrine cells may not express Elavl 2/3/4 or they may switched it on later, possibilities we have not addressed.
113 3.2.1.3.4 Post-G0* markers: Mtap1b, Uchl1, Calb1, GABA
Mtap1b, Uchl1, Calb1 label: ganglion, amacrine, horizontal cells
GABA: amacrine cells
In addition to Neuna60 (NeuN), described in Chapter 2, the staining pattern of six more
markers that label amacrine cells beyond G0* are characterized below.
Microtubule associated protein 1B (Mtap1b, formerly Map1b) is the first
microtubule-associated protein expressed in neural development and is expressed in the
majority if not all retinal neuronal subtypes in the adult mouse and rat retina (Ramon-
Cueto and Avila 1997; Meixner et al. 2000; Pattnaik et al. 2000). Uchl1 (also known as
Pgp 9.5) is a member of the ubiquitin C-terminal hydrolase family (Thompson et al.
1983; Wilkinson et al. 1989) and is expressed in retinal ganglion and horizontal cells of
several mammalian species (Bonfanti et al. 1992). Uchl1-labeled many cells in the GCL
that were Ki67 - at all time points analyzed, and Mtap1b showed identical staining pattern
(Fig 3.15A, Table 3.3). In addition, many Uchl1 +, and some Mtap1b + cells that lacked
Ki67 were also detected in the NBL which are presumably migrating post G0* amacrine and/or horizontal cells. Further, a strong signal did not appear in the GCL for any of these markers until E13, the time when many ganglion cells have already migrated into the GCL, confirming that these three markers label more mature ganglion neurons. In
BrdU pulse chase assays the first Mtap1b + or Uchl1 + cells that were labeled 12-24h after
the pulse; this number approximately doubled after another 24h ( Fig 3.15A; Table 3.3).
Thus, these markers are expressed by neurons approximately 10-22h after the last M- phase.
114 Calb1 is a member of a large family of EF-hand intracellular calcium-binding proteins, (Schwaller et al. 2002; Schwaller 2009). In the rodent retina, Calb1 is expressed in horizontal, amacrine and ganglion cells (Uesugi et al. 1992; Peichl and Gonzalez-
Soriano 1994; Wassle et al. 1998; Haverkamp and Wassle 2000; Sharma et al. 2003). We detected a minute number of Calb1 + cells as early as E16.5 in the NBL. Based on previous studies these are presumed to be differentiating horizontal cells (Uesugi et al.
1992; Sharma et al. 2003). Calb1 + cells were not clearly detectable in the INBL/GCL until after E18.5, in contrast to other amacrine markers (e.g., Elavl 2/3/4, Tfap2a) that are located at the same position at least as early as E14 ( Fig 3.15B, Table 3.3).
γ-Aminobutryic Acid (GABA) is a major inhibitory neurotransmitter present in
approximately half of all amacrine cells (Marc et al. 1995; Crook and Pow 1997; Vaney
2002). GABAergic amacrine cells are born as early as E14 (Cherry et al. 2009).
However, similarly to Calb1, immunostaining detected GABA from E18.5 onwards in the
INBL/GCL, the number and staining intensity increasing with age ( Fig 3.15B ).
Moreover, GABA+ cells never co-stained with Ki67, suggesting that, as Calb1, GABA is
a marker of G0 amacrine cells ( Fig 3.15C; Table 3.3).
3.3 Discussion
Previously we observed that ganglion cell specific neuronal markers can be induced at
variable times relative to cell birth. Here, we analyzed the dynamics of induction of
markers recognizing other retinal cell types including rods and amacrine and bipolar
interneuorns. Using this knowledge we defined a toolbox of markers that can distinguish
115 these cell types from dividing progenitors and mark different stages of their development.
.
3.3.1 Panel of RTC markers
Here we discuss the suitability of markers described in section B to label rod/amacrine and bipolar RTCs and neurons.
Crx clearly detects photoreceptor/bipolar RTCs. However, it does not distinguish
between cones and rods in the early retina. Crx + photoreceptors can be distinguished
from Crx + bipolar cells based on location, i.e., ONL vs. INL, respectively, although co-
staining, e.g. with Chx10, would be required to rigorously distinguish the bulk bipolar
population from rare ectopic photoreceptors in the INL . However, Crx staining becomes
weak after ~P5, thus it is not an ideal bipolar RTC marker. As Crx, Rcvrn also cannot
distinguish between rods and cones and labels them several days after birth. Further,
Rcvrn labels bipolar neurons approximately a week after cell birth. Thus, Crx and Rcvrn
are good markers recognizing photoreceptor RTCs, but immunostaining with an
additional cone or rod specific marker will have to be used to distinguish between these
two cell types. In contrast, Crx and Rcvrn are not useful as bipolar RTC markers.
In contrast to Crx or Rcvrn, Nrl-GFP labels specifically rod photoreceptor RTCs;
in the embryonic retina immediately after cell birth, whereas in the post-natal retina it is
somewhat later. Also, Nr2e3, with the antibody used here, labels rod RTCs at least 20h
after cell birth in both embryonic and post-natal retina. Thus, Nrl-GFP and Nr2e3 are
excellent rod RTC markers.
In striking contrast to early embryonic retina where Isl1 can be detected in small
subset of RPCs (see Chapter 2 ), Isl1 labels G0*/G0 bipolar cells in the post-natal retina. 116 Moreover, it can be easily distinguished Isl1 + amacrine and ganglion cells. As such it can
be used as a bipolar RTC marker.
Gnao1, Cabp5 and Prkca mark post G0* bipolar cells appearing several days after cell birth of this cell type. Also, they are easily distinguishable in the INL from amacrine cell that also stain for these markers. Thus, although, these markers do not label immature bipolar RTCs, they are still useful as markers of developing bipolar neurons.
In the mouse retina Tfap2a exclusively labels amacrine neurons (Bassett et al.
2007). We observed that in contrast to the published results this marker labels a minute
number of E14.5 and E16.5 G0* amacrine RTCs. However, the vast majority of Tfap2a + amacrine neurons are Ki67- post G0* amacrine. Thus, Tfap2a is an excellent marker of amacrine RTCs.
In the embryonic retina, Dcx is in post-mitotitc cells although on its own can not
distinguish ganglion and amacrine cells in the INBL/GCl and migrating recently born
ganglion, amacrine and horizontal cell RTCs in the ONBL. However, in the post-natal
retina, Dcx + horizontal cells in the apical INL and amacrine cells in the basal INL are easily distinguishable, thus Dcx can be used to label these neuronal populations. Elavl
2/3/4 d displays almost identical expression pattern as Dcx, and thus, although not cell type specific, it is, however, a useful marker of post-natal amacrine cells in the basal INL.
3.3.2 Flexibility in expression of Isl1 and Crx during retinal development
In Chapter 2 we introduced the concept of “The Trigger Theory”, which posits that cell cycle progression and cell birth and differentiation run in parallel but are not rigidly connected beyond a common initiation point in an RPC. One of the consequences of this 117 model is that transcription factors that drive certain differentiation programs can be expressed prior to or after cell birth. Indeed, we showed that Isl1 was expressed in ganglion cells in late S-phase RPCs at E12.5, while it was expressed at cell birth at E14.5
(Chapter 2 ). In agreement with this trend we show here that Isl1 is expressed at P3, the
peak of bipolar cell birth, 20-44h after cell birth. This extraordinary range of expression
onset may reflect either different functions of Isl1 during the course of retinal
development in the different cell types, or the lengthening the early stages of RTC
differentiation. Strikingly, the expression pattern of Crx shows a similar trend. At E14.5,
Crx is detectable at cell birth, at E16.5 slightly after (4-8h) cell birth, and at P0/3 it is
detectable only 8-20h after cell birth. The increasing delay in the onset of expression
appears to be related to the general trend of lengthening developmental processes
including S-phase, the entire cell cycle (Alexiades and Cepko 1996), as well as G2/M and
G0* ( Chapter 2).
3.3.3 Insights into photoreceptor development
According to a current model of photoreceptor development Crx is expressed via Otx2
and one of its roles is to help conferring a ‘default’ S cone state upon photoreceptor RTCs
(Swaroop et al. 2010), however, Crx alone does not determine specific photoreceptor cell
fate (Furukawa et al. 1999). Nrl inhibits this default cone state, promoting the rod
photoreceptor fate instead (Swaroop et al. 2010). Nrl activates rod-specific genes and
inhibits cone-specific genes. The first function is carried out partly in tandem with Crx as
these two factors co-regulate some rod genes, and in mice lacking Crx rods do not
differentiate properly as they lack outer segments and display a drastically reduced 118 expression of rod genes including Rho (Furukawa et al. 1999; Mitton et al. 2000; Pittler et al. 2004; Yoshida et al. 2004). Nrl cooperates with Nr2e3 to repress some but not all of the Nrl targeted, as Nrl is able to repress some cone genes by itself or via other unknown factors (Corbo and Cepko 2005; Corbo et al. 2007; Hsiau et al. 2007). In view of these prior data we can predict that a) Nrl should be detectable in photoreceptor RTCs around the same time as Crx and b) the expression of Nrl and Nr2e3 may not be identical.
Indeed Crx and Nrl do appear approximately at the same time. At E16.5, Crx and
Nrl appear in E16.5 photoreceptor RTCs immediately after cell birth. Strikingly, they both appear in the post-natal RTCs much later (8-20h) after birth, suggesting that rod commitment occurs much earlier in the embryonic versus post-natal RTCs.
Regarding the relative expression of Nrl and Nr2e3, while Nrl is detectable as early as E12.5 (Akimoto et al. 2006), we detected weak Nr2e3 signal in the retina only after E16.5, while robust stain was not detected until after E18.5. This is in agreement with a previous study indicating that Nr2e3 mRNA is detectable only after ~E18 (Corbo and Cepko 2005). Also, RTCs that were pulse labeled at E16.5 only became
Nr2e3 +/BrdU + as late as 20-44h later, almost a day after Crx and Nrl. These results indicate that there is a long delay between Nrl induction and when it begins exerting its functions through Nr2e3. Interestingly, in the post-natal RTCs Crx, Nrl and Nr2e3 appeared to be induced at approximately at the same time, ~24h after birth, delaying cell fate specification in these cells. These patterns of transcription factor induction reveal that the expression of genes controlling rod fate and differentiation can be expressed at variable times after cell birth during retinal development. Indeed, it was previously suggested that pre- and post-natally born photoreceptors are two distinct populations with
119 distinct timing of gene (rhodopsin) induction (Morrow et al. 1998). Our results provide a molecular mechanism to explain distinct gene induction in pre and post natal photoreceptors.
3.3.4 Prox1 is expressed already in S-phase RPCs
Prox1 has been implicated in the development of various tissues including the lymphatic system, liver, lens and the retina (Wigle et al. 1999; Wigle and Oliver 1999; Sosa-Pineda et al. 2000; Dyer et al. 2003b). However, the precise roles of Prox1 in these tissues are not clear. Intriguingly, Prox1 affects proliferation, thus one of the roles of Prox1 might involve cell cycle control. The Drosophila homologue of Prox1, Prospero, is a brain tumor suppressor inhibiting neuroblast self-renewal, and decreased Prox1 expression has been observed in hepatocellular carcinomas and biliary duct cancers (Shimoda et al.
