The VBA consists of a tube enclosure (virtual burrow), constrained to slide back and forth along the anterior-posterior axis of the body of a head-fixed mouse (Figure 1A,B). When placed in the virtual burrow mice retreat inside the enclosure, pulling the tube up around their bodies as far as possible—a behavior we have termed ‘ingress’ (Figure 1—video 1). All mice we have tested exhibit this behavior immediately after being introduced into the VBA, remaining fully inside the virtual burrow with only transient excursions from the ingress position (Figure 1C). Neither training nor acclimation to the assay is required to observe this consistent and sustained ingress.

Figure 1 with 4 supplements see all

Download asset Open asset

The assay measures the propensity of an animal to ingress in response to a given stimulus. It is therefore necessary to pull the virtual burrow away from the mouse in order to ask what induces the animal to ingress. When the enclosure is retracted by a motor tethered to the virtual burrow, mice invariably try to get back inside (Figure 1D, Figure 1—video 2). After an initial bout of resistance (<30 s), mice voluntarily maintain an ‘egress’ position and thus can be presented test stimuli. Mice quickly adapt to this scenario, exhibiting within one to three minutes periods of >15 s during which they voluntarily hold the virtual burrow in the egress position. This sequence of behaviors has been observed in all mice tested and been found to persist across repeated tests on four separate days over a 16 day interval (Figure 1—figure supplement 1). No prior acclimation to handling or head fixation is required.

We first determined whether an innately aversive stimulus, such as air puff, induces ingress, and if so, the degree to which this response is rapid and stereotyped (Figure 2A). Strong air puffs delivered to the snout (80 psi, 2 mm distance) elicited short-latency, rapid ingress in all mice tested on all trials (Figure 2B; Figure 2—video 1). Animals generated this behavior by pulling the burrow up to the ingress position in a coordinated, simultaneous movement of their fore- and hind-limbs. The latency of ingress varied little across animals (μ = 18.5 msec, c_v = 0.07, N = 5 mice, three trials per mouse) and across trials (c_v = 0.10). Weak air puffs (2 psi, 15 cm distance) did not elicit ingress but instead a transient movement resulting in no net change in burrow position. This apparent flinch-like startle response (Davis, 1984) can be clearly distinguished from the ongoing movement of the burrow caused by the animal’s breathing (Figure 2C).

Figure 2 with 1 supplement see all

Download asset Open asset

We then demonstrated that ingress in response to air puff reflects flight to shelter inside the virtual burrow, rather than just a backwards movement to avoid the source of the stimulus. First, we compared responses to air puffs delivered to the snout with air puffs delivered to the hindquarters (Figure 2D top, orange vs. green). As observed previously, all animals ingressed on all trials when air puffs were directed at the snout (Figure 2E orange and 2F left, N = 4 mice, three trials each). When air puffs were directed at the hindquarters, mice also invariably ingressed (Figure 2E green and 2F second from left, N = 4 mice, three trials each). Thus the drive to ingress into the virtual burrow overcomes the drive to move in the direction opposite the air puff. Second, we replaced the tube with a flat, open platform that was similarly constrained to move along the anterior-posterior axis of the mouse’s body (Figure 2D bottom). Mice in this configuration exhibited a transient, flinch-like response to air puff delivered to the snout or hindquarters (N = 4 mice each, three trials each) followed, in some cases, by relatively small displacements of the platform with little net preference between the egress or ingress directions (Figure 2E pink and blue, 2F second from right and right, median displacement of 14.9 mm and 17.0 mm in tube vs. 0.2 mm and 2.3 mm on platform, for snout and hindquarters, respectively). These observations, together with the stereotypy, high reproducibility across animals, and low latency of the response to strong air puff, and the fact that these behaviors require no training, suggest that ingress to air puff in the VBA reflects an innate behavioral program to flee to shelter.

