key: cord-0426537-j5ueztpe
authors: Ruhs, Emily Cornelius; Martin, Lynn B.; Downs, Cynthia J.
title: The impacts of body mass on immune cell concentrations in birds
date: 2020-04-25
journal: bioRxiv
DOI: 10.1101/2020.04.23.057794
sha: a5a88d74ac11d30b39b2d59e7c289ae6ed490103
doc_id: 426537
cord_uid: j5ueztpe

Body mass affects many biological traits, but its impacts on immune defenses are fairly unknown. Recent research on mammals found that neutrophil concentrations scaled hypermetrically with body mass, a result not predicted by any existing theory. Although this mammalian model might predict how leukocyte concentrations scale with body mass in other vertebrates, vertebrate classes are distinct in many ways that might affect their current and historic interactions with parasites and hence the evolution of their immune systems. Subsequently, here, we asked which existing scaling hypothesis best-predicted relationships between body mass and lymphocyte, eosinophil, and heterophil concentrations—the avian functional equivalent of neutrophils—among >100 species of birds. We then examined the predictive power of body mass relative to life-history variation, as an extensive literature indicates that the scheduling of key life events has influenced immune system variation among species. Finally, we ask whether these scaling patterns differ from the patterns we observed in mammals. We found that an intercept-only model best-explained lymphocyte and eosinophil concentrations among birds; body mass minimally influenced these two cell types. For heterophils, however, body mass explained over 30% of the variation in concentrations among species, much more than life-history variation (~8%). As with mammalian neutrophils, avian heterophils scaled hypermetrically (b=0.19 ± 0.05), but significantly steeper than mammals (~1.5x). As such, we discuss why birds might require more broadly-protective cells compared to mammals of the same body size. Body mass appears to have strong influences on the architecture of immune systems, which could impact host-parasite coevolution and even zoonotic disease risk for humans.

Body size influences almost all of the behaviors, physiological processes and hence eco-70 evolutionary roles of species in communities [1, 2] . Nevertheless, despite the many life-history 71 traits that we know to scale with body size and that might influence immune cell concentrations, 72 we lack information on how body mass affects defense [3] [4] [5] . Many physiological factors that go 73 hand-in-hand with body size evolution also likely impact exposure to parasites and pathogens 74 and have implications for host immune systems [6, 7] . A growing body of theory and a few data 75

propose that body size influences the immune systems of species, partly because risk of 76 infection varies with body size [6, 7, 8] . In general, higher risk in large species occurs because 77 large species cover greater physical distances with each movement (i.e., per step or beat of 78 wings) and have larger respiratory, digestive and sexual tissue surface areas for infection 79 relative to parasites [4, 8] . Large species also have distinct life-history traits from small ones 80 (e.g. social groups, home range, feeding) [9, 10], as the only way to reach large size is via long 81 development and subsequently delayed maturation, both of which necessitate a long lifespan 82 [11] . Several studies have revealed that such long-living ("slow") species (i.e. elephant; low 83 reproductive rate and long development) have different immune defenses than short-living 84 ("fast") ones (i.e. mouse; high reproductive rate and short development) [12] [13] [14] [15] , such that an 85 individual should invest in specific immune defenses if they live for long periods because they 86 may come into contact with the same pathogens repeatedly. Yet to what degree these life-87 history patterns are genuine or just echoes of body mass effects remains unclear. 88

In the present study we investigated whether and how body mass is related to immune cell 90 concentrations among birds spanning several orders of magnitude in mass (~1180 fold; Fig. 91 S1). Our particular interest was to identify which hypothesis best predicted the scaling 92 relationships for immune cell concentrations among birds (Fig. 1) , and then to test whether 93 these scaling patterns are different from those found in mammals [16] . The main parameter of 94 interest is b, or the slope of the line relating to body mass and leukocyte concentration. For 95 many traits, scaling relationships are isometric, such that changes in traits are directly 96 proportional to changes in body mass. Other traits, though, scale hyper-or hypometrically, such 97 that the trait of interest is disproportionally larger or smaller than would be expected by 98 geometry. Two hypotheses predict isometric scaling for immune cell concentrations (b = 0, 99 black dashed line in Fig. 1 ). The first is the Protecton hypothesis, which proposes that all 100 organisms require similar levels of protection, regardless of size [17, 18] . Similarly, the related 101

