Flea fitness is reduced by high fractional concentrations of CO₂ that simulate levels found in their hosts' burrows

Parasites are ubiquitous in that almost every living organism is parasitized by other organisms. Some parasites live inside the host's body (endoparasites), while others live on the host's surface (ectoparasites). Many ectoparasites of terrestrial vertebrates are nidicolous arthropods that live on burrowing hosts. These ectoparasites spend much of their time in the host's burrow, which provides an environment for living, and often feeding, to pre-imaginal and/or adult stages (Holland, 1964). These ectoparasites usually alternate between spending time on and off the body of their host; thus, they must contend with the two types of environment: (i) the environment on board the host, including its immune and behavioral responses; and (ii) the off-host environment, including the physical conditions inside the host's burrow.

Ectoparasites of small mammals benefit from some of the same advantages of living in burrows as their hosts. For example, the microclimate within the burrow is more stable than that outside the burrow (Kenagy, 1973; Reichman and Smith, 1990; Shenbrot et al., 2002). However, the air in some rodent burrows contains high fractional concentrations of CO₂ (F_CO2), and this is particularly true for burrows of fossorial rodents and those that have blocked entrances (Maclean, 1981b; Kuhnen, 1986; Burda et al., 2007). In contrast, F_CO2 in most burrows of semi-fossorial rodents and those that maintain open burrow entrances is close to the atmospheric value, such as in burrows of Merriam's kangaroo rat (Dipodomys merriami) and banner-tailed kangaroo rat (Dipodomys spectabilis) (Soholt, 1974; Kay and Whitford, 1978). Nonetheless, even in burrows with open entrances, F_CO2 can be higher than atmospheric concentrations. For example, F_CO2 in natural burrows of eastern chipmunks (Tamias striatus) ranged between 0.009 and 0.029, and F_CO2 increased with distance from the burrow entrance (Maclean, 1981b). In the same study, nest chambers in artificial burrows that were inhabited by eastern chipmunks had F_CO2 ranging from 0.012 to 0.064 (Maclean, 1981b). Similarly, F_CO2 as high as 0.022 – that is, 50-fold greater than atmospheric concentrations – occurred in nest chambers of permeable, artificial burrows inhabited by Sundevall's jirds (Meriones crassus Sundevall) (Brickner-Braun, 2014). Although F_CO2 in the tunnels of these artificial burrows is close to zero (Brickner-Braun et al., 2014), these observations highlight the variability of F_CO2 within the burrow and the need to consider microhabitat, such as nest versus tunnel, when investigating conditions experienced by individuals, especially because the nest is where F_CO2 is likely to be highest when occupied by a mother and offspring. In addition, F_CO2 in burrows is expected to be highest near the burrow inhabitant (Wilson and Kilgore, 1978; Burda et al., 2007). Thus, ectoparasites are likely to be exposed to F_CO2 higher than atmospheric F_CO2 when they dwell in the nest chamber, particularly when it is occupied by a host. Furthermore, ectoparasites may be disproportionately exposed to high F_CO2 relative to their mammalian hosts because when on their hosts they inhabit the layer between the host's skin and its hair surface, where F_CO2 may be further elevated because the air is poorly mixed with the surrounding burrow air.

Several rodent species have physiological and behavioral traits that mitigate the effects of inspiring high F_CO2 (Boggs et al., 1984), including respiratory characteristics and blood properties that facilitate increased oxygen delivery to tissues (Foreman, 1954; Hall, 1966) and maintain blood acid–base balance. Given the long, shared evolutionary history of ectoparasites and their hosts (Holland, 1964; Waage, 1979; Balashov, 1984; Barker, 1994; Krasnov, 2008), it is likely that ectoparasites of burrowing rodents, such as fleas, encounter periods of increased F_CO2. We therefore hypothesized that fleas have also evolved adaptations for living in environments with high F_CO2, and thus predicted that high F_CO2 will not affect their fitness.