We next asked whether visual stimuli that evoke flight in freely moving animals also evoke a response in the VBA. For this we employed a predator-like visual ‘looming’ stimulus known to trigger flight in freely moving mice (De Franceschi et al., 2016; Yilmaz and Meister, 2013) (Figure 3A, top left). Mice in the VBA ingressed on 80% of trials in response to an expanding black disk displayed above their heads (Figure 3A, top left, B, top, C, orange, and D, left; Figure 3—video 1). The habituation to repeated stimulation reported in freely moving mice (De Franceschi et al., 2016) was not observed in the VBA (Figure 3B, top). We also presented stimuli to mice in the VBA that do not elicit flight responses in freely moving mice: a contracting black disk and a small black disk sweeping across the visual field (Figure 3A, top middle and top right) (De Franceschi et al., 2016). These stimuli did not reliably elicit ingress, yielding responses on 0% or 13% of trials, respectively (Figure 3B, middle and bottom, C, pink and green, D, middle and right). These data show that a visual stimulus that elicits flight in freely moving mice elicits ingress in the VBA and that control stimuli that do not elicit flight do not elicit ingress.

Figure 3 with 1 supplement see all

Download asset Open asset

We next asked whether the VBA can measure habituation (Figure 4). We observed that mice reliably ingress to initial presentations of an odorant stimulus, and their response decreases over the course of repeated stimulation (ingress on 67% of first three trials; 10% of subsequent trials, N = 5 mice, Figure 4B, left, Figure 4C, orange, and Figure 4D, left). This response is odor-selective: after 15 presentations of the first odorant (Odor 1) we randomly interleaved it with a second odorant (Odor 2) and observed that animals selectively ingressed to the second, but remained habituated to the first odor (ingress on 47% of first three Odor 2 trials; 12% of subsequent trials, Figure 4B,C blocks 16–30, and Figure 4D, middle). After 15 further blocks of Odor 1 and Odor 2, resulting in habituation to the second odor, we randomly interleaved them with a third odorant (Odor 3), and once again observed selective ingress to the new odor (ingress on 60% of first three Odor 3 trials; 8% of subsequent trials, Figure 4B,C blocks 31–45, and Figure 4D, right).

Figure 4 with 1 supplement see all

Download asset Open asset

The selective response to a second odor following habituation to the first (Figure 4B–D) demonstrates that the response decrement is due neither to adaptation of the sensory epithelium nor to effector fatigue (Groves and Thompson, 1970; Rankin et al., 2009; Thompson and Spencer, 1966). A habituated response can transiently increase following presentation of a different stimulus, a phenomenon termed dishabituation (Groves and Thompson, 1970; Thompson and Spencer, 1966). However, we did not observe renewed ingress to Odor 1 following the first presentation of Odor 2 or Odor 3 (See Materials and methods section for discussion).

On early trials mice occasionally exit the virtual burrow before initiating their ingress (data not shown). We speculated that this voluntary egress corresponds to a brief bout of exploration. To test this, we coupled the odor source to the burrow, so that egress from the burrow brought the odor source closer to the animal’s nose, while ingress distanced it (Figure 4E). We reasoned that granting the animal control over its proximity to the odor port would allow it to select between the drive to explore an odorant stimulus (by exiting the burrow) and the drive to remain inside. In contrast to all other experiments, in which trials were initiated while the animal was in a mandatory egress position, here the animals were granted control over burrow position at all times and almost invariably maintained full ingress prior to stimulus delivery. We observed that under this configuration mice exited the virtual burrow for brief (~1 s) bouts before resuming a fully ingressed position (Figure 4F, Figure 4—video 1). This response habituated, with the animals egressing further on earlier than on later trials (median egress 21.9 mm for trials 1–2 vs. 10.5 mm for trials 3–5, N = 5 mice, Figure 4G,H). Together, these results indicate that the VBA measures selective responses to unanticipated stimuli—ingress or egress, depending upon the configuration of the assay—and that these behaviors can be employed to measure habituation.

We next asked whether the VBA can measure conditioned responses following aversive Pavlovian conditioning (Figure 5). One odorant (CS+) was paired with footshock in a conditioning chamber; a second (CS-) was presented in the chamber without footshock; and a third (Odor3) was not presented in the chamber but was tested in the VBA before and after conditioning.

