- Split View
-
Views
-
Cite
Cite
Baojun Wu, Allison I. Macielog, Weilong Hao, Origin and Spread of Spliceosomal Introns: Insights from the Fungal Clade Zymoseptoria, Genome Biology and Evolution, Volume 9, Issue 10, October 2017, Pages 2658–2667, https://doi.org/10.1093/gbe/evx211
- Share Icon Share
Abstract
Spliceosomal introns are a key feature of eukaryote genome architecture and have been proposed to originate from selfish group II introns from an endosymbiotic bacterium, that is, the ancestor of mitochondria. However, the mechanisms underlying the wide spread of spliceosomal introns across eukaryotic genomes have been obscure. In this study, we characterize the dynamic evolution of spliceosomal introns in the fungal genus Zymoseptoria at different evolutionary scales, that is, within a genome, among conspecific strains within species, and between different species. Within the genome, spliceosomal introns can proliferate in unrelated genes and intergenic regions. Among conspecific strains, spliceosomal introns undergo rapid turnover (gains and losses) and frequent sequence exchange between geographically distinct strains. Furthermore, spliceosomal introns could undergo introgression between distinct species, which can further promote intron invasion and proliferation. The dynamic invasion and proliferation processes of spliceosomal introns resemble the life cycles of mobile selfish (group I/II) introns, and these intron movements, at least in part, account for the dramatic processes of intron gain and intron loss during eukaryotic evolution.
Introduction
Since the discovery of spliceosomal introns in eukaryotes (Berget etal. 1977; Gilbert 1978), much progress has been made toward understanding the origin, evolution and function of spliceosomal introns (Cavalier-Smith 1991; Palmer and Logsdon 1991; Koonin 2006; Heyn etal. 2015). Spliceosomal introns are thought to originate from self-splicing group II introns, which are present in eubacteria, archaea, and eukaryotic organelles (Koonin 2006). Such hypothesis is supported by the fact that the spliceosome complex and group II introns share similar structures (Zimmerly and Semper 2015). After the birth of spliceosomal introns, they have undergone extensive gain and loss in a variety of lineages. As a result, intron densities in modern eukaryotic species vary greatly from a few introns in the entire genome of Giardia genus (Morrison etal. 2007) to over six introns per gene in most vertebrates and land plants (Csuros etal. 2011), many of which are either substantially higher or significantly lower than the intron density of the Last Eukaryotic Common Ancestor (estimated at 4.3 introns per gene) (Csuros etal. 2011). Several mechanisms have been proposed for the dynamic intron gain and loss processes (Farlow etal. 2011; Yenerall and Zhou 2012). For instance, variable intron densities among eukaryotic species could result from the interplay between nonhomologous end joining (NHEJ) and homologous recombination (HR), supposedly favoring DNA insertion and intron gain, and intron loss via reverse splicing, respectively.
Much higher (50- to 3,000-fold) rates of intron loss than intron gain have been observed in eukaryote genomes, including many intron-rich lineages (Roy and Gilbert 2005; Csuros etal. 2011). The presence of abundant intron-rich organisms requires mechanisms that can count against intron loss and/or lead to significant intron gain. The fast-growing genomic data have provided direct insights into some of the mechanisms. Particular spliceosomal introns share high similarity with other introns from unrelated genes, and are referred to as introner elements (IEs) (Worden etal. 2009). IEs are commonly present in a wide range of eukaryotes, including fungi, algae, tunicate and nematodes (Roy 2004; Worden etal. 2009; Denoeud etal. 2010; van der Burgt etal. 2012; Verhelst etal. 2013). Many IEs are flanked by 3–8 nt short direct repeats or target site duplications (TSDs), which are hallmarks of NHEJ mediated sequence transposition (Li etal. 2009; Huff etal. 2016). In a recent study, transposition of IEs has been suggested to account for hundreds to thousands of spliceosomal introns in the green alga Micromonas pusilla and the heterokont alga Aureococcus anophagefferens (Huff etal. 2016). Together, these IEs of transposon-like features are potentially important for the transposition and proliferation of modern splicesomal introns.
