We fed four groups of six 8-week-old female mice one of four diets ad libidum
: first, the mouse pellet diet on which they were raised; second, a diet consisting of vegetables, fruit and yogurt identical to the diet fed to chimpanzees in our ape facility; third, a diet consisting of cooked food eaten in our Institute's cafeteria; fourth, a diet consisting exclusively of McDonald's fast food (Table S1
After two weeks, we examined gene expression in liver and brain. Using an ANOVA and permutation test, we find significant expression level differences among mice fed the four diets in liver, but not in brain (one-sided permutation test p
<0.001 and p
0.16, respectively; Table S2
). Similarly, when the effects of particular diets on liver gene expression are compared, all pairs of diets show significant differences from each other (one-sided permutation test p
<0.02) with one exception: The cafeteria and fast food diets are indistinguishable in terms of liver gene expression (p
0.14; Table S2
). We therefore decided to treat these two human diets together. We find that when the human diets are compared to the chimpanzee diet, 830 of the 13,168 expressed genes, or 6.3%, are affected in the mouse liver (one-sided permutation test p
0.030; ; Table S3
The effects diet on gene expression in mice.
We then compared these genes affected by human and chimpanzee diet differences in mouse to the 1,169 orthologous genes that differ in expression when human and chimpanzee livers are compared 
. 10%, or 117 of the genes differentially expressed between human and chimpanzee livers were also among the genes affected by human and chimpanzee diet differences, a proportion larger than expected by chance (one-sided permutation test, p=
0.001; ). Thus, using mice fed just two distinct human diets and one chimpanzee diet, it is possible to replicate some of the expression differences observed between humans and chimpanzees. By contrast, expression differences between the original mouse pellet diet and the two human diets (8.9% of genes; Table S3
) did not overlap significantly with expression differences observed between human and chimpanzee livers (Table S4
). Since mouse pellets, unlike the model chimpanzee diet, have high caloric and protein content and are heat processed, the gene expression differences between humans and chimpanzees seen also in mice fed chimpanzee and human diets () are likely to reflect effects induced in the liver by components of the human and chimpanzee diets, respectively.
Overlap between liver gene expression differences in mice and primates.
A total of 117 genes are differentially expressed both between mice fed human and chimpanzee diets and between humans and chimpanzees in liver (Table S5
). We find that these 117 putatively diet-related genes have higher absolute effect sizes (mean differences between groups in units of standard deviation) for human-chimpanzee expression differences than 1,052 non-diet-related genes differentially expressed between human and chimpanzee livers (one-sided Mann-Whitney U test p
0.022). In other words, diet-related genes exhibit larger expression divergence than most differentially expressed genes in liver. 92 of these 117 genes (78%) show up-regulation under the human diet compared to the chimpanzee diet in mouse, but interestingly, there is no significant correlation between the direction of change in mouse under the two diet conditions and the change seen between humans and chimpanzees (Fisher's exact test p
0.6; Table S6
; Figure S1
). In terms of their functional roles, the 117 diet-related genes are significantly overrepresented in seven biological process categories in the Gene Ontology 
compared to other genes differentially expressed between human and chimpanzee livers. Notably, five of these categories are involved in metabolism in a broad sense (). Furthermore, using orangutan gene expression data as an outgroup 
, we observe that the expression levels of these 117 genes are more similar between orangutan and chimpanzees than between orangutan and humans than is the case for other genes differently expressed between human and chimpanzee livers (one-sided permutation test p=
0.047; Table S7
). This would be expected if the effects of the chimpanzee and the orangutan diets were more similar to each other than either were to the effects of the human diets.
Biological processes significantly enriched in genes potentially involved in human-chimpanzee dietary differences.
In order to gauge the rate of evolution of the 117 genes affected by diet, we compared the DNA sequence divergence in their promoter regions 
between humans and chimpanzees and the inferred amino acid sequences of their encoded proteins 
) all orthologous human and mouse genes that differ in gene expression between human and chimpanzees in liver, (ii
) all human-mouse orthologs expressed in human or chimpanzee livers, and (iii
) all human-mouse orthologs irrespective of their expression in liver. We find that both the promoter sequences (one-sided permutation test p=
0.01, 0.06, 0.04, respectively) and the amino acid sequences (p=
0.002, <0.001, 0.043, respectively) evolve faster in the 117 genes than in the latter sets of genes (; Table S8
). We also tested whether there is any significant overlap between the 117 genes that are affected by diet and differ in expression between humans and chimpanzees and either genes positively selected in their promoters in the human and chimpanzee lineages recently published by Haygood et al. 
or in their amino acid sequences recently published by Bakewell et al. 
. We find no such significant overlaps. This may not be surprising given the presumably high false negative rate pertaining to the identification of relevant genes in our study as well as the other studies.
Sequence divergence of genes potentially involved in human-chimpanzee dietary differences.