Clinical Application of PCR-Restriction Enzyme Pattern Analysis for Rapid Identification of Aerobic Actinomycete Isolates

Abstract

The accuracy and practicality of PCR-restriction enzyme pattern analysis (PRA) for routine identification of aerobic actinomycete clinical isolates were evaluated for 299 cultures submitted to the Mycobacteria/Nocardia Laboratory at the University of Texas Health Center at Tyler. PRA identification using an amplified 439-bp segment (amplicon) of the 65-kDa heat shock protein gene was compared to identification by traditional methods, including growth characteristics, susceptibility patterns, biochemical testing, and high-performance liquid chromatography analysis. Microbiological examination of six cultures ruled out aerobic actinomycetes, and they were omitted from the study. Amplicons were analyzed with BstEII, HaeIII, MspI, HinfI, and BsaHI. When necessary, AciI, HhaI, and NarI were also used. From March 1995 through May 1997 (27 months), 274 of the remaining 293 (93.5%) isolates were accurately identified by PRA. Major diagnostic groups included 170 mycobacteria, 93 nocardiae, and 30 other aerobic actinomycetes. Mixed cultures were readily recognized by PRA, including a wound culture that contained two Nocardia taxa that were indistinguishable morphologically. Mycobacterium mucogenicum was identified in three cultures heavily contaminated with gram-positive cocci. The 19 isolates that produced PRA patterns that did not match those in the current PRA database were differentiated into 8 Mycobacterium species and 11 other aerobic actinomycetes by the presence or absence of BstEII recognition sites. Identification of 15 of these 19 isolates was also equivocal by traditional methods. PRA results were reportable within 2 to 5 working days and were as accurate as and faster and less expensive to obtain than those of traditional methods.

Interest in the identification and taxonomy of aerobic actinomycetes, nocardiae and mycobacteria in particular, has been increasing as a result of the increasing number of immunocompromised individuals in the population who are at greater risk for actinomycoses, especially those with advanced human immunodeficiency virus disease (1, 4, 5). Traditional methods for differentiation of species and taxa of aerobic actinomycetes are laborious and time-consuming and frequently require specialized testing that is beyond the capabilities of clinical laboratories (2–5, 9, 11, 19, 21). The occurrence of clinical isolates of aerobic actinomycetes that are inherently resistant to specific antimicrobials increases the significance of timely and accurate species and taxon recognition (4, 9, 17, 22).

Successful application of molecular biological methodology to the development of protocols for rapid differentiation of mycobacterial species was demonstrated by Telenti et al. in 1993 (13). These authors used PCR-restriction enzyme pattern analysis (PRA) of an amplified 439-bp segment of the 65-kDa heat shock protein (hsp-65) gene and introduced the abbreviation PRA for this method that has now gained wide acceptance (16). Application of this methodology has since been expanded to include 50 commonly encountered pathogenic species and taxa of aerobic actinomycetes comprising the genera Mycobacterium (10, 13); Nocardia (9, 17, 22); and Actinomadura, Gordona, Rhodococcus, Streptomyces, and Tsukamurella (11). The rapidity and accuracy of PRA prompted the current study (9–11, 13, 17). Clinical isolates of aerobic actinomycetes submitted to the Mycobacteria/Nocardia Laboratory at the University of Texas Health Center at Tyler (UTHCT) for identification and susceptibility testing were subjected to PRA for identification in an effort to evaluate the efficacy and cost-effectiveness of this methodology for routine clinical use.

(This study was presented in part at the 97th General Meeting of the American Society for Microbiology, Miami Beach, Fla., 1997.)

MATERIALS AND METHODS

Organisms.

The present study included 293 clinical isolates of aerobic actinomycetes submitted to the Mycobacteria/Nocardia Research Laboratory at the UTHCT for identification and susceptibility testing during the 27-month period from March 1995 through May 1997. All clinical isolates used in this study were subcultured onto Trypticase soy and Middlebrook 7H10 agar plates. One culture of each isolate was used for identification by colonial morphology and antimicrobial susceptibility patterns (1, 17–21). Selected biochemical testing was done in order to differentiate species or taxa with similar susceptibility patterns (6, 10, 14, 15, 17, 19, 21). High-performance liquid chromatography (HPLC) and additional biochemical testing were kindly performed by the Bureau of Laboratories, Texas State Health Department (Austin).