To better understand the origin and spread of spliceosomal introns, we turned to the fungal genus Zymoseptoria, in which transposition of spliceosomal introns between unrelated-genes and IPAPs (intron presence–absence polymorphisms) at homologous loci among conspecific strains have been characterized (Torriani etal. 2011; van der Burgt etal. 2012; Collemare etal. 2013; Brunner etal. 2014). Taking advantage of the population genomics data of this genus, we evaluated intron dynamics at different evolutionary scales, namely, within the genome, and among conspecific and interspecific genomes. Our results suggest that spliceosomal introns can proliferate via intergenic homolog elements within the genome, invade other genomes within the same species and also between distinct species.
Materials and Methods
Introns Verification and Improved Annotation of Intergenic Regions in the Z. tritici IPO323 Genome
The annotated genome (containing 21 chromosomes) of Zymoseptoria tritici IPO323 (accession: ACPE00000000) (Goodwin etal. 2011) was downloaded from the GenBank database (supplementary table S1, Supplementary Material online). Annotated intron sequences were extracted based on their GenBank annotation. The raw transcriptome data (SRA accessions: ERR789217 and ERR789232) (Rudd etal. 2015) were downloaded from the NCBI SRA database and mapped to the IPO323 genome using TopHat (Kim etal. 2013). Splice junctions of each annotated intron were verified based on the mapping results. The expression level of each protein gene was measured by Cufflinks v2.2.1 (Trapnell etal. 2012). Putative RNAs within annotated introns were predicted by tRNAscan (Lowe and Eddy 1997) and Rfam 12.0 (Nawrocki etal. 2015). We considered an annotated intron as a verified intron only if 1) the whole protein coding region is transcribed with FPKM >1; 2) there are at least three reads supporting the intron being spliced out; 3) and there are no detected tRNA/small-ncRNA sequences within the intron region. To obtain a conservative view of intergenic regions, ab initio gene prediction was performed on the annotated intergenic regions using the Augustus program (Stanke and Morgenstern 2005) and regions free of any predicted genes were considered further as verified intergenic regions.
Analysis of Population Genomic Data from the Zymoseptoria Genus
The draft genome sequences of two Z. tritici (synonym Mycosphaerella graminicola) strains isolated from Iran were download from GenBank [accessions: AEYQ00000000 and AFIP00000000 (Stukenbrock etal. 2011)]. Raw sequencing data of 22 Z. tritici strains were downloaded from the NCBI SRA database. Nine were isolated from Switzerland, 13 were isolated from Australia, and their BioProject numbers are PRJNA178194 (Croll etal. 2013) and PRJNA299857 (McDonald etal. 2016), respectively. The draft genome sequences of two other Zymoseptoria species, Z. pseudotritici and Z. ardabiliae, were also analyzed. There are five Z. pseudotritici strains and four Z. ardabiliae strains first described in Stukenbrock etal. (2011). De novo assembly was performed using SPAdes with four different K-mers (21, 33, 55, and 77) (Bankevich etal. 2012). The assembled genomes are available upon request. The statistics of genome assembly is shown in supplementary table S2, Supplementary Material online.
Identification of IIHEs
The verified introns were used as queries to search for potential homologs in the redefined intergenic regions. Only homologous sequences with ≥ 80% sequence identity and ≥ 80% match length were collected. To further minimize the possibility of identified intergenic regions being protein genes, the ±200 nt flanking sequences of each identified IIHE were extracted for BLASTX searches (E-value < 1 ×10−5) against all predicted proteomes of currently available (as of July 2017) species (n = 43) from the Dothideomycetes class in GenBank (supplementary table S3, Supplementary Material online), which Zymoseptoria belongs to.
Sequence Analysis of IPAP among 25 Z. tritici Strains
Fifty-two Z. tritici introns annotated in the Z. tritici genome have been previously identified as IPAP (Torriani etal. 2011). Forty-two introns were supported by available RNA-seq data and EST data (as described earlier). We conducted BLASTN searches (≥80% sequence identity and > 80% match length) using these 42 introns and their flanking exon sequences as query sequences against the assembled Zymoseptoria genomes. An intron was deemed as presence if the intron and flanking exon sequences all have significant matches, and the matched intron sequences is flanked by the matched exon sequences; an intron was deemed as absence if its flanking exon sequences both have significant matches, and the matched exon sequences are adjacent in the examined genome. Orthologous introns were determined by best reciprocal BLASTN searches at intron regions and also flanking exonic sequences. Finally, 20 IPAP introns have high-quality sequence alignment for both the introns and their flanking regions, and are subject to further analyses. The nucleotide diversity (π) of exonic sequences flanking these IPAP introns was calculated by DnaSP (Librado and Rozas 2009) in sliding windows of 50 nt and step sizes of 10 nt.
