This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Genetic Studies of Avermectin Biosynthesis inStreptomyces avermitilis
HARUO IKEDA, HIROMICHI KOTAKI, AND SATOSHI OMURA*
School of Pharmaceutical Sciences, Kitasato University, and The Kitasato Institute, Tokyo 108, Japan
Received 29 April 1987/Accepted 16 September 1987
A genetic recombination study of an industrial strain of Streptomyces avermitilis which produces avermectinis described. A genetic map has been constructed by analysis of haploid recombinants and linkage relationshipsof 16 marker loci. Fifteen avermectin-nonproducing mutants, produced by mutagenesis, were classified intotwo phenotypically different groups, of which one produced avermectin aglycon and the other was able toconvert avermectin aglycon to avermectins. Two different mutants were found to map closely to each other.
Avermectins (1) are a class of 16-membered macrolide(macrocyclic lactone) antibiotics (Fig. 1) with potentanthelminthic and insecticidal activities, which are producedby Streptomyces avermitilis (4). Macrolide antibiotics areformed through two main biosynthetic steps, first the forma-tion of an aglycon moiety, and second, the formation of asugar moiety and its addition to the aglycon. Many of thenonproducing mutants obtained from macrolide antibioticproducers can be divided genetically into two classes (3, 9,18). One class is unable to form the aglycon moiety, and thesecond is unable to form or to bind the sugar moiety. Geneticmapping of mutations of macrolide antibiotic producers hasbeen performed. Genes governing the biosynthesis of tylosinby Streptomyces fradiae are controlled by a self-trans-missible plasmid (2). On the other hand, genes governing thebiosynthesis of erythromycin by Streptomyces erythreus arelocated on the chromosome (18).Our interest was to determine the linkage map location of
the genes governing biosynthesis of the aglycon and thesugar (oleandrose) moieties of avermectin, using the pro-ducer S. avermitilis. In this paper, we describe the order onthe genetic map of 16 marker loci, including two differentloci which affect avermectin biosynthesis in S. avermitilis.
MATERIALS AND METHODS
Organisms. S. avermitilis K139, used in the present study,is a derivative of the original strain KA-325 isolated fromsoil. It produces about 650 ,ug of avermectins per ml. Itsauxotrophic and antibiotic nonproducing mutant strains usedin this study are listed in Table 1.Media and growth. Yeast extract-malt extract-soluble
starch medium (YMS) (H. Ikeda, H. Kotaki, H. Tanaka, andS. Omura, submitted for publication) was used as the com-plete medium and for sporulation. Minimum and avermectinproduction media were described by Hopwood and Sermonti(13) and Burg et al. (4), respectively. All strains were spreadonto YMS plates and incubated at 30°C for at least 5 days,during which spores matured. Then the spores were scraped,washed, suspended in sterile 20% (wt/vol) glycerol, andstored at -30°C.
Isolation of mutants. All mutants of S. avermitilis wereobtained by UV irradiation or by treatment with N-methyl-N'-nitro-N-nitrosoguanidine (NTG). Since photoreactiva-tion was observed, UV irradiation was performed in the
* Corresponding author.
dark. A mutant resistant to streptomycin at 1.25 ,ug/ml onminimal medium was obtained by UV irradiation. Treatmentwith NTG was as described by Delic et al. (7), except that 1mg of the mutagen per ml was used. After mutagenesis, thespores were spread onto YMS plates and incubated at 30°Cuntil spores were formed that expressed the mutation. Afterthe spores were scraped, they were stored as a suspension in20% (wt/vol) glycerol at -30°C. A portion of the sporesuspension was diluted and spread onto YMS medium tomake about 200 colonies per plate. Auxotrophic mutantswere characterized by replica plating. Avermectin-nonpro-ducing mutants were detected by assay of mycelial extractswith silica gel thin-layer chromatography.
Detection of avermectins. A well-separated colony on aYMS plate was inoculated directly into avermectin produc-tion medium. The fermentation and the extraction of prod-ucts were performed as described previously (Ikeda et al.,submitted). The fermentation products were separated bysilica gel thin-layer chromatography which was developedtwice with n-hexane-isopropanol (90:10). Avermectins weredetected by UV light (254 nm).
Bioconversion of avermectin aglycon to avermectins. Anavermectin-nonproducing mutant was cultured in the pro-duction medium. After 2 days, 14C-labeled aglycons of
HO-.
