On the relationship between T-cell mediated immunity in bird species and the establishment success of introduced populations




  • 1The introduction of animals to novel environments has been used as a model system for investigating the factors affecting small populations during their initial period of establishment. Previous studies of introduction success in birds have identified a number of factors associated with successful establishment.
  • 2We predicted that species with strong non-specific immune responses would have an elevated probability of establishment because they would be better able to cope with parasites in the novel environment. Consistent with this prediction, we found that nestling T-cell mediated immune response, as reflected by the response to a challenge with the mitogenic lectin phytohaemagglutinin, was a reliable estimator of establishment success. In multivariate analyses that took previously identified predictors into account, this was only the case when propagule sizes of introduced birds were large.
  • 3These findings suggest that host–parasite interactions can be an important component influencing the fate of small populations in novel environments only when severe disease or virulent parasites are more likely to have been introduced as well.


Introductions of non-native organisms have occurred in large numbers across the globe, often with dramatic effects on native faunas (Vitousek et al. 1997). In particular, avian introductions have become a model system for the study of what occurs when small populations attempt to establish in a novel environment (Lockwood, Moulton & Anderson 1993; McLain, Moulton & Redfearn 1995; Green 1997; Legendre et al. 1999; Blackburn & Duncan 2001; Duncan et al. 2001; McLain, Moulton & Sanderson 1999; Moulton 1985, 1993; Moulton & Pimm 1983; Veltman, Nee & Crawley 1996; Sorci, Møller & Clobert 1998; Moulton, Sanderson & Labisky 2001). A large number of different factors associated with establishment success have been identified, and this has allowed inferences to be made concerning the potential processes involved in the initial establishment of small populations. Non-native introductions can also be considered to represent a model system for the study of what occurs when environmental conditions change rapidly, thereby causing a mismatch between organism and environment. This has obvious implications for studies of global change.

Parasites have been implicated to play a role in causing extinction of small populations (Dobson & May 1986; Dobson 1995; Waldman & Tocher 1998), although the role of disease in the regulation of animal populations in general remains relatively unexplored. The argument has been that such populations may be particularly prone to the fitness costs of parasitism, giving rise to an extinction vortex. However, since transmission of disease is influenced by patterns of aggregation in host populations (Dobson & Poole 1998), with transmission increasing with host density (Møller, Allander & Dufva 1993), small populations with low densities may actually experience reduced risk of disease. A second mechanism causing an increase in risk of extinction of small populations is their degree of inbreeding, which may impact on the ability to raise efficient anti-parasite defences. Several studies have indicated that inbreeding results in a reduction in the ability to raise immune responses, and that this may have serious fitness consequences. For example, pocket gophers Thomomys bottae from isolated populations showed reduced variability in major histocompatibility complex (MHC) haplotypes compared to a larger population (Sanjayan et al. 1996), although the fitness effects of this reduction in genetic variation remained hypothetical. Similarly, studies of MHC in bighorn sheep Ovis canadensis, Australian bush rats Rattus fuscipes and Gila topminnows Poeciliopsis occidentalis all show reduced genetic variability in small populations (Seddon & Baverstock 1999; Hedrick et al. 2001a, 2001b). This may have consequences for the persistence of such populations in the long run.

Hosts defend themselves against parasites by having evolved a number of different defences. The most sophisticated defence system is the immune system (Klein 1990; Roitt, Brostoff & Male 1996; Wakelin 1996). Specific defences are those that provide defence against a particular parasite strain such as particular MHC haplotypes (Klein 1990; Roitt et al. 1996; Wakelin 1996). Non-specific defences are those that provide general defence against many different kinds of parasites such as T-cell mediated immunity (Klein 1990; Roitt et al. 1996; Wakelin 1996). There is some evidence suggesting that individuals with stronger non-specific immune responses indeed have a higher probability of survival than conspecifics with weak responses (Saino, Calza & Møller 1997; Christe, Møller & de Lope 1998; Christe et al. 2001; Birkhead, Fletcher & Pellatt 1999; González et al. 1999; Soler et al. 1999; Merino, Møller & de Lope 2000). Likewise, stronger levels of defence have evolved in species where parasites have a greater impact on reproductive success (Martin et al. 2001; Møller and Erritzøe 2002).

