AutonoMouse: High throughput automated operant conditioning shows progressive behavioural impairment with graded olfactory bulb lesions

Andrew Erskine, Thorsten Bus, Jan T. Herb, Andreas T. SchaeferbioRxiv 2018 · DOI

Abstract

Operant conditioning is a crucial tool in neuroscience research for probing brain function. While molecular, anatomical and even physiological techniques have seen radical increases in throughput, efficiency, and reproducibility in recent years, behavioural tools have seen much less of an improvement. Here we present a fully automated, high-throughput system for self-initiated conditioning of up to 25 group-housed, radio-frequency identification (RFID) tagged mice over periods of several months and >10^6 trials. We validate this “AutonoMouse” system in a series of olfactory behavioural tasks and show that acquired data is comparable to previous semi-manual approaches. Furthermore, we use AutonoMouse to systematically probe the impact of graded olfactory bulb lesions on olfactory behaviour and resolve the long-standing conundrum about the apparent lack of impact of lesions on olfactory abilities. The modular nature and open-source design of AutonoMouse should allow for similar robust and systematic assessments across neuroscience research areas.

Introduction

In neuroscience, molecular and physiological tools have seen radical leaps in throughput and reproducibility. However, behavioral testing—the ultimate readout of brain function—has remained a bottleneck. Standard operant conditioning often requires manual handling (which stresses animals), water restriction (which alters physiological states), and significant experimenter time, limiting the scale and complexity of research questions.

A specific casualty of these limitations is the "olfactory bulb conundrum." For decades, researchers have debated why massive lesions to the olfactory bulb (OB) often result in no apparent behavioral deficit. The authors hypothesized that this was due to a lack of systematic, high-resolution behavioral data across a wide enough parameter space of both lesion size and task complexity.

The AutonoMouse System

The authors developed AutonoMouse, a modular, automated environment consisting of a large home cage connected to a behavioral staging area. This setup allows for high-throughput operant conditioning of up to 25 group-housed mice simultaneously.

Schematic of the AutonoMouse system (a) and the timing diagram of a single behavioral trial (b).

Key technical innovations include:

  • RFID Tracking: Each mouse is tagged with a unique ID chip. When a mouse enters the staging area, an IR beam triggers a door to isolate it, and an RFID coil identifies the individual to deliver a personalized training schedule.
  • No Water Restriction: Instead of dehydrating mice to motivate them, the system provides water ad libitum only through the behavioral task. Mice self-regulate their intake by performing trials whenever they are thirsty.
  • High-Stability Olfactometer: A custom 8-channel air-dilution system ensures that odors are the only salient cues, eliminating non-olfactory artifacts like valve clicks.

Resolving the "Redundancy" Myth

The study generated a massive behavioral dataset, including over 1.3 million trials from 67 animals. This scale revealed that mice are perfectly capable of maintaining their weight and health by "working" for water without being restricted. Data showed that the number of trials performed scales inversely with the reward size, proving that mice regulate their own metabolic demands.

The paper resolves the OB lesion debate by showing that the brain's resilience is task-dependent. The OB appears "redundant" only if testing simple discrimination. Small lesions didn't stop mice from learning new odors, but they severely impaired the recognition of previously learned "templates." This suggests that while the circuitry for discrimination is robust, the storage or retrieval of specific odor identities is highly non-redundant.

Key Results

  • Throughput: Mice performed between 150 and 560 trials per 24 hours (mean ~333). This allowed for the collection of data comparable to months of manual training in just a few days.
  • Circadian Stability: While mice are more active at night, their accuracy remains constant throughout the 24-hour cycle.
  • Lesion Impact:
    • Full Lesions: Produced complete anosmia (inability to smell), though auditory performance remained intact.
  • Partial Lesions: Animals with large lesions could still perform simple binary odor discriminations but failed at mixture discrimination.
  • Minor Lesions: Even the smallest lesions caused significant "forgetting" of familiar odors, though the animals could quickly re-learn them.
Response classification matrix (c) and a typical learning curve (d) showing performance improvement over 600 trials.

Conclusion and Future Work

AutonoMouse represents a significant leap forward in behavioral testing. By removing human handling and water restriction, the system achieves unprecedented data scales and reveals that the "redundancy" of the olfactory bulb is a myth: while simple tasks are robust to damage, complex odor recognition and mixture discrimination are highly sensitive to even minor lesions.

Future versions could integrate respiration monitors to track sniffing patterns or wireless neural recording headers to correlate behavior with real-time brain activity. The modular nature and open-source design of AutonoMouse should allow for similar robust and systematic assessments across various neuroscience research areas.

Summary updated

Abstract

Abstract Operant conditioning is a crucial tool in neuroscience research for probing brain function. While molecular, anatomical and even physiological techniques have seen radical increases in throughput, efficiency, and reproducibility in recent years, behavioural tools have seen much less of an improvement. Here we present a fully automated, high-throughput system for self-initiated conditioning of up to 25 group-housed, radio-frequency identification (RFID) tagged mice over periods of several months and >10^6 trials. We validate this “AutonoMouse” system in a series of olfactory behavioural tasks and show that acquired data is comparable to previous semi-manual approaches. Furthermore, we use AutonoMouse to systematically probe the impact of graded olfactory bulb lesions on olfactory behaviour and resolve the long-standing conundrum about the apparent lack of impact of lesions on olfactory abilities. The modular nature and open-source design of AutonoMouse should allow for similar robust and systematic assessments across neuroscience research areas.
BibTeX
@misc{erskine2018autonomouse,
  author = {Andrew Erskine and Thorsten Bus and Jan T. Herb and Andreas T. Schaefer},
  title = {{AutonoMouse: High throughput automated operant conditioning shows progressive behavioural impairment with graded olfactory bulb lesions}},
  year = {2018},
  doi = {10.1101/291815},
  howpublished = {bioRxiv},
}
Figure