Notch pathway mutants do not equivalently perturb mouse embryonic retinal development

In the vertebrate eye, Notch ligands, receptors, and ternary complex components determine the destiny of retinal progenitor cells in part by regulating Hes effector gene activity. There are multiple paralogues for nearly every node in this pathway, which results in numerous instances of redundancy and compensation during development. To dissect such complexity at the earliest stages of eye development, we used seven germline or conditional mutant mice and two spatiotemporally distinct Cre drivers. We perturbed the Notch ternary complex and multiple Hes genes to understand if Notch regulates optic stalk/nerve head development; and to test intracellular pathway components for their Notch-dependent versus -independent roles during retinal ganglion cell and cone photoreceptor competence and fate acquisition. We confirmed that disrupting Notch signaling universally blocks progenitor cell growth, but delineated specific pathway components that can act independently, such as sustained Hes1 expression in the optic stalk/nerve head. In retinal progenitor cells, we found that among the genes tested, they do not uniformly suppress retinal ganglion cell or cone differentiation; which is not due differences in developmental timing. We discovered that shifts in the earliest cell fates correlate with expression changes for the early photoreceptor factor Otx2, but not with Atoh7, a factor required for retinal ganglion cell formation. During photoreceptor genesis we also better defined multiple and simultaneous activities for Rbpj and Hes1 and identify redundant activities that occur downstream of Notch. Given its unique roles at the retina-optic stalk boundary and cone photoreceptor genesis, our data suggest Hes1 as a hub where Notch-dependent and -independent inputs converge.


Introduction
The central eye field in vertebrate embryos is specified at the end of gastrulation and splits to form bilateral optic vesicles that evaginate from the ventral diencephalon.Multiple signaling pathways regionalize and pattern the growing optic vesicles, demarcating the optic stalk (OS), optic cup (OC) and retinal pigment epithelium (RPE) tissues.The OC gives rise to the neural retina, which is an excellent system for studying cell fate specification and differentiation.The retina is comprised of seven major cell classes that arise in a tightly controlled, but overlapping chronological order: retinal ganglion cells (RGCs), cone photoreceptors, horizontals, and a subset of amacrine neurons-before birth; and amacrines, rods, bipolars and Mu ¨ller gliamainly after birth.These cell types are derived from proliferative multipotent retinal progenitor cells (RPC) that permanently stop diving before differentiating into neurons and glia.Throughout development, RPC pool size must be balanced with neuron and glia production to generate a functional retina [reviewed in 1,2].
Most vertebrate Hes genes are Notch ternary complex targets [7,[35][36][37][38].Hes1, 3 and 5 are important in the nervous system, whereas Hes2, 4 and 7 act in other parts of the body [7,39].The role of Hes6 in development is debatable [reviewed in 7].Both Hes1 and Hes5 can exhibit oscillating expression patterns within stem cells or neural progenitors poised between proliferation and differentiation [39].For example, actively proliferating progenitor cells show high, oscillating Hes1 levels, whereas low Hes1 correlates with differentiation [40].In the mouse spinal cord, Hes5 can be either sustained or oscillatory, with its frequency of oscillation correlating with onset of differentiation [41].Hes1 is an essential gene, whose loss causes prenatal lethality along with embryonic morphogenesis defects characterized by premature differentiation [42].By comparison, complete loss of Hes3 and/or Hes5 has no impact on viability, but can induce discrete defects, suggesting specific contexts when these paralogues are compensated by or redundant with Hes1.This is further supported by the increased severity of Hes1;Hes3;Hes5 triple mutants in other parts of the central nervous system (CNS) [43][44][45][46][47][48].Despite the importance of the Notch pathway in retinal neurogenesis, no functions have been reported for it during mammalian optic vesicle/cup outgrowth, patterning or morphogenesis.Moreover, Hes gene redundancy and compensation have not been explored in the developing retina or adjacent tissues.In the E13.5 mouse eye, both Hes1 expression modes are present.RPCs oscillate while adjacent ONH/OS cells exhibit sustained Hes1 expression [49,50].As a Notch ternary complex target, removing Hes1 is predicted to universally release the block on neuron differentiation, but paradoxically Hes1 retinal mutants simultaneously have excess RGC neurons, but too few cone photoreceptors [14,42,50,51] (S1 Table ).This implies the Hes1 gene is where Notch-independent [52] and Notch-dependent regulation converge, with the latter complicated by Hes gene redundancy or compensation.
Our direct comparisons of Hes TKO versus Rbpj conditional mutants support that Hes genes regulate the balance between RPC growth and neurogenesis progression.We also discovered that Maml cofactor activities are not exclusive to the Notch ternary complex, in that ROSA dn- Maml-GFP/+ retinal mutants have unique nasal-temporal patterning defects.We determined that sustained Hes1 expression is Notch-independent, whereas in the retinal compartment, Hes1 and Hes5 are partially redundant downstream of Notch.Our phenotypic analyses of early neurogenesis reveal both Notch-dependent and -independent functions that influence RPC progression into early competence states, and further highlight directly opposing roles for Rbpj and Hes1 regarding cone fate.Although Hes TKO mutants partially rescue the Hes1 cone phenotype, they do not fully recapitulate those of Notch1 or Rbpj mutants [10,14,16,17].We conclude that unknown genetic inputs, independent from Notch signaling, also impact early neurogenesis and act via competence factors to affect RGC and cone photoreceptor fate determination.