Figure 5 with 2 supplements see all

Download asset Open asset

Prior to conditioning, mice initially ingressed in response to all three stimuli then habituated over subsequent trials (Figure 5D,F,H). After conditioning (48 hr later) the response recovered (Figure 5C,G, first trial), and the animals once again habituated to the CS- and Odor3. However, they continued to ingress to the CS+, exhibiting both a greater ingress likelihood (CS+ 72%, CS- 26%, Odor3 33%, mean across trials 2–7, six trials per mouse, pooled across N = 9 mice) as well as a greater displacement (median ingress 9.98 mm for CS+, 0.27 mm for CS-, and 0.36 mm for Odor3, N = 9 mice, Figure 5B,C,E,G,I) in response to the CS+ than to the CS- or Odor3.

On occasion we noted a second behavioral response evoked by the CS+ following conditioning: an oscillation in burrow position (Figure 5—figure supplement 1). Simultaneous video recording (not shown) indicated that this high frequency oscillatory response was associated with trembling of the animal’s body. While we observed ingress without trembling, on some trials trembling preceded ingress by several seconds (Figure 5—figure supplement 1, top trace), or occurred on ingress-free trials (Figure 5—figure supplement 1, middle trace), and was selective for the CS+ following conditioning. Thus the VBA can measure two kinds of stimulus-selective conditioned responses following aversive conditioning: trembling and ingress.

We took advantage of the high temporal precision of the VBA to measure the exact latency of ingress responses to CS+ presentations. Mice can report discrimination of odor stimuli within ~100 msec (Resulaj and Rinberg, 2015) and we have demonstrated that they are capable of initiating stimulus-evoked ingress within ~20 msec (Figure 2). It was therefore surprising to observe that the median latency to ingress to the CS+ was 709 msec, and 27% of conditioned respones (CRs) were initiated at latencies higher than 1 s following CS+ onset (Figure 6A,B). Precise measurement of CR onset revealed a second unanticipated feature: the frequency of high-latency (>1 s) responses increased significantly as a function of trial number, from 4.2% on the first three trials to 39% on last three trials (Figure 6C). Therefore, extinction is characterized not only by a decrease in the probability of a CR but also by an increase in the latency to produce a CR.

Figure 6

Download asset Open asset

We next tested whether it is possible to obtain stable extracellular recordings of single units in the brains of mice while they are behaving in the VBA. For this we implanted a silicon probe in layer 2 of the piriform cortex and recorded spontaneous spiking activity after head-fixation in the apparatus. No olfactory stimuli were presented during this experiment. After spike sorting we compared both mean and single-unit firing rates to three behavioral quantities: position, absolute velocity, and force. During abrupt transitions in behavioral state, neither mean nor single-unit rates appeared to be affected (Figure 7A). Indeed, the variance in mean firing rate accounted for by position, velocity and force is negligible (Figure 7B, top; linear regression R² = 9.3×10⁻³, 8.9 × 10⁻³, and 3.4 × 10⁻³, respectively). This observation also holds at the single-unit level (Figure 7B, bottom; linear regression median R² = 3.0×10⁻⁴, 1.8 × 10⁻³, and 1.1 × 10⁻³, respectively). We conclude that behavior in the VBA does not sufficiently disrupt simultaneous neurophysiological recording as to prevent or cause spurious detection of spike waveforms. The assay also permits stable measurement of calcium transients in response to odorant stimuli using 2-photon microscopy through a GRIN lens implanted in the basolateral amygdala (O’Neill, P.K., personal communication).

