Molecular Population Genetics of Herbivore-induced Protease Inhibitor Genes in European Aspen (Populus tremula L., Salicaceae)
http://www.100md.com
分子生物学进展 2005年第9期
Ume? Plant Science Centre, Department of Ecology and Environmental Science, University of Ume?, SE-891 87 Ume?, Sweden
E-mail: pelle@eg.umu.se.
Abstract
Plants defend themselves against the attack of natural enemies by using an array of both constitutively expressed and induced defenses. Long-lived woody perennials are overrepresented among plant species that show strong induced defense responses, whereas annual plants and crop species are underrepresented. However, most studies of plant defense genes have been performed on annual or short-lived perennial weeds or crop species. Here I use molecular population genetic methods to survey six wound-inducible protease inhibitors (PIs) in a long-lived woody, perennial plant species, the European aspen (Populus tremula), to evaluate the likelihood of either recurrent selective sweeps or balancing selection maintaining amino acid polymorphisms in these genes. The results show that none of the six PI genes have reduced diversities at synonymous sites, as would be expected in the presence of recurrent selective sweeps. However, several genes show some evidence of nonneutral evolution such as enhanced linkage disequilibrium and a large number of high-frequency–derived mutations. A group of at least four Kunitz trypsin inhibitor genes appear to have experienced elevated levels of nonsynonymous substitutions, indicating allelic turnover on an evolutionary timescale. One gene, TI1, has enhanced levels of intraspecific polymorphism at nonsynonymous sites and also has an unusual haplotype structure characterized by two divergent haplotypes occurring at roughly equal frequencies in the sample. One haplotype has very low levels of intraallelic nucleotide diversity, whereas the other haplotype has levels of diversity comparable to other genes in P. tremula. Patterns of sequence diversity at TI1 do not fit a simple model of either balancing selection or recurrent selective sweeps. This suggests that selection at TI1 is more complex, possibly involving allelic cycling.
Key Words: arms race ? balancing selection ? herbivory ? Populus tremula ? protease inhibitor ? selective sweep ? trench warfare
Introduction
Plants have evolved numerous adaptations to defend themselves against attack by herbivores, including mechanisms to deter, starve, or poison potential herbivores (Marquis 1992). Herbivores are expected, however, to respond to plant defenses by evolving counteradaptations that render defenses less effective or even useless. Such counterselection can ultimately lead to an "arms race" between plants and herbivores, and theoretical models predict that such coevolutionary interactions will lead to an escalation of traits in both plants and herbivores (Bergelson, Dwyer, and Emerson 2001). At the molecular level, such an arms race is expected to lead to the sequential sweep of new defense and counterdefense alleles in the plant and herbivore populations, respectively (Kniskern and Rausher 2001; DeMeux and Mitchell-Olds 2003). A coevolutionary interaction between plants and herbivores characterized by an arms race model therefore predicts that defense genes should exhibit low levels of standing nucleotide sequence variation and high levels of amino acid differentiation (compared to closely related species), induced by strong directional selection for rapid turnover of alleles (Bergelson et al. 2001; DeMeux and Mitchell-Olds 2003).
Alternatively, natural selection may lead to increased diversity in defense genes, either because frequency-dependent selection favors rare alleles or because variability itself is favorable if different alleles confer resistance to different herbivore genotypes (DeMeux and Mitchell-Olds 2003). Such a model for plant-herbivore coevolution, sometimes termed "the trench warfare model" (Stahl et al. 1999) or the "recycling polymorphism" model (Holub 2001), predicts qualitatively different evolutionary dynamics compared to the arms race model. Because different alleles are maintained over long periods in a polymorphic state, balancing selection is expected to result in enhanced levels of sequence diversity because different allelic lineages accumulate mutations more or less independently. Therefore, a trench warfare model of coevolution predicts that defense genes should show enhanced levels of segregating amino acid polymorphisms within populations that sometimes even exceed levels of silent variation (DeMeux and Mitchell-Olds 2003). Such natural selection, favoring allelic variability, has been implicated in the maintenance of hypervariability in both human major histocompatibility complex genes (Ohta 1998) and pathogen recognition genes in many plants (Bergelson et al. 2001).
A growing number of studies are focusing on the molecular population genetic aspects of plant-herbivore and plant-pathogen interactions, and these studies have shown support for both the arms race and trench warfare mode of coevolution (reviewed in DeMeux and Mitchell-Olds 2003). One conclusion emerging from these studies is that specialist defenses, that is, defenses against a small number of natural enemies, are more likely to show deviations from neutral expectations (Tiffin, Hacker, and Gaut 2004). In particular, specialist defenses appear to be more likely to be under the influence of balancing selection (Tiffin, Hacker, and Gaut 2004). In contrast, many generalist defenses, that is, defenses effective against a broad spectrum of enemies, show patterns of intraspecific sequence diversity that either are consistent with a neutral model (Kawabe and Miyashitya 1999; Tiffin and Gaut 2001; Clauss and Mitchell-Olds 2004; Tiffin, Hacker, and Gaut 2004) or show signs of recent positive selection (i.e., selective sweeps, Clauss and Mitchell-Olds 2004; Tiffin 2004).
Plants defend themselves against the attack of natural enemies by using an array of constitutively expressed defenses, such as thorns, spines, and many toxic secondary metabolites (Karban and Baldwin 1997). Plants also actively defend themselves using inducible defense mechanisms (Karban and Baldwin 1997). The active part of a plant's defense arsenal involves, for instance, oxidative enzymes, genes involved in strengthening the cell wall, and/or a wide array of protease inhibitors (PIs) (Constabel 1999). PIs, in particular, are a class of wound-inducible defense genes for which function in herbivore defense is well established and where the physiological mechanisms underlying the defense function are well studied (Karban and Baldwin 1997; Koiwa, Bressan, and Hasegawa 1997; Constabel 1999; Haq, Atif, and Khan 2004). Plant PIs are proteins that function as specific substrates for proteolytic enzymes in the digestive tracts of herbivores (Ryan 1990; Constabel 1999; Haq, Atif, and Khan 2004). However, a PI is not cleaved by a protease, like normal substrates are, but rather forms a stable complex that limits or completely inhibits the proteolytic activity of the protease (Ryan 1990; Constabel 1999; Haq, Atif, and Khan 2004). This results in much reduced proteolysis in the digestive tract, a lack of available amino acids, and ultimately lowered growth rates or starvation of the herbivore (Ryan 1990; Constabel 1999). In addition to their antinutritive effects, PIs may also have toxic or directly lethal effects on herbivores (Duffey and Stout 1996). There are several classes of PIs in plants, each effective against a different class of proteases. The two most common classes of PIs are inhibitors of serine and cysteine proteases, and this is thought to reflect the predominance of these proteases in herbivore digestive systems (Koiwa, Bressan, and Hasegawa 1997; Constabel 1999; Haq, Atif, and Khan 2004). Serine proteases are the predominant digestive enzyme in Lepidoptera, whereas both serine and cysteine PIs are commonly found in Coleoptera (Koiwa, Bressan, and Hasegawa 1997; Haq, Atif, and Khan 2004).
Recent studies of transcriptional patterns following wounding by herbivores have shown that a large number of defense-related genes are upregulated following tissue damage in Populus (Christopher et al. 2004). In particular, a wide range of PIs are among the most upregulated genes following damage by herbivores (Haruta et al. 2001; Christopher et al. 2004). Haruta et al. (2001) identified three wound-induced genes from Populus tremuloides, belonging to the Kunitz trypsin inhibitor class of serine PIs. These genes, TI1, TI2, and TI3, belong to a small gene family and show amino acid sequence similarities ranging from 52% to 83% (Haruta et al. 2001). In particular, TI1 and TI2 appear to be the result of a recent gene duplication as these genes are over 90% identical at the nucelotide level. Christopher et al. (2004) identified two additional Kunitz type PIs from a Populus trichocarpa x Populus deltoides hybrid. These two genes, named TI4 and TI5, represent more diverged members of the Kunitz TI class, sharing less than 30% amino acid similarity with TI1, TI2, and TI3. The transcription of all five TI genes is highly upregulated following mechanical wounding or herbivory, and all five genes can also be induced by external application of the defense signal jasmonate (Haruta et al. 2001; Christopher et al. 2004). However, the patterns of induction of the five genes are quite different and distinct, both locally and systemically throughout the plant (Haruta et al. 2001; Christopher et al. 2004). In addition to the TI genes, several other PIs show strong induction following wounding, including highly upregulated cysteine PI, CI1 (Christopher et al. 2004).
In this paper, I use molecular population genetic analyses to study the evolution of wound-inducible PIs in the European aspen (Populus tremula). I do so by analyzing in more detail previously published surveys of natural variation in TI3 and CI1 (Ingvarsson 2005) and by adding data on patterns of nucleotide diversity in four additional PI genes (TI1, TI2, TI4, and TI5) from the same set of individuals. These six genes are surveyed for departures from neutral expectations, with a special emphasis placed on evaluating the likelihood of either recurrent selective sweeps or balancing selection maintaining amino acid polymorphisms.
Materials and Methods
Sequence Collection
The five serine PIs studied (TI1 to TI5) are all members of the Kunitz TI class, although TI1, TI2, and TI3 appear to form a small gene family of more closely related sequences because they share between 65% and 85% amino acid identity (Haruta et al. 2001). TI4 and TI5 represent more diverged members of the Kunitz TI class, sharing 16%–25% amino acid identity with the TI1–TI3 group of sequences (Christopher et al. 2004). Sequences of TI3 and CI1 were obtained from Ingvarsson (2005). New sequences of TI1, TI2, TI4, and TI5 were collected from the same set of P. tremula trees that were included in Ingvarsson (2005). Leaves were collected from four different sites throughout Europe: Besacon in eastern France, Klagenfurt in southern Austria, F?rjestaden in southeastern Sweden, and Ume? in northern Sweden. TI1 and TI2 were also sequenced from an additional population located close to Halmstad, in southwestern Sweden, about 300 km west of the F?rjestaden population.
Primers to amplify TI1, TI2, TI4, and TI5 from P. tremula were designed from the publicly available GenBank records for these genes (accession numbers AF349441, AF349442, AY378089, and AY378090, respectively). The entire coding sequence was obtained for all genes, including some 5'- and 3'-flanking sequences for TI1, TI2, and CI1. All genes are intronless, except CI which contains two exons, separated by a short intron. Finally, homologous sequences from P. trichocarpa were obtained for all genes by Blast searches of the assembled genome sequence of P. trichocarpa available at http://genome.jgi-psf.org/.
