Acoustic Trapping for Bacteria Identification in Positive Blood Cultures

Article pubs.acs.org/ac

Acoustic Trapping for Bacteria Identification in Positive Blood Cultures with MALDI-TOF MS Björn Hammarström,*,† Bo Nilson,‡,§ Thomas Laurell,†,∥ Johan Nilsson,† and Simon Ekström† †

Department of Biomedical Engineering, Lund University, 221 00 Lund, Sweden Clinical Microbiology, Labmedicin, Region Skåne, 221 85 Lund, Sweden § Department of Laboratory Medicine, Division of Medicinal Microbiology, Lund University, SE-223 62 Lund, Sweden ∥ Department of Biomedical Engineering, Dongguk University, Jung-gu, Seoul 100-715, Korea ‡

S Supporting Information *

ABSTRACT: Matrix assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) is currently changing the clinical routine for identification of microbial pathogens. One important application is the rapid identification of bacteria for the diagnosis of bloodstream infections (BSI). A novel approach based on acoustic trapping and an integrated selective enrichment target (ISET) microchip that improves the sample preparation step for this type of analysis is presented. The method is evaluated on clinically relevant samples in the form of Escherichia coli infected blood cultures. It is shown that noncontact acoustic trapping enables capture, enrichment, and washing of bacteria directly from the complex background of crude blood cultures. The technology replaces centrifugation-based separation with a faster and highly automated sample preparation method that minimizes manual handling of hazardous pathogens. The presented method includes a solid phase extraction step that was optimized for enrichment of the bacterial proteins and peptides that are used for bacterial identification. The acoustic trapping-based assay provided correct identification in 12 out 12 cases of E. coli positive blood cultures with an average score of 2.19 ± 0.09 compared to 1.98 ± 0.08 when using the standard assay. This new technology opens up the possibility to automate and speed up an important and widely used diagnostic assay for bloodstream infections.

T

Despite the great success of MALDI-TOF MS-based identification, several improvements are needed in order to realize the true potential of the technology. A major challenge here is to purify microbes from the wide variety of samples that need to be handled in clinical practice, e.g., blood, urine, and cerebrospinal fluid. Such samples generate an intense and complex background that needs to be removed in order to allow pathogen identification, and extensive manual sample preparation or overnight plate culturing are therefore typically used.1 Examples of other challenges that need to be addressed are providing simple and effective lysis conditions for resilient microbial pathogens13−16 and methods for rapid deduction of antibiotics resistance.17−20 Although MALDI-TOF MS provides a fast readout useful in the analysis of large sample collections, time-consuming sample preparation is often a bottleneck for such progress. To combat these challenges, microfluidic devices may play an important role.21−24 Bacterial infection in the bloodstream can cause sepsis, a lifethreatening condition that is caused by an uncontrolled systemic inflammatory response.25 Sepsis is a leading cause of

he clinical practice for microbe identification is undergoing a major reformation, from phenotypic methods to matrix-assisted laser desorption ionization time-of-flight mass spectrometric (MALDI-TOF MS) identification.1,2 MALDITOF MS has the advantage of being faster and less prone to certain misidentifications3 than phenotypic methods. The ability to use MALDI-TOF MS for rapid identification was realized in the mid-90s of the 20th century4−6 and has since then been clinically verified for identification of a large number of microbes.1 The MALDI-TOF MS identification is based on pattern recognition of highly abundant proteins (e.g., histone and ribosomal proteins) in the bacteria in the mass range of 2 to 20 kDa, which are correlated with bacteria specific spectra in commercially available databases such as Andromas, Vitek MS, or MALDI Biotyper. When compared to molecular methods such as real-time polymerase chain reaction (PCR)7,8 or fluorescence in situ hybridization (FISH),9,10 the MALDI-TOF MS approach has a clear advantage as it allows generic identification at a lower cost without species specific probes. Furthermore, the use of protein patterns can provide identification that is more stable against mutations than those based on short DNA sequences. The most widely used platforms for MALDI-TOF MS identification of bacteria is the MALDI BioTyper from Bruker and the Vitek MS from bioMérieux.11,12 © XXXX American Chemical Society