We can make two different predictions about the importance of immune response for introduction success.

  • 1A large response implies a history of large impact of parasites on host fitness. For example, a study of T-cell mediated immune response in birds has shown that species with greater nestling mortality due to parasites have evolved stronger immune responses (Martin et al. 2001). A similar, but independent relationship has been reported for spleen size in birds (Møller & Erritzøe 2002). Hosts are known to have weakened immune responses during stress (von Holst 1998). Thus, bird species with strong immune responses would be predicted to harbour virulent parasites that could endanger survival during stressful periods such as those encountered during an introduction event.
  • 2A strong T-cell response would imply that individuals have a superior ability to cope with parasites encountered during an introduction. A comparative study of T-cell mediated immunity and dispersal in birds showed that longer natal dispersal distances were associated with stronger T-cell responses (Møller, Martin-Vivaldi & Soler 2003). This effect was clear for natal dispersal, while breeding dispersal, which is generally much shorter than natal dispersal, did not account for additional variance. The age-specific effects of immunity revealed that only nestling T-cell mediated immunity was significantly associated with natal dispersal distance, while adult T-cell response did not have independent effects (Møller et al. 2003a). Thus, age-dependent selection pressures on immunity (Martin et al. 2001) may render some species pre-adapted to colonization. We can predict therefore that species with strong T-cell responses in nestlings would be better colonizers than species with weak responses.

Here we provide a test of these two predictions, using an extensive data set of introduction success of land-bird species (Cassey 2002a). We used previously described predictors of introduction success as independent variables, while adding T-cell mediated immune response of nestlings and adults as additional predictors. The predictions were first tested in univariate analyses, which were followed by a multivariate approach, where we took similarity due to common ancestry and previously identified predictors of non-native avian establishment into account.


data sets

An introduced species is one that has been intentionally transported and introduced free-living outside its naturally occurring geographical range. For the present analysis we include only bird species that are considered to have long-term terrestrial populations and are not dependent on the ocean for feeding. The outcome of each ‘introduction attempt’ was defined by the success or failure of an introduction of a single species to a single mainland state/territory or oceanic island regardless of the number of events or the outcome. For comparability we have included in the present analyses the life history and ecological variables that we have previously found significantly to affect avian introduction success (Cassey 2002a). The dataset includes 38 species from 280 introduction attempts (Appendix 1).

For each introduced species, five variables relating to predictions of introduction success were included in the analyses: average female body mass (measured in grams and log transformed), annual fecundity (the product of modal clutch size and average number of broods and log transformed), habitat generalism (the number of major habitat types described in the range of a species (this value can range from 1 to 7) in no particular order: mixed lowland forest; alpine scrub and forest; grassland; mixed scrub; marsh and wetland; cultivated and farm lands; and urban environments), migratory habit (index of year-round occupancy of a site: 0, sedentary; 1, nomadic and/or local movement; 2, partial migrant; 4, migrant), and sexual plumage dichromatism (dichromatic if there are any differences between sexes in their colour and/or pattern of ornamentation). Propagule size is the total recorded number of individuals released at each location, log transformed for analysis. The enumeration of attempted land bird introductions follows the methods that are described in Cassey (2002b).

