![[PDF]](/pb-assets/Icons/adobe-pdf-icon-1498649896243.png){kind=icon}
Abstract
A fully automated “factory” was developed that uses tobacco plants to produce large quantities of vaccines and other therapeutic biologics within weeks. This first-of-a-kind factory takes advantage of a plant viral vector technology to produce specific proteins within the leaves of rapidly growing plant biomass. The factory’s custom-designed robotic machines plant seeds, nurture the growing plants, introduce a viral vector that directs the plant to produce a target protein, and harvest the biomass once the target protein has accumulated in the plants—all in compliance with Food and Drug Administration (FDA) guidelines (e.g., current Good Manufacturing Practices). The factory was designed to be time, cost, and space efficient. The plants are grown in custom multiplant trays. Robots ride up and down a track, servicing the plants and delivering the trays from the lighted, irrigated growth modules to each processing station as needed. Using preprogrammed robots and processing equipment eliminates the need for human contact, preventing potential contamination of the process and economizing the operation. To quickly produce large quantities of protein-based medicines, we transformed a laboratory-based biological process and scaled it into an industrial process. This enables quick, safe, and cost-effective vaccine production that would be required in case of a pandemic.
Introduction
The vaccine shortages during the 2010 H1N1 swine flu pandemic highlighted the fragility of the vaccine supply chain. Although the reasons for vaccine shortages are multifaceted, including companies leaving the vaccine market, manufacturing or production problems, and insufficient stockpiles,1 the shortages highlight the need for new vaccine production technologies that enable agile, robust, and rapid surge response. The majority of influenza vaccines are currently produced in eggs.2 Egg-based vaccine production is a decades-old well-established technology with existing production facilities.3 However, egg-based vaccine production requires a sufficient supply of eggs, uses live (albeit attenuated) virus, and has a long lead time (6–9 months). Thus, the method has inherent limitations in that it is not scalable, requires long-term planning, and has long annual production times.4,5 The biochemistry revolution has provided alternative production technologies such as recombinant protein production via bacterial and mammalian cell culture, but these still require sophisticated culturing or fermentation equipment and expensive downstream processing.6
In contrast, recombinant protein production in plants is becoming recognized as a viable alternative platform technology.7,8 Plants offer several advantages over other systems in that they can perform eukaryotic posttranslational modifications (glycosylation or disulfide bridging),9,10 have no risks of contamination by animal pathogens, can use light for their main energy source, and are robust and inert. Scaled-up and automated plant-based vaccine production has the potential to compete with other more established manufacturing technologies because plant growth requirements are relatively simple and inexpensive and proteins can be rapidly expressed and produced in them.11 Moreover, the method’s surge capacity offers rapid response to pandemic disease threats or bioterrorism events.8,12
Tobacco (Nicotiana tabacum) is a major platform for molecular farming. Despite the traditionally negative reputation due to its association with smoking and cancer, tobacco is an excellent vehicle for recombinant protein production due to its ease of genetic modification (including transient protein expression modalities), high biomass yield, high soluble protein levels, and ability to produce a wide range of therapeutic proteins. Although different methodologies exist, the process that we chose to scale up uses a proprietary transient protein expression to rapidly produce significant quantities of protein. In brief, the method begins with cloning the desired gene into a plant viral gene expression vector and introducing the vector into Agrobacterium tumefaciens, a common soil microbe. Meanwhile, tobacco plants are cultivated in the automated factory (e.g., seeding, watering, exposure to light). Then, the agrobacteria carrying high copy numbers of vectors are introduced into the leaves of the plants by an automated vacuum infiltration technology. This process results in the simultaneous introduction of the expression vector into all leaf cells of the plant, increasing the efficiency and speed of protein production. Special features of the proprietary vectors ensure its spread to every cell in the stems and leaves where the desired protein is expressed at extremely high levels (gram quantities of target protein per kilogram of fresh plant tissue) over the next 4 to 7 days. (Naive plants are grown and always kept at the ready to minimize vaccine production time.) The plant biomass is harvested, ground, and clarified to produce an extract containing the protein of interest. Proteins are further purified as required using well-established separation and chromatography steps.