Results
During mouse nervous system development, Hes1 appears in the anterior neural plate, optic vesicle and optic cup several days prior to the onset of retinal neurogenesis [42,59].We first compared the expression of multiple Hes genes throughout embryonic eye development (Fig 1).At these early stages, Hes1 mRNA and protein are uniformly expressed (Fig 1A).As the first cohort of retinal progenitor cells (RPCs) cells exit mitosis and differentiate into neurons, there is a switch in Hes1 expression to a "salt-n-pepper" pattern within mitotic RPCs (Fig 1D).However, optic nerve head (ONH) and optic stalk (OS) cells retain uniform Hes1 expression [50] (also see Fig 2A).By contrast, Hes5 mRNA appears later within RPCs just ahead of the first neurons [60].The mouse Hes5-GFP BAC transgene is an accurate reporter of Hes5 expression, enabling direct correlation with Hes1 and other markers during development [60].Hes5-GFP is also found in the diencephalon (Fig 1A -1C), but not in optic stalk cells that express Pax2 (Fig 1B).At E11, there are no Hes5-GFP+ cells in the nasal optic cup as marked by Pax2 and Foxg1 (Fig 1B and 1C).Hes3 functionally overlaps with Hes1 in the brain isthmus and is active in the CNS as early as E9.5 [61].Nonetheless, we did not detect Hes3 mRNA in the retina prior to E18 [62].We conclude that Hes1 is activated well before Hes5, which turns on in a subset of RPCs just prior to the onset of neurogenesis.Hes1 is expressed in distinct modes, appearing to oscillate in RPCs while exhibiting a high sustained level in the optic stalk.
The loss of multiple Hes genes is more catastrophic than loss of Hes1 alone in several regions of the embryo [reviewed in 7].We used the two Cre drivers with Hes TKO mice (Hes1 CKO/CKO ; Hes3 -/-;Hes5 -/-) to test this idea in the optic cup and stalk.We collected litters at E11, E13.5, E16.5, P0 (birth) (S2 Table ).Rax-Cre;Hes TKO mutants were not viable beyond E13, but displayed more severe phenotypes than Hes1 single mutants (S2 Table, Figs 3 and 4) [50].For the surviving Chx10-Cre;Hes TKO mice, we directly compared their P21 ocular phenotypes to Chx10-Cre;Hes1 CKO/CKO single mutants (S3 Fig) .Hes1 single mutants had defective retinal lamination, rosettes, and occasionally a small, vitreal cell mass (S3B Fig boxed area).By contrast, adult Chx10-Cre;Hes TKO eyes had more severe retinal lamination and rosetting defects and conspicuous microphthalmia (S3C and S3D Fig) .In some sections, ectopic tissue in the vitreous appeared contiguous with the ONH (S3C and S3D Fig boxed areas).We performed Tubb3/Endomucin (Emcn) colabeling of the ectopic tissue to assay for neurons and blood vessels, respectively (S3E-S3L' Fig) .Although blood vessels (Emcn+ cell membranes, pink arrows) and autofluorescent red blood cells (asterisks) were obvious, Tubb3+ neurons were difficult to observe, suggesting this ectopic tissue may have a nonneuronal origin.Overall, we observed that Hes TKO mutants are more severe than single Hes1 mutants or Hes3/5 double mutants.Our findings argue that Hes genes act in a complex, yet incompletely redundant fashion during eye development.To unravel this complexity, we initiated a deeper phenotypic evaluation at E13.5, when Rax-Cre triple mutants are viable and the ONH is fully formed.

Hes TKO and Rbpj mutants are the most severe
In theory, combined Hes functions should reflect those of the Notch ternary complex, which transcriptionally activates Hes genes.So we asked to what extent Hes TKO ocular mutants phenocopy the loss of ternary complex gene function.This also allowed us to bypass complexity at the receptor level, as three Notch receptors are expressed in the prenatal mouse eye [11,67].We opted to directly compare conditional mutant phenotypes for Rbpj and dnMAML (dominant allele that creates inactive Notch transcriptional complexes) to those for Hes TKO , using the same Cre drivers (Fig 2).In E11.5 Rax-Cre;Hes TKO eyes, RPC and ONH/OS cells are devoid of Hes1 protein and Hes5 mRNA as expected (Fig 2E).Because Chx10-Cre activates later and only within the retina [50], we expected there would be a loss of Hes1 from RPCs, but not ONH/OS cells.However, in E13.5 Chx10-Cre;Hes TKO eyes, Hes1 clearly persists in both domains (compare Fig 2A and 2G).Since Hes1 is spotty in the retina and dependent upon Cre mediated recombination, we hypothesized this its pattern is due to mosaic Chx10-Cre expression [58,68].This is further supported by immunostaining for Rbpj in Rax-Cre versus Chx10-Cre Rbpj CKO/CKO mutants (S4 .So, subsequent analyses were confined to E13.5, when Rax-Cre mutants are viable and Chx10-Cre mosaicism is less impactful.Next, we examined Hes1 and Hes5 expression in E13.5 Rax-Cre;ROSA dnMAML-GFP/+ and Chx10-Cre;ROSA dnMAML-GFP/+ retinas.Hes1 and Hes5 are only modestly reduced, with a stronger effect seen in the temporal retina (Fig 2M -2P).There was a stronger knockdown in the Rax-Cre;ROSA dnMAML-GFP/+ retinas (Fig 2M -2P).In neither case did we observe a loss of Hes1 in the ONH/OS area, further suggesting that it is independent of Notch signaling.The moderate phenotypes we observed did not fit our expectation that dnMAML misexpression would closely match the loss of Rbpj.Thus, we presume this dnMAML allele exhibits only a partial dominant negative effect in the developing eye.We decided to analyze this allele further to learn when, where and the degree to which it mimics Rbpj CKO/CKO and Hes TKO mutants.

