The relationship between diet, plasma glucose, and cancer prevalence across vertebrates

Could diet and mean plasma glucose concentration (MPGluC) explain the variation in cancer prevalence across species? We collected diet, MPGluC, and neoplasia data for 160 vertebrate species from existing databases. We found that MPGluC negatively correlates with cancer and neoplasia prevalence, mostly of gastrointestinal organs. Trophic level positively correlates with cancer and neoplasia prevalence even after controlling for species MPGluC. Most species with high MPGluC (50/78 species = 64.1%) were birds. Most species in high trophic levels (42/53 species = 79.2%) were reptiles and mammals. Our results may be explained by the evolution of insulin resistance in birds which selected for loss or downregulation of genes related to insulin-mediated glucose import in cells. This led to higher MPGluC, intracellular caloric restriction, production of fewer reactive oxygen species and inflammatory cytokines, and longer telomeres contributing to longer longevity and lower neoplasia prevalence in extant birds relative to other vertebrates.


30
Explaining patterns of cancer susceptibility among multicellular organisms is a major 31 challenge in comparative oncology. Several life-history and environmental factors have been 32 proposed to explain variations in cancer prevalence. Body size and longevity have been hypothesized 33 to be predictors of cancer prevalence, based on the idea that larger animals (i.e. those with more 34 cells), and long-lived animals (i.e. those with a longer time for mutations to accumulate 1 ), should 35 accumulate more carcinogenic mutations and thus have a higher cancer prevalence 2,3 . This has been 36 shown in a study of free-living birds where body mass is positively correlated with tumor prevalence 4 .

37
The majority of studies testing this hypothesis, however, show that larger and longer-lived animals do 38 not get more cancer [5][6][7][8][9][10] . Other life history traits, such as shorter gestation length (across 39 vertebrates 10 ), and large litter size (in mammals 6 ) or clutch size (in birds 8 ), have been correlated with 40 cancer prevalence. The reason for this positive correlation may be that animals may trade off 41 resources between body maintenance (such as protection from DNA damage) and reproduction 11,12 . 6 and n = 50 species, respectively), or diet type (n = 67 species), is not significantly associated with 117 differences in mean plasma glucose concentration (Fig. 1C, 1D).  (Table 1B). Mean plasma glucose concentration is negatively correlated with gastrointestinal cancer 126 prevalence (measured from a total of 242 records of gastrointestinal malignancies across 108 127 species), even after controlling for weight, birds versus non-birds ( Fig. 3; Table 1B). Mean plasma 128 glucose concentration is negatively correlated with non-gastrointestinal neoplasia prevalence only 129 when we control the analyses for weight (Table 1B). Within mammals (Supp. Fig. 5; Table 1C; n = 130 48-68 species), birds (Supp. Fig. 6; Table 1D; n = 31-51 species), and reptiles (Supp. Fig. 7; Table   131 1E; n = 25-35 species) there were no significant correlations between mean plasma glucose 132 concentrations and cancer prevalence or neoplasia prevalence across tissues, gastrointestinal cancer 133 prevalence or gastrointestinal neoplasia prevalence, and non-gastrointestinal cancer prevalence or 134 non-gastrointestinal neoplasia prevalence after applying multiple testing corrections. [Fig. 2 & 3] 136 Relationships Between Cancer and Trophic Levels, Food Type, and Diet Type 137 Trophic levels are positively correlated with cancer prevalence and neoplasia prevalence 138 across tissues among 160 species, with and without controlling for the variance in species' plasma 139 glucose concentrations (Fig. 4A), even after correcting for multiple testing (Table 1F). There is no 140 significant correlation, however, between trophic levels and gastrointestinal malignancy and neoplasia 141 prevalence, non-gastrointestinal cancer prevalence, or non-gastrointestinal neoplasia prevalence after 142 applying corrections for multiple testing (Table 1F). [Fig. 4] 7

143
The percentage of plants in a species' diet is negatively correlated with gastrointestinal 145 neoplasia prevalence among 22 species (Supp. Fig. 2B), but not after controlling for the variation in 146 species plasma glucose concentrations (Supp. Fig. 2B; Table 1F). The percentage of fruit, plants, 147 invertebrates, seeds, endothermic vertebrates, or ectothermic vertebrates in a species' diet is not 148 correlated with cancer prevalence and neoplasia prevalence across tissues ( Fig. 4B; Supp. Fig. 2A), 149 gastrointestinal malignancy prevalence (Fig. 5B), or non-gastrointestinal malignancy prevalence and 150 non-gastrointestinal neoplasia prevalence (Supp. Fig. 2C, D) after correcting the analyses for multiple 151 testing (Table 1F).

152
The percentage of plant-based food in a species's diet is negatively correlated with 153 gastrointestinal cancer prevalence among 29 species even after controlling for the variation in the 154 species' plasma glucose concentration and correcting for multiple testing ( Fig. 5C; Table 1F).