t-cell mediated immunity

For tests of T-cell mediated immunity we obtained data from either the published literature or from adult birds captured in mist nets, while searching for nests with nestlings, during the breeding seasons in May–June 2000–2002 in Northern Jutland, Denmark. T-cell mediated immune response to a challenge with phytohaemagglutinin was used as a measure of immune function. Similar amounts of mitogen were injected to estimate the T-cell mediated immune response and it was measured in exactly the same way in all species. Published estimates were obtained by different research groups, but intraspecific variation in T-cell response is much smaller than interspecific variation (Tella et al. 2002) and comparative studies of T-cell response have shown that estimates of mean responses are significantly repeatable among studies (Tella et al. 2002). This is a standard estimate of the ability to produce a T-cell mediated immune response (Goto et al. 1978; McCorkle, Olah & Glick 1980; Parmentier et al. 1993; Dietert et al. 1996), that in intraspecific studies of birds is strongly correlated with individual survival (Møller & Saino 2004). Comparative studies of T-cell response and parasite-induced mortality show a positive association between immune response and mortality (Martin et al. 2001), implying that species with stronger impact of parasites on hosts have evolved stronger defences against parasites. Injection with phytohaemagglutinin results in local activation and mitogenic proliferation of T-cells, followed by local recruitment of inflammatory cells and major histocompatibility complex molecules (Goto et al. 1978; Abbas, Lichtman & Pober et al. 1994; Parmentier, de Vries Reilingh & Nieuwland 1998). Birds were injected with 0·05 mL of 0·2 mg phytohaemagglutinin (PHA-P) in one wing web and 0·05 mL of physiological water in the other wing web at pre-marked sites indicated by a mark with a water-proof pen. The dose of PHA used in this study is similar to that used in numerous other studies of free-living or captive birds (Lochmiller, Vestey & Boren 1993; Saino et al. 1997; Christe et al. 1998, 2000, 2001; Birkhead et al. 1999; Brinkhof et al. 1999; González et al. 1999; Hõrak et al. 1999; Soler et al. 1999; Merino et al. 2000). We measured the thickness of the patagium injected with phytohaemagglutinin and with physiological water before injection, and again after 6 h in adults in captivity, or after 24 h in nestlings, using a pressure-sensitive caliper (Digimatic Indicator ID-C Mitutoyo Absolute cod. 547–301 Japan), with an accuracy of 0·01 mm. Although estimates of T-cell mediated immune response traditionally have been recorded 24 h post injection, we measured responses after 3, 6, 12, 24, 36, 48 and 72 h in a study of captive house sparrows Passer domesticus, and found no significant increase after 6 h (Navarro et al. 2003). Responses of the birds from the non-breeding season in the present study measured after 6 and 12 h were strongly positively correlated (Pearson r = 0·88, n = 134, P < 0·001), with no significant increase in response after 6 h (paired t-test, t = 0·82, d.f. = 133, NS), justifying the use of a 6-h period for assessment of T-cell response. A similar finding with little evidence of change after an initial swelling has been reported by Goto et al. (1978) for chickens. For six species (Delichon urbica, Hirundo rustica, Parus caeruleus, Parus major, Passer domesticus, Passer montanus) with response measured after 6 h and response measured on free-flying individuals after 24 h revealed a strongly positive correlation (Pearson r = 0·998, n = 6, P < 0·0001), with no significant increase in response after 6 h (paired t-test, t = 1·29, d.f. = 5, NS). The measure of T-cell response has a very high repeatability, as shown by three independent measurements of both wing webs (first measuring the right wing web, then the left wing web, then the right again, etc.). Repeatability analyses are reported in Møller, Erritzøe & Saino (2003b). In the subsequent analyses we used the increase in the thickness of the wing injected with phytohaemagglutinin minus the increase in the thickness of the wing injected with physiological water as a measure of the intensity of the phytohaemagglutinin-induced immune response. These data were combined with published data for additional species (Fair, Hansen & Rickleffs 1999; Smits, Bortolotti & Tella 1999; Møller & Petrie 2002).

