Proc. Nati. Acad. Sci. USA Vol. 84, pp. 8444-8447, December 1987 Botany Two enzymes involved in biosynthesis of the host-selective phytotoxin HC-toxin (cyclic peptide biosynthesis/maize/plant disease) JONATHAN D. WALTON Department of Energy Plant Research Laboratory, Michigan State University, East Lansing, MI 48824 Communicated by Anton Lang, August 24, 1987 (receivedfor review June 2, 1987)

ABSTRACT Cochliobolus carbonum race 1 produces a C. carbonum (imperfect stage Helminthosporium carbo- cyclic tetrapeptide HC-toxin, which is necessary for its excep- num or Bipolaris zeicola) race 1 synthesizes the host- tional virulence on certain varieties of maize. Previous genetic selective toxin HC-toxin (8-12). HC-toxin is a cyclic peptide analysis of HC-toxin production by the fungus has indicated with the structure cyclo(D-Pro-L-Ala-D-Ala-L-AOE), where that a single genetic locus controls HC-toxin production. AOE stands for 2-amino-8-oxo-9,10-epoxidecanoic acid (Fig. Enzymes involved in the biosynthesis of HC-toxin have been 1). The unusual amino acid AOE has been reported in three sought by following the precedents established for the biosyn- other cyclic tetrapeptides from three unrelated filamentous thetic enzymes of cyclic peptide antibiotics. Two enzymatic fungi (13-15). Maize (Zea mays L.) that is homozygous activities from C. carbonum race 1 were found, a D-alanine- and recessive at the nuclear Hm locus is susceptible to C. an L-proline-dependent ATP/PPj exchange, which by biochem- carbonum race 1 and sensitive to HC-toxin. Non-toxin- ical and genetic criteria were shown to be involved in the producing isolates (races 2 and 3) of C. carbonom do not biosynthesis of HC-toxin. These two activities were present in make HC-toxin and are only weakly pathogenic on many all tested race 1 isolates of C. carbonum, which produce maize cultivars (16). HC-toxin, and in none of the tested race 2 and race 3 isolates, A number of the enzymes that catalyze the biosynthesis of which do not produce the toxin. In a genetic cross between two cyclic peptide antibiotics have been studied (17-19). In all isolates of C. carbonum differing at the tox locus, all tox+ cases, the enzymes are large multifunctional proteins, con- progeny had both activities, and all tox- progeny lacked both tain the cofactor pantetheine, and activate the constituent activities. amino acids as thioesters (20-22). These enzymes frequently also catalyze modifications of the amino acids such as Some plant pathogens produce low molecular weight com- racemization (23), methylation (24, 25), acetylation (26), and pounds that selectively affect only those cultivars of the host rearrangements (27). Because cyclic peptide synthetases plant the pathogen is able to infect. Numerous investigators activate each amino acid independently of the others prior to have demonstrated the importance of these host-selective bond formation, it is not necessary to have complete in vitro toxins in the plant disease interactions in which they occur biosynthesis of the cyclic peptide in question to assay them. (1). Thus, in theory it should be possible to identify enzymes Several species of the pathogenic fungus Cochliobolus involved in the biosynthesis of cyclic peptides, such as [also known by the names for its imperfect stage Helmin- HC-toxin, which contain amino acids not easily obtained. thosporium or Bipolaris (2)] make host-selective toxins. In The activation reaction catalyzed by cyclic peptide anti- those species that have been studied, single genetic loci biotic synthetases is control toxin production. A single genetic locus in Cochliobolus carbonum race 1 controls production of the amino acid + ATP = aminoacyl-AMP + PP1. host-selective toxin HC-toxin (3-5). Production of the host- selective toxins HMT-toxin and victorin by Cochliobolus Under the appropriate conditions, this reaction has a net heterostrophus race T and Cochliobolus victoriae, respec- change in free energy ofclose to zero and hence can be driven tively, are also inherited monogenically (5-7). The progeny of in the reverse direction. In the standard amino acid activation crosses between C. carbonum race 1 and C. victoriae assay, radiolabeled pyrophosphate is used, and amino acid- segregate 1:1:1:1 for production of HC-toxin, victorin, both dependent incorporation of 32p into ATP is measured (28). toxins, or neither toxin (5). How single genetic loci control In the work reported here, I describe how the precedents biosynthesis of host-selective toxins is not known (1). established from the study of cyclic peptide antibiotic syn- One approach to understanding the functions of the genes thesis were used to identify two enzymatic activities involved (called the tox loci) that control host-selective toxin produc- in the biosynthesis of HC-toxin. tion would be to identify the products of those genes. Although we cannot at this point exclude the possibility that MATERIALS AND METHODS the tox genetic loci are regulatory, they more likely encode one or several enzymes involved in biosynthesis of the Fungal Isolates. Isolates of C. carbonum were from the toxins. Identification of the enzymes that catalyze the syn- American Type Culture Collection [ATCC 36384, originally thesis of host-selective toxins would thus allow a rational isolated in New York State (4)], R. P. Scheffer (Michigan approach to the study ofthe tox genes from Cochliobolus and State University; isolates 81-64 and 73-4), and K. J. Leonard hence of the molecular structure and evolution of genes (North Carolina State University; isolates 1309, 926, 1368, controlling virulence and host range in this important group and 1274). Isolates SB111 and SB114 (which are siblings) and of plant pathogenic fungi. the progeny derived from a genetic mating between them (see Table 2) were generously provided by S. P. Briggs (Pioneer The publication costs of this article were defrayed in part by page charge Hi-Bred International, Johnston, IA; current address, Cold payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. Abbreviation: AOE, 2-amino-8-oxo-9,10-epoxidecanoic acid.

