Rapid Identification of Bacteria from Positive Blood Cultures by Terminal Restriction Fragment Length Polymorphism Profile Analysis of the 16S rRNA Gene

Clinical Research Center, Marshfield Clinic Research Foundation, Marshfield, Wisconsin 54449

Received 24 February 2003/ Returned for modification 13 April 2003/ Accepted 24 May 2003

	ABSTRACT

Top Abstract Introduction Materials and Methods Results Discussion References

 
Bacteremia results in significant morbidity and mortality, especially among patient populations that are immunocompromised. Broad-spectrum antibiotics are administered to patients suspected to have bloodstream infections that are awaiting diagnosis that depends on blood culture analysis. Significant delays in identification of pathogens can result, primarily due to the dependence on growth-based identification systems. To address these limitations, we took advantage of terminal restriction fragment (TRF) length polymorphisms (T-RFLP) due to 16S ribosomal DNA (rDNA) sequence diversity to rapidly identify bacterial pathogens directly from positive blood culture. TRF profiles for each organism were determined by sizing fragments from restriction digests of PCR products derived from two sets of 16S rDNA-specific fluorescent dye-labeled primers. In addition, we created a TRF profile database (TRFPD) with 5,899 predicted TRF profiles from sequence information representing 2,860 different bacterial species. TRF profiles were experimentally determined for 69 reference organisms and 32 clinical isolates and then compared against the predicted profiles in the TRFPD. The predictive value of the profiles was found to be accurate to the species level with most organisms tested. In addition, identification of 10 different genera was possible with profiles comprising two or three TRFs. Although it was possible to identify Enterobacteriaceae by using a profile of three TRFs, the similarity of the TRF profiles of these organisms makes differentiation of species less reliable with the current method. The ability to rapidly (i.e., within 8 h) identify bacteria from blood cultures has potential for reducing unnecessary use of broad-spectrum antibiotics and promoting more timely prescription of appropriate antibiotics.

	INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion Reerences

 
The detection and identification of bacteria from the blood of patients is one of the most important roles of the clinical microbiology laboratory. Turnaround times for positive results are important because prompt, appropriate treatment of bloodstream infections reduces morbidity and mortality (9). In addition, hospital-acquired infections greatly increase health care costs due to the added expense associated with prolonged hospital stays and antimicrobial therapy (14). Over the past several decades considerable effort has been expended on developing faster and more sensitive methods to detect bacteremia from clinical specimens. This has resulted in the availability of manual and automated systems that have good sensitivity for the initial detection of the most commonly encountered organisms that cause bacteremia.

One limitation that still remains in processing blood cultures is the requirement to subculture positive blood culture bottles in order to perform most biochemical or other tests needed for bacterial identification. This can result in one or more days of delay during the identification process and concurrent treatment with broad-spectrum antimicrobial agents, including the unnecessary burden of antibiotic prescription for coagulase-negative staphylococci contaminants. Additionally, subculture of fastidious organisms may fail to grow on solid media or there may be insufficient growth to allow identification and susceptibility testing with standard methods. These problems would be circumvented if rapid and accurate molecular methods were available for the direct identification of bacteria from positive blood cultures.

A number of procedures have been reported to be successful in the direct identification of the more commonly encountered bacteria from positive blood cultures. Early reports involved biochemical or immunologic methods to identify common agents of bacteremia such as staphylococci, streptococci, and enterococci (4, 8, 12, 26, 33). The primary disadvantage of this approach is the inability to identify organisms less commonly encountered. This is especially problematic for laboratories serving populations that include significant numbers of immunosuppressed patients. More recently, methods such as fluorescent in situ hybridization (24), vibrational spectroscopy (20), single-strand conformation polymorphism (SSCP) analysis of the 16S rRNA gene (32, 34), and sequence-based methods (7, 10, 25) have been described. In general, these methods have the advantage of allowing identification of a broader spectrum of pathogens and have the potential to be at least partially automated in the clinical laboratory.

We describe here terminal restriction fragment (TRF) length polymorphism (T-RFLP) analysis for the identification of bacteria relevant to blood infections. The most commonly described use of T-RFLP is for systematic comparative community analyses of environmental samples (15, 21, 22), but one of the first reports described the identification of clinical strains of Mycobacterium (2). We have extended the utility of this technique by incorporating more than one fluorescent primer in each PCR, by using two nonoverlapping primer sets to amplify a total of 1,300 bases of the 16S rRNA gene, and by developing a searchable TRF profile database (TRFPD) populated with 101 experimentally determined TRF profiles and an additional 5,899 profiles predicted from 16S ribosomal DNA (rDNA) sequence data.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

 
The procedure for identification of bacteria by using TRF profiles is diagrammed in Fig. 1.

fig.ommitted FIG. 1. Outline of procedure for determination of TRF profiles.

