Teneurin-3 regulates the generation of non-image-forming visual circuitry and responsiveness to light in the suprachiasmatic nucleus

The initial processing of visual information occurs in the retina, which is a light sensitive central nervous system (CNS) tissue located at the back of the eye. Photons of light activate photoreceptors, which then pass visual information to retinal ganglion cells (RGCs) via bipolar cells. RGC axons then propagate visual information via the optic nerve to the brain, where higher order processing occurs to produce meaningful visual representations or behavior. Therefore, to ensure that the retina conveys visual perception and tracking, and luminance coding throughout life, it is critical that precise connections between >40 distinct RGC subtypes [1] and over 50 central brain targets [2] are formed during development.

RGC-central target connectivity is established in a stereotyped manner, with neural guidance cues and cell adhesion molecules regulating RGC axon innervation of retinorecipient centers. For example, cadherin-6 is required for innervation of retino-recipient targets such as the olivary pretectal nucleus (OPN) and the medial division of the posterior pretectal nucleus [3]. Further, semaphorin 6A is expressed in subsets of direction-selective RGCs and is required for recognizing plexin receptor ectodomains located in the medial terminal nucleus and for establishing connections critical for accessory optic system-mediated image stabilization on the retina [4]. Additionally, reelin, a secreted extracellular matrix protein, is required in the intergeniculate leaflet (IGL) and ventral lateral geniculate nucleus for normal targeting by intrinsically photosensitive retinal ganglion cells (ipRGCs) [5].

Though progress has been made in identifying the molecules required in some retino-recipient targets for RGC connectivity, very little is known about innervation of most central targets, including the suprachiasmatic nucleus (SCN). The SCN is in the ventral hypothalamus and is entrained to light via ipRGCs to regulate circadian rhythms throughout the body [6]. It is comprised of multiple, spatially segregated, GABAergic neuron subtypes that are largely defined by differences in neuropeptide expression [7]. The principal SCN neuron types are arginine vasopressin (AVP)-secreting neurons located in the shell, as well as gastrin releasing-peptide (GRP)-secreting and vasoactive intestinal peptide (VIP)-secreting neurons located in the core [8]. A recent study shows that in mutants lacking the receptors for AVP, faster entrainment to a jet-lag paradigm is observed [9], suggesting that AVP acts as a brake for fast changes in the phase of the clock in response to light.

It was long thought that VIP neurons were the principal cell type to receive direct retinal input [10]; however, recent evidence shows that AVP and GRP cells are also synaptically connected to ipRGCs [11]. Therefore, it is possible that a diverse set of molecules is required to properly establish ipRGC synaptic connectivity onto a heterogeneous population of retino-recipient cells that are critical for non-image-forming visual functions, such as circadian photoentrainment and pupil constriction.

Here, we consider the role of a teneurin family member, teneurin-3 (Tenm3), in regulating SCN innervation by ipRGCs, intra-SCN connectivity, and aspects of circadian photoentrainment. Teneurins are a family of single-pass, type II transmembrane proteins that are important for many aspects of neuronal wiring across species. In Drosophila, trans-synaptic homophilic interactions between teneurins establish wiring specificity in both the olfactory system [12] and at the neuromuscular junction [13]. Of the 4 mammalian teneurins (teneurins 1–4), Tenm3 homophilic interactions are critical for hippocampal CA1 neuron connectivity to the distal subiculum [14]. These precise limbic system connections are further regulated by Tenm3 repulsive interactions with latrophilin-2, which is expressed in the proximal subiculum [15]. These observations demonstrate the importance of teneurin homophilic and heterophilic interactions during neural circuit development.

Various roles for teneurins in mediating neuronal connectivity during vertebrate visual system development have been appreciated for some time [16]. Tenm3 knockdown in the developing zebrafish visual system leads to neurite lamination errors in both the inner plexiform layer and the tectum [17]. In mammals, Tenm3 homophilic interactions underlie accurate topographic mapping of retinal projections in both the lateral geniculate nucleus and superior colliculus (SC) [18,19]. As a result of these impairments, Tenm3^-/- mice perform poorly in a visual cliff test that assesses an animal’s depth perception [18]. Additionally, Tenm2^-/- mice have severely reduced ipsilateral input to both the dorsal lateral geniculate nucleus and SC [20], suggesting that in addition to Tenm3, Tenm2 is required for proper development of binocular vision. However, whether teneurins play similar roles in establishing neuronal connectivity in non-image-forming visual targets, like the SCN, is unknown.

