Over a century of research has focused on the development of RGCs and their role in image formation. Work over the last decade has revealed an intrinsically photosensitive subset of RGCs (ipRGCs) that signal irradiance information to brain regions that modulate processes such as sleep, circadian rhythms, and PLR [3, 4]. Here, we provide the first direct examination of ipRGC birth and axonal targeting in the brain. Our data reveal a broad diversity of developmental parameters within the ipRGC population, with the development of some ipRGCs diverging from the majority of other RGCs.
A subset of ipRGCs is born later than other RGCs
Previous studies have shown that RGCs are born between E11 and E18 [12, 21]. We find that most ipRGCs are born between E11 and E15, similar to other RGCs [11, 12, 21]. In stark contrast to the Brn3a-positive RGCs, we observe a significant amount of ipRGC neurogenesis after E15 (Figure 1C). Interestingly, Math5, a transcription factor involved in RGC fate determination, is downregulated at E16 . Thus, this late born ipRGC cohort may be specified independent of Math5. In agreement with this idea, some ipRGCs remain in the Math5 knockout . The early born ipRGCs may share an overlapping function with other RGCs, such as the ipRGCs that are capable of supporting pattern vision , and the later born ipRGCs may preferentially target non-image forming areas that are innervated later, such as the SCN and the shell of the OPN. Such a divergence could be determined by the identification of molecular markers for individual ipRGC subtypes.
The SCN is innervated later than image-forming retinal targets
During embryonic development, the majority of RGC axons project through the optic chiasm past the overlying SCN to target visual brain centers such as the LGN and the superior colliculus (Figure 2F) [15, 16, 24, 25]. In contrast, cholera toxin labeling of RGC fibers showed that retinal axons enter the SCN postnatally. The lateral edges of the SCN fill first according to CTB labeling and both the Opn4
;Z/AP reporters, which in turn closely match a previous report that exposure to bright light can induce cFos expression in the lateral edges of the SCN in newborn mice . The spatio-temporal agreement of ipRGC axonal labeling with the induction of cFos expression by light supports the well-established idea that most of the RGCs that innervate the SCN express melanopsin [4–6, 9, 26].
There are at least three possible explanations for this later innervation of the SCN by ipRGCs. First, ipRGC axons may follow a later time course compared to other RGCs. Such a delay in targeting would correspond with the later born ipRGCs we observe. Second, ipRGC axons that innervate the SCN may reach the chiasm earlier in development, but stall there until P0. Though we do observe ipRGC axons in the chiasm ventral to the SCN from E17, it remains to be determined if these specific ipRGC axons subsequently innervate the SCN (Figure 3). It is doubtful that innervation of the SCN is dependent upon functional maturation of target cells since the SCN begins oscillating before birth [27, 28]. Finally, the SCN could be innervated by collaterals from RGC axons that have already passed through the chiasm to more distal targets in the LGN, pretectum or colliculus [29, 30]. Indeed, such a delay in collateralization of RGC axons into the SCN has previously been reported . Since most RGC axons pass by the SCN to enter visual targets embryonically (Figure 2F), this temporal separation could be a means for preventing non-ipRGC axons from aberrantly terminating in the SCN. If this is the case, the ipRGCs that innervate the SCN may be uniquely receptive to a yet undetermined signal from the SCN. Expression of such factors may complement the spatial and temporal characteristics of ipRGC innervation of the SCN during development (Figure 3).
Onset of PLR correlates with emergence of ipRGC axons in the OPN shell
The PLR is mediated by retinal input to the OPN, which can be divided into core and shell regions. The shell is defined by parvalbumin and calbindin-D expression [31, 32], and is innervated by the M1 subset of ipRGCs, which label with Opn4
[6, 9]. This connection of M1 ipRGCs to the OPN shell is crucial to the PLR circuit, since genetic ablation of the Opn4
-labeled ipRGCs results in severe impairment of the pupillary light response . In agreement with these data, the Edinger-Westphal nucleus, which is the brain relay downstream of the OPN in the PLR circuit, is primarily innervated by axons from neurons in the shell of the OPN .
Our data suggest that two subtypes of ipRGCs innervate the OPN with different spatial and temporal profiles. CTB labeling of RGC axons and Opn4
;Z/AP labeling of all ipRGCs reveal robust innervation of the OPN from birth. In contrast, the Opn4
reporter only shows labeling of ipRGC axons in the OPN starting in the second postnatal week. This delayed labeling is limited to the shell region of the OPN, similar to adult labeling with the Opn4
reporter [5, 6]. Thus, it appears that the M1 ipRGCs innervate the OPN shell later than the ipRGCs that innervate the core. We observe a similar temporal separation between Opn4
;Z/AP-labeled ipRGCs that innervate the LGN and the Opn4
-labeled subset that innervate the intergeniculate leaflet (Figure 5). Thus, the Opn4
;Z/AP reveals that the ipRGCs that innervate classical visual targets such as the dLGN follow a developmental paradigm similar to other RGCs, while the Opn4
labeled ipRGCs appear to innervate their specific targets later.
To explore the functional output of the differing ipRGC projections to the OPN, we measured the PLR of postnatal mice. We first detected a rudimentary PLR at P7, coincident with the appearance of the Opn4
-labeled axons in the OPN shell. Although we cannot rule out the role that other relay centers play in determining the onset of PLR, it is striking that the onset of PLR coincides with the appearance of Opn4
-labeled axons in the OPN, a crucial connection for the PLR.
This early PLR must be driven by melanopsin since it accounts for all retinal photoreception through at least P10 [17, 33, 34] and synaptic connections between retinal layers are not functional until eye opening at P12 to P14 . While it remains unclear why PLR begins several days prior to eye opening, it agrees with mounting evidence that ipRGCs comprise the first functioning photoreceptive system during development [17, 34].