H.
Ala
Alb
A2a
A2b
Bia
Bib
B2a
B2b
Ri R2
C2HSCH3
OH C2H5
OH CH3o C2Hs
CH3
R3
CH3
CH3
CH3
CH3H
H
OH C2H5 H
OH CH3 H
FIG. 1. Structural formulae for the avermectins. Where Rl isabsent, the double bond ( -) is present. Both sugars arect-L-oleandrose.
TABLE 1. S. avermitilis strains usedStrain Markers
K297... aveAK299... aveAK329... aveBK1000.. lysAK1001..hisAK1002.. uraAK1003..athAK1004.. trpAK1005.. adeAK1011.. hisD pdxAK1012..hisD proAK1013..adeA uraAK1018.. serA argA strAK1019.. serA cysA strAK1020.. serA adeA strAK1021.. serA argA hisD strAK1022.. serA argA cysA strAK1030.. metB aveAK1031.. hisD aveAK1035..adeA aveBK1036.. hisD aveBK1037.. hisD strA aveB
solventfront
A2aaglycon
Aiaaglycon
AVM AlA2BiB2
Origin
avermectins Ala and A2a, which were prepared biosynthet-ically by feeding [1-'4C]acetate to culture K329, were addedand incubated for another 5 days. The conversion productswere analyzed by radioautography after separation by silicagel thin-layer chromatography as described above.
Cross. The cross procedures used were as described byHopwood and Sermonti (13), except that YMS was used asthe complete medium. A microcomputer program written inPASCAL for PC-9801VM2 (NEC Corp., Tokyo, Japan) wasused for data compilation and analysis of four-factor crosses.
RESULTS
Mutant strains. The marker loci used in mapping studieswith S. avermitilis are listed in Table 2. The concentrationsof NTG tested were from 0.5 to 2 mglml. Treatment with 1 or2 mg of NTG per ml proved efficient for the isolation ofmutants. However, many bald mutants were produced bytreatment with 2 mg of NTG per ml. All of the mutantsevaluated were prepared by treatment with 1 mg of NTG perml, giving approximately a 95% kill, except that the strepto-mycin-resistant mutant was obtained by UV irradiation.
_- N c U)
FIG. 2. Autoradiograms of extracts from mycelia incubated with['4C]avermectin (AVM) aglycons. Silica gel thin-layer chromatogra-phy was developed twice with hexane-isopropanol (90:10). TheX-ray film was exposed for 2 weeks. Lanes: 1, ['4C]avermectins; 2,["4C]aglycons of avermectins Ala and A2a; 3, strain K297; 4, strainK299; 5, strain K582.
Auxotrophic mutants were isolated at a frequency of about3.8 x 10-2 per surviving CFU.Mutants affecting avermectin biosynthesis. Avermectin-
nonproducing phenotypes AveA and AveB were evaluated.Frequencies of appearance of AveA and AveB after treat-ment with NTG were 8.16 x 10-3 (7/858) and 9.32 x 10-3(8/858), respectively, among randomly chosen survivingcolonies. These avermectin-nonproducing mutants werecharacterized.The aveB mutant was found to produce two compounds
with UV absorptions similar to those of avermectins. TheaveA mutant produced no such compounds. From the re-sults of several analyses of the compounds produced by the
TABLE 2. Marker loci usedMutant allele Characteristic
adeA. Requirement for purinesargA. Requirement for arginine, citrulline, or ornithineathA. Requirement for purines plus thiamineaveA. Inability to produce avermectinsaveB. Inability to produce avermectinscysA. Requirement for cysteinehisA. Requirement for histidinehisD. Requirement for histidine or histidinollysA. Requirement for lysinemetB. Requirement for methionine or homocysteinepdxA. Requirement for pyridoxineproA. Requirement for prolineserA. Requirement for serine or glycinestrA. Resistant to streptomycintrpA. Requirement for tryptophan or indoleuraA. Requirement for uracil
TABLE 3. Analysis of a four-factor cross of strains K1018 (serAargA strA) and K1002 (uraA)a
No. of colonies' on selective mediaGenotypes of supplemented with': Avg
selectable progeny Ura, Ser Ser, Arg frequencyArg, Str Ura, Str Ar
+ + + str 64 (156) 44 (98) 68 (172) 69 (161) 147ura + + str 23 (56) 20 (51) 53+ ser + str 13 (29) 12 (30) 30+ + arg str 5 (12) 9 (21) 17+ + + + 25 (56) 22 (51) 54ura ser + str 0 (0) 0+ + arg + 0 (0)ura + arg str 3 (7) 17+ ser + + 12 (27)a See Table 4.b Number of colonies in sample; parentheses indicate total recombinants
per plate.c Ura, Uracil; Arg, arginine; Str, streptomycin; ser, serine.
