Report

A parvalbumin-positive excitatory visual pathway to trigger fear responses in mice

See allHide authors and affiliations

Science  26 Jun 2015:
Vol. 348, Issue 6242, pp. 1472-1477
DOI: 10.1126/science.aaa8694

Looming in on the threat-response circuit

What are the neural elements that transmit threat-relevant inputs in the brain? Shang et al. systematically identified the key neuronal subtypes in the mouse superior colliculus underlying active avoidance and defensive-like behaviors. They found a pathway that responded to looming objects, linking input from the retina to the fear center in the brain.

Science, this issue p. 1472

Abstract

The fear responses to environmental threats play a fundamental role in survival. Little is known about the neural circuits specifically processing threat-relevant sensory information in the mammalian brain. We identified parvalbumin-positive (PV+) excitatory projection neurons in mouse superior colliculus (SC) as a key neuronal subtype for detecting looming objects and triggering fear responses. These neurons, distributed predominantly in the superficial SC, divergently projected to different brain areas, including the parabigeminal nucleus (PBGN), an intermediate station leading to the amygdala. Activation of the PV+ SC-PBGN pathway triggered fear responses, induced conditioned aversion, and caused depression-related behaviors. Approximately 20% of mice subjected to the fear-conditioning paradigm developed a generalized fear memory.

Environmental threats are detected by different sensory organs projecting to central brain areas to trigger fear responses (1, 2). The superior colliculus (SC) is a retinal recipient structure (3, 4) composed of different neuronal subtypes (5, 6), including parvalbumin-positive (PV+), somatostatin-positive (SST+), and vasoactive intestinal peptide–positive (VIP+) neurons (Fig. 1A and fig. S1). In addition to mediating orienting responses (7), the SC contributes to avoidance and defense-like behaviors (811). With an optogenetic approach (1214), we found that activation of neurons expressing channelrhodopsin-2 (ChR2) in mouse SC triggered freezing that lasted 52.8 ± 5.3 s (n = 5 mice) (movie S1). This prompted us to systematically identify the key neuronal subtypes underlying this behavior.

Fig. 1 Neuronal subtypes in the SC to trigger fear responses.

(A) Diagram of different neuronal subtypes in the SC. (B) Ai32 mice were crossed with different Cre lines. (C) Coronal micrographs showing ChR2-EYFP expressed in specific neuronal subtypes. DAPI, 4′,6-diamidino-2-phenylindole. (D) Light-induced spikes from ChR2-EYFP+ neurons in acute slices. (E) Instantaneous locomotion speed before, during, and after light stimulation. (Inset) Locomotion trails from example mice. (F and G) Analyses of peak speed and travel distance during and after light stimulation. Data in (F) and (G) are means ± SEM (error bars); numbers of mice are in bars. Statistical analysis is t test (***P < 0.001; n.s. P > 0.1). Dashed lines in (F) and (G) indicate the control levels measured from behaviors after SC SST+ neuron activation.

By crossing Ai32 (15) with different Cre lines (Fig. 1B) (16, 17), we expressed ChR2–enhanced yellow fluorescent protein (EYFP) in specific neuronal subtypes in the SC (Fig. 1C and fig. S1) and optogenetically elicited spikes in acute slices (Fig. 1D and fig. S1). Activation of SC PV+ neurons, but not SST+ or VIP+ neurons, triggered impulsive escaping (1.18 ± 0.09 s) followed by long-lasting freezing (46.4 ± 2.8 s) (Fig. 1, E to G; fig. S1; and movie S2). To avoid activation of PV+ retinal ganglion cells (RGCs) (18) by ferrule light, we injected adeno-associated virus (AAV) expressing double-floxed ChR2-mCherry (12) into the SC of PV-ires-Cre mice, resulting in specific expression of ChR2-mCherry in SC PV+ neurons but not in PV+ RGCs (Fig. 2A and fig. S2). The light triggered spikes from ChR2-mCherry–positive neurons in SC slices (Fig. 2B and fig. S2), elicited a similar stereotyped locomotor pattern (fig. S2 and movie S3), and increased the heart rate and plasma corticosterone levels that were not observed in mice with SC PV+ neurons expressing mCherry (Fig. 2, C and D).

Fig. 2 Specific activation of SC PV+ neurons induced fear responses.

(A) Specific expression of ChR2-mCherry in SC PV+ neurons of PV-ires-Cre mice. (B) The light-pulse train triggered spikes (red) from ChR2-mCherry–positive neurons and postsynaptic currents (black) from adjacent ChR2-mCherry–negative neurons. (C) Electrocardiographic traces and heart rate analyses from the anaesthetized mice before and after light stimulations. Ctrl, control. (D) Analyses of plasma corticosterone concentration in response to light stimulation. (E to I) Durations of escaping, freezing, and E/F ratios, were plotted as functions of stimulation intensity, duration, frequency, repetition, and sex in mice with SC PV+ neurons expressing ChR2-mCherry. Data in (C) to (I) are means ± SEM (error bars); numbers of mice are in bars. Statistical analysis is t test (***P < 0.001; **P < 0.01; n.s. P > 0.1). Dashed lines indicate the levels measured from control mice. M, male; F, female.