To test this hypothesis, we quantified the effect of an environment with a high F_CO2 (0.04) on fitness of fleas (Xenopsylla ramesis Rothschild) that were raised on their rodent hosts, Sundevall's jirds. Fleas (Siphonaptera) are nidicolous ectoparasites, which live mostly on small and medium-sized burrowing mammal species – only about 6% of flea species are parasites of birds (Krasnov, 2008). All adult fleas are obligate hematophages and most fleas alternate periods when they occur on their host and when they occur in its burrow or nest. Finally, in almost all flea species, with few exceptions, pre-imaginal development is off-host, so both adult and pre-imaginal fleas are strongly affected by environmental factors (Holland, 1964; Marshall, 1981; Krasnov et al., 2001a,b, 2002). For example, survival and development rates of fleas change with ambient temperature and relative humidity (Krasnov et al., 2001a,b, 2002).

We chose to use X. ramesis as a model for this research because we expected that X. ramesis typically experience environments with high F_CO2. First, in the Negev desert, X. ramesis is the most abundant flea of Sundevall's jird, Tristram's jird (Meriones tristrami) and fat sand rats (Psammomys obesus) (Krasnov et al., 1997; Krasnov et al., 1998, 1999) and it spends most of its time in the burrow of the host (Krasnov et al., 2004). Both jird species occasionally plug the entrances to their burrows (Mendelssohn and Yom-Tov, 1999; Shenbrot et al., 2002); this happens particularly often in habitats where these species harbor X. ramesis (Krasnov et al., 1996; Shenbrot et al., 2002). In addition, X. ramesis has been recorded on a rodent species that often seals its burrow entrances, namely Wagner's gerbils (Gerbillus dasyurus) (Harrison, 1956). Furthermore, the tunnels of burrows of these rodents can be long (up to 10 m for Sundevall's jirds; Shenbrot et al., 2002) and deep (up to 80 cm for fat sand rats; Mendelssohn and Yom-Tov, 1999). In general, long burrows and burrows with plugged entrances have elevated F_CO2 (Kay and Whitford, 1978; Maclean, 1981b; Kuhnen, 1986; Burda et al., 2007), thus X. ramesis are likely to encounter high F_CO2 often. Second, the host–parasite system used in the present study serves as a model for other host–parasite pairs, where the host lives in sealed burrows that might have high F_CO2. For example, the fleas Xenopsylla conformis, Xenopsylla gerbilli, Xenopsylla hirtipes and Xenopsylla skrjabini are closely related to X. ramesis. These species are common parasites of jerboas (Dipodidae), hamsters (Cricetus, Mesocricetus, Cricetulus) and ground squirrels (Spermophilus) that seal their burrows during hibernation, as well as fossorial mole voles (Ellobius) in Central Asia (Harrison, 1956; Popova, 1968; Krasnov, 2008) and, thus, might be exposed to high F_CO2.

In our experiments, hosts were housed in an environment with either high or atmospheric F_CO2. We predicted that if fleas are adapted to high F_CO2, then their performance would not differ between environments with high and atmospheric F_CO2. Specifically, we predicted that survival, egg production, the proportion of successfully hatched larvae, and reproductive success (number of imagoes produced) would be unchanged in fleas raised on hosts under high F_CO2. We also explored differential sensitivity of male and female fleas to high F_CO2. Earlier studies on X. ramesis showed that although males and females respond differently to relative humidity and temperature, these factors do not affect overall sex ratio of new imagoes (Krasnov et al., 2001a). Thus, we expected that the proportion of new female imagoes (i.e. the sex ratio) would not differ between the treatments.

In addition, we were interested in how living in a high-F_CO2 environment affected behavior of fleas, specifically mobility. The anti-parasite behavior of a host (i.e. grooming) is an important defense tool (Nelson et al., 1975; Hart, 1994), so reduced mobility may make fleas more susceptible to host defenses. To test the effects of high F_CO2 on activity, we housed fleas without a host in F_CO2=0.04, or in CO₂-free conditions and quantified their mobility. As part of this experiment we also quantified how high CO₂ directly affects survival of fleas.