8444 Downloaded by guest on September 25, 2021 Botany: Walton Proc. Natl. Acad. Sci. USA 84 (1987) 8445 D-Ala High-Pressure Liquid Chromatography (HPLC). Protein CH3 fractionation by HPLC was done on a Beckman system with a model 421 controller, two model 114 pumps, and a model 163 UV a 7.5 CH3 / \ detector (280 nm). Gel filtration was on x 600 XCH NNH L-Aoe mm TSK3000 SW column (Beckman), with a mobile phase of 100 mM KH2PO4, pH 7.0/4 mM dithiothreitol/10% (vol/vol) NH H C- (CH2)5 - CH-C H2 glycerol. Flow rate was 0.5 ml/min. Ion exchange was on a TSK DEAE-5PW anion exchange column (Beckman) at a / 0 o o c >3° flow rate of 0.8 ml/min. Buffer A was 100 mM KH2PO4, pH / \ 7.0/4 mM dithiothreitol/10% (vol/vol) glycerol; buffer B was CH2 CH2 the same except the K142PO4 concentration was 500 mM. The C2 linear gradient started 10 min after injection ofthe sample and 0- Pro proceeded to 100% B in 40 min. The protein standards used to calibrate the gel filtration NC-toxin column (Fig. 2) were from Bio-Rad or Sigma, and were thyroglobulin (Mr, 670,000), apoferritin (Mr, 436,000), 8- FIG. 1. Structure of HC-toxin (9-12). glucuronidase (Mr, 290,000), 83-amylase (Mr, 200,000), and IgG (Mr, 158,000). Spring Harbor Laboratory, Cold Spring Harbor, NY). Cul- Eizyme Assay. ATP/PPi exchange activity was measured tures were maintained on V-8 juice agar. according to Lee and Lipmann (28), except N-tris(hydroxy- Growth ofthe Fungus and Enzyme Extraction. C. carbonum methyl)methyl-2-aminoethanesulfonic acid (TES) buffer (pH was grown for 4-5 days at 22°C on modified Fries' medium 7.0) was used. The assay mixtures were incubated at 30°C for [125 ml per 1000-ml Erlenmeyer flask (8)]. The fungal mats 30 min and stopped by addition ofcharcoal. The charcoal was were collected by filtration, freeze-dried, and ground with a filtered on Whatman GF/A glass fiber filters, washed twice mortar and pestle in liquid nitrogen. The powder was with 5 ml of 0.04 M sodium pyrosphosphate plus 1.4% reground with a small amount of extraction buffer [50 mM (vol/vol) perchloric acid, and once with 5 ml of water. KH2PO4, pH 7.0/10% (vol/vol) glycerol/4 mM dithiothrei- Radioactivity retained on the filters was measured with a tol] plus 1 mg of phenylmethylsulfonyl fluoride per ml and scintillation counter. 0.2 M KCI (24), and then stirred for 1 hr at 4°C. The extracts were centrifuged (36,000 x g for 20 min), made 0.2% (vol/vol) with neutralized polyethylenimine (Sigma), and RESULTS centrifuged (36,000 x g for 10 min). The supernatant was Extracts of C. carbonum race 1 that had been partially Made 40o saturated with solid ammonium sulfate, stirred for purified by precipitation with 40% ammonium sulfate con- - 20 min, and the precipitated protein was collected by cen- tained ATP/PP, exchange activities dependent on L-proline, trifugation (20,000 x g for 15 mim). The protein pellet was D-alanine, and L-alanine. Protein fractions precipitated by redissolved in extraction buffer and either assayed directly or higher concentrations of ammonium sulfate contained the desalted on Sephadex G-25 before chromatography. bulk of amino acid-independent ATP/PP, exchange activity

80 VO 670K 436K 2

o 60 E 40- LUe~40

-c D~s 20

Time (min) FIG. 2. Gel filtration HPLC chromatography of extract from C. carbonum race 1 (isolate SB111) and identification of D-alanine, L-alanine, and L-proline ATP/PP; exchange activities. After precipitation with aipmonium sulfate (40o saturation), 4 mg total protein was loaded. Activity in the absence of exogenous amino acid was subtracted. -, Absorbance at 280 nm; *, D-alanine exchange activity; *, L-alanine activity; o, L-proline activity. Downloaded by guest on September 25, 2021 8446 Botany: Walton