The ATCC type strains of Mycobacterium fortuitum (ATCC 6841) and Nocardia brasiliensis (ATCC 19296) were utilized as internal controls for PRA.

PCR amplification.

DNA was prepared from cells harvested from the initially submitted agar slants and/or the second subculture, when necessary, according to methods previously described (9–11, 13). A 439-bp segment of the hsp-65 gene was amplified from ground cell supernatants by PCR with 1.0 U of Taq DNA polymerase (Boehringer Mannheim, Indianapolis, Ind.) in optimized buffer E (1.5 mM MgCl₂ [pH 9.0]; Invitrogen, San Diego, Calif.) containing 83 μM (each) deoxynucleoside triphosphates, 9% dimethyl sulfoxide, and 1 μM (each) primers TB11 (5′-ACCAACGATGGTGTGTCCAT) and TB12 (5′-CTTGTCGAACCGCATACCCT) (Midland Certified Reagent Co., Midland, Tex.), together with the appropriate positive and negative controls according to a modification of the method of Telenti et al. (13). The PCR mixtures were run for 45 cycles of 94, 55, and 72°C for 1 min each and then for a 10-min extension period at 72°C.

Restriction enzyme analysis.

Data from previous studies (9–11, 13) resulted in the selection of five commercially available restriction endonucleases, BstEII, HaeIII, MspI, HinfI, and BsaHI (New England Biolabs, Beverly, Mass., and Promega, Madison, Wis.), for routine use in the production of PRA band patterns. When indicated (references 10 and 22 and unpublished data), one or more of a secondary set of endonucleases that included AciI, HhaI, and NarI was utilized. Restriction digests were incubated for the appropriate time periods, at the appropriate temperatures, and with the buffers recommended by the manufacturers, with the exception of the temperature and digest mixture for BsaHI. To achieve complete digestion with BsaHI, acetylated bovine serum albumin was substituted for bovine serum albumin and the digestion mixture was incubated at 60°C for 1 h.

Restriction fragments were electrophoresed on 3% Metaphor agarose (4-bp resolution; FMC Bioproducts, Rockland, Maine), containing ethidium bromide (0.625 μg/ml), in a Mini-Sub-Cell electrophoresis system (Bio-Rad Laboratories, Richmond, Calif.) at 95 V for 1.5 to 2.0 h.

Isolate identification.

PRA band sizes (base pairs) from each isolate were estimated visually by comparison with a 100-bp ladder (Life Technologies, Grand Island, N.Y.), a pGEM base pair ladder (Promega), and the PRA patterns obtained for control strains on each gel. Each isolate was then initially identified by one member of the staff (R.W.W.) to the species or taxon level by comparison of visually estimated PRA band sizes with those of species- and taxon-specific patterns contained in the PRA database (9–11, 13, 17, 22). Visual PRA isolate identifications were made prior to, and without knowledge of, identification results by traditional methods.

The PRA database was developed primarily for clinically significant nonpigmented rapidly growing mycobacteria, nocardiae, and other clinically significant aerobic actinomycetes (9–11, 13). With the exception of Mycobacterium avium and Mycobacterium intracellulare, for which 83 and 129 isolates, respectively, have been studied by PRA, the entries in the database for slow-growing mycobacteria have been less extensively developed. This portion of the database (unpublished data) represented PRA patterns obtained from 2 to 10 isolates of each slow-growing mycobacterial species most commonly encountered in clinical samples, including Mycobacterium celatum, Mycobacterium kansasii, Mycobacterium scrofulaceum, and Mycobacterium triviale, in addition to the 10 species listed in Table 1. These patterns were very similar to those published by Telenti et al. (13). To date, the PRA database has not been expanded to include pigmented rapidly growing mycobacterial species owing to their predominantly environmental origin and infrequent clinical occurrence as agents of traumatic wound infections.

