Andrew H. Paterson and J. Mike Chandler

Dept Soil and Crop Science, Texas A & M University, College Station TX 77843-2474; ahp2343@bioch.tamu.edu, fax: (409)845-3773


The primary objective of this project is to investigate the rate and extent of gene flow from cultivated sorghum (Sorghum bicolor L. Moench.) to "johnsongrass" (Sorghum halepense L. Pers.), in order to better define risks associated with use of "transgenic" sorghum. The capability for "transformation" of sorghum with exogenous DNA now exists (Casas et al., 1993), and many genes for resistance to herbicides, diseases, and insects, are available (Fraley et al., 1988). Supplementary objectives include study of the origins of johnsongrass, introduction and spread of johnsongrass across the USA, and possible effects of selection pressures for key traits such as rhizomatousness on the gene pool of johnsongrass.

A close relative of sorghum, johnsongrass (Sorghum halepense), is one of the most noxious weeds in US and world agriculture (Holm et al., 1977). Naturalized populations of johnsongrass are prevalent in the major sorghum production regions of the USA, and extend as far north as southern Ontario (Canada). Johnsongrass reduces yields of maize, soybean, cotton and other major US crops as much as 45% (Millhollen, 1970; McWhorter and Hartwig, 1972), largely by shading, but also by competition for water and nutrients. In addition, johnsongrass serves as an alternate host, and over-wintering site, for insects and diseases which spread to cropping fields during the growing season, and further reduce crop yields (Anderson, 1977).

As a "low-input" crop grown on 8-14 million acres per year in the USA with an annual crop value of ca. $1 billion, considerable economic advantage might derive from genetically-engineered sorghums; however, such potential gains do not compensate for appreciable risk that "transgenes" will make johnsongrass a more serious problem than it already is. It is well established that crosses between sorghum and johnsongrass can occur (cf. Hadley, 1958), but until now, the tools have not been available to investigate whether such crosses actually contribute to the gene pool of johnsongrass. We have developed a large body of tools (Chittenden et al., 1994) and information (Paterson et al., submitted, Appendix A) relevant to this question, and seek to provide new data to support regulatory agency decision-making in regard to potential impacts to the environment over time, of release of genetically-engineered sorghums.


Johnsongrass (Sorghum halepense L. Pers.) is one of the most noxious weeds affecting US and world agriculture (Holm et al., 1977), representing a major contaminant in sorghum seed production, an alternate host and means of over-wintering for pests and pathogens of both monocotyledonous and dicotyledonous crops, and a highly-effective competitor for sunlight and other growth-limiting resources. For example, reductions in economic yield of 45% or greater in monocots such as sugarcane (Millhollen, 1970) and dicots such as soybean (McWhorter and Hartwig, 1972), are specifically attributable to johnsongrass. The name "johnsongrass" came to supplant some 40 common names, and is first documented in an 1874 letter, referring to an Alabama plantation owner, Col. William Johnson, who sowed it on part of his farm (McWhorter, 1971). Johnsongrass may have been intentionally introduced into the USA as a prospective forage, and/or unintentionally introduced as a contaminant of seedlots, and was established in several states by 1830 (McWhorter, 1971). The rapid spread of johnsongrass over the USA and into Canada has been attributed to commercial sale of seed, contamination in seedlots of other species, cavalry movements during the Civil War, intentional planting of rhizomes to resuscitate "worn-out fields" after the Civil War, flooding, and establishment along rights-of-way after escape from railroad boxcars (McWhorter and Hartwig, 1972). As early as 1896, johnsongrass was acknowledged as one of six weeds about which USDA received the greatest number of complaints (Coville, 1896), and in 1900 the first federal appropriation for weed control research targeted johnsongrass (House Bill #121, 1900).

Modern herbicides provide effective control of johnsongrass in maize, soybean, cotton, and many other crops, but at a cost of $12-20 per acre (J. M. Chandler, pers. comm.). However, no herbicide is available which kills johnsongrass but not sorghum; this means that johnsongrass remains uncontrolled on the 8-14 million acres per year in sorghum production.