In this article, we describe the custom machinery developed to scale up and automate the process from seeding to harvesting. The result is a fully automated current Good Manufacturing Practices (cGMP) factory for plant-based recombinant protein production. The factory requires less capital infrastructure and has lower operating costs than other vaccine production techniques because it does not require expensive bioreactors and aseptic liquid-handling technology to generate the production biomass. The factory is designed to be modular, so that multiple products can be produced concurrently. As production cycles are reduced to days instead of months, the system has the potential to greatly affect the future of vaccine (and other biological therapeutics) manufacturing.
Principle of Operation and Factory Design
Overview of Process
Traditional plant-based vaccine production begins with seeding of tobacco plants in a greenhouse, growing the plants under suitable lighting and with appropriate nourishment, infiltrating the plants with a vector that instructs the plants to synthesize a desired protein, further growing the plants as they synthesize the desired protein, harvesting the plants, and finally, extracting and purifying the proteins. With the exception of the Agrobacterium infiltration step and the protein extraction/purification steps, it is essentially an agricultural process, no different from growing tomatoes, for example. To make the process cost-effective and improve yield, the approach taken was to convert what is essentially an agricultural process into an industrial-like process. Doing so enabled the use of automation techniques and strategies commonly employed in industry for the purposes of cost reduction, floor space reduction, and yield improvement.
Since plants have irregular geometries and are not rigid, they are very difficult to handle using industrial automation. To facilitate a completely automated process, a means of easily manipulating the plants had to be developed. The issue was addressed by hydroponically growing an array of plants in a rigid tray. This solved the handling process since the tray is very well defined and can be manipulated by robots and other automated machinery as is common in industrial processes. Figure 1 shows a custom-designed robot carrying a tray of plants.
Once the decision was made to grow the plants hydroponically in trays, a fully automated production system was designed and fabricated. The system consists of five main modules but excludes the protein extraction and purification process, which is accommodated by commercially available equipment. These automated modules include the following:
Seeding
Growth
Infiltration
Harvesting
Robotic transport system
Two custom-designed robots, as shown in Figure 1, transport the trays to each module or station, where the appropriate operation is automatically performed. The process begins with the robots carrying empty trays to the seeding module where the entire array is populated with seeds. The trays are then transported to the growth module where the seeds are automatically watered and nourished for a period of several weeks. The trays are then transported to the infiltration module where an appropriate viral vector is introduced into the plants so they can synthesize the proteins of interest. The trays are then transported to a separate and isolated growth module, where the plants are allowed to grow for an additional period as the proteins are being synthesized. Finally, the trays are transported to the harvesting module where the plants are chopped into small plant matter from which the proteins can be extracted and purified using commercially available equipment.
Hydroponic Growth Trays
To maximize production rates, a feature-rich tray that can grow up to 200 tobacco plants was designed and fabricated. The trays measure 660.2 by 1219.2 mm in size and comprise 20% talc-reinforced black homopolymer polypropylene for rigidity and weight considerations. They were molded using a structural foam molding process with gas counter pressure facilitating their low-cost manufacture. A hydroponic growth medium is loaded into the trays manually, and then the trays are loaded onto a loading rack from which the robot takes trays to the seeder. The plants are seeded and grow in a precise and arrayed pattern that provides accurate location of the base of each plant for ease of locating and manipulating. The tray is covered with a stainless steel plate with holes coinciding with the seeding array. The purpose of the stainless steel plate is to support the plant during upside-down manipulation that is required by some of the downstream operations (see “Infiltration Module” section). Figure 2 shows an assembled tray prior to seeding.
Structural and material considerations played a very important role in the design of the growth trays. These trays have conflicting structural requirements: high rigidity for ease of manipulation and minimal weight for handling purposes. Thus, a fairly complex ribbing system was designed and molded into the trays to maintain rigidity while reducing weight. In addition, homopolymer polypropylene was chosen as it can withstand the temperatures required for sterilization with a sterilizing tunnel washer. In addition to structural features, the growth trays also possess a number of features that ensures homogeneous water distribution.