2006; Laerm et al. 2007). In agreement, the loss of Prox1 in the lens and the retina leads to increased proliferation (Wigle et al. 1999; Dyer et al. 2003a). On the other hand,
Prox1 overexpression in cultured vascular endothelial cells, leads to up-regulation of cyclin E1 and E2 mRNAs; and strikingly, Prox1 functions downstream of the β-
catenin/TCF pathway in the colorectal cancer, promoting dysplasia, tumor growth, and
malignant progression (Petrova et al. 2002; Petrova et al. 2008). Thus, the role of Prox1
in cell cycle control might be context dependent. Indeed, Prox1 functions as a
transcription activator or a repressor in different contexts (Lengler et al. 2001; Petrova et
al. 2002; Qin et al. 2004; Steffensen et al. 2004). The loss of Prox1 in the lens leads to
upregulation of the cell cycle inhibitors p27(Kip1) and p57(Kip2) (Wigle et al. 1999),
120 however the direct targets, and hence the precise roles of Prox1 in the CNS and the retina are currently unknown.
The expression of Prox1 in the CNS appears to coincide with differentiation. In the embryonic mouse and chick spinal cord Prox1 + cells are positioned between dividing
ventricular zone (VZ) progenitors and differentiating mantle zone (MZ) neurons (Misra
et al. 2008). Similarly, in the adult mouse hippocampus, Prox1 expression coincides with
neuronal commitment and differentiation (Steiner et al. 2008). Strikingly, Prox1 can be
detected in a small population of S-phase progenitors as revealed by Prox1/BrdU co-
expression after a short pulse BrdU labeling (Misra et al. 2008; Steiner et al. 2008). A
previous study that used cultured and dissociated retinal tissue concluded that Prox1
expression commences in RPCs at the end of M-phase (Dyer et al. 2003a) ,.
We used a different anti Prox1 antibody, BrdU-labeled mouse embryos in vivo as
opposed to in vitro and stained retinal sections as opposed to dissociated cells. This in
vivo strategy revealed that a large proportion of Prox1+ cells (~20% of E14.5 and ~60% of P3/P5 Prox1 +cells) were labeled by a 1h pulse of BrdU. These cells always displayed
a Prox1 lo signal compared to Prox1 hi /BrdU - cells. Further, Prox1 lo cells co-labeled with
Ccnb1, Ccna2, PH3, Ccnd1 and Vsx2, indicating that Prox1 is detectable in S and G2/M phases of the cell cycle. Thus, contrary to the previous study which placed the onset of
Prox1 expression to G2, we conclude that, Prox1 is already present in many S-phase
RPCs. This finding calls for revising the Prox1 function in the retinal development; especially since Prox1 has been implicated in such diverse functions as energy homeostasis, cell fate control or induction of tumor progression (Charest-Marcotte et al.;
Petrova et al. 2008; Lee et al. 2009).
121 3.3.5 Does Ptf1a label a subset of RPCs biased towards amacrine/horizontal neurons?
Previously, we have described G2/M RPCs that might be biased towards the ganglion cell fate since they already express the ganglion cell fate determinant Pou4f2 and the neuronal marker Tubb3. It is possible that Ptf1a could similarly mark a subset of amacrine cells specified just before cell birth.
Intriguingly, the ability of rat RPCs to be specified towards amacrine cell fate during G2/M has been previously reported, based on the induction of VC1.1 (CD57) and syntaxin (Belliveau and Cepko 1999). Thus, some amacrine cells appear to be specified in the rodent retina prior or during the last M-phase. The extent of this mechanism and which amacrine sub types might be generated in this way awaits further study.
Interestingly, Ptf1a has been recently described to specify fate in pancreatic progenitors in response to changing levels of Notch signaling (Schaffer et al.). Since the levels of Notch signaling affect retinal neuronal vs. progenitor commitment (Livesey and
Cepko 2001), it will be intriguing to investigate whether Ptf1a is also conferring amacrine/horizontal cell fate in response to Notch signaling.
In conclusion, the dynamics of expression of retinal markers, many of which are important transcription factors, with respect to cell birth may vary dramatically over the course of the retinal development. We used this information along with co-expression pattern of these proteins with cell cycle markers to build a panel of markers that allow us to distinguish retinal neurons from proliferating cells. Crucially, these analyses revealed new insights into photoreceptors development and the expression of Prox1 and Ptf1,
122 transcription factors that play critical roles in the differentiation of amacrine and horizontal cells.
123 A. E16 P3 P21 RPE RPE RPE M M
G0* G0* ONL OPL
G1 NBL G1 INBL G2 G2 G0 G0 INL S S
L IPL IPL GCL GCL GCL/ONB
RPC RTC G H C A R B M
Mature retinal cells
B.
G H C A R B M
E10.5 E12.5 E14.5 E16.5 E18.5 P0 P3 P5 P8
Figure 3.1. Overview of retinal structure and development. A. In the embryonic retina, RPCs pass through the cell cycle in a repetitive predictable pattern, S-phase in the apical INBL, M-phase at the ventricular side, with G2 and G1 in between. The ganglion and amacrine cells populate the ONBL which eventually forms a distinct GCL. Throughout the retinal development six retinal subtypes fill the NBL which after ~P3 splits into ONL with the photoreceptors and INL with horizontal, bipolar and amacrine neurons and Müller glia. Note that the number of ganglion cells is reduced approximately by half during the first post-natal week. B. The summary of he time periods during which retinal cell types are generated during development. For example, a few rods are born as early as ~E12 and as late as ~P7, but the peak of births occur ~P0. While amacrine cells are born mostly between E14-18, and only a few are born post-natally. E- embryonic day; P- post-natal day; INBL/ONBL/GCL: inner neuroblastic layer/outer neuroblastic layer/ganglion cell layer; RPC: retinal progenitor cell; RTC- retinal transition cell.
124 RPC S G2 M G0* RTC G0 Neuron Ccnd1, Vsx2 (up-to~P2) BrdU (1-2 hr pulse)
PH3 Perinuclear/nuclear Ccna2a/b1 Ki67 Neurod1, Prox1 RPC A H Ptf1a 0-2h Calb2, A G
E14.5 4-8h Dcx, Elav2/3/4 A
Tfap2a H A G
Neurod1, Otx2 RPC C R
Nrl R 8-12h Crx E16.5 C R 24-48h Nr2e3 R 72-96h Rcvrn C R
Neurod1, Otx2 RPC C R
* Nrl R 12-24h Crx P0 C R 24-48h Nr2e3 R 5-7d Rcvrn C R
Vsx2, Neurod1, Otx2 RPC C R 12-24h Crx B C R
P3 24-48h Isl1, Gnao1, G A B Cabp5 C B ~5d Prkca A B ~10d Rcvrn C R B Figure 3.2. Summary of cell cycle and neuronal markers used in this study and the temporal pattern of their appearance at selected time points. Top: Six cell cycle markers were used to mark the distinct stages of cell cycle. Ki67 is expressed during the cell cycle and remains in RTC for the length of time termed by us as G0*. The disappearance of Ki67 then marks the onset of G0. Bottom: The length of the arrow and the number above the arrow depict the time in hours after cell birth when the marker is first detected, based on BrdU chase. The BrdU chase data are missing for Nrl at P0/P3, but Nrl does not co-label with Ki67, thus it is expressed after G0*. The boxes on the right depict the cell types in which the markers are observed. Purple arrows represent markers that label both RPCs and neurons, and black arrows depict RTC/neuronal markers. Ptf1a is colored green as it appears to label a specific G2/M subset of RPCs. RPC: retinal progenitor cell; G: ganglion cells; A: amacrine cells; C: cones; R: rods; B: bipolar cells. 125 Table 3.1. Analyses of markers that are expressed in both RPCs and RTCs/neurons. The numbers represent percentage of cells positive for the given cell cycle marker at different time points (e.g., (Otx2+, Ki67+) cells / all Otx2+ cells).
Marker/Age KI67 BrdU CCNB1 pH3 CCND1 VSX2 Otx2 E14.5 210/780; 26.6 ± 3.2% (2h) 67/830; 8.2 ± 1.4% 3/1200; 0.2 ± 0.2% 3/1200; 0.2 ± 0.0% P0 382/665; 56.8 ± 2.8% (2h) 54/830; 6.4 ± 0.6% 0/900; 0 ± 0% 0/900; 0 ± 0% P3 198/822; 24.1 ± 0.2% (2h) 44/840; 5.2 ± 0.5% 0/900; 0 ± 0% 0/900; 0 ± 0% Neurod1 E14.5 487/1050; 45.9 ± 3.6% (1h) 123/725; 17.0 ± 2.2% 6/1200; 0.5 ± 0.2% 8/1220; 0.6 ± 0.1% P3 600/800; 75 ± 1.6% (1h) 37/920; 4.0 ± 0.3% 0/900; 0 ± 0% 0/900; 0 ± 0% (2h) 155/700; 22.2 ± 2.5% P5 210/2100; 10 ± 1.4% (1h) 7/2100; 0.3 ± 0.0% 0/900; 0 ± 0% 0/900; 0 ± 0% (2h) 14/2100; 0.6 ± 0.0% Prox1 E14.5 154/766; 19.9 ± 1.1% (1h) 24/760; 3.25 ± 0.5% 3/900; 0.3 ± 0.0% 3/900; 0.3 ± 0.0% 137/800; 17.1 ± 0.4% 132/760; 17.7 ± 2.5% (2h) 38/720; 5.2 ± 0.3% (4h) 91/760; 11.8 ± 0.5% (6h) 111/760; 14.6 ± 0.5% P3 462/770; 59.8 ± 0.9% (1h) 165/900; 18.3 ± 1.6% 12/900; 1.3 ± 0.0% 10/900; 1.1 ± 0.1% (2h) 292/1080; 27.1 ± 1.0% P5 450/760; 58.7 ± 2.1% (1h) 88/780; 11.2 ± 0.3% 10/900; 1.4 ± 0.5% 6/760; 0.8 ± 0.3% (2h) 117/780; 14.8 ± 1.0% Ptf1a E12.5 26/150; 17.5 ± 2.5% (1,2h) 0/200; 0 ± 0% 0/150; 0 ± 0% 0/150; 0 ± 0% 0/150; 0 ± 0% 0/150; 0 ± 0% E14.5 114/600; 19.0 ± 1.0% (1, 2h) 0/1500; 0 ± 0% 3/900; 0.3 ± 0.0% 3/900; 0.3 ± 0.0% 0/900; 0 ± 0% 0/900; 0 ± 0% (4h) 8/1500; 0.5 ± 0.1% (6h) 54/1500; 3.6 ± 0.4% E16.5 198/600; 33.0 ± 1.0% (2h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% E18.5 168/740; 22.7 ± 0.9% (2h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% P0 43/714; 6.0 ± 1.1% (2h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0%
126 A. B. Otx2: E14.5 Otx2: P0
Ki67 Ki67 BrdU (2h)
INBL INBL INBL ONBL
BrdU (2h) Ccnb1 pH3 ONBL
INBL Ccnb1 pH3
INBL C. Otx2: P3
Ki67 BrdU (2hr)
INBL INBL
Figure 3.3. Otx2 is expressed in a subset of RPCs. A. At E14.5 Otx2 (green) is expressed in Ki67+ cells (red) and co-localizes with BrdU, Ccnb1 nd PH3 (all red) which label S and G/M and M phases of cell cycle, respectively, depicted by Ccnb1 pH3 white arrowheads. B, C. At P0 and P3, Otx2 is still detected in Ki67+ cells but is detected only in S- phase RPCs, as it is virtually absent from Ccnb1 and PH3+ G2/M cells. Note that at all timepoints Otx2 also labels post-mitotic photoreceptors/bipolar INBL cells (open arrowheads). Scale bars are 10μm.
