Evaluation of the Integrated Database Network System (IDNS) SmartGene Software for Analysis of 16S rRNA Gene Sequences for Identification of Nocardia Species

Abstract

16S rRNA gene sequences of 102 Nocardia isolates were analyzed using the Integrated Database Network System (IDNS) SmartGene centroid database. A total of 76% of the isolates were correctly identified. Discordant identifications were due to inadequate centroid length (3 species), inaccurate or insufficient entries in the public databases (5 species), and heterogeneous sequences among members of a species (1 species).

Nocardia species are significant human pathogens, especially in immunocompromised patients. Accurate species assignment is important to guide appropriate antibiotic selection, as several species possess high levels of antibiotic resistance (9, 10). 16S rRNA gene sequencing has become the most widely used method for the identification of these organisms.

Noncurated public databases contain a wealth of sequence information but also include a significant amount of information that is compromised by identification errors and low-quality sequence data (8). Searches against such databases and analysis of the results have become increasingly time-consuming, due to the rapidly growing number of submitted sequences. A quality-controlled sequence database that eliminates inaccurate and redundant entries and identifies the best representative sequence for a particular species would facilitate rapid and accurate identification.

SmartGene (SmartGene GmbH, Zug, Switzerland) is a web-based sequence search tool that allows comparison of input sequences with those in two curated databases. Using a proprietary filtering algorithm, the SmartGene 16S rRNA eubacteria database is assembled using sequences deemed acceptable from the public databases. A search using the eubacteria database allows the user to examine the inter- and intraspecies variabilities of a particular species within the SmartGene platform. An additional database, the centroid database, is prepared using a supplementary proprietary algorithm that examines all sequences of a particular species in the eubacteria database and creates a discrete “species group” of sequences for that species. The most representative sequence of that species is designated the “centroid” sequence. Only one sequence per species (the centroid sequence) is included in the centroid database; this sequence may or may not be the type strain of the species, depending on the heterogeneity of that species. Output from a search of the centroid database with an unknown sequence will show the most similar centroid sequences sorted by match score or by the number of base mismatches. The results will also indicate the number of sequences in the centroid group.

A study comparing the use of a previous version of the SmartGene software with the use of conventional identification and with that of another proprietary sequence database found SmartGene to be a time- and labor-saving system that provided a larger percentage of accurate genus or species assignments for a diverse group of bacteria than did the proprietary database (7).

We report here a study comparing the usefulness of the SmartGene centroid and eubacteria databases for accurate species-level identification of a variety of clinically relevant Nocardia isolates. We also include a careful analysis of discrepant results.

Nocardia isolates.

The type strains of 37 Nocardia species and 65 clinical isolates of Nocardia, representing 22 different species, were examined. The clinical isolates were recovered from patients being treated at the Warren Grant Magnuson Clinical Center of the NIH or were referred to the NIH for identification.

Nocardia identification.

All clinical Nocardia isolates were identified by 16S rRNA gene sequence analysis (approximately 1,400 bp) using a BLAST search and by comparison to isolates from an in-house database. In addition, the identification of 49 of the clinical isolates was confirmed by secA1 gene sequencing, as previously described (5).

16S rRNA gene sequence analysis.

Organisms were cultivated, and DNA was extracted as described previously (4). Three overlapping regions of the 16S rRNA gene were amplified and sequenced using primers with tails containing M13 binding sites for reactions 1 and 3, as previously described (3, 4). The primers used for reaction 2 were modified as follows (the M13 binding sites are in boldface): Reaction 2 Forward, 5′-GTA-AAA-CGA-CGG-CCA-GCA-ACT-ACG-TGC-CAG-CAG-CCG-C-3′; Reaction 2 Reverse, 5′-CAG-GAA-ACA-GCT-ATG-ACT-GAC-GAC-AGC-CAT-GCA-CCA-CCT-3′. PCR was performed as previously described (3, 4), except for the addition of 0.4 μg/ml bovine serum albumin to the PCR mix. The PCR amplification products from all regions were purified from gel bands, as previously reported (3), or with the Qiaex II kit (Qiagen, Valencia, CA). Cycle sequencing and sequence analysis were performed as previously reported (3).

SmartGene evaluation.

16S rRNA gene sequences of Nocardia type strains and clinical isolates (approximately 1,400 bp) were individually analyzed using the Integrated Database Network System (IDNS) centroid and 16S rRNA gene eubacteria databases, with results sorted by identities/mismatches. A percent similarity of ≥99.8% (value rounded off) to a centroid sequence (IDNS centroid database) or type strain sequence (eubacteria database) was considered an acceptable identification. This cutoff value was chosen based on our experience with Nocardia; 16S rRNA gene sequences of isolates belonging to a particular species usually are ≥99.8% similar (2). To assess the use of the centroid database with partial 16S rRNA gene sequences, 500-bp sequences (from the 5′ end of the 16S rRNA gene) of all Nocardia type strains were also analyzed.

