Hardware subsystems

Experimental techniques that exploit interactions between light and life, such as light-sensitive proteins, optogenetics, and fluorescent reporters, are ubiquitous in biological research [25, 26]. Chi.Bio contains an array of optical outputs and sensors to support these techniques (Fig 2A). A 650-nm laser is used for OD measurement (calibrated against a Spectrophotometer, S1E Data) and is driven by an analogue feedback circuit to provide stable, temperature-insensitive readings (S1F Data). Optogenetic actuation and excitation of fluorescent proteins employs a focused high-power seven-colour light-emitting diode (LED) with six emission bands across the visible range and a 6500K white output (Fig 2B and 2C), and a separate 280-nm UV LED is included to stress or kill cells in the growing population. Each LED has an independent current-limiting pulse-width modulation (PWM) driver, allowing its intensity to be regulated over three orders of magnitude (S1C Data), and is thermally coupled to the outside of the device to reject heat when operating at high intensity. The combined LED implementation provides optical outputs that exhibit minimal power and spectral variability between devices or environmental conditions (S1G Data).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Hardware subsystems, software, and automation.

(a) The main reactor with sides removed and subsystems labelled. The vertical PCB on the left hosts driving circuitry for many subsystems, as well as power regulation and filtering. (b) Emission spectra of the visible optical outputs in the device (UV LED peak is at 280 nm). (c) Images of optical outputs, with laser set to 50% and LEDs to 5% intensity. LEDs are focused and perpendicular to the spectrometer to maximise faint fluorescence signals. (d) Measurement filter bands of the platform's spectrometer. (e) Magnetic stirring provides a powerful vortex when set to a high rate. (f) Software architecture, which packages multiplexed low-level commands (digital communications following I²C standard) into an easy-to-use web interface accessed from a connected PC or network. (g) A typical 60-second experimental automation cycle. Initially stirring is disabled so liquid can settle, reducing noise in measurements and providing a flat surface for removal of waste media. (h) Media temperature (‘Temp’) controlled to follow a predefined path over 5 hours. Heat input is provided by the heat plate; cooling is passive. (i) The OD of Escherichia coli in exponential growth phase, maintained within approximately 2% of its set-point. (j) OD can be set to follow a dithered waveform (with cells rapidly diluted to a lower OD value whenever the upper limit is reached) to accurately measure growth [27]. Following a change in temperature set-point from 25°C to 37°C growth accelerates significantly. FP, fluorescent protein; LED, light-emitting diode; OD, optical density; OS, operating system; PC, personal computer; PCB, printed circuit board.

https://doi.org/10.1371/journal.pbio.3000794.g002

Measurements of light intensity are performed within the device by a chip-based spectrometer with eight optical filters covering the visual range, as well as an un-filtered ‘Clear’ sensor (Fig 2D). Multiple wavelength bands can be measured simultaneously, each with electronically adjustable gain and integration time. The spectrometer is set up to perform temperature and long-term baseline calibration using a dark photodiode prior to every measurement. Typical spectrometer measurements (e.g., of fluorescence) are reported as the ratio of light intensity measured at the fluorescent protein's emission band to the total intensity of the excitation source; this ratiometric measurement mitigates the impact of differing excitation intensity between devices and spatial variations in culture density.

Culture temperature is measured noninvasively by a medical-grade infrared thermometer, which is accurate to ±0.2°C for temperatures near 37°C. There are also air temperature thermometers (±0.5°C accuracy) within the main reactor and on the control computer for monitoring the surrounding environment. Temperature change is actuated by a PCB-based heat plate, capable of heating a 20-mL culture at up to 2.0°C min⁻¹. Below the heat plate is a magnetic stirring assembly built upon an off-the-shelf fan, which has an adjustable stirring rate (Fig 2E) and can be used with standard laboratory stir bars for aeration and mixing (S1H Data). The main reactor also includes an external expansion port, which provides power and a digital interface for user-built add-ons to Chi.Bio.

External to the main reactor is a pump board, which can house up to four low-cost peristaltic pumps with independently controllable speed (up to 1 mL s⁻¹) and direction. Each pump transfers liquid to/from the culture test tube via standard 4.5-mm silicone laboratory tubing, which can be installed without joints from input to output to assist sterilisation. Typically, two pumps are dedicated to turbidostat functionality (one for input of fresh media, one for removal of waste), leaving the other two free for programmable mixing of media/inducers, or transfer of liquid between reactors. Detailed specifications of Chi.Bio's hardware and electrical subsystems are outlined in S1A–S1D Data, and analysis of their calibration, measurement stability, and interdevice variability is described in S1E–S1H Data.

Software and automation

At the lowest level of Chi.Bio's open-source operating system is a multiplexed I²C bus for digital communications within the device (Fig 2F). Digital signals are in turn controlled by the hardware operating system (Python, S1I Data), which implements automation functions as well as data collection, processing, and storage (in.csv format). Real-time data are output periodically via a web server (accessible from a connected PC or network) that provides an easy-to-use web user interface (built in HTML/JavaScript, S1J Data). The complete software stack lets users set up, control, and monitor standard experiments entirely through their web browser, and more advanced protocols or algorithms that combine any of the platform's measurement/actuation capabilities can be easily implemented using the in-built ‘custom program’ Python framework (S1K Data).