Taxonomic affinity of johnsongrass with cultivated sorghum (Sorghum bicolor , 2n = 2x = 20) is clear. The only morphological feature which consistently distinguishes S. halepense from S. bicolor, is its well-developed rhizomes (Celarier,1958). Both rhizomatous habit, and geographical distribution, suggest that S. halepense is also a descendant of S. propinquum Kunth. (2n = 2x = 20), a strongly rhizomatous species found in southeast Asia, Indonesia, and the Philippines (Doggett, 1976). S. propinquum may account for reports of putative "S. halepense" with ploidy of 2n = 20 (Celarier, 1958). The natural distribution of S. halepense overlaps the distributions of both S. bicolor and S. propinquum (Doggett, 1976). S. bicolor has a center of diversity in Africa; however, the greatest morphological diversity of 2n = 40 sorghums is in southeast Asia, supporting the hypothesis that S. propinquum contributed to S. halepense (Celarier, 1958). Another similar weedy sorghum, S. almum ("Columbus grass", 2n = 4x = 40), is also widespread (Doggett, 1976).

Recent molecular evidence strongly supports the classical inference (Doggett, 1976) that johnsongrass is an inter-specific hybrid of S. bicolor x S. propinquum (Paterson et al., 1995). The proposed inter-specific origin leaves only 6% of johnsongrass alleles unaccounted for, and it is estimated that 1.6% may be novel alleles which have arisen since the divergence of johnsongrass from both the diploid sorghums and S. almum. By contrast, a previously proposed origin within Sorghum bicolor, involving Sorghum bicolor ssp. virgatum (Bhatti et al., 1960) fails to account for more than 24% of alleles found in johnsongrass (Paterson et al., 1995). Another prior investigation (Morden et al., 1990) reported alleles at four isozyme loci in S. halepense and S. almum which could not be accounted for either by S. bicolor or S. propinquum; however, only a single accession of S. propinquum was available. In preliminary results (Paterson et al., 1995), the two accessions of S. propinquum examined were highly-variable, showing different RFLP alleles for 29 (74%) of the 39 DNA probes examined, and harboring almost as many RFLP alleles as the 11 accessions of S. bicolor examined; thus, consideration of a single accession of S. propinquum could not provide conclusive data (as was acknowledged; Morden et al., 1990).

Curiously, the six S. halepense accessions examined in our preliminary study appear to show a greater contribution of alleles from S. bicolor than S. propinquum (Paterson et al., 1995). This is in contrast to the four S. almum accessions examined, which show equal contributions from S. bicolor and S. propinquum. While the evolution of a 2n = 40 polyploid from two 2n = 20 diploids superficially suggests equal contributions from each diploid, many instances of unequal contributions in such cases have been documented (Harlan and deWet, 1975).

In addition to the possibility of "asymmetrical polyploidization", it is conceivable that introgression from S. bicolor into S. halepense, perhaps by association of the weed with cultivated fields, might contribute to the bias in allele composition. Many authors have reported artificial hybrids between the two species (cf. Hadley, 1958; also see list cited in Tang et al., 1991); however, the hybrids tend to be only weakly rhizomatous, and it is unknown whether they serve as an effective conduit for gene flow from sorghum to johnsongrass. Both formal and anecdotal reports have suggested that specific traits in local johnsongrass populations might have been derived from sorghum, however no means has previously existed to investigate such hypotheses in an unequivocal manner. For example, introgression from S. bicolor would be consistent with morphology of non-over-wintering populations of S. halepense at the northerly extreme of its distribution in the Americas (Warwick et al., 1984); however, considerable variation in morphology of johnsongrass is known (cf. Celarier, 1958), thus morphology provides only an equivocal indicator. In view of the possible consequences associated with "escape" of transgenes introduced into sorghum (mentioned earlier in this section), molecular analysis of the possibility of gene flow from sorghum to johnsongrass is warranted.


We will use new DNA tools for study of the extent of gene flow from sorghum to johnsongrass. In order to study gene flow from cultivated sorghums into johnsongrass, it will be necessary to find DNA marker alleles which distinguish cultivated sorghums, from the wild sorghums which gave rise to johnsongrass. Regarding the Sorghum propinquum component of the johnsongrass genome, this is straightforward; S. propinquum and S. bicolor are highly diverged (Chittenden et al., 1994), and show an abundance of species-specific DNA marker alleles. However, a more critical, and more difficult, component is the Sorghum bicolor component of the johnsongrass genome.

DNA marker loci harboring alleles which are common in leading S. bicolor cultivars, but rare elsewhere in the gene pool of S. bicolor, will be identified in a preliminary screen of leading cultivars from the Texas and Nebraska regions under study, and S. bicolor accessions representing the breadth of diversity in the Sorghum genus, using two DNA marker technologies (see techniques). In addition, the screen will include available accessions of S. propinquum, S. halepense, S. almum, S. nitidum, and bulk samples from each of the S. halepense populations to be studied herein.

