Out of the swamp: unidirectional hybridization with weedy species may explain the prevalence of Amaranthus tuberculatus as a weed


Author for correspondence:
P. J. Tranel
Tel: +1 217 333 1531
Email: tranel@illinois.edu


  • Amaranthus tuberculatus represents one of the most dramatic cases of weed invasion documented in the midwestern USA. The species is infamous for evolving resistance to multiple herbicides, and predicting whether these resistances may be transferred to widespread weeds of the Amaranthus hybridus aggregate is a matter of epidemiological concern. Here, we explore the patterns of genetic exchange between Amaranthus tuberculatus and A. hybridus in an effort to understand whether allele introgression occurs throughout the genome and if fecundity penalties are associated with genetic exchange.
  • We evaluated 192 homoploid BC1s at 197 amplified fragment length polymorphism (AFLP) loci, as well as two loci associated with herbicide resistance: ALS and PPO. We also assessed the fecundity of each genotype by evaluation of seed production or pollen development.
  • It was discovered that genetic exchange between the species is unidirectional. Whereas A. hybridus alleles transfer with little or no penalty to A. tuberculatus, the reciprocal exchange is significantly distorted and potentially of limited evolutionary consequence.
  • Our previous hypothesis suggesting unidirectional introgression at ALS owing to circumstantial linkage is now modified to account for the more generalized distortion of genetic exchange observed in this study.


The role of interspecies hybridization in plant evolution has been studied from numerous directions (Anderson & Stebbins, 1954; Barton, 2001). Topics addressed have included the formation of new allopolyploid (Soltis & Soltis, 1999) or homoploid (Rieseberg, 1997) species, the transmission of adaptations (Arnold, 2004) and the onset of invasiveness (Abbott, 1992; Ellstrand & Schierenbeck, 2000). The emphasis of the present study is understanding if and how interspecies hybridization may contribute to the evolution of weediness in plants using weedy Amaranthus species as a model system.

The Amaranthus genus comprises several important weeds grouped into two subgenera. Homonym subgenus Amaranthus includes long-known weeds of cosmopolitan distribution (Mosyakin & Robertson, 2003). The most important among these belong to the Amaranthus hybridus aggregate (Holm et al., 1991), encompassed by monoecious species native of Central and North America, these are predominantly self-pollinated. Subgenus Acnida includes dioecious species restricted to North America (Mosyakin & Robertson, 2003). Sex in these species is controlled by a chromosomal XY-type system (Murray, 1940). Except for Amaranthus palmeri (Sauer, 1957), species of Acnida have gained notoriety as weeds only recently, with Amaranthus tuberculatus/Amaranthus rudis representing one of the most dramatic cases of weed invasion in the midwestern USA.

Amaranthus tuberculatus was previously subdivided into two species, A. tuberculatus and A. rudis (Pratt & Clark, 2001). Amaranthus rudis (initially named Acnida tamariscina) was first described in Oklahoma in 1830 and has since shown continuous northward and eastward accretion into midwestern States, overlapping with A. tuberculatus, of static range, in sandy and muddy stream banks, lakeshores and pond margins, along the Missouri, Mississippi and Ohio river systems (Sauer, 1957, 1972). Although neither A. rudis nor A. tuberculatus were considered significant weeds of agriculture, herbarium samples from crop fields often had intermediates (Sauer, 1957), perhaps suggesting that hybrid forms were favored as weeds. A combination of overlapping ranges and hybridization precludes distinguishing A. tuberculatus and A. rudis and, therefore, a single taxon is presently recognized with A. tuberculatus taking priority in the name (Pratt & Clark, 2001). (Some authors propose A. tuberculatus var. rudis for the more invasive variant; Costea et al., 2005.)

Over the last 20 yr, A. tuberculatus has gone from virtual anonymity to becoming the most significant problem weed in one of the world's premier agricultural regions, the midwestern USA (Steckel, 2007). Its success as a weed is attributed to, among other things, its remarkable ability to evolve resistance to herbicides. Recently, resistance to the herbicide glyphosate, a fundamental herbicide in the production of commodity crops, has been reported in A. tuberculatus (Legleiter & Bradley, 2008). Populations of A. tuberculatus exist in which a single plant can withstand normally lethal doses of at least nine different herbicides, severely restricting chemical weed management (Patzoldt et al., 2005). What is exceptional is that cases of multiple resistances in A. tuberculatus result from the accumulation of independently evolved mutations (i.e. not broad resistances cause by single gene effect), some of which have been reported only in this species (Gressel & Levy, 2006; Patzoldt et al., 2006).