Notch signaling has no impact optic cup patterning
The optic vesicle and cup are patterned along dorsal-ventral (D/V) and nasal-temporal (N/T) axes.Hes1 mutants have no D/V ocular phenotypes [50,59].We checked for mispatterning of the N/T axis, since the Pax2 domain is displaced in Rax-Cre; Hes1 CKO/CKO eyes, and Pax2 germline mutants have abnormal N/T ocular patterning [65].We compared the nasalrestricted marker Foxg1 [71,72]  , we conclude that this change in Foxg1 expression is another indication that the retina has expanded.The other exception is in E16.5 Rax-Cre;ROSA dnMAML-GFP/+ mutants.In this case, Foxg1 was mislocalized to the temporal retina and subretinal space (arrow in S5J Fig), a cell-free zone between the apical retina and RPE.We presume these displaced cells are RPCs, since some Notch pathway mutants lose the outer limiting membrane along the apical side of the optic cup, allowing cells to spill into the subretinal space [73,74].

Hes1 is Notch-independent at the optic cup-stalk boundary
At E12, the neural retina and optic stalk tissues become delineated, also establishing a ring of cells called the optic nerve head (ONH).ONH cells ultimately adopt glial fates and its interface with the retina is delineated by the generally abutting expression of the transcription factors Pax6 (RPCs) and Pax2 (ONH/OS) [66].Although the molecular mechanisms regulating this boundary are not well understood, its formation requires both Hes1 and Pax2 activities [50,65,66] The ONH and brain isthmus share multiple features, including Pax2 and sustained Hes1 expression [49,61,79], Brain isthmus cells have slower cell cycle dynamics than those in adjacent neural compartments with oscillatory expression [80].Cyclin D2 (Ccnd2) is expressed by brain glial cells and intermediate neural progenitors with slow cycling kinetics [81,82] and interestingly, E13.5 ONH cells normally express Ccnd2, which is regulated by Notch signaling in other ocular tissues [83,84].We observed that Ccnd2 is downregulated in Rax-Cre;Hes1 CKO/ CKO and Rax-Cre;Hes TKO mutants with mispositioned Pax2 domains (arrows in Fig 4B and   4C).Interestingly, Chx10-Cre;Hes TKO eyes also downregulate Ccnd2 expression.Because Hes1 encodes a transcriptional repressor, we presume its impact on Ccnd2 expression to be indirect.Once again, only Chx10-Cre;Hes TKO retinal cells ectopically expressed Pax2 (Figs 3O and 4D), consistent with ONH expansion in Pax2 GFP/GFP mutants [65].Without Pax2, retinal cells are unable to lock-in a neural development program expressing both RPC and ONH markers [65].This prompted us to ask whether Hes TKO and Pax2 mutants phenocopy one another regarding the mispatterning of the ONH/OS marker Vax1 [85][86][87](Fig 4E -4H).In Rax-Cre; Hes1 CKO/CKO and Rax-Cre;Hes TKO eyes Vax1 was shifted in the OS (arrows in Fig 4F -4G).But only in Chx10-Cre;Hes TKO eyes had a Vax1 domain that extended in the opposite direction, into the retina (Fig 4H).These data suggest that sustained Hes1 in the ONH helps lock-in the boundary with the retina, whereas multiple Hes genes in adjacent RPCs are necessary for maintaining neurogenic potential.

