https://orcid.org/0000-0002-8363-1777
Abstract
The crested rat, Lophiomys imhausi, is the only mammal known to sequester plant toxins. Found in eastern Africa, this large rodent is thought to defend against predation by coating specialized hairs along its sides with cardenolide toxins from the poison arrow tree, Acokanthera schimperi. To better understand the ecology of this unusual poisonous mammal, we used camera traps, livetrapping, and captive behavioral observations, to study L. imhausi in central Kenya. Although crested rats were rarely detected with camera traps, 25 individuals were caught in live traps, with estimated densities of up to 15 rats/km2 at one of nine trapping sites. Trapping records and behavioral observations suggest that L. imhausi live in male–female pairs, with juveniles that might exhibit delayed dispersal. We observed chewing of A. schimperi and/or anointing in 10 of 22 individuals, confirming the previous poison sequestration observation. We monitored crested rat activity using cameras and found that chewing on A. schimperi and cardenolide exposure had no effect on feeding, movement, or total activity. One crested rat also fed on milkweed (Gomphocarpus physocarpus; Gentaniales: Apocynaceae), but did not anoint with this cardenolide containing plant. This observation, combined with L. imhausi’s selective use of A. schimperi, suggests the potential for use of alternative poison sources. This research provides novel insight into the ecology of L. imhausi, while also suggesting that more field observations, feeding trials, and chemical analyses are needed to understand their behavior and physiology. Furthermore, their complex social interactions, slow life history, and fragmented populations suggest that L. imhausi could be at risk of decline.
Colored like a skunk and thought to be coated in poison, the crested rat, Lophiomys imhausi, has fascinated biologists since its description in 1867 (Milne-Edwards 1867). Known only from eastern Africa, these large rodents are hypothesized to defend against predation using sequestered plant toxins (Kingdon et al. 2011; Happold 2013). As L. imhausi is rarely encountered, most studies either have examined its unusual morphology (e.g., Milne-Edwards 1867; Kingdon 1974; Stoddart 1979; Naumova and Zharova 2003; Kingdon et al. 2011) or documented new localities (e.g., Kock and Künzel 1999; De Beenhouwer et al. 2016). To date, small sample sizes and limited natural history information have made it difficult to assess the ubiquity of L. imhausi’s poison sequestration behavior, and hinder conservation of this unique rodent.
Although long suspected of being poisonous (Goldfinch 1923; Kingdon 1974), L. imhausi’s potential poison source was only identified recently as A. schimperi (Gentianales: Apocyanaceae—Kingdon et al. 2011). Commonly known as the “African poison arrow tree,” this plant is a major component in African arrow poisons due to its high concentrations of cardenolides, including ouabain and acovenoside (Neuwinger 1996). When offered bark, roots, leaves, and green fruit, from A. schimperi, one captive L. imhausi was observed to repeatedly chew bark and apply the masticated material into specialized hairs along its flanks (Kingdon et al. 2011). In addition, spectroscopic analysis on those specialized hairs from a museum specimen strongly suggested that they contained ouabain (Boulet-Audet and Holland 2011; Kingdon et al. 2011). Cardenolides from A. schimperi, like those from milkweeds (e.g., Asclepias spp. and Gomphocarpus spp.) and foxglove (Digitalis spp.), are potent sodium pump inhibitors. At minute concentrations (e.g., ≤ 5 µg/kg/day administered orally), cardenolide drugs such as digoxin can treat congestive heart failure and atrial fibrillation by increasing heart contractile force (Medscape 2020). However, larger exposures induce vomiting, dysrhythmia, convulsions, difficulty breathing, and cardiac arrest (Cheeke 1989; Neuwinger 1996). Most mammals are highly susceptible to cardenolides; however, muroid rodents are more resistant due to sodium-potassium pump modifications that reduce cardenolide binding (Ujvari et al. 2015). Although L. imhausi’s sodium pump structure is unknown, as muroid rodents (Steppan and Schenk 2017), these rats are expected to survive exposures that would be fatal to many mammalian predators.
Interactions between L. imhausi and its natural predators have never been observed directly. However, observed interactions with dogs and humans provide insight into L. imhausi’s antipredator behavior (Kingdon et al. 2011). When threatened, L. imhausi parts the fur along its flanks, revealing both warning coloration and poison-coated specialized hairs (see videos in Weinstein et al. 2020). Oral contact with these hairs can be fatal. Dogs have died following attacks; those animals that have subsequently recovered avoid L. imhausi (Kingdon et al. 2011). For L. imhausi, aposematic coloration and sequestered toxins likely are effective against potential terrestrial predators that attack with their mouths (e.g., honey badger, Mellivora capensis; black-backed jackal, Canis mesomelas; wild dog, Lycaon pictus; caracal, Caracal caracal; serval, Leptailurus serval; hyena, Crocuta crocuta or Hyaena hyaena; and leopard, Panthera pardus). However, L. imhausi carcasses found in eagle owl nests (Bubo capensis—Ogada 2018) suggest that these defenses likely are less effective against avian predators.