Quantification of Intron Turnover and Intron Exchange
Intron turnover and intron exchange at homologous sites were modeled as continuous-time Markov processes. Intron presence and absence were modeled as a two-state process with state 0 for absence and state 1 for presence. For intron exchange at loci of fixed introns, different intron types were treated as different character states. The rates of intron turnover and intron exchange were measured by mapping character states on a phylogeny using the tree branch length as a relative time scale using the R package DiscML (Kim and Hao 2014). As detailed in our previous studies, the rate is relative to nucleotide substitution rate, and the unit is expressed as the number of gains/losses per site per nucleotide substitution (Hao and Golding 2006, 2010; Wu and Hao 2014; Wu etal. 2015).
Identification of Introgression between Different Zymoseptoria Species
Pairwise genetic distance was calculated among the Z. tritici strains, between the Z. tritici and Z. pseudotritici strains, and between the Z. tritici and Z. ardabiliae strains using the Jukes-Cantor model (Jukes and Cantor 1969). If an interspecific (between species) genetic distance is smaller than any conspecific (within-species) distance, this suggests the occurrence of introgression between the two species. To determine whether the introgression phenomenon is more predominant in introns, we compared the results between intron sequences and their protein coding genes.
Phylogenetic Analysis of Species and Strains Relationship
The 1,829 single-copy genes that are universally present in all examined 25 Z. tritici strains, five Z. pseudotritici strains and four Z. ardabiliae strains were used to construct phylogenetic relationships. Intron sequences were removed from the phylogenetic analyses. Each gene was aligned individually using MUSCLE (Edgar 2004). The concatenated sequences of all gene alignments were used to reconstruct the phylogenetic relationship of these strains. Phylogenetic trees were constructed using the RAxML program (Stamatakis 2006) under a GTR + Γ + I substitution model, and 100 bootstrap iterations were performed.
Results
Intergenic Regions as Reservoirs of Modern Spliceosomal Introns
Modern spliceosomal introns have been suggested to have transposon-like features since the discovery of IEs, which are introns sharing high sequence similarity with other introns within the genome (Worden etal. 2009). A recent study has indicated that IEs could transpose between different genes by mechanisms similar to transposons (Huff etal. 2016). If spliceosomal introns are similar to mobile elements, one would expect introns and related elements to transpose to regions beyond gene regions. To test this, we searched for intron homologs in intergenic regions within the only annotated Zymoseptoria tritici genome, IPO323 (Goodwin etal. 2011). In the Z. tritici IPO323 genome, there are 17, 357 annotated introns. Among these introns, 10, 192 introns met our verification criteria (see Materials and Methods) and were considered as verified introns. We then performed ab initio gene predictions on all annotated intergenic sequences, identified 738 putative genes and removed these regions from the intergenic sequences. Through a stringent BLAST search (≥ 80% sequence identity and ≥ 80% match length) using the 10,192 verified introns against the redefined intergenic sequences, we identified 83 introns from 74 genes having homologs in intergenic regions. To further minimize unrecognized genes in intergenic sequences, the flanking sequences (±200 nt) of intron homologs in intergenic regions were used to search against the proteomic sequences of 43 species in the Dothideomycetes class. After the exclusion of the intron homologs whose flanking sequences in intergenic regions match proteomic sequences, 50 introns from 44 genes were considered to have homologs in the re-defined intergenic regions. Hereafter, we call these intron homologs in intergenic regions as intergenic intron homolog elements (IIHEs).
The levels of sequence similarity between introns and their intergenic homologs are often very high, but decrease dramatically in the flanking regions (fig. 1A). We found that insertion sites of introns and their intergenic homologs could be flanked by duplicated sequences (TSD in supplementary fig S1, Supplementary Material online). TSD is a characteristic mark of transposon insertion or remnant of NHEJ, and both processes have been shown to be involved in intron mobility (Li etal. 2009; Huff etal. 2016). Introns having IIHE(s) are of different sequence propensities compared with introns not having IIHE(s). Introns having IIHE(s) are longer (P-value < 0.0001, Wilcoxon Test) and skewed towards the middle and 3′ end of the gene, while the ones not having IIHE(s) are shorter and skewed towards the 5′ end of the gene (fig. 1B and C). This is consistent with the notion that intron loss preferentially takes place toward the 3′-end of the gene (Mourier and Jeffares 2003), resulting in empty intron sites known as protosplice sites where new introns tend to reinvade (Sverdlov etal. 2004).
