key: cord-0284524-n9x404vn
authors: Downs, Cynthia J.; Schoenle, Laura A.; Oakey, Samantha J.; Ball, Ray; Jiang, Rays H.Y.; Klasing, Kirk C.; Martin, Lynn B.
title: Extreme hyperallometry of mammalian antibacterial defenses
date: 2020-09-04
journal: bioRxiv
DOI: 10.1101/2020.09.04.242107
sha: 7e14677b1b6018541beef0ba0263bf69d5d8c9d2
doc_id: 284524
cord_uid: n9x404vn

Terrestrial mammals span 7 orders of magnitude in body size, ranging from the < 2 g pygmy shrew (Suncus etruscus) to the > 3900 kg African elephant (Loxodonta africana). Although body size has profound effects on the behavior, physiology, ecology, and evolution of animals, how body mass affects immune functions is unknown. Here, using data from >160 terrestrial species of mammals, we show that the size-scaling of antibacterial activity in serum exhibits extreme hyperallometry. Compared to small mammals, the serum of large mammals is remarkably inhospitable to bacteria. Such hypermetric scaling of immune defenses provides novel perspectives on the ecology of host-pathogen interactions, and on their co-evolutionary dynamics. These results also have implications for effectively modeling human immunity and for identifying reservoirs of zoonotic pathogens.

Hosts can be thought of as islands whose size impacts how many parasites, including microbes can colonize them (1). Unsurprisingly, bigger hosts represent bigger islands, leading parasite abundance (2) , biomass (3) , diversity (4), size (5) , size diversity (6) , and prevalence (7) all to increase with host body size. Body size is also intimately connected to the relative value of self-defense versus reproduction (8) , as most hosts reach large size via a long lifespan (9, 10) . Subsequently, natural selection should favor robust immune systems in large hosts, but physical or metabolic constraints could prevent the largest hosts from evolving optimal defenses. Ecoevolutionary processes are known to shape the architecture of species' immune systems (11, 12) , but whether body size constrains host immunity is little studied.

Generally, body size scaling relationships are modeled as power laws,

where Y is the focal trait, M is body mass, a is the intercept, and b is the scaling factor. To date, the Protecton and the Complexity of Immunity Hypotheses have received the most attention. Both frameworks predict that immunity should scale in direct proportion to body size (i.e., isometrically), which results in a massinvariant relationship (b = 0) for concentrations (13, 14) . The fractal nature of the circulatory and lymphatic systems was argued to predict isometry because rates of parasite detection and delivery of defenses should be driven by transit of cells and proteins through the vasculature (14) (15) (16) . By contrast, the Rate of Metabolism Hypothesis predicts a hypometric coefficient for immunity (b < 0 for a concentration) because all cell activities are linked to basal metabolic rates (BMR) (17) . As mass-specific BMR scales at b = -0.25, immune defenses involving cell turnover and metabolism should scale similarly (18) . Indirectly, some evidence supports this hypothesis (19) ; the causative agents of West Nile, rabies, and Lyme disease replicate more slowly in large host species (20) , as do endogenous retroviruses over evolutionary time (21) .

The Safety Factor Hypothesis, our focal hypothesis here, invokes performance-safety relationships from biophysics (10) and physiology (22) , positing that large species should evolve disproportionately stronger (i.e., hypermetric, b > 0 for concentrations) damage-mitigation mechanisms than small ones (9, 10) . Immunologically, large hosts should thus evolve exceptionally robust constitutive defenses to protect against the greater infection risks they will experience. Variants of the Safety Factor Hypothesis exist in the literature already, including Peto's paradox for cancer (23) and optimal defense theory (24) . Only recently, however, has direct empirical evidence been provided for immune defenses, namely hypermetric scaling for neutrophil concentrations in mammals (b = 0.11) (9) and heterophil concentrations in birds (b = 0.19) (25) .