Complexity hypothesis [16] assumes that the time for surveillance of tissues and delivery of a 102 single leukocyte is independent of body mass [19, 20] . Another hypothesis, Rate of Metabolism 103

[21], invokes metabolic rate as the key driver of variation in immune cell concentrations [1] . 104 Thus, the rates of proliferation and actions of immune cells should be reliant on the Rate of 105

Metabolism of a host [21] . This hypothesis predicts hypometric relationships between leukocyte 106 concentration and body mass (b = -0.25, red line in Fig. 1) . A final hypothesis is derived from 107 our recent discovery about neutrophil scaling in over 250 species of mammals [16] and prior 108 theory on the evolution of similar organismal functions [22] . In our previous study [16] , we found 109 that neutrophil concentrations scaled hypermetrically with body mass (b=0.11; blue line in Fig.  110 1). This particular outcome led us to offer the Safety factor hypothesis [23], which proposes that 111 hypermetric scaling may have evolved, in particular for early anti-microbial defenses, because 112 larger species require disproportionately more rapidly-acting, broadly-protective defensive cells 113 to combat infections than small species [24] . In other words, we expect that this pattern evolved 114 because large species must prioritize risk reduction over other life-history priorities because 115 they mature more slowly and hence live longer than smaller species [11, 12, 16] . 116 117 Here, we sought to determine which hypothesis best predicted scaling relationships for three 118 leukocyte concentrations (heterophils, eosinophils, and lymphocytes) among avian species. To 119 enable comparison with published values describing how immune cells scale with body mass in 120 mammals, we used the same approach as Downs et al. (2020) . To our knowledge, there is only 121 one other study that has investigated this relationship across various species of birds; however, 122 these data are gathered from wild species and not phylogenetically controlled [25] . Therefore, 123 there is a need to understand how phylogeny and other life-history characteristics contribute to 124 allometric patterns in similarly-housed healthy individuals. Specifically, we conducted three 125 modeling exercises. First we asked which hypothesis best-predicted scaling relationships: the 126 Protecton/Complexity hypothesis (isometric slope), the Rate of Metabolism hypothesis 127 (hypometric slope), or the Safety factory hypothesis (hypermetric slope) [16] . For the Safety 128 factor hypothesis, we did not attempt to fit the same b as discovered in the mammalian study for 129 heterophils, and instead estimated the b from the data. We took this approach because birds 130 and mammals differ in many ways that could influence the magnitude (or even direction) of the 131 slope. Our second modeling exercise sought to discern whether body mass or life-history traits 152 (maximal lifespan, maximal reproductive capacity, and interaction with body mass) best 153 predicted leukocyte counts in birds. We expected body mass to predict variation in all three cell 154 types better than life-history variables, but we expected phylogeny also to be a strong predictor 155 as we and others have observed previously [16, [29] [30] [31] [32] . However, we also expected predictor 156 effects to vary among cell types based on their respective defensive functions. Our third and 157 final modeling exercise directly tested whether the scaling patterns observed in birds and 158 mammals for lymphocytes and heterophils/neutrophils differed from one another. 159

Trait data 162

We extracted species means of heterophil and lymphocyte concentrations (cells L -1 ) from whole 163 blood for 116 avian species and eosinophil concentrations for 88 avian species from 164 Species360 (Table S1) reported for each species (see Dryad for descriptions). We extracted life-history data on body 172 mass, maximum lifespan, age at maturation, inter-laying interval (i.e., how often a species lays a 173 clutch), average clutch size and incubation period from the CRC Handbook of Avian Masses 174

[35] and publicly available databases such as AnAge [36] , and the Animal Diversity Website 175 [37] . Data were compiled as best as possible using a combination of all sources so as to have 176 the most complete dataset. From these life-history data we calculated maximal reproductive 177 capacity for each species, a similar metric to our previous study of mammals [16] : 

