The fleas of house mice (Mus musculus L.) and ship rats (Rattus rattus L.) in forest of the Orongorongo Valley, New Zealand

ABSTRACT

The ectoparasites of introduced rodents in mainland New Zealand forests include several species of cosmopolitan flea that may be important in the population dynamics and future biocontrol of rodents. We describe a 2-rodent, 2-flea system that showed little change over 20 years. Ship rats (Rattus rattus) and house mice (Mus musculus) were snap trapped at fixed sites in the Orongorongo Valley, Wellington, for a study of their population ecology. The fleas Leptopsylla segnis and Nosopsyllus fasciatus were common on mice and rats respectively, and less common on the alternate hosts. Prevalence was described in relation to sex and age of the host, and to season and year. Male mice were more likely than females to carry L. segnis. Prevalence and intensity of infection mostly increased with age of host. Prevalence of both flea species showed modest seasonal variation, with a dip in autumn. Pregnant or lactating female rats and mice were less likely to have fleas than were non-breeding adult females. Prevalence did not vary positively with host density. We detected a slight overall increase in the prevalence of each flea species over the duration of the study. We conclude with some discussion of flea assemblages in New Zealand.

KEYWORDS:

• Fleas
• forest
• house mouse
• Leptopsylla segnis
• long-term study
• Nosopsyllus fasciatus
• Orongorongo Valley
• ship rat
• Siphonaptera

Introduction

Ectoparasites of rodents can be important as vectors or reservoirs of human diseases such as plague. The last recorded case of plague in New Zealand was in 1911 (MacLean 1955), and rodents are now mostly excluded from houses, limiting the spread of zoonotic diseases. However, introduced rodents play a significant and destructive role in New Zealand’s natural ecosystems, and their ectoparasites may in future be critical to the spread of both lethal and non-lethal biocontrol agents. A classic parallel case from Australia concerns myxomatosis and the rabbit flea (Spilopsyllus cuniculi (Dale): Pulicidae) (Sobey et al. 1973). Understanding the ecology of rodent ectoparasites in New Zealand is therefore potentially useful for predicting the transmission of microscopic pathogens that may have a role in rodent control.

The flea fauna of New Zealand is unusual in consisting largely of fleas that are associated with birds (Smit 1979; Pilgrim 1980, 1991). The mammal fleas are cosmopolitan or Australian species associated with mammals introduced to New Zealand, except for an endemic flea (Porribius pacificus Jordan: Ischnopsyllidae) associated with native species of bat. Host-flea associations have been reported by Tenquist and Charleston (1981, 2001). The species of flea on small mammals in various parts of New Zealand have been reported in recent years (King and Moody 1982a; Gibson 1986; Efford et al. 1988; Murphy 1989; Roberts 1991a; Innes et al. 2001). The rodent fleas are from three genera in separate families: Nosopsyllus (Ceratophyllidae), Pygiopsylla (Pygiopsyllidae), and Leptopsylla (Leptopsyllidae).

We conducted a long-term snap-trapping study of rodents (ship rats Rattus rattus L. and house mice Mus musculus L.) in the forest of the Orongorongo Valley. For 20 years all fleas associated with them were collected as part of the autopsy procedures and later identified. Only two species of flea were recorded (Leptopsylla segnis (Schönherr) and Nosopsyllus fasciatus (Bosc.)). The study provided large samples and an opportunity to examine the changes in the flea populations in relation to the changes in the mouse and rat populations (Fitzgerald et al. 2004; Efford et al. 2006). The only other ectoparasite noted was the sarcoptid burrowing mite Notoedres muris (Megnin) on rats. They produce conspicuous lesions on the ears, tail, and scrotum. We did not attempt to include them in this study.

Materials and methods

Study area

The Orongorongo Valley is a steep-sided valley near Wellington, New Zealand, with a Field Station at 41°21'S, 174°58'E. Campbell (1984) described the climate and vegetation. The valley is largely covered in evergreen temperate rainforest, comprising southern beech (mostly Nothofagus truncata (Colenso)), podocarps (Coniferales, especially Dacrydium cupressinum Lamb. and Prumnopitys ferruginea (D. Don)) and hardwood species (Metrosideros robusta A. Cunn., Melicytus ramiflorus J. R. Forst. & G. Forst., and Elaeocarpus dentatus (J. R. Forst. & G. Forst.)). Silver beech Nothofagus menziesii (Hook. f.) is dominant above 600 m.

