A system for automatic recording of social behavior in a free-living wild house mouse population

The study population

Since 2012, we have studied a population of wild house mice in a 72 m
²
barn, situated at the border of a forest near Illnau, Kanton Zürich, Switzerland (Fig. 1; for a detailed description see 17]).

Fig. 1. Our study site in a barn near Zurich, Switzerland. Left the study population of house mice inhabits a 72 m
²
former agricultural building (barn) that is open to dispersal but closed to predators.
Right nest boxes can be opened to monitor reproduction (age and number of pups in litters)

A barn is a natural habitat for a house mouse. House mice in Europe occur in anthropogenic
habitats, commensally with humans, such as grain stores and farm buildings; feral
populations are generally restricted to islands 34], 35].

The barn with the study population was divided into four quarters by aluminum plates,
with holes allowing the passage of mice, and bricks as well as wooden and plastic
barriers providing internal structure and shelters (Fig. 2). Mice could access all parts of the barn, and could leave it under the roof or through
holes in the walls.

Fig. 2. Schematic drawing of the barn. The barn has a floor space of 72 m
²
, and is equipped with 40 artificial nest boxes. Aluminum dividing walls are 75 cm
high, with 11 passages between the four areas. Next to the entrance door into the
barn is a separate area for storage of equipment and for handling of mice—this is
also accessible to the animals. Not shown are the feeding and drinking sites (3 and
5, respectively, per quarter), the position of the Intelliscales (2 per quarter),
and further structuring of the floor with bricks and smaller wooden and plastic barriers
or hides (modified from 17])

The mice nested in 40 artificial nest boxes (ten per quarter; Fig. 2) and were provided with straw as nesting material. The interiors of the nest boxes
were accessible to us, so that offspring could be counted and measured (Fig. 1). Each nest box had openings for two tubes. One was a transparent tunnel through
which mice entered and left a box (it is not necessary to use transparent tubes, but
we found it helpful that we can see inside). The other tube was shorter and opaque,
and was closed with a plug (unplugging those tubes facilitated catching and handling
of mice from nest boxes during our regular population monitoring).

Nest boxes and shelters were monitored weekly for the presence of mice (using handheld
transponder readers) and for new litters. The ages of pups were estimated, and all
litters were measured and sexed shortly before weaning (at 13 days of age, day of
birth of a litter was considered day 1). We also took small tissue samples from the
ear for later genetic analyses. In addition, at approximately 7-week intervals, comprehensive
trapping was conducted to monitor the entire population. Every mouse was weighed,
adult males and females were examined for reproductive state, and those adults lacking
transponders (see below) were tagged. We also monitored the population for remains
of deceased mice. During the last 5 years, the population comprised 250–430 individuals
(we observed seasonal variation, with the lowest numbers during winter; 17]).

Water and food, a 50/50 mixture of oats and commercial rodent food made by Haefliger
AG, were provided ad libitum at twelve feeding trays (three per quarter). We considered
the availability of food within the natural range. The barn itself was free of predators,
but not of parasites, and mice were exposed to predators, including foxes, badgers,
house cats and birds of prey, whenever they exited the barn.

RFID technology used

Male and female mice of minimally 18 g were subcutaneously implanted with PIT tags
of unique radio-frequency identification (RFID). In our study, we used RFID tags from
Euro ID Identifikationssysteme GmbH Co, Germany (trovan
^®
ID 100, 0.1 g weight, 11.5 mm length, 2.1 mm diameter). The transponders provided
a unique 10 digit alpha-numeric code for each mouse, and a means to monitor mice remotely
by transponder readers.

Continuous recording of RFID transponders

In spring 2007, we installed the first automatic transponder reading prototype (NewBehavior
AG, Zurich, Switzerland), which we replaced in December 2012 with the current, more
efficient AniLoc system (FBI Science GmbH, Germany). The system was designed to track
animals entering and leaving nest boxes, or accessing drinking devices.

As mentioned before, mice could only access the artificial nest boxes (cylindrical,
diameter 15 cm, height 15 cm, covered by a tile) through the tunnel (44 mm inner diameter,
6 mm thickness, 250 mm length) that was equipped with two round antennas (distance
between the two antennas is 15–20 cm; Fig. 3). The tunnels were slightly bent (by an angle of 45°) between the two antennas to
allow easier adjustment and to slow down the mice when running into a tunnel.

Fig. 3. Mice access nest boxes through acrylic tunnels. Each tunnel is equipped with 2 antennas
to allow discrimination between an individual entering or leaving a box. Shown here
is (with nest box number 2 as an example) the positioning of the 2 antennas, each
connected to an AniLoc device, with a mouse passing through the “outer” antenna (upper right), and a green light signal indicating proper functioning of the device (lower left)

The antennas registered only mice passing through the tunnel and not tagged mice sitting
outside, directly next to or on top of the antennas (Helmholtz-designed antennas).