127 A. Neurod1: E14.5
Ki67 BrdU (1h) Ccnb1 PH3
INBL INBL ONBL
B. Neurod1: P3 Neurod1: P5
Ki67 BrdU (1h) pH3 Ki67
INBL INBL ONBL BrdU (1h)
Figure 3.4. Neurod1 is expressed a in a subset of RPCs. A. At E14.5 Neurod1 (green) is expressed in Ki67+ cells and co-localizes with BrdU, Ccnb1 and PH3 (red, white arrowheads). B. At P3 and P5, Neurod1 is still present in Ki67+ cells INBL but is detected in only in S-phase RPCs, as it is virtually absent from Ccnb1 and PH3+ cells. Empty arrowheads point to post-mitotic Neurod1+ amacrine/photoreceptor/bipolar cells. Scale bars are 10μm. 128 Prox1: E14.5
Ki67 Ccnd1 Vsx2 Ccnb1 PH3
INBL INBL
BrdU (1h) BrdU (2h) BrdU (4h) BrdU (6h) INBL
Figure 3.5. Prox1 is expressed in S-phase RPCS at E14.5. Two types of Prox1 staining (green) were observed. A bright Prox1hi signal appeared in cells which did not co-localize with Ki67 or any other cell cycle markers (red). These were likely amacrine and horizontal RTCs/neurons (open arrowheads). A weak Proxllo signal was observed in cells that co-labelled with Ki67, Ccnd1, Vsx2, Ccnb1, PH3 and were labelled by a short pulse of BrdU, indicating that Prox1 was detected in RPCs in all cell cycle phases (white arrowheads). Bottom panels show blow ups of the cells to which white arrowheads are pointing above. Note that Prox1 signal increased in Prox1+/BrdU+ cells (white arrows) with the length of the BrdU chase. Scale bars are 10μm. 129 Prox1:P3 Prox1:P5
Ki67 Ki67 INBL ONBL BrdU (2h) BrdU (2h)
INBL
Ccnb1 Ccnb1
pH3 pH3
Figure 3.6. Prox1 is expressed at S-phase RPCs at P3 and P5. As in the embryonic retina, in the postnatal retina two types of Prox1 staining (green) were also observed. Bright Prox1hi signal appeared in cells which did not co-localize with Ki67 or any other cell cycle markers (red). These cells could be amacrine, horizontal or bipolar cells (open arrowheads). Weak Prox1lo signal was observed in cells that co-labelled with Ki67, Ccnd1, Vsx2, Ccnb1, PH3 and were labelled by a short pulse of BrdU (red), indicating that Prox1 was detected in RPCs in all cell cycle phases (white arrowheads). Scale bar is 10μm.
130 E12.5 E14.5 E16.5 E18.5 Ptf1a: E12.5
Ptf1a/Ki67 Vsx2 BrdU (1h) Ccnb1 PH3
INBL INBL
Ptf1a: E14.5
Vsx2 BrdU (2h) Ccnb1 PH3 ONBL
INBL
BrdU (4h)
INBL
BrdU (6h)
Figure 3.7. Ptf1a is expressed in a subset of G2/M RPCs at E14.5. Ptf1a (green) co-localized with Ki67 (red) at all time points, but only co-localized with Ccnb1 and PH3 at E14.5, suggesting that at this time point it was co-expressed in a subset- of G2/M RPCs (white arrowheads). Indeed, at E14.5 Ptf1a+/BrdU+ cells were detected as early as after 4h of BrdU-pulse chase (i.e., at cell birth). The number of Ptf1a+/BrdU+ cells increased with the length of chase. Scale bars are 10μm.
131 Table 3.2. Analyses of markers that are expressed in photoreceptor/bipolar RTCs and neurons. The numbers represent percentage of cells positive for the given cell cycle marker at different time points (e.g., (Crx2+, Ki67+) cells / Crx+ cells). The orange and gray panels represent co-localization or the lack of co-localization of neuronal markers with proliferation markers, respectively. Marker/Age KI67 BrdU CCNB1 pH3 CCND1 VSX2 Crx E14.5 21/762; 2.8 ± 0.8% (2,4,6h) 0/1200; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (ONL) (8h) 21/1200; 1.7 ± 0.2% (10h) 91/1200; 7.5 ± 0.6% E16.5 47/800; 5.8 ± 0.1% (2, 8h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (ONL) (12h) 10/900; 1.1 ± 0.1% (24h) 127/1200; 10.5 ± 0.6% P0 78/743; 10.3 ± 1.2% (12h) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (ONL) (24h) 232/1200; 25.7 ± 0.8% (48h) 435/1200; 48.3 ± 1.6% P3 36/675; 5.46 ± 0.7% (12h) 0/900; 0 ± 0% (ONL) 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (24h) 195/1200; 21.6 ± 1.6% (ONL) (48h) 217/1200; 24.6 ± 1.2% (total) Rcvrn E16.5 (72h) 0/900; 0 ± 0% (ONL) (96h) 50/1200; 4.1 ± 0.3% (ONL) P0 0/900; 0 ± 0% (5d) 0/900; 0 ± 0% (7d) 40/1200; 4.4 ± 0.5% (ONL) (9d) 234/1200; 26 ± 1.2% (ONL) P3 0/900; 0 ± 0% Nrl P0 0/900; 0 ± 0% Nr2e3 (Pnr) E16.5 0/100; 0 ± 0% (24h) 0/900; 0 ± 0% (48h) 420/1200; 35 ± 2.5% E18.5 0/900; 0 ± 0% (24h) 0/900; 0 ± 0% P0 0/900; 0 ± 0% (24h) 0/900; 0 ± 0% (48h) 48/1200; 5.3 ± 1.2% (72h) 140/1200; 15.5 ± 0.9% P3 0/900; 0 ± 0% Isl1 (Islet1) P3 105/2040; 5.1 ± 0.9% (24h) 0/900; 0 ± 0% (total) 0/2100; 0 ± 0% 0/2100; 0 ± 0% 0/2100; 0 ± 0% 0/2100; 0 ± 0% (48h) 660/900; 76.6 ± 8.8% (INL) P5 0/300; 0 ± 0% Gnao1 (Goα) E12.5 0/900; 0 ± 0% E14.5 0/900; 0 ± 0% E18.5 0/900; 0 ± 0% P0 0/900; 0 ± 0% P3 0/900; 0 ± 0% (72h) 135/300; 45 ± 5% (INL) (6d) 250/300; 83.3 ± 2.8% (INL) P5 0/900; 0 ± 0% Cabp5 P3 (48h) 89/180; 49.4 ± 7.5% (INL) (72h) 135/240; 54.1 ± 5.9% (INL) P5 0/900; 0 ± 0% Prkca (PKCα) P3 n/a (5d) 680/900; 75.5 ± 8.3% (INL) P8 0/900; 0 ± 0%
132 A. Crx: E14.5
Ki67 Vsx2 BrdU (8h) BrdU (10h)
← Central Far Peripheral INBL INBL INBL NBL Ccnb1 ONBL
Far Peripheral ONBL
pH3 Central
B. Crx: E16.5+ 24h = E17.5
Crx/BrdU (24h) Crx BrdU (24h) INBL ONBL
Figure 3.8. Crx is expressed in a minute number of G0* RTCs at E14.5. A. A minute number of Crx+/Ki67+ cells were detected in the peripheral retina but not in the central retina (white arrowheads). Further, Crx (green) did not co-localize with any other cell marker (red) suggesting that it was expressed in G0* RTCs. Indeed, Crx+ cells became BrdU+ between 6-8h of chase (i.e., 2-4h after cell birth, white arrowheads). B. At E16.5 the first Crx+/BrdU+ cells were detected between 12-24h of BrdU-pulse chase (8-20h after cell birth, white arrowheads). Scale bars are 10μm.
133 A. P0 P0+24h = P1
Crx/Ki67 Crx/BrdU(24h)
NBL INBL
B. Crx/Vsx2 Crx/Ccnb1 Crx/pH3 P5 P8
Crx/Ki67
ONL
INBL
INL
IPL
Figure 3.9. Crx is expressed in a subset of G0* RTCs in the post-natal retina. A. A small number of Crx+/Ki67+ cells were detected throughout the retina. Crx (green) did not co-localized with any other cell marker (red) suggesting that it was expressed in G0* RTCs. Indeed, Crx+ cells became BrdU+ between 12-24h (8-20h after cell birth, white arrowheads). B. The intensity of the Crx signal in the ONL and INL decreased and became more diffuse and perinuclear after ~ P5. Arrowheads point to Crx+/Ki67+ cells in the INL. Scale bars are 10μm.
134 A. E18.5 P0 P3 E16.5 + 4d = P0
Rcvrn/Ki67 Rcvrn/BrdU
INBL INBL ONL ONL
B. C. P0 P3 E16.5+12h E16.5 + 48h = E18.5 E18.5 P0 P3
Nrl-GFP/Ki67 Nr2e3/BrdU Nr2e3/Ki67
INBL INBL INBL INBL INBL ONL ONL
Figure 3.10. Rcvrn, Nrl-GFP and Nr2e3 are expressed in post G0* RTCs. A. Rcvrn (green) was detected in the retina after E18.5 and never co-localized with Ki67 (red). Indeed, the first Rcvrn+ cells were BrdU+ between 3-4 days of chase (3-4d after birth, white arrowheads). B. Nrl-GFP (green) never co-localized with Ki67(red) in the post-natal retina. C. Nr2e3 (green) appeared in the retina after E16.5 and never co-localized with Ki67 (red). Nr2e3+ cells became BrdU+ between a 24-48h chase (1-2d after birth, whte arrowheads). Scale bar is 10μm.
135 A. Isl1: P5 Isl1: P3+48h=P5
Ki67 BrdU (2h) Ccnb1 pH3 Isl1/BrdU(48h)
Central Peripheral
INL
Amacrine cells ONL
GCL Amacrine cells INL
B. E12.5 E14.5 P0 P3 P5 P8 P3+3d=P6
Gnao1/Ki67 Gnao1/BrdU(3d)
ONL ONL ONL
INBL INBL INBL INBL
INL INL INL
GCL GCL GCL GCL
Figure 3.11. Isl1 and Gnao1 are expressed in post G0* RTCs. A. At P5, Isl1 (green) was expressed in the ganglion and amacrine cells in the GCL and in the amacrine cells in the INL. It was also expressed in the bipolar RTCs in the middle INL. Isl1 was detected in a minute number of Ki67+ cells. It did not however, co-localize with any other cell cycle marker suggesting it was not expressed in RPCs. Indeed, Isl1+ cells became BrdU+ between 24-48h of BrdU chase (1-2d after birth, white arrowheads). B. Gnao1 (green) was expressed in the GCL cells and some photoreceptors, and bipolar cells in the INL. Gnao1 did not co-localize with Ki67 at any time point. At P3, Gnao1+ cells became BrdU+ between 2-3d of BrdU chase (2-3d after birth, white arrowheads). Scale bars are 10μm.