The mean summer temperature (December–February) at the Field Station (130 m a.s.l.) is 15.9°C and mean winter temperature (June–August) is 7.5°C; frosts are uncommon. The annual rainfall over 30 years averaged 2370 mm, with monthly averages ranging from about 100 mm in January to 275 mm in July. The silver beech forest is cooler and moister than the lower forest.

Rattus rattus and Mus musculus were the only rodent species present; other small mammals were feral cats (Felis catus L.), the mustelids stoat (Mustela erminea L.) and weasel (M. nivalis L.), hedgehogs (Erinaceus europaeus L.), and brushtail possums (Trichosurus vulpecula (Kerr)). Fleas are seldom found on possums in New Zealand, the only record being of a cat flea (Ctenocephalides felis (Bouché)) (Tenquist and Charleston 1981).

Intensive studies in part of the long-term study area provided snapshots of absolute host densities. Between July and December 1977, the density of the mouse population was estimated to decline from 3.3 to 0.55 ha⁻¹ (Fitzgerald et al. 1981). Later, in a live trapping study of R. rattus, their density was estimated as being in the range 5–9 ha⁻¹ (Wilson et al. 2007).

Sampling

Rats and mice were snap-trapped on a ‘long line’ of 116 sites at 50-m intervals through the forest near the valley floor from August 1971 to May 1998 (Fitzgerald and Karl 1979, Figure 2; Efford et al. 2006). Southern beech trees were present at 34 sites, designated ‘beech’ sites, and absent at the 82 ‘podocarp’ sites (Fitzgerald et al. 2004). Rats and mice were also trapped on another line, of 36 trap sites, in silver beech forest (> 600 m a.s.l.) from November 1973 to August 1978. For 20 years, from the beginning of the study until May 1991, all fleas found during the autopsy procedures were preserved and identified.

At each trap site, one rat snap-trap and one mouse snap-trap were set under a metal cover. Traps were set for three consecutive nights in late February, May, August and November (summer, autumn, winter and spring). Rodents were collected from the traps each morning, sealed individually in plastic bags, and autopsied on the same day. Bodies were placed in an insect-killing jar with ethyl acetate for a few minutes to kill any fleas. Fleas were collected from the jar and by systematically searching the fur and stored in 70% alcohol. Each rodent was measured and weighed, and its reproductive state was determined. The skulls were saved for determining tooth wear.

Fleas were later prepared as microscope slides using the technique of Smit (1954), modified slightly by R.L.C. Pilgrim (pers. comm.). Fleas were identified and their sex determined, following Smit (1979). Slides were lodged in the Pilgrim flea collection in Te Papa Tongarewa, Museum of New Zealand, Wellington.

Mice were assigned to Lidicker’s (1966) toothwear classes (see also Murphy and Nathan 2021) and rats to Karnoukhova’s (1972) toothwear classes (see also Innes 2005). Toothwear classes represent unequal time intervals. The Lidicker scale may be calibrated approximately for Orongorongo Valley mice as follows (see Fitzgerald et al. 2004): class 1 = 0–1 month, 2 = 1–2 months, 3 = 2–4 months, 4 = 4–6 months, 5 = 6–8 months, 6 = 8–10 months, 7 = 10–14 months, and 8 = > 14 months. Calibration of the Karnoukhova scale is more problematic owing to the greater longevity of rats and to variation within the small known-age sample reported by Efford et al. (2006 Appendix 1); tentatively: classes 1 & 2 = 0–2 months, 3 = 3–9 months, 4 = 10–18 months, 5–7 = > 18 months.

Statistical analysis

In our analyses ‘prevalence’ refers to the proportion of hosts carrying at least one flea, ‘intensity’ is the number of fleas per infected host, and ‘abundance’ is the number of fleas per host whether infected or not (Bush et al. 1997). Results are given as mean ± SE except as indicated. We used χ² tests of homogeneity and t-tests to indicate the statistical support for simple comparisons. Binomial confidence limits of proportions were calculated using the Wilson score interval with continuity correction (Newcombe 1998). For annual analyses each ‘year’ was defined to include the four samples from November in the preceding calendar year to August in the current calendar year (Fitzgerald et al. 2004; Efford et al. 2006).

Trend in annual prevalence was assessed by fitting generalised linear models that controlled for the effects of host sex and toothwear age (as a linear predictor on the logit scale). Significance was assessed by a likelihood ratio test comparing a model with additive logit-linear year effect to a model with no year effect.