Received: June 2, 2014 Accepted: September 30, 2014

A

dx.doi.org/10.1021/ac502020f | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

Figure 1. Manual Sepsityper protocol (a) in comparison with the acoustic trapping-based extraction, where bacteria are captured, washed, and lysed on larger seed-particles in an acoustic trap (b). Further, a solid phase extraction (SPE) step in an integrated selective enrichment target (ISET) microchip is performed to concentrate and purify analytes before MALDI-TOF MS readout. All the assay steps in the trapping scheme (b) can be performed by a pipetting robot in an automated fashion.

■

MATERIALS AND METHODS Chemicals. Unless otherwise specified, all chemicals were purchased from Sigma-Aldrich Co. (St. Louis, MO, US) and used without further purification. In preparation of Luria broth (LB) medium yeast extract (Hy-yeast 412, Sigma Y1001), sodium chloride (Sigma S3014) and Tryptone (Duchefa Biochemie T1332.1000) were used. Pure Bacteria Samples. DH5α strain Escherichia coli was grown in LB medium with 100 μg/mL ampicillin to stationary phase. This culture was used for initial solid phase extraction (SPE) optimization experiments. Positive Blood Cultures. Mock sepsis samples were obtained using BacT/ALERT FA Plus blood culture bottles (BioMérieux, France) containing 30 mL of Fastidious Antimicrobial Neutralization Plus (FAN Plus, BioMerieux, France) media and adsorbent polymeric beads. The specimen consisted of 8 mL of horse blood containing 250 CFU E. coli reference strains DH5α or MB11464_1. The bottles were incubated overnight in a BacT/ALERT three-dimensional (3D) blood culture incubator (BioMérieux, France) until colorimetric bacteria detection occurred at approximately 107 bacteria/mL. Thereafter, a small sample was drawn from the culture bottle using a syringe leaving the large adsorbent beads within the bottle. Acoustic Trapping Device. Acoustic trapping using seedparticles allows submicrometer particles to be captured and retained against fluid flow,30 Figure 2. Herein, precapture of 10 μm silica particles (seed-particles) enabled enrichment and purification of bacteria in a flow-through microfluidic device. Acoustically resonating glass capillaries is an efficient and lowcost approach to realizing such a device.31 Here, borosilicate capillaries (VitroCom, NJ, USA) with cross-sectional inner dimensions of 2000 × 200 μm2 and a wall thickness of 150 μm were used. With a water filled channel at room temperature, these capillary dimensions will give rise to a cross-sectional resonance at 4 MHz, as illustrated by the schematic in Figure 2. This acoustic field pushes the particles away from the surface, into the center pressure node and facilitates noncontact trapping.

death for critically ill patients and incurs great economic cost for the society (estimated to be 17 billion dollars in 200126). The use of MALDI-TOF MS for clinical diagnosis of bloodstream infections (BSI) is increasing.27 Before successful identification with MALDI-TOF MS can be achieved, removal of background from the blood is critical. There are currently two sample preparation methods used in clinical diagnostics to eliminate this background. Both use an initial overnight incubation in blood culture bottles to increase the bacteria concentration. This is subsequently followed either by a second overnight plate culture or by a centrifugation-based protocol to separate blood and bacteria. The manual sample preparation was recently enabled by using either the Sepsityper kit28 or equivalent in-house methods29 and is widely preferred as it removes one entire overnight culture step. However, the method forces additional manual labor as no technology for automation of the centrifugation-based protocol is currently available. A novel method based on acoustic trapping in a microfluidic system is proposed. The setup allows the bacteria purification assay to be performed in an automated and time saving manner (Figure 1) and can reliably be used to determine the identity of a model bacterial species (Escherichia coli). Seed-particle aided acoustic trapping30 in glass capillaries31 was used for the capture and washing of bacteria in the lysed blood culture sample. A recently presented frequency tracking method32 was also utilized to enable both media switching and formic acid (FA) lysis in-trap (despite large differences in speed of sound for the media), as well as providing a novel method for quality assurance. Furthermore, the aspirate/dispense format allows a robot to perform the assay by using the capillary as a pipet tip. Acoustic streaming is also generated in the acoustic trap,33 and this phenomenon has shown to improve mixing and enzymatic turnover in microchannels34 and enable efficient washing in combination with noncontact trapping of a thin particle layer.35 In this study, these benefits were utilized to implement acoustic trapping for accurate identification of E. coli with high score values for infected blood samples. B

