of 22
iScience
Article
Immediate responses to ambient light
in vivo
reveal distinct subpopulations of suprachiasmatic
VIP neurons
Anat Kahan, Karan
Mahe, Sayan
Dutta, Pegah
Kassraian,
Alexander Wang,
Viviana Gradinaru
anat.kahan@mail.huji.ac.il (A.K.)
viviana@caltech.edu (V.G.)
Highlights
Monosynaptic tracing from
SCN
VIP
neurons reveals two
subpopulations
Calcium changes show two
patterns of immediate
responses to light
Manipulating light intensity
and color support the
presence of two
subpopulations
The sensitivity to Opn4
antagonist hints to SCN
VIP
connectivity with M1 and
M2 ipRGCs
Kahan et al., iScience
26
,
107865
October 20, 2023
ª
2023 The
Authors.
https://doi.org/10.1016/
j.isci.2023.107865
ll
OPENACCESS
iScience
Article
Immediate responses to ambient light
in vivo
reveal distinct subpopulations
of suprachiasmatic VIP neurons
Anat Kahan,
1
,
3
,
4
,
*
Karan Mahe,
1
Sayan Dutta,
1
Pegah Kassraian,
1
,
2
Alexander Wang,
1
and Viviana Gradinaru
1
,
*
SUMMARY
The circadian rhythm pacemaker, the suprachiasmatic nucleus (SCN), mediates light entrainment via vaso-
active intestinal peptide (VIP) neurons (SCN
VIP
). Yet, how these neurons uniquely respond and connect to
intrinsically photosensitive retinal ganglion cells (ipRGCs) expressing melanopsin (Opn4) has not been
determined functionally in freely behaving animals. To address this, we first used monosynaptic tracing
from SCN
VIP
neurons in mice and identified two SCN
VIP
subpopulations. Second, we recorded calcium
changes in response to ambient light, at both bulk and single-cell levels, and found two unique activity
patterns in response to high- and low-intensity blue light. The activity patterns of both subpopulations
could be manipulated by application of an Opn4 antagonist. These results suggest that the two SCN
VIP
subpopulations connect to two types of Opn4-expressing ipRGCs, likely M1 and M2, but only one is
responsive to red light. These findings have important implications for our basic understanding of non–im-
age-forming circadian light processing.
INTRODUCTION
The hypothalamic suprachiasmatic nucleus (SCN) is a key structure in non–image-forming (NIF) light processing and circadian behavior.
A specific cell population, SCN vasoactive intestinal peptide (VIP)-expressing neurons (SCN
VIP
), has been shown to be a primary contributor
to SCN light response.
1
,
2
Neuronal activity measurements using a calcium-sensitive fluorescent sensor (GCaMP) in a Cre-dependent manner
using
in vivo
bulk recording, known as fiber photometry (FP), show that SCN
VIP
neurons are active during the day
3
and essential for circadian
regulation, in addition to mediating the rapid response to light solely during the dark phase, as measured by FP or a neuronal activity marker
such as c-Fos.
4
,
5
Recent studies suggest that SCN
VIP
neurons can be divided, either genetically or functionally, into two subpopulations. Based on single-
cell mRNA sequencing data, SCN
VIP
neurons can be genetically separated further into two major neuropeptidergic subpopulations, Vip+/
Grp+ and Vip+/Nms+, with a relatively low percentage of rhythmic genes in the Vip+/Grp+ subgroup compared to the Vip+/Nms+ sub-
group.
6
,
7
Functionally, it was shown that SCN
VIP
neurons could be separated into two subpopulations based on patterns of activity, either
tonic or phasic.
8
,
9
It is unknown, however, whether the two cell populations differ functionally in light information processing.
The primary sensory mediators conveying information about light to the SCN are the retinal opsins. The mouse retina contains two types of
photoreceptive cells: rods and cones. Rods express the photopigment Rhodopsin. Mouse cones express M- and S-opsins, which mediate
color vision, but lack the homolog of human L-opsin. Both rods and cones project indirectly to retinal ganglion cells (RGCs) via a complex
circuit involving bipolar, horizontal, and amacrine cells.
10
About twenty years ago, it was discovered that a small subpopulation (1–3%) of
RGCs are intrinsically photosensitive, therefore termed ipRGCs. The ipRGCs express the Opn4 gene, encoding melanopsin,
11
which has a
unique role as a photo-inducer of circadian phase shifting in response to blue light
12
and is required for NIF responses.
13
IpRGCs form
the retinohypothalamic tract and project directly to the SCN mainly through the M1 subpopulation,
14
,
15
and to other brain nuclei involved
in circadian photo-entrainment.
16–21
Together, opsin-expressing retinal cells are responsible for the photo-entrainment of circadian rhythms,
mainly through ipRGC projections to the SCN.
However, over the years, it has become clear that ipRGCs cannot be the sole retinal input to SCN activity and circadian behavior. First, it
was demonstrated by Gu
̈
ler et al. that ipRGCs transmit photic input transduced by rods and cones.