Figure 7

Download asset Open asset

All procedures were approved by the Columbia University Institutional Animal Care and Use Committee (protocols AC-AAAI8650 and AC-AAAT5466). 8–17 week old, male C57BL/6J mice (Jackson laboratories, Bar Harbor, ME) were fitted with a titanium head plate (27.4 mm x 9.0 mm x 0.8 mm, G. Johnson, Columbia University). Animals were anesthetized with isoflurane (3% induction, 1.5–2% maintenance) and placed within a stereotaxic frame (David Kopf Instruments, Tujunga, CA) on a feedback-controlled heating pad (Fine Science Tools, Foster City, CA). Carprofen (5 mg/kg) was administered via subcutaneous injection as a preoperative analgesic and bupivacaine (2 mg/kg) was delivered underneath the scalp to numb the area of the incision. The skull was exposed, cleaned with sterile cotton swabs and covered in a thin layer of cyanoacrylate adhesive (Krazy Glue, Elmer’s Products, Atlanta, GA). After applying a coating of adhesive luting cement (C and B-Metabond, Parkell, Inc., Edgewood, NY) onto the layer of cyanoacrylate adhesive, the titanium head plate was lowered atop the skull and secured with additional application of luting cement. The headplate was centered about the body’s anterior-posterior axis and equally spaced between bregma and lambda. For mice exposed to visual stimuli, however, head plate position was sufficiently posterior to prevent occlusion of the visual stimuli by the head plate. Mice were allowed at least one full week and typically greater than 4 weeks to recover before any testing was performed (Table 1). All animals were singly housed on a 12 hr/12 hr light/dark cycle and were tested during their dark phase.

The hardware design and control software are freely available for noncommercial use under the Creative Commons License at git.io/vA47E. The Virtual Burrow Assay (VBA) consists of a tube enclosure (virtual burrow) constructed out of a cardboard tube (looming experiments in Figure 3, and aversive learning experiments in Figure 5) or a 3D-printed polylactic acid tube (all other experiments, 45.5 mm inner diameter, 49 mm outer diameter, 7 cm long). In 3D-printed variants, the back of the tube is sealed and a trimmed absorbent underpad (Fisher Scientific, Hampton, NH) is affixed to the bottom. For the air puff (Figure 2), habituation (Figure 4) and aversive learning experiments (Figure 5) the tube included a 1 cm wide extension spanning approximately 1/3 of the tube’s bottom circumference. In all designs, a 4 cm long, 0.5 mm diameter wooden rod is adhered to the front tip of the tube, 1 cm from the bottom, in order for animals to grip and rest their forelimbs. The diversity in tube material and geometry reflects the fact that assay design evolved concurrent with the sequence of experiments described in this manuscript. The final design of the tube (mark 21) can be consulted in the CAD folder on the GitHub repository: 3D-printed, including 1 cm extension (Fink and Schoonover, 2018; copy archived at https://github.com/elifesciences-publications/VBAcmd).

The virtual burrow is coupled to near-frictionless air bushings at 14 psi input pressure (New Way Air Bearings, Aston, PA), which slide along two rails in order to constrain movement to one dimension (rail design and assembly: T. Tabachnik, ZMBBI Advanced Instrumentation, Columbia University: fabrication: Ronal Tool Company, Inc., York, PA) (Figure 1A, right and Figure 1B). The animal is head-fixed via custom-machined stainless steel headplate holders (G. Johnson, Columbia University) that secure the titanium headplate. The entire VBA apparatus rests atop an adjustable platform (Thorlabs, Newton, NJ) to permit precise translation of the position of the tube with respect to the head. With t head thus secured, the animal’s body rests freely inside the virtual burrow, its forepaws resting on the horizontal bar placed at the burrow’s threshold, its hind limbs gripping the burrow’s interior

A linear actuator (Part number: L12-30-50-12-I, Firgelli Automations, Ferndale, WA), tethered to the virtual burrow with fishing line (0.15 mm diameter nylon tippet, 4.75 pound test, Orvis, Sunderland, VT) constrains how far the animal may ingress into the burrow at any given time (Figure 1B,C). This parameter can be manually or programmatically varied over the course of the experiment. A force sensor (Futek FSH02664 load cell with Futek QSH00602 signal conditioner, Futek, Irvine, CA) reports whether, and how strongly, the animal is pulling against the linear actuator in its effort to ingress. Upon head-fixation in the VBA, mice invariably ingress as far as the linear actuator command position permits (Figure 1C). When the linear actuator retracts the burrow away from the ingress position (egress position, 10–20 mm posterior to ingress position), mice resist the translation, pulling against the tether in an effort to move the burrow back up around their body (Figure 1D). This effort typically generates between 0.4 and 1 N of force, corresponding in some cases to more than three times animal’s own body weight (in grams-force). We have not observed any mice that fail to resist retraction of the virtual burrow.