dx.doi.org/10.1021/ac502020f | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

The acoustic trap was seeded by aspirating a small amount of silica particles to form a small seed cluster, as in Figure 2b. To remove any residual particles, the cluster was washed with 20 μL of deionized water. Subsequently, 100 μL of the selectively lysed blood culture sample was aspirated and bacteria were trapped. The bacteria were washed by aspirating 100 μL of deionized water, and bacteria lysis was performed by aspirating 2 μL of 70% formic acid. To release the cluster in a minimal volume, air was aspirated after the formic acid aspiration, drawing the liquid−air meniscus close to the trapping site. The purified bacteria proteins were subsequently dispensed from the trap in a 4−6 μL volume to ensure complete recovery of the lysate. A reverse phase solid phase extraction (RP-SPE) step was included after the acoustic trapping to concentrate and purify the 4−6 μL of protein containing the volume released by acid lysis of the trapped bacteria. The RP-SPE step was performed by dispensing the purified bacteria proteins into the previously described integrated selective enrichment target (ISET) chip.23,24 The ISET device is a MALDI target comprising a silicon chip with a series of perforated nanovials packed with RP-SPE media, shown in Figure 2b. Placing the ISET in a vacuum fixture allowed sample to be drawn through the SPE-bed using a vacuum, as supplied by a vacuum pump (Vacuubrand GMBH, Wertheim, Germany) and regulated with a valve (Vacuum regulator, product NO. 19530, Qiagen, Venlo, Netherlands). Different SPE media were evaluated, but for the final protocol, Poros R1 (50 μm) beads (Carlsbad, CA, US) were selected. For the ISET SPE protocol, each nanovial was loaded with approximately 400 nL of R1 beads in 60%ACN/0.1%TFA and washed at a maximum vacuum with 2.5 μL of 60%ACN/0.1% TFA followed by 0.1% TFA. The bacteria protein samples from the trapping were drawn through the wells, and each well was washed with 3 μL of 0.1% TFA twice, under high vacuum (−10 mmHg). Elution of the proteins onto the backside of the ISET was performed with 2 × 0.3 μL, 80% ACN/0.1%TFA containing 3 mg/mL of cyano-4-hydroxy-cinnamic acid (CHCA) matrix. The elution was done at a low vacuum (−2 mmHg) to generate a MALDI spot on the back-side of the ISET chip. To expose the matrix spots before the final MALDITOF MS analysis, the ISET chip was turned up-side down and placed in a MALDI target having a milled recession with a footprint matching the chip. Mass Spectroscopy. For some of the SPE optimization work, bacteria samples were analyzed with MALDI-TOF MS on a M@ldi LR (Waters/Micromass, Milford, MA, USA) in linear mode. For the MALDI-TOF MS analysis, a spectrum of 100 summed laser shots was acquired for each sample spot. MassLynx 4.1 was used for controlling the Waters instrument. The bacterial identification experiments were performed on an ultrafleXtreme MALDI-TOF/TOF (Bruker Daltronik GmbH, Germany) in linear mode in a mass range of 2 to 20 kDa and were analyzed using FlexControl and MALDI Biotyper 3.1 software (Bruker Daltronik GmbH, Germany). The identification results were evaluated according to the manufacturer’s instructions. Scores values above 2 suggested a probable identification to the species level; scores between 1.7 and 2 indicated genus identification, and scores of 1.7 and below were considered to be unreliable.