Results and discussion.

For 28 of 37 (76%) Nocardia type strains whose 16S rRNA gene sequences were tested using 1,400-bp regions against the centroid database, the closest acceptable centroid was the correct species for that sequence (Table 1). For 6 of 28 species, the centroid database gave more precise results than the eubacteria database, as invalid species or sequences with incorrect identifications present in the eubacteria database were eliminated in the centroid database (Table 1). The results obtained with the clinical isolates of each species were identical to the results obtained with the type strains of those species. Standard BLAST search results were essentially identical to those obtained with the eubacteria database (data not shown).

TABLE 1.

Nocardia species giving acceptable identification with the centroid database^d

Species	Strain	No. of possible IDs meeting similarity threshold in databaseb:		No. of:
		Centroid	Eubacteria	Isolates in centroid group	Clinical isolates tested
N. abscessusa	DSM 44432	1	2	19	5
N. africanaa	DSM 44491	1	1	7	0
N. anaemiaea	DSM 44821	1	1	1	1
N. araoensisa	DSM 44729	1	1	2	0
N. arthritidisa	DSM 44731	1	1	8	3
N. asiaticaa	DSM 44668	1	1	13	2
N. asteroides sensu strictoa	ATCC 19247	1	2	37	1
N. brasiliensisa	ATCC 19296	1	1	16	6
N. brevicatenaa	ATCC 15333	1	1	6	0
N. carneaa	ATCC 6847	1	1	9	0
N. elegansa	JCM 13374	1	1	5	1
N. exalbida	DSM 44883	1	1	2	0
N. farcinica	ATCC 3318	1	2	31	6
N. higoensisa	DSM 44732	1	1	2	0
N. inohanensis	DSM 44667	1	1	4	0
N. mexicana	DSM 44952	1	1	4	2
N. niigatensis	DSM 44670	1	1	6	1
N. nova	ATCC 33726	1	2	39	6
N. paucivorans	ATCC BAA-278	1	1	8	0
N. pseudobrasiliensisa	ATCC 51512	1	1	14	4
N. puris	DSM 44599	1	1	6	3
N. sienataa	DSM 44766	2c	2c	1	0
N. terpenicaa	DSM 44935	1	2	2	0
N. testaceaa	DSM 44765	2c	2c	3	2
N. thailandicaa	DSM 44808	1	1	2	1
N. vermiculataa	DSM 44807	1	2	1	0
N. veterana	DSM 44445	1	1	11	4
N. vinaceaa	JCM 10988	1	1	7	1

^aCentroid is the sequence of the type strain of the species.

^bBased on percent similarity of ≥99.8%. IDs, identifications.

^cN. sienata and N. testacea are known to be 99.9% similar (2 base mismatches).

^dUsing approximately 1,400-bp sequences. Clinical isolates gave results identical to those of the type strains.

Several discordant results were obtained using the centroid database and are summarized in Table 2. In the first error type (error type 1), three type strains were misidentified due to centroids with sequences shorter than those of the input sequences. As with other sequence matching software, one of the criteria for scoring matches is match length; a centroid with a sequence shorter than the input sequence receives a lower score than a longer (although incorrect) sequence. The selection of short centroids is related to the small number of available sequences in the public databases for some species; this error might be corrected by the selection of only full-length (or nearly full-length) sequences as centroid sequences, if such sequences exist.

TABLE 2.

Discordant results from searches of type strain sequences with the centroid database^c

Species	Strain	No. of:		Results
				Centroid database				Best match(es) in eubacteria database
		Clinical isolates tested	Isolates in centroid group	Best match(es)	% similarity of match	Source(s) of discrepant results	Error typeb
N. aobensis	DSM 44805	0	6	N. africana	99.2	Short centroid	1	N. aobensis
N. cyriacigeorgica	DSM 44484	5	52	N. abscessus	98.8	Short centroid	1	N. cyriacigeorgica
N. shimofusensisa	DSM 44733	0	4	N. higoensis	99.2	Short centroid	1	N. shimofusensis
N. transvalensis	ATCC 6865	0	28	N. blacklockiae	98.8	Incorrect centroid	2A	N. transvalensis
				N. wallacei	98.6
				N. transvalensis	98.6
N. wallaceia	ATCC 49873	4	2	N. wallacei	100	Incorrect centroid for N. transvalensis	2B	N. wallacei
				N. transvalensis	99.9
N. beijingensis	DSM 44636	4	24	N. beijingensis	100	Probable incorrect identification of S. gardneri in public database	2B	N. beijingensis
				S. gardneri	99.9			S. gardneri
N. ninaea	DSM 44978	0	1	N. ninae	99.7	Few sequences in public database; errors in those sequences	2C	N. ninae
N. pneumoniaea	DSM 44730	1	2	N. pneumoniae	99.4	Few sequences in public database; errors in those sequences	2C	N. pneumoniae
N. otitidiscaviarum	ATCC 14629	2	24	N. otitidiscaviarum	99.7	Heterogeneous sequences in species group	3	N. otitidiscaviarum

^aCentroid is the sequence of the type strain of the species.

