Identification of Clinical Isolates of Actinomyces Species by Amplified 16S Ribosomal DNA Restriction Analysis

Abstract

Amplified 16S ribosomal DNA (rDNA) restriction analysis (ARDRA), using enzymes HaeIII and HpaII, was applied to 176 fresh and 299 stored clinical isolates of putative Actinomyces spp. referred to the Anaerobe Reference Unit of the Public Health Laboratory Service for confirmation of identity. Results were compared with ARDRA results obtained previously for reference strains and with conventional phenotypic reactions. Identities of some strains were confirmed by analysis of partial 16S rDNA sequences. Of the 475 isolates, 331 (70%) were clearly assigned to recognized Actinomyces species, including 94 isolates assigned to six recently described species. A further 52 isolates in 12 ARDRA profiles were designated as apparently resembling recognized species, and 44 isolates, in 18 novel profiles, were confirmed as members of genera other than Actinomyces. The identities of 48 isolates in nine profiles remain uncertain, and they may represent novel species of Actinomyces. For the majority of species, phenotypic results, published reactions for the species, and ARDRA profiles concurred. However, of 113 stored isolates originally identified as A. meyeri or resembling A. meyeri by phenotypic tests, only 21 were confirmed as A. meyeri by ARDRA; 63 were reassigned as A. turicensis, 7 as other recognized species, and 22 as unidentified actinomycetes. Analyses of incidence and clinical associations of Actinomyces spp. add to the currently sparse knowledge of some recently described species.

The genus Actinomyces comprises a heterogeneous group of anaerobic and facultatively anaerobic, nonspore-forming, nonmotile, non-acid-fast, gram-positive rods that have a G+C content of 55 to 71 mol% (1). Volatile and nonvolatile acid end products of glucose metabolism are acetic, lactic, and succinic acids (11). The natural habitats of many Actinomyces spp. are the mucous membranes of humans and other animals, particularly the oral mucosa. Members of the genus can cause classical actinomycosis, are associated with infections arising from tissue invasion by oral anaerobes, and may be instrumental in the development of periodontal diseases (21).

Historically, classification of Actinomyces spp. was based upon differentiation in a few phenotypic tests (11). However, application of modern chemotaxonomic and genotypic methods has demonstrated heterogeneity within the classic species and led to the recognition of several new species. Furthermore, these powerful tools have demonstrated the existence of several genera within the actinomyces group and the need for a major review of taxonomy (17, 22). Some actinomycetes have been reclassified as Arcanobacterium spp. or as Actinobaculum spp. (14, 17). However, the taxonomic positions of other species remain unsatisfactory, and further revisions are likely in the light of recently described species and additional taxa, as yet unnamed. Recently described actinomycetes isolated from human sources include Actinomyces europaeus (7), Actinomyces graevenitzii (16), Actinomyces neuii (8), Actinomyces radingae (27), Actinomyces turicensis (27), Actinomyces urogenitalis (15), and Actinomyces radicidentis (5). The clinical spectra of infections due to A. turicensis, A. radingae, and A. europaeus have been investigated by Sabbe et al. (20). However, for some species, few strains have been examined, and little is known of their natural habitats, clinical prevalence, and pathogenic potential.

Identification of Actinomyces spp. is notoriously difficult and unreliable (9), yet clinically important. Current taxonomy is based upon delineation in 16S ribosomal DNA (rDNA) sequence analysis, whole-cell protein profiling, and an extensive range of phenotypic characteristics. Clearly, this approach is impractical for routine identification of clinical isolates. Hence, commonly, identification in clinical laboratories is based solely upon a limited range of conventional biochemical tests. Even when a wider range of tests is performed, these lack discrimination at species level and are subject to method-dependent variations, and overlaps in phenotype between isolates of different genospecies may occur, e.g., potential misidentification of A. turicensis as A. meyeri in phenotypic tests has been demonstrated (10). Therefore, identification to species level is often tentative. Furthermore, few clinical laboratories have the ability to perform gas-liquid chromatography for end products of glucose metabolism. This valuable aid to genus-level identification of non-spore-forming gram-positive bacilli enables differentiation of Actinomyces spp. from morphologically similar but nonpathogenic isolates of Propionibacterium, Bifidobacterium, and Lactobacillus (11). Hence, clinical isolates are often misidentified, and many reported Actinomyces spp. are members of other genera. Amplified 16S rDNA restriction analysis (ARDRA) with enzymes HaeIII and HpaII has been shown to be a simple and highly discriminatory method for identification of Actinomyces isolates, including these recently described species (10).