Notch signaling regulates both RPC growth and death
Throughout the CNS, Notch signaling stimulates progenitor cell growth and blocks neurogenesis.Reduced RPC proliferation is common to all mutants in this pathway, although the magnitude of this loss is variable (S1 Table ).We expected proliferation to be reduced in the six mutants and confirmed it by quantifying PhosphoHistone H3 (PH-H3) expression within G 2 and M-phase cells (Fig 5A -5G and 5O).Both Rbpj mutants have the fewest mitotic cells.There was also a modest loss of PH-H3+ cells in Hes TKO mutants for the Chx10-Cre driver, but not Rax-Cre.The opposite outcome was seen in ROSA dnMAML-GFP/+ mutants.Thus, all six mutants do not equivalently lose PH-H3+ cells, which might be due to slight differences in the degree and age of phenotypic onset between Cre mouse lines.
In the E13-E16 retina, Notch1, Rbpj and Hes1 mutants have a significant increase in apoptosis (S1 Table) [14,16,17,50].We used cPARP labeling to quantify dying cells among the six mutants to determine if they were equivalent (Fig 5H -5N and 5P).We observed the anticipated increase in cPARP+ cells in E13.5 Rax-Cre;Rbpj CKO/CKO mutants (Fig 5I -5P), but all other genotypes were unaffected (Fig 5P).This suggests that Rax-Cre;Hes TKO mutants can rescue the apoptosis phenotype previously described for Hes1 single mutants [50].This difference could be attributed to either Hes1 and Hes5 coordinated regulation of RPC target genes, or inherent interactions between retinal and ONH tissues, which impacts cell viability.
Consistent with other studies, we noted defective retinal patterning at E13.5, with rosettes containing Otx2+ or Crx+ cells residing near patches of Atoh7+ RPCs or Rbpms+ RGCs, respectively (Fig 6A -6N).Next, we quantified each nuclear marker and normalized using optic cup area (um 2 , see Methods and S5 Table ) for the different mutants.Surprisingly, the proportion of Atoh7 cells was largely normal, with significantly fewer cells in only Chx10-Cre; Rbpj CKO/CKO and Chx10-Cre;Hes TKO eyes (Fig 6V).This did not correlate with the changes seen for either Rbpms+ RGCs or Crx+ photoreceptors (Fig 6Y).Although excess Pou4f+ RGCs were reported at E16 in a previous Rbpj conditional mutant study [14], here at E13.In contrast to Atoh7, the proportion of Otx2+ cells dramatically increased in Rax-Cre; Rbpj CKO/CKO mutants, and significantly decreased in Hes TKO   Cre;Hes1 CKO/CKO mutants to facilitate direct comparison with both Cre-induced Hes TKO mutants (Fig 6Z).This subset of data is regraphed in Fig 6AA to more easily see the partial rescue for both Hes TKO mutants compared to single Hes1 mutants.There was a simultaneous and significant increase in RGCs for all four Hes TKO or dnMAML mutants (Fig 6Y).The largest increase in RGCs occurred in Rax-Cre;Hes TKO eyes with expanded retinal tissue (Fig 3).Finally, it was surprising that the defects noted for Rbpms or Crx expressing cells correlate with significant changes in the cells expressing Otx2, but negatively correlate to the Atoh7+ population (Fig 6V -6X).
Direct comparison of both qualitative and quantitative defects in RGC versus Crx+ cohorts among the six mutants revealed other allele-specific defects during early retinogenesis.We found that only Rax-Cre;Hes TKO mutants had displaced RGC and cone photoreceptor neurons in tissue that is normally optic stalk (Fig 6K).There were also mislocalized Rbpms+ RGCs in Chx-Cre;ROSA dnMAML-GFP/+ and Chx10-Cre; Hes TKO eyes, akin to interkinetic nuclear migration defects reported other Notch studies (arrow Fig 6M -6N) [74].At E16.5, we noted that only Rax-Cre;ROSA dnMAML-GFP/+ mutants contain more rosettes in the temporal retina (S6J Fig) , suggesting a Notch-independent interaction occurred during N/T patterning that becomes more obvious over time.
It remains unclear why Hes1 appears to promote cone genesis, rather than suppress it like other genes in the Notch pathway (Fig 6Z).One possibility is that Hes1 regulates some aspect of cone versus rod fate choice, since postnatal Hes1 -/-ex vivo retinal cultures were previously described to contain premature rod photoreceptor rosettes and fewer bipolar neurons [42].First, we verified that at E16.5 the ectopic Crx+ cells in rosettes are Thrb2+ cones (S6A-S6N Fig) and not precocious Nr2e3+ rods [96][97][98][99].Then we tested for premature rods within the Crx+ cohort.We collected E17 littermate control and Rax-Cre;Hes1 CKO/CKO retinal sections and colabeled for Crx and Nr2e3, a transcription factor specifically found in nascent rods [96].Nr2e3+ nuclei were evident within the forming outer nuclear layer (ONL) (S6O and S6P Fig) and retinal rosettes.However, the percentage of Nr2e3+Crx+ cells was identical (S6Q Fig) .Therefore, the loss of cones in Hes1 mutants cannot be attributed to accelerated rod genesis.Another explanation is that Hes1 provides temporal restriction to the Otx2 lineage to prevent prenatal bipolar neuron formation [100].Alternatively, RGC development may accelerate in the absence of Hes1, depleting the availability of transitional RPCs to activate Otx2 and adopt a cone fate.
Within the early Otx2 lineage, cells transiting to amacrine or horizontal fates downregulate Otx2 as they activate the transcription factor Ptf1a [reviewed in 104].Ptf1a is both necessary and sufficient for amacrine and horizontal fates, and when retinal cells lose this factor, they erroneously develop as RGCs and photoreceptors [105][106][107].Without Rbpj there was a total loss of Ptf1a+ cells (Fig 7H -7N).By contrast the other mutants had only a partial loss of Ptf1a + cells, likely reflecting a generally reduced pool of RPCs (Fig 7N).The more severe consequences of removing Rbpj on the amacrine pathway agree with previous studies (S1 Table ), and further reinforce that Ptf1a expression depends on Rbpj, similar to Ptf1a target genes [17,105,107].