Lophiomys imhausi currently is categorized by the IUCN as a species of least concern (Schlitter 2016). However, its actual conservation status is unclear because individuals of this species rarely are seen (Happold 2013), making estimation of population size difficult. Lophiomys imhausi is the sole living representative of Lophiomyinae, a rodent subfamily with a prehistoric presence throughout Africa, southern Europe, and possibly southwestern Asia (Carleton and Musser 2013; Steppan and Schenk 2017). Although its historical range likely was larger, the current range of L. imhausi is fragmented and restricted to wooded areas in Djibouti, Eritrea, Ethiopia, Kenya, Somalia, Sudan, Tanzania, and Uganda (Fig. 1 inset; Kock and Künzel 1999; Happold 2013). In these regions, 30–70% of forests have been converted to human use (Riggio et al. 2019) and even protected areas often are threatened by logging and fuelwood collection (e.g., KFS 2019). Because L. imhausi often den in tree cavities (Goldfinch 1923; Kingdon 1974), habitat modifications and widespread forest loss likely impact their populations.
The range of Lophiomys imhausi is fragmented and restricted to eastern Africa (inset map, adapted from Happold 2013 and Kock and Künzel 1999). Camera and tomahawk traps were set at 10 sites northwest of Mount Kenya. For each site, unique L. imhausi captures are listed in the numerator and total trap-nights in the denominator for tomahawk (“Traps”) and camera (“Cams”) trapping.
Although many rodent species persist in degraded and human-modified landscapes (Kingdon 2015), the little that is known about L. imhausi suggests that its life history differs from that of most rodents. These large (~1 kg) rats are long-lived, with individuals surviving over 7 years in captivity (Jones 1977). Females typically have one to two offspring per litter, suggesting a low reproductive rate (Anonymous 1933; Ellerman 1940; Delany 1975), and their population density is thought to be low (Happold 2013; Ogada 2018). In addition, although L. imhausi typically are seen alone, suggesting that they might be solitary (Ogada 2018), observed pairs and females with young suggest more complex social interactions (Kingdon 1974; Kock and Künzel 1999; Butynski and de Yong 2016). Because life history traits impact species resilience (McKinney 1997; Brashares 2003), more information on L. imhausi natural history is needed to assess its conservation needs.
To further our understanding of L. imhausi biology, we studied crested rats at 10 sites in central Kenya over an 8-month period. We used camera and live traps to survey wild populations and examined L. imhausi behavior in captivity. Field observations suggested that L. imhausi live in male and female pairs, and we observed paired behavior in captivity, subsequently testing whether sex and paired status influenced activity. We also provided animals with natural cardenolide sources to confirm toxin sequestration behavior and test whether this exposure altered behavior.
Materials and Methods
Study area.
We set camera and live traps traps at 10 sites in Laikipia and Nyeri counties, Kenya, between March and October, 2018. This equatorial region is characterized by bimodal annual rainfall with long (April–June) and short (October–December) rainy periods. Trapping spanned both wet and dry seasons and occurred primarily in wooded and riparian habitat between 1,880 and 2,150 m in elevation (Fig. 1). We selected trap sites based on evidence of presence of L. imhausi and landowner permission, noting that A. schimperi was abundant at all but the two highest elevation sites.
Camera trapping.
We set 35 camera traps (model A-30, Moultrie Feeders, Birmingham, Alabama) at six sites between March and September, 2018 (Fig. 1). The first 30 cameras were set at sites 1–4 prior to or concurrent with livetrapping. At these sites, we set four to eight camera traps at a time, placing them opportunistically throughout the habitat, preferentially near tree cavities, rocky outcroppings, or other potential den sites. We initially set camera traps to confirm the presence of L. imhausi prior to setting live traps; however, we stopped setting cameras first after we rapidly detected more L. imhausi with live traps than cameras at site 4. At sites 8 and 10, we set an additional five cameras immediately following capture and release of L. imhausi, placing cameras within 2–50 m of release sites. For all sites, we positioned cameras on trees, 0.5–1 m above the ground and angled slightly downward. Cameras were configured to take three photos per trigger and had a passive infrared (PIR) motion sensor, < 0.7-s trigger time, and infrared flash. We placed bait in front of cameras at all sites except at site 10. Bait consisted of a combination of carrot, cabbage, corn, dried fish, and a mixture of oats, peanut-butter, sardines, banana, and vanilla. We reviewed all images and identified mammals using Kingdon (2015). We scored observed animals as unique individuals if at least 5 min had passed since the last record of that species or the individual was clearly different from conspecifics seen in the previous 5 min. Three cameras were stolen or damaged, and we exclude these sets from subsequent analyses and trap effort calculations.
Livetrapping.
We set Tomahawk traps (models 204 and 205, Tomahawk Live Trap, Hazelhurst, Wisconsin) at nine sites between March and October, 2018 (Fig. 1). Traps were spaced 20–100 m apart, and set preferentially near tree cavities, rocky outcroppings, or other potential den sites when available. We baited traps with the same mixture used for cameras, rebaiting as needed. We closed traps during the day (from approximately 0800–1700 h). Depending on site, traplines varied in length from 12 to 40 traps. Once set, traps were left in place for 4–9 days, for a total of 1,550 trap-nights.