TABLE 1.

Clinical isolates identified by PRA in this study

Sp. and taxon	No. of clinical isolates
Genus Mycobacterium
Rapidly growing spp.
M. abscessus	47
M. abscessus-M. chelonae complex	4
M. chelonae	18
M. fortuitum	32
M. fortuitum, third biovariant complex	5
M. fortuitum, third biovar, sorbitol-positive– M. peregrinum Pipr complex	1
M. mucogenicum (MCLO)a	20
M. smegmatis	10
Slow-growing spp.
M. avium	3
M. gordonae	1
M. haemophilum	2
M. intracellulare	4
M. marinum	7
M. simiae	2
M. szulgai	1
M. terrae-M. nonchromogenicum complex	2
M. tuberculosis	3
Mycobacterium sp. (BstEII sites present; taxon not distinguishable)	8
Mycobacterium subtotal	170
Genus Nocardia
Nocardia asteroides complex
Type Ib	9
Type II	1
Type VI	17
Taxon not further differentiated	1
N. brasiliensis	16
N. farcinica	11
N. nova	24
N. otitidiscaviarum	3
N. pseudobrasiliensisc	2
N. transvalensis complex	9
Nocardia subtotal	93
Genera of other aerobic actinomycetes
Actinomadura madurae	1
Gordona sputi	2
Gordona sp.	1
Rhodococcus equi	1
Streptomyces sp.	10
Tsukamurella paurometabolum	4
Other aerobic actinomycete taxa (BstEII sites absent; taxon not distinguishable)	11
Actinomycete subtotal	30
Total clinical isolates	293

^aDescribed by Springer et al. (8). MCLO, M. chelonae-like organism. 

^bAntibiogram types of N. asteroides complex were described by Wallace et al. (20). 

^cDescribed by Ruimy et al. (7) and Wallace et al. (17). 

Species or genus (e.g., Streptomyces) identification was considered conclusive when the PRA pattern matched that of a single species or taxon of aerobic actinomycete in the PRA database and the resulting identification was in agreement with that based on traditional methods.

To test whether precise PRA band size measurements were required for accurate isolate identification, band sizes were measured on a computerized Bio Image system (Millipore, Bedford, Mass.) with the same molecular size standards as noted above, and isolate identifications were made independently by a different member of the staff (V.A.S.) without prior knowledge of the initial visual identifications (R.W.W.). The two independent PRA identifications were then compared to one another and with identifications based on traditional methods including growth characteristics, susceptibility patterns, biochemical tests, and HPLC analysis.

Time studies of the PRA identification protocol were performed in order to provide an estimate of the amount of time and labor involved in applying this methodology under routine clinical conditions.

RESULTS

Organisms.

On microbiological evaluation, six of the cultures submitted to the UTHCT laboratory for identification and susceptibility testing were not aerobic actinomycetes. These cultures did not yield PCR amplification products (amplicons) and were excluded from further study. The distribution of species and taxa identified among the remaining 293 isolates is listed in Table 1. The seven predominant species of aerobic actinomycetes identified in this study comprised over half, 174 of 293 (59%), of the isolates submitted and included the following: Mycobacterium abscessus, 47 of 293 (16%) isolates; M. fortuitum, 32 of 293 (11%) isolates; Nocardia nova, 24 of 293 (8%) isolates; Mycobacterium mucogenicum (formerly Mycobacterium chelonae-like organism) (8), 20 of 293 (7%) isolates; M. chelonae, 18 of 293 (6%) isolates; Nocardia asteroides type VI (20), 17 of 293 (6%) isolates; and N. brasiliensis, 16 of 293 (5%) isolates.

PRA identification.