22
Second, melanopsin is highly sen-
sitive to blue light, with an absorption peak at 480 nm, but it was found that red light (>600 nm) affects circadian rhythm and sleep-wake
behavior
23
,
24
with various effects on behavior and physiology in mice and rats, including altered sleep architecture and locomotor
1
Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA 91125, USA
2
Present address: Department of Neuroscience, The Kavli Institute for Brain Science, Mortimer B. Zuckerman Mind Brain Behavior Institute, Columbia
University, New York,
NY, USA
3
Present address: Department of Animal Sciences, The Faculty of Agriculture, Food and Environment, The Hebrew University, Rehovot, Israel
4
Lead contact
*Correspondence:
anat.kahan@mail.huji.ac.il
(A.K.),
viviana@caltech.edu
(V.G.)
https://doi.org/10.1016/j.isci.2023.107865
iScience
26
, 107865, October 20, 2023
ª
2023 The Authors.
This is an open access article under the CC BY-NC-ND license (
http://creativecommons.org/licenses/by-nc-nd/4.0/
).
1
ll
OPENACCESS
Figure 1. Monosynaptic tracing from SCN
VIP
neurons to the retina supports the existence of two subpopulations
(A) Experimental design: VIP-Cre mice were injected with AAV1 expressing Cre-dependent split-VTA with eGFP (exp, n = 3) or AAV9 expressing Cre-depen
dent
eGFP only (ctrl, n = 3), followed by EnvA-
D
G Rabies expressing mCherry.
(B) Example of expression of the two viruses in the SCN. VIP neurons express eGFP (green), indicating split-VTA expression, and monosynaptic neurons
express
mCherry (white).
(C) Example of a flattened retina, stained with Opn4-ab (red) and expressing Rabies-mCherry (white).
(D) Percentages of cells in the retina which were (left) mCherry+ and labeled by Opn4-ab (black), and (right) labeled by Opn4-ab and mCherry+ (red, ‘bo
xplot’
presentation, MATLAB).
ll
OPENACCESS
2
iScience
26
, 107865, October 20, 2023
iScienc
e
Article
activity.
23
,
25
Third, in mice expressing the human red cone L-opsin, it was shown that SCN responses are sensitive to ambient light con-
ditions.
26
In addition, using circadian and neuronal activity markers, such as Per2 and c-Fos, SCN neuronal activity was shown to be sen-
sitive to a relatively wide spectrum in mice
27
,
28
and hamsters.
29
Lastly, in freely moving mice expressing only cones, SCN electrophysiology
showed that UV light is more effective than green light for photo-entrainment and that the SCN exhibits slower decay times for UV light.
30
Together, this evidence emphasizes that color and intensity perception in the SCN can be indicative of retinal photoreceptors’ sensitivity,
beyond ipRGCs.
To understand this connectivity of SCN
VIP
neurons, and the functional separation of subpopulations, we explored light-responsive inputs
to the SCN. First, we used monosynaptic tracing to identify two types of connectivity to SCN
VIP
neurons. We then used bulk calcium imaging
(fiber photometry, FP) and one-photon single-cell mini-endoscopic imaging of GCaMP6s activity to test SCN
VIP
responsiveness to different
light spectral and intensity conditions
in vivo
. We identified two subpopulations of SCN
VIP
neurons with distinct response profiles. Finally, by
pharmacological manipulation of Opn4 we confirmed that both SCN
VIP
subpopulations are connected to ipRGCs.
RESULTS
Monosynaptic tracing from SCN
VIP
neurons to the retina suggests two subpopulations with distinct connectivity
In order to address the functional properties of SCN
VIP
neurons in response to light, we first aimed to learn their unique connectivity to the
retina. We used an experimental design that allowed us to follow monosynaptic inputs to SCN
VIP
neurons; i.e., direct injection of an adeno-
associated virus (AAV) to deliver a split version of the EnvA receptor protein TVA (experimental group, Exp), or eGFP (control, Ctrl), followed
by EnvA-
D
G Rabies virus injection to deliver mCherry to the SCN of VIP-Cre mice (
Figure 1
A). This approach labels SCN
VIP
with eGFP, and, in
the experimental group, monosynaptically connected neurons with mCherry. Verifying the expression of the starting population in the SCN
(
Figures 1
B,
S1
A and S1B), we observed brain-wide expression, as was reported recently,
6
including in the paraventricular nucleus of the
thalamus (PVT) and the medial preoptic area (MPA) (
Figure S1
C–S1E). We identified 187
G
4 (mean
G
SEM) Opn4+ cells on average (n
mice
=
7) in the RGC layer using immunohistochemistry. Based on TVA-mCherry expression, we identified two subpopulations that project to SCN
VIP
neurons: one that expresses mCherry and is labeled by the Opn4 antibody and one that expresses mCherry only. In the experimental group,
we found 60
G
4 mCherry+ cells (n
mice
= 4, mean
G
SEM). Only a subset of Opn4-ab labeled cells in the retina was co-labeled by mCherry
from SCN
VIP
neurons (3.5
G
1%). Among RGCs identified as projecting to SCN
VIP
neurons (i.e., those that expressed mCherry), a subset of
14
G
2% was co-labeled with Opn4-ab (
Figure 1
D). No mCherry expression was found in the control group (n
mice
=3,
Figure S1
B). These
results suggest that SCN
VIP
neurons can be divided into two functional subpopulations based on retinal connectivity.