We determined the sex, weighed, and ear tagged, each captured L. imhausi, and collected feces, tissues, and hair samples, for future analyses. Initial hair samples were collected by holding animals at one end of the trap using a trap divider and clipping hair through the trap. Outside of traps, animals were handled with leather gloves, but typically were fairly docile and slow moving. Because extensive contact with the neck and flanks caused specialized hairs to be lost, we typically handled individuals by the base of the tail to move animals (e.g., into a bucket to be weighed).
We measured body surface temperatures using a FLIR C2 thermal camera (FLIR Systems, Wilsonville, Oregon). With this camera, we captured thermal images of crested rats with and without their specialized hairs displayed. Surface temperature at displayed specialized hairs and at other body regions was measured during the daytime, on unrestrained individuals housed in the indoor enclosures described below.
After measurements and behavioral observations, most animals were released at their capture sites. Videos of released animals climbing, digging, displaying, and exhibiting other behaviors are available online at the Smithsonian Institution Figshare research repository (https://doi.org/10.25573/data.c.4948773—Weinstein et al. 2020). Animal trapping and handling followed ASM guidelines (Sikes et al. 2016) and was approved by the Smithsonian National Zoological Park’s Animal Care and Use Committee (ACUC protocol #18-01) and Kenya Wildlife Service (Capture Permit KWS/904).
Behavioral observations.
We held a subset of trapped L. imhausi in captivity temporarily to observe their behavior. For short observation periods (typically 1–3 days), we placed animals in large plastic tubs (102 liters, 78 × 52 × 36.5 cm) secured with a metal screen lid. Tubs were lined with hay and provisioned with a wool-lined nest box, food (carrots, cabbage, lettuce, commercial rabbit chow), and water. For longer observations (typically more than 4 days), we held crested rats in large indoor enclosures (75 × 230 × 130 cm). Enclosures were located at site 1, and animals were held an average of 18 ± 22 days. Enclosures were lit with natural light from screened windows on two sides and animals were kept under a natural temperature regime and light cycle (approximately 12:12 h light:dark). Each enclosure was separated from others by wire screen and partially lined with hay. We also provided crested rats with at least one large wool-lined nest box, food (as described above), and water.
To record crested rat behavior, we placed cameras (Moultrie A30) inside enclosures. We monitored the behavior of 10 individual crested rats for a total of 447 daytime and 525 nighttime monitoring hours. We expected crested rats to be solitary, and initially placed the first two observed animals (a male and female from the same trap location) in adjacent enclosures. After observing their interactions for one night, we housed them together, yielding both individual and paired observations for these two individuals. Ultimately, we monitored three male and female pairs (432 h of monitoring) and six individually housed crested rats (546 h). During monitoring periods, cameras were set to record a 15-s video when triggered. We trialed multiple camera positions and found that front-mounted cameras detected more activity than those mounted on the ceiling and were the easiest set with which to differentiate individuals and behaviors. For paired animals, we identified individuals based on size, markings, and ear tag location. We analyzed videos and recorded each activity, its duration to the nearest second, and our confidence in our identification of that behavior. For each activity, we also recorded whether crested rats were displaying poison hairs and, for paired rats, whether rats were together (less than 15 cm apart) or following the other rat. We considered rats to be active when in the frame and outside of the nest box: however, activity rates are minimum estimates due to trigger delays, undetected activity in the nest box, rats climbing out of view, and occasional camera malfunctions. Representative examples of observed L. imhausi behavior are available online (Weinstein et al. 2020).
We examined activity patterns and factors influencing activity rates. For each crested rat, we calculated the duration and number of nocturnal activity periods, considering activity periods to be distinct when separated by at least 20 min without observed activity. Using data only from cameras that recorded a full night of activity (60 nights from 10 rats), we tested for a relationship between the number and duration of nightly activity periods (including rat ID as a random factor). We also calculated average active time per night by summing the duration of each night’s active periods, first calculating an average for each individual before calculating the average across all individuals. We report this value, and all other averages, as mean ± SD.
To examine how sex and paired status altered activity patterns, for each camera set (n = 39), we calculated diurnal and nocturnal activity rates as seconds of diurnal (or nocturnal) activity per minute of diurnal (or nocturnal) camera monitoring. We then tested whether sex or paired status altered diurnal and nocturnal activity rates, including camera position as a fixed effect and rat ID as a random factor. Paired rats frequently followed each other, and we tested whether this behavior differed between sexes or increased on nights when mating was observed. We standardized observations of this following behavior by the duration of each camera set and included camera position and pair ID as a fixed effect, and camera set as a random factor. We constructed all mixed effects models with the nlme package version 3.1 (Pinheiro et al. 2020), using backward model selection to select optimal models using likelihood ratio tests. We carried out all statistical analyses in R version 3.6.1 (R Core Team 2019) and inspected residual plots for deviations from homoscedasticity or normality. We report the full initial models and parameters retained in optimal models in Table 1.