Seeding Module
Tobacco seeds are quite small, on the order of 0.5 mm. They resemble poppy seeds but are smaller and more irregular (see Fig. 3). To maximize production, a functional requirement of the seeding module was that only one seed be planted in each array location in the growth tray. If an array location is accidentally left vacant, then precious real estate is wasted, resulting in lower total protein production. If, on the other hand, multiple seeds are planted in an array location, then multiple plants will compete for water and nourishment, which also results in lower protein production. However, the size and shape of the seeds presented an engineering challenge to separate and manipulate individual seeds.
To address the challenges mentioned above, a vibratory seed pan, in conjunction with vacuum grippers, was employed to isolate and deliver the individual seeds (shown in Fig. 4). A simple pick-and-place robotic mechanism picks up a row of seeds and places them in the corresponding array locations in the trays. This process begins with the robotic transport system loading empty trays into the seeding module. The seeding module then advances the trays one row at a time, prewatering a row of array locations and placing a row of seeds in the prewatered locations. The trays are indexed from row to row until the trays are entirely populated with seeds. The seeds are then watered again. Each tray with 200 array locations can be completely seeded in approximately 3 min. Figure 5 shows the entire seeding module, which has dimensions of 4.7 m long by 0.9 m wide by 2.1 m high. Once fully populated, the trays are taken by the robotic transport system to the growth module, where the plants are grown until they are ready to be infiltrated with Agrobacterium.
Growth Module
In addition to cost-efficient production, an important goal of this automated production system has been to minimize the floor space requirement. It is envisioned that such automated factories can be located close to where the need arises. This may be in highly populated areas with a high cost of factory space. Thus, growing the plants on one level, whether it be in the field or in the factory, was immediately ruled out. Instead, a modular, multilevel growth rack was designed and fabricated, in which trays with plants can be placed to grow not only one above the other but also several rows deep at each level (see Fig. 6). The resulting growth racks measured 3 × 4.3 × 3.2 m and can accommodate up to 48 trays each. The number of growth racks used at any one time depends on the total required protein production rates.
Although this approach minimizes the required floor space, it introduces a number of challenges that had to be resolved. First, ceiling light was no longer an option, since the trays at the higher levels block the light from reaching the lower levels. Thus, a distributed lighting system was developed that ensures that all the plants grow under the same lighting conditions. Second, each tray must receive equivalent and controllable amounts of water. To address this challenge, a sophisticated proprietary water distribution system was developed that ensures consistency between all the plants by delivering the same amount of water and nourishment to each tray. Finally, since there are several rows of trays on each level, automated placement and retrieval presented a challenge as well. As will be discussed in the “Harvesting Module” section, custom robots had to be developed that provide sufficient reach yet can also travel in tight spaces.
Infiltration Module
For the transient protein expression of foreign genes, the plants must be infiltrated with a vector that instructs them to synthesize the protein of interest. The proprietary vector system used in this factory is a hybrid Agrobacterium/viral gene system that does not permanently alter the plant’s genome and does not result in the production of plant virus particles. The agrobacteria (carrying the vector system) are introduced into all aerial parts of the plant by an automated vacuum infiltration technology. This process results in the simultaneous introduction of the expression vector into all leaf cells of the plant, increasing the efficiency and speed of protein production. In the approach taken, the vacuum infiltration is accomplished while the plants are in an upside-down position, as shown in Figure 7. First, the plants are immersed into a vacuum chamber containing the Agrobacterium solution. A vacuum is then applied to draw the air out of the intracellular spaces in the plant cells. When the vacuum is released, the agrobacteria are sucked into the cells. Since for consistency, it is important to infiltrate all the plants in the same point in their growth cycle, the infiltration module performs four process steps in parallel to reduce the cycle time to minimize the time variability between trays of plants: (1) flipping trays upside down, (2) infiltration, (3) rinsing residual Agrobacterium solution off the plants, and (4) flipping the trays upright. The infiltration module can process trays at a rate of two trays per minute, and the full unit is shown in Figure 8.