We used fleas, X. ramesis, from our laboratory colony started in 1999 and collected from wild Sundevall's jirds, fat sand rats or Wagner's gerbils. Details on breeding and maintenance of fleas can be found elsewhere (Krasnov et al., 2002, 2003b; Khokhlova et al., 2009, 2010, 2012). In brief, fleas were maintained on rodents that were kept individually in plastic cages with a wire mesh floor over a pan with a mixture of sand and dried bovine blood (larval nutrient medium) at 25°C air temperature with a photoperiod of 12 h:12 h (light:dark). Every 2 weeks, all substrate and bedding material from the rodent's cage (including nest box and pan) were collected into plastic boxes with perforated lids and transferred to an incubator (FOC225E, Velp Scientifica, Milano, Italy; 25°C air temperature and 75% relative humidity) where the fleas developed. All fleas used in experiments were selected randomly from colonies.

We used Sundevall's jirds from a laboratory colony started in 1997–1999 from progenitors that were captured in the Negev desert (southern Israel). Details on rodent maintenance in colonies can be found elsewhere (Krasnov et al., 2002, 2003a; Khokhlova et al., 2009, 2012; Liberman et al., 2011). In short, rodents were maintained in plastic cages (35×25 cm and 15 cm high) at 25°C ambient temperature and 12 h:12 h (light:dark) light regime. The cages were supplied with wood shavings as bedding, paper towels as nesting material, and a pinecone for chewing. Millet seeds and alfalfa (Medicago sp.) were offered daily ad libitum. Commercial rodent chow was offered once a week. No water was available to the rodents as the alfalfa supplied enough water for their needs. This study was conducted under permits from Ben-Gurion University Committee for the Ethical Care and Use of Animals in Experiments (IL-36-9-2007).

To test for the effects of high F_CO2 on flea fitness, we used 18 female jirds and assigned nine of them to the high-CO₂ treatment and the remaining nine to the atmospheric-CO₂ treatment. To minimize maternal and genetic effects, we used pairs of sisters from nine litters for our experiment and assigned sisters to different treatments. We conducted the experiment in three sequential groups of hosts; each block included three rodents assigned to each treatment. To minimize the effect of age, all jirds were ∼7 months old (144.8±1.5 days, mean±s.d.) when the experiment started.

Experimental cages were 9 l airtight plastic containers (20×26×17 cm, Lock&Lock, Hana Cobi, Korea). To create a high-CO₂ environment, we used two mass flow controllers (MFC-2, Sable Systems, Las Vegas, NV, USA) to pump pure CO₂ and room air at rates that resulted in a final mixed F_CO2 of 0.035. This main airflow was then split into six channels with a series of Y-junctions. The flow of air through each line was ∼300 ml min⁻¹ and was controlled with a needle valve. To maximize mixing of air within the experimental cages, the air inlet and outlet were spaced as far apart from each other as possible (i.e. the air inlet was in the lower right-hand corner of one side, and the air outlet was in the upper left-hand corner of the opposite wall of the cage). At the flow rate used, air in the experimental cage turned over approximately twice an hour. The average CO₂ production of an adult female Sundevall's jird (∼128 g) is 1.58 ml CO₂ min⁻¹ (Brickner-Braun, 2014; I. Brickner-Braun, personal communication), so using equation 11.8 from Lighton (2008), the equilibrium F_CO2 in the high-CO₂ treatment cages was calculated to be 0.0402 (∼0.04). The cages for jirds assigned to the atmospheric-CO₂ treatment were designed similarly, except that they received room air instead of mixed air. Based on the average metabolic rate of a jird and the flow rate of air for the experimental cages, we calculated that these cages would equilibrate at a F_CO2 of 0.005. We verified F_CO2 in the experimental cages by measuring the excurrent air with a CO₂ analyzer (Foxbox, Sable Systems; data not shown). This was higher than atmospheric-CO₂ concentrations, but this level of CO₂ is biologically irrelevant for hosts (Boggs et al., 1984). Cages were fitted with wire mesh floors (5×5 mm) and the bottom of the cage was covered with a 3 mm layer of sand. This cage arrangement provided the fleas (see below) with access to sand so they could lay their eggs and also protected fleas from being killed easily by the host. These experimental cages also contained paper towels for nesting material, and the jirds were provided with millet seed ad libitum and alfalfa leaves.