Aggregation was measured by the size parameter (k) of a negative binomial distribution fitted to the numbers of fleas on individual hosts by maximum likelihood (Venables and Ripley 2002). Small values of k imply high aggregation.

The capture rate of the rodent hosts was converted to a linear index of density that allowed for sprung traps and non-target captures (Fitzgerald et al. 2004): $I={\displaystyle \frac{-100a_2}{a_1+a_2}\mathrm{l}\mathrm{o}\mathrm{g}{\displaystyle \frac{a_0}{a_0+a_1+a_2},}}$where a₀ traps were un-sprung, a₁ traps were sprung or caught a non-target species, and a₂ traps caught the target species. At the low capture rates that we mostly observed, this index is numerically close to the percentage trap success index with ad hoc adjustment for sprung traps (e.g. King 1982). We used the median index value of each host species (mice 2.533, rats 3.510) to separate times of ‘low’ and ‘high’ density. We used the function ‘glm.nb’ of Venables and Ripley (2002) to perform a negative binomial regression of flea abundance on host age class and density index, thereby allowing for aggregation of fleas on hosts.

Results

Between August 1971 and May 1991, flea checks were performed on 1150 rats and 1622 mice from the valley-floor long line and, between May 1974 and August 1978, on 15 rats and 155 mice from the higher-altitude trap line in silver beech. On the long line, 49% of mice and 15% of rats carried L. segnis, while 7% of mice and 25% of rats carried N. fasciatus (Table 1). The prevalence of fleas on mice and rats from the silver beech trap line was lower but followed a similar pattern (Table 1). The following results are for mice and rats from the long line except where it is stated otherwise.

Table 1. Fleas on rats (Rattus rattus) and mice (Mus musculus) snap-trapped in the Orongorongo Valley, Wellington, New Zealand 1971–1991.

					Intensity
Host	Trap line	N	Species	Prevalence	Mean	SD	Max.	k (SE)
Mouse	Long line	1622	L. segnis	0.494	3.98	5.29	70	0.35 (0.02)
			N. fasciatus	0.065	1.24	0.58	4	0.19 (0.05)
	Silver beech	155	L. segnis	0.297	4.15	4.73	26	0.17 (0.03)
			N. fasciatus	0.026	1.50	1.00	3	–
Rat	Long line	1150	L. segnis	0.147	2.23	2.71	17	0.12 (0.01)
			N. fasciatus	0.264	2.68	4.121	50	0.19 (0.02)
	Silver beech	15	L. segnis	0.067	3.00	–	3	–
			N. fasciatus	0.267	1.00	–	1	–

Notes: Prevalence is the proportion of hosts with at least one flea. Intensity is the number of fleas per infected host. k is the aggregation parameter of a negative binomial distribution fitted to per capita abundance (– indicates not calculated because of small sample size).

Number of fleas on live and dead hosts

A small proportion (5%) of mice and rats were still alive when the snap traps were checked. These were killed and included in the sample. The majority were mice caught by the tail in rat traps. The abundance of fleas on live and dead hosts is summarised in Table 2. We used a one-sided t-test of the hypothesis that live hosts carried more fleas than dead hosts, separately for each host and flea species. In no case was there a significant difference (P ≥ 0.2). These data indicate that fleas did not leave their hosts quickly after they died (cf. King and Moody 1982a), and that our counts reflect flea loads on live animals.

Table 2. Numbers of fleas on live and dead hosts (1114 dead and 36 live rats, 1525 dead and 97 live mice).

		Host dead	Host alive
Host	Species	Mean (SE)	Mean (SE)
Mouse	L. segnis	1.95 (0.05)	2.19 (0.20)
	N. fasciatus	0.08 (0.01)	0.05 (0.05)
Rat	L. segnis	0.31 (0.03)	0.72 (0.28)
	N. fasciatus	0.72 (0.05)	0.42 (0.14)

Sex ratio of fleas

The sex ratio of L. segnis was strongly female-biased on both mice (61.2%, N = 3183; X² = 161, 1 df, P < 0.001) and rats (72.1%, N = 377; X² = 74, 1 df, P < 0.001). The observed sex ratio of N. fasciatus was also female-biased on both mice (53.5%, N = 129) and rats (62.9%, N = 808); the bias was not statistically significant for mice (X² = 0.63, 1 df, P = 0.43), but was significant for rats (X² = 54, 1 df, P < 0.001).