Responses of SCN
VIP
neurons to short light intervals are wavelength-dependent
Next, we hypothesized that the different retinal connectivity of SCN
VIP
neurons could give rise to unique responsiveness to light spectral
properties and intensities. To test this, we recorded GCaMP6s signal from freely behaving VIP x Ai162(GCaMP6s) mice implanted with optical
fibers (400
m
m diameter) positioned above the SCN, using a FP setup
31
,
32
(
Figures 2
Ai). Fiber location was visualized using a 3D histology
method we designed for deep targets such as the SCN
31
(
Figures 2
Aii). We applied short intervals of ambient light (15 s) during the dark phase
(ZT15
G
1.5), separated by 30 s intervals of dark to allow response relaxation, while recording GCaMP6s activity in freely behaving mice for a
total of six trials. Light was introduced in a random order of seven different wavelengths: 395, 438, 473, 513, 560, 586, and 650 nm, correspond-
ing to violet, blue, cyan, teal, green, orange, and red light, respectively, covering the major contributions of different opsins (
Figures 2
Aiii).
Applying an equal number of photons at each wavelength (1.55
G
0.05E15 photons/cm
2
/s, measured at the fiber tip, 3.4
G
1E14 photons/
cm
2
/s at the cage bottom), we observed that SCN
VIP
calcium responses differed in their amplitude (
Figures 1
B–1D), showing maximal
response at 438 nm (blue, z-scored dF/F = 5.1
G
0.9) and reduced responses toward red wavelengths (z-scored dF/F = 0.3
G
0.1 at
650 nm). As a side note, the amplitude of the FP signal depends on the location of the fiber relative to the fluorescent target
33
and we found
that the responses to red light, while low in amplitude in this specific experiment, were higher compared to the control (VIP x Ai140(GFP))
when the fiber was closer to the target cells (
Figure S2
). The responses to different wavelengths showed characteristic changes in peak ampli-
tude from first to second repeats (
Figure 2
E, left) and second to third repeats (right, shown as a function of wavelength, as defined in (b)), with a
decrease in amplitude specifically in response to 438 nm (
D
P
1-2
, p = 0.054, and p = 0.016 when excluding one outlier, Kruskal-Wallis Test).
Comparing the spectral sensitivity we observed by FP (
Figure 2
F) to those known for mouse opsins (
Figure 2
G) showed that SCN
VIP
activity
cannot be explained solely by Opn4
34
and M-opsin (expressed in cones
35
), suggesting input from additional opsins, potentially Rhodopsin for
red-shifted sensitivity,
36
,
37
and S-opsin (expressed in cones) for blue-shifted sensitivity.
38
In summary, measuring the spectral response of cal-
cium signals from SCN
VIP
neurons showed broad spectral sensitivity, wider than the sensitivity of individual opsins, which peaks 40 nm away
from the highest sensitivity of Opn4, hinting that mouse Rhodopsin, Opn4, M-opsin, and S-opsin all contribute to the ensemble activity of
SCN
VIP
neurons.
Figure 1.
Continued
(E) Example of ipRGCs labeled with Opn4-ab that do not express mCherry.
(F) Example of mCherry-expressing RGCs that do not show Opn4-ab labeling.
(G) Example of Opn4-ab-labeled RGCs (ipRGCs) expressing mCherry and projecting to SCN
VIP
neurons. Brn-3c staining was used to identify the RGC layer. See
also
Figure S1
.
ll
OPENACCESS
iScience
26
, 107865, October 20, 2023
3
iScienc
e
Article
Figure 2. SCN
VIP
neurons’ GCaMP6s responses to 15 s of light are wavelength-dependent
(A) The fiber photometry experimental setup. (
i
) VIP-Cre x Ai162(GCaMP6s) mice implanted with two 400
m
m optical implants at a 13

angle. (
ii
) LiGS histology to
visualize fiber placement. (
iii
) Top: light exposure patterns. Left: top view of cage as blue light is introduced. Right: spectral specifications of the light exposures.
(B) Examples of individual responses to 15 s of light at the indicated wavelengths (n = 6, 4 females, 2 males).
(C) Averaged responses (6 repeats, z-scored, thick lines) and SEM (thin lines). Yellow bars represent light exposure times.
(D) Normalized z-scored dF/F responses.
(E) The change in peak amplitude from the first to second repeat (left) and second to third repeat (right), as a function of wavelength, as defined in (B).
(F) Comparison between the integrated responses to the seven wavelengths. Light intensities were matched to have an equal number of 1.55E15 photons a
t the
fiber tip.
(G) SCN
VIP
FP responses overlaid with opsin sensitivity spectra: S-opsin, Opn4, M-opsin and Rhodopsin. See also
Figure S2
.
ll
OPENACCESS
4
iScience
26
, 107865, October 20, 2023
iScienc
e
Article
Figure 3. SCN
VIP
neurons show an intensity dependency in their blue and red light response profiles
(A and B) Averaged light response to (a) ambient vivarium white light (n = 9, 6 females, 3 males) and (b) ambient vivarium red light (z-scored, n = 8, 5 fema
les, 3
males).