To determine the ability of body mass to explain concentrations of heterophils, eosinophils, and 208 lymphocytes relative to life-history traits, we fitted three separate omnibus models (one for each 209 cell type) that included log10(body mass), maximum longevity, maximum reproductive capacity, 210 and all two-way interactions between log10 (body mass) and life-history variables as fixed 211 effects. 212

213

Comparing models and determining fit 219

All analyses involved phylogenetic mixed effects models in R (version 3.6.0) using the packages 220 However, species with more log10-transformed heterophils had more log10-transformed 265 eosinophils (r = 0.402, t86 = 4.07, p = 0.0001). We also found correlations between log10-266 transformed body mass and life-history traits (Fig. S3) . 267

Best-fit models for leukocyte allometry 269

The Protecton/Complexity model (Model 1) best-predicted avian lymphocyte and eosinophil 270 concentrations ( Fig. 2A) . First, the inclusion of body 278 mass improved model fit, when the slope coefficient was estimated from the data (Model 3). 279

Model 3 accounted for 77% of the variation in heterophils, with body mass alone accounting for 280 30% of that variation. In this model (but also in the omnibus model; see below), heterophil 281 concentrations scaled hypermetrically (Model 3: b, 95% credible interval = 0.19, 0.14-0.24; Fig.  282 2A). In all models for all leukocyte types, appreciable phylogenetic effects were detected (Table  283 1). 284 285

The lymphocyte and eosinophil omnibus models (model with life-history variables as fixed 287 effects) were not amongst the top models (Table 1 ; DDIC=-7.03 and -2.03, respectively). The 288 heterophil omnibus model was the overall best-fit (Table S2) , accounting for 82% of variation in 289 heterophils (Table 1) . However, the addition of life-history traits increased explanatory power of 290 fixed effects by only 8% (Table 1 ). In light of these subtle effects, we selected Model 3 (mass-291 only; estimating b from the data) as the more informative and simpler alternative about the 292 scaling of heterophils. 293 294

The bird/mammal model for lymphocytes accounted for 80% of the variation (Table S3) ; 296 however, the fixed effects, body mass and class, explained only 2% of the variation (Table S3 ). 297

The model for hetero-/neutrophils explained 92% of the overall variation (Table S3) with body 298 mass and class explaining a higher proportion of the variation (11%). Hetero-/neutrophils scaled 299 hypermetrically with body mass (b, 95% credible interval = 0.19, 0.14-0.24) and the interaction 300 between body mass and class predicted cell concentrations (Table S4) . Taken together, these 301 results suggest that the scaling coefficient for the relationship between body mass and hetero-302 /neutrophil concentrations is steeper in birds (b=0.19) than in mammals (b=0.11) [16] . 303

Here, we investigated the scaling of three types of leukocytes across 116 species of birds. 306 While our database underrepresents small birds due to biases in zoo collections, it does 307 represent 21 orders and a 1180-fold difference in size. We found little evidence for allometry for 308 lymphocyte and eosinophil concentrations; the best-fit models for both cell types did not include 309 body mass. These results are consistent with the Protecton hypothesis, but we encourage 310 caution in interpreting these results as support for it. It is difficult to use null models as evidence 311 of isometry as it is unclear whether the slope is truly zero (isometry) or if there is an absence of 312 a pattern. Heterophils, by contrast, scaled hypermetrically, although the omnibus model 313 performed better and explained more of the variation (82%) than the model with only body mass 314 as a fixed effect (Model 3). These results indicate that body mass explains much more variation 315 in heterophil concentrations (~30%), compared to life-history traits (~8%). Our most striking 316 discovery was the slope of the effects of body mass on heterophil concentration among birds (b, 317 95% CI = 0.19, 0.14:0.24). This coefficient supports the Safety factor hypothesis, but it was 318 also significantly steeper than the estimate we described previously for mammals (b = 0.11; Fig.  319 2B). Below we discuss the ramifications of these results, in particular why avian heterophils 320 scale so much more steeply than mammalian neutrophils. This difference between mammals 321 and birds was also supported by the direction comparison made in this study. The above logic for lymphocytes probably does not transfer to eosinophils. An absence of 343 scaling for eosinophils is less likely derived from functional diversity among members of this 344 class because all eosinophils perform the same type of defense. Eosinophils respond rapidly to 345 local inflammation, mostly when it is induced by macroparasites to which their defensive actions 346 are best attuned [51]. In birds, we detected no effect of body mass on eosinophil concentrations, 347 but this finding is inconsistent with work in bats [48] , carnivores [47], and primates (b=0.05 [32] ), 348 all of which detected hypermetric scaling. The largest study to date in birds also found slight 349 hypermetric scaling; however, the data were not log-transformed; (b=0.048) [25] . This 350 discrepancy could be due to at least three factors. First, as with lymphocytes, body mass might 351 have statistically significant but subtle influences on eosinophil concentrations in some taxa. 352

