We prepared hybrids between 14C-labeled ribosomal ribonucleic acid (rRNA) from either Agrobacterium tumefaciens ICPB TTlll or A . rhizogenes ICPB TR7, and deoxyribonucleic acid (DNA) from a great variety of reference gramnegative and gram-positive bacteria. Each hybrid was described by (i) its TMe,, the temperature at which 50% of the hybrid was denatured, and (ii) percentage of rRNA binding, i.e., micrograms of 14C-labeled rRNA duplexed per 100 pg of filter-fixed DNA. Each taxon occupied a definite area on the rRNA similarity map. The size and shape of this area depended on the phenotypic and genetic heterogeneity of the taxon. There appeared to be a correlation between T M e , of the heterologous hybrids and the overall phenotypic similarities of the organisms and taxa involved. T M e , values above 65°C were taxonomically most meaningful. DNA:rRNA hybridizations condensed all strains from a genus in one narrow cluster; the method had little resolution to distinguish species within a genus, but it seemed to be a very useful approach to detect remote relationships at the inter-and suprageneric level, for taxonomic and identification purposes. The hybrid parameters of Azotomonas fluorescens, Mycoplana bullata, Mycoplana dimorpha, Phyllobacterium, two misnamed "Chromobacterium liuidum" strains from leaf-nodulating plants, two misnamed agrobacteria from the Baltic Sea, and a few misnamed "Achromobacter" strains were all in the vicinity of Agrobacterium and Rhizobium. We suggest that all of these organisms are remote relatives and belong in the family of the Rhizobiaceae. Azotomonas insolita NCIB 9749 is misnamed; it is an Agrobacterium. Several organisms which had been misnamed Agro bacterium formed DNA:rRNA hybrids with properties outside the Agrobacterium area.

We describe and recommend the following improvements of DNA:rRNA membrane filter hybridization methods. One of our aims was to avoid DNA release from filter discs during hybridization. 1. Our hybridization conditions are 2 SSC in aq. dest., with 20% formamide, 50 C, overnight for 16 hr. 2. Duplexing is over in 8-10 hr. 3. Formamide has to be very pure (O.D. less than or equal to 0.2/cm light path at 270 nm). 4. RNAase treatment: 250 mug/5 ml 2 SSC/filter at 37 C for 1 hr. 5. Our conditions for stepwise thermal denaturation are: 5 degrees C steps from 50 C to 90 C in 1.5 SSC in 20% formamide. 6. Single-stranded DNA, fixed on membrane filters, and stored in vacuo at 4C can be used reliably for hybridization for up to 20 months. 7. Concentrated DNA in 0.1 SSC, quick-frozen at -50 C and stored at -90 C for up to 2 years can be used for hybridization without much change. 8. A CsCl gradient purification step yields much purer DNA, but increases the release of DNA from filters by about 20%. Filters with 20% more DAN is a compensation. 9. rRNA can be stored for 20 months in SSC or 2 SSC at -12 C without changing the hybridization results.