Polymerase chain reaction products were cloned into the pCR2.1 vector using a TA-cloning kit from Invitrogen (Carlsbad, Calif.). At least five, and often more, different clones of each fragment were sequenced using BigDye chemistry (Applied Biosystems Inc., Foster City, Calif.) on an ABI377 automated sequencer at the Ume? Plant Science Centre sequencing facility. Sequences were verified manually and contigs were assembled using the computer program Sequencer v 4.0. Multiple sequence alignments were made using ClustalW (Thompson, Higgins, and Gibson 1994) and adjusted manually using BioEdit (http://www.mbio.ncsu.edu/BioEdit/bioedit.html). All sequences described in this paper have been deposited in the EMBL database (accession numbers AJ842907–AJ842952, AJ843666–AJ843713, and AJ936970–AJ937139).
Estimates of nucleotide polymorphism and statistical tests of neutrality were obtained using the computer program DnaSP v4.00.5 (http://www.ub.es/dnasp/) or Jody Hey's SITES program (http://lifesci.rutgers.edu/heylab/HeylabSoftware.htm#SITES). Significance of statistical tests of neutrality were evaluated by determining the null distribution for each test statistic under neutrality by simulating 105 neutral genealogies using Richard Hudson's ms program (available at http://home.uchicago.edu/rhudson1/source/mksamples.html). Briefly, simulations were conditioned on the original sample configuration, sequence lengths, the number of segregating sites observed, and either the empirically determined recombination rate, Chw = 4Nr, using the method of Hey and Wakeley (1997) or Cmin, the minimum recombination rate compatible with the number of inferred recombination events (Rozas et al. 2001). Cmin can be viewed as a lower bound of the recombination rate and will therefore provide an underestimate of the true recombination rate. Because population structure may influence both power and rejection rates of statistical tests of neutrality (Przeworski 2002), simulations were run with samples taken from a subdivided population. The migration parameter, M = 4Nm, was set so that the expected value of FST = 1 – within/total (Charlesworth 1998) matched that observed at each locus. The population structure in the simulations was assumed to follow the Wright's island model with a total number of populations equal to 20.
Data on nucleotide polymorphism and divergence in the six PI genes was compared to several other genes from P. tremula that are not known to be involved in herbivore defense, Adh1, Gapdh, and GA20ox1 (Ingvarsson 2005) and phyB2, ABI1B, ABI1D, and Adh2 (Garcia and Ingvarsson, unpublished data). These genes were sequenced from the same set of individuals that were included in this study.
Results
Structure of Sequenced Regions
Between 33 and 47 alleles were sequenced from the six PI genes. The length of the sequenced regions ranged from 630 bp for TI3 to 802 bp for CI1. The coding regions of the five TI genes range in size from 606 to 684 bp (table 1). The predicted protein sequence of all five TI proteins contains a signal peptide (Nielsen et al. 1997), consisting of 25 amino acids, that is cleaved off to form the active PI. The mature TI protein consists of 10–12 antiparallel ? strands that form the ?-trefoil fold characteristic of Kunitz type PIs (Song and Suh 1998) and carries a reactive region situated on an external loop. The CI1 gene contains a coding region that is 429 bp long and encodes a protein showing strong similarity to other phytocystatins. After cleavage of a 27–amino acid signal peptide (Nielsen et al. 1997), the mature protein forms two hairpin loops and a largely unstructured N-terminal protein end (Brown and Dziegielewska 1997). The first hairpin loop carries the conserved QxSxG motif, thought to be the site active in protein inhibition, and the second hairpin loop carries the conserved PW sequence motif present in all cystatins (Brown and Dziegielewska 1997).
Table 1 Intraspecific Polymorphism in Populus tremula Wound-induced Protease Inhibitor Genes
Intraspecific Patterns of Nucleotide Diversity
There is abundant variation in all genes, with the number of segregating sites ranging from 31 to 62 (table 1, Appendix). Estimates of Watterson's = 4Nμ range from 0.0112 to 0.0191, which is within the range of earlier estimates of obtained from other P. tremula genes (Ingvarsson 2005 and unpublished data). In addition to base pair substitutions, several indels were polymorphic in the sample. Two different indels, each resulting in the deletion of a compete codon, were segregating in TI1. TI2 contained four indels that were polymorphic in the sample, three indels resulting in the deletion of a complete codon each and a larger indel causing the deletion of four contiguous codons. TI4 contained two polymorphic indels, one resulting in the deletion of a complete codon and one causing the deletion of two contiguous codons. TI5 was also polymorphic for a two-codon deletion. Note that none of these indels disrupted the open reading frame, and all alleles at the five TI genes were predicted to be fully functional based on their amino acid sequence. CI1 only contained indels in the intron, and these were all associated with mononucleotide repeats that differed in length.
If the PI genes were under recurrent directional selection due to continual fixation of new beneficial alleles, silent site diversity is expected to be sharply reduced. There does not, however, appear to be any evidence for a reduction in nucleotide diversity at synonymous sites in the PI genes studied (table 1), compared with other gene regions that are not involved in herbivore defense in P. tremula (Ingvarsson 2005, unpublished data). In fact, synonymous site diversity is higher in the six PI genes, sil = 0.0207, compared to nondefense genes, where sil = 0.0134 (table 1). The same is true for nucleotide diversity at replacement sites, where the six PI genes have higher diversities (rep = 0.0117) than nondefense genes (rep = 0.0076). However HKA (Hudson, Kreitman, and Aguade 1987) tests indicate that neither synonymous nor replacement site diversities differ significantly between the six PI genes and nondefense genes (HKA tests, 2 = 0.213, P < 0.645, and 2 = 0.896, P < 0.561 for synonymous and nonsynonymous sites, respectively).
An interesting observation is that TI1 shows such enhanced levels of diversity at replacement sites that rep/syn exceeds unity (table 1). Such high levels of diversity at replacement sites are hard to explain without invoking natural selection. An inspection of table 1 shows that the high rep/syn ratio observed for TI1 is not attributable to unusually low synonymous diversities in this gene but rather to an elevated level of nonsynonymous polymorphism. In fact, levels of polymorphism at replacement sites in TI1 are the highest recorded among the 13 genes for which sequence data is available in P. tremula (table 1, Ingvarsson 2005, unpublished data, range for nondefense genes 2.96 x 10–3–3.44 x 10–3).
There is evidence for an excess of linkage disequilibrium in all five Kunitz trypsin inhibitor genes if the significance of ZnS (Kelly 1997) is evaluated using levels of recombination estimated from the data set (Chw), although if the lower bound on the recombination rate (Cmin) is used, none of the genes have ZnS that are significant (table 2). However, this approach provides an extremely conservative test of excessive linkage disequilibrium because Cmin estimates of the scaled recombination rate are about an order of magnitude lower than those obtained through the method of Hey and Wakeley (1997) (table 2). Enhanced levels of linkage disequilibrium can be generated by directional selection if a favorable mutation has not yet reached fixation or if recombination occurred during a selective sweep, so that variation at linked sites was not completely eradicated during the sweep (Fay and Wu 2000; Przeworski 2002; Kim and Nielsen 2004). An excess of linkage disequilibrium is also expected if natural selection maintains distinct allelic classes at the locus, as is the case with balanced polymorphisms (Charlesworth, Nordborg, and Charlesworth 1997). Finally, demographic scenarios, such as population structure or population growth, can also result in increased linkage disequilibrium compared to neutral expectations (Przeworski 2002). However, effects of demography are expected to be genome wide, and there is no indication of excessive linkage disequilibrium in other P. tremula genes (Ingvarsson 2005), suggesting that the presence of enhanced linkage disequilibrium in TI defense genes is the product of natural selection.
Table 2 Statistical Tests of Neutrality
Three of the defense genes, TI1, TI2, and TI4, also show an excess of high-frequency–derived variants, as measured by Fay and Wu's H, although this excess is only marginally significant in TI4 (P < 0.067, table 2). Under the much more conservative assumption of the low recombination (Cmin in table 2), H remains significant only for TI1 (table 2). An excess of high-frequency–derived variants is also a pattern that is likely to be explained by strong directional selection fixing newly arisen mutations because derived mutations occurring at high frequencies are rare under a neutral model (Fay and Wu 2000; Kim and Nielsen 2004). Again, demographic factors, and in particular population structure, may provide an alternative explanation for an excess of high-frequency–derived sites (Przeworski 2002). However, population subdivision is not likely to be a factor in explaining the negative values of H for TI1, TI2, and TI4 for at least two reasons. First of all, the effect of population structure on H becomes more severe with increasing strength of subdivision (Przeworski 2002), and population subdivision is generally low in P. tremula (table 1 and Ingvarsson 2005). Second, population structure was explicitly taken into account when assessing the significance of H through coalescent simulations. It is possible that the coalescent simulations did not accurately characterize the population structure of P. tremula. However, given the long-distance dispersal of both seeds and pollen seen in P. tremula, the island model used in the coalescent simulations should provide a reasonable approximation for the population structure of P. tremula, and including a model of population structure in the coalescent simulations, even if approximate, is likely better than completely ignoring it.
Polymorphism and Divergence
Divergence at silent sites in the six defense genes is similar or slightly higher than at other P. tremula loci not involved in herbivore defense (fig. 1), whereas nonsynonymous divergence is generally higher at the defense genes than in nondefense genes. In two of the defense genes, TI1 and TI2, there are signs of long-term adaptive evolution, as they show greater rates of divergence at nonsynonymous than synonymous sites, as evidenced by Ka/Ks >1 (fig. 1). Using the modified pairwise Nei-Gojobori method with p distances (Nei and Kumar 2000), as implemented in MEGA 2.1 (http://www.megasoftware.net), the Ka/Ks ratios at both TI1 and TI2 are significantly greater than one (P < 0.037 and P < 0.006, respectively). A Ka/Ks ratio exceeding unity is a clear sign of adaptive evolution, suggesting that nonsynonymous mutations offer a fitness advantage and are fixed at a higher rate than synonymous mutations (Yang 2001).
FIG. 1.— Synonymous and nonsynonymous divergence for six wound-induced protease inhibitor genes (black) and six genes not involved in herbivore defense (light gray).
Surprisingly, the McDonald and Kreitman (1991) (MK) test, which compare rates of divergence with levels of standing polymorphism within species, show no evidence for adaptive evolution in any of the TI genes. However, if polymorphism and divergence data from the five TI genes are pooled, these genes appear to have experienced elevated rates of fixation of nonsynonymous mutations (the neutrality index of Rand and Kann , NI = 0.584, P < 0.016, table 3). Excluding data from TI4, which harbors a nonsignificant excess of intraspecific polymorphism at nonsynonymous sites (table 2), this pattern of an excess of fixed nonsynonymous mutations becomes even stronger (NI = 0.476, P < 0.0019). If the same pooled comparison is made with seven protein-coding genes that are not involved in herbivore defense (Ingvarsson 2005, unpublished data), no deviation from neutral expectations is detected (NI = 1.433, P < 0.155, table 2).