A major epidemiological concern is found in predicting whether rare herbicide resistance alleles found in A. tuberculatus can naturally introgress into widely distributed taxa of the A. hybridus aggregate. Such introgression could exacerbate the occurrence of herbicide-resistant weed populations. Prior research assessing genetic exchange between A. tuberculatus and A. hybridus shows that the two species can hybridize at high frequencies under field conditions (Trucco et al., 2005a,b) and that genetic exchange may occur within a homoploid background (Trucco et al., 2005c). Moreover, the low fecundity of first-generation hybrids may be quickly overcome in the first backcross generation (Trucco et al., 2005c). However, although transfer of a herbicide resistance allele from A. hybridus to A. tuberculatus has been shown experimentally (Tranel et al., 2002), there is no evidence of the reciprocal transfer. In fact, an experimental attempt to transmit an A. tuberculatus herbicide resistance allele at the ALS locus to a monoecious A. hybridus background was unsuccessful. This lack of transfer was hypothesized to be caused by linkage of the ALS locus to some other locus associated with, for example, hybrid sterility or sex identity (Trucco et al., 2005c).

To test this linkage hypothesis, we expanded on our previous study by examining homoploid BC1 individuals for inter-specific introgression at 199 loci. Our goal was to profile the patterns of allele exchange between the two species to: access genetic exchange throughout the genome; determine if fecundity penalties are associated with genetic exchange; and compare the overall pattern of exchange with that of two important herbicide resistance genes, namely the acetolactate synthase (ALS) and the protoporphyrinogen oxidase (PPO) genes. Our findings suggest there is a generalized phenomenon operating after hybrid formation to bring about asymmetric genetic exchange between A. hybridus and A. tuberculatus.

Materials and Methods

Plant material and culture

We used two A. hybridus L. populations and two A. tuberculatus (Moq.) Sauer populations from different regions of the state of Illinois. These populations were known to be polymorphic at least for herbicide resistance loci. An A. hybridus population from Champaign, IL, USA (collected at the University of Illinois’ South Farms), was used maternally in the production of first-generation hybrids (F1s), as this population was susceptible to ALS-inhibiting herbicides. A population of A. tuberculatus from Adams County (IL, USA), resistant to ALS-inhibiting herbicides, was used paternally in the production of F1s. Since ALS-inhibitor resistance is controlled by a single, nuclear-encoded, and dominant gene, hybrid offspring were discriminated from nonhybrid selfings (c. 98% of seed) by surviving a dose of chlorimuron-ethyl lethal to susceptible plants (for more details see Trucco et al., 2005a). Hybridity of these individuals was confirmed further by heterozygosis at ALS, presence of amplified fragment length polymorphism (AFLP) markers from both parents and intermediacy of nuclear DNA content as explained later.

Fourteen female F1s (dioecism is dominant to monoecism) derived from five different hybridization events were backcrossed with A. hybridus pollen (seven of the females) or A. tuberculatus pollen (the remaining seven). For backcrosses, pollen was provided by three individuals from the parental population as well as another three from an additional population of the same species. The additional A. tuberculatus and A. hybridus populations were collected from crop fields in Bond and Edgar counties (IL), respectively. In an additional backcross with A. tuberculatus, pollen was derived from a single male from the Adams County population. The AFLP-genotypes and other molecular and phenotypic characters were evaluated in a total of 192 BC1 individuals, 96 from each backcross direction, with hybrids producing between 7 and 33 BC1 offspring each.

Controlled crosses were performed using pollination tents, and different glasshouses were used for crosses with A. hybridus or A. tuberculatus pollen to minimize contamination with undesired pollen. All BC1 individuals were grown under similar glasshouse conditions and maintained until maturity in similar fashion.