After 3 days in the experimental cages to allow jirds to acclimate to air treatments, we added 100 female and 50 male fleas to each cage to mimic the number of fleas that reside in the burrow of an individual Sundevall's jird in the wild (Goüy de Bellocq et al., 2006). Fleas were combed from rodents and collected from the sand, counted and sexed, and replaced with a new batch of 150 fleas every 3–6 days until the end of the experiment (schedule details in Table 1). We used seven batches of fleas that were put on hosts consecutively, so that each host was exposed to seven batches of fleas in sequence. We did not quantify flea fitness for batch 2 because fleas were not on hosts long enough (<3 days) to lay the necessary eggs (Krasnov et al., 2004) (see below for details of fitness variables measured); thus, we used data from six batches of fleas. We do not have data for fleas on two hosts in the high-CO₂ treatment from the first group. In addition, for the first batch of fleas of the third group, we obtained no data for two of the hosts from the high-CO₂ treatment.

Female fleas were counted then placed in glass Petri dishes (100 mm diameter). We put ∼10 females in each Petri dish (mean±s.d. of 9.3±1.9 female fleas per dish) and we created as many replicate Petri dishes for each host as allowed by the number of fleas collected from that host, up to six dishes (mean±s.d. of 4.1±1.2 dishes). Petri dishes were then placed in an incubator at 25°C and ∼90% relative humidity; CO₂ treatments were not maintained during incubation. After 2 days, the female fleas were removed from the Petri dish and we calculated the mean number of eggs laid per female. The Petri dishes were put back into the incubator and the hatched larvae were counted and discarded 4, 5, 6, 7, 8 and 9 days after females were initially put in the Petri dish. We calculated the proportion of larvae hatched per day per female and the total proportion of eggs that developed into larvae.

When fleas for each batch were collected, the sand from the bottom of the cage was also collected, placed in a plastic box with a perforated lid, and incubated at 25°C and ∼90% relative humidity until adult fleas completed emergence (see below). CO₂ treatments were not maintained during offspring development. Sand from each batch and from each host was incubated in separate boxes for 23 days and then the adult fleas that had emerged were counted. This was continued daily until no more fleas had hatched for 7 days.

The variables we quantified from experiment 1 were daily survival of adults for the duration of time they were on the host, average number of eggs laid per female, and hatching success of eggs. We calculated mean reproductive success, i.e. the mean number of new imago produced per mother flea, from data collected on imagoes emerging from the sand. The number of mother fleas used in this calculation was calculated as the mean number of females added to the experimental cage at the beginning of the flea batch (e.g. batch 1, batch 3, etc.) and the number of females collected from the cage at the end of the same batch. We also calculated the proportion of female imagoes. Average development time from egg to emergence, number of days to first emergence, and number of days to last emergence were calculated to quantify flea development. Note that the data used in these calculations do not include the data from the eggs laid in Petri dishes and that these data were quantified for each batch of fleas separately.

We used general linear mixed effects models to determine the differences between fleas on hosts living in high and atmospheric CO₂. Random effects were batch, host group, host mother ID and host ID. CO₂ treatment (high CO₂ or atmospheric CO₂) was the only fixed effect. We determined whether CO₂ treatment was included in the best fit model using likelihood ratio tests (LRT) as recommended for general linear mixed effects models (Zuur et al., 2009). We used the following transformations to make the residuals normal: number of days to last emergence was log transformed, number of days to first emergence was transformed by raising to the power 0.2, mean development time was square root transformed, and the proportion of eggs hatched was transformed using an angular transformed. Statistics were performed in the program R v2.15.1 (R Development Core Team, 2011) using package lme4.