Table 3 Number of Fixed and Polymorphic Mutations at Synonymous and Replacement Sites in the Coding Regions of Populus tremula Pl Genes
Contrary to the observations for the five TI genes, an MK test indicates a marginally significant (P < 0.075) excess of intraspecific polymorphism at replacement sites in CI1 (table 3). Much of this excess can be explained by the large number of singleton mutations causing amino acid replacements in CI1. There are 13 nonsynonymous mutations found only in single alleles and an additional 7 are found in two to four alleles (Appendix). This is also evident from the Tajima's D, which is significantly negative for CI1 (table 2).
Haplotype Structure at TI1
TI1 has several features suggesting that it has been influenced by natural selection, including an excess of derived mutations occurring at high frequencies, extensive linkage disequilibrium, and a/s and Ka/Ks ratios exceeding one. The gene genealogy at TI1 shows strong haplotype structure, with alleles falling into two distinct haplotypes (haplotypes A and B in fig. 2, also see Appendix) that differ by 14 fixed substitutions, 13 of which occur at replacement sites. The strong haplotype structure also coincides with both an excess of linkage disequilibrium (LD) and high proportions of high-frequency–derived sites (table 2).
FIG. 2.— Gene genealogies for 47 alleles of TI1 in Populus tremula. A sequence from Populus trichocarpa was used as the out-group. TI1-A and TI1-B refer to the two major haplotypes identified. The genealogy was constructed using the neighbor-joining method with p distances in MEGA 2.1.
One of the haplotypes (A in fig. 2) shows very little intrahaplotype variation (S = 5, = 0.00095, Appendix), despite accounting for 25 out of 47 sampled alleles. The remaining sequences, in haplotype B, are more heterogeneous and have a level of polymorphism ( = 0.0117) that is comparable to other loci in P. tremula and about an order of magnitude greater than haplotype A. To test the likelihood of observing two clades with such different levels of intraallelic variation, I used a modified version of the test devised by Hudson et al. (1994). This test evaluates the likelihood, under an equilibrium-neutral model, of observing a haplotype with only five segregating sites making up 25 out of 47 sampled alleles, conditional on the variation observed in the total sample (S = 46). I modified the original approach of Hudson et al. (1994) to explicitly take population subdivision into account in the coalescent simulations used to generate the expected distribution of intraallelic variation in haplotype A. This test shows that under an equilibrium-neutral model the present haplotype configuration is extremely unlikely (P < 0.0001), suggesting that natural selection is responsible for the observed patterns. More specifically, the low sequence diversity seen in haplotype A suggests that this haplotype may have increased in frequency recently. However, there are 14 fixed differences between haplotype A and haplotype B. Comparisons with the out-group species P. trichocarpa reveals that out of these 14 mutations, haplotype A carries the ancestral base at five sites and the derived base at three sites. The ancestral relationship cannot be determined for six of the sites either because the presence of a third base in P. trichocarpa or because the site is deleted in P. trichocarpa. Because the number of derived sites is roughly evenly dispersed between haplotypes A and B, it is unlikely that haplotype A represents a novel haplotype that evolved recently and increased in frequency because of positive selection. A more likely scenario is that the haplotypes A and B have been maintained for long periods of time, through some form of balancing selection. However, patterns of sequence diversity within the two allelic haplotypes are not consistent with a model of balancing selection where the relative frequency of the two haplotypes has been stable over time. Innan and Tajima (1999) showed that if balancing selection maintains two distinct allelic classes at a locus at constant frequencies, the sum of pairwise differences within the two classes is roughly constant and equal to , regardless of the strength and pattern of selection. For TI1 the sum of pairwise diversities within haplotypes A and B equals = 0.01263, which is significantly less (P < 0.05) than the pairwise diversity observed for the total sample (P = 0.0191). A model of constant balancing selection thus appears to provide a poor fit to the TI1 data as well.
Discussion
Sequence Diversity and Evidence for Nonneutral Evolution
In this study I have examined patterns of nucleotide variation in six PIs, five serine PI and one cysteine PI, from European aspen (P. tremula). Although there are differences among the six genes, several signs point to the fact that the evolution of these defense genes is not well characterized by a neutral model. There is no apparent reduction in levels of polymorphism, either at silent or replacement sites in any of the six genes, compared to P. tremula genes that are not involved in herbivore defense (table 1). In particular, the TI1 and TI2 genes stand out by having Ka/Ks ratios that exceed one (fig. 1). A Ka/Ks ratio exceeding unity suggest that nonsynonymous mutations are fixed at a higher rate than synonymous mutations and is a clear sign of adaptive evolution (Yang 2001). Both TI1 and TI2 also have an excess of high-frequency–derived sites (table 2) and excess LD, suggesting that the recent evolutionary histories at these two genes are characterized by one or more selective sweeps that have driven newly arisen mutation to high frequency in the population.
Of the other four genes surveyed, TI3, TI4, and TI5 all show an excess of LD, at least when the significance of the ZnS test is evaluated using the recombination rates inferred from the sequence data (table 2). Such excess LD could result from either directional or balancing selection, although demographic factors such as population structure could also explain these observations. There is no evidence for excess LD in P. tremula genes not involved in herbivore defense (Ingvarsson 2005), The observed patterns could thus be a product of natural selection, although patterns of intraspecific sequence diversities at TI3, TI4, and TI5 show little other evidence of having been influenced by natural selection.
There are, however, indications that the TI genes collectively are evolving in a nonneutral manner because they have an elevated rate of fixation of nonsynonymous mutations (table 3). An evolutionary history of increased amino acid substitution but little current evidence for natural selection could be indicative of episodic selection, where favorable mutations appear and rapidly sweep through the population but are then followed by long periods of selective neutrality (Tiffin and Gaut 2001).
CI1 has an excess of amino acid polymorphisms, much of which can be explained by a large number of singletons. While such an excess of low-frequency amino acid polymorphisms could be an indication of diversifying selection, an excess is also expected under a slightly deleterious model of molecular evolution, where selection is weak enough to allow polymorphism at nonsynonymous sites but where fixation is limited (Ohta 1992).
The different PI genes thus appear to have been influenced by natural selection to different degrees. TI1 and TI2 show several signs of both historical and current action of natural selection, whereas the remaining TI genes and CI1 show sporadic signs of deviations from neutral expectations. In Arabidopsis thaliana, Clauss and Mitchell-Olds (2004) showed that patterns of sequence diversity varied 10-fold across six duplicated trypsin inhibitor genes, located in tandem across a 10-kb segment on chromosome II. Interestingly, they also showed that sequence diversity in the coding regions was positively correlated to variation in expression levels of the different TI genes. They interpreted this among-locus variation in both sequence polymorphism and gene expression to be the result of selection for coping with a diverse set of both specialist and generalist herbivores through subfunctionalization of different loci (Clauss and Mitchell-Olds 2004).
Molecular population genetic data from two PIs in maize (Zea mays), wip1 and mpi, did not show any deviations from neutrality (Tiffin and Gaut 2001; Tiffin, Hacker, and Gaut 2004). Relative rate tests, however, suggested significant rate heterogeneity at wip1 with an increased rate evolution in the Zea lineage; these changes were preferentially located in one of the inhibitory loops of wip1 (Tiffin and Gaut 2001). This suggests, even when molecular population genetic analysis in contemporary populations fails to provide evidence for nonneutral evolution, that defense genes can be characterized by an evolutionary history influenced by natural selection.
Tiffin, Hacker, and Gaut (2004) argue that specialist defenses, that is, defenses active against one or a few natural enemies, are more likely to show evidence for nonneutral evolution compared to generalist defenses. Although PIs as a group are expected to target many different natural enemies, it is not clear whether individual PIs should be viewed as specialist or generalist defenses. The large number of TI genes found in P. tremula, each with their own distinct patterns of gene expression following wounding (Christopher et al. 2004), suggests that these genes might have different targets. It is possible that some of the TI genes have more generalized targets whereas others have more specialized functions, but at present there is no data to shed any light on these ideas. The same view was advocated by Clauss and Mitchell-Olds (2004), who argued that the differences in both sequence diversity and expression pattern seen among the six TI genes in their study are most likely the result of selection for subfunctionalization of the different genes. More studies are clearly needed, both of the patterns of expression following attack by different herbivores and also of the inhibitory functions at the biochemical level of the various PI genes against different herbivore digestive enzymes.
Haplotype Structure at TI1
The TI1 gene has both elevated levels of intraspecific amino acid polymorphism and a Ka/Ks ratio that exceeds one (table 1 and fig. 1). Moreover, the gene genealogy at TI1 shows a haplotype structure with two divergent haplotypes that have different levels of intrahaplotypic variation. The presence of two diverged haplotypes, with one haplotype characterized by a very low level of standing variation and another that carries more normal levels of sequence variation, has been interpreted as a hallmark signature of selective sweeps in a highly recombining genomic region (Parsch, Meiklejohn, and Hartl 2001; Quesada et al. 2003; Kim and Nielsen 2004). However, the two haplotypes at TI1 differ by 14 fixed mutations, 13 of which are nonsynonymous, although there is no difference between the two haplotypes in the number of derived mutations that they carry. The presence of two such diverged haplotypes is therefore probably a result of them having been maintained in the population over evolutionary time, even though the patterns of polymorphism observed in haplotypes A and B do not fit a model of long-term, stable balancing selection (Innan and Tajima 1999). A possible alternative explanation is a recent admixture of two previously isolated populations. This seems unlikely, however, because an admixture should leave a genome-wide signature. First, the population subdivision at TI1 is low (table 1), and both haplotypes are present in all five populations sampled across Europe, reducing the likelihood of recent admixture. Second, although the excess of LD (table 2) could be interpreted as signs of enhanced haplotype structure also at other TI genes, none of the nondefense genes show such patterns (Ingvarsson 2005). Rather, the enhanced LD could be the result of selection maintaining similar although weaker forms of the haplotype structure seen at TI1 also in other PI genes (Appendix).
Patterns of polymorphism and haplotype structure therefore suggest that neither the arms race nor trench warfare model alone is adequate for describing the evolutionary history of TI1. Rather, the data are suggestive of a model that incorporates both directional selection (arms race) and a selective maintenance of intraspecific amino acid polymorphism (trench warfare). One possible explanation that fits both of these criteria is that different resistance haplotypes are maintained through balancing or negative-frequency–dependent selection, but where frequencies of different resistance haplotypes vary over time (Seger 1992).