Since the infiltration process involves the introduction of biological vectors, the infiltration area has to be isolated from the rest of the process, which was accomplished by housing the infiltration unit in a separate room. The trays with plants are transported into the infiltration room by conveyors that enter and leave the room through lift gates. Once the plants are infiltrated with agrobacteria, they must be allowed to grow for an additional period as they synthesize the desired protein. Although this growth phase is facilitated by the same type of growth modules used for preinfiltration growth (described above), they must be distinct physical modules, to ensure that there is no cross-contamination with the plants yet to be infiltrated.
Harvesting Module
Once the desired proteins have been given time to accumulate in the plant tissue, the plants must be harvested so that the protein extraction and purification process can commence. The harvesting process involves chopping the plant into a homogenate that can then be ground and clarified to produce an extract containing the protein of interest. The process begins with the robotic transport system retrieving trays with plants from the postinfiltration growth modules and loading them into the harvesting module. For efficiency and convenience reasons, the plants are harvested upside down, so that the cut plant matter can naturally fall into a collection bin. Thus, the very first step performed by the harvesting module is the inversion of the plant trays so they can be fed past the chopping blades in an upside-down orientation. Figure 9A shows the harvester in the process of inverting a plant tray.
After the plant trays are inverted, they are advanced past the chopping blades by internal conveying belts to sever the green biomass from the roots and the hydroponic medium. Figure 9B shows a plant tray advancing past the chopping blades. Before the trays are unloaded, they are disassembled in the harvesting module, and the hydroponic media and remaining plant matter are discarded into a waste bin for proper disposal. To reduce the burden on the waste processing system, the waste is squeezed through an auger to reduce the volume of the solid waste and to separate the liquid waste. The trays are then automatically unloaded into a cart so they can proceed to cleaning and sterilization for reuse. A solid model rendering of the harvesting module is shown in Figure 10. The harvesting module is a large piece of machinery with dimensions of 7.9 m long, 2.1 m wide, and 3.6 m high when the trays are tipped on end (as in Fig. 9A). The module processes trays with a cycle time of 30 s per tray.
Robotic Transport System
Two robots riding on tracks service the entire factory, transporting plant trays from module to module. Figure 1 shows a photograph of the custom-designed robot holding a tray of plants. These robots interact with the modules on either side of the tracks with very limited clearance (0.82 m) for the track and robots to pass between the processing modules. However, the robots must have a long reach, to either side (0.85 m), to load and unload the growth racks. These conflicting functional requirements required the custom design of the robots since no such robots are commercially available. Figure 11 shows the robot with its arm fully retracted (Fig. 11B) and with the arm fully extended to either side (Fig. 11A,C).
The requirement of an extended reach to either side, coupled with the constraint of minimizing the retracted dimension to pass through constricted spaces, presented a difficult mechanical actuation problem. To address this challenge, we designed a clever belt drive system (Fig. 12). The mechanism is a bidirectional telescoping dual-stage shuttle fork that is capable of extending 120% of its footprint to either direction. A central motor drives the first stage using two symmetrically arranged timing belts. The second stage is then slaved off of the first stage with another set of symmetrical timing belts. The result is an extremely compact and elegant robot end-effecter that handles the plant trays between modules. The resulting robots are driven on the track using a standard rack-and-pinion system. The position and motion profile can be accurately controlled to easily interact with the processing modules on either side of the track.
Factory Control
Five independent PLCs (programmable logic controllers) synchronize the control of the trays throughout the factory—namely, separate controllers for each of the two robots, the seeding module, the infiltration module, and the harvesting module. The two PLCs that control the robots also control the batch operations throughout the factory. The major batch operations are (1) a seeding batch that seeds 48 trays, as well as delivers and loads them into the growth module; (2) an infiltration batch that unloads 48 trays (one at time) from the preinfiltration growth modules through the infiltration module processes and loads them into the postinfiltration growth module; and (3) a harvesting batch that unloads trays from the postinfiltration growth module through the harvester module processes and then loads the empty trays into carts for downstream cleaning.