Table 1. D-Alanine- and L-proline-dependent ATP/PP, exchange in isolates of C. carbonum races 1, 2, and 3 collected in the field ATP/PPj exchange activity, pmol mink'u.g- Isl 0 Isolate Race D-Alanine L-Proline x ATCC 36384 1 3.4 4.2 E

81-64 1 2.2 3.8 I-- 73-4 1 3.0 4.7 0 2W 2 0.2 0.1 0._ 0 1 1309 2 0 0 0 0 926 2 0.1 0 0. cL 0_ 1368 3 0.1 0.1 0. 1274 3 0 0.1 0. (i.e., activity without addition of an amino acid). Additional I- activities dependent on L-proline and L-alanine (but not D-alanine) were precipitated at higher ammonium sulfate concentrations; these probably represent aminoacyl-tRNA synthetases for proline and alanine, respectively. The D- alanine and L-proline activities present in the 40% ammonium sulfate fraction were present in all isolates classified as race Time (min) 1 or tox+ and were absent or present only at much lower levels in the isolates classified as race 2 or 3 or tox- (Tables FIG. 3. Anion exchange HPLC chromatography of extract from 1 and 2). The variation between different tox+ isolates was C. carbonum race 1 (isolate SB111) and identification of D-alanine, extracts from race 2 or tox- isolates L-alanine, and L-proline ATP/PPj exchange activities. Fractions not reproducible. Several from the gel filtration step in Fig. 1 were pooled and chroma- of C. carbonum were further fractionated by gel filtration tographed on a TSK DEAE-5PW HPLC anion exchange column. HPLC; in no case was there any D-alanine or L-proline , Absorbance at 280 nm; *, D-alanine activity; 0, L-alanine exchange activity in the fractions corresponding to the activity; 0, L-proline activity. fractions from race 1 or tox+ preparations that had the activities. Both the D-alanine and the L-proline activities were co- individually had Mrs of 310,000, indicating that in the initial eluted from an HPLC gel filtration column with an apparent gel filtration run they had not been eluted as a complex. Cyclic peptide synthetases but not most aminoacyl-tRNA Mr of310,000 (Fig. 2). A weaker L-alanine-dependent activity transferases can use 2'-dATP instead of ATP as substrate was also coeluted. On an HPLC ion exchange column, the (29-31). The D-alanine and L-proline exchange activities from D-alanine and the L-proline activities were clearly separated C. carbonum used 2'-dATP 100% and 35% as effectively, (Fig. 3). the L-alanine activity remained with the D-alanine respectively,'as ATP. In its relative inability to use 2'-dATP, activity. The D-alanine and L-proline exchange activities the L-proline activity from C. carbonum race 1 differs from were also separated by dye ligand (Affigel-Blue; BioRad) and gramicidin and tyrocidine synthetases (30, 31). hydroxyapatite chromatography. When rerun on gel filtration HPLC after separation by ion exchange, the two activities DISCUSSION Table 2. D-Alanine- and L-proline-dependent ATP/PPj exchange The results indicate that two separable enzyme activities in the random ascospore progeny of a cross between towx and segregate with the Mendelian tox locus, which controls tox- strains HC-toxin biosynthesis. Both of these enzyme activities are ATP/PPj exchange ones that would be expected to be involved in the biosyn- Genotype activity, pmol'min-.'g-I thesis of HC-toxin, based on the precedents of other cyclic Isolate tox D-Alanine L-Proline peptide synthetases. Although L-amino