Direct, rapid antimicrobial susceptibility test from positive blood cultures based on microscopic imaging analysis

Abstract

For the timely treatment of patients with infections in bloodstream and cerebrospinal fluid, a rapid antimicrobial susceptibility test (AST) is urgently needed. Here, we describe a direct and rapid antimicrobial susceptibility testing (dRAST) system, which can determine the antimicrobial susceptibility of bacteria from a positive blood culture bottle (PBCB) in six hours. The positive blood culture sample is directly mixed with agarose and inoculated into a micropatterned plastic microchip with lyophilized antibiotic agents. Using microscopic detection of bacterial colony formation in agarose, the total time to result from a PBCB for dRAST was only six hours for a wide range of bacterial concentrations in PBCBs. The results from the dRAST system were consistent with the results from a standard AST, broth microdilution test. In tests of clinical isolates (n = 206) composed of 16 Gram-negative species and seven Gram-positive species, the dRAST system was accurate compared to the standard broth microdilution test, with rates of 91.11% (2613/2868) categorical agreement, 6.69% (192/2868) minor error, 2.72% (50/1837) major error and 1.45% (13/896) very major error. Thus, the dRAST system can be used to rapidly identify appropriate antimicrobial agents for the treatment of blood stream infection (BSI) and antibiotic-resistant strain infections.

Conflict of interest statement

J.C., Shinhun.H., B.J., D.Y.K., Sangkwon.H. and S.K. (QuantaMatrix Inc.) at the time of manuscript submission, were employed at QuantaMatrix, Inc., which is commercializing the dRAST technology. J.C., Shinhun.H., B.J., D.Y.K., Sangkwon.H. and S.K. (QuantaMatrix Inc.) have equity interest in QuantaMatrix, Inc.

Figures

Figure 1

Device for the direct detection of antibiotic resistance from positive blood culture bottles (PBCBs). (A) The conventional AST system requires three separate culture processes: blood culture for positive infection detection, subculture for separation of bacteria from blood cells, and AST via optical density measurement; the total time to result is approximately 60 hours. The direct and rapid AST (dRAST) system developed herein does not require a subculture process and enables AST to be performed in six hours. The total time to result of dRAST is less than 24 hours. (B) Process of preparing a dRAST chip from a PBCB. A PBCB aliquot was diluted and mixed with liquid-state agarose. The mixture was inoculated into a dRAST chip consisting of 96 test wells containing various antimicrobials at several concentrations. (C) Detailed structure of a dRAST chip well. Each test well consists of a micropatterned radial chamber for agarose matrix molding and a satellite well for freeze-dried antibiotics. Inset images show the micropatterned chamber i) before agarose mixture loading and ii) after agarose mixture loading. (D) Experimental process for detecting bacteria in blood. A sample from a PBCB is mixed with agarose and loaded into the loading chamber in the chip. Freeze-dried antibiotics are rehydrated by adding culture medium (CAMHB). Automated microscopic imaging is used to detect bacterial growth. Inset scanning electron microscope (SEM) images show (i) freeze-dried antibiotics in the satellite well and (ii) the focus mark on the bottom of the chip for automated imaging. (E) In the initial state, bacteria were not detectable using a 20x magnification lens; only blood cells were detectable. After four hours of incubation, in the case of resistance to the antibiotic, a single bacterium divided and formed a microcolony. Filled triangle represents a blood cell. Unfilled triangle represents a bacterial microcolony. The scale bar represents 10 mm in (B), 3 mm in (C), 300 μm in (D), and 100 μm in (E).