The limitation of these and most immune allometry studies is how to interpret them. Hypermetric scaling of cell concentrations could represent greater protection in large animals or simply evidence that large animals must compensate for lower per-cell effectiveness in combatting infections by circulating more cells. Cell size scales isometrically but per-gram metabolism scales hypometrically (26) , so the average large-animal leukocyte might be comparatively ineffective compared to a leukocyte from a small species. Moreover, as selection tends to act on integrated traits (27) , scaling inference based on processes that affect host fitness will inherently be more insightful. Few scaling studies of immunity to date have relied on integrated immune functions, and others either involved few species or found body mass effects but did not estimate allometries (9) .

Our goals here were i) to determine how blood-based antibacterial defenses scaled with body size among terrestrial mammals, and ii) to test whether the Safety Factor hypothesis best-predicts the scaling exponent (b > 0) for this type of immunity. We focused on the antibacterial capacity of sera (28) , as this defense is an important determinant of host health and virulence for certain bacterial pathogens (29) , as well as measurable in a standard, reagent-independent way among diverse species. Briefly, antibacterial capacity measures bacteriostasis and bacteria killing capacity of sera. To describe interspecific variation in this immune function in the most confound-free manner possible, we compared the efficacy of sera of several, independent replicates from adults of mammalian species housed in zoos. To provide generality to any scaling we observed, we quantified the antibacterial capacity of sera of 160+ mammalian species against three distinct bacteria, Escherichia coli (ATCC #8739), Salmonella enterica (ATCC #13311), and Micrococcus luteus (ATCC # 4698) (28) . These bacteria are common pathogens of many vertebrates, some portion of their natural infection timeline occurs in blood (30) , and they are controlled effectively by blood-borne immune defenses (29) . As their last common evolutionary ancestor was ancient, any consistency in allometries for the antibacterial capacity against these particular bacteria should be more attributable to host immunity than parasite phylogenetic history.

To describe constitutive, antibacterial immunity, we fit non-linear functions to 12-step sera dilution series data from each individual animal with extensive individual-level replication across mammalian species (n = 1-67 replicates per species, mean = 5, Fig. S1 ) (31) . We characterized antibacterial dilution series, instead of antibacterial capacity at single dilutions, as the shapes of resultant ex vivo antibacterial functions capture salient aspects of defense (Fig. 1A ). For instance, if a very dilute serum sample was effective against a standard amount of bacteria, an animal of that host species would be relatively adept at controlling a large initial dose of bacteria in its bloodstream. Other meaningful antibacterial parameters included i) the maximum amount of bacteria that serum could protect against, ii) the serum dilution at the point of most rapid change from protected to vulnerable, and iii) the slope at this steepest part of the antibacterial function. This last parameter could be a proxy for either how rapidly serum defenses could be recruited to a bacterial replication event (32) or capture the pattern by which proteins and other molecules were recruited combinatorically to control bacteria. The Safety Factor hypothesis, our focal hypothesis, predicts that larger mammals will have higher maximal antibacterial capacity, higher antibacterial capacity at a specific dilution, and a more rapid shift from vulnerable to complete protection (Fig. 1B) .

To identify any allometries, we first fit non-linear antibacterial functions for each individual separately, then analyzed curve parameters as response variables in a set of multivariate, mixed-effects models (one for each bacterium) (31) . This approach allowed us to estimate the scaling coefficients for several aspects of each antibacterial function while simultaneously accounting for potential correlations among antibacterial curve elements as well as inter-and intra-specific variation (33) . To check whether antibacterial data were affected by the shared evolutionary history of host species, we also generated a separate set of phylogenetic univariate mixed-effects models (31) . We built separate models for each curve parameter with all other curve parameters and body mass included as fixed effects. We estimated a phylogenetic covariance matrix using a tree constructed with National Center for Biotechnology Information molecular data and phyloT ( Fig. S2 ) (34) . Overall, phylogeny explained little variation (Table  S1 ), so we do not address it further.

