Urban areas are the fastest growing land cover globally and are projected to increase a further 30% by 2030 (Seto et al., 2012). The modification of the natural landscape caused by urbanization poses novel challenges to wildlife. For example, urbanization is associated with altered food webs, parasite communities, and chemical, light and noise pollution (Grimm et al., 2008; Alberti, 2015; Isaksson, 2015). To thrive in these novel environments requires changes in behavior and physiology at the phenological and, possibly, genetic level. Such changes have, indeed, been found in urban populations of several species, indicating that they accommodate some of the challenges of city life (Isaksson, 2010; Munshi-South and Kharchenko, 2010; Dominoni et al., 2013a, 2016; Nemeth et al., 2013; Atwell et al., 2014; Gil et al., 2014; Giraudeau et al., 2014; Davies et al., 2016; Watson et al., 2017). Nonetheless, urban environments are linked to reductions in fitness and health in a substantial number of organisms (Chamberlain et al., 2009; Murray et al., 2015). Hence, urban habitats are sometimes considered ecological traps which attract organisms, for example by anthropogenic food availability, but are insufficient for supporting successful rearing of offspring (Plummer et al., 2013; Sumasgutner et al., 2014; Lamb et al., 2017). There is a need, therefore, to understand the mechanisms by which urbanization reduces reproductive success.

Despite their relevance for understanding urban adaptation, the proximate physiological mechanisms that underpin differences in behavior, life histories and fitness between urban and rural populations of wild animals still remain largely elusive (but see Dominoni et al., 2013b; Atwell et al., 2014; Davies et al., 2016; Salmón et al., 2016). Whereas some physiological responses in urban animals have been investigated thoroughly, such as reproductive (Schoech et al., 2004; Partecke et al., 2006; Dominoni et al., 2013b) and stress physiology (Bonier, 2012), others have received little attention. Recently, a study using transcriptome analysis (RNA-seq) has broadly examined the ways physiological systems differ between an urban and rural population of a songbird, the great tit (Parus major) (Watson et al., 2017). Among the systems highlighted in this study, immunity showed particularly clear differences, corroborating reports of substantially reduced health of urban populations in several species (Martin et al., 2010; Isaksson, 2015; Murray et al., 2015). Given the central role of the immune system for determining fitness and given its sensitivity to environmental factors, differences in immunity are thus strong candidates for linking urbanization to reduced reproductive success (Martin et al., 2010; Isaksson, 2015; Watson et al., 2017).

Urban environments can affect the immune system by modulating the structure of the parasite community to which animals are exposed and potentially respond to. Cities have been associated with altered parasite communities and altered infection dynamics (Fokidis et al., 2008; Evans et al., 2009; Giraudeau et al., 2014; Neiderud, 2015). For example, foraging behavior in cities is thought to enhance the prevalence of intestinal coccidians (Giraudeau et al., 2014), potentially driven by increasing contact rates at feeders (Becker et al., 2015). Several other factors associated with urban environments can affect the activation of particular immune pathways of wild animals. Prevalent urban conditions, such as chemical pollution and artificial light at night, can act as environmental stressors influencing the immune system and promoting pro-inflammatory immune profiles (Halliwell and Gutteridge, 2002; Bedrosian et al., 2011; Fonken et al., 2013; Isaksson, 2015; Watson et al., 2017). Low-quality and changed diet in urban areas can also enhance pro-inflammatory immune processes, which in turn can impose behavioral and physiological costs on animals (McGraw et al., 2006; Andersson et al., 2015; Becker et al., 2015; Isaksson, 2015). Furthermore, nutritional stress and resource restriction can exacerbate trade-offs between different biological processes, for example through resource allocation to growth vs. immunity (Sheldon and Verhulst, 1996; Norris and Evans, 2000). The findings of the transcriptomics study by Watson et al. (2017) showed that in an urban songbird population, where some of the above factors are effective, inflammatory immune genes were indeed upregulated. Experimental studies that measure candidate genes are now needed to clarify the origin of such differences, and links to fitness need to be established.

