Bacterial strain, culture medium, and chemicals. Mycobacterium bovis (BCG, Tokyo strain) was purchased from Japan BCG Laboratory (Tokyo, Japan) and maintained in BD BACTEC™ MGIT™ medium supplemented with OADC Enrichment and PANTA antibiotic mixture (Becton Dickinson Co., 245122, Franklin Lakes, NJ) at 37°C. Mid-log-phase BCG was prepared at concentrations of 1,000 CFU/µl for time-lapse microscopy and 10,000 CFU/µl for drug susceptibility testing in MGIT media by using a bacterium counting chamber (A161, SANSYO, Tokyo, Japan). RFP, INH, PZA, SM, EB, LVFX, TH, and CS were obtained from FUJIFILM Wako (Tokyo, Japan). KM and EVM were obtained from Nacalai Tesque (Tokyo, Japan) and Asahi-Kasei Pharma (Tokyo, Japan), respectively. DLM and BDQ were obtained from Selleck (Houston, TX). These chemicals were dissolved in deionized water (INH, PZA, SM, EB, TH, KM, EVM, and CS), methanol (RFP, TH), 0.1 N hydrochloride (LVFX), or dimethyl sulfoxide (DLM, BDQ), and then sterilized using 0.22-µm MILLEX-GV syringe filters (Millipore, Burlington, MA) prior to being added to MGIT media under aseptic conditions. Perfluorocarbon (Fluorinert™, FC-3283) was obtained from 3M (Maplewood, MN). Live/Dead™ BacLight™ bacterial viability reagent was purchased from Thermo Fisher Scientific (L13152, Waltham, MA). Detailed information on the antibacterial agents used in this study is shown in Supplementary Table 1.
Sealed culture and time-lapse microscopy. The bacterial suspension (5,000 CFU of BCG in 5 µl of medium) was placed into a single-well glass dish (IWAKI, 3971 − 101, Tokyo, Japan) and sealed with FCO. The culture well was then covered with glass and sealed with a liquid pressure-sensitive adhesive (BBX909, Cemedine™, Tokyo, Japan). Time-lapse microscopy images (BioStation IM®, Nikon, Tokyo, Japan) were taken every 10 min for 168 h at 37°C. The original magnification was 20×. Microscopic images were constructed using BioStation IM software (version 2.30, Nikon). Slice view images were constructed using NIS-Elements (version 5.10.00, Nikon).
Liquid culture and bacterial viability assay. BCG was cultured with shaking in a BD BBL™ MGIT™ mycobacteria growth indicator tube (4 ml, 37°C, 300 rpm). The increase in CO2 concentration due to bacterial growth was confirmed with a fluorescence indicator. After a pellet of BCG cells was collected via centrifugation at 3,000 rpm for 10 min, the cell concentration was adjusted to 10,000 CFU/µl. RFP, SM, EB, INH, and PZA were added to 500 µl (5,000,000 CFU) of BCG culture at 2× the CDC-recommended critical concentration, and the culture was then placed in an incubator-shaker (37°C, 300 rpm). After 2 or 4 h, the BCG cultures were collected, stained with Live/Dead BacLight bacterial viability reagent in accordance with the manufacturer’s protocol, and subjected to fluorescence microscopy (FITC/Texas red) on a Nikon BioStation.
65-GHz near-field sensor array and culture system. The sensor array consisted of 1,488 LC oscillators made via Complementary metal-oxide-semiconductor (CMOS) technology in a 3-mm square, and its oscillation frequency was designed to be 65 GHz. Bulk water formed a hydrogen bond network with a disappearance time on the order of picoseconds. The theoretical peak of the dielectric constant in the original water molecules was approximately 20 GHz based on the rotational relaxation according to dielectric spectroscopy data. Because the 65-GHz oscillation frequency of this sensor was located on the slope of its large peak, it was sensitive to changes in the amount of bulk water. The surface of the oscillator was covered by SiO2 passivation, and the surface was sensed by contacting the sample. Each oscillator was 50 µm × 110 µm in size (including the inductor), and the electric field was localized in the proximity region approximately 15 µm from the passivation layer. Because the resonance frequency of the oscillator (described below) included dielectric constant in the vicinity of the oscillator, the resonance frequency fres changed according to the change in dielectric constant (\({\epsilon }^{*}={\mathcal{E}}_{0}\left({\mathcal{E}}_{r}-j{\mathcal{E}}_{i}\right)\)) of the sample.
