Team:Paris Bettencourt/Hardware



The minicell-machine

The design

As a final part of our project, we have designed a hardware device, the so-called Minicell-machine. The aim is to develop an autonomous machine that is cheap, biosafe and easy to replicate. It consists of two separate incubation tanks, fixed on a bar so it can be placed anywhere. The original design is using gravity as a mean to transfer liquids from one tank to the other, a simple method that lowers the costs of the machine. However, for this proof of concept version of the bioreactor, we chose to use peristaltic pumps to facilitate the construction and to have higher control over the liquid flow. The third tank (following the filter, where mothercells are retained) is reserved for cleaning and is also not present in our proof-of-concept version.
Both incubation tanks have a tight temperature control thanks to heating jackets and an Arduino microcontroller system. Furthermore, the OD and pH are constantly monitored to ensure proper growth conditions. To start using the machine, bacteria are added to the upper incubation tank that holds 1 liter of self-made M9 culturing media. The beginning OD should be around 0.05. In this tank, the bacteria are cultured at 30°C until they reach an OD of 0.2.

Consequently, IPTG is added with a liquid dispenser in order to start minicell production of the mother cells. After approximately 2 hr, a valve at the bottom part of the first incubation tank opens.

The cell culture with the minicells will flow through a filter. This allows for separation of minicells from mother cells by size. Ideally, this would end up with only the minicells are all transferred to the second tank which is supposed to be free from modified organisms.

To ensure the complete removal of mother cells, the temperature in the lower incubation tank is raised to 42°C. This activates the dormant phage encoded in the genome of the remaining mother cells and causes their lysis.

In the following step, arabinose is added to the media. This will induce the pigment production in the minicells. After approximately 4 hr the minicells containing the pigments can be harvested. While the cells are in the lower incubation tank, the upper part of the machine can be emptied and cleaned already to avoid the accumulation of debris and biofilm formation.

All the steps are automated with 2-3 Arduino microcontrollers. To ensure proper connections between the tubes and the tanks, the specific headpieces were 3D-printed. Stirring and aeration of the media are achieved with the use of an aquarium-air pump. The overall cost of the machine is less than 200€ and all pieces are commercially available or can be re-built with our templates. As this allows for easy replication, the Minicell-machine can serve as the basis for the construction of local, small bioreactors.

To visualize the hardware better, we created some sketches and renders of our minicell-machine design:

The filtering

To ensure biosafety, we want to reduce the amount of mothercells in the last tank as much as possible. The lambda phage in the genome is quite effective already. However, the lysis characterization (see here) showed that it does not lyse all the cells. In fact, some cells are remaining in the sample that can potentially regrow.
Therefore, we also tried to remove the mother cells with a physical method. Affinity chromatography is not an option we can take, as the minicells have very similar membrane compositions as the mother cells. The main difference between them is the size. This is why we went for the filtering option. We wanted to find a new way to filter mothercells from minicells, that is cheap and with easily accessible materials. In literature, the effective use of sand for filtration of bacteria has been stated (1). Hence, we tried out the filtering with aquarium sand, glass beads that have the size of the sand stated in the literature, and also a column stuffed with cotton. A culture of MG1655 with an OD of 0.1 was passed through the different filter techniques.

Figure 1. Setup of filtering experiment with sand, glass and cotton (left). Optical density measurements after the filtering experiment (right). Data was normalized to the unfiltered sample.

The OD after filtering with the 0.8µM filter was removing all the bacteria. The OD of the cell culture was 0. On the contrary, the OD of filtering with sand and glassbeads was even higher than the initial OD of the cell culture. This is probably due to microparticles of sand and glass that were flushed through as well and that increased the turbidity of the sample. The column stuffed with normal cotton was delivering surprisingly good results. The OD of the cell culture was decreased to 12.4% of the initial OD.

To confirm these positive results of the cotton column, we did a second round of measurements where we also performed CFU counts after filtering. We were wondering if FFP2 masks could also be used as a filter, so we included them in the following filtering experiment as well. However, the material of the masks is hydrophobic, which made the filtering quite difficult. We had to separate the 3 layers of the masks and take out the middle, most hydrophobic one. The other two layers had to be soaked in the cell culture for a few hours, in order to be able to pass liquid through. Filtering a MG1655 culture with an OD of 0.1 through one of the layers and both of the layers were tested. Then, the OD of the bacterial culture was measured before and after filtering through the different approaches. Additionally, diluted samples were plated on LB plates for acquiring CFU counts. The data were normalized to the unfiltered cell culture and plotted.