As shown in Table 1, 274 of the 293 (93.5%) isolates produced PRA patterns that matched species- or taxon-specific patterns in the PRA database and resulted in an identification that correlated with the identification by traditional methods. The remaining 19 isolates produced PRA patterns that did not match any of the patterns currently available in the database. Eight of these isolates produced amplicons that contained BstEII recognition sites and were therefore tentatively identified as Mycobacterium species (9–11). Only 4 of these 19 isolates were unequivocally identified to the species level by traditional methods, as shown in Table 2. One isolate (Mo 816) produced an amplicon lacking BstEII recognition sites and an HaeIII pattern resembling that published for Mycobacterium vaccae (13) but was not identified as such due to the lack of adequate data in the PRA database at the time that the isolate was received. A second unique pattern was observed for isolate N 1426, which was identified as Nocardia sp., most likely N. asteroides complex, and which produced an amplicon containing BstEII recognition sites. This sputum isolate gave unique PRA patterns with all enzymes tested and was the first and only Nocardia isolate, among 210 Nocardia isolates examined by PRA in this laboratory, that demonstrated BstEII recognition sites. Traditional methods unequivocally identified 278 of the 293 (94.9%) clinical isolates studied, while PRA correctly identified 274 isolates, resulting in a comparative accuracy of 98.6%.

TABLE 2.

Identification of clinical isolates that exhibited patterns that did not match those in the current PRA database

Isolate no.	Identificationa
Isolates distinguished as Mycobacterium sp. based on the presence of BstEII recognition sitesb,c
Mo 702	Mycobacterium sp. (pigmented rapid grower)
Mo 748	Mycobacterium sp. (pigmented rapid grower)
N 1426	Nocardia sp.
Mo 783	M. simiae
Mo 821	Mycobacterium sp. (pigmented rapid grower)
Mo 830	Mycobacterium sp. (slow grower)
Mo 855	Mycobacterium sp. (pigmented rapid grower)
Mo 875	Mycobacterium sp. (slow grower)
Isolates distinguished as aerobic actinomycetes, other than Mycobacterium sp., based on the absence of BstEII recognition sitesb,d
Mo 728#1	Corynebacterium aquaticum
Mo 728#2	Actinomyces viscosus
N 1420	Unidentifiable; possible N. transvalensis
Mo 780	Rhodococcus sp.
Mo 791	Gordona sp.
Mo 798	N. asteroides complex
Mo 802	Tsukamurella sp.
As 121	Streptomyces sp.
As 124	N. asteroides complex
Mo 816	M. vaccae
Mo 820	Unidentifiable; possible Mycobacterium sp. (slow grower)

^aIdentification by biochemical tests, antimicrobial susceptibility patterns, and HPLC analysis. 

^bWork of Steingrube et al. (9, 11). 

^cTotal of 8. 

^dTotal of 11. 

Three cultures that contained small numbers of gram-positive rods mixed with heavy overgrowth of gram-positive cocci were submitted. Direct PRA was carried out on cells taken from the mixed cultures submitted, and in all three cases, the amplicons obtained produced patterns that were typical of M. mucogenicum. This identification was later confirmed by traditional microbiological methods. Two additional cultures that appeared to be pure produced an excessive number of bands on PRA gels that gave total base pair values in excess of the expected 439 bp. One of these cultures was identified at a reference laboratory as containing both Corynebacterium aquaticum and Actinomyces viscosus (Table 2). The second mixed culture from a multiply infected wound site yielded two distinct PRA patterns that, when reanalyzed on individual colony picks, were typical for isolates of the N. asteroides complex antibiogram type I (20) and the Nocardia transvalensis new taxon 2 of the proposed N. transvalensis complex (9, 22).

Comparison of visual PRA band size estimates and isolate identifications from gel photographs (Fig. 1) with independent isolate identifications based on computer-generated measurements of PRA band sizes resulted in 100% agreement between the two identification methods. Comparison of measured PRA band size (base pair) values from Fig. 1 with published values for the matching species and taxon (9–11) corroborated the visual estimations of PRA band sizes and the identification of the clinical isolates. For example, clinical isolate 96–113 gave PRA band patterns of 235, 115, and 80 bp with BstEII (Fig. 1A, lane 4) and 145 and 125 bp with HaeIII (Fig. 1A, lane 8), which matched the published PRA patterns of M. fortuitum (10). These patterns also matched those of the control strain (ATCC 6841) of M. fortuitum shown in Fig. 1A, lanes 5 and 9, respectively. Likewise, clinical isolates 96–110, 96–111, and 96–112 were identified by matching their PRA band patterns from BstEII, MspI, HinfI, and BsaHI digests with those published (9, 11) for N. nova, N. asteroides complex antibiogram type I, and N. brasiliensis, respectively.

