Information ( ESI ) for : Detection of antimicrobial resistance-associated proteins by titanium dioxide-facilitated intact bacteria mass spectrometry

a. Laboratoire d’Electrochimie Physique et Analytique, Ećole Polytechnique Fédérale de Lausanne, Rue de l'industrie 17, CH-1951 Sion, Switzerland. b. ISIC-GE-VS, Ećole Polytechnique Fédérale de Lausanne, Rue de l'industrie 17, CH-1951 Sion, Switzerland. c. Department of Chemistry, Fudan University, Handan Road 220, 200433 Shanghai, China. d. ICH, Hôpital du Valais, Avenue du Grand Champsec 86, CH-1951 Sion, Switzerland. e. Laboratoire de Chimie Biophysique des Macromolécules, Ećole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland.


Fabrication of TiO2 NPs-modified MALDI target plates
TiO2 NPs aqueous suspension was prepared according to a previously reported method. 1 Briefly, 1 g of P25 TiO2 nanopowder was beforehand heated at 300 °C for 2 h and then separated in a mortar for 3 h. During the separation process, 1 mL of 10% acetic acid aqueous solution was added drop by drop to keep the nanopowder wet. The separated nanoparticles were suspended in an aqueous solution of ethanol (89%, V/V) to reach a concentration of 100 mg·mL -1 , followed by ultra-sonication for 1 h. The suspension was then diluted 25 times with DI water to reach a final concentration of 4 mg·mL -1 . Such obtained TiO2 NPs suspension was stored at 4 °C and was stable for four to six months.
The TiO2 NPs-modified MALDI target plates were prepared by depositing the TiO2 suspension onto the spots of a classic bare stainless steel target plate (here, commercial MSP 96 ground steel target, Bruker Daltonics). The suspension was deposited by drop casting or by dispenser in the following manner: 2 μL of the suspension was firstly dropped onto each spot and air-dried for ~10 min; thereafter, another 2 μL of the suspension was dropped to cover the previous one and again air-dried for ~10 min. Alternatively, the TiO2 suspension could also be dropped as an array of spots onto a piece of stainless steel foil (20 μm thick), which was afterwards affixed onto a commercial bare target plate before MALDI-TOF MS measurement. The dried target plate or steel foil was heated at 400 °C to sinter the nanoparticles. The heating process was accomplished with a three-step automatic program: raising temperature from 25 to 400 °C within Electronic Supplementary Information 1h, keeping the temperature at 400 °C for 1h, cooling down to 25 °C within 4 h. Alternatively, instead of thermal heating at 400 °C, the sintering process could also be completed by photonic curing, which takes only a few miliseconds. The photonic curing was conducted using high intensity light pulses from a xenon flash lamp provided by a PulseForge 1300 photonic curing station (NovaCentrix, USA). The curing parameters were set as: 5 pulses exposure, 450 V bank voltage, 1 ms pulse duration. Through sintering, a stable layer of TiO2 NPs was formed and firmly attached on the steel surface.
The sintering process did not change the crystalline phase of TiO2, which was kept as the mixed rutile (110) and anatase (101) (mainly) phase. Before the usage, prepared TiO2-modified target plate or steel foil pieces were stored in a clean and dry room temperature environment. antimicrobial resistance in each resistant strain, were provided by a local hospital (Hôpital du Valais, Sion, Switzerland). Their detailed antimicrobial susceptibility profiles were measured using a bioMérieux VITEK 2 automated AST system, based on antimicrobial drugs culture method. These strains were directly analyzed with MALDI-TOF MS when they were obtained.

Incubation of bacteria
Concentrations of bacterial cells in bacteria samples were determined by measuring the optical density at 600 nm (OD600nm). 1.0 OD600nm corresponds to ~8 × 10 8 cells·mL -1 .