Our results strongly support the Safety Factor Hypothesis (Table 1 ). In general, results for antibacterial capacity against E. coli, S. enterica, and M. luteus are qualitatively similar (Table 1) . Mammalian species had diverse antibacterial curves (Movie S1). Even with this diversity, large species were much better able to control bacteria than small species and had ), higher antibacterial capacity overall (Fig. 2) . Large species required less serum to obtain the same antibacterial capacity, had a higher maximal antimicrobial capacity, and more steeply escalated antibacterial capacity over a shorter range of serum dilutions (Table 1 , Fig S4) . Large mammals also exhibited generally steeper transitions from minimal to maximal protection ( Fig.  2A) ; one of the smallest species, the cotton mouse (Peromyscus gossypinus), exhibited a gradual decline in the efficacy of antibacterial capacity against E. coli whereas the largest species, the Asian elephant (Loxodonta africana), continued to have complete protection even with very dilute sera (Fig. 2B) . Importantly, b for most dilution series parameters scaled hypermetrically for all three bacteria (Table 1) . Although the directions of allometries were consistent among bacteria and curve parameters, the magnitude of the slope at the steepest part of the antibacterial functions were quite distinct. For this particular parameter, allometric effects were weak for M. luteus (b = 2.4), strong for S. enterica (b = 5.2), and extreme for E. coli (b = 8.7; Fig. S3 A-C).

Asian elephant (3100 kg) and a small species, cotton mouse (25 g, C) demonstrate the general pattern that larger species had higher antibacterial capacity overall (B). Asian elephants have higher maximal capacity, required less serum to obtain the same capacity, and shift more rapidly from minimal to maximal antibacterial capacity.

Our data provide strong evidence for hypermetric scaling of serum antibacterial activity in terrestrial mammals. Hypermetric antibacterial activity might have evolved because large species face greater risk of infection and have more to lose evolutionarily from disease than small species. However, large species have more resources (i.e., space and energy density) on which parasites can capitalize (35) , so we cannot rule out that parasites might sometimes be al ly selected to infect large species. On the other hand, small species provide many more transmission opportunities for many parasites given the comparatively greater densities of small hosts (36) . The forces that lead to immune allometries warrant future investigation, and below we propose briefly some ideas for why and how antibacterial immunity among mammals scales hypermetrically.

The evolution of the vertebrate adaptive immune system (and its functional equivalents) enabled chordates to solve perhaps one of the most perplexing problems of life: how to detect an infinite diversity of enemy markers (i.e., antigens and pathogen-associated molecular patterns) with a limited set of protein-coding genes. Clearly, not only are vertebrates and their immediate ancestors adept at this task, but B and T cell receptor diversity is more constrained by overresponsiveness to self-than under-responsiveness to non-self (37) . Indeed, the probability that a parasite will evade the antigen space covered by B and T cell repertoires is 4x10 -44 (15, 38) . Adaptive immune systems will therefore eventually detect any antigen that ever evolves. The challenge of relying solely on these defenses, though, is that they involve major time lags (i.e., minimally 4-10 days but perhaps longer if the immune system is still developing (39)). Lags for inducible defenses generally leave hosts vulnerable as microbial doubling time proceeds over much shorter period. This incompatibility might explain why among 25 avian species, the production rate of infectious virions and the burst size of cells (i.e., the number of infectious virions released by an infected cell over its infected lifespan) scaled hypometrically, yet the rate of antibody-mediated neutralization of virus was isometric (15) . Perhaps hypermetrically-scaled constitutive immune defenses produced the surprisingly low viral outcomes observed in that study.

We expect that hypermetric immune allometries are likely pervasive in the first, constitutive lines of host defense. When parasites grow rapidly, as bacterial (and viral) infections often do, defenses with short time-delays are favored (24) . Barriers against all-comers are the best option when adversaries can replicate faster than defenses can be honed and mobilized. The downside to these forms of defense is that they impose comparatively large costs, especially when adversaries are absent (40) . Recently, a meta-analysis revealed that small animal species that are relatively long-lived incur the greatest costs of induced immune defenses (41) . Given their tendency to breed prolifically and develop rapidly, hypermetric allometries might manifest as much because small species favor rapidly-mobilizable defenses that are activated only after exposure to a parasite as they do because of mandates of large size.