Stronger differences in immune response have been shown among species than within species when multiple measurements exist for the same species (Møller et al. 2003b). Since the differences among species are older than the age of acclimatization societies, then it is likely that they have remained of a similar magnitude for a very long time.

statistical analyses

We used the GLIMMIX macro in SAS v 8·2 (Littell et al. 1996) to fit generalized linear mixed models (GLMMs) specifying a binomial error distribution and logit link function, with introduction outcome (0 = failure, or 1 = success) modelled as the response variable. To account for clustering of the likely statistical non-independence within the taxonomic hierarchy (Harvey & Pagel 1991) we used a GLMM following the methods established by Blackburn & Duncan (2001). Species were classified using the taxonomy of Sibley & Monroe (1990, 1993). This is the most extensive avian taxonomy available and relatively uncontentious for the species that have been introduced globally (Cassey 2002a). GLMMs provide a framework for analysing data in which observations are likely to be correlated due to clustering and cannot therefore be treated as statistically independent units (Collett 2002). GLMMs incorporate information on such clustering to provide estimates of standard errors corrected for this non-independence, which will generally be more conservative than estimates obtained if the clustering is ignored. We modelled the likely non-independence of introductions of the same taxa by assuming a common positive correlation between introduction outcomes involving the same taxa (species, genera, families and orders), but a zero correlation between introduction outcomes involving different taxa (a variance components model). Clustering of introduction events within island and mainland regions were similarly modelled. The remaining predictor variables were included as fixed effects. Multivariate models were constructed using a manual forward stepwise procedure (α = 0·05).


Having controlled for the effects of taxonomic non-independence, univariate GLMMs revealed that introduction success was significantly related to increases in habitat generalism, propagule size, and nestling T-cell response (Table 1). We subsequently found that the multivariate model that best described the variability in establishment success among introduction attempts included parameters for increasing habitat generalism (estimate = 1·11, SE = 0·25) and increasing propagule size (estimate = 0·71, SE = 0·20).

Table 1.  Fixed effects estimated from univariate generalized linear mixed models (GLMM) with establishment success (or failure) as the response variable for avian introductions. Positive parameter estimates indicate that larger values of the fixed effect are associated with a higher probability of establishment success, accounting for non-independence in the data owing to the taxonomic clustering of introductions as modelled by the random effects
Fixed effectParameter estimateStandard errort statistic
  1. * P < 0·05, **P < 0·01.

Life history
Log10(Body mass)   0·400·69   0·59
Log10(Annual fecundity)   1·551·77   0·88
Habitat generalism   1·110·25   4·39**
Migratory habit−0·400·33−1·21
Sexual monochromatism   0·300·59   0·51
Introduction event
Log10(no. of propagules)   0·730·18   4·15**
Immune response
Nestling T-cell response   0·750·40   1·88*
Adult T-cell response   2·964·35   0·68

Although the present model contains less than 20% of the introduction attempts from previous analyses, the establishment success of attempted introductions does not differ across families (χ2 = 4·42, d.f. = 9, P = 0·88), for the different sized datasets. Thus, the subset of introduction attempts for which indices of T-cell mediated immunity and propagule size data are available do not represent a biased group with regard to their probability of establishment success. Controlling for the effects of habitat generalism (as a fixed covariate) and taxonomy (as nested random effects) we found that nestling T-cell mediated immune response significantly increased establishment success for introductions of more than 100 individuals (n = 26) but not for introductions of either less than 10 individuals (n = 42) or between 10 and 100 individuals (n = 32) (Table 2).

Table 2.  Fixed effect of nestling T-cell response estimated from three multivariate generalized linear mixed models (GLMM) with establishment success (or failure) as the response variable for avian introductions and controlling for the effect of habitat generalism and taxonomic clustering. Positive parameter estimates indicate that larger values of nestling T-cell response are associated with a higher probability of establishment success
Fixed effectParameter estimateStandard errort statistic
  • *

    P < 0·05.