136 A. E14.5 E16.5 P0 P3 P5 P3+48h=P5
Cabp5/Ki67 Cabp5/BrdU(48h)
ONL
ONL INBL
INL INL
B. P3+4d=P7 P3+5d=P8 P8
Prkca/BrdU(4d) Prkca/BrdU(5d) Prkca/Tfap2a
INL
Figure 3.12. Cabp5 and Prkca are expressed in post G0* RTCs. A. Cabp5 (green) is expressed in cones and after P5 also in bipolar cells in the INL. Isl1 did not co-localize with Ki67 (red) at any timepoint. RTCs that were BrdU-pulsed at P3 became BrdU+ between 24-48h of chase (1-2d after birth, white arrowheads). B. Prkca was detected in the INL starting at P7. Cells pulsed at P7 and P8 became Prkca+/BrdU+ between 4 and 5d of chase, indicating that Prkca is expressed by bipolar cells many days after cell birth, white arrowheads. In addition to bipolar cells Prkca is also expressed in a small number of amacrine cells in the basal INL as shown by the co-localization of Prkca and Tfap2a (red). Scale bar is 10μm.
137 Table 3.3. Analyses of markers that are expressed in amacrine neurons. The numbers represent percentage of cells positive for the given cell cycle marker at different time points (e.g., (Tfap2a2+, Ki67+) cells / Tfap2a+ cells. The orange and gray panels represent co-localization or the lack of co-localization of neuronal markers with proliferation markers, respectively.
Marker/Age KI67 BrdU CCNB1 pH3 CCND1 VSX2 Tcfap2a (Ap2α) E12.5 0/100; 0 ± 0% (12h) 0/100; 0 ± 0% E14.5 3/1200; 0.2 ± 0% (0.5, 2 ,8h) 0/1200; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (12h) 3/1200; 0.2 ± 0% (24h) 12/1200; 1 ± 0.2% (48h) 42/1200; 3.5 ± 0.2% E16.5 3/1200; 0.2 ± 0% (2, 8h) 0/300; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% 0/900; 0 ± 0% (12h) 3/1200; 0.2 ± 0% (24h) 51/1200; 4.2 ± 0.2% (48h) 63/1200; 5.2 ± 1.1% P0 0/900; 0 ± 0% Dcx (Doublecortin) E12.5 167/1200; 13.9 ± 1% (1,2,4h) 0/900; 0 ± 0% 0/1010; 0 ± 0% 0/1020; 0 ± 0% 0/1010; 0 ± 0% 0/1010; 0 ± 0% (6h) 3/1300; 0.25 ± 0% (12h) 19/1200; 1.5 ± 0.3% E14.5 41/1200; 3.4 ± 0.3% (1,2,4,6, 8h) 0/900; 0 ± 0% 0/1010; 0 ± 0% 0/1020; 0 ± 0% 0/1010; 0 ± 0% 0/1010; 0 ± 0% (12h) 5/1200; 0.4 ± 0.1% (24h) 8/1200; 0.6 ± 0.1% E16.5 5/1200; 0.3 ± 0.1% (1,2,4,8h) 0/900; 0 ± 0% 0/1010; 0 ± 0% 0/1020; 0 ± 0% 0/1010; 0 ± 0% 0/1010; 0 ± 0% (12h) 5/1500; 0.3 ± 0.1% (24h) 6/1500; 0.4 ± 0% P0 0/1200; 0 ± 0% Elavl2/3/4 (HuB/C/D) E12.5 3/1200; 0.2 ± 0% (1,2,4, 6h) 0/900; 0 ± 0% (8h) 4/1200; 0.3 ± 0.1% (12h) 7/1200; 0.5 ± 0.1% E14.5 3/1200; 0.2 ± 0% (1,2,4,6, 8h) 0/900; 0 ± 0% (12h) 8/1200; 0.6 ± 0% (24h) 13/1200; 1 ± 0.1% E16.5 0/1200; 0 ± 0% P0 0/1200; 0 ± 0% Calb1 E16.5 0/300; 0 ± 0% E18.5 0/300; 0 ± 0% P0 0/300; 0 ± 0% GABA E18.5 0/300; 0 ± 0% P0 0/900; 0 ± 0% P3 0/900; 0 ± 0%
138 A. E16.5 + 12h
TfaP2a/Ki67/BrdU TfaP2a Ki67 BrdU(12h) INBL ONBL
B. E12.5 E14.5 E14.5+12h E16.5
Dcx/Ki67 Dcx/BrdU(12h) Dcx/Ki67 INBL GCL/ONBL
Figure 3.13. Tfap2a and Dcx are expressed in a minute subset of G0* RTCs. A. Tfap2a (green) is an exclusive amacrine cell marker. In the E14.5 retina, a minute number of Tfap2a+ cells co-labelled with Ki67 (red, purple arrowhead). These cells were also labelled by BrdU (red) between an 8-12h chase, revealing that they appeared in RTCs 4-8h after cell birth. B. Dcx (green) labels migrating and mature ganglion, amacrine and horizontal cells. Some Dcx+ cells in the NBL co-label with Ki67 (red ,white arrowheads). An 8-12h BrdU chase in the E14.5 retina revealed that Dcx is detected in RTCs 4-8h after cell birth (white arrowheads). Scale bar is 10μm.
139 E14.5 + 12h E16.5
Elavl2/3/4/Ki67/BrdU Elavl2/3/4 Ki67 BrdU(12h) Elavl2/3/4/Ki67 INBL GCL/ONBL
Figure 3.14. Elav2/3/4 is expressed in a minute number of G0* RTCs. Elavl2/3/4 (green) detect ganglion, amacrine and horizontal cells. In the E14.5 retina, a minute number of Elavl2/3/4+ cells co-labelled with Ki67 (red). These cells were also labelled by BrdU (red) between an 8-12h chase, revealing that Elavl2/3/4 appeared in RTCs 4-8h after cell birth, purple arrowheads. At E16.5 Elavl2/3/4 not longer co-labelled with Ki67, suggesting that it appears in RTCs later than at E14.5. Scale bar is 10μm.
140 A. B. E12.5 E14.5 E16.5 E12.5 + 48h BrdU ( = E14.5) E16.5 E18.5
Mtap1b/Ki67 Mtap1b/BrdU Calb1/Ki67 INBL INBL GCL/ONBL
Uchl1/Ki67 Uchl1/BrdU GCL/ONBL
C. E18.5 P0 INBL GABA/Ki67 INBL GCL/ONBL GCL/ONBL
Figure 3.15. Mtap1b and Uchl1 are expressed in mature post-mitotic neurons. A. Mtap1b and Uchl1 (green) appeared in the retina after ~ E13 labelling ganglion, amacrine and horizontal cells. These markers never co-localized with Ki67 (red) and appeared in RTCs 24-48h after birth as revealed by a 24-48h BrdU chase (white arrowheads). The white dotted line marks the boundary between INBL and ONBL. B. Rare Calb1+ cells (green) appeared in the NBL at E16.5 and became more numerous with age. After E18.5 Calb1 signal could be detected in the basal INL labelling Ki67- amacrine cells and the GCL cells (white arrowheads). C. GABA (green) appeared in the Ki67- amacrine cells at E18.5 and onwards, never co-localizing with Ki67. Scale bar is 10μm.
141
Chapter Four: Rapid Cell Fate Switch in vivo Following an Initiating Oncogenic Event
The work in Chapter 4 was written by Marek Pacal and Dr. Rod Bremner as a manuscript in preparation.
142 4.1 Introduction
Human retinoblastoma is an ideal tumor to study cancer origin because it is one of few tumors where the initiating event, RB loss, is known. The mouse retina has extra protection but retinoblastoma can be initiated here too through loss of Rb and one of its
relatives, p107 or p130 . The first phenotypic effects of Rb or Rb plus p107 absence are
felt when neurons are generated. Thus, whereas normal differentiating cells exit the cell
cycle, neurons lacking Rb family members divide ectopically (Chen et al. 2004). These
abnormal events trigger extensive apoptosis, a valuable anti-cancer defense, and the
tumor-prone Rb;p107 null retina lacks ganglion, rod, cone and bipolar neurons. The two
other neuronal cell types, amacrine and horizontal cells, as well as Müller glia, are more
death-resistant, and seem to avoid tumorigenesis using Rb/p107 independent methods of
exiting the cell cycle. Rare cells escape (presumably by accumulating additional genetic
lesions (Dimaras et al. 2008)) and generate retinoblastoma (Chen et al. 2004). Mature
and early-stage retinoblastoma in Rb/p107 or Rb/p130 null retina has the characteristics of amacrine cells, arguing that tumors arose from these interneurons; indeed (Robanus-
Maandag et al. 1998; Chen et al. 2004; Dannenberg et al. 2004; MacPherson et al. 2004), the intrinsically high death resistance of Rb-deficient amacrine cells provides an attractive feature for a cancer cell-of-origin (Chen et al. 2004; Bremner et al. 2005).
Rather than arising directly from amacrine cells, however, it is also possible that the initiating or post-initiating transforming event causes a fate-switch, conferring amacrine-like properties on another cell type. This possibility, while unproven, is not far- fetched as there are many examples in the retina and beyond of single genes altering fate
(Hatakeyama et al. 2001; Lee and Pfaff 2001; Bertrand et al. 2002). Conceivably, the 143 resulting initiated cell could have hybrid properties that might confer advantages during cancer progression. Rb can influence differentiation, although rather than changing a program from one cell type to another, its loss mainly generate cells that, if they survive, retain their basic original characteristics, but may have abnormal shapes or lack or under- express a subset of differentiation markers(Khidr and Chen 2006; Skapek et al. 2006).
Nevertheless, there are examples of cell type changes in Rb null cells, including the switch from white to brown adipose (Hansen et al. 2004) and in worms a switch from somatic to germ cells (Wang et al. 2005). Very recently, Rb was shown to affect fate in mesenchymal progenitors, favoring adipogenesis over osteogenesis and, critically, this function affected the characteristics of p53 null tumors that arose from these cells (Calo
et al.). In the retina, Rb does not influence progenitors, but is required to ensure
differentiating neurons exit the cell cycle (Chen et al. 2004; MacPherson et al. 2004).
Whether it and/or its relatives might sway fate in these late-stage cells, that are
presumably much less plastic than progenitors, is entirely unclear.
Here, we utilized Rb floxed mice and retroviral Cre expression to generate on a
wild type or p107 null background and follow sporadic Rb or Rb/p107 null clones in the
developing retina, respectively. Our results invoke cell competition as a mechanism
through which tumorigenesis is constrained in the mammalian retina and, most strikingly,
reveal a fate-switch that generates an entirely new population of amacrine cells. These
data challenge current models of retinoblastoma initiation and demonstrate in vivo that
oncogenic events can alter fate even at the earliest stage of tumorigenesis and even in a
terminally differentiating cell. Finally, the effects of Rb on fate choice/differentiation are
144 usually assumed to be E2f-independent. We demonstrate that all the abnormalities in
Rb/p107 DKO clones are restored by the loss of E2f1.