Here we examined how urban life affects avian immunity and fitness traits. As a study species, we used another parid songbird that is common in city and forest environments, the blue tit (Cyanistes caeruleus). Blue tits reportedly show reduced reproductive investment (e.g., clutch size and egg size) and reduced reproductive success in urban compared to rural populations (Chamberlain et al., 2009; Bailly et al., 2016b). We experimentally studied blue tits at two sites, one in an urban park area in Glasgow (UK), and one in a National Park 40 km away. The same populations were simultaneously investigated in a twin study on food resource availability, food provisioning and stable isotope signatures in eggs and blood samples (Pollock et al., 2017). This parallel study design allowed us to link, for the same sites, comparisons of immunity with resource restriction. We first tested the hypothesis that adult urban blue tits show increased expression of genes associated with inflammation compared to blue tits from a forest habitat. Specifically, using RT-qPCR, we measured the transcript levels of TLR4 and LY86, which are involved in anti-bacterial and anti-malarial responses (Medzhitov, 2001). We also tested the type 2 transcription factor GATA3, which is central to innate and adaptive immunity against parasitic helminths and immune homeostasis (Wang et al., 2011; Tindemans et al., 2014) and showed elevated gene expression in urban great tits (Watson et al., 2017).

We then investigated whether any urban-forest differences in immune gene transcript levels are already present during the nestling stage. Urban-forest immune differences in nestlings could arise either through environmental effects during early life, genetic or maternal effects. Thus, we cross-fostered eggs between and within the urban and forest sites to distinguish whether any difference in immune activation of nestlings was induced by the environment or controlled by genetic or maternal mechanisms (Watson et al., 2017). Finally, from the cross-fostering experiment, we quantified the reproductive success at the two sites and, thus, investigated possible associations between environment, immunity and reproductive success. We also examined prevalence of a common avian disease, avian malaria, as a potential driver of differences in expression of our immune markers (Martin et al., 2014; Videvall et al., 2015).

Increased levels of TLR4 and LY86 in adult urban birds are predicted on the basis of the hypothesis that the urban environment enhances inflammation. We also predicted increased levels of GATA3 in adult urban birds as a consequence of high activation of immune pathways against intestinal parasites, whose transmission is thought to be exacerbated by anthropogenic food provisioning at bird feeders (Becker et al., 2015). Expression of these three genes was also reported to be increased in at least one tissue in the transcriptome study on great tits (Watson et al., 2017). For nestlings blue tits, we predicted that urban-raised offspring might also show heightened expression of immune genes, indicating a direct effect of the environment on their immune profiles within their short postnatal life-span. We further predicted that cross-fostered nestlings should also differ in their immune gene expression profiles by origin, either due to micro-evolutionary change or based on differences in maternal investment. Assuming that higher expression levels in immune genes were selected for in urban birds, we expected higher expression levels in nestlings originating from eggs laid the urban environment regardless of their rearing environment. We also expected lower reproductive investment, fledging success and fledgling body mass in blue tits from the urban compared to the forest environment. The differences in reproductive success of urban and forest nests were indeed striking: 88% fledgling success in the forest compared to only 30% in the city. Consequently, in the cross-fostering experiment we obtained robust sample sizes on gene expression only for nestlings reared in the forest. We acknowledge limitations of our study arising from low sample sizes and from a design that involves only two sites (one urban and one rural) in a single year. Our experimental approach, using cross-fostering, addresses some of these concerns by effectively providing within-site replication. However, urban ecology will require studies in multiple sites and years, and on a broad range of organisms, before major, generalizable advances can be achieved (Watson et al., 2017).

Reproductive success of blue tits was dramatically reduced in the city compared to the forest. Our study adds to existing evidence that urban environments commonly impair fitness and health in wild populations (Chamberlain et al., 2009; Murray et al., 2015; Bailly et al., 2016b). In most avian studies, fledging success and nestling weight were lower in urban than in more rural habitats (Chamberlain et al., 2009; Bailly et al., 2016b; Sprau et al., 2017). These findings are matched by observations in this study and in the parallel study of effects of food availability on other nests at our sites (Pollock et al., 2017). In addition to immediate effects of the environment on developing offspring, such differences could also reflect prenatal maternal investment (e.g., differences in egg composition, Toledo et al., 2016) or genetic differences between urban and rural populations (Mueller et al., 2013). Our cross-fostering experiment does not support these latter hypotheses as we found no strong effects of natal origin on reproductive and fitness traits (Figure 3). Slightly larger cross-fostered eggs from both sites were a spurious outcome of our alternating, experimental swaps. Overall, we provide experimental support for the existence of a significant negative postnatal effect of the urban environmental on breeding success in blue tits, although we acknowledge that further replication will be needed to consolidate these results.