'*C-labeled ribosomal ribonucleic acid (rRNA) was prepared from Azoto bacter chroococcum NCIB 8002, Azotobacter paspali 8A, Azomonas agilis NCIB 8636, Azomonas insignis WR 30, Beuerinckia indica NCIB 8712, and Azospirillum brasilense ATCC 29145. These rRNA's were hybridized under stringent conditions with filter-fixed deoxyribonucleic acid from a great variety of gram-negative bacteria. Each hybrid was described by: (i) the temperature at which 50% of the hybrid was denatured, and (ii) the percent rRNA binding (amount in micrograms of rRNA duplexed to 100 pg of deoxyribonucleic acid). These data were used to construct rRNA similarity maps. The following conclusions could be drawn concerning rRNA cistron similarities. (i) Bacterial genera with free-living, aerobic, nitrogen-fixing members are very diverse and belong to different rRNA superfamilies. The present family Azotobacteriaceae is not a biological unit, and its status as a family is highly questionable. (ii) Azotobacter chroococcum, Azoto bacter vinelandii, Azoto bacter beijerinckii, Azoto bacter paspali, Azoto bacter miscellum, Azotobacter armeniae, and Azotobacter nigricans belong in the genus Azotobacter. Any synonymy of these names remains to be determined. Azomonas agilis, Azomonas insignis, and Azomonas macrocytogenes constitute independent branches, which are about equidistant from Azotobacter and section I of Pseudomonas as presented in Bergey 's Manual of Determinative Bacteriology, 8th ed. Xanthomonas, Alteromonas vaga, and Alteromonas communis are located in the same rRNA superfamily. (iii) The genus Beijerinckia appears to be rather heterogeneous. Its closest relatives appear to be Xantho bacter autotrophicus, "Mycobacterium" flavum, "Pseudomonas" azotocolligans, "Pseudomonas" diminuta, the authentic rhodopseudomonads, and some other organisms. These organisms belong in the same rRNA superfamily as Azospirillum, Agrobacterium, Rhizo bium, Aceto bacter, Glucono bacter, and Zymomonas. (iv) Derxia belongs in still another rRNA superfamily , together with Chromobacterium, Janthinobacterium, the Pseudomonas acidovorans and Pseudomonas solanacearum groups, Alcalienes, and a few other taxa. (v) The following organisms were generically misnamed: "Azomonas insignis" ATCC 12523, "Mycobacterium" flavum 301, "Pseudomonas" azotocolligans ATCC 12417, "Pseudomonas" diminuta CCEB 513, and "Rhodopseudomonas" gelatinosa (all strains examined).Molecular biological methods, such as deoxyribonucleic acid (DNA)-DNA or DNA-ribosomal ribonucleic acid (rRNA) hybridizations, which directly compare bacterial genomes, have opened new perspectives for bacterial classification. Many bacterial genera are phylogenetic d y too far removed from each other to form stable DNA-DNA hybrids. DNA-DNA hybridizations are useful either within a genus, such as Agrobacterium (ZO), or between genera which have not diverged too much, such as in the Enterobacteriaceae (12; D. Izard, C. Ferragut, and H. Leclerc, in press). rRNA's are conservative molecules (25, 36, 43). There is a good corre...

S U M M A R YMolecular hybrids were prepared between unlabelled DNA from representative strains of eleven genetic races of Agrobacterium and [I4C]DNA from typical strains of each of the three main races. The thermal stability of each hybrid was determined. The nature of the hybrids formed varied with the incubation temperature and the kind of DNA used. Hybridization in 2 x SSC-30 % dimethylsulphoxide below 59 "C yielded two kinds of hybrids: a labile one of unknown nature, denaturing below 59 "C, and a more or less stable hybrid denaturing above that temperature. The latter was the only one formed in hybridizations at or above 59 "C. There were three kinds of stable hybrids. Within each of the main Agrobacterium races thermal stability of the molecular hybrid was about the same (within 2 "C) as for the homoduplex. Between two races of 50 % DNA relatedness, the duplexes were about 6 "C less stable. Between races of 10 to I 5 % DNA relatedness, the duplexes were weak, and the stability was at least 13 "C lower. The stability of the hybrids decreased concomitantly with the degree of DNA relatedness. The decreased hybrid denaturation curve is not due to AT-rich sequences. The less two races of agrobacteria appeared to be evolutionarily related, the more mutations occurred within the common part. I N T R O D U C T I O NFrom previous studies it appeared that the genus Agrobacterium is genetically very heterogeneous. DNA of cluster I (typical Agrobacterium tumefaciens and A. radiobacter strains), cluster 2 ( A . rhizogenes and atypical A . tumefaciens strains), the 'rubi' group and two very small groups hybridize at about 10 to 15 %. Within cluster I the DNA of the seven groups hybridize at about 50 % and within each group at least 80 % ( Mutational events modified considerable parts of the genomes, preventing in vitro molecular hybridization. Considering the evolutionary history of a bacterial genus, one can wonder whether mutations also occurred within the common DNA parts. For an experimental answer to that question we prepared a number of DNA-hybrids between the genomes of different Agrobacterium groups and clusters, and determined their thermal stability which is interpreted as a measure of base-pairing imperfections, and mutational differences (Brenner & Cowie, 1967). We established that decreased hybrid denaturation curves were not due to preferential binding of heterologous AT-rich DNA sequences.

Volume 27, no. 3, p. 226, Table 1-Continued: Footnote b should read "The G+C values for 14 Arthrobacter strains were taken from Skyring et al. (64). All others were determined in our laboratory on previous occasions by thermal denaturation." Serratia proteamacuZans (Paine and Stansfield) comb. nov., a Senior Subjective Synonym of Serratia Ziquefaciens (Grimes and Hennerty) Bascomb et al.