Transfer of resistance genes into bacteria
Plasmid DNAs carrying a specific antimicrobial resistance gene were transformed into two Escherichia coli (E. coli) Electronic Supplementary Information strains, i.e. two of DH5α, XL1-Blue or BL21, according to the protocol provided by Sambrook and Russel. 2 The plasmids utilized were: pBluescriptIISK(+) carrying resistance against ampicillin (Stratagene, California, USA), pEGFP-N1 carrying resistance against kanamycin (Clontech, California, USA), pEN_TmiRc3 carrying resistance against gentamycin (addgene, Massachusetts, USA) and pOFXT7-2 carrying resistance against chloramphenicol (donated as a gift from University of Lausanne, Lausanne, Switzerland). Specifically, 1.5 mL of each E. coli pre-culture was overnight incubated in 50 mL of LB medium at 37 °C with continuous shaking at 200 rpm. The culture was transferred into an ice-cold polypropylene tube and cooled on ice for 10 min. E. coli cells were separated from the growth medium by centrifugation at 2700 × g for 10 min at 4 °C. After thoroughly removing the growth medium, the cell pellet was suspended in 30 mL of ice-cold MgCl2-CaCl2 solution (80 mM MgCl2, 20 mM CaCl2) by gentle vortexing. The cells were again collected by centrifugation at 2,700 × g for 10 min at 4 °C, and gently resuspended in 2 mL of ice-cold 0.1 M CaCl2. The resulting competent E. coli were either directly used for transformation as described below or dispensed into aliquots and stored at -70 °C.
For transformation, 200 μL of above obtained competent E. coli cells were mixed with 100 ng of purified plasmid DNAs that carried a specific resistance gene. The mixture was incubated on ice for 30 min, and then applied with a heat shock for exactly 90 s in a 42 °C water bath. Thereafter, the mixture was rapidly transferred into an ice bath and chilled for 2 min. After adding 800 μL of LB medium, the mixture was incubated for 45 min in a 37 °C water bath to let the bacteria recover and express the antibiotics resistance marker encoded by the plasmid. The obtained bacteria were spread on selective agar plates containing 20 mM MgSO4 and appropriate antibiotic. Single colonies were picked up and added into 2 mL of LB for overnight incubation at 37 °C with continuously shaking at 200 rpm. The resulting ampicillin-resistant, kanamycin-resistant, gentamicin-resistant or chloramphenicol-resistant E. coli strains were either directly analyzed with MALDI-TOF MS or incubated with antibiotics as described below.

Experimentally controlling the resistance gene expression level within E. coli
In order to experimentally control the resistance gene expression level within bacteria, the above obtained ampicillin-resistant, kanamycin-resistant and chloramphenicol-resistant E. coli DH5α cultures were diluted 100 times with LB medium containing gradually increased concentration of corresponding antibiotic (i.e. ampicillin, kanamycin and chloramphenicol, respectively). The concentrations of each antibiotic were selected as 0, 15, 30, 60 and 120 μg·mL -1 .
The maximum concentration value (120 μg·mL -1 ) was lower than the minimum inhibitory concentration to allow the proliferation of bacterial cells.
To investigate the influence of growth medium, ampicillin-resistant E. coli DH5α were diluted 100 times with different growth media (LB medium, or 2xYT medium) containing 60 μg·mL -1 of ampicillin. Electronic Supplementary Information 6 All of the above mixtures were incubated overnight at 37 °C with continuously shaking at 200 rpm. The obtained fresh cultures were afterwards analyzed with MALDI-TOF MS.

MALDI-TOF MS fingerprinting of intact bacteria with TiO2 NPs-modified target plates or classic bare target plates
For each bacteria sample, bacterial cells were harvested from the growth medium by centrifugation at 7,500 × g for 3 min, then washed with DI water for two times. The cell pellet was finally suspended in DI water to reach a concentration of ~5 × 10 8 cells·mL -1 . The obtained intact bacteria aqueous solutions were deposited onto the spots of a TiO2-modified target plate or a classic bare target plate (1 μL for each spot) by drop casting, then air-dried (~5 min). Each test was performed in triplicates by depositing the bacteria solution on three spots. Sinapinic acid matrix (15 mg·mL -1 in 50/49.9/0.1% acetonitrile/water/ trifluoroacetic acid) was dropped onto the dried sample spots to overlay the bacterial cells (1 μL for each spot). Bacteria/sinapinic acid co-crystals were formed after air-drying for ~5 min. Thereafter, the target plate was loaded into a MALDI-TOF MS instrument for measurement.
In order to compare the behavior of different matrices for intact bacteria fingerprinting, the dried sample spots with E.
Throughout the present work, all MALDI-TOF MS measurements were conducted with aqueous solutions of intact whole bacteria, except that protein extracts from E. coli DH5α was utilized for the characterization of TiO2-modified target plates in Part S7. This protein extracts was prepared according to the typical ethanol/formic acid/acetonitrile extraction protocol, which is described in the Bruker MALDI Biotyper 3.0 user manual (2011). Briefly, E. coli DH5α cells were harvested from 1 mL of fresh culture by centrifugation at 7,500 × g for 3 min and washed two times with DI water.
The cell pellet was suspended in 300 μL of DI water, followed by adding 900 μL of ethanol and vortexing thoroughly. The water and ethanol was completely removed by centrifugation at 8,500 × g for two times (2 min for each time) and air-drying for ~30 min. The obtained cell pellet was resuspended in 50 μL of 70/30% formic acid/water. The mixture was vortexed thoroughly and let stand for 5 min, followed by adding 50 μL of acetonitrile. The final mixture was vortexed thoroughly for another 5 min. Thus, intracellular proteins were extracted into the solvent. After centrifugation at 8,500 × g for 2 min, the supernatant was pipetted onto three spots of a TiO2-modified target plate and a bare target plate (1 μL for each spot). After air-drying for ~5 min, each sample spot was overlaid with 1 μL of sinapinic acid matrix (15 Electronic Supplementary Information mg·mL -1 in 50/49.9/0.1% acetonitrile/water/trifluoroacetic acid) and again air-dried (~5 min).
All measurements were conducted with Bruker MicroFlex LRF MALDI-TOF MS instrument (Bremen, Germany) under linear positive mode at 20 kV accelerating voltage. Instrumental parameters were set as: mass range 2,000-80,000 m/z, laser intensity 70%, laser attenuator with 30% offset and 40% range, 500 laser shots accumulation for each spot, 20.0 Hz laser frequency, 20× detector gain, suppress up to 1000 Da, 350 ns pulsed ion extraction.
At the beginning of each measurement, mass calibration was conducted with a calibration sample containing 1 All experiments related to bacteria and antibiotics were conducted in a biosafety level 1 or 2 (P1 or P2) laboratory.
Experimental supplies including centrifuge tubes and micropipette tips were disposable and sterile. All wastes were autoclaved and disposed properly according to the safety guidelines. Instruments, facilities and benches were wiped with 70/30% ethanol/water when experimental activities were finished.