Distribution of number of fleas per host

The number of fleas per host was much more variable than expected from a Poisson distribution (i.e. variance much greater than mean) in all cases except the small sample of N. fasciatus from mice in silver beech (Table 1). Aggregation of fleas measured by the negative binomial parameter k was towards the high end of the range tabulated by Shaw et al. (1998): a large proportion of the flea population was concentrated on a few host individuals (see also Krasnov 2008, p. 328).

Co-occurrence of flea species

The prevalence of fleas of either species was 51.3% for mice (N = 1622) and 35.2% for rats (N = 1151). Host individuals with fleas of one species were more likely to be infested with the other species (chi-squared test for homogeneity of contingency tables: mice X² = 19.1, 1 df, P < 0.001; rats X² = 19.8, 1 df, P < 0.001). This effect is explained, at least in part, by correlated effects on the two flea species of host sex and age, and seasonal variation, as we document below.

The effects of sex and age of host

Male mice were about 27% more likely to carry L. segnis fleas than were female mice (Table 3). The observed prevalence of L. segnis was about 22% greater among male than female rats, and lack of significance may merely be due to the lower overall prevalence of L. segnis in that species. The prevalence of N. fasciatus on mice, and of either flea species on rats, did not differ significantly with sex of host (Table 3).

Table 3. Sex differences in prevalence of fleas on mice and rats. 14 mice and 6 rats were not sexed. P is significance level of test for sex difference in prevalence.

		Sex of host
Host	Species	Male	Female	P
Mouse	L. segnis	0.546	0.431	<0.001
	N. fasciatus	0.069	0.061	0.61
	N	919	689
Rat	L. segnis	0.161	0.132	0.20
	N. fasciatus	0.279	0.248	0.25
	N	591	553

Prevalence trended upwards with age for each flea/host combination except for L. segnis on rats (Figure 1). This was also true for the number of fleas per infected host with respect to L. segnis on mice and N. fasciatus on rats (Figure 2).

Figure 1. Variation with age in the prevalence of fleas on mice A, and rats B. Bars indicate 95% binomial confidence intervals. Toothwear classes differ between mice and rats; 4 mice in toothwear class 1, and 9 rats in each of toothwear classes 1 and 7, were lumped with the adjacent class for plotting.

Figure 2. Variation with age in the number of fleas per infected host. A, mice B, rats. Sample sizes are numbers of individuals with at least one of the particular flea species. Some toothwear classes combined as in Figure 1. Bars indicate 95% confidence intervals for the mean.

Effect of season and breeding status of host

The prevalence of both flea species was lowest in the autumn and highest in the spring or summer (Table 4). Breeding of rodents was largely restricted to spring and summer (Fitzgerald et al. 2004; Efford et al. 2006). We examined the breeding-season prevalence of fleas on subclasses of adult rodents (potentially mature mice in toothwear class 4 and above, and rats in toothwear class 3 and above). Breeding female mice and rats were less likely to carry L. segnis than non-breeding females (Table 5). No significant difference was detected with N. fasciatus, although the trend was similar (Table 5).

Table 4. Seasonal variation in prevalence of fleas on mice and rats. Samples were collected in February (summer), May (autumn), August (winter) and November (spring). Sample size N is the number of host individuals.

Host	Species	Summer	Autumn	Winter	Spring	P
Mouse	L. segnis	0.548	0.423	0.490	0.517	0.007
	N. fasciatus	0.057	0.038	0.062	0.099	0.006
	N	334	366	516	406
Rat	L. segnis	0.200	0.101	0.102	0.243	< 0.001
	N. fasciatus	0.261	0.163	0.310	0.324	< 0.001
	N	230	307	403	210

Table 5. Prevalence of fleas on mice and rats in the breeding season (spring and summer) in relation to sex and breeding status of the host.

			Females
Host	Species	Males	Non-breeding	Breeding	NB:B	P
Mouse	L. segnis	0.616	0.537	0.371	1.4	0.009
	N. fasciatus	0.094	0.091	0.043	2.1	0.19
	N	372	164	116
Rat	L. segnis	0.288	0.229	0.105	2.2	0.03
	N. fasciatus	0.348	0.312	0.200	1.6	0.10
	N	198	109	95

Notes: ‘Breeding’ females were pregnant or lactating. Mice in toothwear classes 1–3 and rats in toothwear classes 1–2 were excluded as animals in these classes were largely immature (Fitzgerald et al. 2004; Efford et al. 2006). NB:B is the ratio of prevalence in non-breeding and breeding females; the P value is for a test that this ratio equals 1.0.