A laser displacement sensor (Part number: ILD1302-50, Micro-Epsilon, Dorfbach, Germany) is aimed at a flag affixed to the horizontal bar that joins the air bearings in order to measure the linear displacement of the tube along its axis of motion. The readout of the laser displacement sensor yields a continuous, time-dependent, one-dimensional variable. It is this quantity – how far the animal has pulled the virtual burrow around its body – that tracks ingress in response to a given cue.

For all experiments reported here the analog voltage signals from the laser displacement sensor and the force meter were acquired and digitized at 10 kHz using a Cerebus Neural Signal Processor (Blackrock Microsystems, Salt Lake City, UT).

Before each trial the control system pulls the virtual burrow back to the egress position and waits until the force measured by the force meter drops below a user-specified threshold, indicating that the animal has ceased to resist burrow retraction (Figure 1—figure supplement 2). The linear actuator is then advanced to the ingress position, slackening the tether and permitting free movement of the burrow. If the animal spontaneously ingresses prior to stimulus onset, as measured by the laser displacement sensor, the trial is aborted, the burrow is again retracted to the egress position, and the sequence repeats. Once the mouse has maintained the free, egress position without attempting to ingress within a specified duration, and has maintained the standard deviation of the tube position below a user-specified threshold for a specified delay period, the stimulus is delivered. During stimulus presentation, and a set duration following stimulus offset, the control system is switched to open loop, permitting the mouse to pull the burrow up to the ingress position if it wishes.

The burrow position (measured by the laser displacement sensor), burrow force (measured by the force sensor), and the servo position (state of the linear actuator) are analog inputs to a National Instruments card with analog and digital in/out (USB-6008, National Instruments, Austin, TX). The servo position is controlled by the same National Instruments card. (The position and force signals are simultaneously acquired on a separate DAQ, as described in the previous section.)

Prior to testing, naïve mice are head-fixed in the VBA and given 2–10 mn to acclimate to the contingencies in open loop (free movement of the burrow). Without exception, mice maintain the burrow in the ingress position throughout this acclimation period (Figure 1C). Then they are acclimated to the closed loop mode; after an initial period of sustained struggle to maintain the burrow in the ingress position (Figure 1D), mice cease resisting and eventually consent to holding the burrow in the egress position even after the linear actuator has advanced, slackening the tether and granting the mouse control over the burrow. The duration of the closed-loop acclimation period varied across mice in these experiments (1–20 mn) as the experimenters improved upon and acquired experience withthe assay; presently mice do not require more than ~5 mn total acclimation time (typically two mn for open loop and ~3 mn for closed loop). Trial blocks begin once the animal reliably holds the burrow in the egress position for >30 s between spontaneous ingresses. Trial initiation is delayed until after the mouse has held the burrow in the egress position with minimal movement for several seconds so as to ensure that the animal is in a comparable behavioral state prior to each trial.

For experiments examining responses to visual stimuli, nine mice (three per condition) were acclimated to head fixation in the VBA for 3 mn in the open loop configuration. Following a subsequent 10-mn acclimation period with the VBA in the closed loop configuration, the animal was again permitted to freely ingress in open loop for 3–5 mn. The VBA was then returned to the closed loop configuration and once the animal did not spontaneously ingress for periods greater than 30 s (typically after approximately 1–3 mn) visual stimuli were delivered.

