INTRODUCTION

Top Abstract Introduction Materials and Methods Results and Discussion References

Five named species and three unnamed genomospecies were recently added to the genus Citrobacter on the basis of DNA relatedness and biochemical studies (2). Based on this new classification, Janda et al. (8) identified 235 Citrobacter strains from the collection of the Microbial Diseases Laboratory, California Department of Health Services (CDHS), Berkeley. Additional strains of unnamed Citrobacter genomospecies 9 were subsequently identified and studied, resulting in this species being named Citrobacter rodentium (13).

The study reported herein had two goals. We sought to determine the extent to which the biochemical identification of the new, as well as the traditional, species Citrobacter freundii, Citrobacter koseri (Citrobacter malonaticus), and Citrobacter amalonaticus correlated with molecular identification. We also sought to obtain additional strains of the two remaining unnamed Citrobacter genomospecies and to formally name them. For these purposes, 136 Citrobacter strains were identified to species level by DNA hybridization, and the data were compared with the biochemical identifications of these strains. Additional strains of each of the two remaining unnamed Citrobacter genomospecies were identified, resulting in their formal description as Citrobacter gillenii sp. nov. (formerly Citrobacter genomospecies 10) and Citrobacter murliniae sp. nov. (formerly Citrobacter genomospecies 11).

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results and Discussion References

Strains. Included in the study were 136 strains that, prior to the creation of additional Citrobacter species (2), had been identified as C. freundii, C. koseri, and C. amalonaticus on the basis of biochemical reactions. Eighty-six strains were sent to the diagnostic laboratories at the Centers for Disease Control and Prevention (CDC) from 1972 to 1986; 26 of these strains were from the Microbial Diseases Laboratory of the CDHS, having been received from 1970 to 1994; 13 strains were from the Culture Collection, University of Göteborg (CCUG), Göteborg, Sweden; eight strains were from the National Institute of Public Health (NIPH), Prague, Czechoslovakia; and three strains were from the Division of Toxicology and Division of Comparative Medicine, Massachusetts Institute of Technology (MIT), Cambridge, Mass.

Biochemical tests. The methods used for the conventional biochemical tests done at the CDC and for the carbon source utilization tests (using Biotype strips [BioMérieux, La Balme les Grottes, France]) carried out at the Institut Pasteur have been described previously (2, 6, 7). In the present study, Biotype-100 strips were incubated at 30°C for 4 days. The CDC strains were identified both by conventional biochemical tests done at the CDC and by carbon source utilization tests carried out at the Institut Pasteur. The NIPH, CCUG, MIT, and the CDHS strains were also identified by conventional biochemical tests at CDC. The CDHS strains were also identified by conventional biochemical tests at the CDHS (8). In all cases, identification on the basis of either conventional biochemical tests or carbon source utilization tests was made by using the differential tables previously published by Brenner et al. (2).

DNA hybridization. The methods used for DNA extraction and purification and the hydroxyapatite hybridization method for determining levels of DNA relatedness have been described previously (2, 4). Reactions were carried out at 60°C (for optimal DNA reassociation) and at 75°C (for stringent DNA reassociation). Percent divergence was determined by thermal elution of reassociated DNA on the assumption that each 1°C decrease in thermal stability within a reassociated DNA duplex was due to approximately 1% of nucleotide bases within that sequence being unpaired. Percent divergence was calculated to the nearest 0.5%. All DNA relatedness reactions were carried out at least twice.

	RESULTS AND DISCUSSION

Top Abstract Introduction Materials and Methods Results and Discussion References

The standard used to identify strains was the genetic definition of a species on the basis of DNA relatedness, as recommended by Wayne et al. (14). This recommendation states that DNAs from strains of a given species are at least 70% related at optimal conditions for DNA reassociation (60°C incubation temperature in this study) and that divergence (unpaired bases) within related nucleotide sequences is 5% or less. We used the additional criterion that DNA relatedness of strains within a species remains above 60% in reactions carried out under stringent incubation conditions (75°C incubation temperature in this study). A strain was assigned to a given species when the relatedness of its DNA to labeled DNA from the type strain of that species fulfilled the species definition. The DNA relatedness data are summarized in Table 1. Relatedness values obtained with 131 of the 136 strains fully conformed to the molecular definition of a species. Of the five exceptions, two were closest to C. amalonaticus, two were closest to Citrobacter braakii, and one was closest to Citrobacter youngae. They each fulfilled two of the three criteria (percent divergence of less than 5 and relatedness of above 60% in 75°C reactions), but their relatedness in 60°C reactions was slightly under 70%. While it is possible that these five exceptions represent one or more new species, we decided that they were close enough to the species definition to merit provisional assignment to the species to which they were most closely related.