DNA content analysis

Hybridization between A. tuberculatus and A. hybridus occurs mostly within a homoploid background, although triploidy is observed occasionally (Trucco et al., 2005c). To remove triploids from the analysis, the nuclear DNA content of all hybrid progenies (F1 and BC1) was assessed via flow cytometry. Nuclei isolation and staining procedures were conducted as described by Rayburn et al. (1992). A four-color flow cytometer was used to analyse at least 8000 nuclei per plant sample. Less than 2% of all individuals had DNA content values inconsistent with homoploidy, and these individuals were not considered in the following analyses (for DNA content ranges for hybrids and the two types of backcrosses see Trucco et al. (2005c).

AFLP genotyping

AFLP profiles were obtained for 282 individuals, 96 from each of the two backcross directions and 94 from parental populations. Parental populations were represented by 22 F1 hybrids, 23 individuals from the A. hybridus populations and 49 individuals from the A. tuberculatus populations. DNA extractions were conducted following the cetyltrimethylammonium bromide (CTAB) method of Doyle & Doyle (1990) and 500 ng of DNA per sample were used for the AFLP procedure.

The DNA restriction–ligation reactions, as well as preselective and selective amplifications were performed as indicated by Vos et al. (1995), with the exception that primers used were fluorescently labeled to enable reading of AFLP profiles using a capillary-based sequencer. All samples were amplified using eight different primer combinations (MCTA/EACC, MCTA/EACT, MCTA/EACG, MCTA/EACA, MCTC/EACC, MCTC/EACT, MCTC/EACG and MCTC/EACA).

The AFLP chromatograms were converted to binomial (presence/absence) data by setting a fluorescence threshold just above background fluorescence. Parental chromatograms were used to establish the universe of scorable alleles, and only alleles with relative frequencies dissimilar from zero were considered. For this purpose, a frequency limit was established taking the 95% confidence-interval of such frequency to include zero, using a normal approximation to the binomial.

Alleles that showed no segregation (i.e. the marker was always present) in populations of one species, and did not occur in populations of the other, were considered species-specific provided that they were detectable in the hemizygous condition – that is, provided that they were consistently detected in F1s (frequency ≥ 0.95). Also, alleles with frequencies ≥ 0.95 in one species and ≤ 0.05 in the other were considered species-specific if they met the same provision (see the Supporting Information, Table S1).

Species-specific alleles for the recurrent parental species were expected to be homozygous or heterozygous for the AFLP-marker in the backcross individuals. However, if homozygous recessive individuals were observed (i.e. frequency < 0.95 for marker presence), the allele was removed from the species-specific category. Marker elimination in this sense was proportionally greater for A. tuberculatus’ specific alleles than for A. hybridus’ (see Table S2). Cryptic heterozygosis in the dioecious species may explain this discrepancy. Briefly, since AFLP is a dominant marker system, segregation between homozygous dominant and heterozygous individuals cannot be ascertained until the following generation. Population genetics theory predicts this type of segregation to be more frequent in the dioecious species because of obligate outcrossing than in the monoecious species, which predominantly selfs.

Markers selected using the above criteria were applied in the construction of BC1 AFLP genotypes. Genotype distributions in the different backcross populations were tested for normality using the Kolmogorov–Smirnov (KS) test (Stephens, 1974).

Molecular analyses at ALS and PPO

All DNA samples used for AFLP-genotyping were also genotyped at ALS and PPO via PCR-based RFLPs. At both loci, restriction sites exist that are specific to one or the other species, at least among the populations used in this study. A 450 bp fragment of ALS was amplified and digested with EcoRV endonuclease as described by Trucco et al. (2005a). A 1500 bp fragment of PPO was amplified following the procedure of Patzoldt et al. (2006) for PPX2L, an A. tuberculatus PPO isoform. PPO fragments were subjected to digestion with HindIII, which recognizes a restriction site present in A. hybridus but not A. tuberculatus alleles (Fig. S1).