Second, eosinophils are more common in tissues than the blood [51] (these data are from whole 353 blood); therefore, scaling patterns might be more prominent in tissus. Third strive to test these hypotheses, but at the same time, attempts should be made to resolve 375 directly how body mass interacts with life-history traits to influence interspecific variation in 376 immunity. As above, life-history traits have been claimed to drive immune variation among 377 species and populations, but almost never have body mass effects on life-history been 378 disentangled from body mass [7] . Among the bird species studied here, we found some 379 correlations between body mass and life-history traits (Fig. S3) The majority of such differences between classes, however, would be expected to influence 390 only the intercept terms in the models we evaluated, not the slopes. In other words, whereas 391 similar-sized birds tend to live longer than mammals [9], if such a difference influenced 392 leukocyte concentrations, its effects would be expected to do consistently across body masses, 393 not disproportionately at large sizes as we observed. For most differences between birds and 394 mammals (i.e. air sacs versus lungs for breathing, egg-laying versus live birth, organ mass, 395 blood circulation/volume, and many more), differences in avian heterophil concentrations would 396 be consistently different from mammalian neutrophils across all body masses. Our discovery of 397 a fairly striking difference in the steepness of scaling between avian and mammalian 398 granulocytes and mass was unexpected; b for birds in our mass-only model (Model 3) was 399 ~1.5x the mammalian estimate from a comparable model. 400

When data are plotted on log-log scales (Fig. 2a&b) , their biological significance is often 402 obscure, so to convey better the consequence of this difference, we produced conceptual Fig. 3  403 using commonly known species at opposite ends of the extant bird species body mass 404 distribution. Using the equation from Model 3 with published values for blood density and 405 volume for birds [9, 34, 55], we first calculated total whole-animal heterophils for species. We 406 estimated that an ostrich (Struthio camelus; 111000g) circulates 33.9k-fold more heterophils than an American goldfinch (Carduelis tristis; 15g), although their body masses differ by just Table 1 . Best-fit models predicting circulating leukocyte concentrations in songbirds. Top models are represented by 0 ≤ DDIC ≥ 2. Model 1 fit a slope of 0, model 2 a slope of -0.25 and models 3 and 4 were allowed to estimate the slope given the data. Models test for the effects of body mass and life-history metrics on log10-transformed lymphocyte, eosinophil and heterophil concentrations. 

Toward a metabolic theory of ecology

The predominance of quarter-power scaling in biology

The allometry of host-pathogen interactions

Parasites, Disease and the Structure of Ecological 517

Modelling the effects of phylogeny and body 519 size on within-host pathogen replication and immune response

Scaling of Host Competence

Costs of immune responses are related to host body size and 523 lifespan

Infectious disease transmission and behavioural allometry in wild 526 mammals

Size, function, and life history

The Evolution of Life Histories

Do performance-safety tradeoffs cause hypometric metabolic scaling in 530 animals? Trends in ecology & evolution

Linking immune defenses and life history at the levels of the individual and the 532 species

Immune defense and reproductive pace of life 534 in Peromyscus mice

Life history and demographic drivers of reservoir competence for 536 three tick-borne zoonotic pathogens

Living fast and dying of infection: host life history drives 538 interspecific variation in infection and disease risk