Spatial variation in prevalence

The prevalence of L. segnis on mice was greater on the low-elevation long line than the high-elevation silver beech line (Table 1). There was also evidence for variation along the long line itself, with higher prevalence at ‘beech’ sites (Table 6).

Table 6. Prevalence of fleas on mice and rats caught at 82 ‘podocarp’ and 34 ‘beech’ trap sites on the long line (see text).

Host	Species	Beech	Podocarp	B:P	P
Mouse	L. segnis	0.550	0.473	1.16	0.007
	N. fasciatus	0.049	0.071	0.69	0.14
	N	431	1189
Rat	L. segnis	0.210	0.137	1.53	<0.001
	N. fasciatus	0.407	0.240	1.70	<0.001
	N	162	988

Note: B:P is the ratio of prevalence at beech and podocarp sites; the P value is for a test that this ratio equals 1.0.

Annual variation in prevalence

Host density varied substantially over the 20-year sampling period. Mouse density showed irregular multiannual fluctuations and average rat density approximately doubled between the early and later years (Figure 3). Too few hosts were trapped in some seasons to determine the prevalence of fleas, and we therefore plotted prevalence aggregated by year (Figure 3). The prevalence of each flea species showed a weak tendency to increase over time on both hosts (likelihood ratio tests Appendix 1).

Figure 3. Annual variation 1972–1991 in the proportion of A, mice and B, rats with fleas in relation to a quarterly trap-catch index of the host population. Solid symbols Leptopsylla segnis, open symbols Nosopsyllus fasciatus. 95% binomial confidence intervals. N is the number of the host species sampled in each year. Analysis year runs from previous November to September. Breaks in the population index on the log scale indicate seasons in which no host was trapped. Dashed line separates nominal ‘low’ and ‘high’ values of the population index. Shaded band indicates 95% prediction interval for prevalence of dominant flea species (L. segnis on mice and N. fasciatus on rats).

Samples from times of low and high host density did not differ significantly in flea prevalence (P > 0.15) except for L. segnis on mice for which there was a negative relationship (56.4% low, 46.7% high; P = 0.001). Periods of high host density had a preponderance of young hosts (Fitzgerald et al. 2004), so the density effect was possibly an artefact of increasing flea prevalence with age of host (Figure 1). We performed a negative binomial regression of per capita L. segnis abundance on mouse age class and found that adding the host density index as a continuous predictor did not improve the model (likelihood ratio test X² = 1.7, 1 df, P = 0.2; details in Appendix 2).

There was weak evidence that annual variation in the prevalence of L. segnis on mice was linearly correlated with that on rats (r = 0.37, P = 0.1). No correlation was apparent for N. fasciatus (r = 0.18, P = 0.4).

Discussion

We found only two species of flea on rodents in forest of the Orongorongo Valley, Leptopsylla segnis and Nosopsyllus fasciatus. In this ‘two rodent – two flea’ system there was a clear predominance of L. segnis on mice and of N. fasciatus on rats. The life histories of these two fleas are very different: L. segnis is ‘semi-sessile’, often referred to as semi-sedentary, remaining attached to the fur of the host for long periods, whereas N. fasciatus leaves the host after feeding, to conduct much more of its life cycle in litter or nesting material, as is more common among fleas (Krampitz 1980; see also Cole and Koepke in Gibson 1986; and Krasnov et al. 2004).

The majority of mice and more than a third of rats carried fleas. Prevalence remained high through large fluctuations in host density (Figure 3) and despite the low absolute density of hosts. Fleas can suppress the survival of rodent hosts in the laboratory (e.g. Devevey and Christie 2009). High flea prevalence in the wild therefore has the potential to affect host population dynamics, even without the potential for disease transmission.

Influence of host sex, age and season on flea infestation

Both season and host age (indexed by toothwear) had strong effects on prevalence. The two predictors are correlated because rodent breeding was seasonal (Fitzgerald et al. 2004; Efford et al. 2006). Post-breeding host populations in autumn (May) are dominated by recently recruited rats and mice with low flea prevalence (Table 4). The intuitive explanation is that each host accumulates fleas throughout its life, and having acquired fleas is unlikely to become flea-free, especially as habitual refuges and nest sites become infested.