The repertoire and frequency of alleles at these loci will be investigated in 5 johnsongrass populations, from disparate locations across the US, and in different proximity's to cultivated sorghum. In addition, allelic variation and allele frequencies at 100 additional loci, evenly-distributed across the genome of sorghum (based on prior genetic mapping; Chittenden et al., 1994, and AHP et al., unpubl. results), will be determined to assess levels and patterns of DNA polymorphism in the populations.

The following specific investigations will be done, using populations described below:

Determine the levels and patterns of DNA polymorphism (genetic variation) in johnsongrass populations from different areas of the USA.

Identify genetic marker alleles which distinguish cultivated US sorghums, from the wild sorghums likely to have given rise to johnsongrass.

Determine whether DNA marker alleles common in US sorghums but rare in wild sorghums are found in US johnsongrass populations, indicating gene flow from sorghum to johnsongrass.

From each of 5 johnsongrass populations, 100 plants not connected by rhizomes will be sampled, over an area of no more than 1 square mile, and at locations and conditions described below. The 5 johnsongrass populations (1-5) and the Sorghum bicolor, S. halepense, S. almum, and S. propinquum materials to be used (6) are described as follows:

Population 1: South/Sorghum Belt/Sympatric with sorghum. A johnsongrass population will be sampled near Temple, TX (near the southerly end of the sorghum belt), along the fringes of a field which has been in annual or biannual sorghum production for much of the past century. Sampling will be done by the personnel supported (co-PIs, and/or postdoc).

Population 2 South/Sorghum Belt/Allopatric to sorghum. A johnsongrass population growing near Beaumont TX, in a region where rice is the predominant cultigen and virtually no sorghum is produced (or has been for many decades), will be sampled. Although much of this area is wetlands, there are also large areas where soil type is similar to Temple (Population 1); sites will be chosen to minimize differences in edaphic conditions. Rainfall is somewhat different in these regions (ca. 70-80PWz per yr in Beaumont, and ca. 30-40PWz per year in Temple) but does not represent a limiting factor to sorghum growth in either region.

Population 3: North/Sorghum Belt/Sympatric with sorghum. A johnsongrass population growing near Concord, Nebraska will be sampled, along the fringes of a field which has been in annual or biannual sorghum production for over 30 years, approximately the length of time which johnsongrass has inhabited the area. The population will be sampled by a former student of the co-PI (JMC), Dr. David Holshouser (located at University of Nebraska-Lincoln, Northeast Research and Extension Center, Concord, NE 68728).

Population 4: South/outside the Sorghum Belt. A johnsongrass population growing near Tifton, GA, in the region where johnsongrass was first established in the US, but where sorghum is seldom cultivated. The population will be sampled by a colleague of the co-PI (JMC), Dr. Steven Brown (located at Dept. of Crop and Soil Sciences, University of Georgia, Rural Development Center, Tifton, GA 31793).

Population 5: North/outside the Sorghum Belt. A johnsongrass population growing near Bridgeton, NJ will be sampled, in a region where sorghum has never been produced, and johnsongrass has become prominent only in the past 30 years. The population will be sampled by a colleague of the co-PI (JMC), Dr. Brad Majek (located at Rutgers Research and Development Center, 121 Northville Road, Bridgeton NJ 08302).

Additional Sorghum bicolor, S. propinquum, S. halepense, and S. almum materials to be used for identification of diagnostic alleles include representative sorghum cultivars grown near Temple, TX (by recommendation of C. Coffman, Extension Sorghum Specialist, Texas A & M, College Station, TX), and Concord NE (by recommendation of D. Holshouser; University of Nebraska-Lincoln, Northeast Research and Extension Center, Concord, NE 68728), all available accessions of S. halepense, S. almum, and S. propinquum from the USDA Germplasm Collection, as well as 70 accessions of S. bicolor representing the races of Harlan and deWet (1972), and the "working groups" of Snowden (1936), as recommended and provided by the curator of the USDA Sorghum Collection (Dr Jeff Dahlberg, Mayaguez, PR). This collection represents a minimal number of accessions to encompass a maximal portion of genetic diversity within 2n = 20-40 types in the genus Sorghum. Although 2n = 10 types exist, there is no evidence that they contributed to johnsongrass, and only one (S. nitidum) bears close resemblance to 2n = 20-40 types at the molecular level (Sun et al., submitted; provided by S. Hulbert, Kansas State Univ., Manhattan, KS). A single accession of this species, S. nitidum , was available and has been included in the study.