All 136 strains were identified by conventional biochemical tests done at CDC. The 87 strains from the CDC collection were also identified by carbon source utilization tests. Personnel carrying out identification by any method were blinded to the results obtained with other methods.

One hundred nineteen of 136 strains (88%) were correctly identified on the basis of their biochemical profiles. Fourteen strains (10%) were misidentified as other species in the genus Citrobacter, and three strains (2%) were correctly identified as Citrobacter but were not identified to species level. Seven of the misidentified strains and one of the nonidentified strains were biochemically atypical C. freundii. The atypical characteristics most often seen were ornithine decarboxylase production (5 strains); indole production (3 strains); negative growth on citrate, negative fermentation of raffinose, and nonmotility (2 strains each); malonate utilization and fermentation of i-inositol (1 strain each); and negative fermentation of melibiose, negative production of H₂S, and negative gas production from D-glucose (1 strain each).

These results are encouraging, indicating that despite changes in reaction percentages and the inclusion of a large number of atypical strains, the large majority of Citrobacter strains can be identified phenotypically by using conventional biochemical tests. O'Hara et al. (11) investigated the abilities of five commercial identification systems to recognize the newly defined species of Citrobacter by using the 112 strains identified by DNA hybridization in reference 1. Because the eight newly defined species were not included in the databases of any of these systems, most of these strains were identified as C. freundii.

The Vitek GNI+ card (BioMérieux Inc., Hazelwood, Mo.), which contains all 11 named and unnamed Citrobacter species, groups seven species which can be identified by using six additional conventional biochemical tests into the "Citrobacter freundii complex." When evaluated by O'Hara et al. (12), 12 of 16 strains in this complex were correctly identified by using the additional tests. The database of the MicroScan Rapid Neg ID3 panel (Dade Behring, Inc., West Sacramento, Calif.) contains eight of the nine named Citrobacter species (C. rodentium is not included). When evaluated by Bascomb et al. (1), 14 of 15 strains of the newer species were correctly identified.

Revised biochemical test percentages for all Citrobacter species are presented in Table 2 and are compared with the biochemical table presented previously (2). It should be noted that many of the percentages presented in the previous paper were changed when the data were recalculated (Table 2, columns P). Most of the changes are small, but nonetheless, the Hospital Environment Laboratory Branch of the CDC considers these results to be a correction of and a replacement for those presented previously (2). While the addition of strains in the present study did not cause most reaction percentages to change significantly, there are a number of percentage shifts worthy of mention. This is especially true for C. freundii, whose biochemical profile originally included strains of the 10 subsequently recognized Citrobacter species (5). The significant percentage shifts include H₂S production (formerly 96% and now 63%), sucrose fermentation (formerly 18% and now 78%), dulcitol fermentation (formerly 71% and now 24%), and raffinose fermentation (formerly 18% and now 88%). In each case, the actual percentage is probably closer to the former value, since that value is based on a 10-fold-higher number of strains (albeit including many that are not C. freundii) and a much higher percentage of typical strains.

The percentages obtained from the biochemical tests in this study revealed a few small, but significant, changes when compared to the percentages obtained in the recent previous study (2). This is not surprising given the small number of representative strains for most species. For C. braakii, delayed growth on citrate and delayed utilization of sodium acetate, and production of ornithine, and gas from glucose changed from positive to variable, and fermentation of glycerol changed from variable to positive. For Citrobacter werkmanii, production of arginine changed from rapid (within 48 h) positive to delayed positive. For C. rodentium, production of urease changed from positive to delayed positive, and motility changed from negative to delayed variable. For Citrobacter genomospecies 10, fermentation of salicin, raffinose, and esculin changed from negative to variable. For Citrobacter genomospecies 11, growth on citrate changed from positive to delayed positive, production of urease and arginine changed from delayed positive to variable, and fermentation of melibiose changed from delayed positive to delayed variable.