Discussion
The molecular mechanisms integrating Notch with other signaling pathways remain poorly understood.Here we directly compared the genetic requirements for ternary complex components and multiple Hes genes during ONH formation and the onset of retinal neurogenesis (Fig 8A).We found that only Hes1 is required in the ONH.While all genes examined control RPC proliferation, our findings also point to particular Notch-independent activities.Although Hes1 and Hes5 transcriptional repressors have been compared using a variety of tools, their potential redundancy in the eye had not been tested [11,14,18,42,50,51,59,60]. Hes1 maintains optic vesicle and cup growth, the tempo of retinogenesis, and promotes astrocyte development in ONH/OS cells.But paralogues Hes3 and Hes5 have only subtle roles [18,108].In other areas of the CNS, Hes3 is active during oligodendrocyte maturation and interacts with STAT3-Ser and Wnt signaling pathways prior to the initiation of myelination [109,110].Given that Hes3 mRNA is undetectable in the embryonic retina, we propose it is relatively more important postnatally, possibly for retinal astrocyte migration, or optic nerve myelination.

Making and keeping the retinal-glial boundary
The boundary between the retina and OS possesses many characteristics of the brain isthmus, which is comprised of slowly proliferating cells that undergo little to no neurogenesis and act as a signaling hub for adjacent neural tissues [reviewed in 39,111].Consistent with this idea, we found Hes1 is required for Ccnd2 expression, which is associated with prolonged cell cycles.Both the ONH and isthmus require the transcription factors Hes1 and Pax2.In the eye, loss of either gene allows the retina to encroach and displace the ONH.This expansion might be due to a failure to effectively shift from fast to slow cycling kinetics or be driven by ectopic Hes5 plus other early eye factors.In this specific context, Rax-Cre;Hes TKO eyes were not much different than the loss of Hes1 alone.Thus, sustained Hes1 is likely sufficient for ONH formation and maintenance.We found no role for Notch regulation in ONH/OS formation since both Rbpj and ROSA dnMAML-GFP/+ animals retain a recognizable ONH with sustained Hes1 expression (Fig 8A).
Hes1 expression in the ONH/OS must be regulated by other genetic pathways.A strong candidate is the Shh pathway.Shh signaling performs an important feedback mechanism to control RGC population size.Nascent RGCs secrete Shh, which instructs RPCs to remain mitotically active via direct binding of Gli2 to activate Hes1 transcription [52,112].Moreover, at the optic vesicle stage of development, Shh diffuses from the ventral diencephalon midline to stimulate outgrowth of the optic cup and stalk [reviewed in 113].Given that Wnt, Bmp and Retinoic Acid signals also regulate proximoventral optic cup and stalk outgrowth and specification [reviewed in 114], it is tantalizing to speculate that they do so by converging on Hes1 expression and/or activity.It also remains unresolved if the ONH is a signaling hub for the adjacent retina.
To delineate ONH versus retinal phenotypes, we used both Rax-Cre and Chx10-Cre drivers to test for functional redundancy of Hes1 and Hes5 during retinal neurogenesis.Unfortunately, the Chx10-Cre line could only produce a few robust outcomes.This was unanticipated since the Chx10-Cre driver was successfully used in past retinal analyses of Dll1, Notch1, Hes1, Rbpj and Neurog2 function [10,[13][14][15][16]50,115].Due to mosaic expression and because fewer and fewer Cre-GFP+ cells are present as development progresses, we expect that selection pressure favored wild-type cells and their nonautonomous rescue of some of the phenotypes.Nonetheless, we uncovered distinct Hes TKO phenotypes using these Cre drivers at E13.5.Only Rax-Cre; Hes TKO mutants had a specific displacement of retinal tissue into the OS.The Pax6/Pax2 double positive cohort, along the retina-ONH boundary was largely missing, but the size of the mispositioned ONH was relatively normal (Fig 3).By contrast, Chx10-Cre;Hes TKO mutants had a bigger Pax2 domain, further confirmed by expansion of the Vax1 ONH marker into neural retinal territory.Our interpretation is that the earlier, broader Rax-Cre mutant prevented ONH/OS cells from adopting distinct identities, thus cells remained OC-like longer, producing more retinal tissue.This is likely a Hes1-specific process.But in Chx10-Cre mutants, with Cre expression restricted to the neural side of the boundary and acting at a slightly older age, the redundant, neurogenic role of Hes5 was revealed, since the retinal cells coexpressed neuronal and optic-stalk markers.Interestingly both phenotypes are apparent in Pax2 mutants, suggesting that Pax2 is upstream of Hes5, but acts parallel to Hes1.Future multiomic studies that characterize ONH cells, in the absence of Hes1 or Pax2, will be very informative.Finally our data highlight the variable penetrance and severity of Rax-Cre versus Chx10-Cre drivers, which is instructive for future studies.