Particularly important will be leading cultivars from the regions of Texas and Nebraska (respectively) in which the "gene flow" studies will be conducted. A total of 22 cultivars, 11 from each region, have been identified based on recent production records (provided by Extension personnel cited). Although the cultivars from whence introgression would have been derived 100 years ago are no longer available, the gene pool of cultivated sorghum is quite narrow (Morden et al., 1990), and modern cultivars from the same regions are likely to contain a similar range of DNA marker allele variation to obsolete cultivars. Moreover, we have intentionally chosen hybrid sorghums, to maximize the number of alleles represented in these sorghums, and afford a more comprehensive picture of the gene pools of Texas and Nebraska sorghum, respectively.


In this, the first year of the study, we have focussed on collecting plant samples from the locations described above. All five locations have been sampled, collecting 120 samples from each location to allow for any problems with quality or DNA yield from any particular sample. Preparation of genomic DNA for lab analysis is in progress. In addition, seed of the relevant cultivated and exotic sorghums have been collected, and genomic DNA has been prepared from most of these.

In a preliminary screen of 11 S. bicolor accessions with 39 RFLPs (Paterson et al., 1995), we found considerable variation among the races of S. bicolor. Although the specific S. bicolor accession(s) which gave rise to johnsongrass is/are unknown (indeed, it is even unknown whether johnsongrass only originated once), many alleles were identified in our preliminary screen which distinguished S. bicolor from the bulked samples of johnsongrass we examined. Moreover, about 50% of the loci examined showed species-specific alleles useful for further characterization of the genome composition of johnsongrass. Additional species-specific markers are being sought during routine evaluation of new DNA probes on sorghum. The sorghum RFLP map has been increased to more than 1000 loci, providing a large number of mapped DNA probes to choose from for this experiment.

In addition to characterizing RFLP loci, a minimum of 30 highly-polymorphic sequence-tagged microsatellites will be applied to these genotypes -- this technology is much more laborious on a per-sample basis than RFLPs, but affords a higher likelihood of detecting the diagnostic polymorphisms needed to study gene flow from sorghum to johnsongrass. STMs for sorghum are available from Dr. Steve Kresovich, Griffin, GA.


Although it is well-established that cultivated sorghum can, in principle, hybridize to johnsongrass, we seek to evaluate the actual contribution of such hybrids to the gene pool of johnsongrass. Such contributions to the gene pool require not only hybridization, but fitness of hybrid progeny, which is less-well defined. Evaluation of gene pool composition in genomic regions specifically associated with traits such as rhizomatousness, identified by prior QTL mapping, may be of particular importance. Moreover, our data are expected to reveal information about the spread of johnsongrass across North America over the past 150 years, with particular regard to the relative importance of vegetative (rhizomes) and reproductive (seed) propagules.

Sorghum is typically grown in the US on 8-14 million acres per year, with an annual crop value of ca. $1 billion. Gene flow from sorghum to johnsongrass would represent a serious potential risk, in commercial use of bio-engineered sorghum cultivars. The prospect of engineering sorghums resistant to major diseases and insects, as well as tolerant of herbicides, holds promise for improving productivity and profitability of this important dryland grain and feed crop, with particular benefit to dryland agricultural regions of Kansas, Nebraska, Oklahoma, and Texas. However, the promise of sorghum biotechnology may be a two-edged sword; exposure of johnsongrass to sorghum cultivated on more than 8 million acres in the USA alone, presents an enormous opportunity for gene exchange, even if such events are relatively rare. Because of the close relationship between johnsongrass and sorghum, no herbicide is known which can kill johnsongrass but not harm sorghum; thus, the 8-14 million acres of sorghum grown each year provides a "haven" of sorts, for johnsongrass. Gene flow from sorghum to johnsongrass would not only affect weed control in sorghum, but also in maize, soybean, cotton, and other crops.

The primary objective of this proposal is to assess the extent of gene flow from cultivated sorghums into johnsongrass, and study the rate of spread of sorghum genes through johnsongrass populations, to provide a sound basis to support regulatory agency decision-making in regard to potential impacts to the environment over time, of release of bio-engineered sorghums. In addition, the results of this study will also shed light on the origin of North American johnsongrass populations, dispersal of johnsongrass throughout the USA, and levels of genetic variation near genes fundamental to dispersal of johnsongrass (especially rhizomatousness) relative to other regions of the genome. Finally, in the course of identifying alleles diagnostic of cultivated sorghums, a large body of data regarding allelic variation across the Sorghum genus will be assembled. These data will have widespread utility in sorghum breeding and genetics, as well as in the proposed investigation.