^bError type 1, short sequence selected as centroid; error type 2, inaccurate or insufficient entries in the public database (error types 2A and 2B, inaccurate information in the public database; error type 2C, possible sequence errors in type strains); error type 3, heterogeneous sequences of members of a species.

^cClinical isolates have results identical to those of the type strains.

The second error type resulted in discordant results for 5 species. These errors were due to inaccurate or insufficient entries in the public database. The first example (error type 2A) resulted in erroneous identification of the Nocardia transvalensis type strain and is a direct result of inaccurate information in the public database. The centroid for N. transvalensis is based on numerous Nocardia wallacei sequences erroneously identified as N. transvalensis in the public database; the N. transvalensis centroid sequence is identical to the sequence of the N. wallacei type strain and is only 98.6% similar to a reliable sequence of the type strain of N. transvalensis.

The second example of discordant results due to inaccurate or insufficient entries in the public database resulted in the return of centroids of 2 species within the acceptable percent similarity range for searches with N. wallacei and Nocardia beijingensis (error type 2B). The search with the sequence of the type strain of N. wallacei returned identifications of both N. wallacei and N. transvalensis; this error is the result of the inaccurate centroid for N. transvalensis, as noted above. The centroid database search for the type strain of N. beijingensis returned centroids of N. beijingensis and Streptomyces gardneri. The S. gardneri sequence is most likely an incorrect identification in the public database (Table 2).

In the third example of discordant results due to erroneous or insufficient entries in the public database (error type 2C), the top matches in the searches using the sequences of Nocardia ninae and Nocardia pneumoniae were of the correct species but were <99.8% similar to the sequences of the centroid, resulting in an inadequate identification. There are few entries for these species in the public database; the centroids selected for these species were derived from sequences containing some possible sequence errors, as determined by comparison of those sequences with reliable recent GenBank sequence submissions for the same type strains. This illustrates the need for examining the number of sequences included in the species group when analyzing ambiguous results with the centroid database. The developers of this software can eliminate these types of discrepancies by continually monitoring public databases for new entries and modifying centroid sequences as necessary.

The third error type was related to the heterogeneity of sequences assigned to the species Nocardia otitidiscaviarum (error type 3). There are numerous sequences in the public databases for this species, which has been shown to have a high level of intraspecies heterogeneity (6). A sequence different from that of the type strain was chosen as the centroid for that species (Table 2), resulting in an identification for the type strain with a percent similarity that was below the customary threshold.

For all but one of the nine discordant results, searches using the eubacteria database (Table 2) and GenBank (data not shown) returned acceptable identifications.

Because 500-bp 16S rRNA gene sequences are frequently used for bacterial identification, especially with commercial sequencing products, 500-bp sequences of all type strains were also tested using the centroid database. Using these shortened sequences, ambiguous results were obtained for 4 species groups comprised of 10 species. This is due to insufficient base heterogeneity in the first 500 bases of the 16S rRNA genes of some Nocardia species, resulting in inadequate discrimination of some species, as has been previously reported (1). Ambiguous identifications were obtained for the following Nocardia species groups: N. abscessus/N. asiatica/N. arthritidis, N. africana/N. elegans/N. veterana, N. exalbida/N. gamkensis, and N. higoensis/N. shimofusensis.

The SmartGene centroid database has the potential to offer a significant improvement over public databases for assigning Nocardia species identifications based on 1,400-bp sequences of the 16S rRNA gene. Many of the centroids selected for Nocardia species are sequences identical or nearly identical to those of the type strain of the species. The advantage of the centroid approach is that only one entry per species is presented; the search output is relatively short and easily evaluated for sequence similarity. In addition, sufficient information is provided to allow a user with limited experience in Nocardia taxonomy to arrive at a correct identification as long as certain minimal standards of sequence similarity are achieved. For the Nocardia isolates tested using the centroid database, the lowest percentage of similarity indicating conspecificity may be as low as 99.4% (Table 2), although care should be used in the evaluation of species that show <99.8% similarity to a given species. More experienced users may find the eubacteria database useful for examining the intraspecies variability of a particular species.

Continued monitoring of public databases by SmartGene developers and the input of nearly full-length 16S rRNA gene sequences by the user can easily remedy the difficulties encountered with the centroid database for the identification of Nocardia species. This system represents a useful method for sequence comparison by laboratorians at all levels of experience with sequence interpretation.