Overview of Millifluidics
Last updated
Last updated
See the guides for all practical construction and operation.
A key development in microfluidics was the design and fabrication of devices containing integrated (pneumatic) valves that could allow for complex fluidic manipulations with minimal number of control elements27–29. Here, we describe (1) why adapting this technology for the macro scale is valuable for automated cell culture, (2) challenges faced when scaling to larger flow rates, and (3) a new framework for fabrication and bonding of millifluidic devices featuring integrated pneumatic valves. These devices offer a scalable solution to challenges faced by traditional fluidics.
The ability to program complex fluidic tasks could enable entirely new manipulations in automated cell culture applications (see Fig. 5, Supplementary Notes 15 - 17). For example, when growing undomesticated microbes, biofilm may form in the efflux fluidic lines and vials in as little as in 12 hours (e.g. see Fig. 5). The ability to programmatically bypass the vial in order to clean the fluidic lines with a bleach solution, and passage the culture from one vial to the next as a preventative measure, would be critical for long-term continuous growth of these microbes. However, complex fluidic tasks like vial-to-vial transfer, cleaning protocols, and mixed media inputs are extremely difficult with traditional fluidic systems used by current devices. In electronics, custom circuits can be readily created by breadboarding; however, this approach scales poorly to larger, more complex circuits because it relies on tedious manual assembly and leads to limited durability. Similarly, fluidic systems consisting of flexible tubing connecting separate control elements, like pumps and valves, can solve simple fluidic tasks. However, the number of necessary fluidic connections scales with the complexity of the desired task. For example, even with an optimal valving scheme, the ability to perform automated large volume transfers between any two culture vials in a 16-vial eVOLVER unit would require almost 300 fluidic connections and over 20 control elements. As in breadboarding, each connection would need to be routed individually by fluidic tubing and often by hand, a tedious task. Additionally, the tubing is usually fairly long, and each connection introduces dead volume, making the system less robust and impractical. Instead, by creating integrated (pneumatically valved) schematics, we sought to make a millifluidic equivalent of a printed circuit board; the complex fluidic connections are now integrated in a small device that is computer designed, manufacturable, and much easier to reproduce. With the flexibility of CAD, one would be able to customize a fluidic device to fit their particular experimental needs.
The cost of control elements and assembly time of bioreactor units can prove to be a significant burden as one scales fluidic inputs and outputs for high-throughput operations. As previously described, most designs rely on a pinch valve or a peristaltic pump to separately control each of the media sources and another to control waste. For example, a single vial turbidostat unit with 4 different media inputs would require 5 pumps. A hypothetical 48-vial unit with the same capabilities would therefore require 240 pumps, at a cost of $7,000 to $10,000. The key problem is that the number of control elements increases linearly with the number of vials. Our pneumatically-valved devices can leverage concepts developed in microfluidics in order to scale throughput by multiplexing and demultiplexing inputs and outputs27,30. In this scheme, the number of controllable vials scales exponentially to the number of control 23 elements, needing only 30 elements to route up to 16 different fluidic inputs to 48 vials. We project the costs of this new hypothetical 48 vial fluidic schematic to be $1,000 to $2,000, roughly a 90% decrease in cost in comparison to current systems.
Though inspired by microfluidic technologies, integrated millifluidic devices for continuous culture have drastically different design requirements. The following is a list of critical requirements for this technology in the context of continuous culture:
Devices need to be on the decimeter scale. Indicated in the nomenclature, microfluidic devices operate on the nano to micro liter per second flow rate. In contrast, continuous culture in eVOLVER requires flow rates of roughly ~1 mL per second, a 1000-fold increase. As such, in order to increase flow rate with an appropriate safety factor, the flow channels and device needs to be at least 10-fold larger than typical microfluidic devices.
New prototyping framework is necessary for larger devices. Traditional microfluidic prototyping techniques rely on standards from the microelectronics industry, namely patterning photoresist on silicon wafers29,51. Typically, 100 mm (4 inch) circular wafers are the largest size the machinery can pattern designs on, which is still too small for complex millifluidic devices. New fabrication framework is necessary to prototype devices for continuous culture.
Design cycle must be fast and repeatable. The success of microfluidics in the laboratory is, in large part, attributed to rapid and reliable design cycles enabling iteration and testing of prototypes. New design frameworks for millifluidic devices must be equally rapid and reliable.
Device must be transparent. Fluidic systems designed for long-term continuous culture of microbes will be prone to biofilm formation. The ability to monitor flow through the device is critical for debugging experimental issues.
Device needs to interface with pumps, filters, and tubing. Fluidics for laboratory continuous culture typically interface with syringe pumps, pressurized fluids, sterile filters and peristaltic pumps. A robust way to interface dozens of connections between the device and other fluidic elements (e.g. vials, filters) is critical and nontrivial.
Device must be resistant to 10% bleach and 70% ethanol. Sterilization of the device is necessary prior to any experimentation. Fluidic materials and fabrication must be resilient to these chemicals for weeks of continuous usage.
Our pneumatically-valved millifluidic devices enable customizable, programmable routing of liquid at volumetric flow rates of ~ 1 mL/sec. The valves pinch off fluid flow on the flow layer when 10 psi is applied to the control layer and enable flow when vacuum is applied (Supplementary Fig. 11). Improvement of device bonding will enable application of pressures above 10 psi, necessary with higher flow rates.
Fluidic tasks in eVOLVER are enabled by the sequential actuation of valves in a specific fluidic network encoded in the integrated millifluidic device. We demonstrate these fluidic manipulations in a series of experiments (see Fig. 5, Supplementary Notes 15-17). Each experiment utilizes different devices as required to meet the experimental needs. The architectures for most functions are modular (e.g. multiplexer, vial-to-vial router) and can be combined in order to achieve more complex functionalities (Supplementary Fig. 11). For example, simple single media input turbidostat function utilizes multiplexer and demultiplexer modules. The demultiplexer routes the media source to the correct vial and the multiplexer routes the efflux from vial to waste. The same multiplexer and demultiplexer modules are Nature Biotechnology: doi:10.1038/nbt.4151 25 reused in all Fig. 5 applications, but different multiplexed media selectors and vial-to-vial routers are included as needed in different experiments. Software routines to control the control elements (valves and pumps) are also divided into commonly repeated functions, usually in a similar manner to how fluidic modules were divided. The code for each fluidic function is preloaded into the Arduino to coordinate tasks between each fluidic module. For example, a simple dilution event would first actuate valves in the media multiplexer to select media, then actuate a syringe pump for metering the desired volume, followed by valves in the vial demultiplexer and multiplexer to route media into the vial and remove efflux. By loading the routine for abstract functions (e.g. dilute, clean, vial-to-vial transfer) into the Arduino, robust communication can be ensured, with rapid transition between sub-tasks and no skipped steps (which could cause incorrect media routing, mis-priming of the syringe pump, or leaks and other device failures).