These findings converge with a recent study on closely related great tits which also used a cross-fostering experiment and showed that negative implications of urban nesting (in this case, shortened telomeres) arose from the raising environment and not from genetic or maternal factors (Salmón et al., 2016); however, this study did not investigate the possible mechanistic causes of such differences. Our parallel study on food availability and provisioning behavior provided clear evidence of resource limitation and nutritional stress in our urban site compared to our forest location (Pollock et al., 2017). Whereas nestlings in the forest received predominantly caterpillars, the proportion of this preferred, high-quality diet was substantially lower for nestlings in the city. Pollock's data further suggest that parents partly resorted to other food, possibly including anthropogenic sources, to feed their nestlings and presumably themselves (Pollock et al., 2017). Shifts in diet, which were also reported for urban populations in other species (e.g., Murray et al., 2015), can have important effects on the nutritional state of urban animals, and consequently, also on their health (Andersson et al., 2015). Thus, the findings that we present here on immunity have to be interpreted against the backdrop of resource restriction in the urban site.

Our results from adult blue tits on transcript levels of immune genes support some of our initial hypotheses. Urban blue tits showed increased transcript levels of TLR4, a marker of inflammatory processes, although these findings were not paralleled for LY86. Several factors have been proposed to explain how the immune system of urban organisms is expected to change in response to the environment (Isaksson, 2015). Malaria parasite infection has been shown to affect the expression of TLR4 (Martin et al., 2014) and also LY86 in passerines birds (Videvall et al., 2015). In our study, malaria infection status probably had little importance for TLR4 and LY86 levels across habitats. In contrast to other studies (Evans et al., 2009), malaria prevalence was consistently high at both of our sites; however, our statistical power was small and the lack of differences across habitat needs to be considered cautiously. Other intra-cellular pathogens found to vary in prevalence between urban and rural sites (Giraudeau et al., 2014) or in association with anthropogenic food provisioning (Becker et al., 2015), may also be important determinants of TLR4 and LY86 levels. Because we lack information on the wider pathogen assembly in our study sites, we cannot discard differences in other inflammatory pathogens as a cause of the observed gene expression patterns. Additional environmental factors could also explain differences in immunity between urban and forest populations. Low-quality and restricted diets are known to promote pro-inflammatory immune processes (Blount et al., 2003; McGraw et al., 2006; Isaksson, 2015; Nettle et al., 2017). Given our parallel findings of significant diet differences between our study populations (Pollock et al., 2017), it is likely that resource limitation has contributed to elevated inflammation in the city (Larsson et al., 2004; Isaksson, 2015). Costs of inflammation under resource limitation could have contributed to the birds' low reproductive success via exacerbated physiological trade-offs.

In contrast to TLR4, transcript levels of LY86 in adult blue tits did not match our predictions. Because TLR4 and LY86 interact (Lee et al., 2012), we expected to find a correlated pattern of expression between their gene expression. However, the interaction between these molecules is complex and can vary between cell types (Divanovic et al., 2005). The lack of differences in LY86 across habitats, in contrast to our findings regarding TLR4, indicates that these two genes might respond differentially to the urban environment. TLR4 expression may be particularly sensitive to urban-related environmental stressors. As well as acting as a receptor for pathogens (gram negative bacteria molecular patterns), TLR4 is implicated in the recognition of damage-associated molecular patterns that follow tissue damage or cellular apoptosis (Liu et al., 2014). Conceivably, TLR4 could have been additionally enhanced by urban-associated environmental factors that promote oxidative stress and tissue damage, for example, air pollution and artificial light at night (Isaksson, 2010; Fonken et al., 2013). Our findings on TLR4 and LY86 transcript levels can be compared to recent findings of the transcriptomic comparison between urban and rural populations of another parid, the great tit (Watson et al., 2017). In fully grown great tits, expression of TLR4 (gene ID ENSTGUG00000003342) tended to be elevated in blood and liver also in the city compared to the forest (Watson et al., 2017, Supplementary Datasets 1 and 2). Expression of LY86 (gene ID ENSTGUG00000002305) also tended to be higher in liver of urban great tits, but not in blood. Largely, therefore, patterns were similar in the two parid species.