Tiffin, Hacker, and Gaut (2004) also showed that neither the arms race nor the trench warfare model was adequate for explaining patterns of polymorphism at the disease-resistance gene hm2 in Z. mays. Tiffin, Hacker, and Gaut (2004) found several diverged haplotypes with very little intrahaplotype variation and proposed a scenario where these haplotypes were maintained by natural selection, but the frequencies of these haplotypes varied over evolutionary time. In A. thaliana, Rose et al. (2004) also found two major allelic haplotypes at the disease-resistance gene RPP13. They interpreted their results as the consequence of negative-frequency–dependent selection, where single RPP13 alleles have not swept to fixation but have rather been maintained in the population because different alleles encode resistance specificities against different strains of the pathogen (Rose et al. 2004). The RPP13 data is clearly inconsistent with the arms race model and more in line with the predictions made by the trench warfare model. Finally, data from antibacterial peptides in Drosophila melanogaster also do not show clear support for either the arms race or the trench warfare model (Lazzaro and Clark 2003), although these genes show some signs of natural selection, such as relatively high levels of intraspecific amino acid polymorphisms and excesses of derived sites occurring at high frequencies (Lazzaro and Clark 2003).
The scenario proposed above to explain the observed patterns of nucleotide diversity at TI1 is clearly ad hoc. However, current models that have been proposed to explain coevolutionary interactions of plant-herbivore or plant-pathogen interactions at the molecular level, including the trench warfare and arms race models, are rather heuristic, and it is not clear what type of data are needed to either verify or reject these models (see also Tiffin, Hacker, and Gaut for a similar argument). The fact that almost all studies that have examined coevolutionary interactions between plant and pathogens or herbivores at the level of individual genes have found patterns of polymorphism that cannot be explained by either the arms race or trench warfare models suggests that more research is needed to specifically examine patterns of nucleotide polymorphism under a wide range of conditions both in these rather simple and in more complex models of plant-parasite coevolution (Seger 1992).
Appendix
TI1
TI2
TI3
TI4
TI5
CI1
Alignment of polymorphic sites in the six protease inhibitor genes. Residue numbers are indicated above each polymorphic site, and each polymorphic site is indicated as either nonsynonymous (N), synonymous (S), indels (D) or occurring in flanking regions or introns (I). All residues identical to the top sequence are denoted by dots (.) and gaps are given by a dash (–).
Acknowledgements
I would like to thank Carin Olofsson for obtaining the DNA sequences presented in this paper. John McDonald and three anonymous reviewers provided comments that improved the paper. I am also grateful to Barbara Giles who provided help with linguistic corrections. This research has been funded by a grant from the Swedish Research Council (Vetenskapsr?det, VR).
References
Bergelson, J., G. Dwyer, and J. J. Emerson. 2001. Models and data on plant-enemy coevolution. Annu. Rev. Genet. 35:469–499.
Bergelson, J., M. Kreitman, E. A. Stahl, and D. C. Tian. 2001. Evolutionary dynamics of plant R-genes. Science 292:2281–2285.
Brown, W. M., and K. M. Dziegielewska. 1997. Friends and relations of the cystatin superfamily—new members and their evolution. Protein Sci. 6:5–12.
Charlesworth, B. 1998. Measures of divergence between populations and the effects of forces that reduce variability. Mol. Biol. Evol. 15:538–542.
Charlesworth, B., M. Nordborg, and D. Charlesworth. 1997. The effects of local selection, balanced polymorphism and background selection on equilibrium patterns of genetic diversity in subdivided populations. Genet. Res. 70:155–174.
Christopher, M. E., M. Miranda, I. T. Major, and C. P. Constabel. 2004. Gene expression profiling of systemically wound-induced defenses in hybrid poplar. Planta 219:936–947.
Clauss, M. J., and T. M. Mitchell-Olds. 2004. Functional divergence in tandemly duplicated Arabidopsis thaliana trypsin inhibitor genes. Genetics 166:1419–1436.
Constabel, C. P. 1999. A survey of herbivore-induced defensive proteins and phytochemicals. Pp. 137–166 in A. A. Agrawal, S. Tuzun, and E. Bent, eds. Inducible plant defense against pathogens and herbivores: biochemistry, ecology and agriculture. American Phytopathological Society Press, St. Paul, Minn.
DeMeux, J., and T. M. Mitchell-Olds. 2003. Evolution of plant resistance at the molecular level: ecological context of species interactions. Heredity 94:343–352.
Duffey, S. S., and M. J. Stout. 1996. Antinutritive and toxic components of plant defense against insects. Arch. Insect Biochem. Physiol. 32:3–37.
Fay, J., and C.-I. Wu. 2000. Hitchhiking under positive Darwinian selection. Genetics 155:1405–1413.
Haq, S. K., S. M. Atif, and R. H. Khan. 2004. Protein proteinase inhibitors in combat against insects, pests and pathogens: natural and engineered phytoprotection. Arch. Biochem. Biophys. 431:145–159.
Haruta, M., I. T. Major, M. E. Christopher, J. J. Patton, and C. P. Constable. 2001. A Kunitz trypsin inhibitor gene family from trembling aspen (P. tremuloides Michx.): cloning, functional expression and induction by wounding and herbivory. Plant Mol. Biol. 46:347–359.
Hey, J., and J. Wakeley. 1997. A coalescent estimator of the population recombination rate. Genetics 145:833–846.
Holub, E. B. 2001. The arms race is ancient history in Arabidopsis thaliana, the wildflower. Nat. Rev. Genet. 2:516–527.
Hudson, R. R., K. Bailey, D. Skarecky, J. Kwiatowski, and F. J. Ayala. 1994. Evidence for positive selection in the superoxide dismutase (Sod) region of Drosophila melanogaster. Genetics 136:1329–1340.
Hudson, R. R., M. Kreitman, and M. Aguade. 1987. A test of neutral molecular evolution based on nucleotide data. Genetics 116:153–159.
Ingvarsson, P. K. 2005. Nucleotide polymorphism and linkage disequilibrium within and among natural populations of European aspen (P. tremula L., Salicaceae). Genetics 169:945–953.
Innan, H., and Tajima, F. 1999. The effect of selection on the amounts of nucleotide variation within and between allelic classes. Genet. Res. 73:15–28.
Karban, R., and I. T. Baldwin. 1997. Induced responses to herbivory. University of Chicago Press, Chicago, Ill.
Kawabe, A., and N. T. Miyashitya. 1999. DNA variation in the acidic chitinase locus (ChiB) region of the wild plant Arabidopsis thaliana. Genetics 153:1445–1453.
Kelly, J. K. 1997. A test of neutrality based on interlocus associations. Genetics 146:1197–1206.
Kim, Y., and R. Nielsen. 2004. Linkage disequilibrium as a signature of selective sweeps. Genetics 167:1513–1524.
Kniskern, J., and M. D. Rausher. 2001. Two modes of host-enemy coevolution. Popul. Ecol. 43:3–14.
Koiwa, H., R. A. Bressan, and P. M. Hasegawa. 1997. Regulation of protease inhibitors and plant defense. Trends Plant Sci. 2:379–384.
Lazzaro, B. P., and A. G. Clark. 2003. Molecular population genetics of inducible antibacterial peptide genes in Drosophila melanogaster. Mol. Biol. Evol. 20:914–923.
Marquis, R. J. 1992. Selective impact of herbivores. Pp. 203–325 in R. S. Fritz and E.-L. Simms, eds. Plant resistance to herbivores and pathogens—ecology, evolution and genetics. University of Chicago Press, Chicago, Ill.
McDonald, J. H., and M. Kreitman. 1991. Adaptive protein evolution at the Adh locus in Drosophila. Nature 351:652–654.
Nei, M., and S. Kumar. 2000. Molecular evolution and phylogenetics. Oxford University Press, Oxford.
Nielsen, J., J. Engelbrecht, S. Brunak, and G. von Heijne. 1997. Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites. Protein Eng. 10:1–6.
Ohta, T. 1992. The nearly neutral theory of molecular evolution. Annu. Rev. Ecol. Syst. 23:263–286.
———. 1998. On the pattern of polymorphisms at major histocompatibility complex loci. J. Mol. Evol. 46:633–638.
Parsch, J., C. D. Meiklejohn, and D. L. Hartl. 2001. Patterns of DNA sequence variation suggest the recent action of positive selection in the janus-ocnus region of Drosophila simulans. Genetics 159:647–657.
Przeworski, M. 2002. The signature of selection at randomly chosen loci. Genetics 160:1179–1189.
Quesada, H., U. E. M. Ramírez, J. Rozas, and M. Aguade. 2003. Large-scale adaptive hitchhiking upon high recombination in Drosophila simulans. Genetics 165:895–900.
Rand, D. M., and L. M. Kann. 1996. Excess amino acid polymorphism in mitochondrial DNA: contrasts among genes from Drosophila, mice, and humans Mol. Biol. Evol. 13:735–748.
Rose, L. E., P. D. Bittner-Eddy, C. H. Langley, E. B. Holub, R. W. Mitchelmore, and J. L. Beynon. 2004. The maintenance of extreme amino acid diversity at the disease resistance gene, RPP13, in Arabidopsis thaliana. Genetics 166:1517–1527.
Rozas, J., M. Gullaud, G. Blandin, and M. Aguade. 2001. DNA variation at the rp49 gene region in Drosophila simulans: evolutionary inferences from an unusual haplotype structure. Genetics 158:1147–1155.
Ryan, C. A. 1990. Protease inhibitors in plants: genes for improving defenses against insects and pathogens. Annu. Rev. Phytopathol. 28:425–449.
Seger, J. 1992. Evolution of exploiter-victim relationships. Pp. 3–25 in M. J. Crawley ed. Natural enemies: the population biology of predators, parasites and diseases. Blackwell Scientific, London.
Song, H. K., and S. W. Suh. 1998. Kunitz-type soybean trypsin inhibitor revisited: refined structure of its complex with porcine trypsin reveals an insight into the interaction between a homologous inhibitor from Erythrina caffra and tissue-type plasminogen activator. J. Mol. Biol. 275:347–363.
Stahl, E. A., G. Dwyer, R. Mauricio, M. Kreitman, and J. Bergelson. 1999. Dynamics of disease resistance polymorphism at the Rpm1 locus of Arabidopsis. Nature 400:667–671.
Thompson, J. D., D. G. Higgins, and T. J. Gibson. 1994. ClustalW: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22:4673–4680.
Tiffin, P. 2004. Comparative evolutionary histories of chitinase genes in the genus Zea and family Poaceae. Genetics 167:1331–1340.
Tiffin, P., and B. S. Gaut. 2001. Molecular evolution of the wound-induced serine protease inhibitor wip1 in Zea and related genera. Mol. Biol. Evol. 18:2092–2101.
Tiffin, P., R. Hacker, B. S. Gaut. 2004. Population genetic evidence for rapid changes in intraspecific diversity and allelic cycling of a specialist defense gene in Zea. Genetics 168:425–434.