Among the introns having IIHEs, 40% of them share high sequence similarity with at least one other intron (≥80% sequence identity and ≥80% match length). These introns have homologs both in intron and intergenic regions, suggesting a dynamic transposition of intron homologs within a genome. We further investigated the association between IIHEs and intron dynamics at the population level using IPAP as an indicator for recent intron movement. A total of 52 IPAP loci have been previously reported in Torriani etal. (2011), 42 intron loci are verified by available transcriptome data (Rudd etal. 2015) and EST data (Torriani etal. 2011). Among the 42 verified introns, nine of them (21%) have IIHEs. In comparison, among the 10, 150 fixed introns, only 41 introns (0.4%) have IIHEs. This significant contrast (P-value < 0.0001, Fisher’s exact test) suggests that IIHEs are more strongly associated with IPAP than with fixed intron, and likely contribute to recent intron turnover in Z. tritici.
Rapid Intron Turnover and Exchange among Conspecific Strains
We first took advantage of the introns of IPAP to estimate intron turnover rates among 25 conspecific strains. Introns from 20 IPAP loci (supplementary fig. S2, Supplementary Material online) were mapped on the Z. tritici phylogeny (supplementary fig. S3, Supplementary Material online) and the rates of intron gain and intron loss were measured. Our results show a higher overall rate of intron gain (152.7 ± 16.5) than intron loss (83.9 ± 9.1) (fig. 2A) and the likelihood ratio test favors different rates between intron gain and intron loss (2ΔlnL = 4.2, P-value < 0.05, df = 1). As the estimated turnover rates are relative to the nucleotide substitution rate (Hao and Golding 2006; Wu and Hao 2014; Wu etal. 2015), both the rates of intron gain and intron loss are considered to be fast. A difference between rates of intron gain and intron loss is also evident in population genomics and genetic crosses data (Torriani etal. 2011; Brunner etal. 2014). For example, among 43 F1 progenies after cross between ID39491i2-containing and ID39491i2-lacking strains, 30 progenies have intron-present alleles (IPAs) and 13 have intron-absent alleles (IAAs), showing a significant bias towards intron invasion (P-value = 0.009, χ2 test).
To gain insights into the molecular mechanisms underlying the spread of introns, we carefully analyzed the ±500 nt (nucleotides) exonic sequences immediately flanking intron insertion sites from 20 IPAP loci (supplementary fig. S2, Supplementary Material online). Overall, nucleotide diversity (π) of exonic sequences is low near intron insertion sites (flanking ±100 nt) among IPAs, and also among IAAs (fig. 2B). Strikingly, however, the same regions are among regions with the highest nucleotide diversity between IPAs and IAAs (fig. 2B), suggesting a tight association between the absence/presence of introns and the evolutionary history of exonic sequences immediately flanking intron insertion sites. We then sought to map the patterns of intron presence/absence and ±100 nt exonic sequences (only nucleotide sites that have variation among strains are shown) immediately flanking intron-insertion sites onto the organismal tree (supplementary fig. S3, Supplementary Material online). Exonic sequences near intron insertion sites show a strong association with the presence or absence of the corresponding intron, but little association with the organismal phylogenetic relationship or geographic distribution (fig. 2C and supplementary fig. S2, Supplementary Material online). The narrow flanking regions (±100 nt) associated with intron presence/absence are unlikely due to chromosomal recombination events during meiosis, whose interval distances in Z. tritici have been measured to be 3 kb or longer (Croll etal. 2015). Such narrow regions flanking intron presence/absence resemble the co-conversion tracts flanking selfish group I/II introns generated by gene conversion (Lazowska etal. 1994; Sanchez-Puerta etal. 2008; Wu and Hao 2014).