The visual stimuli employed were based on those described in De Franceschi et al. (2016). Briefly, the stimuli were presented on a Dell 1707FP 17’ LCD monitor, 1280 × 1024, 60 Hz, elevated 30 cm and centered above the animal’s head. The three stimuli, generated using the Psychophysics Toolbox Version 3 in MATLAB (Mathworks, Natick, MA), consisted of a black disk presented against a gray background: expanding disk (‘loom’), widening from 2° to 50° over 250 msec, holding the 50° disk for 500 msec; contracting disk (‘recede’), diminishing from 50° to 2° over 250 msec, holding the 2° disk for 500 msec; and sweeping disk (‘sweep’), a 5° disk sweeping smoothly across the diagonal of the screen at a rate of 21°/sec. In order to permit synchronization of stimulus timing with burrow position measurement, the software controlling the visual stimulus also controlled a PWM signal (generated by an Arduino Uno, Adafruit, New York, NY; acquired as an analog voltage input digitized at 10 KHz simultaneous to the position and force signals) that encoded the identity and timing of the visual stimuli.

We divided nine mice into three groups of three animals, one group per stimulus type, and presented each mouse only one of the stimulus types in a single session of five stimulus presentations separated by a 10-mn ITI. The data for each stimulus type are pooled across animals for each group.

In all experiments involving odorant stimuli we avoided using molecules known to elicit systematic attraction or aversion in freely moving animals, such as trimethyl-thiazoline, a volatile cue secreted from the anal gland of fox that elicits avoidance and fear responses (Hebb et al., 2004) or 2-phenylethanol, a component of rose oil that elicits attraction (Root et al., 2014).

We used a custom-built olfactometer to deliver odorant stimuli. Briefly, a nose port constructed of polyether ether ketone (PEEK) was placed approximately 1 mm away from the animal’s nose. When no odorant stimulus was given, the port delivered a steady stream of air (one liter per minute, controlled by a mass flow controller, GFCS-010201 from Aalborg, Orangeburg, New York) that had bubbled through a 50 ml glass bottle containing 15 ml dipropylene glycol (DPG, Part number: D215554, Sigma-Aldrich, St. Louis, MO). To deliver an odorant stimulus, a four-way valve (Part number: LSH360T041, NResearch Inc., West Caldwell, NJ) routed the air stream to exhaust, replacing it with a stream of odorized air; the odorant stimulus was switched off by the four-way valve routing the odorized air back to exhaust. Monomolecular odorants (cis-3-Hexen-1-ol, catalog number W256307; (R)-(+)-Limonene, catalog number 183164; Octanal, catalog number O5608; Ethyl trans-3-hexenoate, catalog number W334200, all from Sigma-Aldrich, St. Louis, MO) were dissolved in 15 ml DPG at a concentration of 2% volume/volume for Hexenol, Limonene, and Octanal and 4% volume/volume for Ethyl trans-3-hexenoate in separate 50 ml glass bottles. After passing through the nose port all gas was routed to a photo-ionization detector (miniPID, Aurora Scientific, Aurora, ON, Canada) to permit constant monitoring of odorant concentration. To avoid contamination, all material in contact with the odorized air stream was constructed in either Teflon, Tefzel, or PEEK. The flow of the air and odor streams were equalized before each experiment (using mass flow meter GFMS-010786 from Aalborg, Orangeburg, NY) and the tubing carrying the two streams from the four way valve was set to equal length and impedance to minimize variation in flow rate upon switching between the air and odor streams.

For odor habituation experiments, five mice were head-fixed in the VBA and allowed to acclimate in the open loop configuration for 5 mn. In order to promote all-or-nothing behavioral responses, we reduced the input pressure to the air bushings from the customary 14 psi to 2 psi, thereby requiring the animal to generate greater force to initiate an ingress than under near-frictionless conditions. After acclimation the VBA was set to closed loop for 10–15 mn. Following acclimation to the closed-loop mode, the animal was then presented with odorant stimuli with the VBA in the closed loop mode. A single odorant, Odor 1 was presented 15 times. Then, a second odorant, Odor 2, was introduced and the two odorants were presented 15 times each, pseudo-randomly interleaved within blocks in which each of the two odorants was presented in every block. Finally, a third odorant, Odor 3, was added and all three odorants were presented in 15 final blocks of three trials each. Each odorant stimulus was presented once per block. All odorant stimuli were 8 s in duration and the ITI was 40 s. Limonene, Octanal, and Hexenol were used as odorant stimuli with different animals receiving different odorants for the Odor 1, Odor 2, and Odor three stimuli (Table 1). Because we refrained from using odors known to elicit innate attraction or aversion such as trimethyl-thiazoline, or 2-phenylethanol, it is unlikely that the egress and ingress responses we observed in early trials were due to the odorants having an innately attractive or aversive quality beyond their novelty.

