The Genus Arthrobacter

Conn (1928) described a group of bacteria, extremely numerous in certain soils, which were unusual in that they appeared as Gram-negative rods in young cultures and as Gram-positive cocci in older cultures. For these bacteria, Conn (1928) created the species Bacterium globiforme, which, as Arthrobacter globiformis, was later to become the type species of the genus Arthrobacter. The abundance in soil of bacteria similar to Conn’s organism, and of other coryneform bacteria, was confirmed later by Jensen (1933, 1934) and Topping (1937, 1938), who, however, referred to them as soil corynebacteria, and by Taylor and Lochhead (1937), who used the name Bacterium globiforme. Jensen (1934) considered that these soil bacteria should be classified in the genus Corynebacterium because of their morphological resemblance to corynebacteria of animal origin. However, Conn (1947) vigorously opposed this view and created the genus Arthrobacter (by reviving an old name), with A. globiformis as the type species and with two of Jensen’s soil corynebacteria as additional species (Conn and Dimmick, 1947).

In addition to their characteristic morphology and staining reactions, members of the genus Arthrobacter were originally described as being highly aerobic, nutritionally nonexacting, and capable of liquefying gelatin slowly (Conn and Dimmick, 1947). These features were chosen mainly to distinguish Arthrobacter from Corynebacterium as represented by C. diphtheriae and similar animal parasitic species. However, because of its poor circumscription (see Gibson, 1953; Jensen, 1952), the genus Arthrobacter was not widely accepted until it was included as a member of the family Corynebacteriaceae in the seventh edition of Bergey’s Manual of Determinative Bacteriology (Breed et al., 1957). But by that time, the genus had been extended to include the two nutritionally exacting species A. terregens (Lochhead and Burton, 1953) and A. citreus (Sacks, 1954), and shortly afterwards two others were added (Lochhead, 1958a). Indeed, one of Conn’s strains of A. globiformis was shown subsequently to require biotin for growth (Chan and Stevenson, 1962; Morris, 1960). Thus the concept had developed of Arthrobacter as a genus of soil bacteria whose major distinguishing feature was a growth cycle in which the irregular rods in young cultures were replaced by coccoid forms in older cultures; these coccoid forms, when transferred to fresh medium, produced outgrowths (“germinated”) to give irregular rods again, and so the cycle was repeated (Fig. 1).

This dependence on morphological features and habitat in the circumscription led to a great deal of confusion in the classification of the genus Arthrobacter and thus created considerable problems in the identification of new isolates as arthrobacters. Thus, isolates from soil and, more especially, those from other habitats have frequently been referred to in the literature as arthrobacters on the basis of morphological features alone, even though they were not necessarily similar to A. globiformis in other respects.

It was not surprising that when representatives of the genus were examined by more modern taxonomic methods such as numerical taxonomy (Jones, 1978), various chemotaxonomic techniques (Bowie et al., 1972; Keddie and Cure, 1977, 1978; Minnikin et al., 1978b; Schleifer and Kandler, 1972), and determinations of DNA base ratios (see Skyring and Quadling, 1970; Skyring et al., 1971), it was found to be heterogeneous.

The genus Arthrobacter as defined in the eighth edition of Bergey’s Manual (Keddie, 1974) was heterogeneous, as was noted by Keddie and Jones (1981) in the first edition of The Prokaryotes. They referred to Arthrobacter in this broad sense as Arthrobacter sensu lato (Keddie and Jones, 1981). In Bergey’s Manual of Systematic Bacteriology (Keddie et al., 1986), the genus was limited to those species which, like the type, A. globiformis, contain lysine as the cell wall diamino acid, i.e., Arthrobacter sensu stricto (Keddie and Jones, 1981). Thus some species formerly considered to be arthrobacters (Keddie, 1974) have now been removed from the genus.

The two species formerly named A. terregens and A. flavescens, which contain ornithine as the cell wall diamino acid (Schleifer and Kandler, 1972; Keddie and Cure, 1977), have been transferred to the genus Aureobacterium as Aur. terregens and Aur. flavescens (Collins et al., 1983). The species Arthrobacter radiotolerans has now been transferred to the new genus Rubrobacter as R. radiotolerans (Suzuki et al., 1988).