Each antenna was coupled with the animal identification system AniLoc (square black
box visible in Fig. 3), and had a unique identification number. The AniLoc device continuously generated
a close-range electromagnetic field within its double coil antenna (666 ?H inductivity,
50 mm diameter; the two Helmholtz coils had a distance of 30 mm). The 125 kHz low-frequency
radio signal emitted by the device activated transponders located in the range of
an antenna. The activated transponder then sent its unique code back to the AniLoc
reader. Once the complete sequence of the transponder was received, the AniLoc system
transmitted the decoded RFID number together with the antenna identification and a
timestamp (in ms) through the CAN-bus that interconnected all antennas and the power
supply of 12 V. We used rodent proof cables (ALMI PREXTHAN-VA 4G1, 5 mm
²
with reinforced steel coating, AlMi GmbH Co. KG, Mülheim, Germany), since mice otherwise
may gnaw at and damage unprotected cables.

Optical indication of proper operation is provided with each AniLoc (LED indicators
at the bottom of the AniLoc box signaling green versus red light, see lower left picture
in Fig. 3).

The AniLoc devices were equipped with two types of auto-calibration. The first was
carried out during power-up and measured for each antenna inductive static objects
in its surrounding area, like metal or other antennas, which influenced the electromagnetic
field. The device automatically adjusted the power of the electromagnetic field and
the sensitivity of the antenna (signal amplification and filters) to an optimum. Such
calibration minimized the problem of interference with nearby metal- or water-based
objects. The second automatic procedure interrupted the operation of an antenna every
20 s for 50 ms, and checked for its correct impedance to detect short circuits or
interruptions. Additionally, that procedure provided a short resting time for the
operation of the electromagnetic field, since otherwise the transmitter was in continuous
operation (unlike a transponder handheld reader or other RFID products).

The CAN-bus loop terminated on both ends with an interface (CAN2USB-interface “IXXAT”,
IXXAT Automation GmbH, Germany) to a laptop computer (located in the entrance area
of the barn, and protected against mice) running the software OLCUS (FBI Science GmbH,
Germany) within the Microsoft operating system “Windows XP”. OLCUS analyzed the data
stream online in conjunction with the topology of the nest boxes and their antennas.
Besides recording the data, OLCUS supervised the proper operation and carried out
basic statistical analysis. All data were continuously registered in a log file stored
in the onsite laptop. Every 24 h, a copy of the file was automatically sent to a server
at the University of Zurich and uploaded into a database.

Furthermore, 8 weighing scales (Intelliscale, FBI Science GmbH, Germany) with a small
freely movable platform (5 cm diameter) were mounted below water bottles equipped
with an antenna below each drinking nipple (Fig. 4). In combination with a PIR (infrared motion sensor), a mouse was detected when moving
onto the platform and was identified by its RFID transponder. The weight was registered
while drinking water. As soon as the mouse left the platform, the scale was automatically
tared (set to zero) to adjust for the weight of any debris or water drops left on
top of the platform before the next measurement. The Intelliscales were connected
to the same CAN-bus and were fully configurable. For financial and maintenance reasons,
we had not equipped all feeding and drinking sites with Intelliscales, and thus collected
body weight data rather opportunistically so far.

Fig. 4. Intelliscale device (FBI Science GmbH, Germany). A mouse drinks from a water bottle,
with its head inside the field of an antenna. It sits on a white, freely movable platform that is connected to a scale in the lower black box. A PIR (infrared motion sensor) is located in the middle of the platform, here
partly covered by the mouse

The four sections of the barn (Fig. 2; with 10 nest boxes and 2 Intelliscales each) could be supplied with power independently
of each other, which had been designed for improved failover and maintenance purposes.

Time resolution of RFID transponder readings

RFID transponder detection speed was 30 ms, so that AniLoc registered a transponder
if a tagged mouse ran through the field of the antenna with a speed up to 1 m/s (3.6 km/h).
Bending of the tunnels, as mentioned before, was intended to slow down the mice when
running in or out of the tunnels (to avoid missing a reading of an RFID transponder).
AniLoc signaled the first event as soon as the transponder entered the electromagnetic
field, which triggered an entry message with RFID code and timestamp, as described
before (for an example see Table 1). If the registered mouse left the electromagnetic field again within 200 ms, or
another tagged mouse entered the same antenna while the previous was still in it,
an exit message was given. If a mouse spent a longer time inside the field of an antenna,
this resulted in a sequence of messages with the same RFID code.

Table 1. Example of raw data output

The technology used could not simultaneously read several transponders (collision
detection). In case a second mouse entered the field of an antenna already occupied
by a tagged conspecific, RFID codes of the two mice were read randomly. The rather
small inner diameter of a tunnel was expected to minimize the probability that two
mice entered the field of an antenna in parallel. Nevertheless, we occasionally observed
that one mouse squeezed itself on top of another mouse already sitting in a tube.

Ethical note

Injection and use of PIT tags as well as any other manipulations require a permit
according to the Swiss Animal Welfare Ordinance (TSchV). When applying with the authorities
for a license to perform animal experiments, methods, practices and reasons for the
animal experiment have to be described in detail and are evaluated. Data collection
of the project described here was approved by the Veterinary Office Zurich, Switzerland
(Kantonales Veterinäramt Zürich, no 151/2010, 56/2013).