(C and D) Averaged light response to different intensities of (c) 650 nm red light, 6.32E14 to 1.82E14 photons/cm
2
/s and (d) 438 nm blue light, 1.4E15 to 1.76E13
photons/cm
2
/s (n = 4, 1 females, 3 males).
ll
OPENACCESS
iScience
26
, 107865, October 20, 2023
5
iScienc
e
Article
SCN
VIP
neurons show an intensity dependence in their blue and red light response profiles
Inspired by previous work showing that spectral responses in the SCN can indicate retinal connectivity,
30
,
37
,
39
we tested whether light inten-
sities at two wavelengths, blue (438 nm) and red (650 nm), resulted in distinct functional neuronal activity. We used an ambient light exposure
of 15 s at several intensities, from 1.76E13 to 1.40E15 photons/cm
2
/s, depending on the wavelength of light, including vivarium white and red
light, for comparison (
Figures 3
A–3D, see detailed response profiles in
Figure S3
). Overall, responses to 650 nm light were lower in amplitude
and not detectable at intensities at or lower than 1.82E14 photons/cm
2
/s (
Figure 3
C). Responses to blue light were easily detectable over the
applied range (
Figure 3
D). Comparing the averaged response profiles (
Figure 3
E1, 3E2) showed that high-intensity light induced a unique
response profile, i.e., a fast initial rise at stimulus onset, decreasing by 50–60% within 10–20 s and then sustaining for the duration of the light
exposure. The correlation of the temporal profiles of the responses to 438 nm light decreased, with correlation coefficients (CCs) reducing
from 1 to 0.92, indicating loss of the unique response profile with decreasing intensity. We observed a more substantial decline in correlation
in the responses to red light. Over a small range of intensities (6.32E14 to 1.95E14 photons/cm
2
/s), the CC values decreased from 1 to 0.63, as
the ‘‘high-intensity’’ response profile flattened (
Figure 3
E3). Ambient white room light at 4.49E14 photons/cm
2
/s caused the maximal ampli-
tude of dF/F peak values (dF/F = 10 a.u.). Blue light (438 nm) induced relatively high median dF/F peak values (6.9 a.u.), which were non-linearly
correlated to the light intensity (increasing from 3.1 to 6.9 a.u. in response to 1.76E13 and 4.27E14 photons/cm
2
/s, respectively), as was found
by Walmsley and colleagues.
39
Overall, the responses to 650 nm light were low-amplitude, between 0.4 and 1.3 (dF/F, a.u.) (
Figure 3
F1). The
light intensity also affected the rise time in response to both 650 nm and 438 nm stimuli (
Figure 3
F2). Turning off the light induced an
exponential decay, with decay times unaffected by stimulus wavelength or intensity (
Figure 3
F3).
In addition to the averaged response, we found that repeated stimuli (over the six repeats) induced two non-identical temporal responses.
Examples of responses to white, blue (438 nm), and red (650 nm) light are presented (
Figure 3
G). By quantifying the change in peak responses
in the first three repeats, we observed that the second response had a lower amplitude than the first (
D
P
1-2
< 1). This phenomenon occurred
when the mice were exposed to white room light, as well as when they were exposed to blue (438 nm) light at intensities higher than 1.3E14
photons/cm
2
/s, and was not observed at lower intensities or 650 nm light exposure (
Figure 3
h1). This property disappeared after the second
light exposure (
D
P
2-3
z
1,
Figure 3
h2). These results show that SCN
VIP
neurons are sensitive to both wavelength and intensity and emphasize
that the neurons differ in their immediate responses to 650 nm vs. 438 nm light. Combining these results with our previous spectral compar-
ison (
Figure 2
G) strengthens the assumption that these multiphasic responses could indicate multiple pathways contributing to SCN
VIP
neurons’ activity at the presence of light.
Opn4 antagonist changes the transient response of SCN
VIP
neurons to blue but not red light
We hypothesized that blue light contributes to SCN
VIP
responses mainly through melanopsin-expressing ipRGCs via the retinohypothalamic
tract, whereas red light exerts its effects through a red-shifted opsin. To test this idea, we used a pharmacologic approach, a unique oppor-
tunity afforded by our
in vivo
recording. This approach bypasses the need for transgenic mice, which could have developmental compensa-
tions. We performed calcium recording from SCN
VIP
neurons while the mice were in the dark phase, exposed to 15 s of ambient blue (438 nm)
or red (650 nm) light, before and 20 min after intraperitoneal (i.p.) administration of sulfonamide AA92593 (1-(4-methoxy-3-methyl-benzene-
sulfonyl)-piperidine), a melanopsin antagonist
40
(
Figure 4
A, n
mice
= 6). Following AA92593 administration, exposure to blue light resulted in
SCN
VIP
neuronal responses which were flattened relative to their double-humped profile before Opn4 antagonist administration (
Figure 4
B-C
i-iii). The responses to blue light were similar before and after administering the solvent (DMSO) alone (
Figure S4
). By contrast, the temporal
responses to red light were not affected by AA92593 (
Figure 4
C iv-vi). Comparing the quantified parameters of the responses revealed that
the mean width of the response peaks significantly increased following AA92593 application (
Figure 4
D). As we saw before, responses to
650 nm red light and 438 nm blue light were significantly different, based on the mean area under the curve (AUC) and rise-time
(
Figures 4
E and 4F). Overall, these results show that while SCN
VIP
neuronal activity is sensitive to both blue and red light, an Opn4 antagonist
alters SCN
VIP
GCaMP response profiles to blue light but not to red light, supporting the idea that SCN
VIP
receives input from both Opn4-
expressing cells as well as non-Opn4 cells, which mediate the response to red light.
