Abstract

TEXT

Bloodstream infections caused by Gram-negative bacteria are a leading cause of morbidity and mortality among hospitalized patients (12, 14). Inasmuch as decreasing the time to identify the causative organism and determine its antibiotic susceptibility can significantly improve outcomes in patients with bacteremia or sepsis (1, 7, 16), we validated a method to rapidly identify and perform susceptibility testing for Gram-negative bacteria directly recovered from positive blood culture medium. Our strategy used matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry (MALDI Biotyper; Bruker Daltonics, Billerica, MA) and automated antibiotic susceptibility testing (Phoenix, BD, Sparks, MD) instruments.

The study was performed in two parts. First, we created a training test set of 50 simulated bacteremia specimens. The 10 Gram-negative species included in the training set were distributed in a ratio that is representative of the bacteremia specimens processed in our clinical microbiology laboratory. Organisms were selected from cryopreserved stocks of strains purchased from the American Type Culture Collection (ATCC, Manassas, VA), strains recovered from College of American Pathologists (CAP, Northfield, IL) proficiency testing surveys, and patient isolates that were collected in accordance with a protocol approved by the institutional review board at The Methodist Hospital. Each simulated bacteremia specimen was prepared at a clinically relevant concentration (~10³ to 10⁵ CFU/ml) using known Gram-negative strains (Table 1). Briefly, each strain was grown overnight on tryptic soy agar supplemented with 5% sheep blood (BD), colonies were transferred to 5 ml sterile saline using an inoculating loop, and the suspension was vortexed for 30 s. The concentration of bacteria in each suspension was estimated using a nephelometer (PhoenixSpec, BD) and McFarland equivalence turbidity standards (Remel, Lenexa, KS). A 10-ml simulated sample was then prepared using an appropriate dilution factor, inoculated into an aerobic blood culture bottle (Bactec Standard 10 Aerobic F Medium; BD), and incubated in an automated instrument (Bactec FX; BD) according to the manufacturer's instructions. Second, we created a validation test set of 60 positive patient blood cultures in which the initial Gram stain indicated a Gram-negative organism. Polymicrobial cultures were excluded from the study. The organism identification and susceptibility testing were performed by blinded laboratory personnel.

After being flagged as positive for growth by the blood culture instrument, 6 ml of broth was transferred from the blood culture bottle with a syringe to an 8.5-ml serum separator tube (Vacutainer SST Plus; BD) and centrifuged at 2,000 rpm for 15 min at room temperature. The supernatant was aspirated, using caution to not disrupt the loosely formed pellet of bacteria present at the surface of the polymeric gel. The MALDI-TOF mass spectrometry identification and antibiotic susceptibility panels were then prepared simultaneously. For rapid organism identification, an aliquot of the pelleted bacteria was collected by gently touching it with a cotton swab (Pur-Wraps; Puritan Medical Products Company, Guilford, ME) and spotted onto the stainless steel target. Importantly, in evaluating several different swabs, we determined that a loosely wound, coarsely textured cotton tip provided the most consistent transfer of bacteria to the target. The bacteria were then allowed to dry at room temperature, overlaid with 1 μl of α-cyano-4-hydroxycinnamic acid (HCCA) matrix, and analyzed using the MALDI-TOF mass spectrometer (MALDI Biotyper with FlexControl software; Bruker Daltonics) according to the manufacturer's instructions. Since MALDI-TOF mass spectrometry cannot reliably distinguish between Escherichia coli and Shigella spp., a spot indole test was performed to confirm the identification. Each unknown organism was also identified using conventional biochemical phenotyping methods (Phoenix; BD) on colonies isolated from subcultures of the positive blood culture medium. Any discordant results from the two identification methods were resolved by 16S rRNA gene sequencing (ARUP Laboratories, Salt Lake City, UT). For rapid antibiotic susceptibility testing, an aliquot of the pelleted bacteria was resuspended (Phoenix ID broth; BD), processed using the autoinoculator (Phoenix AP; BD), and analyzed on an automated instrument (Phoenix; BD) according to the manufacturer's instructions. The MICs of 2 to 20 antimicrobial agents, selected as appropriate for each organism (Table 2), were determined for each strain. Antibiotic susceptibility results using the direct pellet method were compared to those generated by the conventional method (Phoenix; BD) that tests colonies isolated from subcultures of the positive blood culture medium. Any strain-antibiotic combination demonstrating a >1-fold dilution difference in the MIC or a change in the categorical drug-susceptible/resistant interpretation between the two methods was resolved by performing an epsilometer test (bioMérieux, Durham, NC). Categorical disagreements were scored as very major errors (susceptible by the direct method and resistant by the reference method), major errors (resistant by the direct method and susceptible by the reference method), or minor errors (intermediate by one method and susceptible or resistant by the other method) as previously described (6).

Results of the rapid identification study using the MALDI-TOF mass spectrometer were highly concordant with those based on conventional biochemical phenotypes and 16S rRNA gene sequences (108/110 identifications were concordant [98%]) (Table 1). Both discordant identifications occurred in the training set. The MALDI-TOF mass spectrometer did not generate a reliable identification score for one Enterobacter cloacae complex strain, and it identified one Stenotrophomonas maltophilia strain to only the genus (Table 1). A similar level of diagnostic accuracy for identifying Gram-negative bacteria directly from positive blood culture bottles has been previously reported (4, 11, 15). Results of the rapid susceptibility study using pelleted bacteria directly recovered from the positive blood culture medium were also highly concordant with those of the conventional method using colonies isolated by subculture (Table 1). A similar level of accuracy for susceptibility testing directly from positive blood culture bottles has been previously reported (3, 5, 8–10). The training set and validation set included totals of 821 and 1,074 organism-antibiotic combinations, respectively. In total, 5 very major errors (0.26%), 6 major errors (0.32%), and 26 minor errors (1.37%) were recorded. These data are consistent with accepted error rates for clinical microbiology laboratories (2, 13). No significant trend in the errors was observed for any organism or antibiotic agent tested (chi-square test, Prism 4; GraphPad Software, La Jolla, CA). Importantly, by bypassing the time-consuming subculture steps included in the conventional method, results from the direct testing method were generated a full day earlier (mean, 24.06 h; standard error of the mean [SEM], 0.84 h; range, 9.57 to 35.07 h).

In summary, the data demonstrate that Gram-negative organisms recovered from positive blood cultures can be accurately identified and rapidly assessed for antibiotic susceptibility directly from the primary broth. Although other investigators have also recently reported the successful identification of bacteria directly from positive blood culture medium using MALDI-TOF mass spectrometry (4, 11, 15), we are the first to couple rapid identification with rapid antibiotic susceptibility testing. To have a maximum benefit to patient care, particularly in Gram-negative sepsis, appropriate antibiotic therapy must be administered as soon as possible (12, 14). To this end, our strategy markedly reduced turnaround times for species identification and susceptibility testing. Due to operational constraints in our laboratory, specimens are processed in batches at predetermined time points. However, if the workflow were modified to immediately process positive blood cultures for Gram stain interpretation, MALDI-TOF analysis, and susceptibility testing when flagged by the instrument, clinically actionable results could be generated even more quickly. The direct testing method described herein has been successfully implemented in our clinical microbiology laboratory that serves as the reference laboratory for The Methodist Hospital System. Validation studies are now under way to expand the strategy of rapid identification and antibiotic susceptibility testing to bacteremia caused by Gram-positive bacteria and yeast.

ACKNOWLEDGMENTS

Footnotes

REFERENCES