FIG. 3. Proposed biosynthetic pathway of avermectins. Where Rl is absent, the double bond ( -) is present. AVM, Avermectin.
aveB mutant, performed by nuclear magnetic resonance,mass spectrum, and elemental analyses, the compoundswere determined to be aglycons of avermectins Ala and A2a(14, 17) (data not shown). Consequently, the aveB markerindicated an effect on oleandrose biosynthesis or addition ofthe sugar to the avermectin aglycon. Since all of theavermectins produced were accumulated in the mycelium,the method of cosynthesis in vivo could not be applied to S.avermitilis.The characterization of aveA was performed by examining
the conversion of the precursor (avermectin aglycon) toavermectins. The aveA mutant possessed the ability toconvert the aglycons of avermectins Ala and A2a toavermectins Ala and A2a. The conversion rate, however,was extremely low. The mutant was not able to convertaglycons of avermectins Ala and A2a to avermectin Bcomponents. The above data indicate that the aveA markereffects avermectin aglycon biosynthesis. From the aboveresults and the previous report that the aglycon was formedfrom acetate, propionate, and a branched amino acid, asdescribed by Cane et al. (5), we formulated a proposedbiosynthetic pathway of avermectin (Fig. 3).
Cross analysis. Crosses were made between several dif-ferent combinations of the strains listed in Table 1. Recom-bination frequencies ranging from 3 x 10-5 to 6 x 10-7 were
observed, and polarities were not detected in any crosses. Inthe beginning, four-factor crosses were analyzed by themethod recommended by Hopwood (11). An initial assump-tion was made that the genome of S. avermitilis was circularor circularly permuted. Haploid recombinants were isolatedfrom four different media which did not permit the growth ofeither parent, and their phenotypes were determined onappropriate diagnostic media. In a typical cross betweenstrains K1018 (serA argA strA) and K1002 (uraA), sporesfrom the mixed culture, when plated on selective and diag-nostic media, gave the results shown in Tables 3 and 4.Chi-square analysis and the number of crossovers obtainedindicated a dependent segregation, and therefore adjacentpositions, for the unselected markers on each medium. Thedata indicated that alleles serA uraA and argA strA were
adjacent.Similar analysis was applied to crosses between strains
K1018 (serA argA strA) and K1001 (hisA) (Tables 5 and 6)and between strains K1018 (serA argA strA) and K1000
TABLE 4. Segregation of pairs of nonselected alleles'
a See Table 3. Relative recombination frequency in each interval: ser-arg, 64; ser-str, 234; ser-ura, 218; arg-ura, 248; ura-str, 124; arg-str, 230.b Quadruple crossover.
TABLE 5. Analysis of a four-factor cross of strains K1018 (serAargA strA) and K1001 (hisA)'
No. of coloniesb on selective mediaGenotypes of supplemented with': Avg
selectable progeny His, Ser Ser frequencyArg, Str His, Str Ar
+ + + str 3 (7) 4 (9) 6 (16) 41 (8) 10his + + str 71 (156) 74 (193) 175+ ser + str 57 (133) 20 (52) 93+ + arg str 2 (4) 14 (3) 4+ + + + 8 (19) 45 (9) 14his ser + str 0 (0) 0+ + arg + 0 (0)his + arg str 24 (53) 62+ ser + + 30 (70)a See Table 6.b Number of colonies in sample; parentheses indicate total recombinants
per plate.c His, Histidine; Arg, arginine; Str, streptomycin; Ser, serine.
(lysA) (Tables 7 and 8). The results of representative crossesshown in Tables 3 through 8 indicate a circular linkage maphaving the six markers in the sequence shown in Fig. 4.Further marker loci were also determined by the strategydescribed by Hopwood (10). For example, the adeA allelewas mapped by the data from crosses between K1021 (serAargA hisD strA) and K1005 (adeA) (Fig. 5). Two possiblepositions for adeA were indicated by the allele frequencies of
the markers. The data in Table 9 show that hypothesis I inFig. 5 should be chosen because of the lowest number ofquadruple crossovers. Furthermore, other marker loci (seeFig. 8) were determined by the same procedures as de-scribed above or by four-factor cross.