Single-cell calcium responses to light reveal at least two SCN
VIP
subpopulations
To test whether individual SCN
VIP
neurons could be separated into at least two subpopulations with distinct response profiles, we used a
mini-endoscope
41
to image single-cell calcium signals in SCN
VIP
neurons in response to 15 s of ambient white room light (ZT15
G
1.5,
Figure 3.
Continued
(E) Normalized response profiles to (
e1
) white light (black), vivarium red light (dark red), and 650 nm red light at different intensities (red) and (
e2
) 438 nm blue light
at different intensities (blue: > 1E14 and black: < 1E14 photons/cm
2
/s). The normalized response to the 650 nm light exposure at 1.82E14 photons/cm
2
/s was at
noise level and is not displayed. (
e3
) Matrix of correlation coefficients (CCs) between response profiles.
(F) Detected parameters across intensities and light sources. Black: white room light; dark red: red room light; blue: 438 nm light source; and red: 65
0 nm light
source. All values represent median
G
SEM. (
f1
) Maximum amplitude of non z-scored dF/F. (
f2
) Rise time, defined as the time from the beginning of the light
exposure to the peak response. (
f3
) Decay time after turning the light off.
(G) Examples of dF/F responses to white light, 438 nm blue light, and 650 nm red light, six repeats each.
(H) The change in peak amplitude from the (
h1
) first to second repeat and (
h2
) second to third repeat. All responses are averaged over six repeats (except (g)). See
vivarium white and red light spectra in
Figure S3
.
ll
OPENACCESS
6
iScience
26
, 107865, October 20, 2023
iScienc
e
Article
Figure 4. Opn4 antagonist changes the transient profile of SCN
VIP
neurons’ GCaMP6s responses to blue light but not to red light
(A) Experimental design.
(B) Individual responses to ambient blue light, before and 20 min after AA92593 application (colored lines: individual repeats, black line: mean).
ll
OPENACCESS
iScience
26
, 107865, October 20, 2023
7
iScienc
e
Article
Figures5
A–5C; n
cells
=120;n
mice
=7). Non-parametric numerical clustering of theresponseprofiles separated the SCN
VIP
neuronsinto three
subpopulations (k-means clustering,
Figures 5
D and 5E). The first two (clusters #1 and #2, n
cells
= 51 and n
cells
= 36, respectively) differed in
their mean AUC, and both showed higher AUC and amplitude than cluster #3. The responses contributing to cluster #3 (n
cells
= 31) were
noisy, some were inhibitory, and all were characterized by low amplitude and AUC (28
G
8 a.u., median
G
SEM) and by a relatively slow
rise time (5.8
G
0.5 s). Comparing the classification with two vs. three clusters, we found that three clusters resulted in a cleaner classification,
but the PCA plots do not strongly suggest that cluster #3 is a unique cell population (
Figure S5
). Instead, based on the large distribution of
the rise times, and the low ratio between the response intensities at 5 and 15 s, we think it represents mainly noise (
Figure 5
G). At the single-
cell level, while the averaged
D
P
1-2
(asdefined in
Figure 2
B)was lowerincluster#1 vs. #2, itwasnotsignificantly different,norwas
D
P
2-3.
More
importantly, the responses in cluster #1 had a slower rise time on average than in cluster #2 (6.8
G
0.2 s vs. 4.4
G
0.3 s, respectively) (
Fig-
ure 5
F), a key parameter we identified above that distinguishes unique light responses (
Figure 3
F). We identified a small bias between sexes,
with 43% and 56% clusters #1 and #2, respectively, in males and 67% and 33% in females (see details in
Table S2
). We also applied a Gaussian
clustering approach (3 clusters), which revealed similar results, showing increased amplitude and faster rise time of cluster #2 compared to
cluster #1 (
Figure S5
). Therefore, we concluded that the single-cell responses of the two clusters are highly correlated with the two FP
response profiles we observed at different intensities and with Opn4 antagonist manipulation. We next aimed to reveal how each
subpopulation responds to different light conditions.
Single-cell calcium responses to different light intensities support two SCN
VIP
subpopulations with intrinsic response
profiles
Following our identification of SCN
VIP
subpopulations at the single-cell level, we next tested the consistency of their responses across varying
stimulus conditions. Since various intensities of blue light (438 nm) all produced a detectable signal by both bulk FP and single-cell recording
in our previous experiments, we used the same light exposure paradigm and recorded SCN
VIP
neuronal responses to blue light at three
different intensities, 1.4E15, 1.3E14, and 1.7E13 photons/cm
2
/s, as well as to white room light for comparison. We recorded 56 neurons in
total (
Figure 6
A, four light conditions, n
mice
= 4). We then applied the same clustering approach as in the previous experiment (
Figure 5
)
to each light condition, limiting the classification to two clusters due to the smaller number of cells. Representative normalized mean re-
sponses of the two clusters in two light conditions are presented in
Figure 6
B. To understand if the cell response properties are unique to
each cell or change with intensity, we compared the cell identification in each cluster across the four light conditions. We found that 70%
of the neurons (39/56) kept their cluster identification in all different light conditions. Comparing the clustering of the high-intensity light re-
sponses, white and blue, we found that only three neurons changed their cluster identity. This observation affirms the power of the clustering
analysis and indicates that blue light is sufficient to induce responses similar to those induced by white room light (
Figure 6
C). Representative
traces from three cells, two that preserved cluster identity (one from each cluster), and one that changed identity, are shown in
Figure 6
D (see
also
Figure S6
). As these examples show, the amplitude was the response property that differed the most between conditions. We next tested
whether other response properties also changed by analyzing the response profiles across the entire dataset, focusing on amplitude, rise
time, and decay time. We observed that cells in both clusters showed a significant decrease in mean AUC when blue light intensity was
reduced from 1.4E15 to 1.7E13 photons/cm
2
/s, which recovered when white light was applied. Cluster #1, characterized by the relatively
flat response profile, was more sensitive to light intensity (
Figure 6
E, light gray). Rise times also changed and were significantly longer for
blue light (7.2
G
0.5 s, mean
G
SEM) than white light (4.8
G
0.4 s), with a negligible effect of blue light intensity. Cluster #1 showed a trend
of slower rise times compared to cluster #2 across light conditions, but the differences between the two clusters were insignificant (
Figure 6
F).