The effects of body mass on immune cell concentrations of 541 mammals

A comparative study of leukocyte counts and disease risk in primates

ISIS physiological reference intervals for captive wildlife: a CD-ROM resource

Bloomington (MN): International Species Information System

Body masses of birds of the world. CRC Handbook of Avian Body 581

Human Ageing Genomic Resources: integrated databases and tools 583 for the biology and genetics of ageing

The animal diversity web

APE: analyses of phylogenetics and evolution 587 in R language

MCMC methods for multi-response generalized linear mixed models: the 589

The evolution of energetic scaling across the vertebrate tree of life

A general and simple method for obtaining R 2 from 593 generalized linear mixed-effects models

The phylogenetic mixed model. The 596

7 -Avian Innate Immune Responses

phyloT: phylogenetic tree generator

The percentage of T-and B-Cells in the lymphomyeloid tissue 601 of hyperimmunized bursaless chickens

Scaling aspects of lymphocyte trafficking

A comparative study of white blood cell 605 counts and disease risk in carnivores

Series B-Biological Sciences

Measures of the constitutive immune 608 system are linked to diet and roosting habits of neotropical bats

Circulating white blood cell counts in captive and wild rodents are 611 influenced by body mass rather than testes mass, a correlate of mating promiscuity

Comparative morphology, development, and function of 614 blood cells in nonmammalian vertebrates

The immunobiology of eosinophils

The avian heterophil

Oxidative stress in ecology and evolution: lessons from avian studies

Does cellular metabolism 624 from primary fibroblasts and oxidative stress in blood differ between mammals and 625 birds? The (lack-thereof) scaling of oxidative stress

Ecological scaling: mammals and birds

Mammalian skin evolution: a reevaluation

408 7.4k-fold. To capture the consequences of this allometric difference from what we observed 409 previously in mammals [16] , we tallied the number of cells in avian and mammalian species of 410 the same size. We found that whereas a mouse (Mus musculus) and a finch would maintain 411 roughly the same number of neutrophils/heterophils in their blood, an ostrich would circulate 412 2.2-fold more heterophils (1.17x10 20 cells) than a giant panda (Ailuropoda melanoleuca) 413 circulates neutrophils (5.43x10 19 cells; Fig. 3C ). 414 415 Our approach cannot reveal why large birds circulate so many more heterophils than large 416 mammals circulate neutrophils, but there are a few conspicuous possibilities. The first is that 417 the two cell types are not functionally similar after all, and for some reason, avian heterophils 418 might require disproportionate compensation for less effective function via increases in cell 419 number as birds increase in size. The second possibility involves mode of locomotion. As most 420 birds fly, a bird would presumably expose itself to more risk space for infection than a similarly-421 sized mammal. One way to test this hypothesis would be to investigate neutrophil scaling in 422 bats; such research has occurred, but allometric slopes were not estimated. Of course, though, 423 the largest birds do not fly, and investigation of their position on scaling curves does not support 424 this perspective. A third possibility entails the lack of lymph nodes and leukocyte pools in birds. 425Because birds lack defined lymph nodes, they might not be able to shuttle heterophils as 426 efficiently around their bodies, particularly at large body sizes, as mammals can do for 427 neutrophils. Until we know how lymphoid tissue is organized relative to body mass in birds, we 428 can only speculate that it might scale more hypometrically than mammals. In mammals, we 429 know that myelocytes (leukocyte progenitors) can reside in a "lazy pool," and mature neutrophils 430 can persist in a "rapid mobilizable pool" [28] . The existence of heterophil pools in birds is 431 unknown, but as we proposed for lymphoid tissue, steep hypometric scaling of the size or 432 numbers of these pools could explain hypermetric scaling of heterophils in birds. We are far from resolving why birds require so much more heterophil-mediated protection at 440 large body sizes, but we encourage additional work on the topic. It is also obscure whether the 441 patterns we detected truly represent evidence for the Safety factor hypothesis, so we advocate 442 that future work try to address this open question. Such insight will be useful for resolving how 443 and why body size affects disease ecology and evolution. We expect that additional 444 mechanistic studies will be useful, particularly the development of more precise tools for 445 distinguishing lymphocyte classes or methods that capture better the functional immune 446 variation (e.g., direct control of infections