 
Bacterial strains. For the present study we analyzed 69 reference organisms and 32 clinical blood isolates by TRF profiling (Table 1), including 40 of the most common bacterial species isolated from positive blood cultures according to Marshfield Laboratories incidence reports for January 1999 through July 2002. Marshfield Laboratories uses the BacT/Alert blood culturing system (Organon Teknika Corp., Durham, N.C.) and performs Gram staining for all positive blood cultures. All reference organisms and positive blood bottles were cultured on Trypticase soy broth-5% sheep blood or chocolate agar at 37°C under the appropriate atmospheric conditions. Blood bottle isolates were identified by using the automated Vitek System (bioMérieux Vitek, Inc., Hazelwood, Mo.) and standard biochemical methods.

fig.ommitted TABLE 1. Bacterial strains analyzed by TRF profiling

 
MICs for Staphylococcus aureus were determined by broth microdilution (Vitek System). Confirmation of oxacillin resistance was done by using Mueller-Hinton agar with 4% sodium chloride and 6 µg of oxacillin sodium/ml. Isolates were considered methicillin resistant if the MIC for oxacillin was 4 µg/ml and growth was present on oxacillin confirmatory media.

DNA extraction. DNA was extracted from cultured colonies by using a modification of the tissue protocol for the QIAmp DNA Mini kit (Qiagen, Inc., Valencia, Calif.). Briefly, 200 µl of a cell suspension in H₂O (containing ca. 5 to 10 mg of cells) was mixed with 200 µl of QIAmp Buffer AL and 10 µl of proteinase K (20 mg/ml). Samples were incubated for 15 min at 56°C and then for 15 min at 95°C. Then, 200 µl of ethanol was added and mixed by vortexing. The samples were loaded, washed, and eluted in 200 µl from QIAmp spin columns as indicated in the remainder of the manufacturer's protocol.

DNA was extracted from growth-positive blood culture bottles by using a modification of the QIAmp DNA Mini kit tissue protocol and incorporating a benzyl alcohol extraction (11). Briefly, 100 µl of blood culture was mixed with 100 µl of QIAmp Buffer AL and 10 µl of proteinase K (20 mg/ml). Samples were incubated for 15 min at 56°C and then for 15 min at 95°C. Samples were then diluted with 600 µl of filter sterile H₂O to reduce the viscosity. Then, 500 µl of benzyl alcohol was added to each sample, mixed by vortexing for 10 s, and centrifuged at 20,000 x g for 5 min. Next, 400 µl of the aqueous supernatant was transferred to a new 1.5-ml tube and 200 µl of ethanol was added and mixed by vortexing. The samples were loaded, washed, and eluted in 200 µl of buffer from QIAmp spin columns as indicated in the remainder of the manufacturer's protocol.

PCR primers. Independent reactions with two sets of fluorescent primers targeting the 16S rRNA gene were used for PCR with chromosomal preparations from reference strains and blood isolates. The primers were chosen based on previously published evaluations of domain specificity (1, 5, 6, 27, 35) and maximization of matches received from queries by using the T-RFLP analysis program (TAP T-RFLP) of the Ribosomal Database Project II (RDP-II) (19). As indicated, primers were labeled at the 5' end with the dyes 6-carboxyfluorescein (6FAM), 4,7,2',4',5',7'-hexachloro-6-carboxyfluorescein (HEX), or N,N,N',N'-tetramethyl-6-carboxyrhodamine (TAMRA). All primers were purchased from Sigma-Genosys (The Woodlands, Tex.).

The 5' region (encompassing hypervariable regions V1 to V3) of the 16S rRNA gene was amplified with S-D-Bact-0045-b-S-20 (5'-6FAM-GCY TAA CAC ATG CAA GTY GA) and S-D-Bact-0785-a-A-19 (5'-HEX-CTA CCA GGG TAT CTA ATC C). The 3' region of the 16S rRNA gene was amplified with S-D-Bact-0785-a-S-19 (5'-6FAM-GGA TTA GAT ACC CTG GTA G) and S-D-Bact-1371-a-A-20 (5'-HEX-AGG CCC GGG AAC GTA TTC AC). PCR products from the primer pairs were predicted to be 750 and 600 bp, respectively, with some variability depending on the organism from which the template DNA is extracted.

The primers for the detection of nuc, referred to here as TAMRA-nuc-S (5'-TAMRA-GCG ATT GAT GGT GAT ACG GTT) and nuc-A (5'-AGC CAA GCC TTG ACG AAC TAA AGC), were reported previously and correlated with coagulase phenotype in staphylococci (3, 17, 18). The predicted PCR product from the targeted template was 279 bp in length and did not contain an HhaI restriction site.

Primers for detection of the mecA gene of staphylococci, referred to here as TAMRA-mecA-S (5'-TAMRA-AAA ATC GAT GGT AAA GGT TGG C) and mecA-A (5'-AGT TCT GCA GTA CCG GAT TTG C), were previously reported and evaluated for accuracy of identification of methicillin resistance in staphylococci (18, 23). The predicted PCR product from the targeted template was 530 or 533 bp in length and contains an HhaI restriction site, resulting in a TAMRA-labeled fragment of 331 or 334 bases, respectively. The PCR amplification of the 3' region of the 16S rRNA gene (6FAM-labeled S-D-Bact-0785-a-S-19 and HEX-labeled S-D-Bact-1371-a-A-20), nuc, and mecA were accomplished in a triplex reaction, similar to previous reports (16, 18), under conditions described below.

Controls for terminal fragment size and restriction digest were prepared by using TAMRA-labeled S-D-Bact-0785-a-A-19 (5'-TAMRA-CTA CCA GGG TAT CTA ATC C) and unlabeled S-D-Bact-0045-b-S-20 (5'-GCY TAA CAC ATG CAA GTY GA). PCR was performed with chromosomal DNA preparations from four organisms (Haemophilus influenzae, Micrococcus luteus, Salmonella enterica serovar Enteritidis, and Stenotrophomonas maltophilia) for which the TRF sizes of the TAMRA-labeled products were known for each restriction enzyme (AluI, HhaI, MspI, and RsaI).