We noted that we did not observe renewed ingress to familiar odorants following presentation of unexpected ones. This absence of dishabituation (Groves and Thompson, 1970; Thompson and Spencer, 1966) may reflect the use of lower input pressure to the air bushings in this experiment, which promotes all-or-nothing responses by increasing friction between the bearings and the rails but also dramatically decreases the sensitivity of the assay.

For the exploration experiments, the VBA control software was set to open loop mode and the fixed odor port was replaced with an odor port coupled to the virtual burrow, such that the animal was required to egress approximately 30 mm from the fully ingressed position to touch the port with its nose. The odorant stimuli (Ethyl trans-3-hexenoate) were 8 s in duration and the ITI was 60–120 s. In the rare instances during which the animal was not ingressed at the conclusion of the ITI, stimulus delivery was delayed until the animal had resumed a fully ingressed position.

Electrophysiology was performed using silicon probes (A1 × 32-Poly3-5mm-25s-177-H32_21 mm, NeuroNexus, Ann Arbor, MI) chronically implanted into anterior piriform cortex (APC). At least one week following headplate attachment, mice were anesthetized with ketamine/xylezine and a craniotomy centered 1,150 µm posterior to the rostral confluence of the dorsal sinuses and 2,250 µm lateral to the midline was performed using a dental drill (Osada Success 40, Osada Electric Company, Tokyo, Japan). The probe was lowered until the dense spiking characteristic of layer 2 of APC was detected, at which point it was cemented in place. Animals were allowed to recover for at least four weeks before any recording took place to allow for the tissue to fully settle around the implanted probe.

Neural signals were acquired simultaneous to behavior using a Cerebus Neural Signal Processor (Blackrock Microsystems, Salt Lake City, UT). Custom written scripts in Matlab were then used to filter and preprocess the data, which were then automatically spike sorted using Kilosort (Pachitariu et al., 2016) and manually curated using phy (https://github.com/kwikteam/phy).

To determine whether responses in the VBA differed across experimental conditions, we asked whether the likelihood of ingress was larger in one condition than another. For the purposes of this test we define an ingress as a maximum change in burrow position greater 2 mm (the results of the statistical tests are robust to the choice of threshold; see Figure 5—figure supplement 2). For each condition we pooled all ingress responses across mice and used a two-proportion z-test with the null hypothesis that the probability of ingress in the tested condition was less than or equal to the probability of ingress in the other condition. In figures, one star (*) indicates p<0.05, two stars (**) indicate p<0.01, and three stars (***) indicate p<0.001.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We are grateful to Tanya Tabachnik for the design and fabrication of the frictionless rail; Hopi E Hoekstra and Caroline K Hu for helpful conversations regarding burrowing behavior, as well as for providing unpublished video of mice in the wild; Ya-tang Li, Markus Meister and Samuel G Solomon for advice on flight-inducing visual stimuli; John P Cunningham for advice on statistical analyses; Gary W Johnson for precision machining; J Andrew Miri for suggesting the use of a frictionless rail; Paul R Stegall and Damiano Zanatto for advice on mechanical engineering; Alberto Hernandez for supplying cardboard tubes; Annegret L Falkner for Friday night science; D Caroline Blanchard, Rui M Costa, Walter M Fischler, John W Krakauer, Eve Marder, and Catherine H Rankin for critical comments on the manuscript; Ethan S Bromberg-Martin, Michael E Goldberg, and Alla Y Karpova for insightful comments; and the 2016 Janelia Junior Scientist Workshop on Neural Circuits and Behavior.

Animal experimentation: All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocols AC-AAAI8650 and AC-AAAT5466 of Columbia University Medical Center. All surgery was performed under Isoflurane anesthesia, and every effort was made to minimize suffering.

© 2019, Fink et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.