By using the species profiles generated in our previous study, 65 of the 86 strains (76%) tested by carbon source utilization were correctly identified. Fifteen strains (17%) were not identified to species level, and six (7%) were misidentified as other species in the genus Citrobacter. After adjustments were made in the identification program for carbon sources, all strains were retested and reidentified. Upon retesting, 84 of the strains (98%) were correctly identified and two (2%) were misidentified as other Citrobacter species. On the basis of carbon source utilization patterns, C. freundii strains formed seven clusters (biotypes) and C. braakii strains formed two clusters. Profiles for the biotypes and revised carbon source utilization patterns for all species are shown in Table 3. The C. braakii biotypes were separable by analysis of their abilities to utilize 4-aminobutyrate, lactose, D-lyxose, maltitol, 1-O-methyl--galactoside, 3-phenylpropionate, and propionate. C. freundii biotypes were separable on the basis of their profiles for use of the following carbon sources: trans-aconitate, 4-aminobutyrate, 5-aminovalerate, dulcitol, ethanolamine, 1-glutamate, myo-inositol, maltitol, 3-O-methyl-D-glucose, 1-O-methyl--D-glucoside, palatinose, 3-phenylpropionate, propionate, putrescine, sucrose, meso-tartrate, and D-turanose.

Additional strains of each unnamed Citrobacter genomospecies were identified on the basis of DNA relatedness and biochemical studies. Three additional strains confirmed to be Citrobacter genomospecies 9 were obtained from David B. Schauer. As with the three initially reported strains of this species, the additional strains were all from mice and were causative agents of transmissible murine colonic hyperplasia (13). The six strains of Citrobacter genomospecies 9 were recently described as the new species C. rodentium (13).

Eleven additional strains of Citrobacter genomospecies 10 were identified, for a total of 14, and seven additional strains of Citrobacter genomospecies 11 were identified, for a total of 10. The names "Citrobacter gillenii" sp. nov. and "Citrobacter murliniae" sp. nov. are proposed below for Citrobacter genomospecies 10 and for Citrobacter genomospecies 11, respectively.

Where known, the sources of isolation of the strains used in the present and in the previous studies (2) are shown in Table 4. It seems clear that all Citrobacter species other than C. rodentium are predominantly isolated from human clinical specimens.

Description of Citrobacter gillenii sp. nov. Citrobacter gillenii (gil.len'i.i. N.L. gen. n. gillenii, to honor George Francis Gillen, an American microbiologist, who, along with C. H. Werkman, studied the fermentative production of trimethylene glycol from glycerol and proposed the genus Citrobacter [15]). Formerly called Citrobacter genomospecies 10. It is negative for the production of indole and ornithine decarboxylase, positive for utilization of malonate, and delayed positive for growth on citrate and usually for production of arginine dihydrolase. Other biochemical characteristics useful for differentiation are negative reactions for production of urease and fermentation of dulcitol and the inability to utilize gentisate, 3-hydroxybenzoate, 3-O-CH₃-D-glucose, L-sorbose, and tricarballylate as sole carbon sources (2). Complete results of routine biochemical reactions are given in Table 2, and complete carbon source utilization reactions are given in Table 3.

Description of Citrobacter murliniae sp. nov. Citrobacter murliniae (mur.lin'i.ae. N.L. gen. n. murliniae, to honor Alma C. McWhorter-Murlin, an American microbiologist, who, during her 39-year career at the Centers for Disease Control and Prevention, made substantial contributions to our knowledge of Salmonella serotyping and Citrobacter and to the taxonomy of a number of species in the family Enterobacteriaceae [3, 4, 9, 10]). Formerly called Citrobacter genomospecies 11. It is positive or delayed positive for production of indole and growth on citrate, usually delayed positive for production of arginine dihydrolase, and negative for production of ornithine decarboxylase. Other biochemical tests useful for differentiation are acid production from dulcitol and esculin (delayed), growth on sodium acetate (usually delayed) but not on malonate, and the ability to utilize dulcitol, D-lyxose, 1-O-CH₃--galactoside (delayed), and L-tyrosine, but not malonate and protocatechuate, as sole carbon sources. Complete results of routine biochemical reactions are given in Table 2, and complete carbon source utilization reactions are given in Table 3.