The aims of this study were to evaluate ARDRA for identification of clinical isolates of putative Actinomyces spp., to create a robust library of ARDRA profiles for Actinomyces spp. of clinical origin, and to explore the prevalence of species and their associations with specific natural habitats, sites of infection, or particular pathogenic potential. ARDRA was applied to 176 fresh and 299 stored clinical isolates of putative Actinomyces spp. referred to the Anaerobe Reference Unit (ARU) of the Public Health Laboratory Service for confirmation of identity. Isolates were referred between 1983 and 1999 from hospital laboratories throughout England and Wales. Results were compared with those obtained for reference strains (10) and with conventional phenotypic reactions. Identities of some strains were confirmed by analysis of partial 16S rDNA sequences.

MATERIALS AND METHODS

Bacterial strains.

A total of 475 strains, comprising 176 fresh and 299 stored clinical isolates, were examined. The fresh isolates were all of the putative Actinomyces spp. received by the ARU for confirmation of identity in 1998 and 1999. For these strains, ARDRA was performed blind, in parallel with conventional phenotypic tests. Stored isolates were selected from those referred during 1983 to 1997 and were identified at the time of submission by conventional phenotypic tests as Actinomyces spp. or as gram-positive rods of uncertain identity. Strains were selected to represent the range of morphological and biochemical diversity within the genus, and in light of previous findings, all strains (n = 113) previously identified as A. meyeri or resembling A. meyeri were examined. Strains were stored at −80°C on Microbank beads (Pro-lab Diagnostics, Wirral, United Kingdom) and were recovered on Fastidious Anaerobe Agar (LabM, Bury, United Kingdom) incubated anaerobically at 37°C for 48 h.

ARDRA.

Tests were performed and analyzed as described previously (10). Each isolate was assigned a six-digit code, the first and second groups of three digits representing banding patterns obtained in HaeIII and HpaII digests, respectively. Novel banding patterns were assigned three-digit codes, as encountered, and added to existing HAE and HPA libraries (10). Resulting ARDRA profiles were equated with species where they contained a reference strain and clinical isolates with concordent phenotypic reactions, or where the identity of one or more representative strains was confirmed by 16S rDNA sequencing. For isolates that yielded ARDRA profile 001/003 (a pattern obtained for two species), a supplementary digest with EarI was performed as described for HaeIII and HpaII digests (10). Resulting patterns were analyzed as above and assigned a three-digit code, listed after the HAE/HPA code, e.g., 001/003/001.

Conventional tests.

Fresh clinical isolates were identified as members of the genus Actinomyces on the basis of volatile and nonvolatile fatty acid end products of glucose metabolism, detected by gas-liquid chromatography (11).

Cell and colonial morphologies, pigment production, fluorescence under long-wave UV illumination, ability to grow in air and in air plus 5% CO₂, and production of catalase and indole were recorded. Hydrolysis of esculin and starch and production of acid from amygdalin, arabinose, cellobiose, glucose, mannitol, raffinose, ribose, salicin, sucrose, trehalose, and xylose were tested by the method of Phillips (18). Production of nitrate reductase, urease, pyrazinamidase, β-galactosidase, α-glucosidase, and β-N-acetylglucosaminidase were detected after incubation for 18 to 24 h with Rosco diagnostic tablets (BioConnections, Leeds, United Kingdom). Results were interpreted with reference to the identification tables of Funke et al. (7, 8), Lawson et al. (14), and Pascual Ramos et al. (16). Stored isolates had been similarly identified at the time of submission to the ARU, and results were interpreted according to the Anaerobe Laboratory Manual (11) and other publications available at the time of testing (8, 13).

Partial 16S rDNA sequencing.