PCR conditions. The reaction mixtures for PCR contained 1x PCR buffer, 200 µM concentrations of each deoxynucleoside triphosphate, 3.0 mM MgCl₂, 0.5 µM concentrations of each primer species, 1.0 U of Taq DNA polymerase, and 1.0 µl of chromosomal preparation per 15 µl of reaction. Due to an 4-fold-higher detection sensitivity of the fluorescence emission from the 6FAM dye relative to the HEX dye, a ratio of one part 6FAM-labeled primer (0.125 µM) to three parts unlabeled primer (0.375 µM) was used for primers S-D-Bact-0045-b-S-20 and S-D-Bact-0785-a-S-19. DNA was amplified by using a model 9600 thermal cycler (Perkin-Elmer, Norwalk, Conn.) with the following program: 15 min at 94°C for denaturation and Taq activation; 35 cycles of denaturation (45 s at 94°C), annealing (30 s at 52°C), extension (60 s at 72°C); and a final extension for 5 min at 72°C.

Thermal gradient PCR with various magnesium concentrations was performed with reference strains to determine conditions that provided optimal product formation with minimal nonspecific products. All PCRs were analyzed by gel electrophoresis and stained with ethidium bromide to assure product formation prior to restriction digest.

Restriction digests of PCR products. PCR products were digested with restriction enzymes (Invitrogen, Carlsbad, Calif., and New England BioLabs, Beverly, Mass.) without further purification. Each 10-µl restriction digest contained 1 µl of PCR product, 0.5 µl of restriction enzyme, and 8.5 µl of prepared stock containing 2.5 ng/µl each appropriate digest/size control DNA in 1x buffer. The PCR product of 6FAM-labeled S-D-Bact-0045-b-S-20 and HEX-labeled S-D-Bact-0785-a-A-19 (amplifying the 5' region of the 16S rRNA gene) was digested in separate reactions with AluI (5'-AGCT-3'), HhaI (5'-GCGC-3'), MspI (5'-CCGG-3'), and RsaI (5'-GTAC-3'). The PCR product of 6FAM-labeled S-D-Bact-0785-a-S-19 and HEX-labeled S-D-Bact-1371-a-A-20 (amplifying the 3' region of the 16S rRNA gene), or the product(s) of the triplex reaction, were only digested with HhaI. Restriction digests were incubated for 2 h at 37°C, followed by treatment for 20 min at 65°C for enzyme inactivation.

16S rDNA TRF length analysis. The lengths of TRFs derived from amplified DNA products were determined by electrophoresis with a model 377 automated sequencer (Applied Biosystems Instruments, Foster City, Calif.). Sample were prepared by combining 2.0 µl of restriction digest product, 2.5 µl of deionized formamide, 0.5 µl of 25 mM EDTA (pH 8.0) containing 5% (wt/vol) blue dextran, and 0.50 µl of X-rhodamine MapMarker 1000 XL (BioVenture, Inc., Murfreesboro, Tenn.). The size standard contains single strands of DNA with a single ROX fluorophore at 50, 75, 100, 125, 150, 200, 250, 300, 350, 400, 450, 475, 500, 550, 600, 650, 700, 750, 800, 850, 900, 950, and 1,000 bases. Samples were mixed by pipetting, denatured at 94°C, and immediately cooled to 4°C. Aliquots of 1.0 µl were loaded onto a 36-cm, 4.75% denaturing polyacrylamide gel and electrophoresed at 51°C for 4.0 h with limits of 3 kV and 60 mA. Data were collected in Genescan mode, and the lengths of control and sample TRFs were calculated by comparison with the internal standard by using the Local Southern method (31). Fragments lengths of 20 but <50 bases were calculated by linear extrapolation from the migration times of the 50- and 75-base standards.

Construction of the TRFLP. A program and searchable TRFPD were developed to support calculation and comparison, respectively, of raw data files containing 16S rDNA gene TRF size information exported from Genescan analysis. The program performs mean and standard deviation calculations of control fragment lengths, providing data for quality control from every sample lane. The maximum-area 6FAM and HEX peak data for each lane are identified and organized according to the respective primer set and restriction digest combination. The program is then used to compare the sample TRF profile, within an adjustable search window for each fragment, against organism profiles in the TRFPD. The results from a Gram stain can also be selected in the data input window to limit the search. The result of a search includes the number of matching fragment lengths (within chosen windows), ordering of closest matches, and links to best match TRF profiles.