A typical automated experiment follows a 60-second cycle (Fig 2G) during which the stirring vortex is allowed to settle prior to measurement, media/waste is added/removed, and data are processed to calculate new control inputs. This automation protocol can be adjusted in real time to change controller set-points or measurement setup, all of which are recorded throughout an experiment. Automated temperature and OD regulation are implemented using proportional-integral-derivative (PID) and model predictive control (MPC) algorithms [28]. The temperature control algorithm (S1L Data) facilitates rapid temperature changes with minimal error and set-point overshoot (Fig 2H). The OD control algorithm (S1M Data) typically maintains OD within approximately 2% of its set-point (Fig 2I) and can also be used to dither a culture's OD about a central set-point (Fig 2J) for precise measurement of growth rate at near-constant density [27].

Strain and plasmid selection

E. coli strain MG1655 was used for growth/biofilm characterisation experiments. Fluorescent protein experiments utilised BL21(DE3) E. coli, apart from the optogenetics experiments that utilised BW29655. Bacterial strains were made chemically competent by treatment with calcium chloride and transformations were performed via heat shock. All E. coli strains used in this study are freely available (S1R Data), as are plasmids (S1S Data).

Culture techniques

Cells were cultured in either LB (growth/biofilm experiments) or EZ rich defined media (Teknova; cat: M2105) supplemented with 1% (v/v) glycerol (for optogenetic/fluorescence experiments). All experiments were performed in 20-mL volume in 30-mL clear borosilicate glass test tubes (Fisher Scientific, 11593532) open to atmosphere. Stirring employed disc stir bars (Fisher Scientific, 11878892) at speed setting 0.6. Culture temperature was maintained at 37°C unless otherwise stated. For all fluorescent protein–measurement and optogenetic-control experiments, growing cells were maintained at an OD of 0.4 using the system's turbidostat functionality. Carbenicillin (a semisynthetic analogue of ampicillin with greater stability), chloramphenicol, and spectinomycin were used at final concentrations of 100 μg mL⁻¹, 25 μg mL⁻¹, and 50 μg mL⁻¹, respectively.

Experiment setup procedure

Prior to experiments, test tubes and stir bars were sterilised by autoclave. Silicone tubing was sterilised by pumping 70% ethanol for 10 seconds and media for 5 seconds. For experiments that required chemical induction (Application 3), an additional 500 mL of tap water was pumped through each tube prior to the experiment to prevent inducer cross contamination. Tubing exterior was sterilised via swabbing with 70% ethanol. For each media type, each reactor's OD zero point was calibrated with a test tube of fresh media as described in S1E Data. Test tubes were filled with 20 mL media and inoculated prior to insertion into each reactor. All experiments were run directly through the Chi.Bio web interface, which was supplemented with ‘custom programs’ where indicated.

Biofilm experiments

In each trial, cells were initially diluted to an OD of 0.1 and (once reached) maintained at an OD of 0.5. Each experiment's zero-time is defined as the time when the culture first reached OD 0.5 (typically about 2 hours post inoculation). To develop a biofilm-forming phenotype, E. coli was grown at high density (OD >1) in LB media in a reactor in chemostat mode for 120 hours. Samples from this culture were subsequently used to inoculate each of the ‘adapted’ biofilm trials.

Fluorescence measurements

GFP was excited by the 457-nm LED at power setting 0.1 and measured at ×512 gain and 0.7-second integration time using the Clear filter as base-band and 550-nm filter as emission band. RFP was excited by the 595-nm LED at power setting 0.1 and measured at ×512 gain and 0.7-second integration time using the Clear filter as base-band and 670-nm filter as emission band. A baseline fluorescence (corresponding to the fluorescence measured for wild-type cells without plasmid) was subtracted from each measurement. Measurements were then smoothed using a moving mean filter with 10-minute width. Examples of raw (unprocessed) data and a detailed analysis of the fluorescence measurement procedure are presented in S1P Data.

Optogenetic control

In optogenetic experiments that required full activation/deactivation (Fig 4A and 4B), either the green (525 nm) or red (625nm) LED was enabled individually at power setting 0.1. In experiments in which the precise induction level was controlled (Fig 4C and 4D), the procedure suggested by Olson and colleagues [5] was followed; the red LED was illuminated at a fixed power setting of 0.1, and the green LED's intensity was varied.

Chemically induced reporters

In fluorescence experiments that used chemically inducible fluorescent reporters, L-Rhamnose concentrations of 0.1 mg mL⁻¹ and aTc concentrations of 300 nM were used. At the time of induction, each inducer was added manually to both the reaction chamber and the fresh media supply (to provide a sharp step increase) without stopping the experiment.