The control architecture is extensive in that it requires the coordination of hundreds of inputs and outputs. The PLCs are involved in synchronizing the action of numerous subcomponents—for example, servo motors, pneumatic cylinders, lights, water pumps, doors, conveyor belts, and UV sterilizers. Inputs to the PLCs include a sensor for tray position, temperature, water pressure, and vacuum level. Of note, the communication bus allows each of the five PLCs to read and write on the memory of the other PLCs, facilitating overall synchronization. The main interface to the computer system is through four HMI (touchscreens), which each can read and write memory space of the five PLCs.
cGMP-Driven Design
To be approved by the Food and Drug Administration (FDA) for vaccine production, all of the modules described above had to be designed such that they meet cGMP guidelines. This entailed a number of design considerations relating to the choice of materials, ease of cleaning, and avoidance of contamination sources. Generally, parts that have the potential of coming in contact with the plants at any time during the growth phase were made from passivated stainless steel. All materials had to be able to withstand harsh cleaning agents, such as vaporized hydrogen peroxide, bleach products, and solvents, so that the equipment could be sterilized. Lubricants were largely avoided, but where necessary, FDA-approved food grade products were applied. Exposed threads on screws, as well as other small inaccessible areas where bacteria can grow easily, were avoided as much as possible. In the infiltration station, all processing vessels were designed to be cleaned in place using a commercial clean-in-place (CIP) skid and custom-designed spray balls for the tanks. The chopping blades in the harvesting module were designed for easy disassembly and automated cleaning in a commercial dishwasher after a completed run.
Results and Discussion
The plant-based pharmaceuticals factory described in this article is the first of its kind, fully automated factory that meets cGMP guidelines. Its development was an integrated effort between engineers working at the Fraunhofer Center for Manufacturing Innovation at Boston University (CMI) and biologists working at the Fraunhofer Center for Molecular Biology (CMB). All of the individual modules described herein were developed, fabricated, and debugged at CMI, as was the overall factory control architecture and software. The modules were then shipped to CMB and installed in that facility (Fig. 13). Once debugged and fully operational, the factory was taken through a number of qualification and verification procedures on the path to FDA certification under cGMP guidelines. To date, the factory has been used to provide clinical trial material for multiple FDA-approved studies. In full production mode, the factory produces a nominal output of 50 kg of plant biomass per week. In combination with the proprietary transient protein expression technology, the factory produces target protein in the range of grams per kilogram harvested plant biomass.8,13,14
Figure 13. Solid model of overall factory showing relative placement of equipment. The plants are processed from left to right. Each module is color-coded: The seeding module is red, the growth modules are gray, the infiltration module is tan, the harvesting module is deep blue, and the handling robots and tracks are light blue. A dotted line outlines the clean rooms, separated from the other parts of the facility by lift gates.
The development of this plant-based pharmaceutical factory was a very successful example of translational research, at the intersection between engineering and biology. It was a cross-disciplinary undertaking that required a team of engineers and biologists to work closely together. At the outset, communication was not ideal because the engineers and biologists spoke different “languages.” With time, the team coalesced as the engineers learned some biology and the biologists learned some engineering. The end result is a testament to the value of the translational research that is needed to solve large, challenging problems and to take research all the way to commercialization.
In the past decade, new discovery technologies have accelerated the pace of vaccine discovery, but vaccine manufacturing technologies have essentially remained the same primarily because little incentive existed to change the postdiscovery downstream processes. However, the recent pandemic and biothreat assessment has provided incentive to overhaul and innovate these manufacturing processes. The automated plant-based factory presented here offers numerous advantages over alternative technologies, including very low overall costs, short production timelines (weeks), and low contamination risks. Unlike cell culture systems, our automated method does not require the time-consuming identification and isolation of high-producing cell clones that need to be scaled up for manufacture. In addition, the plants are grown in a soil-free, hydroponic medium under strict climatic and lighting controls for the highest degrees of reproducibility.
Although the purpose of the factory described in this article was to produce plant-based vaccines and biologic therapeutics, the underlying technologies and factory architecture can be used in plant-based food production as well. This would be especially beneficial in populated urban areas, in which there is insufficient land for agriculture.