The 16S rDNA was extracted and amplified as for ARDRA (10). Amplified products were purified by Qiaquick spin-column (Qiagen, Crawley, United Kingdom) according to the manufacturer's instructions. A variable region of approximately 450 bp of the 16S gene was sequenced with a reverse primer targeting positions 536 to 516 (Escherichia coli numbering) and the ABI-PRISM Big Dye Terminator sequencing kit (Perkin Elmer, Warrington, United Kingdom) according to the manufacturer's instructions but in half volume. Sequences were compared to those in the EMBL database with BlastN 2.1.1 and analyzed further in DNASIS (Hitachi Software, Yokohama, Japan). The efficacy of this method as a screening tool, representative of full 16S gene sequence similarities, was demonstrated in preliminary experiments by application to some well-characterized clinical isolates of Actinomyces spp. (authors' data; not presented). Similarities of >98% were obtained with reference strains of the same species.

RESULTS

ARDRA profiles of fresh and stored isolates are listed in Table 1. The 475 clinical isolates produced 62 profiles containing from 1 to 59 strains (Table 1); 331 isolates (70%) were clearly assigned to recognized Actinomyces species, confirmed either by concurrence with ARDRA profiles obtained for reference strains and supportive results in conventional phenotypic tests or by >98% partial 16S rDNA sequence similarities of representative strains with those of reference strains (Table 2). Representative strains of 52 isolates in 12 ARDRA profiles showed ∼N97% partial 16S rDNA sequence similarities with those listed for reference strains, so isolates in these groups were designated as apparently resembling recognized species (Tables 1 and 2). Forty-four isolates, in 18 novel profiles, were confirmed as members of genera other than Actinomyces (Tables 1 and 2). The identities of 48 isolates in nine profiles remain uncertain, and these may represent novel species of the actinomyces group. Thirty-nine of these isolates were contained in three ARDRA profiles, and within each group, isolates were morphologically and phenotypically similar. Group 1 strains (n = 12, ARDRA 001/016) resembled Actinomyces odontolyticus but grew poorly or not at all in air, produced pink rather than red colonies, and failed to ferment lactose. Group 2 strains (n = 19, ARDRA 025/026) resembled A. meyeri but grew moderately well in air plus CO₂ and poorly or not at all in air. Colonies were grey rather than white, and most strains reduced nitrate to nitrite. Group 3 strains (n = 8, ARDRA 035/023) were obligately anaerobic, yielding tiny grey colonies, were weakly saccharolytic, and reduced nitrate to nitrite. Partial 16S rDNA sequence analysis of representative strains of each group confirmed them as members of the genus Actinomyces but distinct from currently recognized species.

TABLE 1.

ARDRA identification of fresh and stored isolates of putative actinomycetes^a

Identification	No. of isolates	ARDRA profile	No. of isolates
			Fresh	Stored
Resembling A. denticolens	5	011/027	1	4
A. europaeus	5	012/014	2	3
Resembling A. europaeus	3	042/014	0	3
A. georgiae	4	001/004	2	2
A. gerencseriae	41	006/008	20	21
A. graevenitzii	3	013/013	1	2
Resembling A. graevenitzii	6	031/013	2	4
A. israelii	72	008/009	18	15
		020/009	30	8
		040/009	1	0
Resembling A. israelii	6	018/009	2	4
A. meyeri	28	001/003/002	5	23
A. naeslundii genospecies 1	16	003/005	2	7
		009/005	1	6
A. naeslundii genospecies 2	37	014/005	2	5
		014/010	1	2
		017/005	0	4
		021/005	7	16
Resembling A. naeslundii/ A. viscosus	8	011/005	1	2
		017/010	0	1
		021/010	1	1
		083/005	1	0
		054/040	1	0
A. neuii	8	022/019	3	5
A. odontolyticus	31	001/001	8	21
		073/001	0	1
		074/001	1	0
Resembling A. odontolyticus	24	001/003/001	5	2
		025/001	4	11
		025/003	0	2
A. radingae	3	028/023	2	1
A. turicensis	80	015/016	4	55
		015/032	3	10
		016/016	0	8

A. urogenitalis	3	024/011	1	2
Unidentified actinomycete group	48	001/016	1	11
		001/026	0	1
		019/017	0	1
		023/020	0	1
		025/026	6	13
		034/014	0	1
		035/023	3	5
		036/019	1	2
		041/020	0	2
A. schaalii	3	004/015	1	2
A. bernardiae	3	026/021	2	1
A. haemolyticum	10	027/022	3	7
A. vaginae	1	039/033	0	1
Bifidobacterium adolescentis group	1	060/045	1	0
Bifidobacterium bifidum	1	061/046	1	0
Bifidobacterium infantis group	1	058/041	1	0
Bifidobacterium longum	2	054/043	2	0
Bifidobacterium sp.	1	043/055	1	0
Lactobacillus acidophilus	1	049/037	1	0
Lactobacillus rhamnosus	1	064/047	1	0
Lactobacillus sp.	1	067/053	1	0
Propionibacterium acnes	2	033/028	2	0
Propionibacterium avidum	1	048/036	1	0
Propionibacterium granulosum	6	038/031	6	0
P. propionicum	8	052/038	7	0
		053/039	1	0
Streptococcus mutans	1	032/029	1	0
All isolates	475		176	299