Decay times did not change between light applications and were consistent with the previous dataset (

0.5 s

1
,
Figure 6
G). These results
show that different intensities of blue light mainly altered response amplitude while the overall architecture of the responses was preserved.
Finally, we compared the response to red light at the single-cell level, revealing that the two clusters differ in their response to red light, with
cluster #1 not responsive to red light, and both clusters responsive to blue light (
Figure S6
). Overall, these results support the hypothesis that
the response profiles are intrinsic properties of the cells, and could indicate connectivity to rods and ipRGCs.
Opn4 antagonist application changes the single-cell response profiles of both SCN
VIP
clusters
Finally, we wanted to test whether the clusters we identified differ in their response to the Opn4 antagonist AA92593, at the single-cell level.
We recorded SCN
VIP
neuronal responses to blue light at 4.3E14 photons/cm
2
/s, before and 20 min after AA92593 application, from 84
neurons (
Figure 6
H, n
mice
= 4). At the averaged population level, the overall responses before and after AA92593 application (
Figure 6
I)
were similar to what we observed by bulk FP, affirming those results (
Figure 4
). However, single cells responded differently to the drug
Figure 4.
Continued
(C) Averaged z-scored dF/F response to blue light (
i
) before and (
ii
) after application of AA92593 (n = 6, 4 females, 2 males), as well as averaged z-scored dF/F
responses to red light (
iv
) before and (
v
) after application of AA92593 (thick lines: average over 6 repeats, colored background: SEM) (
iii
) and (
vi
) Averaged
responses.
(D–G) Comparisons between dF/F parameters. (d) Mean peak width. (e) Mean area under the curve (AUC). (F) Rise time. (G) Exponential decay constant.
Additional details and parameter comparisons are shown in
Figure S4
. Data is shown in MATLAB ‘‘boxplot’’ presentation, showing sample median and the
25th (bottom) and 75th (top) percentiles of the sample. Nonparametric Kruskal-Wallis test, followed by Bonferroni correction (*p < 0.05, **p < 0.01)
. See also
Figure S4
.
ll
OPENACCESS
8
iScience
26
, 107865, October 20, 2023
iScienc
e
Article
(examples shown in
Figure 6
J). To better understand this, we applied k-means clustering, revealing two clusters (
Figures 6
K and 6A third clus-
ter was excluded as noise) separated by the same features, such as AUC and rise-time (
Figure 6
L), as we found previously. Testing the effect of
AA92593 on each of the clusters revealed that none of the parameters changed significantly. These results clarify that both RGC subpopu-
lations connected to SCN
VIP
neurons express Opn4, even though only one was robustly labeled by Opn4-ab.
Figure 5. Single-cell calcium responses to light with a one-photon miniscope reveal at least two SCN
VIP
subpopulations
(A) Experimental setup.
(B) Example of identified cells.
(C) Responses of 120 individual cells to 15 s of ambient white room light, repeated six times. Top: mean of all cells. Bottom: heatmap of z-scored dF/F (n
=7,6
females, 1 male).
(D) Cell population distributions of three clusters, defined by k-means clustering. For comparison, we also applied Gaussian clustering (
Figure S5
).
(E) Averaged temporal pattern of the three clusters shown in (D) (thick lines: average over 6 repeats, colored background: SEM).
(F and G) Parameters of the response profiles for each cluster. Nonparametric Kruskal-Wallis test, followed by Bonferroni correction (#p < 0.0005, al
l clusters,
***p < 0.005, **p < 0.01, ‘boxplot’ MATLAB presentation next to full data displayed as asterisks). See also
Figure S5
.