Another conspicuous opportunity in light of our findings is to resolve how large mammals achieve hypermetric antibacterial capacity. We expect (but could not measure) that complement and other circulating proteins (i.e., lysozyme, mannan-binding protein, betadefensins) mediate most of what we quantified. Complement provides protection against diverse bacteria via three pathways, but here the alternative complement pathway is implicated because we used serum for assays (32) . Large animals might achieve higher antibacterial capacity because they mobilize pathways more rapidly, assemble functional membrane attack proteins more swiftly, or maintain higher concentrations of particular immune proteins. For example, large species may express more of a key complement element, C3, and disproportionately more decay-accelerating factor, CD55, or Factors H or B to minimize any detrimental effects of high C3 activity (32) . Factor D might also be a critical step in the pathway because it circulates in an active form whereas the others circulate as zymogens (32) . Alternatively, these patterns might arise in part because of trade-offs between the various functional responsibilities of the blood and liver. Many of the constitutive proteins involved in the antibacterial defenses we measured in serum are produced in the liver (42) . The liver also exports proteins, glucose, and lipoproteins to support cellular maintenance and respiration, among other important functions. The high massspecific metabolic rates rate of small animals and consequent high demand for these hepatic inputs may limit the liver's capacity to produce antibacterial proteins. Additionally, there is a direct relationship between the amount of protein in the blood and its viscosity (43) . This limits the amount of protein that blood can transport because increased viscosity decreases microvascular perfusion and delivery of nutrients and oxygen to cells. The high mass-specific metabolic rate of small animals results in the challenge of needing to deliver more nutrients and oxygen to cells without increasing blood viscosity, which may result in lower concentrations of antibacterial proteins (44) , although this perspective would not explain why large species would be so adept at antibacterial defense. In the future, it would be interesting to reveal why the magnitude of antibacterial allometries varied among bacterial species. Whereas M. luteus grows slowly and causes modest disease relative to the other two bacteria, both S. enterica and E. coli can cause sepsis (45) , infect via the diet, and have similar replication rates in culture. The scaling disparities we detected (Table 1) suggest that risk of morbidity or death to E. coli is enhanced relative to the other two bacteria, but too few data on sepsis yet exist and too many interspecific disparities remain to resolve this issue (46, 47) .

Our findings have ramifications for human medicine and management of disease cycles, especially zoonoses. In regard to medicine, we might revisit how we choose species as models of human immunity (48) . One option is to use the scaling effects described here as a form of null model from which to extrapolate currently favored rodent species responses to human applications. A greater understanding of immune scaling could also help develop better epidemiological models. Because "the rate that pathogens spread through populations is influenced by the rate of spread through individual hosts" (49), we must understand better how host traits affect their propensity to maintain and transmit viable parasites (50, 51) .

We expect that many more important immune allometries await description, and in particular, we encourage future scaling work on Order Chiroptera, as bats are implicated in the emergence of many zoonoses including SARS-COV2 (52, 53) . Unfortunately, bats are rare in zoos, so few bat species were in our analysis. We also encourage efforts to investigate immune allometries in ectotherms, and especially vectors, as their immune defenses are more temperature-dependent (17) , an important condition for a planet undergoing extensive climate change. We also hope creative approaches can be applied to scaling studies of more pathogenic microbes, ideally in more organismal contexts. Such work is challenging, as pathogenicity differs among species as well as route of infection (29) . By continuing to merge cellular and molecular efforts with the behavioral, ecological, and evolutionary concepts, we can produce a more integrative immunology and thus more effective options to predict and manage infectious disease risk.