To determine how F_CO2 directly affected survival and activity of fleas, we put fleas in high-CO₂ (F_CO2=0.04) and CO₂-free environments without a host. We used 10 replicates for each treatment, and for each replicate we put 10 fleas in a glass jar (20 ml) with a substrate of sand. The jars were covered with cloth that allowed gas exchange but prevented fleas from escaping. The jars for each treatment were then put inside a 9 l plastic experimental cage (described above) and these cages were put inside a temperature-controlled cabinet (Refritherm-5, Struers, Denmark) set at 25°C. To maintain the desired F_CO2, experimental cages were attached to air pumps. For the CO₂-free air treatment, incurrent air was pumped into the system through a purge gas generator (Pure Gas, Broomfield, CO, USA, model no. PCDA-1-12-m-32-C) to remove CO₂ and water vapor, both to less than 1 ppm. For the high-CO₂ treatment, we mixed dry, CO₂-free air with bottled CO₂ with a two-way gas mixing pump (Wösthoff DIGAMIX 5KA27, Bochum, Germany) set to a F_CO2 of 0.04. For both treatments the flow rate of incurrent air was set to 300 ml min⁻¹.

On days 5 and 6 we counted the surviving fleas. On day 5 we returned the fleas to their jars, and returned the jars to their experimental treatments. On day 6 we also quantified activity by pouring the sand and fleas into the bottom of a Petri dish (100 mm diameter) elevated 6 cm. Fleas were allowed to jump out of the dish for 5 min and then the top was put on the Petri dish to prevent the remaining fleas from escaping. We then counted the number of fleas that had jumped out of the dish and used it as a proxy for mobility.

We tested for differences in survival using a general linear mixed model. Survival was measured as the cumulative proportion of fleas surviving on days 5 and 6 of the experiment. We transformed both survival and activity with an angular transformation to normalize the distribution of residuals. CO₂ treatment (high F_CO2 or CO₂ free), day (5 or 6 days in treatment), and interaction between CO₂ treatment and day were potential fixed effects. Replicate (i.e. jar) was the only random effect. We determined whether CO₂ treatment was included in the best fit model using LRT as recommended for general linear mixed effects models (Zuur et al., 2009). We tested for differences in mobility for fleas in high-CO₂ and CO₂-free treatments using a t-test. Statistics were performed in program R v2.15.1 (R Development Core Team, 2011). General linear mixed models were run using package lme4.

In high F_CO2, 25.7±2.3 (estimated mean±s.e.) fleas died per day, whereas only 20.3±2.3 fleas died per day in atmospheric CO₂. Thus, 27% more fleas died per day when on hosts in high F_CO2 than in atmospheric CO₂ (LTR between model with CO₂ treatment as a fixed effect and null model, LR₁=21.6, P<0.001). On average, females in high F_CO2 produced 0.3 less of an egg per day than those in atmospheric CO₂ (LR₁=6.7, P=0.009; Fig. 1). The hatching success of eggs was the same regardless of CO₂ treatment (LR₁=1.4, P=0.24; Fig. 1).

Mean reproductive success was 25% lower for fleas in high F_CO2 than in atmospheric CO₂ (LR₁=7.0, P<0.001; Fig. 2). The proportion of newly emerged imagoes that were female did not differ between treatments (LR₁=1.36, P=0.24; Fig. 2). Time of development until the first new imago emerged was, on average, 1.9 days longer in high CO₂ than in atmospheric CO₂ (LR₁=850.1, P<0.001; Table 2) and mean development time of the new generation of fleas was 2 days shorter in high F_CO2 (LR₁=10.2, P<0.001; Table 2) than in atmospheric CO₂. However, the time of development assessed as the time from the first egg laid to the last adult flea to emerge did not differ between treatments (LR₁=0.007, P=0.93; Table 2).