To identify novel candidate molecules with the potential to regulate ipRGC-SCN connectivity, we conducted an in silico screen using the Allen Brain Atlas, a publicly available resource that includes in situ hybridization data at several stages of mouse brain development [21]. We began by searching for genes with high expression in the developing SCN that encode neural guidance or cell adhesion molecules. From this analysis, we observed robust expression of the teneurin family member Tenm3 in the postnatal SCN that persisted into adulthood. Given the requirement for Tenm3 in other aspects of mammalian visual system development and function [18,19], we chose to characterize Tenm3 expression in the SCN and to investigate roles for Tenm3 in regulating ipRGC-SCN connectivity and function.

Our in silico observations suggested that Tenm3 expression in the developing SCN is highest in the core region. We confirmed this by performing RNAscope in situ hybridization [22] on P5 C57Bl/6J brain sections with probes designed to detect Tenm3, Vip, and Avp, observing that Tenm3 expression is strongest in the SCN core region and is highly expressed in VIP cells (Fig 1A–1C and S1 Data). AVP cells that occupy the shell of the SCN also expressed Tenm3, but at much lower levels (Fig 1A–1C and S1 Data). Additionally, we found that a subset of non-AVP, non-VIP cells in the central SCN express Tenm3. We speculate that these cells are likely GRP cells, which are known to reside in this region. Taken together, we show that although multiple cell types in the postnatal SCN express Tenm3, it is expressed at much higher levels in VIP neurons.

Since Tenm3 homophilic interactions are important for axon targeting and topographic mapping in other regions of the developing visual system [18,19], we investigated this possibility for the ipRGC-SCN neural circuit. Using RNAscope in situ hybridization, we found that Tenm3 is very weakly expressed in the SCN-innervating Pou4f2-negative ipRGCs and is modestly expressed in Pou4f2-positive ipRGCs (S1A Fig and S1 Data). We confirmed our in situ expression data by performing transcriptomic profiling using single-cell RNA sequencing (scRNA-seq) on GFP-positive cells from Opn4^Cre/+; Brn3b^zDta/+; Rosa26^fsTRAP/+ retinas isolated by FACS at 2 early postnatal time points. Our scRNA-seq data reveals minimal Tenm3 expression in the Pou4f2-negative cluster of sequenced P1 and P5 ipRGCs (S1B–S1E Fig). In contrast, we found that expression of the latrophilins (Adgrl1, Adgrl2, and Adgrl3) is variable across ipRGC types. In the Pou4f2-negative cluster of sequenced P1 and P5 ipRGCs, almost all cells express Adgrl1, whereas expression of Adgrl2 and Adgrl3 is more heterogenous. These findings suggest that any role for Tenm3 in establishing ipRGC-SCN connectivity is likely to occur through heterophilic interactions with latrophilins, or with an as yet unidentified binding partner, rather than through homophilic Tenm3 associations.

Our observation that Tenm3 is highly expressed in the developing SCN led us to ask whether Tenm3 is required for M1 ipRGC innervation of the SCN. Therefore, we acquired Tenm3^+/- mice [18], generated Tenm3^-/-; Opn4^lacZ/+ mice, and quantified M1 ipRGC axon innervation of the SCN. We found that the ventro-medial (core) region of the SCN in Tenm3^-/-; Opn4^lacZ/+ mice shows reduced M1 ipRGC innervation (Fig 2A and 2B and S1 Data), whereas innervation of the ventro-lateral SCN is normal (Fig 2C and S1 Data). This suggests that Tenm3 is only required in specific cell populations of the SCN, most likely VIP neurons, for proper axon targeting. Importantly, we found that M1 ipRGC innervation of other central targets, such as the IGL and OPN shell, is normal (S2 Fig and S1 Data). As suggested by our expression data, conditional removal of Tenm3 from ipRGCs does not affect ipRGC innervation of the SCN (S3A–S3C Fig and S1 Data), supporting a role for SCN-derived Tenm3 in regulating ipRGC innervation. Taken together, Tenm3 is not required for central target innervation by all M1 ipRGCs but is essential for M1 ipRGC axon innervation of select cell types within the SCN.

One explanation for these innervation deficits is that in the absence of Tenm3, conventional RGCs lose their responsiveness to a repulsive factor and instead target the SCN, replacing ipRGCs. To test this, we injected a retrograde virus into the SCN of Tenm3^-/- mice and performed immunohistochemistry on retrogradely labeled retinas to identify ipRGCs. We quantified the percentage of retrogradely labeled melanopsin-positive and melanopsin-negative RGCs and found no difference between Tenm3^+/- and Tenm3^-/- mice (S4 Fig and S1 Data).