FIG. 1.

PRA patterns from BstEII, HaeIII, MspI, HinfI, and BsaHI digests of amplicons from clinical isolates and reference strains of aerobic actinomycetes. (A) Lanes 1 to 5, BstEII digests; lanes 8 and 9, HaeIII digests. Amplicon digests: lanes 1 to 5, clinical isolates 96–110, 96–111, 96–112, and 96–113 and reference strain ATCC 6841, respectively; lanes 6 and 7, size markers (100-bp and pGEM-bp ladders, respectively); lanes 8 and 9, clinical isolate 96–113 and reference strain ATCC 6841, respectively. (B) Lanes 1 to 4, MspI digests; lanes 5 to 8, HinfI digests; lanes 11 to 14, BsaHI digests. Amplicon digests: lanes 1 to 4, clinical isolates 96–110, 96–111, and 96–112 and reference strain ATCC 19296, respectively; lanes 5 to 8, clinical isolates 96–110, 96–111, and 96–112 and reference strain ATCC 19296, respectively; lanes 9 and 10, size markers (100-bp and pGEM-bp ladders, respectively); lanes 11 to 14, clinical isolates 96–110, 96–111, and 96–112 and reference strain ATCC 19296, respectively. Clinical isolates 96–110, 96–111, 96–112, and 96–113 were identified as N. nova, N. asteroides complex antibiogram type I (20), N. brasiliensis, and M. fortuitum, respectively, by matching each of their PRA band patterns with published values (9–11).

A time course study was performed with the PRA procedure in order to better define the amount of time and labor involved and to provide a stepwise description of the procedure (Table 3). Final identification results can be achieved within 24 h of receiving a culture under optimal conditions. As a routine practice, however, separate analysis of individual isolates was neither cost-effective nor practical. As a practical routine, identification results based on PRA were generally achievable on a 2- to 5-day schedule.

TABLE 3.

Timed protocol for aerobic actinomycete clinical isolate identification by PRA^a

PRA step no.	PRA procedure	Time required (h)
1	Cell harvest, suspension in buffer, heat kill, grind in Mini-Beadbeater	1.5
2	Set up PCRs and program Temp-cycler	1
3	PCR amplification in Temp-cycler with 4°C holding mode	Overnight
4	Preparation of four agarose gels (one for step 5 and three for step 8)	1
5	Gel loading and electrophoresis of PCR products (amplicons)	1.75
6	Set up restriction endonuclease digests (three restriction endonucleases and up to 36 digests)	1
7	Incubation of restriction endonuclease digests	1
8	Gel loading and electrophoresis of restriction endonuclease digests (three gels)	2–2.5
9	Photograph gels, evaluate PRA patterns, and identify clinical isolates	2.25

^aTime required to analyze 12 clinical isolates per procedure. 

DISCUSSION

PRA correctly identified 274 of 293 (93.5%) aerobic actinomycete clinical isolates, compared to traditional identification methods that unequivocally identified 278 of those isolates (94.9%). Comparison of the two identification methods resulted in a 98.6% relative accuracy for the molecular biological identification method. PRA proved highly specific, enabling identification of aerobic actinomycetes from cultures heavily contaminated with other bacteria as well as identifying mixed cultures that contained more than one taxon of aerobic actinomycete that was not readily recognizable on isolation plates. The specificity of PRA was further demonstrated with six cultures that failed to yield amplicons on PCR and were found not to contain aerobic actinomycetes on further microbiological evaluation.