Four raw data files containing TRF sizes extracted for bacterial sequences with individual primer and restriction enzyme combinations (described above) were obtained by special request from the RDP-II (19). The search parameters we chose allowed as many as five mismatches in the first 15 bases from the 5' end of each primer. The data for in silico digest products derived from each primer were combined and sorted by accession number or other unique identifier. The sorted data set was then cleared of all organism entries that did not have restriction digest information for each of the four 16S rDNA-specific primers. The remaining profiles each contained 10 TRFs for 5,899 organisms, representing 2,860 unique species, and were used to populate the TRFPD.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

 
PCR products with 16S rDNA-specific primers. As expected, PCR products from (6FAM)-S-D-Bact-0045-b-S-20 and (HEX)-S-D-Bact-0785-a-A-19 were 750 bp, whereas products from (6FAM)-S-D-Bact-0785-a-S-19 and (HEX)-S-D-Bact-1371-a-A-20 were 600 bp (Fig. 2). The DNA bands obtained from clinical isolates and their respective extractions from the positive blood culture bottles were indistinguishable.

fig.ommitted FIG. 2. Representative PCR products from 16S rDNA-specific primers. Chromosomal preparations of the following organisms were used for PCR: lanes 1a and 1b, E. coli ATCC 35218; lanes 2a and 2b, E. faecalis ATCC 51299; lanes 3a and 3b, H. influenzae ATCC 49247; lanes 4a and 4b, K. pneumoniae ATCC 700603; lanes 5a and 5b, M. luteus ATCC 4698; lanes 6a and 6b, S. enterica serovar Enteritidis MC 24R; lanes 7a and 7b, S. aureus ATCC 43300; and lanes 8a and 8b, S. maltophilia ATCC 13637. Lanes 9a and 9b contain the PCR blanks and lanes M contain the molecular size markers (1-kb ladder, values to the left are kilobases). The letters with each lane number denote the primer pair used for PCR as follows: a, (6FAM)-S-D-Bact-0045-b-S-20 and (HEX)-S-D-Bact-0785-a-A-19; and b, (6FAM)-S-D-Bact-0785-a-S-19 and (HEX)-S-D-Bact-1371-a-A-20. The PCR products were separated in a 1.0% agarose gel at 120 V for 20 min.

 
TRF digest or size controls. Validation of TAMRA-labeled TRF control sizes, three fragments for each restriction digest, was performed by calculation from duplicate loads of triplicate digests done on two consecutive days (12 values for each control [data not shown]). Under the conditions used for the restriction digests there were no detectable undigested TAMRA-labeled fragments in the sizing gels. The control values calculated for TAMRA-labeled TRF sizes during determination of organism profiles are compiled in Table 2. In order for TRFs to be included in the profile for a given organism, the three internal digest control sizes were required to fall within a ±3-base window for the expected size. Although only five of the mean TRF values rounded to exactly the expected size, all 12 control TRF sizes were within one base of the expected value. The precision of fragment size determination is apparent from the range and standard deviation calculations, for which the greatest range was for two of the MspI digest controls (3.4 bases). The average range for all of the controls in the present study, weighted for frequency, was ±1.1 base from the mean.

fig.ommitted TABLE 2. Statistical evaluation of TAMRA-labeled TRF digest/size controls

 
TRF profiles for reference strains and clinical isolates. The TRF profiles for 69 reference strains and 32 clinical isolates were determined and incorporated into the TRFPD. Combined with the original predicted TRF profiles there are currently a total of 2,435 TRF profiles representing 921 different species of potential relevance to bacteremia. Representative composite electropherograms for several organisms relevant to bacteremia are presented in Fig. 3.

fig.ommitted FIG. 3. TRF profiles for several strains isolated from positive blood culture bottles. Each panel is a compilation of the five PCR product or restriction digest electropherograms for each organism and the applicable range of the standard curve (50 to 800 bases; red peaks). The single major 6FAM fragments (blue peaks) and HEX fragments (green peaks) of each digest are identified by the letters A to J. TRFs derived from PCR products of (6FAM)-S-D-Bact-0045-b-S-20 and (HEX)-S-D-Bact-0785-a-A-19 are identified by A and B (AluI), C and D (HhaI), E and F (MspI), or G and H (RsaI), respectively. TRFs derived from PCR products of (6FAM)-S-D-Bact-0785-a-S-19 and (HEX)-S-D-Bact-1371-a-A-20 are identified by I and J (HhaI), respectively. Tabulated fragment size information for each organism is available from Table 3.

 
The 33 experimentally determined TRF profiles in Table 3 are presented in the approximate order of incidence for isolation at Marshfield Laboratories. For any given organism, all fragment sizes initially within a three-base range were averaged and rounded to the nearest base to derive the reported cumulative TRF profiles (as with S. aureus and Escherichia coli). When an experimental TRF was >3 bases different than a duplicate, or another TRF from the same species, the entire procedure from the initial PCR was repeated for that isolate to confirm or correct the sizing of a fragment. As a result, the TRFs derived from products of (6FAM)-S-D-Bact-0045-b-S-20 and (HEX)-S-D-Bact-0785-a-A-19 for four different Enterococcus faecalis strains were found to have reproducible size diversity for two of the TRFs (6FAM fragment with AluI and MspI). In contrast, all of the staphylococcal species have essentially the same TRF profile and classification is described below. The majority of the experimentally determined TRF profiles matched the sequence predicted TRF profiles for the same species.

fig.ommitted TABLE 3. TRF profiles for reference and clinical isolates compared to sequence predicted TRF profiles

 
Accurate identification of organisms at the genus level was successful through searches with partial TRF profiles (Table 4). For example, all 84 staphylococcal profiles in the TRFPD match one of two minimal TRF profiles containing a 200 ± 3 (81 of 84) (Fig. 3, S. aureus, peak C)- or 539 ± 3 (3 of 84)-base TRF derived from (6FAM)-S-D-Bact-0045-b-S-20 and a 104-base TRF (Fig. 3, S. aureus, peak D) derived from (HEX)-S-D-Bact-0785-a-A-19 digested with HhaI, without any matches for another genus. In addition, 96% of the corynebacteria and rhodococci (124 of 129) matched a profile containing only two TRFs with only two other organisms from any other genus (Brevibacterium helvolum and Tropheryma whippelii).

