Interregional Coevolution Analysis Revealing Functional and Structural Interrelatedness between Different Genomic Regions in Human Mastadenovirus D

Abstract

Human mastadenovirus D (HAdV-D) is exceptionally rich in type among the seven human adenovirus species. This feature is attributed to frequent intertypic recombination events that have reshuffled orthologous genomic regions between different HAdV-D types. However, this trend appears to be paradoxical, as it has been demonstrated that the replacement of some of the interacting proteins for a specific function with other orthologues causes malfunction, indicating that intertypic recombination events may be deleterious. In order to understand why the paradoxical trend has been possible in HAdV-D evolution, we conducted an interregional coevolution analysis between different genomic regions of 45 different HAdV-D types and found that ca. 70% of the genome has coevolved, even though these are fragmented into several pieces via short intertypic recombination hot spot regions. Since it is statistically and biologically unlikely that all of the coevolving fragments have synchronously recombined between different genomes, it is probable that these regions have stayed in their original genomes during evolution as a platform for frequent intertypic recombination events in limited regions. It is also unlikely that the same genomic regions have remained almost untouched during frequent recombination events, independently, in all different types, by chance. In addition, the coevolving regions contain the coding regions of physically interacting proteins for important functions. Therefore, the coevolution of these regions should be attributed at least in part to natural selection due to common biological constraints operating on all types, including protein-protein interactions for essential functions. Our results predict additional unknown protein interactions.

Importance: Human mastadenovirus D, an exceptionally type-rich human adenovirus species and causative agent of different diseases in a wide variety of tissues, including that of ocular region and digestive tract, as well as an opportunistic infection in immunocompromised patients, is known to have highly diverged through frequent intertypic recombination events; however, it has also been demonstrated that the replacement of a component protein of a multiprotein system with a homologous protein causes malfunction. The present study solved this apparent paradox by looking at which genomic parts have coevolved using a newly developed method. The results revealed that intertypic recombination events have occurred in limited genomic regions and been avoided in the genomic regions encoding proteins that physically interact for a given function. This approach detects purifying selection against recombination events causing the replacement of partial components of multiprotein systems and therefore predicts physical and functional interactions between different proteins and/or genomic elements.

FIG 1

Regional recombination and coevolution. The abscissae of all panels represent the positions adjusted to HAdV-8 (accession number AB448767) as a reference. (A) Positions of the 195 identified recombined regions. Each black line represents a recombined region. (B) Number of recombination events in each 200-bp window. Green and magenta dots represent basal and nonbasal regions, respectively. The upper dots in a lighter color and the lower dots in a darker color represent the counts of all recombined regions and those between distant types only, respectively. (C) Merged coevolution (upper left) and partial coevolution (lower right) matrices. The values of significant correlation/partial correlation coefficients (P < 0.0025) are shown using a color gradient ranging from yellow (near 0) to red (1.0). The diagonal is shown in gray. The basal (green), nonbasal (magenta), and invariant (gray) regions are indicated at the bottom of the matrix (details are in Table 1). Genes (thick arrows) and protein-coding regions (black arrows) are shown around the matrix. (D) Ratio of the windows showing significant correlations to the window at each position. (E) Correlation coefficient of each window against the entire DNA polymerase-coding region. Significant (P < 0.0025) correlation coefficients are shown with green circles, corresponding to the basal regions, and the others are shown with magenta circles.

FIG 2

Distribution of recombined segment lengths. The abscissa shows the different lengths, while the ordinate shows the frequency by each size category. Lengths of the recombined segments were adjusted to match Fig. 1A.

FIG 3

Evolutionary correlation on simulated sequences. The simulated multiple-genome alignment of 45 artificial genomic sequences was generated by simulating sequence evolution using Mesquite version 2.75 under the following conditions: the tree topology = the genome tree; the number of characters = 33,645; the ratio of invariant sites = 0.65; the alpha parameter of the gamma distribution of rate variance = 0.477; nucleotide frequencies of A, T, G, and C = 0.22, 0.21, 0.29, and 0.28, respectively; and the transition/transversion ratio = 1.57. The abscissa and ordinate (the x and y axes) of this matrix represent the physical positions in the simulated MGA, and each point (x, y) of the matrix shows the Mantel's correlation coefficient between windows x and y. The correlation coefficient ranges from near 0 (yellow) to near 1 (red). Independent and diagonal windows are colored black and gray, respectively.

FIG 4

Histogram of significant correlation ratios. The abscissa is the ratio of the number of windows that show a significant correlation coefficient to a specific window against the total number of windows (=167). The left and right ordinates are for the absolute frequencies (bars) and relative cumulative frequencies (lines), respectively. The gray bars and line are for the simulated data. The absolute frequencies in basal and nonbasal regions of the real data are shown in black and cross-hatched bars, respectively, together with cumulative frequencies in the black line.

FIG 5

Finer-scale coevolution analysis. The results of the finer-scale analyses are depicted for two highlighted regions, fiber (A) and E3 region (B). The abscissae show the position in the HAdV-8 genome (accession number AB448767). The left ordinates represent the correlation coefficients between each 100-bp window and the entire DNA polymerase-coding region. Significantly correlated windows (P < 0.0025) are shown with solid circles, equivalent to a basal region, and the others are presented with open circles. Protein-coding regions are shown with arrows below each plot. The predicted extracellular (O), transmembrane (M), and cytoplasmic (I) regions are shown in the coding regions of CR1α, gp19K, CR1β, and CR1γ.