When facing threats, animals can either fight or flee. To test whether SC PV+ neurons were involved in this behavioral dichotomy, we measured the durations of light-induced escaping (E) and freezing (F) and calculated their ratio (E/F ratio). We conducted a series of tests spanning 5 days (table S1). First, light stimulations with higher intensity or longer duration enhanced E/F ratios in the same male mice by prolonging escaping more strongly than freezing (Fig. 2, E and F). Second, light stimulations with higher frequency but similar total illumination time prolonged escaping and freezing proportionally (Fig. 2G). Third, both responses showed strong adaptation to repetitive light stimulations (every 5 min), with no significant change in E/F ratios across each stimulation (Fig. 2H). Finally, the same light stimulations elicited longer escaping and shorter freezing in female versus male mice, resulting in higher E/F ratios in females (Fig. 2I). The origin of these sexually dimorphic behaviors was further examined (supplementary text).

We next characterized the morphological and physiological properties of SC PV+ neurons. They were predominantly but not exclusively distributed in the superficial gray (SuG) layer of the SC (Fig. 3A and fig. S3). Whole-cell recording of tdTomato-expressing PV+ neurons in SC slices from PV-ires-Cre; Ai9 mice (19) demonstrated that, distinct from V1 PV+ interneurons with slow frequency adaptation (20), the SuG PV+ neurons responded to depolarizing currents in a faster adaptation mode (Figs. 3, C and D, and fig. S3). SuG PV+ neurons labeled with neurobiotin had parallel dendrites extending to the SC surface, presumably receiving inputs from RGCs (Fig. 3B and fig. S3). The postsynaptic currents from PV-negative neurons induced by optogenetic activation of PV+ neurons expressing ChR2-mCherry were blocked by d-(–)-2-amino-5-phosphonopentanoic acid (APV) and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), but not by picrotoxin (Fig. 3, E and F, and fig. S3), suggesting that these neurons were glutamatergic but did not release γ-aminobutyric acid. The SC PV+ neurons in intermediate and deep layers of the SC were examined (supplementary text).

Fig. 3 Morphological and physiological properties of SC PV+ neurons.

(A) Layer-specific distribution of SC PV+ neurons. (B) Neurobiotin-labeled PV+ neurons in the SC SuG layer and V1. (C) Spike firings of PV+ neurons in the SC and V1 to depolarizing currents. (D) Analyses of spike number and spiking phase as a function of current intensity. n, number of cells. (E and F) Effects of CNQX (20 μM)/APV (50 μM) (E) and picrotoxin (50 μM) (F) on the postsynaptic currents (PSC) induced by light stimulation. n, number of cells. (G) Single-unit activity recorded from a putative SC PV+ neuron triggered by light pulses (arrows, 1 ms at 10 Hz). (H) Raster plot showing the latency of light-evoked spikes relative to the light pulses (0 ms). (I) Distributional plot (left) and example spikes (right) evoked by visual stimuli and light showing quantitative identification of PV-positive and PV-negative units based on the waveform correlation and energy of light-evoked spikes. (J) A virtual soccer ball flying toward the eye of an anaesthetized mouse. (K) Example single-unit traces from a putative SC PV+ neuron in response to the soccer ball (20 cm in diameter) moving in six directions at 2 m/s. (L) Peristimulus time histograms of a PV+ neuron to looming stimuli with controlled velocities (V) (2 and 5 m/sec) and diameters (D) (20 and 40 cm). Arrows indicate response onset time. (M and N) Correlation analyses of response onset time (M) and response peak time (N) of SC PV+ neurons and the square root of diameter/velocity of the looming ball. Data are means ± SEM (error bars); numbers of cells or units are in graphs. Statistical analyses are t test and one-way analysis of variance (***P < 0.001; n.s. P > 0.1). R, correlation coefficient.

To test whether SC PV+ neurons were involved in detecting collision threats in the visual field (2123), we displayed a virtual soccer ball moving in controlled velocities and directions to the anaesthetized mice (Fig. 3J and fig. S4). The single-unit activity recorded with optrodes was quantitatively identified (24, 25) as putative SC PV+ neurons expressing ChR2-mCherry (Fig. 3, G to I). These putative PV+ neurons (n = 9 cells) were strongly activated by the ball moving toward the animal but not by the motion in the other five directions (Fig. 3K). The response onset time before collision depended on the size and velocity of the ball (Fig. 3L) and was linearly correlated with the square root of the diameter/velocity (Fig. 3M). The response peak was close to the time to collision and was independent of the size and velocity of the soccer ball (Fig. 3N). In freely behaving mice, the escaping triggered by SC PV+ neuron activation pointed to the side of the SC receiving light stimulation (movie S4).