Assessment of fecundity

All plants used in the study were grown under similar glasshouse conditions and harvested at maturity. Seed heads were dried at room temperature for 1 month before threshing. The number of seeds produced by each plant was estimated by dividing total seed weight by the weight of 100 seeds and multiplying by 100. If fewer than 100 seeds were produced, then the actual seed output was taken. When male hybrids occurred, inflorescences from some of these individuals (n = 22) were sampled early in the reproductive phase for cytological evaluation of pollen grains. Anthers were dissected out of flowers previously fixed in a 3 : 1 ethanol–acetic acid solution, placed on a slide, stained with 1% aceto-orcein and squashed. Pollen grains were evaluated at ×10 magnification, and the numbers of micro-sized and normal-sized pollen grains were recorded. All data were normalized using a natural log conversion, and correlation coefficients were calculated using the method of Pearson for parametric data (Moore, 2006). Correlations were obtained between frequency of introgressed alleles and fecundity, and for single locus associations between allele presence and fecundity.

Genetic linkage analysis

The QTX analysis platform described by Chmielewicz & Manly (2002) was used to establish loci associations. In the populations obtained by backcrossing with A. hybridus, a majority (68 out of 96) of individuals showed limited (i.e. only one locus), if any, introgression. These individuals could not be used to establish marker associations and therefore were not considered in the analysis. In the reciprocal exchange all individuals were considered. Associations were established using Kosambi's function (α = 0.0001), using G-statistic-based testing to allow for segregation distortion (Garcia-Dorado & Gallego, 1992).


Identification of species-specific markers

Between 92 and 183 AFLP markers were detected for each of the eight primer combinations used (Table S1). On average, 23% of these markers had frequencies near zero among all of the parental populations and were removed from the analysis. Of the 814 remaining markers, 59% were monomorphic, 5% were always present and only present in A. hybridus, and 1% always present and only present in A. tuberculatus. The remaining markers segregated in both species; of these, 14% were ‘almost’ always present in A. hybridus and absent in A. tuberculatus, and 4% behaved in the inverse manner. Markers that were always or almost always present in one parent and absent in the other were considered ‘species-specific’ provided that they were detectable in the hemizygous condition and showed no segregation in backcross progeny with the species from which they originated (Table S2). See the Materials and Methods section for details on the criteria used in these classifications.

Marked segregation distortion in backcross progeny with A. hybridus

We monitored the segregation of 64 A. hybridus-specific AFLP markers in progeny from backcrosses to A. tuberculatus and observed that 50% of these markers showed Mendelian segregation (Fig. 1a). By contrast, all 13 of the A. tuberculatus-specific markers monitored in the reciprocal backcross showed segregation distortion (Fig. 1b). Compared with A. hybridus, fewer A. tuberculatus-specific markers were identified (Table S2). This is not surprising, as greater homozygosity is expected in the predominantly self-pollinated A. hybridus relative to the dioecious A. tuberculatus (and, therefore, fewer markers did not segregate in A. tuberculatus, which was a criterion for marker selection). Consequently, we expanded our analysis to consider additional A. tuberculatus markers. In this instance, we considered markers that were polymorphic among the exact parents used in the cross but that could not be categorized strictly as species-specific (we named these as more permissive species-specific markers). Of 120 such markers considered, 97% showed strong negative distortion (Fig. 1c).

Figure 1.

 Amplified fragment length polymorphism (AFLP) markers (alleles) grouped by their segregation frequencies. Frequencies for Amaranthus hybridus specific markers (n = 64) in backcrosses with Amaranthus tuberculatus (a), and frequencies for A. tuberculatus strictly specific (n = 13) (b) and more permissive markers (n = 120) (c) in backcrosses with A. hybridus.

Reciprocal backcross progeny show different genotype distributions

Progeny from backcrosses to A. tuberculatus exhibited genotypes varying in their proportions of A. hybridus-specific markers, with a distribution of frequencies approaching normality (Fig. 2a, KS = 0.108, P > 0.1). Progeny from reciprocal backcrosses showed a dramatically skewed distribution of genotypes, with most individuals showing few if any A. tuberculatus-specific markers (Fig. 2b, KS = 0.321, P < 0.0001). The same skewed tendency is observed when more permissive species-specific polymorphisms are used in the analysis (Fig. 2c, KS = 0.338, P < 0.0001). Briefly, whereas offspring from backcrosses with A. tuberculatus show a continuous array of genotypes, two discrete categories were observed in offspring from backcrosses with A. hybridus: a majority of offspring with no evidence of introgression at most of 133 monitored AFLP loci, and a smaller group with evidence of introgression at varying proportions, with up to 85% of the A. tuberculatus alleles present.