fig.ommitted TABLE 4. Genus specificity of partial TRF profiles for blood pathogens

 
Detection of nuc and mecA. The presence of nuc was confirmed in all tested strains of S. aureus (ATCC 25923, ATCC 29213, ATCC 43300, MC 53R, MC 91, MC 97R, MC 98R, MC 99R, and MC 100R). A 279-base TAMRA-labeled TRF from the HhaI-digested product of a triplex PCR, derived from the primers TAMRA-nuc-S and nuc-A, was detected only in S. aureus strains (see the representative electropherograms in Fig. 4). The other staphylococci (Staphylococcus epidermidis ATCC 35983, Staphylococcus saprophyticus MC 28R, and Staphylococcus haemolyticus ATCC 29970) did not have a TAMRA-labeled TRF above the background level in the (279 ± 3)-base window, indicating the absence of nuc in these strains.

fig.ommitted   FIG. 4. TRF profiles for detection of nuc and mecA from staphylococci. Each panel includes the TAMRA-labeled PCR product or restriction digest (black peaks) for each representative staphylococcal strain and the applicable range of the ROX standard curve (red peaks). The presence of nuc (279 bases) and mecA (331 to 334 bases) were screened from the HhaI digest of PCR products from the primers TAMRA-nuc-S and TAMRA-mecA-S, respectively. The TAMRA-labeled HhaI restriction digest or fragment size controls are indicated as Ctl 228 and Ctl 425.

 
Several S. aureus strains with previously determined methicillin phenotypes were screened for the presence of mecA by TRF sizing. A 331- or 334-base TAMRA-labeled TRF from the HhaI-digested product of a triplex PCR, derived from the primers TAMRA-mecA-S and mecA-A, was detected only in S. aureus strains determined to be methicillin resistant (ATCC 43300, MC 53R, MC 97R, and MC 98R) (see representative electropherograms in Fig. 4). The methicillin-sensitive strains of S. aureus (ATCC 25923, ATCC 29213, MC 91, MC 99R, and MC 100R) and other staphylococci (S. epidermidis ATCC 35983, S. saprophyticus MC 28R, and S. haemolyticus ATCC 29970) did not have a TAMRA-labeled TRF above background level in the (333 ± 3)-base window, suggesting the absence of mecA in these strains.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Dicussion References

 
This study describes the use of TRF profiling for the identification of bacteria. The primary objective was to develop and validate a procedure for the identification of bacteria from blood culture by using 16S rDNA-specific PCR primer sets, restriction digest combinations, and sizing of the resulting TRF. A program was developed that allowed us to search the TRFPD and score the results on the basis of the fragment length similarity and the frequency of each profile. TRF profiles represent a highly reproducible and predictive source for identification of many organisms associated with bacteremia.

Due to the fact that samples can be obtained directly from positive blood bottles, significant delays that result from the need for isolated colonies for biochemical identification methods are avoided. Commercially available biochemical systems that require isolation of a microorganism after growth in a blood culture bottle generally require an additional 1 to 3 days before organism identification. In contrast, analysis of 16S rRNA genes by TRF profiling allows the identification process to be completed within 8 h of obtaining a positive blood bottle sample.

The procedures described here could be readily adapted to 96-well templates for DNA extractions, PCR, restriction digest, and capillary-based fragment analysis (including TRF calculations and predictive scoring). The modular nature of the TRF procedure allows it to be adapted as necessary to primer set and restriction enzyme combinations as desired. For example, primer sets with specificity to fungi or enteric bacteria could be included to diversify and clarify the identification capability of the current procedure.

When an organism is identified for which antibiotic susceptibilities are required, knowledge of the genus or species of the infecting bacteria can still be useful to a clinician for the determination of empirical therapy, resulting in more rapid prescription of appropriate antibiotics to the patient. In addition, it is possible that susceptibility testing could be adapted to inoculations of cultures directly from positive blood culture bottles, thereby decreasing the time required to determine antibiotic susceptibilities once the organism has been identified by TRF profiling. Reduced diagnostic time has implications for duration of infection, the cost of patient care, the length of hospitalization, the development of broad-spectrum antibiotic resistance, and mortality due to bacteremia.

The 16S rDNA-specific primers were chosen on the basis of previous evaluations of domain specificity and their predicted potential to provide species-specific TRF profiles (±3 bases) in combination with appropriate restriction digests. Modification to include C or T at the third position from 3' in (6FAM)-S-D-Bact-0045-b-S-20 was predicted to enhance annealing to the streptococci and closely related organisms. While we obtained PCR products from all of the chromosomal preparations for organisms used in the present study, the yield of product for (6FAM)-S-D-Bact-0045-b-S-20 and (HEX)-S-D-Bact-0785-a-A-19 from streptococci was typically lower than for all other bacteria, as indicated by the intensity of the bands visualized from ethidium bromide-stained agarose gels. This may be due in part to a lower affinity of the primer for template sequence or to the propensity for the lysis of streptococci at the stationary phase in the blood culture media, resulting in decreased bacterial DNA yields from extractions. However, we obtained adequate yields for TRF identification of all organisms tested in the present study.