Samples. We used the simulations performed by Dingemanse and Dochtermann to guide our sample sizes (Fig. 1 of 33) . We obtained serum from 212 species of healthy, zoo-and labhoused animals ranging in body size from 16 g -3,600 kg, which covers the range of body sizes of extant terrestrial mammals (Fig. S4) (54) . Most zoo samples were collected as part of routine wellness checkups whereas lab samples were collected at the termination of experiments. We supplemented zoo samples, which comprised the vast majority, with samples from lab-housed animals to augment small-bodied species (< 2 kg) in our analysis, which are rare in most zoos. We could not run all assays against all bacterial species for all samples because of sample volumes. Final sample sizes of species were 212, 199, and 167 for antibacterial capacity against E. coli, S. enterica, and M. luteus, respectively. We assayed a mean of 5 samples per species for each microbe (Fig. S1 ). We also obtained >15 samples from a subset of species representing the full range of body masses to help ensure that a single individual from a large or small species was not leveraging results (Fig. S5 ) Samples were stored at the collection location at -80 to -20 °C until shipped on dry ice to USF or Hamilton College where they were stored at 80°C until assays. Samples were used in assays within 24h of thawing. We restricted our analysis to samples collected in the years 2005-2019 because preliminary work indicated that samples collected before 2005 had reduced antibacterial ability. Use of these samples was approved by the Institutional Animal Care and Use Committees at Hamilton College (memo of understanding), SUNY ESF (memo of understanding), and University of South Florida (Protocol # 4920).

Antibacterial capacity. We measured antibacterial capacity against E. coli (ATCC #8739), S. enterica (ATCC #13311), and M. luteus (ATCC # 4698) using an adaptation of the microbiocidal assay (a.k.a. bacteria killing assay, BKA, microbiocidal activity) developed by French and Neuman-Lee (28) . Briefly, we plated a 12-point dilution curve of each animal sample in triplicate on a round-bottomed 96-well plate. E. coli had two serial dilutions: 6 points from raw to 1:64, and 6 points from 3:4 to 3:256. S. enterica also had two serial dilutions: 6 points from raw to 1:64, and 6 points from 3:4 to 3:128. M. luteus had a single serial dilution ranging from raw serum to 1:2048. Super antibacterial capacity curves-curves for species with 100% antibacterial capacity at the lowest dilution of the normal curves-for E. coli were 6 points from 1:4 to 1:256 and 6 points from 3:16 to 3:512 and for were S. enterica: 6 points from 1:16 to 1:512 and 6 points from 3:4 to 3:128. We also plated a 4-point dilution curve in Dulbecco's Phosphate Buffered Saline (PBS, Sigma-Aldrich #D8537) of commercially available cow serum (Innovative Research Novi, MI 48377, #IBV-Compl) as our inter-assay control. Cow curve dilutions were 1:32, 1:64, 1:128, 1:256 for E. coli; 1:16, 1:32, 1:64, 1:128 for S. enterica; and 1;20, 1:80, 1:320, 1:1280 for M. luteus. Our final volume of each serum dilution was 18 μ l. We plated three replicates of a 20 μ l negative control of PBS and 9 replicates of an 18 μ l positive control. We added 2 μ l of a standard concentration (10 4 CFU ml -1 ) of bacteria to all wells except the negative controls. We incubated plates for 30 min. at 37 °C. We next added 125 μ l of tryptic soy broth (BD #211825) to all wells, and then shook the plate for 1 min at 300 rpm. We then measured baseline absorbance of all wells at 300 nm using a plate reader (Biotek Synergy HTX Multi-Mode Reader) to serve as an internal control. We incubated the covered plate for 12 h for E. coli, 10 h for S. enterica, and 48 h for M. luteus at an incubation temperature of 37 °C. We measured final absorbance of all wells again at 300 nm. Antibacterial capacity for each well was calculated as

. We observed little evidence of sample degradation due to preservation method or duration, nor did we observe high inter-assay (9.02%) or intra-assay variation (0.02%). Key parameters of the antibacterial capacity assay for each microbe are summarized in Table S1 and full details about the procedures used to develop and validate our assays are available in the Supplementary Text. Assay protocols are deposited on FigShare (55) .