Multiple modes regulating retinal histogenesis
Another important goal of this study was to understand how precisely Hes1 and Hes5 activities mirror the Notch ternary complex, which directly activates Hes gene transcription [reviewed in 4].Because there are multiple ligands and Notch receptors expressed in the developing retina (Fig 8B ), we focused on the requirements for Rbpj (Fig 8C ) and to a lesser extent Maml.There are three Mastermind-like paralogues (Maml genes), but germline mutant analyses failed to uncover individual gene functions during embryogenesis [reviewed in 116].Subsequently, a dominant negative isoform of MAML1 (dnMAML) was created, in which the MAML1 N-terminus forms ternary complexes with NICD and Rbpj, but cannot further interact with obligate transcriptional coactivators (e.g., p300, histone acetyltransferases) [54][55][56][57].This has been a powerful tool in cancer biology and immunology research [54], but during retinal neurogenesis, dnMAML is less effective at blocking Notch signaling.This might be attributed to differences in expression levels relative to other studies (in vivo Cre-mediated induction here, versus plasmid or viral delivery).However, several dnMAML eye defects, namely temporal retina-specific downregulation of Hes1 and Hes5, Foxg1 mislocalization and an unequal appearance of photoreceptor rosettes (Figs 2, S5 and S6J) suggest that Rax-Cre;ROSA dnMAML-GFP/+ mutants have Notch-independent genetic interactions.In vitro proteomic studies support this idea, where dnMAML can bind to Gli and Tcf/Lef proteins [117,118].This implies that ROSA dnMAML-GFP/+ retinal phenotypes may represent composite outcomes of simultaneously interfering with Notch, Shh, and/or Wnt signaling.
Rbpj also has Notch-independent functions (Fig 8C ), the most common being its role in co-repressor protein complexes to silence transcription via DNA methylation [reviewed in 4].Another activity is through Rbpj interactions with Ptf1a-E47 in a higher order PTF1 complex that has been studied in the pancreas, spinal cord and retina [104].In the pancreas, PTF1 complexes can activate Dll1, suggesting as a feedback loop from postmitotic to mitotic cells, via Dll1 binding to Notch1 [119].PTF1 can also directly antagonize Notch signaling in a cell autonomous and dose-dependent manner, since Ptf1a and NICD bind to the same site on the Rbpj protein [104,120].The second scenario is likely more relevant here.It is plausible that in the retina, when a critical threshold of Rbpj protein is bound up in PTF1 complexes, it not only impacts Rbpj availability for active Notch ternary complexes, but diverts cells from photoreceptor fate choice.We conclude that Rbpj activity regulates early photoreceptor development in at least two ways.First, in the Notch-dependent ternary complex, Rbpj controls RPC division versus differentiation into neurons like photoreceptors.Second, independent of Notch, Rbpj prevents cells normally destined to become amacrines from erroneously developing as photoreceptors via regulation of and independent physical interaction with Ptf1a (Fig 8B and 8C).
These additional Rbpj and Hes1 functions significantly complicate meaningful interpretation of our genetic data concerning Notch signaling regulation of Otx2.For Rbpj mutants, the expansion of Otx2+, Crx+, Prdm1+ cells, and cones, at the expense of Ptf1a and amacrine neurons, fits current models of mutual exclusion mentioned above [reviewed in 104].Conversely, Hes1 mutants produce excess RGCs and too few cones, which is essentially the opposite of Rbpj mutants.This might be attributed to Hes1 loss being relatively more efficient than Rbpj, facilitating RPC adoption of RGC fate, which also depletes the pool available for photoreceptor formation.Alternatively, Hes1 and Rbpj may simultaneously regulate (via distinct Notch-independent activities), competence or differentiation factors, for example Atoh7 [14,50,121].Here we found that Atoh7 protein expression is not correlative with RGC differentiation, in agreement with, single cell transcriptomics data [24].Instead, other competence factors, like Otx2, fluctuate as transitional RPCs adopt RGC or cone fates, with Otx2 expression becoming permanent in nascent and differentiated photoreceptors [33,34].Does this mean that the absence of Otx2 is needed for RGC fate?We propose that the Notch genes tested here, via different modes of action, act upstream of Otx2, to influence cell cycle status while also potentially targeting other genes that enable or limit RGC formation.
When considered together, our data and other studies, point to Hes1 as a signal integration point (Fig 8D).Hes1 mRNA and protein are dynamic, and likely important for the establishment of cellular heterogeneity.Hes1 might convey pulsatile feedback to other oscillating molecules like Dll1, Neurog2 or Ascl1 [40,41], which could occur upstream of Otx2.Although circumstantial, Prdm1+ cells and rods are specifically reduced in postnatal Neurog2 mutants, but how directly these events are linked remains to be determined [88,115].Future studies that apply short-lived Hes reporters and single cell imaging and sequencing modalities to remaining questions about when and where Notch signaling is required will be illuminating.