The prestudy validation of the TAMRA-labeled digest or size controls was performed to establish parameters for data quality. TRF data from test organisms was accepted only when all three size controls in the sample fell within a ±3-base window from their validated values. This requirement was actually more liberal than necessary since all of the study controls (n = 760) were calculated to be within a maximum range of 3.4 bases. The precision of the data for controls indicates that the values obtained for TRFs from test organisms can be confidently predicted within a ±3-base window. Since the TRF profiles for the chosen primer set or restriction combinations did not rely on precision better than ±3 bases for most species identifications, this window was chosen for all profile comparisons with the predicted profiles in the TRFPD. One of the strengths in the design of the present study is the emphasis on identification based on relatively large differences in the lengths of fragments making the TRF profiles versus relying on the ability to differentiate smaller differences in TRF length.

The predictive value of the TRF profiles for identification of several bacterial species was strong. For example, 11 of the species in Table 3 (Pseudomonas aeruginosa, Bacteroides fragilis, Proteus mirabilis, Serratia marcescens, Micrococcus luteus, Acinetobacter baumannii, Rhodococcus equi, Bartonella henselae, Listeria monocytogenes, Moraxella osloensis, and Neisseria gonorrhoeae) have profiles that match 10 of 10 fragment sizes (±3 bases) without ambiguity for any other species.

Occasionally, the TRF profile also matched organisms from different genera, most notably organisms within the Enterobacteriaceae family. For example, 41 of the 42 TRF profiles for E. coli in the TRFPD match for all 10 fragment lengths (±3 bases), but this profile also matches all 7 Shigella profiles and 9 of 28 Salmonella profiles of various species. The similarity in profiles between the Escherichia and Shigella genera was expected due to the considerable sequence identity between these genera. In addition, one of the two profiles for Klebsiella pneumoniae matches with 4 other Klebsiella species, 12 of 15 Citrobacter profiles, and 4 of 18 Enterobacter profiles in the TRFPD. Finally, one of two profiles for Enterobacter cloacae matches with 4 other Enterobacter species and 11 of 28 Salmonella profiles in the TRFPD.

Although some ambiguity exists with other organism profiles, the confidence in identification of the enteric bacteria was the most significant limitation. However, due to the profile similarity of the Enterobacteriaceae, a partial profile of three TRFs was found to be sufficient to identify 170 of 180 (94%) strains in the TRFPD within the genera Citrobacter (14 of 15), Enterobacter (18 of 18), Erwinia (2 of 3), Escherichia (45 of 46), Hafnia (3 of 3), Klebsiella (31 of 31), Kluyvera (2 of 2), Proteus (3 of 5), Rahnella (8 of 10), Salmonella (25 of 28), Serratia (12 of 12), and Shigella (7 of 7) (see Table 4). In contrast, the same enteric profile only identifies 9 of 229 (4%) nonenteric strains in the TRFPD in the genera Neisseria, Pseudomonas, and Vibrio.

Organisms belonging to 10 different genera or related genera were frequently identifiable by using no more than three of the TRFs in a profile (Table 4). The enterococci and the streptococci, two relatively common genera from positive blood cultures, can be grouped confidently with only the TRF data for HhaI digest of (6FAM)-S-D-Bact-0045-b-S-20 and (HEX)-S-D-Bact-0785-a-A-19 products. The streptococci relevant to positive blood culture include many species with variable TRFs for at least 5 of the 10 profile categories. The determination of accuracy of speciation of the streptococci by TRF profiling was limited in part by the fact that Lancefield grouping (28) was the primary classification of these organisms.

All staphylococci were not differentiated to the species level by the complete TRF profile (10 TRF sizes). However, from a clinical perspective, the determination of a coagulase-positive phenotype is adequate for the identification of S. aureus. Previous reports have confirmed a high correlation between production of coagulase and an extracellular thermostable nuclease, encoded by nuc in S. aureus (3, 17, 18). In addition, the presence of mecA strongly correlates with methicillin resistance (13, 16-18). Therefore, with the current assay design, it is possible to rapidly and accurately identify staphylococci, predict the coagulase phenotype, and predict methicillin resistance on the basis of TRF sizes from HhaI digests of products from two PCR samples. More rapid dissemination of this information to the physician would result in decreased use of unnecessary antibiotics for coagulase-negative staphylococci and more appropriate use based on the predicted susceptibility to methicillin.

While this issue was not specifically addressed in the present study, we found the system capable of identifying organisms in a mixed culture (data not shown). For example, TRFs for all restriction digest products from both coagulase-negative staphylococci and P. aeruginosa were easily identified from a mixed culture. This was due in part to the distinct TRF patterns exhibited by these two organisms. However, other factors that could affect the ability to detect minor species from mixed cultures other than TRF profile differentiation include the relative ratio of organisms in culture, the extraction efficiency, and the PCR amplification efficiency. One distinct advantage of our data analysis program is the ability to incorporate a search for the profile of a secondary organism, thereby automatically determining the TRFs amidst the predominant profile. The ability to detect multiple organisms from mixed culture could reduce the time to identification since classical methods and sequencing first require subculturing of isolates.