^aFresh isolates comprise all putative Actinomyces spp. referred in 1998 and 1999, selected solely by submission from referring laboratories. Stored isolates were past referrals (1983 to 1997) selected for morphological and biochemical diversity.  

TABLE 2.

Similarities to reference strains of partial 16S rDNA sequences of clinical isolates^a

ARDRA profile	Strain (ARU no.)	Reference strain having highest similarity	Sequence similarity
			EMOL accession no.	% similarity
011/027	R12391	Actinomyces sp. oral strain B19SCb	AF287748	97.5
		Actinomyces denticolens NCTC 11490T	X80412	97.2
042/014	R4119	A. europaeus CCUG 32789AT	Y08828	97.0
031/013	R12575	A. graevenitzii CCUG 27294T	Y09589	97.3
008/009	R10167, R9548	A. israelii ATCC 12102T	X82450	99.6
018/009	R3358	A. israelii ATCC 12102T	X82450	97.4
020/009	R5753	A. israelii ATCC 12102T	X82450	99.6
040/009	R11968	A. israelii ATCC 12102T	X82450	99.3
003/005	R7710	A. naeslundii ser I NCTC 10301T	X81062	98.4
009/005	R8152	A. naeslundii ser I NCTC 10301T	X81062	98.4
011/005	R9108	A. naeslundii ser. I NCTC 10301T	X81062	95.5
014/005	R4479	A. naeslundii ser. I NCTC 10301T	X81062	96.8
014/005	R1265	A. naeslundii ser. II ATCC 49339T	Authors' data	96.4
		Actinomyces sp. oral clone EP011b	AY008315	96.5
017/005	R2242	A. viscosus ser. II ATCC 27044T	Authors' data	96.5
		Actinomyces sp. oral clone EP005b	AY008314	96.5
021/005	R2589	A. viscosus ser. II ATCC 27044T	Authors' data	96.9
021/005	R11430	A. viscosus ser. II ATCC 27044T	Authors' data	95.7
		Actinomyces sp. oral clone AP064b	AF287749	97.2
083/005	R13569	A. naeslundii ser. I NCTC 10301T	X81062	97.0
		Actinomyces sp. oral clone EP011b	AY008315	97.0
054/040	R13724	A. naeslundii ser. I NCTC 10301T	X81062	97.0
		Actinomyces sp. oral clone EP011b	AY008315	97.0
014/010	R1284	Actinomyces sp. oral clone AG004b	AF287747	100
		A. naeslundii ser. I ATCC 49339T	Authors' data	99.0
017/010	R11372	A. viscosus ser. I NCTC 10951T	X82453	95.3
021/010	R7437	A. naeslundii-like ATCC 49338	X81063	99.4
001/001	R5969	A. odontolyticus CCUG 20536T	AJ234040	99.1
025/001	R6084	A. odontolyticus CCUG 20536T	AJ234040	97.6
073/001	R5568	A. odontolyticus CCUG 20536T	AJ234040	98.0
074/001	R13009	A. odontolyticus CCUG 28084	AJ234041	99.1
001/003/00	R10717	A. odontolyticus-like CCUG 32402	X78721	98.7
025/003	R10146	A. odontolyticus-like CCUG 32402	X78721	99.0
015/016	R8614	A. turicensis DSM 9168T	X78720	99.8
016/016	R10672	A. turicensis DSM 9168T	X78720	100
015/032	R5978	A. turicensis DSM 9168T	X78720	99.8
024/011	R6344	A. urogenitalis CCUG 38702T	AJ243791	99.4
001/016	R5571, R10394	A. meyeri ATCC 35568T	X82451	91.1
001/026	R5040	A. odontolyticus CCUG 20536T	AJ234040	96.0
019/017	R5292	A. naeslundii ser. I NCTC 10301T	X81062	91.6
		A. viscosus ser. I ATCC 27044T	Authors' data	91.6
		Actinomyces sp. oral clone AP064b	AF287749	91.9
023/020	R7773	A. israelii ATCC 12102T	X82450	93.5
025/026	R5307, R10236	A. hyovaginalis NCFB 2983T	X69616	94.0
035/023	R5231	A. neuii DSM 8576T	X71861	90.8
034/014	R5638	A. neuii DSM 8576T	X71861	88.7
035/023	R12359	A. neuii DSM 8576T	X71861	90.8
036/019	R11881	A. neuii DSM 8576T	X71861	87.3
041/020	R7252	A. israelii ATCC 12102T	X82450	94.9
039/033	R10307	A. vaginae CCUG 38953T	Y17195	100
032/029	R2334	S. mutans NCTC 10449	AJ243965	99.2

