Challenges to understanding the consequences of population bottlenecks for the conservation of endangered wildlife


K. A. Hale, School of Biological Sciences, University of Canterbury, Christchurch, New Zealand. Emails: and

The widespread decline of many species has brought new urgency to understanding the consequences of population bottlenecks. Even if conservation efforts succeed in returning endangered species to a less threatened status, their brief ‘flirt’ with extinction raises questions about any lingering effects of passing through a severe bottleneck. Although the genetic consequences of population bottlenecks are well documented (e.g. Nei, Maruyama & Chakraborty, 1975), it is unclear whether bottlenecks also lead to declines in the fitness of the recovered population that reduces its survival and ability to adapt to changing environments. As pointed out by Tompkins (2007), it would be useful for planning conservation initiatives if we knew whether the fitness effects of bottlenecks were just temporary, and which aspects of an animal's life history were most affected. If the fitness consequences of bottlenecks are relatively minor or short term, then more attention could be focused on other urgent conservation problems. We agree with Hawley (2007) that conservation biologists indeed need to ‘look beyond genetics’ to determine whether bottlenecks warrant their serious attention.

Our examination of two New Zealand robin Petroica australis populations, in which we compared a severely bottlenecked population with its ancestral source population, supports predictions that bottlenecks can lead to declines in the immune response of individuals in a post-bottlenecked population (Hale & Briskie, 2007). Our conclusion is based on combined analyses of current parasite loads, differences in blood cell composition and immunocompetence assays in both populations over two seasons. The importance of using a multiple approach is illustrated by the difficulty in interpreting any single measure of fitness. For example, the similar ectoparasite levels in both populations of robins would suggest that bottlenecks have little detrimental effect on parasite loads (Hale & Briskie, 2007). However, as Hawley (2007) points out, the key measure is not simply current parasite load but parasite resistance, and a number of other explanations for our results are possible (including the ones we did not address). Perhaps the greatest drawback of using current parasite loads to assess immune response is the assumption that each population has been exposed to potential parasites at similar rates and other potential environment factors are equal. Such conditions are unlikely to be met, and without experimental exposures of birds to parasites under controlled conditions, one cannot be sure whether the lack of a difference in current parasite load is due to some confounding factor. The experimental exposure of endangered animals to novel parasites or pathogens would allow a direct way to test a decrease in resistance to parasites or lowered immune response, but such methods raise ethical issues that may be difficult to address with the current methodology. Despite potential problems, it is clear that future workers need to approach the study of fitness costs of bottlenecks with a variety of tests as not all aspects of an animal's biology are likely to be affected equally or consistently over time.

As highlighted by Smits (2007), measuring whether a bottleneck induces fitness costs on individuals in the post-bottlenecked population requires a careful selection of tests. A poor selection of methods may lead to studies that wrongly conclude that the fitness effects of bottlenecks are minimal or even exaggerated. In our study, we experimentally measured the response of robins to phytohaemagglutinin (PHA) during two different times of the year (pre- and post-breeding) to minimize this risk and found differences in the strength of response to the antigen. If we had tested birds only during the spring, we would have concluded that there was no fitness cost (Hale & Briskie, 2007). Smits (2007) suggests that as we tested robins only during ‘benign’ periods of the annual cycle, we underestimated the fitness costs. Although stresses to birds likely differ over their annual cycle, our finding that robins in autumn show a compromised response to PHA should nonetheless raise concerns. If robins cannot cope with a foreign antigen during a supposed ‘benign’ period, how will they cope during the most stressful periods of their annual cycle?

On the other hand, if we had not found any differences in PHA response between our two populations in either season, we could rule out an effect during the seasons we tested, but would be unsure that differences might be present during more stressful periods, as Smits (2007) suggests. Negative results in this case do not necessarily mean an absence of fitness costs and conservation biologists should be cautious in extrapolating the results of studies beyond the boundaries defined by the researchers. As we suggest, measures of the fitness consequences of bottlenecks need to be performed throughout all stages of an animal's life (seasonally and with increasing age) if we are not to underestimate the risks. Nevertheless, tests undertaken during a more stressful period, such as the breeding season, may involve sacrificing the breeding success of the birds involved. In our case, the differences in PHA response between populations in the autumn were sufficient to demonstrate a potential fitness cost of severe bottlenecks and we could not justify repeating the study at a time that might jeopardize the survival of dependent young.

Once a test is selected, then questions can arise as to whether the methods can be applied unquestioningly to a new species. Immunocompetence tests initially designed for one species may not be applicable for taxonomically different species without modification. Smits (2007) suggests that the results of our PHA tests should be treated cautiously as we altered the standard protocol given by Smits et al. (1999) from 24 to 6 h, and this invalidates our conclusions as this would not be enough time for a T-cell-mediated response to be initiated. However, recent work on five species of passerine birds has found peak responses to PHA within 6 h, with no significant increase in response after this time period (Navarro et al., 2003; Møller & Cassey, 2004), and that this time period coincides with the start of the main period of T lymphocyte accumulation at the site (Goto et al., 1978). As it is often necessary to hold birds during the post-injection period (as we did in our study), this shorter time frame is also likely to be less stressful to the birds, which can be important when working with endangered species (Tompkins, Mitchell & Bryant, 2006). We suggest that workers review whether current protocols are acceptable for their species before implementing them to a new species or new situations.