acids are considered to be the natural substrates of cyclic peptide synthetases, Parents generally if a cyclic peptide contains a D-amino acid, the SB111 + 4.5 3.3 corresponding synthetase will activate both isomers (23). SB114 0.1 0.1 Epimerization occurs after amino acid activation. D-Proline Progeny is the isomer of proline found in HC-toxin (Fig. 1), but a 1 0.1 0 D-proline ATP/PPj exchange activity was not found in C. 3 0 0 carbonum race 1, only an L-proline activity. Bacitracin 4 0.1 0.1 contains D-ornithine but bacitracin synthetase activates only 5 + 7.2 1.9 L-ornithine (27). Likewise, actinomycin contains D-valine but 8 0 0 actinomycin synthetase II recognizes only L-valine (29). In its 10 + 6.0 1.9 inability to activate the D-isomer of proline the enzyme 11 0 0.1 described here resembles bacitracin and actinomycin 12 0 0 synthetases. 13 + 11.5 1.1 Several hypotheses have been proposed to account for 14 + 7.3 1.4 control of toxin production by single loci in the species of 15 0.2 0 Cochliobolus that have been studied genetically (1). One 16 + 4.4 1.3 hypothesis is that a tox locus is a mutation that by blocking 17 + 8.3 ND some pathway allows accumulation ofa phytotoxic metabolic The crosses were done as described (3), and the genotypes of the intermediate. However, demonstration-that the tox locus in progeny were determined by ability to infect maize susceptible to C. C. heterostrophus is dominant (32) argues against this inter- carbonum race 1. ND, not determined. pretation for that system, and the recently elucidated struc- Downloaded by guest on September 25, 2021 Botany: Walton Proc. Natl. Acad. Sci. USA 84 (1987) 8447 tures of HC-toxin (9-12) and victorin (33) argue against these 7. Yoder, 0. C. & Gracen, V. E. (1975) Phytopathology 65, toxins being metabolic intermediates. A second possible 273-276. explanation is that toxin biosynthesis only requires a single 8. Scheffer, R. P. & Ullstrup, A. J. (1965) Phytopathology 55, 1037-1038. enzyme encoded by a single gene. HMT-toxin is a polyketide; 9. Walton, J. D., Earle, E. D. & Gibson, B. W. (1982) Biochem. the polyketide 6-methylsalicylic acid is synthesized from Biophys. Res. Commun. 107, 785-794. acetate by a multienzyme composed of identical subunits 10. Kawai, M., Rich, D. H. & Walton, J. D. (1983) Biochem. (34). The cyclic depsipeptide enniatin B is synthesized by a Biophys. Res. Commun. 111, 398-403. single multifunctional polypeptide (24). Both compounds are 11. Liesch, J. M., Sweeley, C. C., Staffeld, G. D., Anderson, from filamentous fungi. A third possible explanation of the M. S., Weber, D. J. & Scheffer, R. P. (1982) Tetrahedron 38, genetic data is that tox loci are complex, composed ofseveral 45-48. contiguous genes that together encode all the enzymes 12. Gross, M. L., McCrery, D., Crow, F., Tomer, K. B., Pope, necessary to synthesize the toxins. My findings (Tables 1 and M. R., Ciuffetti, L. M., Knoche, H. W., Daly, J. M. & Dun- kle, L. D. (1982) Tetrahedron Lett. 23, 5381-5384. 2) favor this last possibility as the simplest explanation 13. Hirota, A., Suzuki, A., Aizawa, K. & Tamura, S. (1974) consistent with the data. Biomed. Mass Spectrom. 