We appreciate the support of the USDA-Biotechnology Risk Assessment Program for our research.


Anon. 1900. House Bill No. 121, 56th Congress, first session, appropriation for U.S. Department of Agriculture.

Casas, A.M., Konowicz, A.K., Zehr, U.B., Tomes, D.T., Axtell, J.D., Butler, L.G., Bressan, R.A., Hasegawa, P.M. 1993. Proc. Nat. Acad. Sci. USA 90: 11212-11216.

Celarier, R.P. 1958. Cytotaxonomic notes on the subsection Halepensia of the genus Sorghum. Bull. Torr. Bot. Club 85: 49-62.

Chittenden, L.M., Schertz, K.F., Lin, Y-R., Wing, R.A., Paterson, A.H. 1994. Theor. Appl. Genet. 87: 925-933.

Coville, F.V. 1896. Weed investigations, p. 96-97. In: Ann. Rep. of the Secretary of Agr. (U.S. Government Printing Office, Washington, D.C.).

Doggett, H. 1976. Sorghum. In Simmonds, N. W., Evolution of Crop Plants. Longman Press, Essex, UK. pp. 112-117.

Doggett, H. 1988. Sorghum, 2nd ed. John Wiley and Sons, Inc., NY. 512 pp.

Egley, G.H., and J.M. Chandler. 1978. Germination and viability of weeds after 2.5 years in a 50-year buried seed study. Weed Science 26: 230-239.

Fraley, R.T., N.M. Frey, J. Schell. 1988. Genetic improvement of agriculturally important crops. (Cold Spring Harbor Laboratory, Cold Spring Harbor NY). pgs. 13-20, 37-40, 47-54, 55-61, 77-82.

Gizmawy, I., J. Kigel, D. Koller, M. Ofir. 1985. Initiation, orientation, and early development of primary rhizomes in Sorghum halepense (L.) Pers. Ann. Bot. 55: 343-350.

Harlan, J.R. and J.M.J. deWet. 1972. A simplified classification of cultivated sorghum. Crop Sci. 12: 172-176.

Harlan, J.R., and J.M.J. deWet. 1975. On O. winge and a prayer: The origins of polyploidy. Bot. Rev. 41: 361-390.

Holm, L.G., Plucknett, D.L., Pancho, J.V., Herberger, J.P. 1977. The world's worst weeds: distribution and biology (University Press of Hawaii, Honolulu), pp. 54-61.

McWhorter, C.G. and Hartwig, E.E. 1972. Competition of johnsongrass and cocklebur with six soybean varieties. Weed. Sci. 20:56-59

McWhorter, C.G. 1971. Introduction and spread of johnsongrass in the United States. Weed Sci. 5: 496-500.

Morden, C.W., Doebley, J., and Schertz, K.F. 1990. Allozyme variation of sorghum lines and hybrids in the United States. Theor. Appl. Genet. 80: 296-304.

Paterson, AH, KF Schertz, YR Lin, SC Liu, YL Chang. 1995. The weediness of wild plants: molecular analysis of genes responsible for dispersal and persistence of johnsongrass (Sorghum halepense L. Pers.). Proc. Nat. Acad. Sci USA, 92: 6127-6131.

Schertz, K.F., Stec, A. and J. Doebley. 1990. Isozyme genotypes of sorghum lines and hybrids in the United States. Texas Agr. Expt. Stn. MP1719.

Snowden, J.D. 1936. The cultivated races of sorghum. Allard and Son, London.

Suneson, C.A., K.O. Rachie, G.S. Khush. 1969. A dynamic population of weedy rye. Crop Sci. 9:121-124.

Tang, H., and G.H. Liang. 1988. The genomic relationship between cultivated sorghum [Sorghum bicolor (L.) Moench] and johnsongrass [S. halepense (L.) Pers.]: a re-evaluation. Theor. Appl. Genet. 76: 277-284.

Tang, H., S.K. Dure, G.H. Liang, S. Kung. 1991. Possible repetitive DNA markers for Eusorghum and Parasorghum and their potential use in examining phylogenetic hypotheses on the origin of Sorghum species. Genome 34: 241-250.

Warwick, S.I., B.K. Thompson, L.D. Black. 1984. Population variation in Sorghum halepense, johnsongrass, at the northern limits of its range. Can. J. Bot. 62: 1781-1790.

Warwick, S.I., D. Phillips, C. Andrews. 1986. Rhizome depth: the critical factor in winter survival of Sorghum halepense (L.) Pers. (johnsongrass). Weed Res. 26: 381-387.