Since there is known synaptic connectivity among SCN subtypes, and since our observations suggest heterophilic interactions involving Tenm3 are required for normal ipRGC innervation of the SCN, we wondered whether synaptic connectivity between cells within the SCN (intra-SCN connectivity) is also perturbed in the absence of Tenm3. Since VIP cells express the highest levels of Tenm3 in the SCN, we assessed the consequences of conditional Tenm3 deletion from VIP cells on intra-SCN connectivity. We generated VIP^Cre/+; Tenm3^fl/-; ROSA^{Synaptophysin-tdTomato/+} mice to label sites of very close contact between VIP cells and other cell types within the SCN, an association that likely reflects synaptic connectivity (Fig 2D). We quantified VIP-AVP close contacts and observed an increase in their number in VIP^Cre/+; Tenm3^fl/-; ROSA^{Synaptophysin-tdTomato/+} mice compared to controls (Fig 2E and S1 Data). This result suggests that Tenm3 also plays a critical role in regulating connectivity within the SCN.

Our observations of altered ipRGC-SCN innervation and intra-SCN connectivity in Tenm3^-/- mice prompted us to ask whether there are functional differences in how light is processed within the SCN between wild-type and Tenm3^-/- mice. First, we investigated whether light-induced cellular activation is affected by the loss of Tenm3. We entrained Tenm3^+/- and Tenm3^-/- mice to a 12-h light-dark cycle. Once entrained, mice were released into darkness for 1 day and then presented with a 15-min light pulse at circadian time (CT) 22. We used the phosphorylation of histone H3 (pH3) as a readout since recent work shows that this rapid phosphorylation event in the SCN does not exhibit a circadian component and is only activated by light [23,24]. At first, we found no difference in pH3 immunoreactivity between Tenm3^+/-and Tenm3^-/- mice when they were presented with a light pulse of 100 lux (Fig 3A and 3C and S1 Data). However, we reasoned that this intensity could be saturating, and so we next presented mice with a light pulse of 10 lux and observed significantly increased pH3 immunoreactivity in the core and the shell of the SCN in Tenm3^-/- mice (Figs 3B, 3C, and S5 and S1 Data). Interestingly, pH3-positive cell density in Tenm3^-/- mice was elevated under both dim and saturating light intensities (Fig 3B and 3C and S1 Data), suggesting that the SCN has increased sensitivity to external light cues in the absence of Tenm3.

Increased SCN cell activation in response to light stimuli in Tenm3^-/- mice led us to hypothesize that the loss of Tenm3 may also affect how the SCN adjusts circadian rhythms in response to changes in the light environment. Therefore, we employed an artificial jet-lag wheel-running assay to assess whether Tenm3^-/- mice can properly re-entrain their endogenous circadian rhythms to the external light environment compared to heterozygous controls. Tenm3^-/- mice showed normal circadian photoentrainment (Fig 4A and S2 Data). However, when the light cycle was advanced by 6 h, Tenm3^-/- mice re-entrained to the new light pattern within an average of 2 days, whereas it took Tenm3^+/- mice approximately 4 days to re-entrain, similar to the normal re-entrainment speed of published wild-type animals [25,26] (Figs 4A, 4B, S6A, S6C, and S6E and S2 Data). When the light cycle was delayed by 6 h, we observed no differences in the time required to re-entrain between Tenm3^+/- and Tenm3^-/- mice (Figs 4A, 4B, S6F, and S6G and S2 Data). Furthermore, Tenm3^-/- animals exhibited a normal circadian period under constant dark conditions, indicating that the circadian clock is not disrupted in these mutants (S6H Fig and S2 Data).

We next reasoned that if loss of Tenm3 results in increased sensitivity to external light cues, then Tenm3^-/- mice would still be able to re-entrain to low light phase advances at an accelerated rate. To test this, we performed a wheel-running jet-lag assay under 10 lux, 2 orders of magnitude dimmer than the previous light cycle we tested. We observed that all mice were still able to entrain to the 10 lux light-dark cycle (Figs 4C and S7 and S3 Data). Additionally, we found that Tenm3^-/- mice re-entrain to phase advances significantly faster, requiring an average of 4.5 days compared to Tenm3^+/- mice, which took an average of 8 days to re-entrain to a 6-h phase advance under 10 lux (Fig 4D and S3 Data). Consistent with previous results, we observed no significant difference in re-entrainment to delays (Fig 4D and S3 Data).