Cumulative survival of fleas was 25.4% lower in the high-F_CO2 treatment than in the CO₂-free treatment (LRT between model with just treatment and day plus treatment, LR₁=4.7, P=0.03; Fig. 3A) and decreased as the experiment progressed (LRT between model with just day and day plus treatment, LR₁=29.9, P<0.001; Fig. 3B). We found no interaction between number of days since the beginning of the experiment and CO₂ treatment on cumulative survival (LRT between model with interaction and additive model, LR₁=0.49, P=0.48). Fleas in the high-F_CO2 treatment were also less active than those in the CO₂-free treatment (t_1,17=−5.53, P<0.001; Fig. 4).

The results of our first experiment indicate that the fleas we studied are not physiologically equipped to persist in environments with high F_CO2 because fleas in such conditions have reduced fitness relative to those in low F_CO2; both survival and reproductive output were reduced when F_CO2 was 0.04. The second experiment demonstrated that flea survival was reduced when fleas were exposed to high F_CO2, as opposed to being killed by hosts, which is consistent with studies on other insects (Nicolas and Sillans, 1989). High F_CO2 may directly affect survival of fleas by altering gas exchange rates. Elevated levels of CO₂ cause insects to increase respiration and spiracles to dilate, which in turn increases desiccation and may cause mortality (Bursell, 1974; Nicolas and Sillans, 1989). Specifically, spiracles remain open when CO₂ tensions are above a threshold and the spiracle is very sensitive to CO₂ (Burkett and Schneiderman, 1967; Förster and Hetz, 2010). In the flea X. cheopis, opening of the spiracles is controlled by want of oxygen, but the duration that spiracles remain open is determined by how long it takes CO₂ to diffuse out of the spiracle (Wigglesworth, 1935). Diffusion time increases in environments with higher F_CO2, and X. cheopis kept their spiracles open constantly when respiring air with a F_CO2 of 0.02 (Wigglesworth, 1935). In addition, water loss increases when spiracles are open (Mellanby, 1935). Thus, fleas in our study that were respiring air with a F_CO2 of 0.04 likely kept their spiracles open continuously, leading to a higher rate of evaporative water loss. Indeed, high levels of CO₂ caused tsetse flies (Glossina morsitans) to keep their spiracles open and the rate of water loss increased as the percentage CO₂ in the air increased (Bursell, 1957). Experimentally reduced relative humidity negatively affects the survival of many insects, including X. ramesis, presumably because desiccation rates increase at low relative humidity (e.g. Krasnov et al., 2001a; Hulasare et al., 2005). Indeed, critical threshold experiments indicate that X. ramesis lose water to the environment when relative humidity drops below some point between 60% and 65%, and they lose water at a rate of 1.4–2.4% mass h⁻¹ when exposed to dry air, depending on the life history stage (Fielden et al., 2002). Small insects, such as fleas, would be particularly susceptible to water loss caused by high F_CO2 because resistance to desiccation decreases with decreasing body size (Hood and Tschinkel, 1990; Lighton et al., 1994; Lehmann et al., 2000), although some insects are capable of shifting patterns of gas exchange in low relative humidity to minimize water loss (Duncan et al., 2002).

In addition, our results indicate that high F_CO2 indirectly reduces flea survival because fleas in high F_CO2 were less active, which might increase the probability of being dislodged or killed by a host. This finding is consistent with the common practice of using sublethal doses of CO₂ to anesthetize insects in the laboratory. Mechanistically, flea activity may be reduced because CO₂ may block synaptic transmission at the skeletal neuromuscular junction, as was found in Drosophila melanogaster (Badre et al., 2005).