Figure 2.

 Amplified fragment length polymorphism (AFLP) genotypes grouped by frequencies of nonrecurrent-species alleles. Genotypes grouped by occurrence of Amaranthus hybridus specific alleles (a), Amaranthus tuberculatus’ strictly (b) or more permissive species-specific (c) alleles among backcrosses with A. tuberculatus (a) or A. hybridus (b,c). In each backcross direction, 96 individuals were considered.

Recombination in backcross progeny with A. tuberculatus

Linkage analysis was performed to assess if positive genetic movement was the outcome of a limited number of allelic combinations, or if a majority of markers were introgressant independent of one another. This analysis was restricted in backcross offspring with A. hybridus, as only a small portion of the progeny were recombinant (see Fig. 2b) and very few markers could be categorized strictly as species-specific. Nonetheless, of the 13 markers considered for the analysis, four showed correlated segregations conforming to two linkage groups (Table S3). A more comprehensive analysis could be conducted with progeny from the reciprocal backcross. In this case, although c. 11% of the markers could not be clustered into linkage groups, the remaining AFLP markers distributed into 14 linkage groups (Table S4). Knowing that the two species have a haploid complement of 16 chromosomes (Grant, 1959), recombination can be considered extensive in the A. tuberculatus direction, and introgression is unlikely precluded to a few chromosomes or chromosomal regions.

Contrasting fecundity penalties with increased introgression

We examined seed production in progeny from both backcrosses and normalized the data by natural log-transformation. We established correlations between the proportion of introgressant alleles in a genotype and its fecundity. In backcrosses with A. tuberculatus, although a general negative tendency is observed, no correlation could be established between the two parameters. That is, no fecundity penalty manifested as increasing proportions of A. hybridus alleles occurred in a genotype (Fig. 3a). Conversely, a significant negative correlation was observed between the proportion of A. tuberculatus alleles present in a genotype and the fecundity of backcrossed offspring with A. hybridus (Fig. 3b). This negative correlation remained unaltered when only recombinant progeny were considered in the analysis (r = −0.83, R2 = 0.69, P < 0.01). Upon backcrossing with A. tuberculatus, male progeny are obtained. We assessed the fertility of some of these individuals by evaluation of pollen grains and, unlike what was observed in females from the same backcross direction, a slightly significant correlation was established between number of introgressant loci and aberrant pollen formation (Fig. 3c). Evaluating each locus independently, c. 5% of evaluated loci showed a significant correlation with fecundity in male and female offspring produced with A. tuberculatus as the recurrent parent (data not shown). These loci were different for male and female individuals.

Figure 3.

 Correlation between introgression and fecundity. Association between occurring proportion of Amaranthus hybridus specific alleles and normalized fecundity (seed output) (a) or per cent micropollen (c) in female or male progenies obtained from backcrosses with Amaranthus tuberculatus, respectively. Correlation between normalized fecundity and occurring proportion of A. tuberculatus specific alleles in backcrosses with A. hybridus is shown in (b).

Asymmetric transfer of herbicide resistance alleles

We monitored allele movement at two loci involved with monogenic herbicide resistance, those for the ALS and PPO genes. Whereas A. hybridus alleles for these loci exhibited Mendelian segregation in backcross progeny with A. tuberculatus, this behavior was not observed in the reciprocal direction (Table 1). Moreover, introgression in the A. hybridus direction occurred only in progeny identified as recombinant with AFLP markers, and A. tuberculatus’ ALS and PPO alleles showed Mendelian segregation among these individuals (Table 1).

Table 1.   Allele introgression at loci associated with herbicide resistance
Recurrent parentIntrogression at ALS locusIntrogression at PPO locus
Frequencyχ2 (1 : 1)PFrequencyχ2 (1 : 1)P
  1. Allele frequencies, χ2 statistics for Mendelian segregation and corresponding probabilities are given for introgression at the ALS and PPO loci, where herbicide resistance-conferring mutations have been identified. Calculations were conducted using all of the analysed backcross offspring produced with A. hybridus (all) or only those identified as recombinant (recomb.) using amplified fragment length polymorphism (AFLP) markers.