Yang, Z. 2001. Adaptive molecular evolution. Pp. 327–350 in D. J. Balding, M. Bishop, and C. Cannings eds. Handbook of statistical genetics. Wiley, Chichester, United Kingdom.(P?r K. Ingvarsson)
E-mail: pelle@eg.umu.se.
Abstract
Plants defend themselves against the attack of natural enemies by using an array of both constitutively expressed and induced defenses. Long-lived woody perennials are overrepresented among plant species that show strong induced defense responses, whereas annual plants and crop species are underrepresented. However, most studies of plant defense genes have been performed on annual or short-lived perennial weeds or crop species. Here I use molecular population genetic methods to survey six wound-inducible protease inhibitors (PIs) in a long-lived woody, perennial plant species, the European aspen (Populus tremula), to evaluate the likelihood of either recurrent selective sweeps or balancing selection maintaining amino acid polymorphisms in these genes. The results show that none of the six PI genes have reduced diversities at synonymous sites, as would be expected in the presence of recurrent selective sweeps. However, several genes show some evidence of nonneutral evolution such as enhanced linkage disequilibrium and a large number of high-frequency–derived mutations. A group of at least four Kunitz trypsin inhibitor genes appear to have experienced elevated levels of nonsynonymous substitutions, indicating allelic turnover on an evolutionary timescale. One gene, TI1, has enhanced levels of intraspecific polymorphism at nonsynonymous sites and also has an unusual haplotype structure characterized by two divergent haplotypes occurring at roughly equal frequencies in the sample. One haplotype has very low levels of intraallelic nucleotide diversity, whereas the other haplotype has levels of diversity comparable to other genes in P. tremula. Patterns of sequence diversity at TI1 do not fit a simple model of either balancing selection or recurrent selective sweeps. This suggests that selection at TI1 is more complex, possibly involving allelic cycling.
Key Words: arms race ? balancing selection ? herbivory ? Populus tremula ? protease inhibitor ? selective sweep ? trench warfare
Introduction
Plants have evolved numerous adaptations to defend themselves against attack by herbivores, including mechanisms to deter, starve, or poison potential herbivores (Marquis 1992). Herbivores are expected, however, to respond to plant defenses by evolving counteradaptations that render defenses less effective or even useless. Such counterselection can ultimately lead to an "arms race" between plants and herbivores, and theoretical models predict that such coevolutionary interactions will lead to an escalation of traits in both plants and herbivores (Bergelson, Dwyer, and Emerson 2001). At the molecular level, such an arms race is expected to lead to the sequential sweep of new defense and counterdefense alleles in the plant and herbivore populations, respectively (Kniskern and Rausher 2001; DeMeux and Mitchell-Olds 2003). A coevolutionary interaction between plants and herbivores characterized by an arms race model therefore predicts that defense genes should exhibit low levels of standing nucleotide sequence variation and high levels of amino acid differentiation (compared to closely related species), induced by strong directional selection for rapid turnover of alleles (Bergelson et al. 2001; DeMeux and Mitchell-Olds 2003).
Alternatively, natural selection may lead to increased diversity in defense genes, either because frequency-dependent selection favors rare alleles or because variability itself is favorable if different alleles confer resistance to different herbivore genotypes (DeMeux and Mitchell-Olds 2003). Such a model for plant-herbivore coevolution, sometimes termed "the trench warfare model" (Stahl et al. 1999) or the "recycling polymorphism" model (Holub 2001), predicts qualitatively different evolutionary dynamics compared to the arms race model. Because different alleles are maintained over long periods in a polymorphic state, balancing selection is expected to result in enhanced levels of sequence diversity because different allelic lineages accumulate mutations more or less independently. Therefore, a trench warfare model of coevolution predicts that defense genes should show enhanced levels of segregating amino acid polymorphisms within populations that sometimes even exceed levels of silent variation (DeMeux and Mitchell-Olds 2003). Such natural selection, favoring allelic variability, has been implicated in the maintenance of hypervariability in both human major histocompatibility complex genes (Ohta 1998) and pathogen recognition genes in many plants (Bergelson et al. 2001).
A growing number of studies are focusing on the molecular population genetic aspects of plant-herbivore and plant-pathogen interactions, and these studies have shown support for both the arms race and trench warfare mode of coevolution (reviewed in DeMeux and Mitchell-Olds 2003). One conclusion emerging from these studies is that specialist defenses, that is, defenses against a small number of natural enemies, are more likely to show deviations from neutral expectations (Tiffin, Hacker, and Gaut 2004). In particular, specialist defenses appear to be more likely to be under the influence of balancing selection (Tiffin, Hacker, and Gaut 2004). In contrast, many generalist defenses, that is, defenses effective against a broad spectrum of enemies, show patterns of intraspecific sequence diversity that either are consistent with a neutral model (Kawabe and Miyashitya 1999; Tiffin and Gaut 2001; Clauss and Mitchell-Olds 2004; Tiffin, Hacker, and Gaut 2004) or show signs of recent positive selection (i.e., selective sweeps, Clauss and Mitchell-Olds 2004; Tiffin 2004).
Plants defend themselves against the attack of natural enemies by using an array of constitutively expressed defenses, such as thorns, spines, and many toxic secondary metabolites (Karban and Baldwin 1997). Plants also actively defend themselves using inducible defense mechanisms (Karban and Baldwin 1997). The active part of a plant's defense arsenal involves, for instance, oxidative enzymes, genes involved in strengthening the cell wall, and/or a wide array of protease inhibitors (PIs) (Constabel 1999). PIs, in particular, are a class of wound-inducible defense genes for which function in herbivore defense is well established and where the physiological mechanisms underlying the defense function are well studied (Karban and Baldwin 1997; Koiwa, Bressan, and Hasegawa 1997; Constabel 1999; Haq, Atif, and Khan 2004). Plant PIs are proteins that function as specific substrates for proteolytic enzymes in the digestive tracts of herbivores (Ryan 1990; Constabel 1999; Haq, Atif, and Khan 2004). However, a PI is not cleaved by a protease, like normal substrates are, but rather forms a stable complex that limits or completely inhibits the proteolytic activity of the protease (Ryan 1990; Constabel 1999; Haq, Atif, and Khan 2004). This results in much reduced proteolysis in the digestive tract, a lack of available amino acids, and ultimately lowered growth rates or starvation of the herbivore (Ryan 1990; Constabel 1999). In addition to their antinutritive effects, PIs may also have toxic or directly lethal effects on herbivores (Duffey and Stout 1996). There are several classes of PIs in plants, each effective against a different class of proteases. The two most common classes of PIs are inhibitors of serine and cysteine proteases, and this is thought to reflect the predominance of these proteases in herbivore digestive systems (Koiwa, Bressan, and Hasegawa 1997; Constabel 1999; Haq, Atif, and Khan 2004). Serine proteases are the predominant digestive enzyme in Lepidoptera, whereas both serine and cysteine PIs are commonly found in Coleoptera (Koiwa, Bressan, and Hasegawa 1997; Haq, Atif, and Khan 2004).
Recent studies of transcriptional patterns following wounding by herbivores have shown that a large number of defense-related genes are upregulated following tissue damage in Populus (Christopher et al. 2004). In particular, a wide range of PIs are among the most upregulated genes following damage by herbivores (Haruta et al. 2001; Christopher et al. 2004). Haruta et al. (2001) identified three wound-induced genes from Populus tremuloides, belonging to the Kunitz trypsin inhibitor class of serine PIs. These genes, TI1, TI2, and TI3, belong to a small gene family and show amino acid sequence similarities ranging from 52% to 83% (Haruta et al. 2001). In particular, TI1 and TI2 appear to be the result of a recent gene duplication as these genes are over 90% identical at the nucelotide level. Christopher et al. (2004) identified two additional Kunitz type PIs from a Populus trichocarpa x Populus deltoides hybrid. These two genes, named TI4 and TI5, represent more diverged members of the Kunitz TI class, sharing less than 30% amino acid similarity with TI1, TI2, and TI3. The transcription of all five TI genes is highly upregulated following mechanical wounding or herbivory, and all five genes can also be induced by external application of the defense signal jasmonate (Haruta et al. 2001; Christopher et al. 2004). However, the patterns of induction of the five genes are quite different and distinct, both locally and systemically throughout the plant (Haruta et al. 2001; Christopher et al. 2004). In addition to the TI genes, several other PIs show strong induction following wounding, including highly upregulated cysteine PI, CI1 (Christopher et al. 2004).
In this paper, I use molecular population genetic analyses to study the evolution of wound-inducible PIs in the European aspen (Populus tremula). I do so by analyzing in more detail previously published surveys of natural variation in TI3 and CI1 (Ingvarsson 2005) and by adding data on patterns of nucleotide diversity in four additional PI genes (TI1, TI2, TI4, and TI5) from the same set of individuals. These six genes are surveyed for departures from neutral expectations, with a special emphasis placed on evaluating the likelihood of either recurrent selective sweeps or balancing selection maintaining amino acid polymorphisms.
Materials and Methods
Sequence Collection
The five serine PIs studied (TI1 to TI5) are all members of the Kunitz TI class, although TI1, TI2, and TI3 appear to form a small gene family of more closely related sequences because they share between 65% and 85% amino acid identity (Haruta et al. 2001). TI4 and TI5 represent more diverged members of the Kunitz TI class, sharing 16%–25% amino acid identity with the TI1–TI3 group of sequences (Christopher et al. 2004). Sequences of TI3 and CI1 were obtained from Ingvarsson (2005). New sequences of TI1, TI2, TI4, and TI5 were collected from the same set of P. tremula trees that were included in Ingvarsson (2005). Leaves were collected from four different sites throughout Europe: Besacon in eastern France, Klagenfurt in southern Austria, F?rjestaden in southeastern Sweden, and Ume? in northern Sweden. TI1 and TI2 were also sequenced from an additional population located close to Halmstad, in southwestern Sweden, about 300 km west of the F?rjestaden population.
Primers to amplify TI1, TI2, TI4, and TI5 from P. tremula were designed from the publicly available GenBank records for these genes (accession numbers AF349441, AF349442, AY378089, and AY378090, respectively). The entire coding sequence was obtained for all genes, including some 5'- and 3'-flanking sequences for TI1, TI2, and CI1. All genes are intronless, except CI which contains two exons, separated by a short intron. Finally, homologous sequences from P. trichocarpa were obtained for all genes by Blast searches of the assembled genome sequence of P. trichocarpa available at http://genome.jgi-psf.org/.