4.2 Results
4.2.1 Sporadic Rb inactivation
Previously, we used the α-Cre transgene to delete Rb at E10 in all peripheral RPCs (Chen
et al. 2004). This blanket KO model cannot reveal the cell-autonomous effects of Rb
loss. To generate Rb KO or Rb/p107 DKO clones we injected a retrovirus (MXIE-Cre;
Fig 4.1A ) expressing Cre recombinase and enhanced green fluorescent protein (eGFP)
into the sub-retinal space of Rb loxP/loxP or Rb loxP/loxP ;p107 -/- mice at postnatal day 0 (P0),
generating sporadic Rb KO or Rb/p107 DKO RPCs surrounded by WT cells. Contra-
lateral eyes received control virus (MXIE). As an additional control, Rb loxP/+ ;p107 +/- retinas received Cre virus. On a ROSA26R indicator background, β-galactosidase +/GFP +
double-labeled clones confirmed Cre activity ( Fig 4.1B ). Control virus generated GFP + clones lacking β-gal ( Fig 4.1B ). To determine the onset of the retroviral cassette expression in the retinal cells, we followed GFP expression in the virus-injected P0 retinas. We harvested retinas at 2, 4, 8, 12 and 24h after injection and immunostained retinal sections with anti-GFP antibodies. No signal was observed after 2h, but 4h after the injection a few GFP+ cells were observed and the number of these cells approximately
doubled with each time point ( Fig 4.1C ). To determine if Cre was active in these GFP + cells we replaced immunofluorescent secondary antibody with an antibody conjugated to
HRP and detected GFP immunohistochemically, then used laser capture microscopy
145 (LCM) to isolate single HRP + cells from Rb loxP/loxP retinas transduced by either GFP- virus or Cre-virus, and performed PCR to detect recombination of the floxed Rb alleles
(Fig 4.1D,E ). Recombined Rb alleles were detected in 16/21 (76.2%) HRP + cells collected 4h after transduction and 32/36 (88.5%) HRP + cells collected 24h after
transduction ( Fig 4.1E ). Thus, recombination is rapid and efficient following retroviral
integration and expression.
4.2.2 DKO clones expand but contract precisely to wild type sizes
We quantified the effect of Rb loss at P0 on the size of GFP + clones in the mature (P21) retina. Rb blocks division in differentiating cells, so we expected an increase in Rb KO
clone size, and more so in Rb/p107 DKO clones. Unexpectedly, the average size and
even the size distribution of P21 Rb KO or Rb/p107 DKO clones was the same as WT
(Fig 4.2A, B ). In blanket models of Rb deletion a subset of ectopically dividing neurons are susceptible to apoptosis (Chen et al. 2004; MacPherson et al. 2004). Thus, we wondered if normal Rb KO and Rb/p107 DKO clone size at P21 might mean that ectopic division and apoptosis is perfectly balanced. First we determined if clone size was altered prior to P21. Relative to WT clones, the average Rb KO or Rb/p107 DKO clone
size was 1.5 or 2.8 times larger at P8, which dropped to no difference or 1.6-fold by P14,
respectively ( Fig 4.2A, C ). At P8, increased Rb KO or DKO clone size was due to more
4 and ≥5-cell clones and fewer 1 cell clones, and 2 cell DKO clones were also reduced
(Fig 4.2C ). Thus, Rb KO clones are larger initially but are pruned by P14. Rb/p107 DKO clones start out even larger, but are also normal by P21.
146 The above perturbations provide additional evidence that Cre retrovirus efficiently removes Rb . Retroviruses integrate into chromosomal DNA following nuclear breakdown at mitosis. One cell clones arise when viral DNA integrates into a chromosome in one of the two RTC daughters from an RPC undergoing final mitosis
(Turner and Cepko 1987). Fewer single cell Rb KO and Rb/p107 DKO clones at P8 proves that Cre has an effect even when the virus integrates as an RTC is born ( Fig
4.2C ).
To directly measure whether early increases in clone size was due to elevated division, mice injected with Cre at P0 received BrdU two hours before sacrifice at P3, P5, or P8, and BrdU +/GFP + cells quantified. Division slows in the post-natal retina and ceases at P8 in the central regions and P11 in the far periphery (Young 1985b; Alexiades and Cepko 1996), thus few BrdU +/GFP + WT cells were detected at P3 or P5 and virtually none were seen at P8 ( Fig 4.2D, E) . Ectopic division in Rb KO and Rb/p107 DKO cells was evident at P3 and by P5 and P8, 8.4% and 5.7% of Rb KO cells were BrdU +, while
26% and 34.2% of DKO cells were BrdU +, respectively ( Fig 4.2D, E ). TUNEL revealed
a parallel rise in apoptosis (compare Fig 4.2D and F ). Apoptosis returned to WT levels
by P14 or P21 in Rb KO or Rb/p107 DKO clones, respectively ( Fig 4.2F, G and data not
shown). Thus, ectopic division causes an early increase in Rb KO or DKO clone size
which apoptosis corrects.
“Cell competition” balances excessive proliferation or apoptosis of mutant clones
by death or division of normal neighbors (Diaz and Moreno 2005; Khare and Shaulsky
2006). Strikingly, we found no difference in BrdU+ or TUNEL + cells in a four-cell diameter around WT, Rb KO and Rb/p107 DKO GFP + P5 or P8 clones (Fig 4.3 and data
147 not shown). Thus, unlike classic cell competition, only mutant cell numbers are affected in Rb KO or Rb/p107 DKO retinal clones.
4.2.3 Ectopic division of specific differentiating cell types
To define which Rb KO or DKO cells divide ectopically we used markers characterized in
Chapter 2 and 3 . Calb2 and Tcap2a were used to label differentiating amacrine cells.
We scored rods both by position (i.e., ONL cells) and with the rod markers Nr2e3 and
Arr3. Further, we used Cabp5 and Prkca to label bipolar cells. Importantly, Tcap2a,
Nr2e3, Arr3, Cabp5 and Prkca never co-label with Ki67 or any other cell cycle marker in the P0 retina, while approximately 2% of post-mitotic (G0*) Calb2 + cells contain Ki67 at
P0 ( Chapter 3). We also scored differentiating Müller cells based on the expression of cyclin D3. The Müller marker Cyclin D3 was undetectable at P3 or P5, and exclusive to post-mitotic cells at P8 (data not shown). Glutathione synthase (GS) was also specific to post-mitotic Müller glia (data not shown).
Next, we quantified the proportion of each type of Cre-expressing (GFP +) RTC
(marker +) that was also in S-phase BrdU + or Ki67 + 8 days after virus delivery at P0.
Even though Ki67 is in dividing cells but also G0* cells, most of the markers never
colocalize with Ki67. RTCs in control infected clones were BrdU - and Ki67 - ( Fig 4.4,
4.5 ). In contrast, ectopically dividing Rb KO or DKO rod, bipolar amacrine and Müller
RTCs were detected ( Fig 4.4, 4.5 ). Loss of p107 together with Rb enhanced ectopic amacrine and rod RTC division further ( Fig 4.4, 4.5 ) while Rb KO or DKO bipolar RTCs displayed similar level of ectopic division. In summary, 8/8 antibodies that mark three
148 distinct differentiating RTCs revealed that Rb ensures differentiating RTCs exit the cell cycle.
4.2.4 Rb/p107 loss generates more amacrine-containing clones
Next, we tested if the cell type composition of pruned mutant clones was, like size, also preserved. The well-established criteria of cell position and morphology (Turner and
Cepko 1987; Livne-Bar et al. 2006) were used to examine cell-type composition of mutant clones at P21. After P0, RTCs are born that generate rod, bipolar, Müller and a few amacrine cells (Young 1985a). Cell-type proportions in control clones were identical to those in Rb loxP/+ ;p107 +/- retina transduced with MXIE-Cre virus ( Fig 4.6A ), thus, as in other studies (Chen et al. 2004; MacPherson et al. 2004; Zhang et al. 2004),
Cre alone did not affect retinal development and aberrations required the presence of a floxed Rb locus. Below we discuss effects on the four post-natal cell types in Rb KO or
Rb/p107 DKO clones.
Rod numbers were unaffected in any genotype (p < 0.05) ( Fig 4.6A ), which is remarkable since ~ half of Rb KO and all Rb/p107 DKO rods are missing in a blanket KO model (Chen et al. 2004). Unlike a prior study (Zhang et al. 2004), we did not observe stunted rods in Rb KO clones ( Fig 4.6Bb and Fig 4.7 ). Even Rb/p107 DKO rods had normal morphology ( Fig 4.6B ). Where truncated rods were detected, the remaining portion was always present in an adjacent section (Fig 4.7 ).
In Rb KO clones, Müller cell number, morphology and expression of the glial cell marker cyclin D3 was normal ( Fig 4.6A , Fig 4.8A and data not shown). However, in
DKO clones cells with normal glial morphology were reduced from 8.2% to 1.6%, which 149 was paralleled by an increase in INL cells with unusual morphology from 0.02% to 8.8%
(Fig 4.6A,B ). Some (but not all) of these abnormally shaped cells had cell bodies in the
center of the INL and stunted glial-like morphology ( Fig 4.6B ), which stained lightly for
or lacked cyclin D3, but were always glutamine synthase (GS) positive, confirming their
glial identity ( Fig 4.8A ). Thus, loss of both Rb and p107 impairs Müller glia
differentiation.
Bipolar cells dropped from 10.3% in WT clones to 4.4% in Rb KO clones, and
further to 1.4% in Rb/p107 DKO clones ( Fig 4.8A ). Bipolar cells were also severely
depleted in a blanket DKO model (Chen et al. 2004). In Rb/p107 DKO clones a few of
the INL cells with unusual morphology mentioned above had soma in the outermost INL
where bipolar cells reside ( Fig 4.6B ). They had fine, barely visible processes consistent
with bipolar morphology ( Fig 4.6B ) and PKC α staining confirmed their identify ( Fig
4.8B ).
Amacrine cells made up only 1.5% of WT clones and this was preserved in Rb KO
clones. Remarkably, however, this increased 7.2 fold to 10.8% in Rb/p107 DKO clones
(Fig 4.6A,B ). Their identity was confirmed with several markers, including Tfap2a and
γ-amino-butyric acid (GABA) (Fig 4.8C ). Finally, some of the unidentified INL cells
mentioned above were in the inner INL and had faint amacrine-like processes ( Fig 4.6B) .
Logically, an increase in the proportion of Rb/p107 null amacrine cells might
reflect expansion of pre-existing amacrine cells through ectopic division. This
mechanism would generate more amacrine cells per clone and, unexpectedly, the average
frequency of amacrine cells in WT, Rb KO or Rb/p107 DKO amacrine-containing clones
was always ~ 1 ( Fig 4.9A). The frequency was similar in amacrine-containing clones
150 irrespective of clone size ( Fig 4.9A). Rods increase in number as clone size expands, but exactly like amacrine cells Rb or Rb/p107 absence did not affect rod frequency per clone
(Fig 4.9B ). Thus, even though there are ectopically dividing amacrine cells in DKO clones ( Fig 4.4, 4.5 ) they are, like rods, pruned to normal proportions within each clone.
The alternative explanation for higher amacrine cell proportions is an excess of
amacrine containing clones. Indeed, the fraction of amacrine-containing clones was
elevated >7-fold ( Fig 4.9C ), matching the overall increase in amacrine cell proportions
(Fig 4.6A ). There are three scenarios that could explain this striking pattern, two of
which can be dismissed outright. First, amacrine containing DKO clones might survive
better than DKO clones lacking amacrine cells. The same virus preps were used for all
genotypes therefore the starting number of clones was similar, and the huge (> 7-fold)
decrease in DKO clones would have been obvious, yet was not observed (data not
shown). Second, Rb/p107 deletion could improve survival of amacrine cells that are normally deleted in WT or Rb KO clones. There are no reports of massive (>7-fold)
amacrine cell pruning between P0 and P21 in the WT retina, and we did not observe such
loss of amacrine cells WT or Rb KO clones. The remaining mechanism to explain extra amacrine-containing DKO clones is a fate-switch. Here, ectopically dividing cells that should become rod, bipolar and/or Müller glia instead, lacking Rb and p107, differentiate into amacrine neurons. One prediction of this model is that the probability of a switch would rise with clone size. Indeed this was the case; e.g. the chances of an amacrine cell being present in a two cell DKO clone was ~1 in 5, which rose to ~1 in 2 in DKO clones with ≥ four cells ( Fig 4.9D ). Overall, an Rb/p107 -deficient RTC has about a 1 in 10 chance of switching to an amacrine cell. The ectopic division and apoptosis in clones
151 makes it difficult to define whether the new amacrine cells arise from rod, bipolar and/or
Müller glia. Irrespective, this striking result shows that an oncogenic event can alter fate in the initiated cell, even before neoplastic transformation.