1, 15-19. By purification of these enzymes and application of estab- 14. Closse, A. & Huguenin, R. (1974) Helv. Chim. Acta 57, lished techniques, it should be possible to isolate the genes 533-545. encoding them. This would allow study of the molecular 15. Kawai, M., Pottorf, R. S. & Rich, D. H. (1986) J. Med. Chem. organization of the tox locus in C. carbonum. Ultimately, it 29, 2409-2411. should be possible to study the evolutionary relationships 16. Nelson, R. R., Blanco, M., Dalmacio, S. & Moore, B. S. between races of C. carbonum that can and cannot make (1973) Plant Dis. Rep. 57, 822-823. HC-toxin and between C. carbonum and other species of 17. Kleinkauf, H. & von Dohren, H. (1981) Curr. Top. Microbiol. Immunol. 91, 129-177. fungi that make cyclic peptides closely related to HC-toxin. 18. Laland, S. G. & Zimmer, T.-L. (1973) Essays Biochem. 9, Clearly, enzyme activities in addition to the three de- 31-57. scribed here (activation of L-alanine, D-alanine, and -L- 19. Vining, L. C. & Wright, J. L. C. (1977) Biosynthesis 5, 240- proline) are required for the complete synthesis ofHC-toxin. 305. These include those necessary for the epimerization of 20. Gilhuus-Moe, C. C., Kristensen, T., Bredesen, J. E., Zimmer, L-proline and L-alanine, for the biosynthesis and activation of T.-L. & Laland, S. J. (1970) FEBS Lett. 7, 287-291. AOE, and for cyclization of the peptide. Whether races and 21. Lipmann, F. (1971) Science 173, 875-884. isolates of C. carbonum that do not make HC-toxin have any 22. Gevers, W., Kleinkauf, H. & Lipmann, F. (1969) Proc. Natl. or all ofthese activities and whether these activities reside on Acad. Sci. USA 63, 1335-1342. 23. Vater, J. & Kleinkauf, H. (1976) Biochim. Biophys. Acta 429, polypeptides distinct from the two described here remain to 1062-1072. be determined. 24. Zocher, R., Keller, U. & Kleinkauf, H. (1982) Biochemistry The techniques developed during the study of the biosyn- 21, 43-48. thesis of cyclic peptide antibiotics that have been used here 25. Zocher, R., Nihara, T., Paul, E., Madry, N., Peeters, H., could also be applied to the study ofthe biosynthesis ofother Kleinkauf, H. & Keller, U. (1986) Biochemistry 25, 550-553. peptidic phytotoxins with a known role in plant disease, such 26. Mohr, H. & Kleinkauf, H. (1978) Biochim. Biophys. Acta 526, as victorin, tentoxin, AM-toxin, phaseolotoxin, tabtoxin, and 375-386. PC-toxin (35). 27. Fr0yshov, 0. & Laland, S. G. (1974) Eur. J. Biochem. 46, 235-242. 28. Lee, S. G. & Lipmann, F. (1975) Methods Enzymol. 43, The initial stages ofthis work were done at the Plant Cell Research 585-602. Institute, Inc., Dublin, CA, in the laboratory of Dr. J. B. Mudd, 29. Keller, U. (1987) J. Biol. Chem. 262, 5852-5856. whose support is gratefully acknowledged. This work was funded in 30. Krause, M., Marahiel, M. A., von D1ihren, H. & Kleinkauf, part by the U.S. Department ofEnergy Division of Biological Energy H. (1985) J. Bacteriol. 162, 1120-1125. Research under Contract DE-AC02-76ER01338. 31. Marahiel, M. A., Krause, M. & Skarpeid, H.-J. (1985) Mol. Getz. Genet. 201, 231-236. 1. Yoder, 0. C. (1980) Annu. Rev. Phytopathol. 18, 103-129. 32. Leach, J., Tegtmeier, K. J., Daly, J. M. & Yoder, 0. C. (1982) 2. Alcorn, J. L. (1983) Mycotaxon 17, 1-86. Physiol. Plant Pathol. 21, 327-333. 3. Nelson, R. R. & Ullstrup, A. J. (1%1) Phytopathology 51, 1-2. 33. Wolpert, T. J., Macko, V., Acklin, W., Jaun, B., Seibl, J., 4. Kline, D. M. & Nelson, R. R. (1969) Phytopathology 59, Meili, J. & Arigoni, D. (1985) Experientia 41, 1524-1529. 1133-1135. 34. Dimroth, P., Walter, H. & Lynen, F. (1970) Eur. J. Biochem. 5. Scheffer, R. P., Nelson, R. R. & Ullstrup, A. J. (1967) Phyto- 13, 98-110. pathology 57, 1288-1291. 35. Durbin, R. D., ed. (1981) Toxins in Plant Disease (Academic, 6. Lim, S. M. & Hooker, A. L. (1971) Genetics 69, 115-117. New York). Downloaded by guest on September 25, 2021