Figure 2. Rectangular cross-section 2000 × 200 μm2 trapping capillary and transducer device (a) used like a pipet in aspirate/dispense-mode by leaving one end of the capillary open and connecting the other end to a pump. Localized actuation enables acoustic trapping of a controlled amount of particles (b) above the millimeter sized ultrasonic transducer. The schematic side-view (c) shows the ultrasonic standing wave that retains particles in the center of the channel.

The miniaturized ultrasonic transducer was used to locally actuate the cross-sectional mode, providing an acoustic force potential minima in the center of the channel extending along the length of the transducer. In order to enable automatic frequency tracking of the resonance mode, a kerfed lead zirconate titanate (PZT) transducer was used and the optimal frequency was continuously tracked using LabVIEW (National Instruments, TX, US) to implement an algorithm that maximizes the power input to the transducer.32 Automatic Sample Handling Platform. The capillary trap was operated in aspirate/dispense-mode like a regular pipet allowing automation of the bacteria purification using standard laboratory robotics. One end of the capillary was connected to a computer controlled syringe pump (neMESYS, Cetoni, Korbußen, Germany) via flexible tubing (Tygone, Saint-Gobain Performance Plastics, France), and one end was left open for aspirate/dispense purposes. An XYZ-stage with LabVIEW (National instruments) controlled stepper motors (Parker Hannifin, OH, US) was used to translate the entire device to different liquids, e.g., tubes in a rack or a microtiter plate. Reference Protocol. For comparative purposes, all the positive blood cultures were simultaneously processed using the standard Bruker Sepsityper Kit (part # 8270170, Bruker Daltronik GmbH, Germany) according to the manufacturer sample preparation protocol (revision 2, 2013). The protocol is based on selective lysis of blood cells using a lysis buffer, washing of the bacteria, and subsequent lysis of the bacteria using formic acid with a total of five centrifugation steps (Figure 1a). Acoustic Trapping Protocol. In the presented method, seed-particle aided acoustic trapping was used to capture and wash bacteria from positive blood culture samples (prepared as described above). A blood culture sample (100 μL) was dispensed into a vial containing lysis buffer (20 μL) from the standard kit (MALDI Sepsityper Kit, Bruker Daltronik GmbH, Germany) for the selective lysis of blood components.

■

RESULTS AND DISCUSSION Selection of Seed-Particles. As bacteria are single micrometer scale organisms, fluidic drag both from acoustic

C

dx.doi.org/10.1021/ac502020f | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

Figure 3. Assessment of changes in the trapping frequency during the in-trap assay can allow quality assurance of the sample preparation. When the cluster of seed-particles is captured, the frequency drops rapidly (a); as water in the channel is subsequently replaced with lysed sample, the frequency gradually increases (b). During capture of bacteria in the lysed blood, the frequency again decreases (c); as the lysed sample is replaced with deionized water during the wash, the frequency first decreases and is subsequently stable (d). Finally, as FA-lysis buffer is aspirated, a large frequency decrease is observed (e).

streaming33 and channel flow will dominate over the acoustic radiation forces, making trapping difficult or impossible. As previously shown for pure E. coli, direct capture does not occur below a critical concentration in the range of (0.5−5) × 107 bacteria/mL.30 Below this concentration, capture can, however, be enabled by using larger seed-particles that retain the bacteria by acoustic particle−particle interaction. This seed-particle aided trapping allowed for continuous enrichment of bacteria (E. coli) in pure buffer at a concentration of 105 bacteria/mL with a capture efficiency of 95%. Due to the viscous sample and the fact that bacteria concentration produced during blood culture (∼107 bacteria/ mL) is not above the critical concentration range for direct capture, 10 μm seed-particles were used to ensure capture. Employing these techniques allowed bacteria to be captured, retained, and enriched in a thin layer against fluid flows around 50 μL/min. The application of this technique for bacteria purification from blood culture bottles required careful selection of seed-particle material in order to avoid background introduced by binding of proteins from plasma and blood cell lysate to the seed-particles. Here, it was found that silica particles provided lower background as compared to polymeric beads of equal size. Frequency Tracking for Quality Assurance. To enable the washing and extraction of bacteria from positive blood culture samples, a recent improvement in the acoustic trapping platform is the use of kerfed transducers to allow continuous optimization of the driving frequency in the acoustic trap.32 This is enabled by online electrical impedance analysis that maximizes the power input to the system. One of the major advantages with frequency tracking in this application is that it also provides a means to monitor the trapping as it takes place, thereby facilitating quality assurance (QA) of the automated sample preparation. As the acoustic properties (compressibility and density) of particles and liquids in the trap will influence the properties of the entire resonator, each step in the protocol will generate a specific frequency signature. The frequency tracker was used to