Reproductive success was also reduced in high F_CO2, and this reduction was the result of fleas laying fewer eggs in the environment with a high F_CO2 as opposed to that environment rendering eggs unviable. Furthermore, the proportion of new female imagoes did not differ between treatments, suggesting that sensitivity to high CO₂ was not sex dependent. Our results partially concur with those from experiments with D. melanogaster where it was found that flies laid fewer eggs and the fraction of eggs that hatched decreased when in a F_CO2 of 0.13; the decrease was even greater when F_CO2 was 0.20 (Helenius et al., 2009). In addition, exposure of insects to sublethal doses of CO₂ (often F_CO2=1) for short durations for anesthetic purposes resulted in insects laying fewer eggs and in reduced hatching success of those eggs (Press et al., 1973; Mbata et al., 1998). Differences regarding the viability of eggs between the present study and other studies might result from differences in experimental F_CO2 and duration of exposure. Sub-lethal doses of CO₂ reduce production by preventing maturation of oocytes in female red flour beetles (Tribolium castaneum) (Press et al., 1973). In addition, breathing air with a high F_CO2 could affect reproductive performance by reducing feeding and drinking, as was found in house cricket (Acheta domesticus) larva (Woodring et al., 1978), which would reduce the resources available for egg production. However, we did not investigate the mechanism by which breathing air with a high F_CO2 reduced reproductive success, so we cannot draw a general conclusion at this point.

High F_CO2 could also indirectly affect flea fitness through changes in the chemistry of the host blood. For example, a host inspiring air with a high F_CO2 might develop hypercapnia with high levels of CO₂ and bicarbonate in the blood (Henderson and Haggard, 1918), and inspiring high F_CO2 also changes gene expression of hosts, including genes associated with immune function (Cummins et al., 2010; Taylor and Cummins, 2011). These and other changes in host physiology might affect parasites by changing the nutritional value of the blood, similar to what happens to phytophagous insects when high F_CO2 indirectly affects insect fitness by altering the nutritional quality of host plants (Fajer et al., 1989; Coviella and Trumble, 1999), but this idea would have to be tested directly.

The duration of development time to emergence of imagoes increased in the high-F_CO2 treatment, which concurs with results found in other studies of invertebrates in which high F_CO2 was reported to increase development time (Nicolas and Sillans, 1989; Bouletreau-Merle and Sillans, 1996). The duration of development encompasses the effects of changes during the egg, larval or pupal stages of flea development. However, fleas for which development time was calculated were only exposed to different F_CO2 treatments as eggs, indicating that exposure during egg development within the mother or the egg stage underlies differences in the duration of development. Further evidence to support this idea is the lack of a difference between CO₂ treatments for the duration of time until the last imago emerged (Table 2). All eggs from each batch of fleas were collected from the host cages at the same time, so eggs laid later would have been exposed to their CO₂ treatment for less time, resulting in less time for plastic traits associated with development to have been altered.

The result concerning the duration of development time to emergence of imagoes indicates that conditions during early development alter traits throughout an individual flea's lifetime. The change in duration of development could have occurred through at least three mechanisms. Most simply, eggs in high F_CO2 could take longer to hatch, delaying the whole development cycle. Alternatively, female fleas exposed to high F_CO2 might alter offspring development time via maternal effects such as investment in resources provided to the eggs, resulting in slower development overall. Finally, exposure of eggs to high F_CO2 could push the individual into a ‘slow-development’ phenotype through changes in gene expression, resulting in an increased overall duration of development – duration of development being the result of developmental canalization (Cohen et al., 2012). We are inclined to believe that the first explanation is correct because it is the simplest. However, CO₂ exposure can alter expression of transcription factors (Cummins et al., 2010; Taylor and Cummins, 2011) and can alter biochemical pathways including NADPH production (Friedlander et al., 1984). Even short-term exposure to high F_CO2 can cause long-term disruption of neuroendocrine function in house crickets (Woodring et al., 1978). Thus, any of these proposed explanations might be correct and experimental work is needed to distinguish among them.

