Mini-scale turbidostat for pathway optimization:
a combination of microfluidic device and electronic sensor


In order to optimize the bio-polymer production at genetic level, we need to choose a strain with the gene construct that has highest yields.

One way is to choose the gene construct that has the steepest slope of protein expression. However, while the understanding of growth kinetics from a genomic level has been the foundation for the prediction of protein expression rate, there is still great difficulty in connecting the gap between genetic analysis and accurate verification with controlled cell growth experiments, or cell cultures (Lee, Boccazzi, Sinskey, & Ram, 2011).

The ideal way to bridge such gap is to study the effects of various genetic parameters on protein production by maintaining a set of condition and screen various genetic variance, like optical density and fluorescence emittance simultaneously, which is not easy to be done in a 96-well plates. This difficulty is due to the fact that each measurement is taken using a separate piece of equipment. That separation increases the risks of unexpected noise in gathered data stemming from from the variance of the culture environment and condition.

Our goal is to integrate multiple measurements into a device that could facilitate the validation and prediction of gene expression rate in synthetic biology by: (1) maintaining ideal optical density and (2) take fluorescent measurement at different time points. Such integration is expected to reduce the impact of ambient factors on the data of interest and yield fast and reliable data due to the device’s small scale.

Design parameters

This design needs to perform the following tasks:
  • Suitable for short-term bacterial cell culturing
  • Measure Optical density at different time points
  • Maintain homologous density of the cell culture before taking such measurement
  • Pump in more media when the optical density is almost beyond that of log phase
  • Measure fluorescence simultaneously with optical density
  • Calculate the rate of protein expression with gathered optical density and fluorescence


Flow scheme of an ideal design

This design is a collaboration between our team and the Boston University’s iGEM hardware team. Our minimal goal for iGEM semester is to get the proof of concept of the mixing mechanism and optical density measurement.

Designs Specifications

OD sensor

The purpose of OD board is to gain data to measure cell density. This is a DIY version of a 600nm OD. The data will be used to maintain ideal cell density, whereby maintaining log phase cell growth. Ideal OD600 is in the 0.5-0.6 range.
Essentially, the OD measurement is the absorbance of the cell culture with an incoming light with wavelength. The formula is given as follow:


Summary for circuit design
  • 5V supply
  • 12 inch wire
  • Decoupled by a 0.01-μF to 0.1-μF capacitor at VDD leg

Brief description from the datasheet (Lee et al. 2011)

“Output is a square wave (50% duty cycle) with frequency directly proportional to light intensity (irradiance) on the photodiode. The digital output allows direct interface to a microcontroller or other logic circuitry. The device has been temperature compensated for the ultraviolet-to-visible light range of 320 nm to 700 nm and responds over the light range of 320 nm to 1050 nm. The TSL235R is characterized for operation over the temperature range of −25°C to 70°C”
block diagram

Image taken from the datasheet (Lee et al. 2011)

The left graph shows that with the 600nm wavelength, the normalized responsitivity is ~0.9 (quite a good number). The right graph shows that the normalized output frequency is appeared raw at 5V.

Circuit design

“Power-supply lines must be decoupled by a 0.01-μF to 0.1-μF capacitor with short leads placed close to the TSL235R (Figure 6). The output of the device is designed to drive a standard TTL or CMOS logic input over short distances. If lines greater than 12 inches are used on the output, a buffer or line driver is recommended.”

Sensor calibration

Standard References
  • OD measurements taken from a nanodrop equipment using a cuvette
  • The Characteristic Curve: consider offset, slope and linearity
Image taken from the datasheet (Lee et al. 2011)

The graph is pretty linear so ideally we only need to use a 2 point calibration to characterize the slope. E = freq/1000

CorrectedValue = (((RawValue – RawLow) * ReferenceRange) / RawRange) + ReferenceLow
Referenced from

Sensor readings

Essentially, the L2F sensor is the digital version of the photodiode. The data for analysis is the frequency of the square waveform generated by the sensor. To read this, we use Arduino to measure the total time taken to complete one high-low cycle, in other words, measuring the period of the square waves generated.

Image taken from the datasheet (Lee et al. 2011)

Based on the site, the method that we use Works on pulses from 10 microseconds to 3 minutes in length. This means that we could measure the range of frequency from 0.005 Hz to 10^7 Hz or 10^4 kHz. This range is well characterized in the datasheet.

Using the period, we could either convert it into frequency for calculate irradiance directly. For the latter method, we expect this relationship as between the radiance and the period:


Expected relationship between period and irradiance

Vapor polishing

The milling process leaves an opaque area at the milled parts. This opacity varies each milling process. Vapor polishing would make such area transparent, whereby increases the reproducibility of the design.
For the prototype, we heat up methylene chloride in a bunsen flask and place the milled area over the vapor. This was done inside a fume hood.


3D models

Initial design: 2 inlets (to the left of the images), 1 for cell culture and 1 for media input. 1 Outlet (to the right) for waste output. The 2 chambers serve as the mixing mechanism. The device pumps the culture from one chamber to the other, whereby creating a turbulence to mix the culture and the media.

Initial design

base with dimension v3

base with dimension v32

smooth flow layer v3

smooth flow layer v32

control layer v3

control layer v32

Mixing mechanism
The mixing mechanism worked as expected without breaking the PDMS membrane:
Mixing mechanism
Image taken by Boston U hardware team.


Prototype before vapor polishing

Vapor polishing


Areas with and without vapor polishing

IMG 7814

Vapor polishing using methyline chloride yielded expected transparent surface.

Setup for the OD sensor readers: the light source (a 600nm LED) is placed perpendicular to the surface of the chamber so that the incoming light is also perpendicular to the light sensor placed right beneath. The system will then be covered with black paper or black tape so that ambient noise from the environment would not affect the measurements.

IMG 7827

After testing the OD sensor with the prototype, there are several factors that affect the measurement:

  1. Distance from led to sensor.
  2. Position of led vs. sensor (more of less orthogonal): If we move the sensor just a little bit away from the sensor, to the left or right but still on the same original plane, the value changes.
  3. Volume filled in the chamber: the data from the sensor reading changes based on whether the chamber is fully filled or not.

OD sensor experiment results

OD calibration graph

Point to point curve-fitting graph showing OD - measured frequency from the sensor
This is not what we expected.
We have the following calculation:
E = freqency/1000
OD = log (Eblank/Eculture) = log (Eblank) - log (Eculture) = const - log (Eculture) = const - log (freq/1000) = const - log (freq) + 3
Whose graph should have a shape similar to this:
OD expected freq graph

Expected curve of frequency measured from the sensor

We found out that during measurement, the cell liquid culture bound to the walls of the chamber and created an air hole in the chamber. Because we need to cover the measuring area entirely with black tape, we could not prevent this. Air bubbles could be the reason why the OD measurements are not as expected.
Next steps are to (1) use the same OD reader system but measure optical density in the cuvette and (2) calibrate the readings of the sensor using a well-defined equipment to measure light intensity. This makes sure that there are no other technical problems that cause the unexpected behaviors of the sensors.

Future development

We hope that future iGEM teams that are interested in hardware design or any applications of this device could use our designs as a foundation to build on.


Lee, K. S., Boccazzi, P., Sinskey, A. J., & Ram, R. J. (2011). Microfluidic chemostat and turbidostat with flow rate, oxygen, and temperature control for dynamic continuous culture. Lab on a Chip, 11(10), 1730–1739.