Mixed effects model results, including coefficient estimate (β), standard error (SE), t-value, and P-value. For each model, the full initial model is listed as the header, followed by the factors retained in the best model, with significant factors shown in bold.
| Model | Factor | β | SE | t-value | P-value |
|---|---|---|---|---|---|
| M1: Diurnal activity ~ sex + camera position + paired + 1|Rat ID | |||||
| Paired (single) | 1.52 | 0.43 | 3.5 | 0.001 | |
| Camera (front) | 1.20 | 0.57 | 2.13 | 0.04 | |
| M2: Number of active periods ~ active period duration + 1|Rat ID | |||||
| Duration | −5.0 | 0.65 | −7.71 | <0.0005 | |
| M3: Nocturnal activity ~ sex + camera position + paired + 1|Rat ID | |||||
| Camera (front) | 1.16 | 0.46 | 2.54 | 0.015 | |
| M4: Following rate ~ sex + mating + camera position + pair ID + 1|Set ID | |||||
| Sex | 0.79 | 0.38 | 2.06 | 0.057 | |
| M5: Total activity ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model | |||||
| M6: Feeding ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model | |||||
| M7: Movement ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model |
| Model | Factor | β | SE | t-value | P-value |
|---|---|---|---|---|---|
| M1: Diurnal activity ~ sex + camera position + paired + 1|Rat ID | |||||
| Paired (single) | 1.52 | 0.43 | 3.5 | 0.001 | |
| Camera (front) | 1.20 | 0.57 | 2.13 | 0.04 | |
| M2: Number of active periods ~ active period duration + 1|Rat ID | |||||
| Duration | −5.0 | 0.65 | −7.71 | <0.0005 | |
| M3: Nocturnal activity ~ sex + camera position + paired + 1|Rat ID | |||||
| Camera (front) | 1.16 | 0.46 | 2.54 | 0.015 | |
| M4: Following rate ~ sex + mating + camera position + pair ID + 1|Set ID | |||||
| Sex | 0.79 | 0.38 | 2.06 | 0.057 | |
| M5: Total activity ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model | |||||
| M6: Feeding ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model | |||||
| M7: Movement ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model |
Mixed effects model results, including coefficient estimate (β), standard error (SE), t-value, and P-value. For each model, the full initial model is listed as the header, followed by the factors retained in the best model, with significant factors shown in bold.
| Model | Factor | β | SE | t-value | P-value |
|---|---|---|---|---|---|
| M1: Diurnal activity ~ sex + camera position + paired + 1|Rat ID | |||||
| Paired (single) | 1.52 | 0.43 | 3.5 | 0.001 | |
| Camera (front) | 1.20 | 0.57 | 2.13 | 0.04 | |
| M2: Number of active periods ~ active period duration + 1|Rat ID | |||||
| Duration | −5.0 | 0.65 | −7.71 | <0.0005 | |
| M3: Nocturnal activity ~ sex + camera position + paired + 1|Rat ID | |||||
| Camera (front) | 1.16 | 0.46 | 2.54 | 0.015 | |
| M4: Following rate ~ sex + mating + camera position + pair ID + 1|Set ID | |||||
| Sex | 0.79 | 0.38 | 2.06 | 0.057 | |
| M5: Total activity ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model | |||||
| M6: Feeding ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model | |||||
| M7: Movement ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model |
| Model | Factor | β | SE | t-value | P-value |
|---|---|---|---|---|---|
| M1: Diurnal activity ~ sex + camera position + paired + 1|Rat ID | |||||
| Paired (single) | 1.52 | 0.43 | 3.5 | 0.001 | |
| Camera (front) | 1.20 | 0.57 | 2.13 | 0.04 | |
| M2: Number of active periods ~ active period duration + 1|Rat ID | |||||
| Duration | −5.0 | 0.65 | −7.71 | <0.0005 | |
| M3: Nocturnal activity ~ sex + camera position + paired + 1|Rat ID | |||||
| Camera (front) | 1.16 | 0.46 | 2.54 | 0.015 | |
| M4: Following rate ~ sex + mating + camera position + pair ID + 1|Set ID | |||||
| Sex | 0.79 | 0.38 | 2.06 | 0.057 | |
| M5: Total activity ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model | |||||
| M6: Feeding ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model | |||||
| M7: Movement ~ Time interval (pre-, post-, or matched-chew) + 1|Chew event | |||||
| Not significantly different from null model |
Cardenolide interactions.
To confirm poison sequestration and examine cardenolide exposure impacts, we provided crested rats (n = 22) with freshly collected A. schimperi branches (n = 45). Most branches had leaves and some had fruit or flowers, depending on the season. We also tested whether crested rats (n = 2) would use cardenolide containing milkweed (Gomphocarpus physocarpus) as an alternative poison source. For both A. schimperi and G. physocarpus trials, we either observed crested rats directly or monitored their behavior overnight with cameras set and analyzed as described for general behavior observations.