4.2.5 All clonal Rb/p107 defects, including fate change, are E2f1 -dependent
Next, we examined the mechanism underlying Rb/p107 DKO clone phenotypes.
Different E2fs drive ectopic division and death in distinct Rb KO tissues (Tsai et al. 1998;
Liu and Zacksenhaus 2000; Ziebold et al. 2001; Saavedra et al. 2002; Chen et al. 2007).
Effects of Rb on differentiation are usually ascribed to its ability to potentiate cell-
specific transcriptional activators (Skapek et al. 2006; Burkhart and Sage 2008). In most
cases E2fs are not investigated, yet when studied, the results have been striking (Fajas et
al. 2002; Chen et al. 2007; McClellan et al. 2007; Asp et al. 2009), thus we wondered if
even the differentiation defects and fate change detected in Rb/p107 DKO clones might
be E2f-dependent.
First, we examined E2f1 function in Rb KO clones by introducing Cre retrovirus into Rb f/f ;E2f1 -/- retina at P0. In line with a blanket Rb KO model (Chen et al. 2007), deleting E2f1 alone completely blocked ectopic division and apoptosis ( Fig 4.2d, e, f, g;
Fig 4.4 ). In agreement, Rb/E2f1 DKO clones displayed normal average sizes and size distribution at all times ( Fig 4.2 a, b, c; Fig 6 ). Next, we asked whether E2f1 also drives
the more pronounced ectopic division and death seen in Rb/p107 DKO clones. E2f1 and
p107 are only 2 Mb apart on mouse chromosome 2, but extensive backcrossing yielded a
p107 -;E2f1 - haplotype. Subsequently, Rb loxP/loxP ;p107 +/-;E2f1+/-, Rb loxP/loxP ;p107 -/-
;E2f1+/+ , and Rb loxP/loxP ;p107 -/-;E2f1-/- littermates were injected at P0 with either Cre or 152 control virus. Ectopic division and apoptosis at P3, P5 and P8 in the Rb/p107 DKO
RTCs was rescued in Rb/p107/E2f1 triple KO (TKO) RTCs ( Fig 4.2d, e, f, g; Fig 4.4 ), as were both average clone size and clone size distribution at all times ( Fig 4.2 a, b, c;
Fig 4.6). Thus, E2f1 drives ectopic division and death in both Rb KO and Rb/p107 DKO
RTCs.
Next, we turned to the differentiation defects. As noted earlier, we did not
observe rod differentiation defects in Rb KO or Rb/p107 DKO clones, but found novel
abnormalities in DKO Müller and bipolar cells ( Fig 4.6B ). Our E2f1 results provided a
unique opportunity to investigate whether these differentiation defects persist in the
absence of ectopic division and death. Remarkably, the morphology of Rb/E2f1 DKO or
even Rb/p107/E2f1 TKO Müller glia, bipolar cells, amacrine cells and rods was
indistinguishable from that of WT cells and the reduction in cyclin D3 in glia was also
reversed ( Fig 4.6B; Fig 4.8A ). Müller glia had long, complex processes terminating at
the outer and inner limiting membranes, bipolar cells had a downward and upward
process terminating in the OPL and IPL, respectively, amacrine cells had a downward
process that arborized in the IPL, and all rods had a downward process terminating in the
OPL and an upward process that sprouted inner and outer segments. Clearly, Rb and
p107 are essential to repress E2f1, but not to potentiate retinal-specific differentiation
factors. E2f1 could perturb Müller or bipolar cell differentiation indirectly by promoting
proliferation and apoptosis, and/or directly by altering expression of differentiation
regulators.
Next, we assessed the role of E2f1 in driving the formation of new amacrine cells.
Strikingly, all cell type proportions were normal in Rb/p107/E2f1 TKO clones, including
153 amacrine neurons ( Fig 4.6A ). Moreover, the proportion of amacrine-containing clones was also corrected ( Fig 4.9C ). Thus, in addition to driving more subtle differentiation defects in bipolar and Müller cells, E2f1 is also required for the more dramatic fate- switch that generates extra amacrine cells in Rb/p107 null clones.
4.3 Discussion
4.3.1 The Origin of Retinoblastoma: a Moving Target
The origin of retinoblastoma has long been debated, and virtually every retinal cell type has been suggested as a potential starting point for the disease (Dyer and Bremner 2005;
Pacal and Bremner 2006; Bremner 2009). The standard approach is to study marker expression in tumors and extrapolate backwards, but this strategy suffers from the caveat that accumulated oncogenic events may alter the cancer cell epigenome, transcriptome, and/or proteome, making it impossible to use endpoints to define origin (Gonzalez-
Fernandez et al. 1992; Nork et al. 1995; Sakata and Yanagi 2008; Xu et al. 2009a).
Whether these changes occur and exactly how early in the stepwise process towards malignant retinoblastoma is unclear. Here, using a sporadic deletion model to match the clonal evolution of natural cancers, we discovered a fate-switch at the earliest point possible, concurrent with the first phenotypic effects of oncogenic initiation, and well before transformation. Specifically, Rb/p107 deletion, the initiating neoplastic event in
mouse retina, generated a large (> 7-fold) increase in amacrine cells. These new cells
were not the result of ectopic division, since the number of amacrine cells per clone was
always ~1 irrespective of genotype, and instead were derived from other mutant cell
154 types, evidenced by the appearance of new amacrine-containing clones. Larger clones were more likely to contain a new amacrine cell, consistent with an increased probability of fate switch. This finding is particularly striking given that amacrine cells are the suspected cell-of-origin in multiple mouse models of retinoblastoma (Pacal and Bremner
2006). Emerging tumors already have features of these cell (Chen et al. 2004), but our new data mean that instead of or as well as RTCs originally destined for the amacrine fate, even very early stage tumors could arise from another type of differentiating RTC that switched to the amacrine lineage. Conversion might even generate hybrid initiated cells, providing unusual gene combinations that facilitate transformation and/or later stages of tumor progression. In this regard it is of interest that ~8% of Rb/p107 null
clones contained aberrant INL cells that did not exactly match standard amacrine, bipolar
or Müller cell morphology. The precise source of new amacrine cells (or aberrant INL
cells) in Rb/p107 null clones remains to be identified, but irrespective, our data demonstrate that an oncogenic signal can alter fate in the developing retina even at a precancerous stage of tumorigenesis. These data, and related findings in other tissues emphasize that the characteristics of transformed cells cannot be used reliably to define the origins of any cancer (Calo et al. 2010; Goldstein et al. 2010).
4.3.2 Deregulated E2f1, not tissue-specific factors, drives fate and differentiation defects
Deleting one activating E2f partially or wholly rescues ectopic division and apoptosis in
various Rb KO cell types (Tsai et al. 1998; Liu and Zacksenhaus 2000; Ziebold et al.
2001; Saavedra et al. 2002). We previously showed that E2f1 drives defects in Rb KO
155 retinal cells (Chen et al. 2007), however whether E2f1 drives defects also in any Rb;p107
DKO cell type, was unknown.
Here we showed, that clones containing Rb KO and Rb;p107 DKO cells expanded
within a week after Rb deletion due to ectopic division of differentiating neurons. While
the effect in Rb KO clones was marginal, DKO clones expanded up to ~3-fold the size of
WT clones. Strikingly, by P21 apoptosis corrected the DKO clone size to WT levels.
Ectopic division and death were rescued in Rb;p107; E2f1 TKO clones, revealing that
these phenotypes were driven by E2f1.
In addition to ectopic neuronal proliferation, we noted a fate switch leading to
increased amacrine cell population in Rb;p107 DKO clones. While, the number of rods
was not affected in DKO clones, the numbers of bipolar and Müller glia cells were much
reduced ( Fig 4.6 ). Moreover, bipolar and Müller cells which appeared to be specified
correctly displayed abnormal morphology; bipolar cells were missing both processes and
dendritic trees and Müller cells were always stunted ( Fig 4.6B ).
Rb binds >100 proteins (Morris and Dyson 2001) and might promote
differentiation by potentiating tissue-specific transcription factors that promote programs
such as myogenesis, adipogenesis and osteogenesis (Gu et al. 1993; Chen et al. 1996;
Thomas et al. 2001). Further, Rb controls cell fate in mesenchymal progenitors by
regulating gene expression of key differentiation factors (Calo et al. 2010). We expected
that Rb or p107 exploit this mechanism to affect retinal cell fate and differentiation.
Surprisingly, deleting E2f1 suppressed all defects in Rb/p107 DKO clones; the
proportions of cell types were restored and no differentiation defects were observed (Fig
4.6B; Fig 4.8A, B ). E2f1 might trigger defects either because it directly regulates genes
156 that influence fate and differentiation (Muller et al. 2001; Fajas et al. 2002; Dimova et al.
2003; Korenjak et al. 2004; Korenjak and Brehm 2005), and/or because of the secondary consequences of ectopic proliferation and/or death of neighbors. Irrespective, our data reveals that retina-specific factors do not require Rb/p107 to promote differentiation of rod, Müller, bipolar or amacrine cells, and that amacrine cell fate switch is driven by
E2f1.
Does Rb have any role in retinal development beyond repressing E2f1? In a parallel study using a blanket model of Rb deletion we found that E2f1 deletion rescues all defects bar one (Chen et al. 2007). A subset of amacrine cells exhibit defects in the level and transport of synaptic proteins. Remarkably, this defect was suppressed by deleting E2f3 (Chen et al. 2007). Together, our data raise the possibility that the only critical Rb activity in retinal development is to quench E2f family protein function in
RTCs.
It was proposed that Rb promotes rod photoreceptor differentiation (Zhang et al.
2004). We have shown previously that the rod differentiation and functional defects in the blanket Rb KO model can be restored by removing E2f1 (Chen et al. 2007).
However, here we observed only morphologically normal rods in Rb KO or even Rb/p107
deficient clones ( Fig 4.6B; Fig 4.7 ). Therefore, we showed using two different models
that photoreceptor specification and differentiation does not require Rb or p107 and any
defects can be rescued by the loss of E2f1.
Differentiation defects were reported in other Rb KO neurons (Jacks et al. 1992;
Lee et al. 1992; Lee et al. 1994; Zhang et al. 2004), but because Rb is critical for mitotic
exit (this work & (Slack et al. 1998; Ferguson and Slack 2001; Ferguson et al. 2002;
157 MacPherson et al. 2003; Chen et al. 2004; MacPherson et al. 2004)), it is unclear whether it has a direct differentiation activity in these contexts. One group argued that mutated
Rb proteins that do not bind E2f still stimulate differentiation (Sellers et al. 1998).