Ethics statement
All mice were housed and cared for in accordance with guidelines provided by the National Institutes of Health and the Association for Research in Vision and Ophthalmology, and conducted with approval and oversight from the UC Davis Institutional Animal Care and Use Committee (Protocols #20065 and #21839).

Histology and immunofluorescent labeling
P21 eyes were dissected and fixed in 4% paraformaldehyde/PBS overnight at 4˚C then processed through standard dehydration steps and paraffin embedding.Four micron sections were deparaffinized using Histoclear II (National Diagnostics HS200), hydrated through graded ethanol series and either stained with Hematoxylin and Eosin (H&E), or underwent antigen unmasking in hot (95˚C) 0.01M sodium citrate for 20 minutes, prior to immunofluorescent staining and imaging.For cryosection immunofluorescence, embryonic heads were fixed in 4% paraformaldehyde/PBS for 1 hour on ice, processed by stepwise sucrose/PBS incubations, and embedded in Tissue-Tek OCT.Ten micron frozen sections were labeled as in [78] with primary and secondary antibodies listed in S3 and S4 Tables.Nuclei were counterstained with DAPI.

Microscopy and statistical analysis
Histologic and in situ hybridization sections were imaged with a Zeiss Axio Imager M.2 microscope, color camera and Zen software (v2.6).Antibody-labeled cryosections were imaged using a Leica DM5500 microscope, equipped with a SPEII solid state laser scanning confocal and processed using Leica LASX (v.5) plus Navigator tiling subprogram, FIJI/Image J Software (NIH) and Adobe Photoshop (CS5) software programs.All images were equivalently adjusted for brightness, contrast, and pseudo-coloring.At least 3 biologic replicates per age and genotype were analyzed for every marker, and 1-2 sections per individual were quantified via cell counting and retinal tissue area measurements (S5 Table ).Sections were judged to be of equivalent depth by presence of or proximity to the optic nerve and/or characteristics of the adjacent forming lens.To normalize marker quantifications relative to tissue morphology changes, we calculated the square area (um 2 ) of retinas from E13 sections, using FIJI (NIH) to trace a polygon, excluding the opening for the optic nerve [125].The average number of marker+ cells were divided by the square micron area of the retina and graphed using Prism (GraphPadv9).For E17 retina, 11 tile scanned retinal sections for each of 3 biologic replicates/genotype were quantified, using the count tool in Adobe Photoshop CS5.Statistical analyses were performed on cells counts (S5