ll
OPENACCESS
iScience
26
, 107865, October 20, 2023
9
iScienc
e
Article
Figure 6. Single-cell calcium responses to light at different intensities and Opn4 antagonist support two SCN
VIP
subpopulations, both affected by Opn4
antagonist
(A–G) Responses of 56 individual cells to 15 s of blue light (438 nm) at intensities of 1.4E15, 1.3E14, and 1.7E13 photons/cm
2
/s and to ambient white room light,
repeated six times. (A) Heatmaps of individually z-scored dF/F (n = 4, 1 female, 3 males). (B) Clustering results, showing two conditions, 1.4E15 and 1
.3E14
photons/cm
2
/s. (C) Comparison of individual cells’ cluster identity across intensity changes. Green: cells that preserved cluster identity; gray: cells that
changed cluster. (D) Responses of three representative cells. Top and middle: cells which retained cluster classification across conditions, from c
lusters #1
ll
OPENACCESS
10
iScience
26
, 107865, October 20, 2023
iScienc
e
Article
DISCUSSION
In this work, we test the idea that SCN
VIP
neurons, previously reported to comprise two subpopulations, can also be separated based on their
response to light, indicating a possible role of the retina and underlying neural connectivity in modulating SCN
VIP
functional activity. Mono-
synaptic tracing from SCN
VIP
neurons to the retina supports the existence of two subpopulations, one clearly labeled as ipRGCs, and one not,
as presented above. To further investigate this separation, we used an
in vivo
approach in which mice were exposed to repeated 15-s rounds
of ambient light at the beginning of the dark phase (ZT15
G
1.5) under various light and pharmacological conditions. At the same time, we
recorded changes in calcium responses using GCaMP6s at both bulk and single-cell levels. By spectral analysis of SCN
VIP
neurons’ bulk
GCaMP6s responses to these short light exposures, we found that SCN
VIP
neurons show an intensity-dependent response to both blue
and red light, and an Opn4 antagonist specifically affected blue light responses, indicating two distinct functional response profiles. Indeed,
single-cell calcium responses to light revealed the existence of at least two SCN
VIP
subpopulations, supported by their responses to different
intensities of blue light. We found that both clusters are sensitive to light intensity but differ in their overall AUC. In addition, we found that the
Opn4 antagonist AA92593 significantly changes the averaged response profiles, but this significance is lost when the clusters are separated,
hinting at an effect on both subpopulations. Based on these results, we conclude that light inputs through the retina modulate two response
profiles of SCN
VIP
neural activity, reflecting distinct underlying retina-SCN connectivity, likely through ipRGC types M1 and M2.
42
,
43
Based on GCaMP6s
in vivo
imaging and quantification, we characterized two SCN
VIP
light response profiles–one relatively flat and highly
sensitive to light intensity, and the other bimodal, with a faster rise time, which were also identified at the single-cell level. What could explain
the distinction of these responses in spectral and intensity dependence, Opn4 dependence, and light adaptation? We suggest that the two
clusters represent distinct SCN
VIP
subpopulations. Our results show that cluster #2 has a relatively fast rise time and a dominant response to
blue light. This fast-rising response profile is missing when Opn4 antagonist is present. The response profile of cluster #1 has a higher
dependency on the light intensity, slower rise time, and also appears in response to red light. At the single-cell level, Opn4 antagonist
has an effect on the aggregate of both clusters. In addition, Opn4 antagonist does not change the response to red light, at least at the level
of bulk recording. These observations, together with our monosynaptic tracing results, lead us to suggest that cells of cluster #1 are connected
to ipRGCs that are not stained well by Opn4-ab, possibly type M2, and cluster #2 cells are connected to Opn4-ab-labeled ipRGCs, likely type
M1, which are also highly sensitive to blue light, the peak sensitivity of Opn4 (see also.
42
,
43
Supporting this idea, Beier et al. showed that the
SCN receives nearly exclusive input from ipRGCs,
44
which we can now trace specifically to SCN
VIP
neurons.
Several electrophysiology studies support the existence of subpopulations in the light-sensitive neurons in the SCN, including SCN
VIP
neurons, as demonstrated here and in previous work.
4
,
6
First, Drouyer et al. demonstrated two subpopulations that differ in their kinetics.
37
Interestingly, they observed an intensity-dependent initial fast transient, which disappeared when light intensity was reduced from 2E14 to
2E12 photons/cm
2
/s. These findings align with our observations of intensity-dependent response to blue light and the change in the response
profile, in both bulk and single-cell imaging (
Figures 3
and
6
). The two profiles are consistent with typical rod- and cone-driven retinal mech-
anisms and intensity-dependent dynamics,
17
where rods are responsible for slow kinetics and low intensities. Additional support for two types
of retinal connectivity comes from the work of Walmsley et al., which divided the light-responsive neurons in the SCN based on color and
irradiance sensitivity.
39
Although our work differs in experimental design (they replaced the mouse cone with a human long-wavelength-sen-
sitive cone), we found similar temporal profiles of the responses to light. Lastly, SCN activity in slices from the diurnal rodent
Rhabdomys
pumilio
shows two profiles in response to dim and bright light.
45
While electrophysiology recordings lack cell-type specificity, the similarity
in the response profiles supports the existence of two SCN
VIP
subpopulations.
An alternative explanation for the two profiles could be pupil adaptation, which would explain the difference in response to subsequent
applications of high but not low-intensity blue light. In addition, genetic manipulations of Opn4 expression have been shown to prevent full
pupil constriction at high intensities,
22
,
46
and the AA92593 Opn4 antagonist we used here caused a delay in pupil diameter reduction on a
timescale of 5–30 s,
40
which may be due to Opn4 expression in the mouse iris sphincter muscle.
47
However, pupil constriction does not explain
the presence of two functional clusters at the single-cell level. Therefore, we think this pathway has a minor contribution to the neuronal
response.