adapt the driving frequency and display the frequency signature in order to monitor the assay. A characteristic frequency signature for bacteria purification from a blood culture sample is shown in Figure 3 where five characteristic features (a−e) can be observed. As previously published,32 the capture of particles reduces the resonance frequency of the trap. This was again noted during formation of the seed-cluster of silica particles where a drastic frequency decrease was observed (a). Replacing water with the bacteria rich cell lysate was observed as a gradual frequency increase (b), while during capture of bacteria another frequency decrease was noted (c). When the lysed sample is subsequently replaced with deionized water during the wash, the frequency first decreases and then subsequently becomes stable (d). The formic acid used to lyse the bacteria introduces a large frequency shift in the final stage (e). As seen in Figure 3, the total change in frequency from the beginning to the end of the protocol is approximately 200 kHz. It has been previously shown for similar devices that deviating more than approximately 50 kHz from the optimal frequency can have very detrimental effects on the trapping performance.32 A skilled operator may in theory have been able to manually select a frequency in the center of the experimental range. However, on the basis of data from the previous characterization, the 100 kHz deviation from this frequency necessary for this assay implies that flow rate would have had to be reduced at least 1 order of magnitude. For this reason, the frequency tracking not only is useful for QA but also significantly increases the speed and usability of the assay, especially during the FA-lysis. Optimizing Solid Phase Extraction of Bacteria Lysate. As the processed sample is transferred from the trapping site to the open end of the capillary, some dispersion will occur. For this reason, the minimum volume in which the sample could reliably be extracted was in the range of 4−6 μL. This volume is too large to be deposited directly on a standard MALDI plate spot and hence needs to be concentrated down to a volume of 1 μL or less. Initially, direct deposition of both intact and lysed D

dx.doi.org/10.1021/ac502020f | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

Figure 4. Initially, trapped bacteria were washed and released directly on top of the SPE beads together with polymeric seed-beads in the ISET (a). The trapped bacteria were released by dispensing a 10 μL water volume containing the bacteria cluster into the ISET sample preparation platform packed with 20 μm C18 RP beads, in order to decrease the sample volume, followed by on-ISET formic acid lysis of the bacteria. The bottom spectra (b) results after in-trap formic acid lysis and applying the optimized ISET SPE conditions. The optimized SPE protocol used provided much less plasma/blood background and more bacterial protein peaks.

protein amounts resulting after trapping and lysis. Also, it was important to use an SPE media that introduced the least possible amount of bias in the observed peaks, due to the fact that current MALDI Biotyping software is based on reference spectra from pure samples analyzed without the use of any SPE. Due to the strong retention of proteins on the C18 material, a C4 SPE matrix in combination with elution using 80% ACN in the elution buffer provided the best mix of reproducibility and sensitivity. When initially performing the acoustic trapping protocol on positive blood cultures, the trapped bacteria were washed and released directly on top of C18 SPE beads together with polymeric seed-beads in the ISET. This resulted in an identification success rate of roughly 50% compared to the standard Bruker Sepsityper protocol. By applying a combination of in-trap formic acid lysis of the bacteria with the optimized SPE conditions (C4 SPE and 80% ACN elution), without releasing the seed beads, the identification rate of the sample became equal to or even slightly better than that of the standard protocol. Figure 4 shows the difference between deposition of the bacteria directly on top of C18 SPE (no intrap lysis) in the ISET and the protocol using in-trap lysis followed by C4 SPE in the ISET. The observed improvement is probably a result of both avoiding the blood plasma background with the in-trap lysis as the seed-beads are not released into the SPE step and a better recovery in the ISET SPE step. Bacterial Identification of Positive Blood Cultures. Both the acoustic trapping assay and the reference Sepsityper assay were performed on simulated blood cultures where the used incubator detected bacteria at a concentration of approximately 107 bacteria/mL. After selective lysis of the blood cells in the positive blood culture, 100 μL of the sample solution was used for each analysis with the acoustic trap. In the case of the standard Bruker Sepsityper protocol, 1 mL of blood was used for each sample. When the optimized protocol was applied to E. coli grown in blood culture bottles, some preferential binding of larger