Figure 2

Consistent AST results from a wide range of inoculum sizes. (A) Colony formation of bacteria at different inoculum sizes. In all inoculum sizes from 5.0 × 107 to 5.0 × 105 CFU/ml, there was microcolony formation at 0, 0.25 and 0.5 μg/ml gentamicin. At 1 μg/ml gentamicin, there was no colony formation at any inoculum size. The scale bar represents 100 μm. (B) MIC values from different E. coli ATCC 25922 inoculum sizes of 5.0 × 107, 5.0 × 106 and 5.0 × 105 CFU/ml incubated with 17 clinically important antimicrobials and analyzed using dRAST and broth microdilution test. For all inoculum sizes, the MIC values were in the middle of the CLSI quality control ranges. (C) Summary of a comparison of AST results from various inoculum sizes of E. coli ATCC 25922, P. aeruginosa ATCC 27853, S. aureus ATCC 29213 and E. faecalis ATCC 29212. “Correct” means a case with identical MIC values from different inoculum sizes. “One two-fold” and “Two two-fold” denote cases with MIC values from different inoculum sizes that were different by two-fold and four-fold, respectively. “Essential agreement” refers to the proportion of cases that were correct or had one two-fold difference. The essential agreement rates were 99.33% in E. coli ATCC 25922, 100% in P. aeruginosa ATCC 27853, 95.37% in S. aureus ATCC 29213 and 100% in E. faecalis ATCC 29212.

Figure 3

Automatic AST using colony-forming area detection. (A) Time-lapse images were acquired from the automated imaging system, and the image data were transferred to the analysis system. (B) Time-lapse images of K. pneumoniae with several concentrations of amoxicillin/clavulanic acid (A/C). (C) The raw images from (B) were processed to binary format images. The number of white pixels in the image represents the microcolony area. (D) Graph of the microcolony-forming area in (C). From low A/C concentrations until 16/8 μg/ml, there was a substantial increase in microcolony-forming area in the images. However, from 16/8 μg/ml to 32/16 μg/ml, there was no substantial change in the area of bacteria. (E) The normalized growth rates at all the concentrations were calculated. Values higher than 20% were regarded as growth, and lower values were regarded as non-growth. From low A/C concentrations until 16/8 μg/ml, the normalized growth rates were higher than 20% and regarded as growth. However, at 32/16 μg/ml, the normalized growth rate was 2.3% and lower than 20%; thus, it was regarded as non-growth. Therefore, the MIC value was determined to be 32/16 μg/ml. (F) Final AST report. Using the ID information and MIC interpretive criteria from CLSI, antibiotic susceptibility was determined in a full antimicrobial panel. The scale bars in (B) represent 100 μm.

Figure 4

Summary of clinical testing with dRAST using colony-forming area detection. (A) Bacterial ID of 206 clinically isolated strains. There were 16 species of 105 Gram-negative strains and 7 species of 101 Gram-positive strains. (B) Classification of clinical samples for MIC interpretive standards. (C) Discrepancy rates and CA rates for dRAST using clinical samples. The dRAST results were compared with the BMD results to calculate the discrepancy rates. For the BMD test: S, susceptible; I, intermediate; R, resistant. For dRAST: mE, minor error; ME, major error; VME, very major error; CA, categorical agreement. (D) CA rates according to the main classified strains: Enterobacteriaceae spp. (n = 87, 89.9%), P. aeruginosa (n = 9, 88.9%), Staphylococcus spp. (n = 66, 92.1%), Enterococcus spp. (n = 35, 95.4%). (E) Discrepancy rates according to the main classified strains: Enterobacteriaceae spp. (mE = 7.9%, ME = 2.6%, VME = 1.2%), P. aeruginosa (mE = 11.1%, ME = 0.0%, VME = 0.0%), Staphylococcus spp. (mE = 4.7%, ME = 3.8%, VME = 2.4%), and Enterococcus spp. (mE = 4.3%, ME = 0.0%, VME = 0.6%). The error rates were calculated by comparing the AST results from each method with the BMD test. Error bars represent 95% confidence intervals.

Figure 5

Average time to results of AST from (A) time of positive detection of blood culture bottles or (B) time of blood collection from patient. The time required for AST using dRAST and a conventional AST method requiring subculture (Vitek 2 or MicroScan) was calculated.