We found a somewhat greater prevalence of fleas on male hosts than on female hosts, but the difference was pronounced only for L. segnis on mice (Table 3). Male-biased parasitism is widespread (Krasnov et al. 2012). The larger home ranges and greater activity of male mice may be a sufficient explanation (e.g. Fitzgerald et al. 1981).

Spatial variation

The prevalence of the dominant flea species on each host along the low elevation long line was greater at ‘beech’ trap sites than at ‘podocarp’ sites (Table 6). Mice from the high elevation silver beech trap line carried relatively few fleas (Table 1). Long-line beech sites, mostly with Nothofagus truncata, tended to be on drier slopes, but also showed a less complex forest structure than ‘podocarp’ sites (Campbell 1984). Differences in flea abundance may therefore be either a direct response to environmental conditions or a reflection of host behaviour in the differing habitats.

Sex ratio of fleas

Female fleas predominate in samples from their hosts in New Zealand (Table 7). A female bias is common in fleas (Krasnov et al. 2008, p. 93), although the sex ratio of L. segnis on M. musculus in California was only slightly female-biased (51% of 251 fleas; Linsdale and Davis 1956). Despite the substantial behavioural differences between L. segnis and N. fasciatus (Krampitz 1980), the overall sex ratios of fleas residing on the rodent hosts were similar (62.4% and 61.6% respectively in our study). Gibson (1986) noted a male bias in 24 N. fasciatus collected from rat nests, but collections from host burrows are rarely male-biased (Krasnov et al. 2008, Table 7.2).

Table 7. Sex ratio of fleas on mice and rats in New Zealand (samples of 10 or more individuals).

Flea	Host	Proportion
species	species	female	N	P	Source
L. segnis	M. musculus	0.612	3183	<0.001	This study
	R. rattus	0.721	377	<0.001	This study
	M. musculus	0.611	216	0.001	Mana I. Efford et al. (1988)
	M. musculus	0.714	28	0.02	Mana I. May 1989 BMF unpubl.
	R. rattus	0.714	14	0.11	Great Mercury I. May 1984 BMF unpubl.
	Rodent species	0.704	142	<0.001	Gibson (1986)
N. fasciatus	M. musculus	0.535	129	0.42	This study
	R. rattus	0.629	808	<0.001	This study
	M. musculus	0.585	53	0.22	Mana I. Efford et al. (1988)
	M. musculus	0.692	13	0.21	Mana I. May 1989 BMF unpubl.
	Mustela erminea	0.684	662	<0.001	King and Moody (1982a)
	R. rattus	0.606	33	0.22	Pureora. Innes et al. (2001)
	R. rattus	0.679	28	0.06	Great Mercury I. May 1984 BMF unpubl.
	Rodent species	0.560	218	0.08	Gibson (1986)
Pygiopsylla phiola	R. rattus	0.581	43	0.28	Great Mercury I. May 1984 BMF unpubl.

Note: P is the significance level of a chi-squared test for equal sex ratio.

Flea assemblages on rodent hosts in New Zealand

The lack of any other rodent flea species on the rodents of the Orongorongo Valley is unexplained; the rodent fleas Pygiopsylla hoplia Jordan & Rothschild, P. phiola Smit and Nosopsyllus londiniensis (Rothschild) have all been recorded within 20 km of the study area – P. hoplia in Stokes Valley, P. phiola at Lower Hutt and N. londiniensis in Stokes Valley and Belmont (Smit 1979; Pilgrim Collection Te Papa Tongarewa).

King and Moody (1982a) reported rather similar rodent–flea relationships in Craigieburn Forest Park and the Eglinton and Hollyford Valleys in Fiordland National Park. At Craigieburn, of 97 mouse carcasses searched, 18 had fleas (43 L. segnis and one N. fasciatus). No rats were trapped there, but of 164 stoats, 12 had fleas (22 N. fasciatus and five L. segnis). However, in Fiordland National Park none of 480 mice, 20 rats and 368 stoats had L. segnis, while N. fasciatus was present on all host species. A possible explanation for the absence of L. segnis there may be provided by Krampitz (1980, p. 377) who states that L. segnis ‘infests only synanthropic murines, and then only in restricted areas. Like its most important host, the house mouse, L. segnis is well adapted for residence in the drier parts of the world’ and ‘laboratory colonies do best in relatively low levels of humidity’. Average rainfall was 4250 mm in the Hollyford Valley and 2300 mm in the Eglinton Valley (King and Moody 1982b). However, the other New Zealand localities could barely be considered dry (Craigieburn 1450 mm (King and Moody 1982b), Pureora 1829 mm (King et al. 1996), and Orongorongo Valley 2370 mm (Campbell 1984)).