The present study was designed to allow us to differentiate a wide variety of bacteria known to be isolated from blood culture. TRF profiles represent sequence predicted information, thereby offering an advantage over SSCP analysis, since additional primer sets can be more easily incorporated into the identification scheme. TRF profiling for bacterial identification is more rapid than sequencing of the 16S rRNA gene, primarily because it does not require a second thermal cycler reaction to produce sequenceable terminated products. However, no attempts were made to identify nonbacterial organisms. Also, as indicated above, there were a few instances when TRF profiles for organism identification were not specific to a given genus. The modular nature of TRF determinations allows the current system to be expanded to address both of these issues. For example, primer sets specific to medically important yeasts could be incorporated into the assay design as a separate PCR and a TRFPD of species-specific profiles derived from sequence and experimental data (7, 29, 30). Similarly, for more accurate identification of the Enterobacteriaceae it may be possible to incorporate an additional PCR with primers specific to this group, but with differentiating TRF sizes after restriction digest, or to explore TRFs derived from the 16S-23S intergenic spacer region. These examples are being explored in order to enhance the diagnostic capabilities of the current system.

　

	ACKNOWLEDGMENTS

 
We thank Pamela J. Squires at Marshfield Clinic Research Foundation for development of the TRFPD infrastructure and data extraction program. We thank the staff at Marshfield Laboratories for provision of blood cultures and isolates for analysis. We thank Terrance L. Marsh at Michigan State University for technical assistance with T-RFLP. We thank Rana M. Nasser of Marshfield Clinic and Graig E. Eldred of Marshfield Clinic Research Foundation for helpful comments and criticisms of the manuscript.

This project was funded by grants obtained through the Marshfield Clinic Research Foundation.

	REFERENCES

Top Abstract Introduction Materials and Methods Results Discussion References

 

1. Amann, R. I., W. Ludwig, and K. H. Schleifer. 1995. Phylogenetic identification and in situ detection of individual microbial cells without cultivation. Microbiol. Rev. 59:143-169.

2. Avaniss-Aghajani, E., K. Jones, A. Holtzman, T. Aronson, N. Glover, M. Boian, S. Froman, and C. F. Brunk. 1996. Molecular technique for rapid identification of mycobacteria. J. Clin. Microbiol. 34:98-102.

3. Brakstad, O. G., K. Aasbakk, and J. A. Maeland. 1992. Detection of Staphylococcus aureus by polymerase chain reaction amplification of the nuc gene. J. Clin. Microbiol. 30:1654-1660.

4. Browne, K., J. Miegel, and K. D. Stottmeier. 1984. Detection of pneumococci in blood cultures by latex agglutination. J. Clin. Microbiol. 19:649-650.

5. Brunk, C. F., E. Avaniss-Aghajani, and C. A. Brunk. 1996. A computer analysis of primer and probe hybridization potential with bacterial small-subunit rRNA sequences. Appl. Environ. Microbiol. 62:872-879.

6. Chen, K., H. Neimark, P. Rumore, and C. R. Steinman. 1989. Broad range DNA probes for detecting and amplifying eubacterial nucleic acids. FEMS Microbiol. Lett. 48:19-24.

7. Chen, Y. C., J. D. Eisner, M. M. Kattar, S. L. Rassoulian-Barrett, K. Lafe, U. Bui, A. P. Limaye, and B. T. Cookson. 2001. Polymorphic internal transcribed spacer region 1 DNA sequences identify medically important yeasts. J. Clin. Microbiol. 39:4042-4051.

8. Denis, F., J. Fleurette, G. Laurans, A. Moulin, M. Mounier, J. Orfila, and M. E. Reverdy. 1984. A latex agglutination technique for rapid, direct identification of pneumococci in blood cultures. Eur. J. Clin. Microbiol. 3:321-322.

9. Doern, G. V., R. Vautour, M. Gaudet, and B. Levy. 1994. Clinical impact of rapid in vitro susceptibility testing and bacterial identification. J. Clin. Microbiol. 32:1757-1762.

10. Fredricks, D. N., and D. A. Relman. 1999. Application of polymerase chain reaction to the diagnosis of infectious diseases. Clin. Infect. Dis. 29:475-486.

11. Fredricks, D. N., and D. A. Relman. 1998. Improved amplification of microbial DNA from blood cultures by removal of the PCR inhibitor sodium polyanetholesulfonate. J. Clin. Microbiol. 36:2810-2816.

12. Gordon, L. P., M. A. Damm, and J. D. Anderson. 1987. Rapid presumptive identification of streptococci directly from blood cultures by serologic tests and the L-pyrrolidonyl-ß-naphthylamide reaction. J. Clin. Microbiol. 25:238-241.

13. Grisold, A. J., E. Leitner, G. Muhlbauer, E. Marth, and H. H. Kessler. 2002. Detection of methicillin-resistant Staphylococcus aureus and simultaneous confirmation by automated nucleic acid extraction and real-time PCR. J. Clin. Microbiol. 40:2392-2397.

14. Jarvis, W. R. 1996. Selected aspects of the socioeconomic impact of nosocomial infections: morbidity, mortality, cost, and prevention. Infect. Control Hosp. Epidemiol. 17:552-557.

15. Kitts, C. L. 2001. Terminal restriction fragment patterns: a tool for comparing microbial communities and assessing community dynamics. Curr. Issues Intest. Microbiol. 2:17-25.

16. Lem, P., J. Spiegelman, B. Toye, and K. Ramotar. 2001. Direct detection of mecA, nuc and 16S rRNA genes in BacT/Alert blood culture bottles. Diagn. Microbiol. Infect. Dis. 41:165-168.