Log propagule size (< 1·0)
Nestling T-cell response0·570·960·59
Log propagule size (< 2·0)
Nestling T-cell response1·002·050·49
Log propagule size (> 2·0)
Nestling T-cell response7·603·911·94*


We investigated the relationship between an estimate of T-cell mediated immune response in nestling and adult birds and establishment success of these species during non-native introduction events. Univariate analyses indicated that species with stronger nestling T-cell responses were more successful in becoming established. In addition, when previously identified variables were accounted for it was revealed that T-cell response of nestlings was only important for influencing the establishment success of introduction attempts with large propagule sizes. These findings show that the ability of hosts to raise a non-specific immune response when challenged is associated with increased probability of establishment in a novel environment albeit only as an interaction with increasing propagule size.

Previous studies have shown that intraspecific variation in T-cell mediated immune response in both nestling and adult birds is a significant predictor of survival (Saino et al. 1997; Christe et al. 1998, 2001; Birkhead et al. 1999; González et al. 1999; Soler et al. 1999; Merino et al. 2000; Møller & Saino 2004). This implies that individuals are more likely to survive when they are able to produce strong T-cell responses. In addition, a comparative analysis of variation in T-cell response among species of birds has demonstrated that species with strong responses suffer from greater parasite-induced nestling mortality than species with weak responses (Martin et al. 2001). This finding suggests that hosts have evolved strong immune responses as a consequence of selection pressures imposed by parasites. Apparently, parasites are ahead of hosts in their coevolutionary interactions because hosts with strong immune responses still suffer more from the impact of parasites than hosts with weak responses.

We tested two predictions concerning the relationship between immune response and establishment success. First, since a strong immune response indicates a long evolutionary history of parasite-mediated natural selection, such species would be particularly susceptible to parasitism during periods of stress, as encountered during an introduction event. von Holst (1998) provides an extensive review of the evidence of immune suppression during periods of stress. Remember that many species were introduced to a novel environment after having spent months in a cage onboard a ship that provided transport from one continent to another. Indeed, some introductions, like that of the nightingale (Luscinia megarhynchos) to New Zealand (Thomson 1922), actually failed during transport before the species could even be successfully released. Second, hosts with an ability to raise a strong immune response to a novel challenge were predicted to be better able to cope with a novel environment, as encountered during an introduction. Our data clearly support the second prediction, while rejecting the first one. This implies that parasites that normally impose strong selection pressures on hosts do no cause particularly severe mortality during an introduction event, since bird hosts with strong T-cell immune responses suffered more from parasite-mediated selection than hosts with weak responses, but were still better able to cope during introduction.

We found evidence of T-cell mediated immune response by nestlings being a better predictor of establishment success than T-cell response of adults of the same species. This finding suggests that different selection pressures have moulded immune responses of nestling and adult birds. Age-specific selection pressures are also suggested by an analysis of the relationship between life history and immunity in birds (Martin et al. 2001). We suggest that the mechanism, which causes superior establishment success of bird species with strong T-cell responses, is a pre-adaptation to dispersal. Møller et al. (2003a) have found that T-cell mediated immunity in different species of birds predict natal dispersal distance much better than breeding dispersal distance. Since natal dispersal distance is much longer than breeding dispersal distance, we can argue that adaptation to dispersal in terms of ability to raise a non-specific immune response would mainly have acted in association with natal dispersal. This suggestion is also consistent with the observation that immune defence organs that disappear during juvenile life such as the bursa of Fabricius and the thymus do so after natal dispersal, but before breeding dispersal once juveniles have encountered almost all their environments (Møller & Erritzøe 2001). Thus, species that have evolved strong immune responses because they disperse far are also better able to endure the challenges of parasitism in a novel environment, and therefore become successfully established.

Our result has been deduced theoretically by Drake (2003), who showed that differences in the probability of establishment should only occur for large propagule sizes and in cases of very severe parasite loading. It is clear that no trait universally predicts establishment success of introduced birds, at low propagule sizes the influences of random factors are well documented (Moulton & Sanderson 1997; Cassey 2002a). At larger propagule sizes there is a greater probability (given that the prevalance of virulent parasites is generally low, e.g. Combes 2001) that severe disease or virulent parasites are more likely to have been introduced as well. This is in accordance with the observation that host species with very weak immune responses were less successful during introduction to environments where parasites are predicted to be relatively benign, and suggests that the main cause of introduction failure among large inoculates was chronic infection that may have re-emerged during the introduction event (Steadman et al. 1990).