We used video data to test whether cardenolide exposure altered crested rat behavior. If L. imhausi experience acute toxicity following cardenolide exposure, we expect activity after chewing to differ from typical activity. To test for behavioral changes, for each recorded A. schimperi chewing event (chewing lasting at least 2 s with greater than 70% certainty), we summarized the hour of activity immediately preceding chewing (pre-chew), immediately after chewing (post-chew), and in a matched post-chew time interval on a day with no chewing (post-matched). For matched time intervals, we selected the closest nonchewing date with a camera set in the same position as on the chewing night. Using these data, we tested whether duration of total activity, feeding, or movement (walking and climbing), differed between pre-chew, post-chew, and post-matched, time intervals using mixed effects models with chewing event as a random factor (Table 1), testing models as previously described.
Results
Camera trapping.
Cameras produced 3,160 vertebrate wildlife images from 441 days of trapping. Individual cameras recorded for an average of 14 days (SD ± 6) and captured an average of 7.1 (± 5.6) individual vertebrates per day. In addition to the target species, L. imhausi, we detected at least 32 other mammal species, representing 715 unique individual records. Observed species included small mammals such as Atelerix albiventris (Eulipotyphla: Erinaceidae), Grammomys dolichurus (Rodentia: Muridae), Graphiurus murinus (Rodentia: Gliridae), and Cricetomys gambianus (Rodentia: Nesomyidae; see Supplementary Data SD1 for complete camera trapping records). We observed L. imhausi four times on three cameras: once concurrent with livetrapping and three times immediately following the release of livetrapped animals (Fig. 2). The four L. imhausi observations occurred between 1900 and 0400 h with no individuals observed during daylight hours.
Lophiomys imhausi from a camera trap at site 8, captured within 24 h of two rats being released at that location. See Supplementary Data SD1 for additional L. imhausi camera trap images.
Livetrapping.
We trapped 25 individual L. imhausi, with traps open an average of 3.5 nights before capturing a crested rat. We typically removed traps when releasing animals, but at sites where we continued trapping after releasing animals, we recaptured three individuals. Although we did not design surveys to estimate L. imhausi density, data from more intensively trapped sites provide rough density estimates. We captured nine crested rats in a 2-km2 forest strip along the Timau river (site 4), and seven in a 0.4-km forested patch along the Burguret River (site 8). Calculating density as the number of crested rats divided by patch area suggests a minimum of 4 and 15 crested rats per km2, respectively, at these sites. We also noted that two captured crested rats (a male from site 4 and female from site 6) had wounds on the head, body, and specialized hair regions, suggestive of recent predator encounters.
Behavioral observations.
Male (n = 12) and female (n = 13) crested rats differed both in behavior and size. The smallest captured crested rats were two immature females weighing 520 and 565 g, and a 600 g female that we also suspect was immature. Adult females were 12% larger than adult males (adult female: 891 g ± 93, n = 10; adult male: 791 g ± 92, n = 12, t19.3 = 2.5, P = 0.02). Females also were more aggressive and were more likely to hiss, bite, display specialized hairs, and scratch their noses, while in traps. When handled or placed near another male, males often produced high-pitched cries that we did not hear from females (“squeak.mp3” in Supplementary Data SD2). We also noted that crested rats differed in the amount of white on their tails, and although we used this to distinguish individuals in paired video analyses, we did not systematically record pelage information.
We repeatedly captured adult male and female crested rats at the same trap location, suggesting that adults occupy overlapping ranges. In total, 36% of crested rats were captured at trap locations that had recently captured another individual, including a reproductively active adult male (820 g, scrotal) and an adult female (960 g) caught together in the same trap. For traps replaced after capturing one L. imhausi (n = 12), two caught one more crested rat, and three caught two additional individuals. We caught one group of three in July (a 950 g adult female, 700 g adult scrotal male, followed by a 565 g immature female) and another in August (a 930 g adult female with enlarged nipples, 850 g adult scrotal male, followed by a 520 g immature female; Fig. 3). In these trios, we trapped females first; however, in two other pairs we caught males before females. When we placed male and female trap-pairs (n = 3) together either in larger enclosures or smaller observation tubs, rats produced a purring sound (“purr.mp3” in Supplementary Data SD2), groomed each other, and used the nest box together. We also observed similar behavior in a male/female/juvenile trio (Fig. 3; see Weinstein et al. 2020 for videos). Although most animals caught at the same trap location either were male and female pairs (n = 3) or male, female, and juvenile, groups (n = 2), we also caught one male in a trap that had already caught an adult male and female, and another male in a trap that had caught an adult male 3 days prior.
An immature Lophiomys imhausi displaying specialized hairs and warning coloration (A) and the same individual (center) with an adult male (top right), and female (top left), from the same trap location. The juvenile is being groomed by the male (B).
Video monitoring of 10 captive animals confirmed that L. imhausi are primarily nocturnal (Fig. 4A). However, approximately 3.5% of recorded activity did occur during the 447 monitored daylight hours, with single crested rats significantly more active during the day (Table 1, M1). Crested rats typically emerged from nest boxes between 1800 and 2300 h (average 1947 h), and then moved in and out of the box throughout the night (Fig. 4B). Between sunset and sunrise, crested rats averaged 6.1 ± 2.4 active periods, each lasting an average of 23 ± 17 min. Crested rats were active an average of 128 ± 77 min per night, with a negative correlation between the number and duration of active periods (Table 1, M2). Nocturnal activity rates did not differ between male and female or single and paired individuals, but were influenced by camera position, with front-positioned cameras recording significantly more activity than ceiling-mounted cameras (Table 1, M3).