Data processing and analysis. Statistics were analyses using protocols in Program R (56). We compiled data using a workflow developed using Program R (62) (all code files are available on Figshare (57)). We visually checked each sample dilution curve for outliers (i.e., replicates that were distinct from the other two replicates and/or fell off the expected sigmoidal curve). We determined whether positive controls, negative controls, and inter-assay controls (i.e., cow sera) were within the expected range, and discarded the plate if data fell outside expected ranges. To determine how antibacterial capacity (i.e., the shape of dilution curves across all 12serum concentrations) scaled with body mass, we fit 5-parameter logistic regression growth curves to the cleaned dilution curves for each sample using package nplr (58) . To aid in curvefitting, we log 10 -transformed serum concentrations (i.e., the dilutions) and converted the antibacterial ability from a percent to a proportion. Curves could only be fit to values between 0 and 1, so percent inhibited values > 100 were forced to a random value between 99 and 100 and percent inhibited values < 0 were forced to a random value between 0 and 1 prior to conversion to proportions. This approached restricted values to a range near maximal and minimal antibacterial activity for all other curves. We extracted the curve parameters (inflection dilution, inflection antibacterial capacity, bottom asymptote antibacterial capacity, top asymptote antibacterial capacity, slope, asymptote coefficient) for use in univariate and multivariate general linear models (Fig. 1A) . Slope and inflection dilution were already on log 10 -scale. We added 1 to Inflection antibacterial capacity, bottom asymptote antibacterial capacity, top asymptote antibacterial capacity and then log 10 -transfromed them. We also log 10 -transformed body mass. All analyses were performed using transformed data.

To check for effects of mammalian phylogeny on antibacterial capacity, we constructed a separate phylogenetic univariate mixed model for each curve parameter for each bacterium. We included body mass and all other curve parameters as fixed effects then fit models using MCMCglmm (59, 60) . The phylogenetic covariance matrix for this analysis was estimated using a phylogenetic tree constructed with NCBI molecular data and phyloT (Fig S2) (40) . All mixed models were fit using a weak inverse-Gamma prior with shape and scale parameters set to 0.01 for the random effect of phylogenetic variance. Default priors for all other fixed effects were used. Model chains were run for 7.8 × 10 5 iterations, with an 180,00-iteration burn-in and a 600iteration thinning interval. We estimated Pagel's lambda as a measure of how much of the total observed variation was explained by phylogeny (61) .

Once strong effects of phylogeny were ruled out, we generated separate multivariate mixed models to query body mass effects on antibacterial capacity. Log 10 -transformed body mass was included as a fixed effect and species was included as a random effect. This model allowed species to have different intercepts and slopes. These models were fit using the MCMCglmm package (59, 60) . All mixed models were fit using a weak inverse-Gamma prior with shape and scale parameters set to 1.002 for the random effect. Default priors for all other fixed effects were used. Model chains were run for 1.82 × 10 6 iterations, a 420,000-iteration burn-in, and a 1400 iteration thinning interval. Results were robust across alternative priors, and chain length was sufficient to yield negligible autocorrelation. We extracted the slopes describing the relationship between each parameter and body mass across all species.

We included bacterial capacity at the bottom asymptote antibacterial capacity and asymmetric coefficient in our models because they helped describe antibacterial capacity curves although we did not make predictions about their biological functions. In our study, we did not always collect data at the lower asymptote because we limited our sampling to 12-dilutions. As a result, the bottom asymptote estimated by the general linear model was based on incomplete data and not biologically relevant to our scaling hypotheses. The dilution at the inflection point was also include in our models because it informed our understanding of the antibacterial capacity at the inflection point, but it was not biologically meaningful for testing scaling of immune defenses by itself. Results for these parameters can be found in Table S2 . 

Hosts as islands

Allometry of animal-microbe interactions and global census of animal-associated microbes

The scaling of total parasite biomass with host body mass

Interactions among host diet, nutritional status and gastrointestinal parasite infection in wild bovids

Mallophaga from Apteryx , and their Significance

Evolutionary ecology of parasites

Seasonal dynamics in mosquito abundance and temperature do not influence avian malaria prevalence in the Himalayan foothills

Size, function, and life history

The Effects of Body Mass on Immune Cell Concentrations of Mammals

Do Performance-Safety Tradeoffs Cause Hypometric Metabolic Scaling in Animals?