Finally, conclusions derived from comparative analyses have to be treated with caution due to the effects of potentially confounding variables. T-cell mediated immune response has been shown to be related to nest site (stronger responses in hole nesters than in open nesters), sociality (stronger responses in colonial than in solitary species), and hemisphere (stronger responses in the southern than the northern hemisphere) (Martin et al. 2001; Møller et al. 2001). Therefore, we included these variables in the analysis and tested for evidence of multicollinearity and predictive significance with population establishment success. Consideration of these variables, however, did not change the overall conclusions in Table 1 in any way. Thus, we can be confident in our finding of an immune response measure being a genuine predictor of establishment success of birds populations introduced to novel environments. This effect was independent of previously identified predictors, only when the size of the introduced propagule was large. Nestling immune response was a better predictor of success than adult response. We suggest that immunological adaptations of hosts to long-distance dispersal make certain species more likely to succeed in becoming established in novel environments.


We are grateful to Tim Blackburn and John Ewen for comments on this manuscript. Ken Norris and two anonymous referees greatly improved an earlier version. P.C. acknowledges assistance from the French Ministry of Education and Research, Action Concertee Incitative ‘Jeunes Chercheurs 2000’, awarded to the group ‘Eco-Evolution Mathematique’ and the French Ministry for Environment, Action Concertee Incitative ‘Invasions Biologiques’, and the Leverhulme Trust (grant F/00094/AA).


Appendix 1

Table 3.  Species composition (n = 38) used in the analysis of establishment success of introduced populations. Species-specific data on nestling T-cell mediated immunity was available for species marked by an asterisk. Number of events (n = 280) are the sample sizes for which introduction-specific information was available and establishment success is the global success of the species from Cassey (2002a). Estimates and standard errors are given for the species random effect solutions in the multivariate model that best described establishment success
GeneraSpeciesNumber of eventsEstablishment successEstimateSE
Family: Phasianidae
Alectorisbarbara 80·26−0·110·50
Alectorisgraeca 60·71   0·180·49
Bambusicolathoracica 20·27−0·020·52
Chrysolophuspictus 70·14   0·180·52
Pavocristatus* 40·50   0·180·52
Perdixperdix250·37   0·440·40
Phasianuscolchicus300·71   0·260·41
Tetraourogallus 90·00−0·090·51
Family: Odontophoridae
Oreortyxpictus 30·20−0·050·51
Family: Anatidae
Anasplatyrrhynchos 20·67−0·030·52
Cygnusolor 50·93−0·190·50
Family: Psittacidae
Psittaculakrameri 10·79   0·100·51
Family: Tytonidae
Tytoalba* 60·75   0·080·49
Family: Columbidae
Geopeliastriata* 90·91   0·220·49
Family: Muscicapidae
Lusciniamegarhynchos 50·00−0·180·49
Turdusmerula* 70·62−0·090·46
Turdusphilomelos 80·42   0·050·45
Family: Sturnidae    0·51
Graculareligiosa 10·33   0·150·51
Family: Sylviidae
Leiothrixlutea 20·45−0·110·50
Family: Alaudidae    0·42
Alaudaarvensis*140·75   0·370·42
Family: Passeridae
Euplectesorix 10·50−0·020·52
Lonchuracastaneothorax 20·55−0·050·51
Lonchuramalacca 20·67−0·070·51
Passermontanus* 50·69   0·050·47
Family: Fringillidae
Cardueliscannabina 50·00−0·310·48
Cardueliscarduelis* 80·46   0·120·45
Carduelischloris* 60·64   0·050·47
Emberizacitrinella* 40·33   0·050·48
Fringillacoelebs* 60·30   0·110·48
Pyrrhulapyrrhula 20·00−0·050·51
Serinusmozambicus* 10·63   0·130·51