Captive Lophiomys imhausi were most active between sunset and sunrise (A). Crested rats, housed singly and in pairs, alternated between inactive and active periods each night. Here, a single full night of activity is shown for three pairs and three single individuals (B). When active, rats spent most of their time feeding, walking, and standing (C). Each bar represents one of the 10 observed captive crested rats and displays the percentage of active time spent in each activity. “Other” includes unidentified activity, urinating, defecating, fighting, and other rarely observed behaviors.
Crested rats moved around the enclosures and exhibited a variety of behaviors including feeding, grooming, drinking, mating, and watching animals in adjacent enclosures (Figs. 4C and 5; videos available in Weinstein et al. 2020). Crested rats spent 40 ± 12% of active time feeding, and typically ate food where it had been placed in the middle of the enclosure. One female repeatedly carried carrots and cabbage into the nest box; however, nest box checks showed no evidence of caching. We also observed two females collecting hay. Grooming behavior included both chewing and scratching, and although rats typically groomed themselves (Fig. 5C), we also observed paired males and females occasionally grooming the other individual in their pair. We rarely observed urination or defecation because rats typically deposited feces at the front of the enclosure next to the camera and away from their food and nest box.
Captive Lophiomys imhausi exhibited a variety of behaviors including following each other (A), feeding together (B), and grooming (C). Males often bit the female’s back to hold on during mounting (D). After mating (E), animals occasionally nipped or swatted at each other (F). See Weinstein et al. (2020) for videos of L. imhausi behavior.
Male, female, single, and paired, rats all occasionally displayed their specialized hairs; however, these hairs were visible during less than 0.5% of active time. When rats parted their fur for these displays, they produced a stronger heat signature (Fig. 6). Crested rat surface temperatures (~30°C) in the sparsely furred specialized hair strips were ~10°C warmer than the rest of the body surface. In more densely furred body regions, surface temperatures were only 3–4°C warmer than ambient temperatures (~16–17°C) in the enclosure at the time of measurement.
Lophiomys imhausi produced different heat signatures when displaying (A) and not displaying specialized hairs (B). Surface temperatures (top left of image) in densely furred body regions (measured in white circle) were approximately 10°C cooler than the face and exposed specialized hair regions.
Paired individuals spent 55% of observed active time within approximately 15 cm of the other rat. In all pairs, males and females followed each other through the enclosure (Fig. 5A). Males followed females more frequently (68% of following observations) than females followed males (32% of observations); however, neither sex nor mating behavior significantly influenced following rates (Table 1, M4).
All three pairs mated during the July through September observation period, with matings observed in July, August, and September. Paired crested rats mated on six of 25 observation nights, with mating seen 1–4 times per night when it occurred. Mating often was initiated by the male nudging the female, circling around her, and then often biting and holding onto her back before mounting (Figs. 5D and 5E; see videos in Weinstein et al. 2020). When mating occurred, mounting typically lasted less than 15 s and, in total, mating-associated behaviors typically lasted less than a minute, and always less than five minutes. We also occasionally observed males unsuccessfully attempt to initiate mating and noted that approximately one-third of successful matings ended with a brief, less than 15 s, aggressive response in which females nipped or swatted at males (Fig. 5F).
We observed aggressive behavior (e.g., nipping, swatting) within all pairs on nights with and without mating; however, this behavior was rare. Aggressive interactions typically lasted less than 10 s, with animals immediately resuming normal activities, often together. We did not observe aggression between individuals in adjacent enclosures, although adjacent crested rats often watched and interacted with each other. However, placing a new male in an enclosure, near a nest box site with a paired male and female, did result in aggression (e.g., hissing, attempted bites) and distress calls.
Cardenolide interactions.
Crested rats did not interact with A. schimperi or G. physocarpus each time the plants were offered. Animals often ignored A. schimperi, but also sniffed the plant, ate the flowers, and nibbled branches or leaves without removing a noticeable amount of plant material (see images and videos in Weinstein et al. 2020). One L. imhausi ate more than 5 g (dry weight) of milkweed leaves and stems without exhibiting any anointing behavior or behavioral changes indicative of toxicity.
We observed 17 chewing events on A. schimperi from 10 different male and female crested rats. Three instances were based on chew marks on branches left overnight with crested rats (see example, chewed_branch.jpg, in Weinstein et al. 2020), three were directly observed, and 11 were captured using cameras. All but one of the chewing events recorded by the cameras occurred at night, with some individuals chewing multiple times in the same night. Paired crested rats chewed both separately and together; however, we only observed animals anointing themselves.