Although now resolved, the position of the two species A. simplex and A. tumescens has been more problematical. They were shown by Cummins and Harris (1959) to differ from A. globiformis in containing LL-diaminopimelic acid (LL-A₂pm) as the cell wall diamino acid, and many other taxonomic differences were detected subsequently (see Keddie et al., 1986, for further details). In the case of A. simplex, 16S rRNA cataloging studies showed this species to be only distantly related to A. globiformis (Stackebrandt et al., 1980). Conversely, 5S rRNA sequencing indicated that Pimelobacter simplex (A. simplex—see below) clearly belonged to an “Arthrobacter—Micrococcus—Cellulomonas” subgroup of the coryneform bacteria (Park et al., 1987). While there was general agreement that A. simplex and A. tumescens should be removed from Arthrobacter, there was disagreement about where they should be accommodated. Suzuki and Komagata (1983) created the genus Pimelobacter for the LL-A₂pm-containing coryneform bacteria and distinguished three species by use of DNA-DNA base-pairing techniques. The first species, Pimelobacter simplex, contained most strains of A. simplex (and also strains named “Brevibacterium lipolyticum”), the second, P. tumescens, was for strains formerly called A. tumescens, and a third species, P. jensenii, was created for a single strain originally identified as an A. simplex strain (Gundersen and Jensen, 1956). The same authors also concluded from their DNA homology studies that A. simplex and A. tumescens were only distantly related to Nocardioides albus, a nocardioform organism. However, O’Donnell et al. (1982) considered that A. simplex (but not A. tumescens) closely resembled Nocardioides species in chemotaxonomic (particularly lipid) characters and proposed that it be transferred to that genus as Nocardioides simplex. This view received support from other studies (reviewed by Keddie et al., 1986). In an attempt to resolve the problem, Collins et al. 1989 made a comparative study of the 16S rRNA from the type strains of Nocardioides albus, N. luteus, Pimelobacter simplex, P. tumescens, and P. jensenii by reverse transcriptase sequencing and compared the results with those from 18 previously studied actinomycetes from 14 different genera. The study confirmed the reclassification of P. (Arthrobacter) simplex in the genus Nocardioides as N. simplex as proposed by O’Donnell et al. (1982) and also showed that P. jensenii should be transferred to that genus as N. jensenii (Collins et al., 1989). However, P. (Arthrobacter) tumescens was so distinct from all the other actinomycete taxa that Collins et al. (1989) proposed its reclassification in a new genus, Terrabacter, as T. tumescens. The taxonomic status of the single strain of the species A. duodecadis remains unresolved. Details of this and some other species named Arthrobacter, but now excluded from the genus, are given by Keddie et al. 1986 and for A. siderocapsulatus and A. viscosus by Collins (1986).

However, Collins (1986) has shown that the lipid composition of “A. sialophilus” (Tanenbaum and Flashner, 1977) and the phytopathogen Agrobacterium pseudotsugae is consistent with their being members of Arthrobacter sensu stricto (Keddie and Jones, 1981), but no formal proposal has been made to include them in the genus.

As presently circumscribed, the genus Arthrobacter contains two “groups of species” referred to as the A. globiformis/A. citreus group and the A. nicotianae group. These groups differ in their peptidoglycan structure, teichoic acid content, and lipid composition (see “Further Identification of Arthrobacters” and Tables 2 and 3 for details). It has been suggested that the genus should be restricted to those bacteria that exhibit the characteristics of the A. globiformis/A. citreus group (Minnikin et al., 1978a; Collins and Kroppenstedt, 1983). However, on the basis of DNA-DNA homology studies of representative arthrobacters (Stackebrandt and Fiedler, 1979) and later 16S rRNA cataloging studies, Stackebrandt et al. (1983) concluded that the genus Arthrobacter contained two “nuclei,” one represented by the A. globiformis/A. citreus group of species and the other by the A. nicotianae group. This view was adopted by Keddie et al. (1986) and is the one accepted here.

Table 2. Characters^a most useful in differentiating the Arthrobacter species which have peptidoglycans of the A4α variation^b and MK-8^c (or MK-9^d) as major menaquinones—the A. nicotianae group.^{e, f}

Table 3. Characters^a differentiating Arthrobacter from similar genera which either have a rod-coccus growth cycle or have lysine as the cell wall diamino acid.^b

The results of 16S rRNA cataloging studies (Stackebrandt et al., 1980; Stackebrandt and Woese, 1981) indicate that the genus Arthrobacter is related to the other coryneform genera, Aureobacterium, Cellulomonas, Curtobacterium, and Microbacterium, and is more distantly related to Brevibacterium. All of these genera are members of the high GC “actinomycete” branch of the Gram-positive eubacteria (Stackebrandt and Woese, 1981). The studies of Stackebrandt et al. (1980) also showed that on a phylogenetic basis the Arthrobacter species could not be separated from members of the genus Micrococcus. While accepting that the genera Arthrobacter and Micrococcus are very closely related phylogenetically, we treat them here as distinct taxa for practical purposes.