17. Louie, L., J. Goodfellow, P. Mathieu, A. Glatt, M. Louie, and A. E. Simor. 2002. Rapid detection of methicillin-resistant staphylococci from blood culture bottles by using a multiplex PCR assay. J. Clin. Microbiol. 40:2786-2790.

18. Maes, N., J. Magdalena, S. Rottiers, Y. De Gheldre, and M. J. Struelens. 2002. Evaluation of a triplex PCR assay to discriminate Staphylococcus aureus from coagulase-negative staphylococci and determine methicillin resistance from blood cultures. J. Clin. Microbiol. 40:1514-1517.

19. Maidak, B. L., J. R. Cole, T. G. Lilburn, C. T. Parker, Jr., P. R. Saxman, R. J. Farris, G. M. Garrity, G. J. Olsen, T. M. Schmidt, and J. M. Tiedje. 2001. The RDP-II (Ribosomal Database Project). Nucleic Acids Res. 29:173-174.

20. Maquelin, K., C. Kirschner, L. P. Choo-Smith, N. A. Ngo-Thi, T. van Vreeswijk, M. Stammler, H. P. Endtz, H. A. Bruining, D. Naumann, and G. J. Puppels. 2003. Prospective study of the performance of vibrational spectroscopies for rapid identification of bacterial and fungal pathogens recovered from blood cultures. J. Clin. Microbiol. 41:324-329.

21. Marsh, T. L. 1999. Terminal restriction fragment length polymorphism (T-RFLP): an emerging method for characterizing diversity among homologous populations of amplification products. Curr. Opin. Microbiol. 2:323-327.

22. Marsh, T. L., P. Saxman, J. Cole, and J. Tiedje. 2000. Terminal restriction fragment length polymorphism analysis program, a web-based research tool for microbial community analysis. Appl. Environ. Microbiol. 66:3616-3620.

23. Murakami, K., W. Minamide, K. Wada, E. Nakamura, H. Teraoka, and S. Watanabe. 1991. Identification of methicillin-resistant strains of staphylococci by polymerase chain reaction. J. Clin. Microbiol. 29:2240-2244.

24. Oliveira, K., S. M. Brecher, A. Durbin, D. S. Shapiro, D. R. Schwartz, P. C. De Girolami, J. Dakos, G. W. Procop, D. Wilson, C. S. Hanna, G. Haase, H. Peltroche-Llacsahuanga, K. C. Chapin, M. C. Musgnug, M. H. Levi, C. Shoemaker, and H. Stender. 2003. Direct identification of Staphylococcus aureus from positive blood culture bottles. J. Clin. Microbiol. 41:889-891.

25. Qian, Q., Y. W. Tang, C. P. Kolbert, C. A. Torgerson, J. G. Hughes, E. A. Vetter, W. S. Harmsen, S. O. Montgomery, F. R. Cockerill III, and D. H. Persing. 2001. Direct identification of bacteria from positive blood cultures by amplification and sequencing of the 16S rRNA gene: evaluation of BACTEC 9240 instrument true-positive and false-positive results. J. Clin. Microbiol. 39:3578-3582.

26. Ratner, H. B., and C. W. Stratton. 1985. Thermonuclease test for same-day identification of Staphylococcus aureus in blood cultures. J. Clin. Microbiol. 21:995-996.

27. Relman, D. A., J. S. Loutit, T. M. Schmidt, S. Falkow, and L. S. Tompkins. 1990. The agent of bacillary angiomatosis: an approach to the identification of uncultured pathogens. N. Engl. J. Med. 323:1573-1580.

28. Ruoff, K. L., R. A. Whiley, and D. Beighton. 1999. Streptococcus, p. 283-296. In P. R. Murray, E. J. Baron, M. A. Pfaller, F. C. Tenover, and R. H. Yolken (ed.), Manual of clinical microbiology, 7th ed. ASM Press, Washington, D.C.

29. Shin, J. H., F. S. Nolte, B. P. Holloway, and C. J. Morrison. 1999. Rapid identification of up to three Candida species in a single reaction tube by a 5' exonuclease assay using fluorescent DNA probes. J. Clin. Microbiol. 37:165-170.

30. Shin, J. H., F. S. Nolte, and C. J. Morrison. 1997. Rapid identification of Candida species in blood cultures by a clinically useful PCR method. J. Clin. Microbiol. 35:1454-1459.

31. Southern, E. M. 1979. Measurement of DNA length by gel electrophoresis. Anal. Biochem. 100:319-323.

32. Turenne, C. Y., E. Witwicki, D. J. Hoban, J. A. Karlowsky, and A. M. Kabani. 2000. Rapid identification of bacteria from positive blood cultures by fluorescence-based PCR-single-strand conformation polymorphism analysis of the 16S rRNA gene. J. Clin. Microbiol. 38:513-520.

33. Wellstood, S. A. 1987. Rapid, cost-effective identification of group A streptococci and enterococci by pyrrolidonyl-beta-naphthylamide hydrolysis. J. Clin. Microbiol. 25:1805-1806.

34. Widjojoatmodjo, M. N., A. C. Fluit, and J. Verhoef. 1995. Molecular identification of bacteria by fluorescence-based PCR-single-strand conformation polymorphism analysis of the 16S rRNA gene. J. Clin. Microbiol. 33:2601-2606.

35. Woese, C. R. 1987. Bacterial evolution. Microbiol. Rev. 51:221-271.