Mapping of loci affecting avermectin biosynthesis. Themarker aveA designates a lesion which blocks the formationof the aglycon moiety. The results of a cross between strainsK1021 (serA argA hisD strA) and K1030 (metB aveA) (Fig. 6)indicate two possible positions for aveA. One of these(hypothesis I) was chosen on the basis of the lowest numberof quadruple crossovers (Table 10). Further, the markeraveB, which blocks the formation of oleandrose or theaddition of oleandrose to the aglycon, was analyzed by a
cross between strains K1022 (serA argA cysA strA) andK1035 (adeA aveB) (Fig. 7). Hypothesis I in Fig. 7 was
adopted on the basis of the lowest number of quadruplecrossovers (Table 11). The chromosomal locations of aveAand aveB were consistent in the serA metB and the cysAadeA intervals, respectively. Four aveA-aveB crosses were
carried out to determine their locations. Recombinants car-
rying the aveA and aveB alleles produced no detectable ave+recombinants among randomly chosen samples of selectedprototrophic or prototrophic Strr progeny (Table 12). There-
fore, aveA and aveB are likely to be close to each other on
the chromosome.Linkage map of S. avermitilis. Marker sequences were
established by crosses, in various combinations (more than
TABLE 6. Segregation of pairs of nonselected alleles'
arg his- his+ Segregation str ser- ser+ Segregation his ser ser+ Segregation str arg- arg+ Segregation
arg- 24 2 str- 57 4 his 0b 74 str- 14 41Not adjacent Not adjacent Adjacent Adjacent
a See Table 5. Relative recombination frequency in each interval: ser-arg, 159; ser-str, 251; ser-ura, 28; arg-uira, 179; ura-str, 251; arg-str, 278.b Quadruple crossover.
TABLE 7. Analysis of a four-factor cross of strains K1018 (serA argA strA) and K1000 (lysA)"
Genotypes of No. of coloniesb on selective media supplemented with": Avgselectablefrqecprogeny Lys Arg Str Ser Ser Lys Str Arg frequency
+ + + str 4 (11) 19 (28) 18 (25) 29 (31) 24lys + + str 87 (249) 69 (97) 173+ ser + str 14 (20) 13 (18) 19+ + arg str 1(3) 7 (7) 5+ + + + 36(52) 64(67) 60lys ser + str 0 (0) 0+ + arg + 0 (0)lys + arg str 8 (23) 34+ ser + + 31(45)a See Table 8.b Number of colonies in sample; parentheses indicate total recombinants per plate.Lys, Lysine; Arg, arginine; Str, streptomycin; Ser, serine.
a See Table 7. Relative recombination frequency in each interval: ser-arg, 58; ser-str, 236; ser-Ihs. 89; arg-lys, 137; Iys-str, 267; arg-str, 216.b Quadruple crossover.
FIG. 4. Preliminary linkage map of S. avermitilis for six markers.Marker positions were determined by data from four-factor crosses.
20), among the strains listed in Table 1. All of the mappingdata accumulated to date on S. avermitilis are summarized inthe linkage map shown in Fig. 8. The placement of markerloci' on the map does not reflect actual map distance,although some' markers, e.g., hisA and serA, appeared to beclosely linked. In contrast, serA and strA appeared to bequite distant because the two lesions were simultaneouslycorrected in crosses only in rare instances. The positions ofaveA and aveB were located in the serA metB and cysA adeAintervals, respectively, and aveA and aveB were very closeto each other. There is no direct evidence for the linkage ofaveB and uraA; however, from the results of K1021 (serAargA hisD strA) x K1035 (adeA aveB) and K1021 (serA argAhisD strA) x K1013 (adeA uraA) crosses (data not shown),the distance of the aveB adeA interval would be longer thanthat of the uraA adeA interval.