bacteria on a MALDI target was tested but resulted in no data or inadequate data quality. Concentration and purification of the proteins resulting from in-trap lysed bacteria is most conveniently done by solid phase extraction (SPE). Here, there are many different options available, such as tip-based SPE,36,37 magnetic beads,38,39 or miniaturized technologies. The integrated selective enrichment target (ISET) SPE platform applied here has previously been demonstrated as a fast, sensitive, and automatable sample preparation method in applications such as purification of immuno-captured peptides35 or screening of recombinant proteins.40 One caveat with the application of SPE for bacterial protein sample preparation is that there are very little published data on using SPE in combination with bacteria MALDI Biotyping. In order to validate the application of a SPE step in this setting, a thorough investigation of the impact of the SPE on the mass spectrometric read-out was undertaken. In a first series of experiments, four different SPE phases (C4, SCX, HILIC, and C18) were tested in the ISET. This was done in order to see if any beneficial complementary peaks would be generated due to the use of different affinities.41−43 Regardless of the type of SPE media used, the peaks observed in the resulting spectra where surprisingly similar (Supporting Information, Figure S1); a reasonable explanation for this is the high abundance of the proteins used for bacteria identification. Using the micro Bradford assay (Thermo Scientific, Rockford, IL, USA), the amount of protein available after applying the standard Bruker protocol with the used E. coli was determined to be approximately 200 μg/mL. Due to the seed-beads and aggregation, the number of bacteria captured in the acoustic trap was not possible to obtain using conventional counting, but from the MALDI spectra intensities observed, there was very little difference between 1 μL of standard protocol prepared sample and the amount obtained by trapping. This would give a rough estimate of 200 000−500 000 bacteria in the trap. Hence, the 10 picomole SPE capacity afforded by the ISET platform was sufficient to handle the E

dx.doi.org/10.1021/ac502020f | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

Figure 5. A comparison of the reference peaks in the MALDI Biotyper-database for E. coli DH5α (a), the peaks obtained using the standard sample preparation method (b), and the peaks from the acoustic trapping ISET SPE method (c). As shown by overlaying standard and acoustic trapping spectra (d), the peaks observed in the standard sample preparation can also be found after the acoustic trapping.

Table 1. Successful Identification of Escherichia coli from Blood Samples in 12 out of 12 Cases Using Acoustic Trappinga sample ID

organism (first match)

score value

Trap1 (DH5α) Trap2 (DH5α) Trap3 (DH5α) Trap4 (DH5α) Trap5 (DH5α) Trap6 (DH5α) Trap7 (DH5α) Trap8 (DH5α) Trap9 (DH5α) Trap10 (DH5α) Trap11 (DH5α) Trap12 (DH5α)

E. E. E. E. E. E. E. E. E. E. E. E.

coli coli coli coli coli coli coli coli coli coli coli coli

2.011 2.065 2.256 2.292 2.211 2.219 2.144 2.312 2.124 2.22 2.167 2.245

STD STD STD STD STD

PR PR PR PR PR

(DH5α) (DH5α) (DH5α) (DH5α) (DH5α)

E. coli E. coli E. coli E. coli no peaks

2.098 1.953 1.959 1.92