Polymerase chain reaction products were cloned into the pCR2.1 vector using a TA-cloning kit from Invitrogen (Carlsbad, Calif.). At least five, and often more, different clones of each fragment were sequenced using BigDye chemistry (Applied Biosystems Inc., Foster City, Calif.) on an ABI377 automated sequencer at the Ume? Plant Science Centre sequencing facility. Sequences were verified manually and contigs were assembled using the computer program Sequencer v 4.0. Multiple sequence alignments were made using ClustalW (Thompson, Higgins, and Gibson 1994) and adjusted manually using BioEdit (http://www.mbio.ncsu.edu/BioEdit/bioedit.html). All sequences described in this paper have been deposited in the EMBL database (accession numbers AJ842907–AJ842952, AJ843666–AJ843713, and AJ936970–AJ937139).
Estimates of nucleotide polymorphism and statistical tests of neutrality were obtained using the computer program DnaSP v4.00.5 (http://www.ub.es/dnasp/) or Jody Hey's SITES program (http://lifesci.rutgers.edu/heylab/HeylabSoftware.htm#SITES). Significance of statistical tests of neutrality were evaluated by determining the null distribution for each test statistic under neutrality by simulating 105 neutral genealogies using Richard Hudson's ms program (available at http://home.uchicago.edu/rhudson1/source/mksamples.html). Briefly, simulations were conditioned on the original sample configuration, sequence lengths, the number of segregating sites observed, and either the empirically determined recombination rate, Chw = 4Nr, using the method of Hey and Wakeley (1997) or Cmin, the minimum recombination rate compatible with the number of inferred recombination events (Rozas et al. 2001). Cmin can be viewed as a lower bound of the recombination rate and will therefore provide an underestimate of the true recombination rate. Because population structure may influence both power and rejection rates of statistical tests of neutrality (Przeworski 2002), simulations were run with samples taken from a subdivided population. The migration parameter, M = 4Nm, was set so that the expected value of FST = 1 – within/total (Charlesworth 1998) matched that observed at each locus. The population structure in the simulations was assumed to follow the Wright's island model with a total number of populations equal to 20.
Data on nucleotide polymorphism and divergence in the six PI genes was compared to several other genes from P. tremula that are not known to be involved in herbivore defense, Adh1, Gapdh, and GA20ox1 (Ingvarsson 2005) and phyB2, ABI1B, ABI1D, and Adh2 (Garcia and Ingvarsson, unpublished data). These genes were sequenced from the same set of individuals that were included in this study.
Results
Structure of Sequenced Regions
Between 33 and 47 alleles were sequenced from the six PI genes. The length of the sequenced regions ranged from 630 bp for TI3 to 802 bp for CI1. The coding regions of the five TI genes range in size from 606 to 684 bp (table 1). The predicted protein sequence of all five TI proteins contains a signal peptide (Nielsen et al. 1997), consisting of 25 amino acids, that is cleaved off to form the active PI. The mature TI protein consists of 10–12 antiparallel ? strands that form the ?-trefoil fold characteristic of Kunitz type PIs (Song and Suh 1998) and carries a reactive region situated on an external loop. The CI1 gene contains a coding region that is 429 bp long and encodes a protein showing strong similarity to other phytocystatins. After cleavage of a 27–amino acid signal peptide (Nielsen et al. 1997), the mature protein forms two hairpin loops and a largely unstructured N-terminal protein end (Brown and Dziegielewska 1997). The first hairpin loop carries the conserved QxSxG motif, thought to be the site active in protein inhibition, and the second hairpin loop carries the conserved PW sequence motif present in all cystatins (Brown and Dziegielewska 1997).
Table 1 Intraspecific Polymorphism in Populus tremula Wound-induced Protease Inhibitor Genes
Intraspecific Patterns of Nucleotide Diversity
There is abundant variation in all genes, with the number of segregating sites ranging from 31 to 62 (table 1, Appendix). Estimates of Watterson's = 4Nμ range from 0.0112 to 0.0191, which is within the range of earlier estimates of obtained from other P. tremula genes (Ingvarsson 2005 and unpublished data). In addition to base pair substitutions, several indels were polymorphic in the sample. Two different indels, each resulting in the deletion of a compete codon, were segregating in TI1. TI2 contained four indels that were polymorphic in the sample, three indels resulting in the deletion of a complete codon each and a larger indel causing the deletion of four contiguous codons. TI4 contained two polymorphic indels, one resulting in the deletion of a complete codon and one causing the deletion of two contiguous codons. TI5 was also polymorphic for a two-codon deletion. Note that none of these indels disrupted the open reading frame, and all alleles at the five TI genes were predicted to be fully functional based on their amino acid sequence. CI1 only contained indels in the intron, and these were all associated with mononucleotide repeats that differed in length.
If the PI genes were under recurrent directional selection due to continual fixation of new beneficial alleles, silent site diversity is expected to be sharply reduced. There does not, however, appear to be any evidence for a reduction in nucleotide diversity at synonymous sites in the PI genes studied (table 1), compared with other gene regions that are not involved in herbivore defense in P. tremula (Ingvarsson 2005, unpublished data). In fact, synonymous site diversity is higher in the six PI genes, sil = 0.0207, compared to nondefense genes, where sil = 0.0134 (table 1). The same is true for nucleotide diversity at replacement sites, where the six PI genes have higher diversities (rep = 0.0117) than nondefense genes (rep = 0.0076). However HKA (Hudson, Kreitman, and Aguade 1987) tests indicate that neither synonymous nor replacement site diversities differ significantly between the six PI genes and nondefense genes (HKA tests, 2 = 0.213, P < 0.645, and 2 = 0.896, P < 0.561 for synonymous and nonsynonymous sites, respectively).
An interesting observation is that TI1 shows such enhanced levels of diversity at replacement sites that rep/syn exceeds unity (table 1). Such high levels of diversity at replacement sites are hard to explain without invoking natural selection. An inspection of table 1 shows that the high rep/syn ratio observed for TI1 is not attributable to unusually low synonymous diversities in this gene but rather to an elevated level of nonsynonymous polymorphism. In fact, levels of polymorphism at replacement sites in TI1 are the highest recorded among the 13 genes for which sequence data is available in P. tremula (table 1, Ingvarsson 2005, unpublished data, range for nondefense genes 2.96 x 10–3–3.44 x 10–3).
There is evidence for an excess of linkage disequilibrium in all five Kunitz trypsin inhibitor genes if the significance of ZnS (Kelly 1997) is evaluated using levels of recombination estimated from the data set (Chw), although if the lower bound on the recombination rate (Cmin) is used, none of the genes have ZnS that are significant (table 2). However, this approach provides an extremely conservative test of excessive linkage disequilibrium because Cmin estimates of the scaled recombination rate are about an order of magnitude lower than those obtained through the method of Hey and Wakeley (1997) (table 2). Enhanced levels of linkage disequilibrium can be generated by directional selection if a favorable mutation has not yet reached fixation or if recombination occurred during a selective sweep, so that variation at linked sites was not completely eradicated during the sweep (Fay and Wu 2000; Przeworski 2002; Kim and Nielsen 2004). An excess of linkage disequilibrium is also expected if natural selection maintains distinct allelic classes at the locus, as is the case with balanced polymorphisms (Charlesworth, Nordborg, and Charlesworth 1997). Finally, demographic scenarios, such as population structure or population growth, can also result in increased linkage disequilibrium compared to neutral expectations (Przeworski 2002). However, effects of demography are expected to be genome wide, and there is no indication of excessive linkage disequilibrium in other P. tremula genes (Ingvarsson 2005), suggesting that the presence of enhanced linkage disequilibrium in TI defense genes is the product of natural selection.
Table 2 Statistical Tests of Neutrality
Three of the defense genes, TI1, TI2, and TI4, also show an excess of high-frequency–derived variants, as measured by Fay and Wu's H, although this excess is only marginally significant in TI4 (P < 0.067, table 2). Under the much more conservative assumption of the low recombination (Cmin in table 2), H remains significant only for TI1 (table 2). An excess of high-frequency–derived variants is also a pattern that is likely to be explained by strong directional selection fixing newly arisen mutations because derived mutations occurring at high frequencies are rare under a neutral model (Fay and Wu 2000; Kim and Nielsen 2004). Again, demographic factors, and in particular population structure, may provide an alternative explanation for an excess of high-frequency–derived sites (Przeworski 2002). However, population subdivision is not likely to be a factor in explaining the negative values of H for TI1, TI2, and TI4 for at least two reasons. First of all, the effect of population structure on H becomes more severe with increasing strength of subdivision (Przeworski 2002), and population subdivision is generally low in P. tremula (table 1 and Ingvarsson 2005). Second, population structure was explicitly taken into account when assessing the significance of H through coalescent simulations. It is possible that the coalescent simulations did not accurately characterize the population structure of P. tremula. However, given the long-distance dispersal of both seeds and pollen seen in P. tremula, the island model used in the coalescent simulations should provide a reasonable approximation for the population structure of P. tremula, and including a model of population structure in the coalescent simulations, even if approximate, is likely better than completely ignoring it.
Polymorphism and Divergence
Divergence at silent sites in the six defense genes is similar or slightly higher than at other P. tremula loci not involved in herbivore defense (fig. 1), whereas nonsynonymous divergence is generally higher at the defense genes than in nondefense genes. In two of the defense genes, TI1 and TI2, there are signs of long-term adaptive evolution, as they show greater rates of divergence at nonsynonymous than synonymous sites, as evidenced by Ka/Ks >1 (fig. 1). Using the modified pairwise Nei-Gojobori method with p distances (Nei and Kumar 2000), as implemented in MEGA 2.1 (http://www.megasoftware.net), the Ka/Ks ratios at both TI1 and TI2 are significantly greater than one (P < 0.037 and P < 0.006, respectively). A Ka/Ks ratio exceeding unity is a clear sign of adaptive evolution, suggesting that nonsynonymous mutations offer a fitness advantage and are fixed at a higher rate than synonymous mutations (Yang 2001).
FIG. 1.— Synonymous and nonsynonymous divergence for six wound-induced protease inhibitor genes (black) and six genes not involved in herbivore defense (light gray).
Surprisingly, the McDonald and Kreitman (1991) (MK) test, which compare rates of divergence with levels of standing polymorphism within species, show no evidence for adaptive evolution in any of the TI genes. However, if polymorphism and divergence data from the five TI genes are pooled, these genes appear to have experienced elevated rates of fixation of nonsynonymous mutations (the neutrality index of Rand and Kann , NI = 0.584, P < 0.016, table 3). Excluding data from TI4, which harbors a nonsignificant excess of intraspecific polymorphism at nonsynonymous sites (table 2), this pattern of an excess of fixed nonsynonymous mutations becomes even stronger (NI = 0.476, P < 0.0019). If the same pooled comparison is made with seven protein-coding genes that are not involved in herbivore defense (Ingvarsson 2005, unpublished data), no deviation from neutral expectations is detected (NI = 1.433, P < 0.155, table 2).