The findings of higher GATA3 transcript levels in adult urban compared to forest blue tits also followed our initial predictions. The type 2 transcription factor GATA3 constitutes a major regulatory component of the immunity against helminth parasites (Tindemans et al., 2014). Although we were unable to monitor intestinal parasite load in our study, our results could be explained by potentially higher helminth pressure in our urban location. Aggregation in feeding stations causes high rates of horizontal transmission of directly-transmitted parasites and, hence, affects parasite distribution (Becker et al., 2015). If increased activation of GATA3-induced immunity in our urban birds was, indeed, caused by increased helminth infection, it might thus reflect a parasitological cost imposed by supplementary feeding and contribute to some of the negative effects of supplementary feeding on fitness (Plummer et al., 2013). Results for GATA3 expression (gene ID ENSTGUG00000002134) from an urban great tit study in Sweden were comparable for liver, with higher levels in the city, whereas for blood expression did not differ between sites (Watson et al., 2017).

For the nestling stage, our findings from the cross-fostering experiment provide no evidence that urban as compared to forest origin was associated with higher immune transcript levels in nestlings (Figure 2). Unfortunately, our sparse data from the urban rearing environment do not allow comprehensive conclusions from the full cross-fostering experiment (Figure S2). However, overall, transcript levels of nestling immune genes appeared to be similar across habitats (Figure S2). Thus, pending confirmation by a larger dataset, our findings suggest a differential impact of urbanization on the immune system, becoming more pronounced in later life-history stages. The contrasting patterns among age groups might be explained by chronic effects of long-term exposure to urban-specific environmental factors in adults (e.g., Bedrosian et al., 2011). Genetic or prenatal maternal causes for increased TLR4 and GATA3 transcript levels in adult blue tits are made unlikely by our findings that among nestlings reared in the forest, origin (city or forest) did not strongly affect transcript levels of any gene (Figure 2). Hence, the cross-fostering experiment indicates that urban-forest differences in transcript levels of adult blue tits were most likely driven by the environment (Salmón et al., 2016). However, we cannot fully exclude the possibility that genetic or maternal factors are only identifiable in adulthood and not in nestlings. In contrast to our findings, a recent study on great tits did report differences in immune status between urban and rural nestlings (Bailly et al., 2016a). Interestingly, the differences were opposite to those we found for adults blue tits: using physiological assays, rather than gene expression studies, Bailly et al. (2016a) show that urban nestlings produce less haptoglobin (a marker of inflammation) than forest birds, and offer as an explanation that haptoglobin production is compromised by food resource availability in the urban site. Although caution is needed when comparing different species and inflammatory markers, the opposite findings for the age groups might reflect different trade-offs between immunity and development: under resource restriction, growing nestlings might not be able to mount costly responses in the same way as fully grown adults (Sheldon and Verhulst, 1996; Norris and Evans, 2000; Alonso-Alvarez and Tella, 2001).

Our results contribute to a body of evidence suggesting that urban living has reproductive costs and can impact the health of wild animals by altering their immune system (Audet et al., 2016; Bailly et al., 2016a; Watson et al., 2017). We acknowledge that caution is needed when interpreting our result given our sample size and the existence of only one urban-forest study pair and one study year. However, our experimental approach, with within-forest and within-city controls, allows us to draw conclusions on the causal links between immunity, fitness and the urban environment in our study system. Furthermore, the results generally confirm our original predictions and are in line with widespread evidence of reductions in fitness in urban environments (Chamberlain et al., 2009; Bailly et al., 2016b) and with recent discoveries on the effect of the urban environment on gene expression profiles in wild birds (Watson et al., 2017). We find evidence for elevated expression of immune genes in adults of our urban compared to our forest population, and data from our cross-fostering experiment indicate that such changes are best explained by environmental factors. In combination with our twin study on reduced food availability and modified provisioning in the city (Pollock et al., 2017), we tentatively identify links between resource limitation and altered immunity. To better understand the impact of urbanization on the immune system of wild animals, we need fine-scale characterization of the urban environment including parasite assemblies, coupled with a broader immune assessment of wild populations and information on fitness across life stages. Such information would reveal the role of the immune system for adaptation to urban life, as well as the long-term demographic consequences of altered immunity for urban-dwelling species.

PC-L, BH, DD, SB, and TB conceived the original study and designed experimental procedures. PC-L, DD, CP, and BH carried out field work, and PC-L, PO, MM, LW, and FB performed laboratory analysis. Statistical analyses were done by PC-L with advice from DD and BH. PC-L, BH, and DD wrote the manuscript with input from all other co-authors.