However, binding on chromatin was not assessed, and we find an excellent correlation between strength of E2f binding on chromatin with biological activity (T. Yu and R.B, unpublished). Thus, defects attributed to reduced activity of tissue-specific differentiation factors might actually reflect deregulated E2F activity. It will be important to define the specific combinations of E2fs that drive ectopic division and/or apoptosis in different Rb -deficient tissues, then reassess whether differentiation defects
persist when these problems are completely suppressed.
4.3.3 Cell competition in the retina
Rb or Rb and p107 loss perturbed cell cycle exit and survival to different degrees, thus we
expected WT, Rb KO and/or Rb/p107 DKO clones to exhibit distinct sizes.
Unexpectedly, clone size distribution was genotype-independent. Initially, mutant clones
had fewer 1 cell and more multi-cell clones due to RTC division, but apoptosis pruned
clones precisely to normal sizes. How do ectopically dividing clones made up of
different RTCs know their correct size?
In “cell competition”, clones of fitter cells (e.g. that proliferate faster or are larger)
may kill weaker neighbors to maintain normal compartment size ( Fig 5b ) (Abrams and
White 2004; Diaz and Moreno 2005). In the Drosophila wing imaginal disc, Myc-
expressing clones trigger the death of slower growing WT neighbors, (de la Cova et al.
2004; Moreno and Basler 2004). Here, we reported that mutant cells did not affect 158 division and survival of WT cells ( Fig 4.3 ). However, we currently do not know whether
WT cells curb the mutant cell proliferation and/or kill the mutant cells, thus bringing the
final clone size back to WT levels.
The Rb/p107 clone size seem to be corrected by “mutant cell competition,” i.e.,
competition of mutant cells for survival within a clone. Cell competition typically
involves genetically non-identical cells (mutant v normal), although competition between
genetically identical cells may be broader than previously appreciated (Khare and
Shaulsky 2006); Rb KO or Rb/p107 clones might provide a striking example. Mutant competition may be especially pertinent in neuronal tissues where there is a highly ordered radial structure to progenitor-derived clones (Turner and Cepko 1987).
The effect of Rb or Rb/p107 loss in mouse retina contrasts that of over-expressing
the Rb inhibitory cyclin D/CDK4 complex in Drosophila wing. In that case, mutant
clones grow larger without killing normal neighbors or pruning mutant kin (de la Cova et
al. 2004). Thus, fitter cells do not always induce cell competition, and when they do it
may affect either their neighbors, or only themselves ( Fig 4.5b).
Our loss-of-function study also contrasts with the effect of retroviral over- expression of E1A or activated Smoothened (Smo), which enlarge retinal clone size
(Zhang et al. 2004; Yu et al. 2006). Expressing E1A/Smo in RPCs likely drives extra division in both RPCs and daughter RTCs, whereas Rb/p107 loss-of-function specifically drives ectopic RTC division. Expanding RPCs may not trigger cell competition, thus increased clone size may be the result of more RPCs rather than RTCs. A direct comparison with our study would require expression of E1A/Smo specifically in RTCs.
It is also equally possible that E1A and Smo override cell competition. Indeed, several
159 studies have shown that oncogenic signals can engage pathways that facilitate clonal overgrowth and tumorigenesis (de la Cova et al. 2004; Moreno and Basler 2004; Tyler et al. 2007; Menendez et al. 2010).
E2f1 is central to the mechanism of retinal cell competition as it drives the apoptosis that pares back oversized clones. It remains to be shown how deletion and expansion of different RTCs is coordinated to generate WT-sized clones. Drosophila cell
competition can be mediated by secreted factors such as Decapentaplegic (Dpp) (Moreno
et al. 2002; Moreno and Basler 2004). If secreted signals regulate cell competition in the
retina, then their receptors and/or cognate signaling pathways must be distributed
differently among Rb KO and Rb/p107 DKO mutant cells versus unaffected normal
neighbors. Alternatively, signals might be conveyed within a clone of cells connected via
surface receptors and/or diffusible molecules that pass through gap junctions.
160
A D 5’ LTR MCS IRES GFP 3’ LTR 4 h 24 h MXIE Before LCM After LCM Before LCM After LCM MXIE-Cre Cre
B Z/AP reporter mice:
βgeo hAP CMV β-actin HRP
MXIE MXIE-Cre E
610 bp 240 bp
N2 N1 N2 N1 Cre N GFP AP GFP AP AP Exon19 1
212 19E 18 212 18 ROSA26R reporter mice: 283 bp 260 bp 670 bp Rosa 26 Stop LacZ
MXIE MXIE-Cre 670 ONL ONL INL INL
283 GCL GFP GCL GFP X-gal Merge 240
C cell cell +
Infect with Harvest retinas + virus Water Tail DNA cells Wt β -gal Water Tail DNA cells Wt β -gal
Primers: 19E + 18 212 + 18 P0 4h 8h 12h 24h N2 + N1 GFP/DAPI KO /Total cells 4rhs 4 h 8 h 12 h 24 h 670 (wt) 4rh 16/21 (76.2%)
24h 32/36 (88.5%) 240 (KO) 1 2 3 4 5 6 7 8 9 10 11 12
1,3,4,6-11: Rb KO cells 2,5: no PCR product ; 12: wt cell
Figure 4.1. A sporadic model of Rb deletion. A. Schematics of the MXIE retroviral vectors expressing either GFP alone (control vector), or Cre recombinase and GFP. B. Control or Cre virus was injected into newborn ROSA26R reporter mouse retina. GFP marks cells with virus and X-gal (blue) indicates Cre-mediated recombination and induction of the β- galactosidase reporter. C. To determine the onset of retroviral cassette activity, virally transduced retinas were immunostained for GFP. GFP signal was detected 4h after transduction. D. In some sections we stained for GFP using HRP immunohistochemistry. Single HRP+ cells were isolated using laser capture microscopy (LCM). E. DNA from tail (lanes 1 & 5), LCM-extracted WT (Rbf/f) cells (lanes 2 & 6), LCM-extracted Cre;Rbf/f cells (lanes 3 and 7) or water was amplified using primers 19E and 18 generating a 240 bp band in WT DNA (lanes 1-4), or primers 212 and 18 followed by nested PCR with primers N1 and N2 that amplify a 670 bp band in WT DNA, or a 240 bp band at the recombined Rb locus (lanes 5-8). The lower panel shows a series of 11 (lanes 1-11) Cre;Rbf/f cells and one WT cell (lane 12) collected 4h after the viral transduction; lanes 1,3,4,6-11 indicate Rb KO cells, 2,5 gave no PCR product. ONL: outer nuclear layer; OPL: outer plexiform layer; INL: inner nuclear layer; IPL: inner plexiform layer: GCL: ganglion cell layer. Scale bar in b and c 20µm. 161 A B E Clone Size Clone Size Distribution (P21) GFP/BrdU/DAPI 70 6 * 60 s 50 4 * * 40
# ONL % Cell 2 30
% Clones 20 0 10 P8 P14 P21 0 Control 1234≥5 INL Cells per Clone Control virus: Rbff C Clone Size Distribution (P8) GCL 60 E2F1-/- 50 Cre virus: s 40 Rb KO * 30 * Rb/p107 DKO ** 20 * * % Clone *
Rb/E2F1 DKO KO 10
Rb/p107/E2F1 TKO Rb 0 12345+ Cells per Clone D BrdU 50 40 * s 30 * 20 % Cell * 10 * * * 0 G DKO P3 P5 P8 GFP/TUNEL/DAPI Control DKO F TUNEL 5
4 * INL ONL
s 3
2 % Cell * * INL * TKO 1 * * 0 GCL P3 P5 P8 P14
Figure 4.2. Ectopic division increases mutant clone size, but apoptosis rapidly restores cell numbers to WT levels. Control or Cre virus was injected subretinally in P0 mice of the indicated genotypes and measurements made at the indicated times. A. Clone size at P8-P21. B. Clone size distribution at P21. C. Clone size distribution at P8. D, E. Two hours prior to sacrifice, mice received BrdU, and GFP+/BrdU+ cells were quantified at multiple times; quantification is shown in D and examples at P8 are shown in E. F, G. GFP+/TUNEL+ cells were also determined at multiple times. Quantification is shown in F, and an example at P8 in G. Asterisks in graphs indicate significant difference from Rbf/f mice injected with control virus (A: p < 0.0001; B: 1 cell: p<0.0001, 2 cells: p<0.001, 4 and 5+ cells: p < 0.0001; D: p<0.0001; E: P3: p<0.01, P5, P8 and P14: p<0.0001). Scale bars in A and G are 20 μm. 162 A C
BrdU GFP/TUNEL/DAPI 0.2 Cells
- Control Rb/p107 DKO 0.15 Z-stack Z-stack 0.1 ONL 0.05 0 P5 P8 % Wild-type GFP
B P5 INL TUNEL 0.8 Cells - 0.6 0.4 GCL 0.2 0 Section Section
% Wild-type GFP P5 P8 ONL Control virus: Rbff Cre virus: P8 Rb/p107 DKO INL
Figure 4.3. Normal levels of dividing and dying cells around Rb/p107 DKO clones. Control or Cre virus was injected into the subretinal space of mice of the indicated genotypes at P0 and tissue harvested at the indicated time points. Mice received BrdU 2 hours prior to sacrifice. Retinal sections were then triple labeled with DAPI, GFP antibodies and either (A) BrdU antibodies or (B) TUNEL. The percentage of dividing (BrdU+) or apoptotic (TUNEL+) uninfected normal (GFP- ) cells anywhere within four nuclei from the GFP+ clone was assessed. C. Typical P5 and P8 clones showing similar levels of TUNEL-labeling around WT or Rb/p107 DKO GFP+ clones (white arrows). Natural levels of bipolar cell apoptosis are highest at P8. In the P8 sections some of the GFP+ clone cells are also apoptotic (magenta arrows); the level is much higher in KO clones (see Fig 6e & 8a). Scale bar is 10 μm
163 A C BrdU 45 * DKO 40 Control Virus: GFP/BrdU/INL BrdU/ INL cell- * Rbff GFP/BrdU INL cell-marker 35 cell-marker marker Cre Virus: 30 e Rb KO 25 Rb/p107 DKO Rb/E2F1 DKO * PKC 20 * Rb/p107/E2F1 TKO α
% Cell Typ 15 * 10 * * * 5 * * * * * *
0 Calb2 PRC Bipolar Müller Amacrine P5 P8 P5 P8 P8 P3 P5 P8
B
DKO AP-2 GFP/Arrestin/
GFP/BrdU α Co-localized pixels
Z-stack Z-stack CyclinD3
ONL
INL GS
Figure 4.4. Rb or Rb/p107 loss triggers ectopic division of each type of RTC. A. Control or Cre virus was injected subretinally to mice of the indicated genotypes at P0. Mice received BrdU 2 hours prior to sacrifice at the indicated times. Retinal sections were quadruple labeled with DAPI and antibodies to GFP, BrdU or Ki67 and markers for specific RTCs (see text). A subset of criteria was used for quantification (A), and additional markers were used to confirm the identity of dividing cells (see B and C). For quantification, we identified photoreceptors (PRCs) by their location in the ONL and characteristic shape, as well as by arrestin or Nr2e3 staining, bipolar cells by Cabp5 or Prkca staining, Müller glia by cyclin D3 expression and amacrine cells by Cal2 or Tfap2a staining. Asterisks indicate values that differ significantly from control: PRC, bipolar and Müller: p < 0.0001, amacrine: P5: p<0.01, P8: p<0.001. B. Division of PRCs was confirmed using rod arrestin and Nr2e3. The examples show Rb/p107 GFP+/BrdU+ rods that are also rod arrestin+ or Rb/p107 GFP+/Ki67+ rods that are also Nr2e3+. C. Examples of dividing Rb/p107 DKO Prkca+ bipolar, GS+ and cyclin D3+ Müller and Calb2+ and Tfap2a+ amacrine cells, and confirmation that each of these cells divide using PKCa, GS, and AP-2a, respectively. In B and C, arrows indicate GFP+/BrdU+/marker+ triple-labeled cells. Scale bar is 10 μm. 164 CaCalb2 et CalretininCalb2 Tfap2aAP2 Tfap2aAP20 50 80 2.5 40 40 60 2 30 30 1. 5 40 20 20 1 20 10 0.5 10 Ki67(%) 0 0 0 Ki67(%) 0