Fig 4 .
Fig 4. Hes1 and Hes TKO retina-ONH boundary phenotypes.(A-D") Pax2 and Ccnd2 immunolabeling at E13.5.Normally, Pax2 and Ccnd2 are coexpressed in ONH cells.In Rax-Cre;Hes1 CKO/CKO and Rax-Cre;Hes TKO eyes, the Pax2 OS domain is elongated, with Ccnd2 expression dramatically downregulated in the optic stalk or mislocalized into the RPE (arrows in A,A", B,B",C,C").Intriguingly, in Chx10-Cre;Hes TKO eyes, both Pax2 and Ccnd2 domains expanded into the optic cup (arrows in D, D").(E-H) Vax1 mRNA expression in the ONH/OS (arrows).Eyes in F-H are albino and the retina is outlined with dotted lines.The Vax1 domain shifted toward the brain in Rax-Cre;Hes1 CKO/CKO and Rax-Cre;Hes TKO eyes, but in Chx10-Cre;Hes TKO eyes it was expanded both into the retina and towards the brain.All panels oriented nasal up (noted in A) and the diencephalon to the right; n = 3 biologic replicates/genotype. https://doi.org/10.1371/journal.pgen.1010928.g004 Fig).Moreover, we observed that E13.5 Rax-Cre;Rbpj CKO/CKO mutants had a cell autonomous loss of Rbpj from RPC, ONH/OS, and RPE cells as expected (compare S4A, S4B and S4B' Fig).Although Hes1 was absent from the optic cup and RPE (compare S4A' and S4B" Fig), ONH/OS cells still express Hes1.Thus, we conclude sustained Hes1 expression in the ONH/OS is independent of Notch, whereas its expression in the retina depends upon Rbpj and Notch signaling.We took advantage of a Cre-GFP fusion protein within the Chx10-Cre driver to directly compare GFP and Rbpj coexpression in Chx10-Cre;Rbpj CKO/CKO and control Chx10-Cre; Rbpj CKO/+ retinal sections (S4C, S4E, S4G and S4I Fig).This Chx10-Cre BAC transgene encodes a Cre-GFP fusion protein, allowing us to test cell autonomy in the GFP+ cell population [58].At E13.5 we noted a strong autonomous knockdown of Rbpj protein (S4C" Fig vs S4E" Fig), yet at E16.5 there were more Rbpj-expressing retinal cells that lacked GFP, identifying them as wild type (compare S4G' Fig to S4I' Fig).Hes1 was partially autonomously downregulated at both ages, mirroring what was seen with Rbpj (S4D, S4F, S4H, S4J Fig).Thus, we concluded that Chx10-Cre phenotypes generated through E13.5 are informative, but beyond this stage the wild type cohort (GFP-neg) outcompetes mutant (GFP+) cells [69,70], providing ample levels of Notch signaling and partially rescuing development.Evaluation of Hes5 mRNA further confirmed Rax-Cre as the more effective driver, since we could still detect Hes5 in Chx10-Cre;Rbpj CKO/CKO retinas (compare Fig 2J to Fig 2L) among the six Rax-Cre or Chx10-Cre-induced mutants at E13.5 and E16.5 (S5 Fig).We noted normal Foxg1 retinal expression, with two exceptions.At E13.5 Rax-Cre;Hes TKO eyes, the Foxg1 nasal retinal domain was contiguous with the nasal optic stalk (S5D Fig).This is reminiscent of younger stages (Fig 1C), since at E13.5 Foxg1 in the wild type condition is no longer made in the nasal OS domain (S5A Fig).Based on RPC domain expansion into the optic stalk (Fig 1G, see below) . To understand whether Notch signaling controls formation of this boundary, we performed Pax6/Pax2 colabeling at E13.5 among all mutants (Fig 3I-3N).The Rax-Cre;Hes TKO eyes, were the most severe, with Pax6+ retinal tissue extending into the optic stalk territory, displacing the Pax2 domain (boxed area in Fig 3K).Although the Pax6-Pax2 boundary is intact in Rax-Cre;Rbpj CKO/CKO eyes, the shape of the ONH was attenuated compared to controls (Fig 3I).Interestingly, the proximal-most optic cup cells, those lacking Vsx2, still expressed Pax6 (compare Fig 3B to Fig 3I), suggesting these cells may have differentiated into neurons, since Pax6 is also expressed by nascent RGCs [69,78].The Rax-Cre;ROSA dnMAML-GFP/+ eyes were largely unaffected, but ONH shape was abnormal (Fig 3J).In all three Chx10-Cre generated mutants, a Pax6-Pax2 boundary was clearly discernable (Fig 3L-3N).But for Chx10-Cre; Hes TKO mutants, Pax2 was uniquely ectopic within the retinal territory (box in Fig 3N), demonstrating overlapping Hes gene function at this boundary (Fig 3D, 3K and 3N).We quantified the total number of Pax2+ cells per section (Fig 3O) and the small number of Pax6-Pax2 coexpressing "boundary" cells (Fig 3P).These data confirmed that although displaced, the Pax2-expressing ONH is of typical size in Rax-Cre;Hes TKO eyes (Fig 3K, 3K', 3K"-3O), and there was a significant loss of Pax6-Pax2 coexpressing boundary cells, most likely due to retinal extension (Pax6-only cell, Fig 3P).Moreover, only Chx10-Cre;Hes TKO mutant eyes had an expanded Pax2 domain (Fig 3N, 3N'-3N" and 3O), but with normal boundary cell composition (Fig 3P).
5 we found no difference in Rbpms+ RGCs, for either Rbpj mutant (Fig 6Y).It is plausible that ectopic RGCs in Rax-Cre;Rbpj CKO/CKO eyes might rapidly die (Fig 5P), and/or there is nonautonomous rescue in Chx10-Cre;Rbpj CKO/CKO eyes (S4E Fig).Alternatively, these RPCs may erroneously differentiate into neurons without fully committing to be an RGC, since there are obviously more Tubb3+ neurons in both Rbpj mutants, compared to control (Fig 6O-6U).Consistent with past studies of RGC genesis after blocking Notch signaling, we saw a significant increase in Rbpms+ cells for both sets of dnMAML and Hes TKO mutants (Fig 6Y).

Fig 8 .
Fig 8. Notch pathway activities and integration points with other genetic pathways in the embryonic eye.(A) Sustained Hes1 expression in the ONH/OS does not require ternary complex gene activities (Notch-independent).RPC status and early retinal fates are Notch-dependent, but also require other inputs.(B) Canonical Notch signal (blue) blocks premature RGC differentiation, while noncanonical Rbpj activity (yellow) utilizes Notch-independent modes to also regulate photoreceptor versus amacrine/horizontal fate (A/H).(C) The Rbpj protein forms distinct protein complexes (three shown) that uniquely regulate transcription and cell fates.By sequestering Rbpj into the different complexes, the production of one cell type also impacts its availability to regulate the other early cell types.(D) Distinct Hes1 transcriptional regulators influence oscillating Hes1 expression and activity during retinal neurogenesis.Since Hes1 encodes a repressor protein, its positive effect on cones is predicted to be indirect, presumably blocking and unknown factor X that normally suppresses cone genesis.https://doi.org/10.1371/journal.pgen.1010928.g008 Table) using Prism (GraphPad v9) or Excel (v16.16.11) software, with pvalues determined using one-way ANOVA and pair-wise Dunnett or pair-wise Whitney test or a Student's T-test.p-values less than 0.05 were considered statistically significant.