Data analysis
Mass spectral fingerprint patterns of bacteria were visually compared using the mMass Open Source Mass Spectrometry Tool (http://www.mmass.org/). To facilitate data interpretation, the patterns were compared in three separate sections, i.e. 2,000-15,000 m/z, 15,000-29,000 m/z and 29,000-80,000 m/z. The peak numbers in each section were automatically countered by using the mMass tool, with "peak picking" parameters set as: S/N threshold 3.0, relative intensity threshold 2.0% for 2,000-15,000 m/z and 0.1% for 15,000-80,000 m/z, apply smoothing, remove shoulder peaks.
The 36 fingerprint peaks newly detected from Bacillus subtilis (strain 168) using a TiO2 NPs-modified target plate in contrast with a classic bare one were analyzed by proteome database search, according to the method provided by Fenselau et al. 3,4 The search was conducted against Bacillus subtilis 168 proteome database (UniProtKB, proteome ID UP000001570) on UniProtKB/Swiss-Prot/TrEMBL database platform (http://www.uniprot.org/). The primary search parameter was the experimental m/z value of each peak (averaged value from three replicates). During the search process, the mass window was selected as 300 ppm (± m/z experimental × 0.03%) for the mass range of 2,000-80,000 m/z.
Information about subcellular location of each tentatively assigned protein was directly obtained from the database.
Their isoelectric point (pI) and grand average value of hydropathicity (GRAVY) was obtained using the Electronic Supplementary Information 8 ExPASy-ProtParam computation tool (http://web.expasy.org/protparam/) by entering the UniProtKB/ Swiss-Prot/TrEMBL accession number of each protein and following the links to "Compute parameters" and later "Submit".
Fingerprint patterns of a testing bacteria and a reference bacteria were mathematically compared using a public bacteria identification service platform Bacteria MS (http://bacteriams.fudan.edu.cn/#/). Raw data files (in '.txt' format) of three replicated patterns from a testing and a reference bacteria were uploaded onto this platform. The system then automatically averaged the three replicates, and the averaged patterns of the testing and the reference bacteria were compared by choosing the cosine correlation method. Information including identical peaks, different peaks, their normalized intensities and the similarity score of the two patterns were thus obtained. As illustrated in our previous work, 5 with the cosine correlation algorithm, the similarity score between two mass spectra (i and j) was defined as: where y is the normalized intensity of a peak appearing in both spectra i and j (identical peak), l is the number of identical peaks in the two spectra, Y is the normalized intensity of a peak appearing in a spectrum, n is the number of peaks in a spectrum. Only peaks with S/N ≥ 3 were taken into account. Peaks appearing in different spectra with Δ(m/z)/(m/z) ≤ 300 ppm were considered as identical peaks. This 300 ppm tolerance was chosen according to the resolving power of linear mode TOF analysis. It has been demonstrated that a similarity score of ≥ 0.8 ensures a successful identification at the species level.