^aAbbreviations: ATCC, American Type Culture Collection; NCTC, National Collection of Type Cultures, London; England; CCUG, Culture Collection, University of Goteborg, Goteborg, Sweden; DSM, Deutsche Sammlung von Mikroorganismen, Braunschweig, Germany; NCFB, National Collection of Food Bacteria, Reading, England; ser., serotype. 

^bSequences submitted by Paster et al. (17a). 

Profile 001/003 comprised isolates identified in phenotypic tests as A. meyeri (n = 28) and A. odontolyticus (n = 7) plus reference strains ATCC 35568 (A. meyeri type strain) and CCUG 32402 (APL 11), described by Wust et al. as resembling A. odontolyticus (27, 28) (ARDRA data unpublished). In EarI digests, the two taxa were clearly differentiated, in concurrence with phenotypic tests, and as predicted by analysis of 16S rDNA sequences for reference strains; isolates resembling A. odontolyticus were not cut (profile 001), whereas A. meyeri isolates produced a band of approximately 130 bp (profile 002). Partial 16S rDNA sequencing of representative strains from this group and reference strains of A. odontolyticus and A. meyeri demonstrated a high level of similarity between the two species and the intermediate position of strain CCUG 32402 and clinical isolates R6084, R10717, and R10146 (Table 2).

Precise correlation of identities in ARDRA with those obtained in phenotypic tests was not possible, as the latter often lacked discrimination at species level, and some isolates produced variable carbohydrate fermentation reactions on repeat testing. For most fresh isolates, phenotypic reactions were consistent with published reactions and with identities obtained in ARDRA. For stored isolates identified in phenotypic tests as Actinomyces israelii, Actinomyces gerencseriae, or Actinomyces naeslundii or A. viscosus, 61 of 72 were confirmed in ARDRA. Four isolates were identified in ARDRA as A. israelii (profile 008/009) and in conventional tests as A. gerencseriae by virtue of their inability to ferment arabinose. Two of these fermented arabinose on repeat testing. Partial 16S rDNA sequences of the other two isolates (ARU strains R10167 and R9548) confirmed both to be A. israelii (Table 2). For three isolates, identified in ARDRA as A. neuii but originally identified as A. viscosus, phenotypic reactions were reassessed, and all were confirmed as A. neuii by their ability to ferment mannitol and xylose.

ARDRA confirmed the identities of 35 of 59 stored isolates identified in phenotypic tests as A. odontolyticus. The remainder were found to be A. israelii (n = 5), A. naeslundii or resembling naeslundii (n = 4), A. geraevenitzii (n = 3), and one each of A. gerencseriae, A. meyeri, and Actinomyces georgiae, and unidentified actinomycetes (n = 9, of which eight belonged to ARDRA group 1, profile 001/016). Dark pigmentation was recorded in some mature cultures of A. graevenitzii and A. georgiae, and pink pigmentation was found in some A. israelii and A. naeslundii isolates. These findings, and the several carbohydrate fermentation tests listed as variable for A. odontolyticus, may account for some of these misidentifications in phenotypic tests.

Of 113 stored isolates originally identified as A. meyeri or resembling A. meyeri by phenotypic tests, only 21 were confirmed as A. meyeri by ARDRA. Seventy were reassigned as A. turicensis (n = 63), A. europaeus (n = 1), apparently resembling A. europaeus (n = 2), A. georgiae (n = 1), A. israelii (n = 1), Actinobaculum schaalii (n = 1), and Arcanobacterium bernardiae (n = 1). Twenty-two isolates in eight novel ARDRA profiles remain unidentified, but partial 16S rDNA sequencing of representative strains suggests that these are Actinomyces spp.