Calb2/GFP(%) Wt RbKO Rb/p107 RbKO Rb/p107
Tfap2a/GFP(%) Wt RbKO Rb/p107 RbKO Rb/p107 DKO DKO DKO DKO Nr2e3 Nr2e3PNR Cabp5 Cabp5 15 0 40 30 60 30 10 0 20 40 20 50 10 10 20 Ki67(%) Ki67(%) 0 0 0 0 Nr2e3/GFP(%) Wt RbKO Rb/p107 RbKO Rb/p107 Cabp5/GFP(%) Wt RbKO Rb/p107 RbKO Rb/p107 DKO DKO DKO DKO
Merge GFP Calb2 Ki67 Merge GFP Tfap2a Ki67 Control Control RbKO RbKO Rb/p107DKO Rb/p107DKO
Merge GFP Nr2e3 Ki67 Merge GFP Cabp5 Ki67 Control Control RbKO RbKO Rb/p107DKO Rb/p107DKO
Figure 4.5. Additional markers confirm that Rb or Rb/p107 loss triggers ectopic division of each type of RTC. Control or Cre virus was injected subretinally to mice of the indicated genotypes at P0. P8 retinal sections were quadruple labeled with DAPI and antibodies to GFP, Ki67 and markers for specific RTCs. Here we used Nr2e3 to identify rods, Tcap2a and Calb2 to identify amacrine cells and Cabp5 to identify bipolar cells. The arrows indicate GFP+/BrdU+/marker+ triple-labeled cells. Scale bar is 10 μm. 165 A Cell type proportions (P21) 90
80 Control virus: Rbff 70 E2F1-/- 60 Cre virus: 50 Rbf/+;p107+/- p107 KO % Cells % 40 Rb KO 30 Rb/p107 DKO 20 Rb/E2F1 DKO * * Rb/p107/E2F1 TKO 10 * * * 0 PRC Bipolar Müller Amacrine Unidentified INL B Control Rb/p107E2F1 TKO
Rod Rod Rod ONL Rod Un
A M M INL BP BP A
GCL
Rb/p107
Rod Rod ONL
M A M Un INL BP A Un
GCL
Figure 4.6. Clone size, cell proportions and differentiation in P21 WT, RbKO, Rb/p107 DKO and Rb/p107/E2f1 TKO clones. Control or Cre virus was delivered subretinally to newborn mice of the indicated genotypes. A. The cell type proportions of GFP+ clones were assessed at P21. Asterisks indicate significant difference from control virus in Rbf/f mice (p < 0.0001). B. Examples of clones displaying rods, bipolar, amacrine and Müller glia cells. Rods had normal morphology in all genotypes. Müller glia (M) were stunted in Rb/p107 DKO clones; bipolar cells (BP) have a cell body in the upper INL and a descending process that arborizes in the IPL (arrows), but processes were truncated in Rb/p107 DKO clones (arrow shows cell body, bottom panel); amacrine cell (A) bodies are located in the inner INL, and many Rb/p107 DKO clones contained this cell type. Many Rb/p107 DKO clones also contained cells unidentifiable based on morpholy (UN). E2F1 deletion (TKO) rescued all these defects. Scale bar is 40 μm. 166 Control Rb KO
A D B C E ONL
INL Section 1
GCL
A B E C D Section 2
GFP/DAPI
Figure 4.7. The photoreceptors in RbKO clones display normal morphology. Cre virus was injected into the subretinal space of mice of the indicated genotypes at P0 and tissue harvested at P21. Retinal sections were labeled with DAPI and antibodies to GFP. No stunted or aberrantly shaped rods were observed in RbKO clones. Two adjacent sections (1 and 2) illustrate five photoreceptors (A-E) where one or more rod appears truncated in one section, but the remainder of the cell is present in the next section. This pattern was observed in all genotypes studied. Scale bar is 50 μm.
167 A C Z-stack Section Z-stack Section
GFP/DAPI D3 GFP/D3 GFP/DAPI Tfap2a GFP/Tfap2a Control KO
Rb Rb GFP/DAPI GABA GFP/GABA DKO DKO TKO
GFP/DAPI D3/GS GFP/D3/GS
Figure 4.8. Confirmation of INL cell type identity. Clones at P21 of the indicated genotypes, derived as in
DKO Figure 2, were stained with the markers shown, and confocal images gathered. Compressed Z-stack images are shown in some cases to reveal morphology of the complete cell. A. Cyclin D3 staining confirmed the B Z-stack Section identity of WT or RbKO Müller glia, which are also characterized by the position of their cell bodies in the GFP Prkca GFP/Prkca middle of the INL. Cyclin D3 was reduced or absent in abnormally shaped glia (only the cell body is shown here, for full length view of an oddly shaped glial cell see Fig 4.6B). Glutamate synthase (GS) staining confirmed the glial identity of these cells (bottom DKO panel set). E2f1 deletion suppressed both the
DKO morphological defect and the absence of cyclin D3 (Rb/p107/E2f1 TKO). B. PKCa-staining confirmed the identity of surviving Rb/p107 DKO bipolar cells in the outer INL that lacked processes. E2f1 deletion rescued morphology (Rb/p107/E2f1 TKO). C. Tfap2a and GABA staining confirmed the identity of Rb/p107 DKO amacrine cells in the inner INL. Arrows in A-C indicate GFP+ cells, arrowhead in C indicates the downward bipolar cell process. Scale bars in A and C are 20μm, in B 10 mm. TKO
168 A B
# Amacrine Cells per Amacrine Cell-containing Clones # Rods per Rod-containing Clones
2.5 8 2 6 1.5 4 1 0.5 2 per clone (#) clone per Amacrine cells cells Amacrine 0 (#) clone per Rods 0 All 2 Cell 3 Cell 4Cell ≥5 Cell All 2 Cell 3 Cell 4Cell ≥5 Cell Clones Clones Clones Clones Clones Clones Clones Clones Clones Clones
C D Fraction of clones that contain a specific cell type Proportion of Amacrine-containing clones 120 70 60 100 50 40 80 30 20 % All Clones % All 60 10 0
% All Clones % All 40 All Clones 1 Cell 2 Cell 3Cell ≥4 Cell Clones Clones Clones Clones 20 Control virus: 0 Rbff Rods Muller Bipolar Amacrine Unidentified INL Cre virus: Rb KO Rb/p107 DKO Rb/E2F1 DKO Rb/p107/E2F1 TKO
Figure 4.9. Rb/p107 deletion increases the production of amacrine containing clones. A, B. The average number of amacrine cells per amacrine containing clones (A) or rod cells per rod containing clone is plotted (B). C. The fraction of clones containing at least one of the indicated cell type is plotted. Note the >7-fold increase of amacrine-containing clones with the Rb/p107 DKO genotoype. D. The fraction of clones of the indicated size containing at least one amacrine cell is plotted. In A-D the virus delivered and the clone genotype is indicated using the colors shown in the key.
169
Chapter Five: Materials and Methods
170 5.1 Mice
Timed pregnat C57/BL/6 (The Jakson labs) mice were used for all experiments. Rb floxed mice and p107 null mice were genotyped as previusly described (Wu et al. 2001;
Chen et al. 2004). The noon of the day the vaginal plug was observed was considered E
0.5. All mice were treated in accordance with institutional and national guidelines.
5.2 BrdU labeling
Mice were injected I.P. with 10mg /ml BrdU in PBS (Roche) at 10mg/kg of body weight.
BrdU was detected using antigen retrieval protocol described bellow for immunostaing
(10mM sodium citrate (pH 6.0), boiled for 1h).
5.3 Immunostaining
Retinas were fixed in 4% PFA for 24h, dehydrated in 30% sucrose for 24h, frozen and cryo-sectioned to15 µm thickness on Superfrost plus slides (VWR), on a Leica cryostat.
Slides were dried for ~6h and re-hydrated and washed in PBST (PBS, 0.05% Tween).
Slides were immersed in 10mM sodium citrate (pH 6.0) in coplin jars and boiled for 30m.
Sections were cooled to room temperateure (~30m) and blocked in 4% serum (depending on the species of 2 o antibody) in PBS-T for 1hr at RT. Primary antibodies were diluted in
PBST and kept on the sections for 12h at 4 oC. Secondary antibodies (Alexa fluor,
Invitrogen) were diluted in PBST containing concentration of 200 ng/ml DAPI (Sigma) and kept on sections for 1h at room temperature. Slides washed in ample volumes of
171 PBST between primary and secondary antibodies. Slides were mounted with MOWIOL
(Calbiochem). Images were taken with Zeiss laser confocal system and images were handled with Image J (Sigma filter, stack manipulation) and Adobe Photoshop (image cropping, rotation, final overall intensity adjustment).
5.4 Retroviral constructs
Cre recombinase cDNA (L. van Parijs) was cloned into the BglII site of pMXIE vector
(D. van der Kooy, permission from G. Nolan). Retrovirus was produced using Phoenix- eco cell line as described (Livne-Bar et al. 2006). Cre recombinase activity was confirmed both in Z/AP (not shown) and ROSA26R mice, and by PCR detecting recombined Rb flox alleles (see bellow).
5.5 Morphological Identification of GFP + Cell Types
Cell-type quantification was performed on 20 µM sections at the optic nerve level stained for eGFP and DAPI. Only cells in which the nucleus was clearly visible were included.
Cell types were identified on the basis of their position and morphology. Photoreceptors have their cell body in the ONL, an upward process ending in inner and outer segments, and a downward process terminating in the OPL. Cells located in the upper half of the
INL with an upward process that ends in the OPL and a process descending to the IPL were identified as bipolar neurons. Müller cells have a cell body in the middle of the INL and long bidirectional processes that span the entire width of the retina. Cells in the lower half on the INL with a process descending to the IPL are amacrine neurons. 172 5.6 Retroviral Injections
Newborn mouse pups (P0) were anesthetized on ice. Cre-MXIE or control MXIE were
MMLV- based retroviruses prepared using phoenix eco packaging cell line (G. Nolan,
Stanford) as described previously (Livne-Bar et al. 2006). Retrovirus (2 l) was injected into the subretinal space through a small corneal incision. At P21, rats were killed by cervical dislocation, and the eyes were removed. The cornea was nicked to allow fixative penetration, and the eyes were immersed in 4% paraformaldehyde (pH 7.2) for 1 h at room temperature, then equilibrated in 30% sucrose, and frozen in embedding medium for cryosectioning. Sections (20 m) were used for immunostaining. At least three retinas
from three different litters were used for clonal analysis.
5.7 TUNEL staining
Immunostained sections were incubated for 1 h at 37 °C with 75 l of mixture solution consisting of 0.5 l of terminal deoxynucleotide transferase, 1 l of biotin-16-dUTP, 7.5