Figure S1
Comparison of different matrices for intact bacteria MALDI-TOF MS fingerprinting. The patterns were generated from intact Escherichia coli (strain DH5α) in the mass range of 2,000-80,000 m/z with the utilization of four matrices: α-cyano-4-hydroxycinnamic acid (CHCA), 2,5-dihydroxybenzoic acid (DHB), sinapinic acid (SA) and trans-ferulic acid (FA). A classic bare stainless steel MALDI target plate was employed. Number of bacterial cells on each spot was around 5×10 5 . Each pattern was averaged from three replicates.
As shown in Fig. S1. CHCA, DHB and SA matrices exhibited high detection sensitivity in the typical mass range for bacteria identification (2,000-15,000 m/z). They also provided satisfying reproducibility, with pattern similarity scores > 0.99 for three replicates. The similarity scores between MS patterns were calculated using a public bacteria identification service platform BacteriaMS (http://bacteriams.fudan.edu.cn/#/) by choosing the cosine correlation algorithm, which gives the maximum score as 1.0. Proteins detectable with CHCA and DHB were smaller than ~15,000 Da, and the ones detectable with SA were a little larger. Meanwhile, FA matrix gave more peaks in the mass range of Electronic Supplementary Information 15,000-80,000 m/z, but the sensitivity in 2,000-15,000 m/z was drastically decreased together with the reproducibility, giving pattern similarity scores lower than 0.5 for three replicates. Thus, SA matrix was selected for further study in this work, due to its ability to generate MS patterns with high quality within 2,000-15,000 m/z and higher for detection of large bacterial components. TiO2 has a unique electronic structure, which is characterized by a filled valence band and an empty conduction band.

Electronic Supplementary Information
With the band gap of 3.0-3.2 eV (3.0 eV for rutile and 3.2 eV for anatase), TiO2 have strong absorption in UV range (< 400 nm). MALDI techniques typically use UV lasers such as nitrogen lasers (337.1 nm) and frequency-tripled and quadrupled Nd: YAG lasers (355 nm and 266 nm, respectively). Thus, TiO2 NPs on the surface of MALDI target plate can absorb energy from the laser source during MALDI-TOF MS measurement.

S5. TiO2-triggered photocatalytic reactions related to the generation of reactive oxygen species 6
TiO 2 + ℎ → e − + h + (1) On each sample spot, there were 1 µL of protein mixture and 1 µL of sinapinic acid matrix. Each pattern was the overlay of three replicates (in red, blue and black color, respectively). In the present work, each MALDI-TOF MS test was repeated three times; in each replicate, a freshly cultured bacteria strain was measured. Collected bacterial fingerprint patterns demonstrated high reproducibility. For example, in Fig.S11, the pattern similarity score is 0.9967 between replicate 1 and 2, 0.9978 between replicate 1 and 3, and 0.9989 between replicate 2 and 3. The similarity scores were calculated using the BacteriaMS platform with cosine correlation algorithm.

S12. A list of proteome database search result
Herein, we show the detailed proteome database search result of the 36 fingerprint peaks newly detected from Bacillus subtilis (strain 168) using a TiO2 NPs-modified target plate in comparison with a classic bare one (in main article Fig. 3c)  In the main article, it has been demonstrated that five different types of antimicrobial resistance-associated proteins were successfully detected from intact bacteria cells by MADLI-TOF MS fingerprinting method when TiO2-modified target plates were used, as shown in Fig.4. As a comparison, all the measurements in Fig.4 were repeated with classic bare stainless steel plates. The mass spectra in corresponding mass range were shown in Figure S14. Clearly, none of the protein markers were detected with classic bare stainless steel plates, confirming the advantages of TiO2-modifed target plates.

S16. Lists of antimicrobial susceptibility profiles
Results of antimicrobial susceptibility testing conducted with bioMérieux VITEK 2 automated AST system, based on antimicrobial drugs culture method, are listed in Table S1-S12.             In the main article, a peak at 13,080±2 m/z was exclusively detected from a MRSA strain (Fig.6 c). This peak could come from the characteristic fragment of PBP 2a protein, which confers the methicillin-resistance. To further confirm this assumption, two more MRSA strains (provided by Sion Hospital, Switzerland) were tested. As shown in Figure S17, the peak around 13 kDa was also successfully detected from the two MRSA strains, at 13,083±3 and 13,081±4 m/z, respectively. The detailed antimicrobial susceptibility testing profiles of the two MRSA strains are listed in Table S11-12, Part S16 in this supporting document.