Of 55 isolates originally identified only as Actinomyces sp. or as gram-positive rods of uncertain identity, 47 were assigned in ARDRA to various recognized Actinomyces spp. (n = 38), of which 18 were in recently described species, Arcanobacterium haemolyticum (n = 7), A. schaalii (n = 1), or Atopobium vaginae (n = 1); eight remain unidentified. Failure to identify isolates confirmed in ARDRA as A. israelii (n = 5) and A. gerencseriae (n = 2) had been due to their weak or negative carbohydrate fermentation reactions.

All of the eight isolates identified in ARDRA as A. europaeus or apparently resembling A. europaeus reduced nitrate to nitrite, as did the A. europaeus type strain (CCUG 32789A) in our hands (10); three A. europaeus isolates and the type strain were asaccharolytic.

Clinical sources of Actinomyces spp. are summarized in Table 3. The classic species A. gereneseriae, A. israelii, A. odontolyticus, and the A. naeslundii/viscosus complex were isolated principally from cervicofacial sites and intrauterine contraceptive devices (IUCDs).

TABLE 3.

Sources of Actinomyces spp.

Sourceb	Actinomyces spp.a
	Den.	Eur.	Geo.	Ger.	Gra.	Isr.c	Mey.	Nae.	Neu.	Odo.	Rad.	Tur.	Uro.	Gp. 1	Gp. 2	Gp. 3	Oth.	Total
Neck-face	1	2	2	11	6	17	6	17	2	12		4		2	1	2	3	88
Eye			1	5		4		2		3								15
Thorax					3	3	5	2		4		1		1	1		1	21
Abdomen				3		7	3	2		1		4						20
Pelvis		1		2		4	1					1		1	1			11
IUCD	3	1		13		41	2	18	1	18	1	21		5	4	1	1	130
Vagina/penis	1			2				1		1		26	1	1	2	1	1	37
Superficial		2				1	2	1	5	4	2	15	1		7	3	2	45
Blood			1					9		7		3						20
Brain/CSF		1		1			7	1		1		2		1	2	1		17
Other/unknown		1		4		2	2	8		4		3	1	1	1		1	28
Total	5	8	4	41	9	79	28	61	8	55	3	80	3	12	19	8	9	432

^aAbbreviations for Actinomyces spp.: Den., resembling A. denticolens; Eur., A. europaeus and resembling A. europaeus; Geo., A. georgiae; Ger., A. gerencseriae; Gra., A. graevenitzii and resembling A. graevenitzii; Isr., A. israelii and resembling A. israelii; Mey., A. meyeri; Nae., A. naeslundii genospecies 1 and 2 and resembling A. naeslundii/A. viscosus; Neu., A. neuii; Odo. A. odontolyticus and resembling A. odontolyticus; Rad., A. radingae; Tur., A. turicensis; Uro., A. urogenitalis; Gp. 1, unidentified actinomycetes ARDRA profile 001/016; Gp. 2, unidentified actinomycetes ARDRA profile 025/026; Gp. 3, unidentified actinomycetes ARDRA profile 035/023; Oth., unidentified actinomycetes ARDRA profiles 001/026, 019/017, 023/020, 034/014, 036/019, and 041/020. 

^bSuperficial, soft tissue lesions including breast, axillary, limb, groin, buttock, rectal, and pilonidal abscesses; CSF, cerebrospinal fluid. 

^cOne A. israelii isolate from lung and liver included in thoracic and abdominal sources. 

A. meyeri strains (n = 28) were from brain abscesses (n = 7), cervicofacial lesions (n = 6), pleural fluids (n = 4), and a chest abscess (n = 1). A. europaeus and organisms apparently resembling this species were from breast abscesses (n = 2) and one from each of brain, neck, and mastoid abscesses, Pouch of Douglas fluid, IUCD, and necrotizing fasciitis of the thigh. A. graevenitzii and organisms resembling this species were found in intraoral sources (n = 4), and one each in lung, bronchoalveolar lavage, sputum, neck abscess, and osteomyelitis of the jaw. A. neuii occurred in breast (n = 3), dental (n = 2), axillary (n = 1), and buttock (n = 1) abscesses and an IUCD. A. radingae was found in one each of breast and axillary abscesses and an IUCD. A. urogenitalis was isolated from a vaginal swab, a groin abscess, and a case of osteomyelitis, site unspecified. Principal sources of A. turicensis were penile lesions, mainly balanitis (n = 24), IUCDs (n = 21), abscesses of groin or rectal areas (n = 6), and pilonidal abscesses (n = 5). Six of eight isolates identified as Propionibacterium propionicum were from cases of lacrimal canaliculitis. The strain identified as A. vaginae was isolated from kidney tissue.