Figure 2. Optical density measurements and CFU counts after the second filtering experiment. Data was normalized to the unfiltered sample.

The filtering through one layer of the FFP2 mask decreased the OD to 71.4% of the initial OD. The addition of a second layer decreased the OD a little further to 66.3%. On the other side, the CFU counts were indicating higher cell numbers than in the unfiltered sample. We have not used a sterile mask, bacteria being already on the surface of the masks were most likely responsible for those increased CFU counts. The masks should have been sterilized before and LB should have been passed through them and analyzed to account for contaminations.

Another important property the filter must-have is the permeability for minicells. For this, we tested only the layer of the mask and the column stuffed with cotton with a purified minicell sample. The initial OD was 2.25, the data was normalized again to the unfiltered sample and plotted.

Figure 3. Optical density measurements after filtering a purified Minicell sample. Data was normalized to the unfiltered sample.

Filtering through the mask only decreased the OD to 98.7% of the initial one. It lets us assume that almost all of the minicells were passing through. The sample which went through the cotton column had a decreased OD of 40.7%. It seemed like approximately half of the minicell population was retained in the filter.

To confirm the presence of minicells in the filtered samples, microscopy images were taken.


Filtering with cotton column

Filtering with 0.8µM filter

Figure 4. Microscopy data of unfiltered, cotton- and 0.8µM filtered samples in order to analyse the permeability for minicells.

The microscopy images firstly show that the minicell sample was not completely purified. There was still some bacteria present. After filtering with the cotton column the amount of normal-sized E. coli is lower and there are also minicells visible in the image. Surprisingly, also in the sample that was passed through the 0.8µM filter, a lot of minicells are visible, even if the OD measured was 0.

The microscopy was delivering important data, as it confirmed the presence of minicells in the samples after filtering. It has also shown that the decreased OD in the minicell-filtering experiment of the cotton column might not be due to a loss of minicells, but due to partial removal of the mothercell population. The experiment also demonstrated that the 0.8µM filter has optimal filtering properties with complete removal of mothercells and permeability for minicells. But also the self-made cotton stuffed column was providing good results. It is able to filter about 90% of bacteria while being permeable for at least 50% of minicells. For our proof of concept in our hardware, we could therefore use either of them. As the cotton column is cheaper and easier to implement, as no pressure is required to pass the liquid through, we decided to use this method in our minicell-machine design.

Our choice of materials

The actual construction of the Minicell-Machine confronted us with some challenges. Firstly, the choice of the material for the tank. It has to be tightly closed to contain the bacteria, but still accessible. Our first approach was to use big glass flasks and only have one opening at the top. To be able to connect the tubes and all the sensors, we designed a 3D-printed lid. It closes the flask and allows us at the same time to take measurements. As the flask is closed at the bottom, we have to use a peristaltic pump for this prototype to move the liquid into the second tank.

Second challenge was to find a proper heating system. After consideration of different heating methods, like heating plates and heating devices that are inserted directly into the tank, we chose to go with an external heating source. Heating mats that are used for reptiles have the perfect size to go once around our flasks. This allows for a more distributed heating than having only a heating plate on the bottom. As they are not in touch with the liquid, also the cleaning is not complicated and we don't have to worry about biofilm formation on it. Additionally, this was an affordable solution and potentially less dangerous than other heating sources.

Another important part in the hardware is a proper stirring system and sufficient aeration for optimal growth of the bacterial culture. To combine those two elements, we decided to use an aquarium air pump. The bubbles blown into the liquid will stir the culture, while it will also provide enough oxygen for the bacteria to grow. Additionally, the added air will facilitate the oxidation of indoxyl to indigo - an important step for the pigment production.

Last but not least, the whole system is automatized to limit human interaction with the bacteria and to increase reproducibility. We used Arduino microcontrollers not only to automatize the valve openings, heating and bubbling, but also to gather information about various sensors. The temperature is constantly monitored with a thermometer and the heating mat regulated accordingly to keep the temperature of the cell culture at 30°C. Furthermore, the OD is monitored every 10min in the first tank with a turbidity sensor. Whenever the OD measurement is taken, the bubbling system is turned off for a minute to not disturb the turbidity measurement. If the OD reaches 0.5, the valve is opened / the peristaltic pump is activated in order to move the liquid through the filter in the second tank. There, a very similar arduino system is keeping the temperature at 42°C. Bubbling with a second air pump ensures stirring.