Amaranthus hybridus (all)0.1060.2< 0.0010.0862.2< 0.001
A. hybridus (recomb.)0.362.30.1290.351.80.180
Amaranthus tuberculatus0.432.00.1570.630.80.371

Interestingly, introgression was never observed in individuals exhibiting A. hybridus’ sexual condition and taxonomically discriminating character: monoecism. As noted previously for backcross progeny with A. hybridus, recombinant individuals exhibited a heavy fecundity penalty relative to nonrecombinant siblings, that is, fecundity decreased with increased introgression (Fig. 3b). The ALS and PPO loci are no exception to this phenomenon, with significant fecundity penalties for heterozygosity in both instances (r = −0.61 and r = −0.67 for ALS and PPO, respectively; P < 0.0001 for both loci). Consequently, continued introgression of ALS and PPO alleles is disfavored in the A. hybridus direction. By contrast, introgression at either loci in the A. tuberculatus direction showed no association with fecundity (P = 0.3049 and P = 0.1288 for ALS and PPO, respectively).


As stated previously, A. tuberculatus is known for evolving resistance to herbicides including triazines, ALS-inhibitors, PPO-inhibitors and glyphosate (Heap, 2009). However, some of these resistances have not been reported in widely distributed weeds of the A. hybridus complex, despite high rates of interspecies hybridization with A. tuberculatus (Trucco et al., 2005a,b) and some reproductive viability in the F1 (Trucco et al., 2006). Moreover, sympatric populations of A. tuberculatus and A. hybridus show independent evolution for ALS-inhibitor resistance alleles (P. J. Tranel, unpublished), even though one would predict resistance to evolve via gene flow before selection of de novo mutations (Jasieniuk et al., 1996). This raises the question of whether gene introgression can proceed after the F1 generation.

In a previous study, we used DNA content analysis to examine the patterns of gene introgression after the F1 generation and observed that most progeny produced upon backcrossing hybrids with A. tuberculatus or A. hybridus were homoploid and recombinant (Trucco et al., 2005c). However, just as observed herein, A. tuberculatus ALS alleles did not show introgression into the monoecious A. hybridus background. We suggested then that lack of introgression resulted from linkage of ALS to a hybrid sterility locus associated with sex determination, a taxonomically discriminating character for these species. It is evident now, after failing to transfer a second herbicide resistance allele – in this instance at the PPO locus – that genetic exchange likely is limited by a phenomenon beyond circumstantial linkage.

Indeed, we were unable to transfer most of 133 AFLP markers from A. tuberculatus to A. hybridus, with the exception of introgression in a smaller group of nonmonoecious BC1s characterized by anomalous phenotypes and F1-like fecundities, or worse. This observation is striking as we uncovered a very different scenario in the reciprocal exchange. We were able to transfer from A. hybridus to A. tuberculatus not only ALS and PPO alleles but also most of our A. hybridus-specific AFLP markers. Although introgression at some loci appeared to be disfavored (i.e. showed negative segregation distortion), ALS and PPO alleles as well as many of the AFLP-markers showed Mendelian segregation in backcross progeny with A. tuberculatus. This was not observed in progeny from reciprocal backcrosses, where almost all markers (strictly species-specific or more permissive markers) showed strong negative distortion and increased introgression was associated with reduced reproductive viability.

Taking monoecism and dioecism as the taxonomic distinguishing characters for A. hybridus and A. tuberculatus, respectively, we may say that contrary to previous speculations (Trucco et al., 2005c), genetic exchange between these species is unidirectional. Even if we considered nonmonoecious recombinant BC1s obtained with A. hybridus, the evolutionary contribution of these individuals seems restricted by the dramatic fitness disadvantage compared with their reciprocals. By contrast, there is no significant association between the level of introgression measured in A. tuberculatus BC1s and their seed output. We should point out however, that a small but significant penalty was observed when we measured micropollen occurrence, an indicator of male sterility (Srivastava et al., 1977), and level of introgression in males. This difference could result from the differential sensitivity of the parameters measured, with micropollen data more closely reflecting meiotic dysfunctionalities than seed yield. Also, there may more hybrid incompatibilities that affect male fertility than those that affect female fertility. Alternatively, or in addition, the difference observed between male and female reproductive fitness may be a reflection of Haldane's rule (Haldane, 1922), as males are the heterogametic sex (Murray, 1940). If this is the case, ovules will be favored over pollen for continued introgression. However, this is a matter for further investigation.