We then determined the circuit mechanism underlying the fear responses mediated by SC PV+ neurons. By injecting AAV expressing double-floxed monomeric green fluorescent protein (mGFP) into the SC of PV-ires-Cre mice, we specifically labeled SC PV+ neurons (Fig. 4B and figs. S5 and S6) and observed axon terminals in the parabigeminal nucleus (PBGN), the pontine nucleus (Pn), and the dorsal lateral geniculate nucleus (DLGN) (Fig. 4, A and C). These projections were confirmed by retrograde tracing with cholera toxin B with Alexa Fluor-594 (CTB-594). CTB-594 injection into the PBGN (Fig. 4D) retrogradely labeled SC neurons predominantly in the ipsilateral SuG layer (Fig. 4E and fig. S7). A considerable proportion of CTB-labeled SC neurons (SC-PBGN: 52 ± 4.3%; SC-Pn: 31 ± 4.5%; SC-DLGN: 33 ± 3.8%, n = 3 mice) were positive for PV (Fig. 4F and fig. S7).

Fig. 4 PV+ SC-PBGN pathway mediated fear responses.

(A to C) Specific expression of mGFP in SC PV+ neurons (B) of PV-ires-Cre mice resulted in labeling of their axon terminals in the PBGN, Pn, and DLGN [(A) and (C)]. (D to F) CTB-594 injected in the PBGN (D) retrogradely labeled cells in the SC (E), a large proportion of which were PV+ (denoted by arrowheads) (F). DpG, deep gray layer; InW, intermediate white layer; Ing, intermediate gray layer; Op, optic nerve layer; Zo, zonal layer. (G) Diagrams showing the optic fibers implanted either above the PBGN or the Pn to stimulate ChR2-mCherry–positive axon terminals. (H and I) Locomotion analyses during and after the activation of PV+ SC-PBGN and SC-Pn pathways. (J) Analyses of escaping, freezing, and E/F ratio in mice receiving activation of the PV+ SC-PBGN pathway and SC PV+ neurons. Data in (H) to (J) are means ± SEM (error bars); numbers of mice are in bars. Statistical analysis is t test (***P < 0.001; n.s. P > 0.1).

To determine which of these parallel projections (26) participated in the fear responses, we injected AAV expressing double-floxed ChR2-mCherry into the SC and implanted optic fibers in the PBGN or Pn (Fig. 4G) to locally stimulate the ChR2-mCherry–positive axon terminals. Activation of the PV+ SC-PBGN pathway, but not the PV+ SC-Pn pathway, triggered the stereotyped escaping-freezing locomotor pattern (Fig. 4, H to J; fig. S9, and movie S5). We examined whether PBGN projected to the amygdala by anterograde and retrograde tracings. Local injection of AAV-SynaptoTag (27) in the PBGN and its adjacent region strongly labeled axon terminals positive for synaptobrevin-2–EGFP in the central amygdaloid nucleus (28, 29), whereas CTB-594 injection in the amygdala retrogradely labeled neurons in the PBGN (fig. S8). Finally, the relation between PV+ SC-PBGN pathway activation and the affective state of mice was explored (supplementary text and figs. S10 to S12). Taken together, these data revealed a PV+ excitatory visual pathway to trigger stereotyped fear responses in mice.

Our data lead to the following conclusions. First, the SC PV+ neurons form a subcortical visual pathway that transmits threat-relevant visual information to the amygdala to trigger fear responses. These data, in alliance with earlier studies (18, 30, 31), suggest a “retina-SC-PBGN-amygdala-hypothalamus” pathway for vision-induced fear responses. Second, the SC PV+ neurons in the SuG layer are predominantly glutamatergic projection neurons with spiking patterns distinct from those of their counterparts in cortical regions. Thus, this finding broadens the concept of PV+ neurons (32) and adds another perspective to understanding their functions. Third, the SC PV+ neurons may belong to type-ρ looming detector, supporting the notion that mathematically defined computational units correspond to specific neuronal subtypes (33).

Supplementary Materials

www.sciencemag.org/content/348/6242/1472/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S12

Table S1 and S2

Reference (34)

Movies S1 to S5

Data S1

References and Notes

  1. Experimental procedures are explained in the supplementary materials on Science Online.
  2. Acknowledgments: We thank T. Südhof, K. Deisseroth, Y. Wang, B. Li, and M. Luo for providing plasmids, instruments, and technical support for this study. This work was supported by the Thousand Young Talents Program of China. We declare no conflicts of interest. All data are archived in the Institute of Biophysics, Chinese Academy of Sciences.
View Abstract

Stay Connected to Science

Navigate This Article