List of materials

  • Two 1.5L glass tanks

    The tanks can be purchased on amazon or Ikea.

  • Two 3D-printed lids

    For each of the tanks of the hardware. Intructions to print the lids can be found here.

  • A laser cut panel

    A panel holding all the electronic components (Arduino, peristaltic pumps, relays, power and sensor connectors) was designed. More detailed instructions can be found here.

  • Two heating mats

    Bought in a pet shop. The heating mats were wrapped around both tanks and hooked up to individual channels of the relay. This setup proved to reach the required temperature (up to 42°C from 21°C ambient room temperature).

  • Two DS18B20 temperature sensors (one in each tank)

    They were chosen for their temperature range (From -55°C to 125°C ), their price (2€ in a package with stainless steel probe and water proof cabling pre-setup), their availability (available at low cost at most online suppliers and local electronic/hobbyist shop), and their convenience to use (these sensors work on with the One-Wire protocol, and, therefore require less installation).
    The sensor and the cable were fit through a 4mm Internal diameter stainless tube to insure stability of the system against shaking, possible transport and to increase overall robustness of the design.

  • An OD sensor

    To measure the optical density (OD) of the growing bacterial culture (triggering the transfer of fully grown bacterias to the pigment expression tank), the DFRobot Gravity Analog Turbidity Sensor was used. It has been tested and shows sufficient sensitivity to the changes of turbidity specific to the growth of our bacterial cultures.

  • Peristaltic pump

    DIY peristaltic pumps were purchased from and proved sufficient for liquid handling and stirring through bubbling.

  • Arduino Uno

    Arduino is an open-source electronics platform based on easy-to-use hardware and software that is very popular among DIY communities, as a tool to teach and learn STEAMS as well as in research labs across the globe for its low cost and high flexibility/modularity characteristics. We based the control mechanism of our design on an Arduino Uno

  • Graphical User interface

    We developed a graphical user interface (GUI) that displays the current state of the sensor's measurement, of the relays controlling the pumps and the recorded data. It offers the possibility to overwrite the automated program and to manually control the liquid handling pump (for example to extract liquid from the growth tank in order to perform some measurement). The software also comes with a Python Base Data Logger recording the state of the system at a customizable rate on a CSV file.
    The GUI also offers the possibility to showcase the recorded data in real-time and to load files and instantaneously graph data from previous experiments. The implementation was based on TouchDesigner, a node-based programming environment primarily designed for the interaction design community allowing for extremely fast prototyping through a high level of modularisation. If a TouchDesigner license has to be purchased for commercial purpose, the GUI has been designed in order to run on the free version of touchdesigner (available both for Windows and MacOS)

  • Relays

    8 channel optocoupler driven 5v Relay array to control the heating mats and the 3 peristaltic pumps. All together, these components takes 5 of the 8 channels leaving some free to hook up other devices (such as other pumps or valves for chemical inputs)

  • OLED screen

    An SSD 1306 OLED Screen has been mounted on the panel to display the sensor's data in real time. This display has been chosen for its availability (available at low cost in most online suppliers and local electronic/hobbyist shops), its convenience (simple integration with the Adafruit_SSD1306 Library and simple 4 pin wiring) and its price (less than 5€ on amazon).

  • Instructions to build the minicell-machine

    1. 3D-Print the lid
    For the growth thank
    The optical density (OD) sensor is held on a floating 3D printed tube allowing the sensor to always sit right at the surface of the culture 4 holes were placed on the lid to fit cable glands holding stainless steel tubing. A 4mm internal diameter (ID) stainless steel tube is maintaining the temperature sensor in place (hence always measuring temperature from the same location). The 3, 8mm ID stainless steel tubes are carrying stirring through bubbling and liquid handling from one tank to the other. The file for 3D-printed lid can be found here.

    For the enzyme production tank
    The design of the lid for the enzyme production tank is similar to the one of the growth tank except that optical density is not measured here, hence, the mechanism holding the OD sensor has been removed. Here, stainless steel tubes were used as they were readily available but Polyvinyl chloride (PVC) tubes might be considered to lower the costs The file for 3D-printed lid can be found here.