Ecological immunology: costly parasite defences and trade-offs in evolutionary ecology

Mechanisms and Methods in Ecoimmunology: Integrating Within-Organism and Between-Organism Processes. Integrative and Comparative Biology

The E-T (Elephant-Tadpole) paradox necessitates the concept of a unity of B-cell function: The protecton

Some Scaling Principles for the Immune System

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

A mathematical framework for understanding how lymph node architecture scales with host body size to produce an efficient immune response

Toward a Metabolic Theory of Ecology

Allometric Scaling of the Active Hematopoietic Stem Cell Pool across Mammals

Scaling of Host Competence

The Allometry of Host-Pathogen Interactions

Larger Mammalian Body Size Leads to Lower Retroviral Activity

Quantitative evolutionary design

Origins of human cancer

Inducible Defense against Pathogens and Parasites: Optimal Choice among Multiple Options

The impact of body mass on immune cell concentrations in birds. bioRxiv (2020)

Scaling: why is animal size so important?

Studying the evolution of physiological performance. Oxford surveys in evolutionary biology

Improved ex vivo method for microbiocidal activity across vertebrate species

Bactericidal and bacteriolytic activity of serum against gram-negative bacteria

Immune response to infection with Salmonella typhimurium in mice

Quantifying individual variation in behaviour: mixed-effect modelling approaches

phyloT: phylogenetic tree generator

A Metabolic and Body-Size Scaling Framework for Parasite Within-Host Abundance, Biomass, and Energy Flux

Life History and Demographic Drivers of Reservoir Competence for Three Tick-Borne Zoonotic Pathogens

How diverse should the immune system be?

Theoretical Studies of Clonal Selection: Minimal Antibody Repertoire Size and Reliability of Self-Non-self Discriminationt

Embryonic development period and the prevalence of avian blood parasites

Individual variation in thermogenic capacity affects above-ground activity of high-altitude Deer Mice

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

Hepatocytes: a key cell type for innate immunity

Role of plasma proteins in whole blood viscosity: A brief clinical review

Basal Metabolic Rate: History, Composition, Regulation, and Usefulness

Host-pathogen interactions in sepsis. The Lancet Infectious Diseases

Animal models of sepsis

The disconnect between animal models of sepsis and human sepsis

Of mice, macaques and men: scaling of virus dynamics and immune responses

Allometry and simple epidemic models for microparaites

Body-size scaling in an SEI model of wildlife diseases

Body size and meta-community structure: the allometric scaling of parasitic worm communities in their mammalian hosts

Host and viral traits predict zoonotic spillover from mammals

Global Patterns of Zoonotic Disease in Mammals

How big should a mammal be? A macroecological look at mammalian body size over space and time

Protocols for 12-dilution antibacterial capacity curves for interspecific comparisons

R: A language and environment for statistical computing (R Foundation for Statistical Computing

Code to compile data and analyzing data for a comparative antimicrobial analysis. figshare

R package nplr n-parameter logistic regressions

General quantitative genetic methods for comparative biology: phylogenies, taxonomies and multi-trait models for continuous and categorical characters

MCMC Methods for Multi-Response Generalized Linear Mixed Models: The MCMCglmm R Package

The Phylogenetic Mixed Model

Beyond phytohaemagglutinin: assessing vertebrate immune function across ecological contexts: Assessing vertebrate immune function across ecological contexts

No simple answers for ecological immunology: relationships among immune indices at the individual level break down at the species level in waterfowl

Constitutive innate immunity is a component of the pace-of-life syndrome in tropical birds

Nashville Zoo at Grassmere, Oregon Zoo (Mitch Finegan & David Shepherdson), Sedgwick County Zoo, Smithsonian's National Zoo and Conservation Biological Institute, Topeka Zoo & Conservation Center

Uyeda for statistical guidance, and the Martin Lab and HA Woods for discussion about an early version of the manuscript. Funding: This work was supported by State University of New York College of Environmental Science and Forestry's Provost Office, the Levitt Center, Hamilton College's Dean of Faculty Office, University of South Florida's College of Public Health, and the National Science Foundation

One file containing:

Figures S1-S8Tables S1-S3