Whether individuals in bottlenecked populations suffer from increased loads of parasites is important both for understanding the fitness effects of bottlenecks and for the management of such potential threats. The wide diversity of potential parasites and variation in their presence and abundance over the seasons and with the age of an animal can mean that it is difficult to detect consistent patterns that might be indicative of potential problems in an endangered species. For species on isolated islands like the New Zealand robins in our study, there is the further complication that the absence of parasites may not be an indication of high immunocompetence but rather a result of isolation from the larger pools of parasites on the mainland (Hawley, 2007). Whether a particular parasite induces a fitness cost to the host is also likely to vary from species to species, with ectoparasites like hippoboscid flies exerting a higher immunological cost than feather mites (Smits, 2007). For example, Smits (2007) cites evidence that feather mites are commensal organisms with no pathological cost to the host, although this view is not universal (Poulin, 1991; Poiani, 1992; Thompson et al., 1997; Harper, 1999). In contrast, hippoboscid flies appear to pose a greater threat, although Smits (2007) also raises concerns that our estimation of hippoboscid flies would have been more reliable had we used insecticidal dusting, rather than counts of flies visible to the observer during direct searches of the plumage. Although dusting is commonly used to assess ectoparasite loads, trials of dusting robins yielded few hippoboscid flies as all flew off before the insecticidal powders took effect (K. A. Hale & J. V. Briskie, unpubl. data). Alternative methods involve gassing birds in enclosed chambers but such techniques are not without biases and the long holding times required would have interfered with the PHA tests. The small size of robins, and the relatively large size of the hippoboscids, meant that the plumage of robins could be searched within a few minutes, thus limiting stress that would be associated with long processing times. Workers studying larger birds, or with interests in other ectoparasites, may find that dusting or gassing are the best methods. Again, some critical assessment and preliminary trials of potential methods need to be considered before applying standard methods to new species or new situations.

When faced with a selection of tests that vary in their efficacy, a researcher needs to weigh the relative merits of each protocol and assess any trade-off between accuracy and feasibility. Estimating counts of white blood cells (WBC) and ratios of heterophils to lymphocytes (H:L) are one such class of fitness measures, and Smits (2007) outlines three techniques and their relative levels of accuracy. Our method of counting WBCs from blood smears requires minimal equipment and is thus attractive when sampling in remote locations without electricity or reliable transport to laboratory services. However, this method can be subject to errors from poor smear quality and interpretation of changes in leukocyte types (Smits, 2007). Methods of sampling WBCs that rely on rapid post-sampling analyses using hematocytometers may overcome these shortfalls but are themselves subject to errors in sample quality due to variation in shipping times, and use of dilutents and anticoagulants (Fudge, 2000). WBC counts from smears tend to be more variable under controlled lab conditions than hematocytometers (Fudge, 2000), but this would make tests between populations more conservative. Thus, the consistent differences that we found between robin populations likely underestimate the real effect of a bottleneck. Although researchers opting for use of blood smears may lose power in statistical tests, they remain an attractive option when other tests are logistically difficult. Nevertheless, as Smits (2007) warns, such results are best viewed in conjunction with background information on healthy individuals. For species that have passed through severe bottlenecks, this may be impossible as no populations may be left to act as healthy controls and comparisons may need to be made with closely related and non-bottlenecked species.

The majority of tests currently available to measure the fitness costs of bottlenecks rely on indirect measures. Estimates of parasite loads, blood cell counts and assays to foreign antigens are only surrogates of how an animal might respond when faced with a life-threatening pathogen. Although it has been assumed that an animal responding weakly to PHA would fare poorly against a real immunological threat, direct tests of the response of bottlenecked birds to a live pathogens are lacking and raise ethical dilemmas: Should endangered animals be subjected to direct tests of disease resistance that might led to unacceptable levels of mortality or introduce novel pathogens into their environment? Such tests need to be conducted but perhaps captive populations of common species or bottlenecked populations of non-endangered species may provide more acceptable models (Briskie, 2006).

Perhaps the greatest limitation of our study is that we sampled the immune response of robins in only one bottlenecked population, which limits our ability to generalize (Hawley, 2007; Tompkins, 2007). Ideally, multiple populations of known demographic history need to be sampled as replicates, but such opportunities are limited in wild animals. Other island populations of robins occur in New Zealand, although variation in the number of founders, island size, climate, age of the population and differing subspecies involved might make it difficult to disentangle bottleneck effects from these other factors. Other species, both native (e.g. saddleback Philesturnus carunculatus) and introduced (Briskie, 2006), could provide other opportunities to study multiple populations, although they likewise suffer from a variety of confounding factors. Captive populations may ultimately be most useful for providing the most robust estimates of how fitness is affected by population bottlenecks, although studies of wild animals in their native habitat are still needed to put theory into practice. The challenge for conservation biologists trying to understand the consequences of bottlenecks is how to conduct the needed experiments on a wide variety of endangered animals without compromising either the science or the welfare of the animals.