DISCUSSION

Analysis of 475 clinical isolates of putative Actinomyces spp. confirmed the efficacy of ARDRA for identification of members of this genus; 70% of strains were identified as Actinomyces to species level with confidence, 11% apparently resembled recognized species, 9% were members of other genera, and 10% were deemed to be unidentified actinomycetes. In general, phenotypic data supported identifications made in ARDRA but, in many cases, lacked discrimination between species. Thus, identities of some isolates were amended upon reassessment of phenotypic data in the light of ARDRA findings. This was most notable in the reassignment to other species of 81% of isolates originally identified as A. meyeri or resembling A. meyeri. These data underline those of other studies in which isolates subsequently found to be A. turicensis were variously described on the basis of phenotypic data as resembling A. meyeri (2, 10), Actinomyces (now Arcanobacterium) pyogenes (28), or Gardnerella vaginalis (25, 26).

Reassuringly, all of 44 isolates found to be members of genera other than Actinomyces yielded distinct ARDRA profiles, and thus none was misidentified as an Actinomyces sp. Numbers tested were small, but the profiles obtained indicate the potential of this method for identification of a broader range of genera. These findings and the authors' data for reference strains (not presented) demonstrate the ability of ARDRA to differentiate Actinomyces spp. from other non-spore-forming gram-positive bacilli, obviating the need for gas-liquid chromatography and conventional biochemical tests. Furthermore, when molecular expertise and equipment are available, the simplicity and cost-effectiveness of ARDRA compare favorably with conventional tests and have enabled the examination of a large number of strains in this study. At the ARU, the current strategy for identifying unknown, clinically significant anaerobic or microaerophilic actinomycetes comprises observation of cell and colony morphologies and atmospheric requirements plus ARDRA. When the ARDRA profile is consistent with a recognized Actinomyces sp. and morphology concurs, no further tests are necessary. When the ARDRA profile is distinct or suggests that the isolate is an unidentified actinomycete or a member of another genus, appropriate conventional biochemical tests, including gas-liquid chromatography for end products of glucose metabolism, are performed.

Difficulties in identification of actinomycetes in the clinical laboratory are demonstrated by outcomes for putative Actinomyces spp. referred to the ARU in 1998 and 1999. Only 65% of the 176 isolates were confirmed as members of recognized Actinomyces spp. A further 10% were identified as apparently resembling recognized Actinomyces spp., and 6% were deemed to be unidentified actinomycetes. The remainder (19%) were found to be members of other genera. Of interest, 9% of the 176 isolates were members of recently described species A. turicensis (n = 7), A. neuii (n = 3), A. europaeus (n = 2), A. radingae (n = 2), and A. graevenitzii (n = 1).

A. neuii subsp. neuii and A. neuii subsp. anitratus were not distinguished in ARDRA but can be differentiated by nitrate reductase reactions. Isolates of A. gerencseriae (n = 41) and A. meyeri (n = 28) formed homogeneous taxa in ARDRA, but in some species, subspecies variations were seen. Distinct profiles obtained within the species A. israelii and A. odontolyticus may relate to biochemical, serological, or molecular differentiations noted by other workers (1, 12, 23, 24). The three ARDRA profiles obtained for members of A. turicensis may denote similar variation within this species. Indeed, the dendrogram derived from whole-cell protein patterns, published by Vandamme et al. (25), indicates several distinct groups within A. turicensis. Within the A. naeslundii/A. viscosus complex, numerous distinct ARDRA profiles and relatively low similarities in 16S rDNA sequences confirmed the wide diversity of this group. Interestingly, some of these taxa showed high homology with 16S rDNA sequences of oral clones catalogued recently in EMBL by Paster et al. (see Table 1, footnote b) but not otherwise published.