    2. Laser cut the Panel
    Laser-cut out of 3mm thick transparent Acrylic. The panel should be mounted on 20x20mm V slot aluminum extrusion. The DXF file for laser cutting the panel can be found here or here.

    The design of the lid for the enzyme production tank is similar to the one of the growth tank except that optical density is not measured here, hence, the mechanism holding the OD sensor has been removed. Here, stainless steel tubes were used as they were readily available but Polyvinyl chloride (PVC) tubes might be considered to lower the costs The file for 3D-printed lid can be found here.

    3. Assemble the parts
    Wrap the heating mats around the tanks, put the sensors and the airpump hose through the lids and fix the lids on the tanks. Make sure that it is tightly closed and connect the tubes via the peristaltic pump with the two tanks.

    4. Mount the parts to the panel
    An SSD 1306 OLED Screen has to be mounted on the panel to display the sensor's data in real-time. The breadboard, relays, pump and other parts have to be fixed on the panel as well before wiring the parts. Make sure that the parts are not in direct contact with the metal. A little piece of cardboard can for example be but in between.

    5. Connect the electronics
    Connect the relays, different sensors, Arduino Uno and the GUI. Also do not forget to connect the system to an external energy source.

    6. Test the minicell-machine

    Future Improvements

    1. Heating system
    Two key learning points were extracted from experimenting with this setup:

  • Even if the required temperature is ultimately achieved, this process remains slow (up to 3 hours)

  • As we used relays to control the heating pads, the only available control mechanisms were through thresholding at specific temperatures. This process is hard to tune and does not encompass the inertia of the system (even if turned off, the pad remain hot for some times) and only allow for binary values (on or off)

  • To approach these problems we propose to replace the heating pads with Peletier Elements (allowing to heat the tank faster). The trade-off for such change is that a special frame holding several Peletier elements as well as heatsink systems will need to be designed, hence increasing cost and rendering the entire setup heavier and harder to build. Such change should, therefore, only be considered for use cases where heating faster is required.

    To improve control of the temperature, a Mosfet-based system should be considered to allow for non-binary switching of the heating mat (Using a Mosfet-based system would allow to turn on the heating pad at any arbitrarily defined temperature). With an analog temperature control system, a proper proportional–integral–derivative (PID) controller could be implemented in the control firmware

    2. Peristaltic pumps
    In the actual setup, the liquid handling and stirring pumps are controlled through individual relays. They are, therefore, unidirectional. Being able to control the direction of the pump would allow for more precise liquid handling and better stirring. To achieve that, an H-Bridge component can be added to the system to change the direction of the pumps.
    Additionally, a future improvement of the hardware would be to apply our initial idea of using gravity with valves instead of pumps.

    3. Arduino
    In its current state, wiring of all the components (The temperature and Optical Density sensors, the relays and the OLED screen displaying real-time data) is done through Jumper cables and breadboards. The design of a custom Printed Circuit board (PCB) in the form of an Arduino shield would greatly help reproduce the setup as much less wiring would be required, thus, preventing human mistakes, saving time and cost in production as well as providing a cleaner workbench.

    4. Relays
    As they get triggered these mechanical relays produce a clicking sound. It is subtle but can be heard and problematic to implement in specific places (such as retail or other sound-sensitive environments). As an alternative, Solid State relays might be used.

    5. OLED screen
    Despite its many advantages, this display remains small and with a limited resolution (124x64 px). Reading information might be challenging and another display can be considered.

    6. Liquid dispensers
    One of the next steps would be to add two liquid dispensers for IPTG, arabinose and tryptophan respectively tubed with the first and the second tank


    Design and prototyping a bioreactor is a valuable resource for other makers or labs that are interested in improving the prototype or just doing the montage on their own. Since our hardware was designed to be a pretty general device for bioproduction using minicells, we wanted to make it open access. Here is the link of our github repository, where you can find all the details about the machine.


    (1) Cawst. (2012). Biosand Filter Construction Manual. Centre for Affordable Water and Sanitation Technology.

    © iGEM Paris Bettencourt 2021

    Twitter - iGEM Paris Bettencourt Instagram - iGEM Paris Bettencourt