We then analyzed whether neuronal and endothelial transcriptomes might differ in their rates of evolution across species. For this, we used two data sets (Khaitovich et al. 2005
; Khaitovich, Tang, et al. 2006
) which consist of measurements from 6 humans and 5 chimpanzees for Brodmann Area 9 (BA9) and of 10 humans, 6 chimpanzees, and 6 rhesus macaques for Brodmann Area 46 (BA46), respectively. Both BA9 and BA46 are dorsolateral prefrontal cortex regions.
In all within-species comparisons, the transcriptome diversity in NEX genes is significantly lower than in ENDEX genes (all P values < 10−4), showing that neuronal transcriptomes are more constrained than endothelial transcriptomes also in chimpanzees and rhesus macaques.
In contrast, the divergence between species does not significantly differ between NEX and ENDEX genes in the BA46 data set, although the human–chimpanzee difference is of borderline significance (P
human–chimpanzee BA46 = 0.06; P
human–rhesus BA46 = 0.63; P
chimpanzee–rhesus BA46 = 0.84). In the BA9 data set, transcriptome divergence for NEX genes is significantly lower (P
human–chimpanzee BA9 = 0.02). The point estimates and 95% bootstrap derived confidence intervals (CIs) for divergence and diversity are reported in supplementary table 2
We next compared evolutionary rates by analyzing the ratios of divergence relative with diversity (D) of NEX and ENDEX genes. Endothelial D in BA9 is 1.59, whereas D for neurons is 3.02 and thus higher than for the endothelium (Pbootstrap < 10−5; ). Endothelial D calculated for BA46 between humans and chimpanzees is 2.06, whereas neuronal D is 2.75 and thus larger than endothelial D (Pbootstrap = 0.0135; ). When neuronal D is compared with endothelial D for the human–macaque comparisons and for the chimpanzee–macaque comparisons, neuronal D is found to be 1.9 times larger (6.3 vs. 3.35) than endothelial D for the human comparison (Pbootstrap < 10−5) and 1.5 times larger (3.71 vs. 2.48) for the chimpanzee comparison (Pbootstrap = 0.0007), respectively (). Thus, both these data sets suggest that D is higher in neuronal transcriptomes than in endothelial transcriptomes among primates.
FIG. 4.— Pairwise comparisons of transcriptome divergence to diversity estimates (D) for different primate species in the BA9 and BA46 of the prefrontal cortex for NEX and ENDEX genes. The error bars indicate the 95% CIs of the D estimate as determined by 10,000 (more ...)
To analyze if D for NEX and ENDEX genes differ significantly from D estimates for other transcripts expressed in the brain, we randomly assigned genes expressed in BA9 and BA46 data sets to 10,000 groups of the same size as the NEX and ENDEX genes. The D values estimated for NEX and ENDEX genes in the two frontal cortex data sets were then compared with the distributions of D in the 10,000 random groups of genes. In all four comparisons was the observed neuronal D significantly higher than D calculated for the random gene sets (). In fact, for BA46, none of the 10,000 random groups of genes had as high a D as the neuronal D found for the human–chimpanzee and the human–macaque comparisons. By contrast, the observed endothelial D was significantly lower than for the random genes in three out of four cases. Thus, D measured for neuronal and endothelial transcriptomes are located at two extremes of the distribution of possible D values—with the neuronal D being exceptionally high and the endothelial D exceptionally low.
FIG. 5.— Diversity (x axis) and divergence values (y axis) for bootstrap replicates of genes expressed in BA46 and BA9. Each small blue dot indicates an outcome of bootstrapping the same number of genes as there are in the list of NEX genes; each small red dot (more ...)
As mentioned above in the text, the differences in D
values between NEX and ENDEX genes are mainly driven by differences in diversities and not by differences in divergence (; supplementary table 2
, Supplementary Material
online). It remains to be elucidated why expression of NEX genes show excess divergence given the low levels of diversity. One possible explanation would be that NEX genes are more often regulated by the same transcription factors than ENDEX genes. This would allow the NEX genes to evolve in a more concerted manner (in expression) than ENDEX genes, and expression diversity would remain low—even if divergence between species increases.
Other explanations for these patterns could be that the magnitude of environmental influences on gene expression in neurons is lower than on other cells—and the magnitude of environmental influence on endothelial cells higher than on other cells in the brain. This is conceivable considering that endothelial cells are a part of the blood brain barrier and thus have a function of controlling the influx of molecules into the brain and therefore might have to react more often to environmental changes. This, however, would not explain why neuron-related genes are more constrained in their expression in nonbrain tissues or why they show more constraints in their amino acid sequences.
The study presented with this paper points at the fact that different transcriptomes of different cell types within a tissue can show different evolutionary patterns. Thus, there is a potential in better understanding transcriptome evolution if transcriptomes of more closely defined cell populations are analyzed. For primate tissue transcriptome comparisons (other than blood), laser capture microdissection provides currently the only technological solution for isolating specific cell populations from complex tissues. This is because the tissue material is collected several hours after the death of donors, which compromises tissue integrity and thus other technologies for isolating specific cell types out of a complex mixture, such as fluorescence-activated cell sorting (Arlotta et al. 2005
), can therefore not be applied to such samples. Future work will aim at measuring and comparing different cell types isolated from human and chimpanzee brains and other tissues.