Sequestration took up to 10 min with crested rats alternating between chewing and anointing. To anoint, crested rats stripped bark or chewed leaves (Fig. 7A), chewing the material before applying it to their specialized hair tracks using motions similar to grooming (Fig. 7B, see Anointing.mp4, in Weinstein et al. 2020). Although anointing motions were similar to some movements seen in grooming, grooming was characterized by short bouts of chewing and scratching directed at multiple body regions (see grooming videos in Weinstein et al. 2020). In contrast, anointing occurred in concert with Acokanthera chewing and involved extensive application of material to the specialized hair tracks along the animal’s neck and sides. We observed no obvious behavioral changes following Acokanthera chewing and no significant difference in total activity, feeding, or movement between pre-, post-, and matched-chew, intervals (Table 1, M5–7; Fig. 7C).
Lophiomys imhausi chewing Acokanthera schimperi (A) and anointing after chewing (B). Chewing on A. schimperi did not alter activity. There was no difference in activity rates measured before (“pre”) or after (“post”) chewing, or in a matched (“matched”) nonchewing time interval on another date (C). See Weinstein et al. (2020) for videos of L. imhausi anointing.
Discussion
Crested rats possess a suite of unusual traits that render them both fascinating and challenging to study. Here we expand upon previous observations on sequestration of A. schimperi and demonstrate that this toxin exposure has no impact on crested rat behavior. Behavioral observations further suggest that adults live in male and female pairs with juveniles that might exhibit delayed dispersal. Complex social behavior, combined with a slow life history, fragmented populations, and habitat loss, potentially put L. imhausi at conservation risk, and low camera trap detection rates make monitoring difficult.
Camera traps are an efficient and noninvasive tool for mammal surveys (Wearn and Glover-Kapfer 2019), but appear to underestimate abundance of L. imhausi. In complex forest environments, slow movements and low surface temperatures likely reduce L. imhausi detection in cameras with PIR sensors. These sensors are triggered by rapid changes in temperature (Welbourne et al. 2016), and small body size and temperature differences less than 4–5°C reduce trigger rates (Glen et al. 2013; Welbourne 2014; Lerone et al. 2015; Anile and Devillard 2016). In addition to potential camera trigger issues, L. imhausi only are active for short periods during the night and might have relatively small home ranges, further reducing detection rates if cameras are set to accommodate large mammal home ranges (Meek et al. 2014). Because L. imhausi likely are poorly detected in camera traps set for larger species but too large to detect in live traps set for smaller mammals, mapping their range and habitat preferences will require carefully designed and targeted surveys.
Livetrapping revealed high densities of L. imhausi in riparian forests, in what is likely a suitable habitat for these large herbivorous rodents. These densities (4–15 per km2) are higher than previous estimates (1 per km2) in more degraded agricultural habitats. As previous estimates were calculated from carcasses associated with owl nests (Ogada 2018), differences also could be due to sampling methodology. Alternatively, agricultural landscapes might support smaller populations or act as sinks for adjacent forests. Crested rats occur both in pristine and human-modified landscapes and as forests continue to decline, understanding population dynamics across habitat types will improve conservation and management plans.
Trapping records and behavioral observations from this study suggest that L. imhausi are more social than previously thought, and form family groups or pairs. Although monogamy is rare in mammals (Kleiman 1977), it does occur in rodents such as the California mouse (Peromyscus californicus—Ribble and Marco 1990), prairie vole (Microtus ochrogaster—Getz et al. 1993), and crested porcupine (Hystrix cristata—Mori et al. 2014). Lophiomys imhausi has many traits often associated with monogamy, including large size, long life span, low reproductive rate, cavity dwelling, and larger, more aggressive, females (Kleiman 1977). In other mammals, monogamy also often is associated with long maturation periods (Kleiman 1977). Information on fetal development in L. imhausi is limited (Delany 1975), but trapping patterns suggest that juveniles remain with adults for an extended period, possibly akin to the delayed dispersal seen in the monogamous Malagasy giant jumping rats (Hypogeomys antimena; Rodentia: Nesomyidae—Sommer 1997). Although we suspect that L. imhausi live in male and female pairs, occasionally with offspring, it also is possible that they live in extended family groups or that pairs only are together for the breeding season. In this equatorial region, rodents often breed in both wet seasons (April–June and October–December—Martin 1985). Pregnant L. imhausi have been trapped in March, July, and late August (S. B. Weinstein and B. Agwanda, pers. obs.), and captive crested rats mated July through September, suggesting that breeding and paired behavior span both wet and dry seasons and might occur year-round. Where L. imhausi occur, they are easily trapped (and re-trapped) and future mark–recapture or telemetry studies could offer more insight into breeding behavior, paired interactions, and juvenile dispersal.