The response to red light indicates contributions to SCN
VIP
neuronal activity from other light receptors, such as Rhodopsin (see also
43
,
48
). In
particular, our observation that the FP responses to red light and low-intensity blue light have slower rise times supports the conclusion that
the response to red light (and possibly blue light) is mediated by a photoreceptor other than melanopsin. This concurs with previous findings
that Opn4 is not involved in locomotor activity responses to dim, 20 lux, light,
49
and that rods and cones are both involved in dim light
processing (see also
48
). In addition,
in vivo
SCN electrophysiology in freely moving mice has demonstrated that cone opsins contribute to
Figure 6.
Continued
and #2, respectively. Bottom: a cell whose classification switched between conditions. (E-G) Quantification of response profiles over sessions (left
to right) and
clusters (dark and light gray). (E) Mean AUC. (F) Mean rise time. (G) Decay time.
(H–N) Responses of 56 individual cells to 15 s of blue light (438 nm) at an intensity of 4.3E14, repeated six times, before and 20 min after AA92593 applic
ation. (h)
Heatmaps of individually z-scored dF/F (n = 4, 2 females, 2 males). (I) Averaged cell response, before (orange, solid line), and after (purple, dashed
line) AA92593
application. (J) Individual cell responses sorted by k-means cluster, before (orange, solid line), and after (purple, dashed line) AA92593 applica
tion. (K) Averaged
cell responses sorted by k-means cluster. (L
–N
) Quantification of response profile parameters: (L) mean AUC, (M) mean rise time, and (N) decay time. In (B), (D), (I–
K): thick lines: average over 6 repeats, colored background: SEM. In (E–G), (L–N): The bars and lines represent mean
G
SEM. Statistics within clusters (orange, red)
and between clusters (black): Nonparametric Kruskal-Wallis test, followed by Bonferroni correction (*p < 0.05, **p < 0.01, ***p < 0.005, #p < 0.0005)
. See also
Figure S6
.
ll
OPENACCESS
iScience
26
, 107865, October 20, 2023
11
iScienc
e
Article
circadian regulation.
30
The response to red light at 650 nm should raise a concern about the usage of red light in vivarium, particularly in
circadian and sleep related experiments.
Our recordings of GCaMP signals of SCN
VIP
neurons in response to light, also done previously by us and others using FP,
4
,
6
,
31
highlight the
relative advantages and disadvantages of bulk and single-cell calcium imaging. FP is relatively easy to perform and robust, while single-cell
imaging from the SCN has become feasible only recently.
50
,
51
Acquiring high-quality data is challenging because the SCN is a small (400
m
m
ML, 700
m
m AP) and deep (5.5mm) target that suffers from high levels of motion artifacts. Performing both analyses, as we do here, allows
comparison of the results to achieve a better understanding. For example, we can see that the rise-time dependency on the blue light intensity
(FP,
Figure 3
) is mainly due to an increased contribution of cluster #2 rather than to a change in individual cells’ response profile (single-cell,
Figures 5
E and 5F).
Sensitivity to light properties is fundamental for the critical role of the SCN in mediating naturally relevant information, such as day vs.
night, the specific time of day, such as twilight vs. midday,
39
as well as seasonal information.
52
In addition to keeping the daily rhythm, the
SCN’s ability to relay information about various light properties is essential for hormonal modulations
53
and immune responses.
54
Therefore,
understanding the mechanism that connects light properties to the SCN and beyond is highly important for better detecting and treating
circadian-related ailments. We find that SCN
VIP
neuronal response profiles to light reflect at least two subpopulations. In the future, it will
be of great interest to examine how this relates to other genetic and functional subpopulation definitions,
6–8
increasing our understanding
of how retinal connectivity and sensitivity affect the SCN in particular, and circadian behavior in general.
Limitations of the study
One difficulty in calcium imaging is understanding the cellular source of the calcium signal. Calcium flux is required for circadian rhythm gen-
eration in mammalian pacemaker neurons,
55
and it has been noted that the intracellular circadian calcium rhythm in the SCN has a dual origin:
light-induced Ca+ flux and action potentials.
56
Thus, the signal we measured here may be a combined response of GCaMP to Ca+ increase.
However, extracellular electrophysiology of SCN neurons from anesthetized mice showed an immediate increase in firing rates in response to
high intensities of light,
26
,
57
suggesting that the immediate response of GCaMP follows the increased firing rates. In addition, similar response
profiles appear in other NIF nuclei, the pretectal olivary nuclei (PON).
58
STAR
+
METHODS
Detailed methods are provided in the online version of this paper and include the following:
d
KEY RESOURCES TABLE
d
RESOURCE AVAILABILITY
B
Lead contact
B
Materials availability
B
Data and code availability
d
EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS
B
Mice
d
METHOD DETAILS
B
Surgery
B
Fiber photometry recording
B
Single-cell imaging
B
Behavioral assays
B
Tissue collection and processing
B
Light-guided sectioning (LiGS) tissue preparation
B
LiGS cryo-sectioning
B
LiGS staining and clearing
B
Monosynaptic tracing tissue preparation
B
mRNA detection
B
VIP probes
B
Immunofluorescence imaging
d
QUANTIFICATION AND STATISTICAL ANALYSIS
B
Fiber photometry
B
Single-cell imaging
B
K-means clustering of calcium transients
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at
https://doi.org/10.1016/j.isci.2023.107865
.
ll
OPENACCESS
12
iScience
26
, 107865, October 20, 2023
iScienc
e
Article