Finally, we asked whether the responses to phase shifting in Tenm3^-/- mice are due to melanopsin-mediated phototransduction. We generated Tenm3^-/-; Opn4^Lacz/Lacz double knockout (DKO) mice and found that the combined loss of melanopsin and Tenm3 in the DKO mice does not result in slower re-entrainment to an advanced light pulse (S6D Fig and S2 Data), despite findings in Opn4^-/- mice [27,28] showing a trend toward slower re-entrainment to both advances and delays (S6B, S6E–S6G Fig and S2 Data). These data suggest that melanopsin is not required for the Tenm3-based regulation of SCN responses to phase advances.

Data presented in this study suggest that the cellular adhesion molecule Tenm3 plays an important role in SCN connectivity without disrupting the circadian clock and, specifically, influences the response to phase-advancing light stimuli. These results are the first to show that Tenm3 is critical for establishing M1 ipRGC axon innervation to select SCN cell types and for regulating VIP-AVP connectivity within the SCN. Loss of Tenm3 has profound effects on the sensitivity of the SCN since Tenm3^-/- mice exhibit increased apparent VIP-AVP synaptic connectivity and increased sensitivity to dim light stimuli. Thus, a unique behavioral response of Tenm3^-/- mice is resistance to “jet-lag” induced by phase-advancing, but not phase-delaying, light cues.

If Tenm3 is required for SCN development along with maintenance and/or function in the adult, why do Tenm3^-/- mice rapidly re-entrain to phase advances? Interestingly, removal of the AVP receptors V1a and V1b leads to rapid re-entrainment similar to what we observe in Tenm3^-/- mice [9]. In contrast to our observations in Tenm3^-/- mice, however, V1a^-/-; V1b^-/- mice rapidly re-entrain to both phase advances and delays. This can be mimicked through pharmacological blockade using V1a and V1b antagonists, suggesting that AVP-mediated inter-neuronal communication underlies the timing of re-entrainment observed in wild-type animals to a jet-lag paradigm [9]. If Tenm3 is only weakly expressed in AVP neurons, how then could it play a significant role in AVP signaling? Recent studies identify a high degree of synaptic connectivity not only among similar SCN cell types, but also between distinct neuronal populations in the SCN core and shell [29]. Notably, almost all VIP neurons form multiple synaptic contacts onto AVP cells [29]. Therefore, one hypothesis is that loss of Tenm3 in VIP cells alters VIP-AVP connectivity, consistent with observations in this study, thus influencing AVP signaling and leading to rapid re-entrainment following phase shifts.

AVP signaling is equally important for re-entrainment following both phase advances and delays [9]. Our unexpected observation that Tenm3^-/- mice rapidly re-entrain to phase advances, but not delays, raises the following question: How can the distinct behavioral responses to phase shifts we observe in Tenm3^-/- mice be explained at the molecular level? One possibility that is gaining recent support in the field is that there are multiple signals and/or circuits which mediate the phase advancing versus delaying effects of light on the circadian clock [30,31]. Consistent with this idea, only phase-advancing, but not delaying, light stimuli activate the SCN shell, where the AVP neurons reside [23]. This suggests that AVP cells found in the SCN shell are preferentially activated only in response to phase-advancing light and is consistent with activating more AVP neurons having the opposite effect as observed in V1a^-/-; V1b^-/- mice. Therefore, our prediction is that the increased synaptic inhibitory input from VIP onto AVP neurons we observe in Tenm3^-/- mice should lower AVP activation and thus mimic the phase-advancing effects of light observed in V1a^-/-; V1b^-/- mice. How re-entrainment to delays is enhanced in V1a^-/-; V1b^-/- mice is still not known but it appears to be independent of Tenm3.

VIP-AVP connectivity plays a critical role in regulating the sensitivity of the SCN to light. Under normal circumstances, the master circadian clock processes light information and responds with gradual changes, coordinating numerous tissues and hormones to work in synchrony [6]. It would be evolutionarily problematic for the SCN pacemaker to be hypersensitive to subtle changes in light signaling. Here, we demonstrate that Tenm3 is a necessary component of VIP-AVP signaling, maintaining SCN sensitivity to light which is critical for physiological interactions with the environmental light/dark cycle. Our data support the notion that Tenm3 is not exclusive to image-forming visual system circuit assembly and function, and we show here for the first time it also plays a key role in establishing neuronal connectivity in a non-image-forming visual brain region.