Chemotaxonomic methods and 16S rDNA sequencing have demonstrated the existence of several genera within the genus Actinomyces as currently recognized (17, 22) and classification of this actinomycete group is still in flux. In this study, 23% of isolates deemed to be members of the actinomycete group could not be assigned to recognized species with confidence. Further clarification of taxonomy of the group is necessary, including a review of classification at the genus level.

Analyses of incidence and clinical associations of Actinomyces spp. must be undertaken with caution, as the study strains were a selected population; first by referral from clinical laboratories to the ARU and, second by selection of stored isolates to represent phenotypic diversity. Furthermore, for some species, the numbers of strains were small. However, certain trends are apparent.

In terms of incidence, the classic species were well represented, and strains belonging to six recently described species were identified. Numbers of A. europaeus, A. graevenitzii, A. neuii, A. radingae, and A. urogenitalis were small, but A. turicensis represented 17% of study strains. Similarly, Sabbe et al. (20) found that A. turicensis was isolated much more frequently than A. radingae or A. europaeus. Given that A. neuii is an aerotolerant catalase producer, it is probable that many clinical isolates of this species are dismissed as Corynebacterium spp. We found no isolates of A. radicidentis. To date, this species has been isolated only from infected root canals of teeth. It may be specific to that site and not associated with the wide range of clinical sites from which isolates were received in this study. Of interest among strains designated as apparently resembling recognized Actinomyces spp., five resembled Actinomyces denticolens, a species isolated from dental plaque of cattle (6) but not previously reported from human sources. Three ARDRA groups of unidentified actinomycetes contained 19, 12, and 8 isolates, suggesting that these organisms are at least as common in clinical material as some currently recognized species.

Clinical sources of the classic species A. gerencseriae, A. israelii, A. odontolyticus, and the A. naeslundii/A. viscosus complex have been well documented and were, as expected, principally cervicofacial sites and IUCDs. However, 25% of A. meyeri strains, now reliably identified, were isolated from brain abscesses, 21% from cervicofacial lesions, and 17% from thoracic sites. These appear to be real associations, as the study included all A. meyeri isolates in the ARU collection, and the sources of these are remarkably similar to those of the 16 strains reported by Cato et al. (4). Conversely, A. turicensis was isolated predominantly from genital or skin-related sources, similar to those reported previously for this organism (20, 25, 27). This species was also isolated from bacteremias in this and other studies (20, 25).

With the exception of A. turicensis, few isolates of recently described species have been reported to date. Therefore, though numbers in this study were small, our data add significantly to current knowledge of clinical sources of these species. Unfortunately, little information regarding possible significance and concomitant organisms was available to us. The skin-related and urogenital tract sources of A. europaeus and organisms apparently resembling this species were similar to those reported previously for this species (7, 20). However, one isolate was from a left temporal brain abscess; concomitant organisms were Fusobacterium nucleatum, Peptostreptococcus magnus, Prevotella loescheii, and coliforms. We believe this to be the first report of A. europaeus from a brain abscess. A. graevenitzii and organisms resembling this species were found exclusively in head, neck, and thoracic sites. These data support those of Pascual Ramos et al. (16) and suggest an oral niche for this organism. P. propionicum was associated with cases of lacrimal canaliculitis, as previously noted (3). Among unidentified actinomycetes, three ARDRA groups of 12, 19, and 8 isolates were prominent. Isolates in groups 1 and 3 occurred in sources similar to those of classic species, whereas group 2 isolates were found in sites similar to those of A. turicensis. Further investigations of these and other unidentified actinomycetes are in progress.

ARDRA is a valuable tool to elucidate the incidence in clinical material of currently recognized Actinomyces spp. and to screen for novel species. Currently, the method is suitable principally for specialist laboratories. However, rapid advances in molecular technologies, including DNA chips (microarrays), to which ARDRA may be readily adaptable, may bring the technique within the abilities of routine clinical laboratories.

ADDENDUM IN PROOF

Partial 16S rDNA sequences of representatives (R5307 and R10236) of the 19 strains in ARDRA profile 025/026 (ARDRA group 2) show 99% similarity to that published for the recently described species Actinomyces funkei (GenBank accession no. AJ404889) (P. A. Lawson, N. Nikolaitchouk, E. Falsen, K. Westling, and M. D. Collins, Int. J. Syst. Evol. Microbiol. 51:853–855, 2001).