We extended our investigation on whether intron exchange takes place at loci of fixed introns (introns present in all Z. tritici strains), which would not result in IPAP. We employed BLASTClust (100% sequence coverage and 100% nucleotide identity) to group intron sequences from each homologous locus. Introns with no sequence variation and the ones forming more than four BLASTClust groups were excluded. Finally, we found introns from 463 loci (5% of total verified introns) showing evidence of sequences exchange between strains according to their phylogenetic positions on the organismal tree (supplementary fig. S3, Supplementary Material online). Among the Z. tritici strains isolated from Australia, nine of them show low levels of overall intron sequence diversity (the top nine strains in fig. 3A) due to a recent and perhaps severe population bottleneck (Banke etal. 2004). Nevertheless, among the 463 intron loci, 186 loci (40%) clearly support that at least one of the nine postbottleneck strains is different from other postbottleneck strains, but identical with some “organismally” more distantly related strains. Thus, there is a significant amount of intron exchange between some of the postbottleneck strains and prebottleneck strains. For example, the ID64923i1 intron in the WAI147 from Australia is identical with the one in the STIR04A48b strain from Iran but very different from the other eight postbottleneck strains (fig. 3C and supplementary fig. S4, Supplementary Material online). Furthermore, we modeled different intron types as different states and estimated the rates of intron exchange at each of the 463 loci of fixed introns. Under an equal rate model, the rates of intron exchange (± SE) were estimated from 14.8 ± 10.7 to 555.6 ± 249.1 with a median of 98 (fig. 3B). This suggests that fixed introns can undergo intron exchange at rates comparable to intron turnover at IPAP loci.
Intron Introgression among Different Zymoseptoria Species
We expanded comparative analysis of intron homologs among the Z. tritici strains to their closely related species Zymoseptoria pseudotritici (five strains) and Z. ardabiliae (four strains). Twenty-two intron loci were found to be of IPAP among the three Zymoseptoria species (Torriani etal. 2011). As illustrated in three intron loci, the exonic sequences immediately flanking intron insertion site are tightly associated with the presence or absence of the intron in different strains, but independent of their organismal relationship (supplementary fig. S5, Supplementary Material online). For example, the exonic sequences immediately flanking the ID77869i3 intron in Z. tritici break into two utterly different groups, with the 13 intron-containing Z. tritici strains clustered together with seven intron-containing strains in Z. pseudotritici and Z. ardabiliae. In contrast, the entire coding sequences of the ID77869 gene support a clear separation of the three well-defined species (fig. 4). The different phylogenetic relationships between the intron and coding sequences suggest that the introgression of the ID77869i3 intron is restricted to a very short region, not likely mediated by exchange in large sequences fragments. The strikingly similar or identical sequences immediately flanking the intron insertion site but not the entire protein gene cannot be simply explained by random point mutation, but favor the notion that introgression takes place at regions restricted to introns and their short flanking regions. This type of fine-scale sequence exchange between different species, as sources for diversity, can also explain the high sequence diversity flanking intron insertion sites between IAAs and IPAs among conspecific strains (fig. 2C and supplementary fig. S2, Supplementary Material online).
Intron introgression also takes place extensively at loci of fixed introns. Among 5,506 orthologous introns commonly shared by all 25 examined Z. tritici and 4 Z. pseudotritici strains, 824 introns support a higher interspecific sequence similarity than between conspecific ones. The 824 introns involved in introgression are from 712 protein genes, but 544 (76%) of these protein genes support a monophyletic relationship of all Z. tritici strains based on phylogenetic analysis of entire protein gene. For instance, the five orthologs of the ID97336i1 intron in Z. pseudotritici share 100% identity with two Z. tritici strains (fig. 5A), but the sequences of the entire ID97336 gene show a clear separation between Z. pseudotritici and Z. tritici (fig. 5B). Similarly, 484 introns were involved in introgression between Z. tritici and Z. ardabiliae. These introgressed introns are from 438 protein genes but 371 (85%) of these genes support a monophyletic relationship of all Z. tritici strains (one case shown in supplementary fig. S6, Supplementary Material online). These results suggest that intron introgression is highly specific to the intron regions. Such targeted and extensive intron introgression is important in the wide spread of introns among a broad spectrum of organisms.
Discussions
Our findings show that IIHEs are associated with intron movement within the genome and rapid intron turnover among different genomes in Z. tritici. It is noteworthy that the detected introns having IIHEs account for only a small fraction (∼0.5%) of the total number of verified introns. This is due to our stringent criteria used to define IIHEs, and also the fact that many introns are lost during the processes of rapid intron turnover and exchange. The detected introns having IIHEs may be viewed as the visible tip of the largely hidden iceberg of all (both extant and extinct) introns once having IIHEs. IIHE-mediated intron movement is perhaps a general phenomenon during intron evolution. In fact, intergenic intron homologs have also been observed in algae species Micromonas (van Baren etal. 2016), which has undergone recent massive intron invasion (Simmons etal. 2015).