The effect of a high F_CO2 on parasite fitness likely depends on species, microclimatic conditions (e.g. relative humidity, temperature, F_CO2) and life stage. For example, D. melanogaster had increased development time and reduced viability when continuously exposed to a F_CO2 of 0.2 (Bouletreau-Merle and Sillans, 1996), but not when exposed to a F_CO2 of 0.018 (MacAlpine et al., 2011). Similarly, a F_CO2 of 0.05 improved fecundity in D. melanogaster (Bouletreau-Merle and Sillans, 1996), but a F_CO2 of 0.018 did not alter fecundity (MacAlpine et al., 2011). Furthermore, the duration of exposure to CO₂ that causes mortality depends on temperature and relative humidity (Nicolas and Sillans, 1989). Thus, caution should be exercised when generalizing the results of this study.

Furthermore, caution should also be exercised when drawing conclusions about the ecological implications of our results because we investigated the role of a single F_CO2 value, 0.04, on flea fitness. Mean F_CO2 in burrows of semi-fossorial rodents typically ranges between 0.005 and 0.064 (Maclean, 1981a,b; Kuhnen, 1986; Roper et al., 2001), although values as high as 0.108 have been recorded in an artificial burrow with sealed entrances (golden hamster, Mesocricetus auratus; Kuhnen, 1986). Furthermore, F_CO2 ranged from 0.002 to 0.08 in nest chambers of artificial burrows designed to match the geometric properties of natural burrows of Sundevall's jirds (Brickner-Braun, 2014). Thus, the F_CO2 we chose was within the range of F_CO2 found in borrows and is in the middle of the range found in artificial nest chambers of jirds. The F_CO2 used, however, is the upper end of the typical range and is potentially outside the range found in natural burrows of jirds, which has never been measured. More experimental work is needed to determine how more moderate F_CO2 affects fitness of fleas and to better understand the ecological context of these results. If the effects of inspiring high F_CO2 on the fitness of fleas are generalizable, the results of our experiment suggest an interesting hypothesis about the benefits for hosts of high F_CO2 in their burrows.

Studies of burrow-dwelling animals, including rodents, report that the architecture of the animals’ burrow may reduce the build-up of CO₂ in some cases (Vogel et al., 1973; Kleineidam and Roces, 2000; Brickner-Braun et al., 2014). Indeed, this behavior is expected if there are costs associated with high F_CO2, because burrows are the product of natural selection acting on the builder (Turner, 2000). Some animals, however, live in burrows with high F_CO2, particularly those with sealed entrances (Maclean, 1981b; Kuhnen, 1986; Mendelssohn and Yom-Tov, 1999; Shenbrot et al., 2002; Burda et al., 2007). This leads to the question, do rodents build burrows designed so they do not to completely eliminate the elevated F_CO2? Our results provide a potential answer: if our results are generalizable and high F_CO2 reduces parasite fitness and thus reduces the fitness costs of parasite infection on the host more than elevated F_CO2 reduces the host’s fitness, then selection should favor hosts that build burrows with slightly elevated F_CO2. This explanation does not exclude other explanations about how the balance of other costs – e.g. digging costs (Vleck, 1979) – and benefits results in selection on burrow design. Rather, reduction of parasite load may be just one of the many factors contributing to the complex costs and benefits equation of burrow design.

For this explanation, involving reducing parasite pressure, to be true, the F_CO2 found in natural burrows must bring about reduced fitness of the flea that feeds on the host, but simultaneously minimize the costs to the host. Previously, it was reported that rodents that inhabit burrows are physiologically adapted to living in them, and those physiological attributes allow rodents to tolerate high inspired F_CO2 (Boggs et al., 1984; Ar, 1987). Those studies suggest that rodents can tolerate high F_CO2 with little negative effect on their physiology and therefore high F_CO2 might not affect fitness. The results of the present study indicate that exposure to a F_CO2 of 0.04 reduces flea survival and reproductive success. Combined, these observations support the notion that high F_CO2 in burrows may benefit some hosts by reducing the fitness of ectoparasites.

This is publication no. 882 of the Mitrani Department of Desert Ecology. We thank Inbal Brickner-Braun for helpful advice and discussion during the planning of this experiment, and thank two anonymous reviewers for their constructive feedback.