160
There is no association between diet type and cancer prevalence among 67 species (  (Table 1F). 167 [Fig. 5] 168 Discussion 169 Similar to previous studies testing glucose concentrations in the plasma and whole blood 170 across vertebrates 42,43 , we found that birds have the highest mean concentration of plasma glucose, 171 followed by mammals, reptiles, and amphibia. In our cross-vertebrates diet, plasma glucose, and 172 cancer association studies we found two main associations. Our results do not support our first 173 8 hypothesis, as we found that diet is not correlated with mean plasma glucose concentration across 174 vertebrates. Our results support our second hypothesis that there is a negative correlation between 175 mean plasma glucose concentrations and neoplasia and cancer prevalence. Specifically, we found a 176 negative correlation between mean plasma glucose concentrations and gastrointestinal neoplasia and 177 cancer prevalence across vertebrates. We also found a negative correlation between mean plasma 178 glucose concentration and non-gastrointestinal neoplasia prevalence only when controlling the 179 analyses for weight. Our results also support our third hypothesis that trophic levels are positively

200
The reduced import of glucose inside avian cells has multiple intracellular effects that 201 contribute in stabilizing and protecting DNA, suppressing cellular growth, and slowing down ageing 202 and ageing-related diseases in birds relative to mammals 63

218
Eleventh, there are higher levels of the potential antioxidant "uric acid" in birds than mammals 79,80 .

219
Twelvth, the antioxidant NRF2 is constitutively expressed in most birds as an evolutionary 220 consequence of the loss of a binding domain in its repressor KEAP1 during the Permian-Triassic 221 period 60 . These features may explain the increased lifespan 44,81 and lower cancer prevalence 18,45,82,83 222 in birds relative to mammals and reptiles, and our findings that vertebrate species with higher plasma

226
The relationship between higher plasma glucose levels and lower gastrointestinal neoplasia 227 and cancer prevalence, however, may not be entirely explained by phylogenetic relatedness. As seen 228 in previous studies, a mammal, the Egyptian fruit bat, has around 5X higher blood glucose levels than 229 a healthy rat 84 . Further explanations for this negative correlation between glucose levels and 230 gastrointestinal neoplasia and cancer prevalence, apart from phylogeny, may be a shared ancestral 231 environment by some of these species. This may mean that they share a similar microbiome, which is 232 known to affect metabolic rates in some taxa [85][86][87] , and thus may also affect blood glucose levels. A 233 10 shared environment may also mean that the animals are exposed to similar pollutants affecting their 234 cancer prevalence. Also, the less host-specific gut microbiome of birds and bats, and the potential 235 subsequent higher proportion of nutrients absorbed directly by bird and bat cells, relative to other 236 vertebrates 88 , may explain the relatively constant high blood glucose levels in birds and bats.

237
Additionally, a non-selective force of evolution, random genetic drift acting during domestication may   (Table 1F). Additionary, it is known that toxins 253 bioaccumulate at higher concentrations in higher trophic levels [91][92][93] . In other words, lower trophic 254 levels have relatively fewer bioaccumulated toxins. Also, a study in 95 birds has shown that eating 255 invertebrates or a higher seed-to-fruit ratio was positively correlated with higher Trolox-equivalent 256 antioxidant capacity and higher concentration of the potential antioxidant uric acid 94 . Overall, 1) the 257 relative abundance of reptiles and mammals in higher trophic levels; 2) the biomagnification of toxins 258 in higher trophic levels; and the 3) oxidative stress and DNA damage associated with eating red meat, 259 may explain the higher cancer prevalence in higher trophic levels 13 (Table 1F).

260
Limitations & Future directions 261 11 There may be some mismatches between the diet, mean plasma glucose concentration, and 262 cancer prevalence of species in our dataset versus the actual values of diet, mean plasma glucose 263 concentration, and cancer prevalence in a species because each of these three components were 264 most likely collected from different individuals of the same species and we do not known the exact 265 health status of every individual at the time of data collection. For example, the mean plasma glucose 266 concentrations in our study were collected from anesthetized animals (data from ZIMS), some of 267 which may have had infections that affected their blood glucose levels, may have had higher blood 268 glucose levels due to diabetes 95 , and/or may have been stressed during the method of capture (e.g.,

269
being darted, netted, or grabbed). Anaesthesia or capture can mildly to moderately elevate the blood    We identified trophic levels (herbivore, invertivore, primary carnivore, or secondary carnivore) 312 for n=160 species based on records from previous studies (classified for each species based on their 313 primary diet) 13 . We collected food type percentages (endothermic vertebrates, ectothermic

428
(A) Trophic level is not significantly correlated with gastrointestinal malignancy prevalence for 108 429 species after correcting for multiple testing (Table 1B;

452
The data used in this study will be made available upon acceptance of the manuscript for publication.   (Table   1F). There is no significant correlation between the percentage of fruit, invertebrates, seeds,

Competing interests
We declare we do not have any conflicts of interest. Tables   Table 1. Phylogenetic generalised least squares (PGLS) regression summary results of the figures in the main text and in the supplementary materials. High values of lambda indicate that the signals can be mainly explained by common ancestry among species. In rows where all columns show NA, there were <10 species to conduct a powerful analysis. In the "Type of Association" column we report whether there is a positive (+) or negative (-) association between the independent variable A and the dependent variable. When the independent variable is categorical, we report the sign (+ or -) of the majority of between-group comparisons. If 50% of the between-group comparisons have a positive (+) association and 50% of the between-group comparisons have a negative (-) association, we report both signs. For multivariate analyses, in the 1st P-value column we report the P-value of the independent variable A, in the 2nd P-value column we report the P-value of the independent variable B, in the 3rd P-value column we report the P-value of independent variable C, in the 4th P-value column we report the P-value of variable D, and in the F-statistics column we report the F-statistics of the independent variable A. We mark P-values that passed the False Discovery Rate (FDR) correction in column "P-value of variable A" with an asterisk (*).
Based on our hypotheses, we performed a separate FDR correction in the group of analyses as separated in the tables below (A-F).