DISCUSSIONWe have constructed the first simple genetic linkage map
of S. avermitilis, which includes loci of representative mu-tations affecting the biosynthesis of avermectin. The systemof genetic recombination observed in S. avermitilis closely
004c
--
100
TABLE 9. Location of adeA locus by least-quadruple crossoveranalysisa
Crossover intervals forNo. of position:Genotype recombinants
a Results derived from a cross between strains K1021 (serA argA hisD strA)and K1005 (adeA).
resembles that described for Streptomyces coelicolor A3(2)by Hopwood and Sermonti (13). The polarity described byCoats and Roeser (6) in Streptomyces bikiniensis subsp.zorbonensis was not observed in our different crosses, andthe recombination associated with the fertility characteristic,described by Hopwood et al. (12) for S. coelicolor A3(2),also was not detected in S. avermitilis.
It is interesting that the series of 14 markers on the S.avermitilis linkage map (Fig. 8) has an exact counterpart inS. coelicolor A3(2): hisA lysA proA cysA uraA athA adeAmetB strA hisD pdxA argA serA trpA. This provides afurther example of the conservation of linkage relationshipsin streptomycetes first noted in a comparison of the maps ofS. coelicolor A3(2) and Streptomyces rimosus (8).Although two different types of ave mutations, Qf which
one influences the formation of the aglycon and the secondinfluences the formation of oleandrose or the addition of thissugar to the aglycon, were characterized by analysis of theproducts and of the bioconversion reaction, at least twoadditional steps, methylation of the hydroxyl group at theC-5 position and dehydration at the C-22 and C-23 positions,are required for avermectin biosynthesis. The sequences ofthe biosynthetic steps encoded O,y aveA and aveB wereconfirmed by bioconversion study. The methylation of thehydroxyl group at C-5 and the dehydration at C-22,23 couldtake place either before or after the formation of the aglycon.Since the aglycons of avermectins Ala and A2a were con-verted to avermectins Ala and A2a without demethylation ofthe C-5 methoxyl group of the aglycon moiety, and since theavermectin B O-methyltransferase which catalyzes the
0
51
-C41100100
Hypothesis I Hypothesis x
FIG. 5. Analysis of the data in Table 9. The numbers next to themarkers on the diagramns are percentage frequencies of alleles; thenumbers between the circles designate map intervals. The trianglesindicate selected alleles. Hypothesis I is chosen.
+51
-
100
Hypothesis I Hypothesis It
FIG. 6. Analysis of the data in Table 10. The triangles indicateselected alleles. Hypothesis I is chosen.
transfer of the methyl of S-adenosyl-L-methionine to the C-5hydroxyl group utilizes avermectin B aglycon as a substraterather than avermectin B (M. D. Schulman, D. Valentino,and C. Ruby, Fed. Proc. 44:931, 1985), the methylation musttake place before the addition of oleandrose to the aglycon.The aveB mutant produced avermectin Ala aglycon, whichis the dehydrated product of A2a aglycon. Because the
enzyme involved in the dehydration at C-22,23 has not beendetected yet, it is not possible to state whether the reactioninvolved in dehydration at C-22,23 takes place before orafter the formation of the aglycon.Assuming random distribution of ave mutations, gene size
will be related to observed frequency of mutation. Becausethe frequencies of aveA and aveB mutations were compara-ble, it is assumed that the respective gene sizes are similar.Interestingly, the results for the avermectin producer aresignificantly different from those for other macrolide antibi-otic producers. Mutants analogous to those reported herehave been obtained in studies of S. fradiae (3) and Strepto-myces platensis subsp. malvinus (9), which were blocked inproduction of the 16-membered macrolide antibiotics tylosinand platenomycin, respectively. A majority of the antibiotic-nonproducing mutants of S. fradiae and of S. platensissubsp. malvinus were blocked in the formation of theaglycon. On the other hand, the frequencies of appearance ofmutants blocked in the formation of the aglycon (aveA) andmutants blocked in the formation of sugar moiety or of theaddition of the sugar to aglycon (aveB) in S. avermitilis weresimilar. This suggests that the biosynthesis of the aglyconmoiety of avermectin may be genetically and biochemicallydifferent from those of 16-membered macrolide antibiotics ingeneral, including tylosin and platenomycin.The fact that the ave genes appear to reside close together
on the linkage map is not surprising since other antibioticbiosynthesis genes on linkage maps appear as single clusters
on chromosomes (15, 16, 18). Perhaps other expected muta-tions for methylation and dehydration steps will be found inthis cluster. In the future, it is possible that the entire set ofgenes required for avermectin biosynthesis will be isolatedon one DNA fragment by genetic manipulation, since thegenes appear to form a cluster on the chromosome.