Table 3 Number of Fixed and Polymorphic Mutations at Synonymous and Replacement Sites in the Coding Regions of Populus tremula Pl Genes
Contrary to the observations for the five TI genes, an MK test indicates a marginally significant (P < 0.075) excess of intraspecific polymorphism at replacement sites in CI1 (table 3). Much of this excess can be explained by the large number of singleton mutations causing amino acid replacements in CI1. There are 13 nonsynonymous mutations found only in single alleles and an additional 7 are found in two to four alleles (Appendix). This is also evident from the Tajima's D, which is significantly negative for CI1 (table 2).
Haplotype Structure at TI1
TI1 has several features suggesting that it has been influenced by natural selection, including an excess of derived mutations occurring at high frequencies, extensive linkage disequilibrium, and a/s and Ka/Ks ratios exceeding one. The gene genealogy at TI1 shows strong haplotype structure, with alleles falling into two distinct haplotypes (haplotypes A and B in fig. 2, also see Appendix) that differ by 14 fixed substitutions, 13 of which occur at replacement sites. The strong haplotype structure also coincides with both an excess of linkage disequilibrium (LD) and high proportions of high-frequency–derived sites (table 2).
FIG. 2.— Gene genealogies for 47 alleles of TI1 in Populus tremula. A sequence from Populus trichocarpa was used as the out-group. TI1-A and TI1-B refer to the two major haplotypes identified. The genealogy was constructed using the neighbor-joining method with p distances in MEGA 2.1.
One of the haplotypes (A in fig. 2) shows very little intrahaplotype variation (S = 5, = 0.00095, Appendix), despite accounting for 25 out of 47 sampled alleles. The remaining sequences, in haplotype B, are more heterogeneous and have a level of polymorphism ( = 0.0117) that is comparable to other loci in P. tremula and about an order of magnitude greater than haplotype A. To test the likelihood of observing two clades with such different levels of intraallelic variation, I used a modified version of the test devised by Hudson et al. (1994). This test evaluates the likelihood, under an equilibrium-neutral model, of observing a haplotype with only five segregating sites making up 25 out of 47 sampled alleles, conditional on the variation observed in the total sample (S = 46). I modified the original approach of Hudson et al. (1994) to explicitly take population subdivision into account in the coalescent simulations used to generate the expected distribution of intraallelic variation in haplotype A. This test shows that under an equilibrium-neutral model the present haplotype configuration is extremely unlikely (P < 0.0001), suggesting that natural selection is responsible for the observed patterns. More specifically, the low sequence diversity seen in haplotype A suggests that this haplotype may have increased in frequency recently. However, there are 14 fixed differences between haplotype A and haplotype B. Comparisons with the out-group species P. trichocarpa reveals that out of these 14 mutations, haplotype A carries the ancestral base at five sites and the derived base at three sites. The ancestral relationship cannot be determined for six of the sites either because the presence of a third base in P. trichocarpa or because the site is deleted in P. trichocarpa. Because the number of derived sites is roughly evenly dispersed between haplotypes A and B, it is unlikely that haplotype A represents a novel haplotype that evolved recently and increased in frequency because of positive selection. A more likely scenario is that the haplotypes A and B have been maintained for long periods of time, through some form of balancing selection. However, patterns of sequence diversity within the two allelic haplotypes are not consistent with a model of balancing selection where the relative frequency of the two haplotypes has been stable over time. Innan and Tajima (1999) showed that if balancing selection maintains two distinct allelic classes at a locus at constant frequencies, the sum of pairwise differences within the two classes is roughly constant and equal to , regardless of the strength and pattern of selection. For TI1 the sum of pairwise diversities within haplotypes A and B equals = 0.01263, which is significantly less (P < 0.05) than the pairwise diversity observed for the total sample (P = 0.0191). A model of constant balancing selection thus appears to provide a poor fit to the TI1 data as well.
Discussion
Sequence Diversity and Evidence for Nonneutral Evolution
In this study I have examined patterns of nucleotide variation in six PIs, five serine PI and one cysteine PI, from European aspen (P. tremula). Although there are differences among the six genes, several signs point to the fact that the evolution of these defense genes is not well characterized by a neutral model. There is no apparent reduction in levels of polymorphism, either at silent or replacement sites in any of the six genes, compared to P. tremula genes that are not involved in herbivore defense (table 1). In particular, the TI1 and TI2 genes stand out by having Ka/Ks ratios that exceed one (fig. 1). A Ka/Ks ratio exceeding unity suggest that nonsynonymous mutations are fixed at a higher rate than synonymous mutations and is a clear sign of adaptive evolution (Yang 2001). Both TI1 and TI2 also have an excess of high-frequency–derived sites (table 2) and excess LD, suggesting that the recent evolutionary histories at these two genes are characterized by one or more selective sweeps that have driven newly arisen mutation to high frequency in the population.
Of the other four genes surveyed, TI3, TI4, and TI5 all show an excess of LD, at least when the significance of the ZnS test is evaluated using the recombination rates inferred from the sequence data (table 2). Such excess LD could result from either directional or balancing selection, although demographic factors such as population structure could also explain these observations. There is no evidence for excess LD in P. tremula genes not involved in herbivore defense (Ingvarsson 2005), The observed patterns could thus be a product of natural selection, although patterns of intraspecific sequence diversities at TI3, TI4, and TI5 show little other evidence of having been influenced by natural selection.
There are, however, indications that the TI genes collectively are evolving in a nonneutral manner because they have an elevated rate of fixation of nonsynonymous mutations (table 3). An evolutionary history of increased amino acid substitution but little current evidence for natural selection could be indicative of episodic selection, where favorable mutations appear and rapidly sweep through the population but are then followed by long periods of selective neutrality (Tiffin and Gaut 2001).
CI1 has an excess of amino acid polymorphisms, much of which can be explained by a large number of singletons. While such an excess of low-frequency amino acid polymorphisms could be an indication of diversifying selection, an excess is also expected under a slightly deleterious model of molecular evolution, where selection is weak enough to allow polymorphism at nonsynonymous sites but where fixation is limited (Ohta 1992).
The different PI genes thus appear to have been influenced by natural selection to different degrees. TI1 and TI2 show several signs of both historical and current action of natural selection, whereas the remaining TI genes and CI1 show sporadic signs of deviations from neutral expectations. In Arabidopsis thaliana, Clauss and Mitchell-Olds (2004) showed that patterns of sequence diversity varied 10-fold across six duplicated trypsin inhibitor genes, located in tandem across a 10-kb segment on chromosome II. Interestingly, they also showed that sequence diversity in the coding regions was positively correlated to variation in expression levels of the different TI genes. They interpreted this among-locus variation in both sequence polymorphism and gene expression to be the result of selection for coping with a diverse set of both specialist and generalist herbivores through subfunctionalization of different loci (Clauss and Mitchell-Olds 2004).
Molecular population genetic data from two PIs in maize (Zea mays), wip1 and mpi, did not show any deviations from neutrality (Tiffin and Gaut 2001; Tiffin, Hacker, and Gaut 2004). Relative rate tests, however, suggested significant rate heterogeneity at wip1 with an increased rate evolution in the Zea lineage; these changes were preferentially located in one of the inhibitory loops of wip1 (Tiffin and Gaut 2001). This suggests, even when molecular population genetic analysis in contemporary populations fails to provide evidence for nonneutral evolution, that defense genes can be characterized by an evolutionary history influenced by natural selection.
Tiffin, Hacker, and Gaut (2004) argue that specialist defenses, that is, defenses active against one or a few natural enemies, are more likely to show evidence for nonneutral evolution compared to generalist defenses. Although PIs as a group are expected to target many different natural enemies, it is not clear whether individual PIs should be viewed as specialist or generalist defenses. The large number of TI genes found in P. tremula, each with their own distinct patterns of gene expression following wounding (Christopher et al. 2004), suggests that these genes might have different targets. It is possible that some of the TI genes have more generalized targets whereas others have more specialized functions, but at present there is no data to shed any light on these ideas. The same view was advocated by Clauss and Mitchell-Olds (2004), who argued that the differences in both sequence diversity and expression pattern seen among the six TI genes in their study are most likely the result of selection for subfunctionalization of the different genes. More studies are clearly needed, both of the patterns of expression following attack by different herbivores and also of the inhibitory functions at the biochemical level of the various PI genes against different herbivore digestive enzymes.
Haplotype Structure at TI1
The TI1 gene has both elevated levels of intraspecific amino acid polymorphism and a Ka/Ks ratio that exceeds one (table 1 and fig. 1). Moreover, the gene genealogy at TI1 shows a haplotype structure with two divergent haplotypes that have different levels of intrahaplotypic variation. The presence of two diverged haplotypes, with one haplotype characterized by a very low level of standing variation and another that carries more normal levels of sequence variation, has been interpreted as a hallmark signature of selective sweeps in a highly recombining genomic region (Parsch, Meiklejohn, and Hartl 2001; Quesada et al. 2003; Kim and Nielsen 2004). However, the two haplotypes at TI1 differ by 14 fixed mutations, 13 of which are nonsynonymous, although there is no difference between the two haplotypes in the number of derived mutations that they carry. The presence of two such diverged haplotypes is therefore probably a result of them having been maintained in the population over evolutionary time, even though the patterns of polymorphism observed in haplotypes A and B do not fit a model of long-term, stable balancing selection (Innan and Tajima 1999). A possible alternative explanation is a recent admixture of two previously isolated populations. This seems unlikely, however, because an admixture should leave a genome-wide signature. First, the population subdivision at TI1 is low (table 1), and both haplotypes are present in all five populations sampled across Europe, reducing the likelihood of recent admixture. Second, although the excess of LD (table 2) could be interpreted as signs of enhanced haplotype structure also at other TI genes, none of the nondefense genes show such patterns (Ingvarsson 2005). Rather, the enhanced LD could be the result of selection maintaining similar although weaker forms of the haplotype structure seen at TI1 also in other PI genes (Appendix).
Patterns of polymorphism and haplotype structure therefore suggest that neither the arms race nor trench warfare model alone is adequate for describing the evolutionary history of TI1. Rather, the data are suggestive of a model that incorporates both directional selection (arms race) and a selective maintenance of intraspecific amino acid polymorphism (trench warfare). One possible explanation that fits both of these criteria is that different resistance haplotypes are maintained through balancing or negative-frequency–dependent selection, but where frequencies of different resistance haplotypes vary over time (Seger 1992).