Identification of species and taxa of aerobic actinomycete isolates commonly encountered in clinical specimens was accomplished with virtually 100% accuracy. Only very rarely occurring species presented difficulties for identification, most frequently as a result of the unavailability of reference patterns in the PRA database. As noted above, the current PRA database has been well developed for species and taxa of nonpigmented rapidly growing mycobacteria, nocardiae, and other clinically significant aerobic actinomycetes (9–11). PRA data have not been as well developed for species of slow-growing mycobacteria, and to date, data from PRA of pigmented rapidly growing mycobacteria remain to be developed. As this study has progressed over the past 2 years, data for less frequently occurring species have been constantly added to the database as sufficient numbers of isolates and reference strains have been accumulated and examined by PRA. This steady development of the scope and breadth of the PRA database should continually expand the number of aerobic actinomycete species identifiable by this method.

Seven categories listed in Table 1 appear as complex or genus identifications and warrant further consideration. The M. abscessus-M. chelonae complex; M. fortuitum third biovariant complex; M. fortuitum third biovar, sorbitol-positive–Mycobacterium peregrinum Pip^r complex (10); and the Mycobacterium terrae-Mycobacterium nonchromogenicum complex each represent two groups containing very closely related mycobacterial taxa that are not readily differentiated by either traditional methods or PRA (10, 13). Three isolates listed as M. abscessus-M. chelonae complex could not be differentiated further by traditional methods and exhibited hybrid PRA patterns that contained features of both M. abscessus and M. chelonae. These isolates may represent a heretofore-unrecognized taxon closely related to these mycobacterial species. The isolate listed as an undifferentiated taxon of the N. asteroides complex gave patterns with MspI and BsaHI that resembled those of the N. asteroides complex antibiogram type I (20) but produced a pattern with HinfI that was not represented in the PRA database and, therefore, may represent a minor pattern for isolates within this taxon, similar to the occurrence of major and minor patterns previously observed for isolates of this taxon with BsaHI (9). The single isolate listed as Gordona sp. exhibited a PRA pattern with HinfI that was unique to this genus but gave patterns with all other endonucleases tested that matched those of the previously reported isolate Mo 315 (11). These two isolates could be identified only to the genus level at the Centers for Disease Control and Prevention (Atlanta, Ga.) and were neither Gordona bronchialis nor Gordona sputi. As previously discussed (11), identification of Streptomyces isolates beyond the genus level was considered to be clinically irrelevant and was not pursued in this study.

Identification of clinical isolates of aerobic actinomycetes by PRA, performed on a continuing daily basis common to the routine of clinical laboratories, provided final identification results within 2 to 5 working days, compared to traditional identification methods that required from 2 to 6 weeks for final results. Prompt and accurate identification of pathogenic aerobic actinomycete isolates is particularly important when invasive species such as Nocardia pseudobrasiliensis (7, 17) are encountered or when innately drug-resistant species such as those of the N. transvalensis complex, which are resistant to all aminoglycosides (9, 20, 22), and Nocardia farcinica, which is resistant to all extended-spectrum cephalosporins (12, 21), are involved. This is particularly significant when immunocompromised patients such as those with advanced human immunodeficiency virus disease are infected with these organisms (4, 5, 9, 17, 22).

Although absolute PRA band sizes have been found to vary from 5 to 10 bp between laboratories (10, 13), the overall patterns have proven highly reproducible and species- and taxon-specific. Visual comparison of PRA patterns with molecular size standards and patterns produced by internal control isolates, such as M. fortuitum and N. brasiliensis reference strains used in this study, resulted in successful clinical isolate identifications that correlated perfectly with identifications based on measured band size values. Consequently, there is no requirement for costly computerized measurement systems, a major expense consideration, in implementing this methodology for routine clinical use.

PRA was cost-effective, with the expenses of specialized equipment and reagents being more than compensated for by savings in time and labor, and could be economically incorporated into the clinical laboratory setting. This methodology has proven both practical and cost-effective as a rapid, efficient, and highly accurate identification system for use in identifying clinically significant species and taxa of aerobic actinomycetes.