\({f}_{\text{r}\text{e}\text{s}}=\frac{1}{2\pi \sqrt{{L}_{\text{0}}{C}_{\text{eff}}}}\) \({C}_{\text{eff}}={C}_{0}+{{C}_{1}C}_{\text{a}\text{i}\text{r}}\frac{{{\epsilon }_{r}C}_{1}+{C}_{\text{a}\text{i}\text{r}}({\epsilon }_{r}^{2}+{\epsilon }_{i}^{2})}{{\left({C}_{1}+{\epsilon }_{r}{C}_{\text{a}\text{i}\text{r}}\right)}^{2}+{\left({\epsilon }_{i}{C}_{\text{a}\text{i}\text{r}}\right)}^{2}}\) (1)
where \({C}_{\text{eff}}\), \({L}_{\text{0}}\), \({C}_{\text{0}}\), \({C}_{1},\) and \({C}_{\text{a}\text{i}\text{r}}\) are, respectively, the effective capacitance (including the passivation layer on the sensor surface), the inductance, the circuit capacitance, the capacitance of the passivation layer, and the capacitance with vacuum permittivity (8.85 pF/m) calculated by electromagnetic field simulation.
When one oscillator captured data with a gate time of 200 µs, the time to collect data for all 1,488 elements was 0.5 s or less. The frequency resolution of each oscillator under these conditions was estimated to be 0.33 MHz. Because the dielectric constant of water highly depends on the temperature in the 65-GHz band, temperature sensors were mounted at the four corners of the sensor; when the 1,488 data points were captured, the temperature at that time was also recorded. The temperature change and the change in resonance frequency using ultrapure water were measured at − 20 MHz per + 1°C. Therefore, to control the housing of the sensor, an air circulation incubator (Cool Incubator ICI-1, ASONE Co., Ltd., Osaka, Japan), a Peltier temperature controller (VPE-35, VICS Co., Ltd., Musashino, Japan), and a panel glass heater (ST-H070, BLAST Co., Ltd., Kawasaki, Japan) were used. The temperatures were set to 36°C, 40°C, and 42°C for the air incubator, Peltier controller, and panel glass heater, respectively, and the temperature was controlled such that the thermometer with a built-in sensor would reach 36°C. The culture system of the sensor is shown in Supplementary Fig. 1. The measured data were saved to a computer via a USB cable.
Sealed culture on sensor and drug susceptibility testing. BCG (30,000 CFU/3 µl medium) was placed on the sensor. FCO (200 µl) and MGIT medium (100 µl) were then layered, and an adhesive film (Ultra Amp plate seal 36590, Sorenson BioScience, Inc., Salt Lake City, UT) was used to ensure an air-tight seal. A layer of MGIT medium was used to avoid water evaporating from the bacterial suspension (Supplementary Fig. 1). BCG was cultured at 37°C for 3 h. The difference in the resonance frequency (MHz) was measured at 5-min intervals for up to 36 h. The BCG growth ability was confirmed if the difference increased by + 5 MHz or greater during culturing. Next, an antibacterial drug or MGIT medium (2 µl) was injected into the bacterial suspension. The final concentration of each antibacterial drug is listed in Supplementary Table 1. For the first-line therapeutic agents (RFP, INH, EB, and PZA) and SM, DST was attempted using drug concentrations of 2× the CDC-recommended critical concentration in the MGIT liquid culture. For the other tested drugs, susceptibility testing was conducted at two different drug concentrations (shown in Supplementary Table 1). Each drug was diluted in BD BACTEC MGIT medium (pH 6.74) in accordance with the manufacturer’s instructions. Because the manufacturer of the PZA drug susceptibility test recommended that it be performed at pH 5.85, this test was performed at pH 5.85 as well as at the pH 6.74 used testing the other drugs. For drug-free controls, an equal volume (2 µl) of MGIT liquid medium was administered instead of an antibacterial drug.
Measurements and data analysis. The magnitude of change in resonance frequency of an element is indicated by the average difference in frequency (MHz) and SD (shown in figures as the error bar). From the start of culture to the end of measurement, the entire region where the bacterial suspension was present on the element and for which no change of − 10 MHz or less was observed before or after injection of the drug solution was used as the measurement area. Of the 1,488 elements on the sensor array, 745–1,260 elements (average: 1,039; SD: 150) were automatically selected by the computer program for use in analysis. In accordance with these criteria, sensor elements that lost bacteria from the near-field after the injection of antibacterial drug or liquid medium were removed by an automated program. The growth ability was indicated by the mean difference in resonance frequency (MHz) starting from 0 h and the SD. DST was performed at two concentrations for some drugs (Supplementary Table 1). There were three control experiments, and one experiment for each drug. Because the temperature was controlled within ± 0.1°C, the effect of each drug on BCG was determined from the time course of the difference (MHz) of each sensor element without correcting for changes in temperature. A histogram analysis was performed at 16 h after drug administration by using elemental measurements from each sensor (Fig. 3). The data analysis and graphs were prepared by using GraphPad Prism™ (version 9.2.0, GraphPad Software, Inc., San Diego, CA).