ACKNOWLEDGMENTS
We thank H. B. Woodruff for critical reading of the manuscript.We also thank M. Uchida for isolation of some auxotrophic mutants.
LITERATURE CITED1. Albers-Schonberg, G., B. H. Arison, J. C. Chabala, A. W.
Douglas, P. Eskola, M. H. Fisher, A. Lusi, H. Mrozik, J. L.Smith, and R. L. Tolman. 1981. Avermectins: structure deter-mination. J. Am. Chem. Soc. 103:4216-4221.
2. Baltz, R. H. 1980. Genetic recombination by protoplast fusion inStreptomyces. Dev. Ind. Microbiol. 21:43-54.
3. Baltz, R. H., and E. T. Seno. 1981. Properties of Streptomycesfradiae mutants blocked in biosynthesis of the macrolide anti-biotic tylosin. Antimicrob. Agents Chemother. 20:214-225.
4. Burg, R. W., B. M. Miller, E. E. Baker, J. Birnbaum, S. A.Currie, R. Hartman, Y. Kong, R. L. Monaghan, G. Olson, I.
Putter, J. B. Tunac, H. Wallick, E. 0. Stapley, R. Oiwa, and S.Omura. 1979. Avermectins, new family of potent anthelminthicagents: producing organism and fermentation. Antimicrob.Agents Chemother. 15:361-367.
5. Cane, D. E., T.-C. Liang, L. Kaplan, M. K. Nallin, M. D.Schulman, 0. D. Hensens, A. W. Douglas, and G. Albert-Schonberg. 1983. Biosynthetic origin of the carbon skeleton andoxygen atoms of the avermectins. J. Am. Chem. Soc. 105:4110-4112.
6. Coats, J. H., and J. Roeser. 1971. Genetic recombination in
Streptomyces bikiniensis var. zorbonensis. J. Bacteriol.105:880-885.
7. Delic, V., D. A. Hopwood, and E. J. Friend. 1970. Mutagenesisby N-methyl-N'-nitro-N-nitrosoguanidine (NTG) in Streptomy-ces coelicolor. Mutat. Res. 9:167-182.
8. Friend, E. J., and D. A. Hopwood. 1971. The linkage map ofStreptomyces rimosus. J. Gen. Microbiol. 68:187-197.
9. Furumai, T., and M. Suzuki. 1975. Studies on the biosynthesisof basic 16-membered macrolide antibiotic, platenomycins. I.Selection of and cosynthesis by non-platenomycin-producingmutants. J. Antibiot. 28:770-774.
10. Hopwood, D. A. 1967. Genetic analysis and genome structure inStreptomyces coelicolor. Bacteriol. Rev. 31:373-403.
11. Hopwood, D. A. 1972. Genetic analysis in microorganisms.Methods Microbiol. 7B:363-433.
12. Hopwood, D. A., R. J. Harold, A. Vivian, and H. M. Ferguson.1969. A new kind of fertility variant in Streptomyces coelicolor.Genetics 62:461-477.
13. Hopwood, D. A., and G. Sermonti. 1962. The genetics ofStreptomyces coelicolor. Adv. Genet. 11:273-342.
14. Mrozik, H., P. Eskola, B. H. Arison, G. Albers-Schonberg, andM. Fisher. 1982. Avermectin aglycons. J. Org. Chem. 47:489-492.
15. Rudd, B. A. M., and D. A. Hopwood. 1979. Genetics ofactinorhodin biosynthesis by Streptomyces coelicolor A3(2). J.Gen. Microbiol. 114:35-43.
16. Rudd, B. A. M., and D. A. Hopwood. 1980. A pigmentedmycelial antibiotic in Streptomyces coelicolor: control by achromosomal gene cluster. J. Gen. Microbiol. 119:333-340.
17. Schulman, M. D., D. Valentino, 0. D. Hensens, D. Zink, M.Nallin, L. Kaplan, and D. A. Ostlind. 1985. Demethyl-avermectins: biosynthesis, isolation and characterization. J.Antibiot. 38:1494-1498.
18. Weber, J. M., C. K. Wierman, and R. Hutchinson. 1985. Geneticanalysis of erythromycin production in Streptomyces erythreus.J. Bacteriol. 164:425-433.