Tiffin, Hacker, and Gaut (2004) also showed that neither the arms race nor the trench warfare model was adequate for explaining patterns of polymorphism at the disease-resistance gene hm2 in Z. mays. Tiffin, Hacker, and Gaut (2004) found several diverged haplotypes with very little intrahaplotype variation and proposed a scenario where these haplotypes were maintained by natural selection, but the frequencies of these haplotypes varied over evolutionary time. In A. thaliana, Rose et al. (2004) also found two major allelic haplotypes at the disease-resistance gene RPP13. They interpreted their results as the consequence of negative-frequency–dependent selection, where single RPP13 alleles have not swept to fixation but have rather been maintained in the population because different alleles encode resistance specificities against different strains of the pathogen (Rose et al. 2004). The RPP13 data is clearly inconsistent with the arms race model and more in line with the predictions made by the trench warfare model. Finally, data from antibacterial peptides in Drosophila melanogaster also do not show clear support for either the arms race or the trench warfare model (Lazzaro and Clark 2003), although these genes show some signs of natural selection, such as relatively high levels of intraspecific amino acid polymorphisms and excesses of derived sites occurring at high frequencies (Lazzaro and Clark 2003).
The scenario proposed above to explain the observed patterns of nucleotide diversity at TI1 is clearly ad hoc. However, current models that have been proposed to explain coevolutionary interactions of plant-herbivore or plant-pathogen interactions at the molecular level, including the trench warfare and arms race models, are rather heuristic, and it is not clear what type of data are needed to either verify or reject these models (see also Tiffin, Hacker, and Gaut for a similar argument). The fact that almost all studies that have examined coevolutionary interactions between plant and pathogens or herbivores at the level of individual genes have found patterns of polymorphism that cannot be explained by either the arms race or trench warfare models suggests that more research is needed to specifically examine patterns of nucleotide polymorphism under a wide range of conditions both in these rather simple and in more complex models of plant-parasite coevolution (Seger 1992).
Appendix
TI1
TI2
TI3
TI4
TI5
CI1
Alignment of polymorphic sites in the six protease inhibitor genes. Residue numbers are indicated above each polymorphic site, and each polymorphic site is indicated as either nonsynonymous (N), synonymous (S), indels (D) or occurring in flanking regions or introns (I). All residues identical to the top sequence are denoted by dots (.) and gaps are given by a dash (–).
Acknowledgements
I would like to thank Carin Olofsson for obtaining the DNA sequences presented in this paper. John McDonald and three anonymous reviewers provided comments that improved the paper. I am also grateful to Barbara Giles who provided help with linguistic corrections. This research has been funded by a grant from the Swedish Research Council (Vetenskapsr?det, VR).
References
Bergelson, J., G. Dwyer, and J. J. Emerson. 2001. Models and data on plant-enemy coevolution. Annu. Rev. Genet. 35:469–499.
Bergelson, J., M. Kreitman, E. A. Stahl, and D. C. Tian. 2001. Evolutionary dynamics of plant R-genes. Science 292:2281–2285.
Brown, W. M., and K. M. Dziegielewska. 1997. Friends and relations of the cystatin superfamily—new members and their evolution. Protein Sci. 6:5–12.
Charlesworth, B. 1998. Measures of divergence between populations and the effects of forces that reduce variability. Mol. Biol. Evol. 15:538–542.
Charlesworth, B., M. Nordborg, and D. Charlesworth. 1997. The effects of local selection, balanced polymorphism and background selection on equilibrium patterns of genetic diversity in subdivided populations. Genet. Res. 70:155–174.
Christopher, M. E., M. Miranda, I. T. Major, and C. P. Constabel. 2004. Gene expression profiling of systemically wound-induced defenses in hybrid poplar. Planta 219:936–947.
Clauss, M. J., and T. M. Mitchell-Olds. 2004. Functional divergence in tandemly duplicated Arabidopsis thaliana trypsin inhibitor genes. Genetics 166:1419–1436.
Constabel, C. P. 1999. A survey of herbivore-induced defensive proteins and phytochemicals. Pp. 137–166 in A. A. Agrawal, S. Tuzun, and E. Bent, eds. Inducible plant defense against pathogens and herbivores: biochemistry, ecology and agriculture. American Phytopathological Society Press, St. Paul, Minn.
DeMeux, J., and T. M. Mitchell-Olds. 2003. Evolution of plant resistance at the molecular level: ecological context of species interactions. Heredity 94:343–352.
Duffey, S. S., and M. J. Stout. 1996. Antinutritive and toxic components of plant defense against insects. Arch. Insect Biochem. Physiol. 32:3–37.
Fay, J., and C.-I. Wu. 2000. Hitchhiking under positive Darwinian selection. Genetics 155:1405–1413.
Haq, S. K., S. M. Atif, and R. H. Khan. 2004. Protein proteinase inhibitors in combat against insects, pests and pathogens: natural and engineered phytoprotection. Arch. Biochem. Biophys. 431:145–159.
Haruta, M., I. T. Major, M. E. Christopher, J. J. Patton, and C. P. Constable. 2001. A Kunitz trypsin inhibitor gene family from trembling aspen (P. tremuloides Michx.): cloning, functional expression and induction by wounding and herbivory. Plant Mol. Biol. 46:347–359.
Hey, J., and J. Wakeley. 1997. A coalescent estimator of the population recombination rate. Genetics 145:833–846.
Holub, E. B. 2001. The arms race is ancient history in Arabidopsis thaliana, the wildflower. Nat. Rev. Genet. 2:516–527.
Hudson, R. R., K. Bailey, D. Skarecky, J. Kwiatowski, and F. J. Ayala. 1994. Evidence for positive selection in the superoxide dismutase (Sod) region of Drosophila melanogaster. Genetics 136:1329–1340.
Hudson, R. R., M. Kreitman, and M. Aguade. 1987. A test of neutral molecular evolution based on nucleotide data. Genetics 116:153–159.
Ingvarsson, P. K. 2005. Nucleotide polymorphism and linkage disequilibrium within and among natural populations of European aspen (P. tremula L., Salicaceae). Genetics 169:945–953.
Innan, H., and Tajima, F. 1999. The effect of selection on the amounts of nucleotide variation within and between allelic classes. Genet. Res. 73:15–28.
Karban, R., and I. T. Baldwin. 1997. Induced responses to herbivory. University of Chicago Press, Chicago, Ill.
Kawabe, A., and N. T. Miyashitya. 1999. DNA variation in the acidic chitinase locus (ChiB) region of the wild plant Arabidopsis thaliana. Genetics 153:1445–1453.
Kelly, J. K. 1997. A test of neutrality based on interlocus associations. Genetics 146:1197–1206.
Kim, Y., and R. Nielsen. 2004. Linkage disequilibrium as a signature of selective sweeps. Genetics 167:1513–1524.
Kniskern, J., and M. D. Rausher. 2001. Two modes of host-enemy coevolution. Popul. Ecol. 43:3–14.
Koiwa, H., R. A. Bressan, and P. M. Hasegawa. 1997. Regulation of protease inhibitors and plant defense. Trends Plant Sci. 2:379–384.
Lazzaro, B. P., and A. G. Clark. 2003. Molecular population genetics of inducible antibacterial peptide genes in Drosophila melanogaster. Mol. Biol. Evol. 20:914–923.
Marquis, R. J. 1992. Selective impact of herbivores. Pp. 203–325 in R. S. Fritz and E.-L. Simms, eds. Plant resistance to herbivores and pathogens—ecology, evolution and genetics. University of Chicago Press, Chicago, Ill.
McDonald, J. H., and M. Kreitman. 1991. Adaptive protein evolution at the Adh locus in Drosophila. Nature 351:652–654.
Nei, M., and S. Kumar. 2000. Molecular evolution and phylogenetics. Oxford University Press, Oxford.
Nielsen, J., J. Engelbrecht, S. Brunak, and G. von Heijne. 1997. Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites. Protein Eng. 10:1–6.
Ohta, T. 1992. The nearly neutral theory of molecular evolution. Annu. Rev. Ecol. Syst. 23:263–286.
———. 1998. On the pattern of polymorphisms at major histocompatibility complex loci. J. Mol. Evol. 46:633–638.
Parsch, J., C. D. Meiklejohn, and D. L. Hartl. 2001. Patterns of DNA sequence variation suggest the recent action of positive selection in the janus-ocnus region of Drosophila simulans. Genetics 159:647–657.
Przeworski, M. 2002. The signature of selection at randomly chosen loci. Genetics 160:1179–1189.
Quesada, H., U. E. M. Ramírez, J. Rozas, and M. Aguade. 2003. Large-scale adaptive hitchhiking upon high recombination in Drosophila simulans. Genetics 165:895–900.
Rand, D. M., and L. M. Kann. 1996. Excess amino acid polymorphism in mitochondrial DNA: contrasts among genes from Drosophila, mice, and humans Mol. Biol. Evol. 13:735–748.
Rose, L. E., P. D. Bittner-Eddy, C. H. Langley, E. B. Holub, R. W. Mitchelmore, and J. L. Beynon. 2004. The maintenance of extreme amino acid diversity at the disease resistance gene, RPP13, in Arabidopsis thaliana. Genetics 166:1517–1527.
Rozas, J., M. Gullaud, G. Blandin, and M. Aguade. 2001. DNA variation at the rp49 gene region in Drosophila simulans: evolutionary inferences from an unusual haplotype structure. Genetics 158:1147–1155.
Ryan, C. A. 1990. Protease inhibitors in plants: genes for improving defenses against insects and pathogens. Annu. Rev. Phytopathol. 28:425–449.
Seger, J. 1992. Evolution of exploiter-victim relationships. Pp. 3–25 in M. J. Crawley ed. Natural enemies: the population biology of predators, parasites and diseases. Blackwell Scientific, London.
Song, H. K., and S. W. Suh. 1998. Kunitz-type soybean trypsin inhibitor revisited: refined structure of its complex with porcine trypsin reveals an insight into the interaction between a homologous inhibitor from Erythrina caffra and tissue-type plasminogen activator. J. Mol. Biol. 275:347–363.
Stahl, E. A., G. Dwyer, R. Mauricio, M. Kreitman, and J. Bergelson. 1999. Dynamics of disease resistance polymorphism at the Rpm1 locus of Arabidopsis. Nature 400:667–671.
Thompson, J. D., D. G. Higgins, and T. J. Gibson. 1994. ClustalW: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22:4673–4680.
Tiffin, P. 2004. Comparative evolutionary histories of chitinase genes in the genus Zea and family Poaceae. Genetics 167:1331–1340.
Tiffin, P., and B. S. Gaut. 2001. Molecular evolution of the wound-induced serine protease inhibitor wip1 in Zea and related genera. Mol. Biol. Evol. 18:2092–2101.
Tiffin, P., R. Hacker, B. S. Gaut. 2004. Population genetic evidence for rapid changes in intraspecific diversity and allelic cycling of a specialist defense gene in Zea. Genetics 168:425–434.
Yang, Z. 2001. Adaptive molecular evolution. Pp. 327–350 in D. J. Balding, M. Bishop, and C. Cannings eds. Handbook of statistical genetics. Wiley, Chichester, United Kingdom.(P?r K. Ingvarsson)