It has been acknowledged that direct introduction of group II introns into nuclear-encoded protein genes is deleterious and unlikely to reach fixation (Doolittle 2014; Qu etal. 2014). IIHEs can fill the gap between the proposed origin of likely deleterious group II intron and observed wide-distribution of modern spliceosomal introns. With intergenic homologs as intermediate stage, insertion of group II introns into intergenic regions in eukaryotes would be less deleterious, provide a buffered environment for subsequent mutations to accumulate, lead to minimal deleterious effects and ultimately facilitate intron proliferation. Unlike typical group II introns, which often encode reverse transcriptase genes, sequence analysis of IIHEs in this study revealed no transposase or reverse transcriptase homologs. The IIHEs are rather similar to the miniature group II introns or miniature transposons that do not encode genes for their transposition. Unfortunately, the genes responsive for the transposition of spliceosomal introns and related IIHEs are still unknown. Further studies are needed to determine the key genes and underlying molecular mechanisms.
We measured very fast rates of intron gain and loss at IPAP loci in Z. tritici, which have been suggested in Neurospora tetrasperma and Daphnia (Li etal. 2014; Sun etal. 2015). Similar to the dynamic IPAP loci, the loci of fixed introns also show fast rates of intron exchange. Furthermore, a substantial amount of spliceosomal introns are involved in introgression between different species. It is important to note that these introgression cases of spliceosomal introns are fundamentally different from the published horizontal transfer cases of nuclear introns in fungi and algae (Hibbett 1996; Nikoh and Fukatsu 2001; Coates etal. 2002; Simon etal. 2005), which are self-splicing group I introns. As spliceosomal introns are believed to originate from group II introns (Rogers 1990; Lambowitz and Zimmerly 2004), similar features between sequences and structures have been well documented (Zimmerly and Semper 2015). Here, we observed additional features shared between spliceosomal introns and self-splicing introns with respect to their mobility. 1) Both spliceosomal introns (e.g., in Z. tritici) and group I/II introns (Dai and Zimmerly 2002; Wu etal. 2015) may have IPAP among conspecific strains and show high mobility. 2) Splicesomal introns and self-splicing introns could have co-conversion tracts flanking their insertion sites, which are strongly associated with intron invasion and recombination (Lazowska etal. 1994; Moran etal. 1995; Sanchez-Puerta etal. 2008; Wu and Hao 2014). 3) Both spliceosomal introns and self-splicing introns could show invasive nonmendelian transmission with a significant bias towards intron gain in genetic crosses between intron-containing and intron-lacking strains. This has been seen in a spliceosomal intron (the ID39491i2 intron) in Z. tritici (Torriani etal. 2011) and in group I/II introns in Saccharomyces cerevisiae (Jacquier and Dujon 1985; Moran etal. 1995).
Conclusion
In this study, we demonstrated three crucial stages in the lifecycle of spliceosomal introns, where intron elements can proliferate within the genome, spread horizontally among conspecific strains, and introgress between related species. These processes are very similar to those involving mobile introns and transposable elements (Goddard and Burt 1999; Williams 1999; Schaack etal. 2010), and also consistent with the hypothesis that spliceosomal introns are of a transposon origin (Roy 2004). These processes can partially account for the paradox that rich modern introns exist in a variety of eukaryotic genomes under generally very high rates of intron loss (Csuros etal. 2011). Given the highly variable tempo of spliceosomal intron movement among eukaryotic organisms (Worden etal. 2009; Torriani etal. 2011; van der Burgt etal. 2012; van Baren etal. 2016), we expect the rapidly growing population genomics data, high-quality genome assembly and annotation from diverse eukaryotic organisms to shed new lights on the evolution of modern spliceosomal introns.
Supplementary Material
Supplementary data are available at Genome Biology and Evolution online.
Acknowledgments
The authors thank three anonymous reviewers for their helpful comments. They are grateful for the grid computing service from Computing and Information Technology of Wayne State University. This work was supported by funds from Wayne State University (to W.H.).
Author Contributions
B.W. and W.H. designed the study. B.W. and A.I.M. performed experiments and analyzed data. B.W., A.I.M., and W.H. wrote the paper.
Literature Cited
Author notes
Associate editor: Dennis Lavrov