The study of natural populations from hybrid zones has provided evidence on the occurrence of asymmetric introgression among different hybridizing plant (Sweigart & Willis, 2003; Burgess et al., 2005) and animal species (Johannesen et al., 2006; Kraus et al., 2007). Turelli & Moyle (2007) indicate that Darwin was the first to call attention to this phenomenon, which the authors proposed to name ‘Darwin's Corollary’. Asymmetric hybridization is commonly observed in the generation of interspecific hybrids (Lewis & Crowe, 1958; Rick, 1963), especially as influenced by prezygotic reproductive barriers, such as ecological factors and differences in flowering times and characters (Levin, 1978). The occurrence of differential viability in reciprocal hybrids (i.e. postzygotic asymmetries) is also well documented (Tiffin et al., 2001), though it is seldom analysed beyond the first generation. Our study is novel in that postzygotic asymmetry is observed not in F1 viability (i.e. F1 pollen or seed production with either parental species) but in the array of genotypes and fitness values generated upon backcrossing.

Nonsymmetrical postzygotic isolation is commonly attributed to nuclear–cytoplasmic interactions (Levin, 1978; Tiffin et al., 2001), though this type of asymmetry is usually observed in the F1 generation. Interestingly, the F1 hybrids used in our study were produced with A. hybridus as the seed parent. Therefore, contrary to the actual observation, we would expect greater compatibility in the A. hybridus direction, as deleterious A. tuberculatus-A. hybridus nuclear–cytoplasmic interactions should be minimized in this cross.

Chromosomal rearrangement models of postzygotic reproductive isolation predict a symmetrical outcome, as both parents are considered to contribute equally to the chromosomal composition of hybrids (Tiffin et al., 2001). A symmetrical outcome may also be predicted based on classical Dobzhansky–Muller nuclear interactions (Levin, 1978). However, this model may predict asymmetry when sex chromosomes are involved, as is the case with A. tuberculatus (Murray, 1940). For example, consider the existence of a deleterious interaction between an X-linked gene from A. tuberculatus and a recessive gene from A. hybridus. This interaction will not manifest in the F1, as first-generation hybrids will be heterozygous at the recessive gene. Upon backcrossing with A. hybridus, the recessive condition could be obtained and BC1 viability will be observed only in individuals without the X-linked gene (presumably monoecious) or in individuals heterozygous at the recessive gene. Since A. hybridus’ homozygous recessive condition cannot be restituted upon backcrossing with A. tuberculatus, the lethal interaction cannot be obtained in this direction.

From a practical standpoint, results of this research indicate that it may be difficult for A. hybridus to acquire herbicide resistance from A. tuberculatus, despite the species’ cohabitation and propensity to hybridize with each other. This is significant in that numerous populations of A. tuberculatus have evolved resistance to PPO-inhibiting herbicides and/or to the even more widely used herbicide, glyphosate, whereas A. hybridus has not yet evolved these resistances (Heap, 2009). By contrast, our data indicate that A. tuberculatus can readily acquire genetic material from A. hybridus. Did such genetic acquisition foster expansion of A. tuberculatus from its primarily riparian habitat to agricultural fields throughout the midwestern USA? Although our results do not directly address this question, they add to a growing body of evidence that interspecific hybridization has the potential to play a significant role in the evolution of at least some Amaranthus species as weeds. Research is now needed to find a causative relationship between gene introgression and enhanced weediness in A. tuberculatus, and to determine the genomic basis for the herein discovered asymmetric genetic exchange. Such research provides an excellent opportunity for collaboration between evolutionary ecologists and crop scientists.


This work was supported by a Research Grant from Instituto de Agrobiotecnología Rosario (INDEAR) SA. We thank Jonathan Gressel for inspiring the title of this paper and three anonymous reviewers for many valuable comments.