Acknowledgements
We would like to acknowledge our colleagues at Fraunhofer CMB who developed the plant-based technology for vaccine production and defined the functional requirements for this factory. We would also like to thank Matthew Lipsitz and Patrick Sercel for their contributions to the design of the factory.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded in part by Defense Advanced Research Projects Agency (DARPA).
References
| 1. |
Centers for Disease Control and Prevention (CDC) . Vaccines and Preventable Diseases: Current Vaccine Shortages & Delays; CDC: Atlanta, GA, 2012. Google Scholar |
| 2. |
Matthews, J. T. Egg-Based Production of Influenza Vaccine: 30 Years of Commercial Experience. The Bridge 2006, 36(3), 17–24. Google Scholar |
| 3. |
Scannon, P. J. Pharmaceutical Preparedness for a Pandemic. The Bridge 2006, 36(3), 10–16. Google Scholar |
| 4. |
Fedson, D. S. NEW Technologies for Meeting the Global Demand for Pandemic Influenza Vaccines. Biologicals 2008, 36(6), 346–349. Google Scholar | Crossref | Medline | ISI |
| 5. |
Gerdil, C. The Annual Production Cycle for Influenza Vaccine. Vaccine 2003, 21(16), 1776–1779. Google Scholar | Crossref | Medline | ISI |
| 6. |
Tremblay, R., Wang, D., Jevnikar, A. M., Ma, S. Tobacco, a Highly Efficient Green Bioreactor for Production of Therapeutic Proteins. Biotechnol. Adv. 2010, 28(2), 214–221. Google Scholar | Crossref | Medline | ISI |
| 7. |
Obembe, O. O., Popoola, J. O., Leelavathi, S., Reddy, S. V. Advances in Plant Molecular Farming. Biotechnol. Adv. 2011, 29(2), 210–222. Google Scholar | Crossref | Medline | ISI |
| 8. |
Yusibov, V., Streatfield, S. J., Kushnir, N. Clinical Development of Plant-Produced Recombinant Pharmaceuticals: Vaccines, Antibodies and Beyond. Hum. Vaccin. 2011, 7(3), 313–321. Google Scholar | Crossref | Medline |
| 9. |
Ma, J. K. C., Drake, P. M. W., Christou, P. The Production of Recombinant Pharmaceutical Proteins in Plants. Nat. Rev. Genet. 2003, 4(10), 794–805. Google Scholar | Crossref | Medline | ISI |
| 10. |
Horn, M. E., Woodard, S. L., Howard, J. A. Plant Molecular Farming: Systems and Products. Plant Cell Rep. 2004, 22(10), 711–720. Google Scholar | Crossref | Medline | ISI |
| 11. |
D’Aoust, M. A., Couture, M. M. J., Charland, N., Trepanier, S., Landry, N., Ors, F., Vezina, L. P. The Production of Hemagglutinin-Based Virus-Like Particles in Plants: A Rapid, Efficient and Safe Response to Pandemic Influenza. Plant Biotechnol. J. 2010, 8(5), 607–619. Google Scholar | Crossref | Medline | ISI |
| 12. |
Penney, C. A., Thomas, D. R., Deen, S. S., Walmsley, A. M. Plant-Made Vaccines in Support of the Millennium Development Goals. Plant Cell Rep. 2011, 30(5), 789–798. Google Scholar | Crossref | Medline | ISI |
| 13. |
Shoji, Y., Farrance, C. E., Bautista, J., Bi, H., Musiychuk, K., Horsey, A., Park, H., Jaje, J., Green, B. J., Shamloul, M. A Plant-Based System for Rapid Production of Influenza Vaccine Antigens. Influenza Other Respi. Viruses. 2012, 6(3), 204–210. Google Scholar | Crossref | Medline | ISI |
| 14. |
Jul-Larsen, A., Madhun, A., Brokstad, K., Montomoli, E., Yusibov, V., Cox, R. The Human Potential of a Recombinant Pandemic Influenza Vaccine Produced in Tobacco Plants. Hum. Vaccin. Immunother. 2012, 8(5), 653–661. Google Scholar | Crossref | Medline | ISI |