Sequestration of A. schimperi was common in captive L. imhausi, but did not occur each time the plant was offered. There are several reasons why crested rats might be selective in their sequestration. Crested rats might only sequester toxins in response to stress and predation threat. However, human presence typically is stressful for wild animals (Sikes et al. 2016), and captured crested rats did not always chew when A. schimperi first was offered. Alternatively, crested rats might apply cardenolides only as needed to maintain baseline poison levels. Cardenolides are stable toxins, with arrow poisons made from A. schimperi remaining potent for decades (Neuwinger 1996). As crested rats rarely expose their specialized hairs, anointed cardenolides also might persist for extended periods. Crested rats also might prefer plants with certain chemical concentrations or compositions, with nibbling and sniffing used to ascertain plant quality. In A. schimperi, toxicity varies between plants and plant parts (Neuwinger 1996). Like other herbivores (e.g., Zalucki et al. 1990; Torregrossa and Dearing 2009), L. imhausi might be able to modulate plant interactions based on plant chemistry. If L. imhausi can detect cardenolides, animals at sites without A. schimperi (e.g., Ogada 2018; M. Sommerlatte, Ragati Conservancy, Kenya, pers. comm.) might be able to use alternative toxin sources. Although L. imhausi did not anoint when offered G. physocarpus, trials were limited to two animals only. Interactions with other toxic plants have not yet been investigated. Additional field observations, captive feeding trials, and chemical analyses of hair samples could provide insight into the cues that trigger sequestration and reveal whether additional plant species are used.
As expected, L. imhausi appears to be resistant to cardenolides. Interactions both with milkweed and A. schimperi produced no behavioral changes associated with acute cardenolide poisoning. However, we note that the vomiting commonly seen in birds and some mammals is not physically possible for rodents (Fink and Brower 1981; Neuwinger 1996; Horn et al. 2013). Although we do not know how much A. schimperi is ingested during the anointing process, the L. imhausi that ate over 5 g of milkweed likely swallowed a substantial quantity of cardenolides. The absence of behavioral changes following this and other exposures suggests that crested rats are resistant to these compounds. Modified sodium pumps likely contribute to resistance (Ujvari et al. 2015), but other factors also might reduce cardenolide toxicity in L. imhausi. For example, in laboratory rats (Rattus norvegicus), cytochrome p450 enzymes metabolize the cardenolide digoxin in the liver (Salphati and Benet 1999) and in humans, p-glycoproteins influence digoxin absorption and secretion in the intestine (Van Asperen et al. 1998). Microbial symbionts also could metabolize cardenolides. Herbivores often host toxin-degrading gut bacteria (Kohl et al. 2014), and one human gut bacterium, Eggerthella lenta, is known to metabolize cardenolides (Haiser et al. 2013). Thought to be a foregut fermenter, L. imhausi has an unusual four-chambered stomach with a dense bacterial community that could contain cardenolide-degrading microbes (Naumova and Zharova 2003). Further research on L. imhausi’s microbiome and genetics could provide insight into its unusual behavior while enhancing our understanding of mammalian cardenolide interactions.
Chemical defenses are rare in mammals (Savitzky et al. 2012), and among mammals, L. imhausi is unique in its ability to sequester cardenolides from A. schimperi. In addition to their unusual defense behavior, these large aposematic rodents also might be monogamous. Unfortunately, their complex social interactions, slow life history, and fragmented populations suggest that L. imhausi populations could be at risk. As forests continue to decline, assessing the conservation status of L. imhausi will require targeted surveys to establish baseline densities and monitor changes.
Supplementary Data
Supplementary data are available at Journal of Mammalogy online.
Supplementary Data SD1.—Summary data for camera traps set to detect Lophiomys imhausi at six sites in central Kenya, including additional camera trap images of L. imhausi, and a satellite imagery map of camera and trapping sites.
Supplementary Data SD2.—Lophiomys imhausi vocalizations: audio files (MP3) of squeaking, purring, and hissing, vocalizations.
Acknowledgments
We thank Peter Lotira, Jackson Mathee, Stephanie Higgins, Samuel Njuki, Zahra Kahn, Esir, Duncan Kago, Enock Kenyati, Charles Maina Wachira, and Joseph Wamai, for assistance in the field and laboratory; Hillary Young, Jake Goheen, Adam Ferguson, and Georgia Titcomb, for research equipment; and Thomas M. Butynski, Yvonne A. de Jong, Malte Sommerlatte, Dino Martins, Scott Miller, Laura Morse, Jennifer Donato, Fardosa Hassan, Fred Omengo, and Thadeus Obari, for valuable advice and permitting assistance. We thank Lolldaiga Hills, Elizabeth Coverdale, Juliet King, Karl Amman, Phillipa Bengough, Trout Tree Restaurant, Mount Kenya Game Reserve, Mount Kenya Safari Club, and Mpala Research Centre, for site access and research support. We also thank Darcy Ogada and Rafael Reyna for comments that improved the quality of this manuscript. This work was permitted through Kenya Wildlife Services (KWS/904, KWS/BRM/5001, KWS/BRP/5001), conducted under NACOSTI/P/18/5880/20631 and NEMA/AGR/104/2018, and approved by the Smithsonian National Zoological Park’s Animal Care and Use Committee (ACUC protocol #18-01). Research was supported by NSF-1601362, NSF-1342615, NIH T32AI055434 from the National Institute of Allergy and Infectious Diseases, National Science Foundation, and a Smithsonian-Mpala Postdoctoral Fellowship. The authors declare no conflicts of interest.