A three rodent – three flea system was recorded in Pureora Forest Park, central North Island, where an Australian rodent flea Pygiopsylla hoplia was present in addition to L. segnis and N. fasciatus (Innes et al. 2001). R. rattus had all three fleas, R. norvegicus had N. fasciatus and Pygiopsylla hoplia and M. musculus had L. segnis and N. fasciatus.

N. fasciatus is the most widespread flea of rodents in New Zealand, and it is also recorded on almost all the other species of small mammal in New Zealand – hedgehog, rabbit, cat, dog, stoat, ferret and weasel (Smit 1979). It was the predominant flea on stoats from New Zealand National Parks: 181 stoats yielded 670 rodent fleas; 99% were N. fasciatus, and the remainder were L. segnis (King and Moody 1982a). They suggested that stoats acquire N. fasciatus by using rat dens or when preying on rats. However, as Haddow et al. (1983, pp. 118–119) noted, N. fasciatus ‘may parasitise virtually any host with which it comes in contact, whether rodent … insectivores, carnivores, man or birds’. Their list of host records from specimens in the Rothschild Collection, British Museum includes many species of rodent, especially murids, but also ‘birds like doves, swallows, thrushes and wrens’. The list of avian hosts of N. fasciatus in New Zealand is probably even more eclectic, including a dead albatross on Campbell Island, Apteryx australis Shaw, Columba livia Gmelin, Rhipidura fuliginosa (Sparrman), Leucocarbo colensoi (Buller), Gallirallus australis (Sparrman), Strigops habroptila G. R. Gray, and Sturnus vulgaris L. (Smit 1964, 1965, 1979; Pilgrim Collection, Te Papa Tongarewa Museum of New Zealand).

Comments on the flea fauna of offshore and outlying islands

In our two-rodent system, there was a clear predominance of L. segnis on mice and N. fasciatus on rats. This raises the possibility that fleas merely ‘spill over’ to the alternate host, but there is evidence from islands with restricted host faunas that N. fasciatus persists on mice in the absence of rats (Mana Island from 1977–1978 (Efford et al. 1988) to 1989 (Fitzgerald and Guo 1989), Allports Island (Murphy unpubl. in Murphy and Pickard 1990, p. 241), and Antipodes Island (Smit 1979, Pilgrim Collection Te Papa Tongarewa Museum of New Zealand)). The only evidence of which we are aware that a population of L. segnis may persist on rats in the absence of mice, comes from mouse-free islands in northern New Zealand – Korapuki Island (Smit 1979), where there was only Rattus exulans (Peale) (Towns and Broome 2003), Motukahaua (Gibson 1986) where there was only R. rattus (Innes 2005), and Great Mercury Island where there were both R. rattus and R. exulans (BMF pers. obs.).

R. exulans presumably arrived in New Zealand without the cosmopolitan rodent fleas: those fleas are not known from the eastern Pacific, although the data are sparse (Gibson 1986; Roberts 1991a, 1991b). Roberts (1991c) advanced the hypothesis that the fleas of Rattus exulans on New Zealand islands where it is now the only rodent host had originally arrived with R. rattus, R. norvegicus or M. musculus that subsequently died out. However, spontaneous extinction of these cosmopolitan rodents has not been documented and, given the frequency with which N. fasciatus has been found on avian hosts (references above), it seems more likely to us that it hitched a ride on a bird. Starlings and gulls that nest or roost on small offshore islands would be candidates. Roberts’ hypothesis may apply in the case of L. segnis on Korapuki, or the fleas may have transferred to R. exulans from another source. However, the hypothesis may be difficult to test in future as rodents are eradicated from more and more islands to conserve the endemic fauna.

Acknowledgements

We are indebted to the late Prof. R.L.C. Pilgrim for guidance on flea preparation and identification. We thank Te Papa Tongarewa Museum of New Zealand for access to the Pilgrim Siphonaptera Collection. We thank two reviewers for their comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data analysed in this paper are openly available in Zenodo at https://doi.org/10.5281/zenodo.8151598.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.