Biotechnology Advances 35 (2017) 458–465

Contents lists available at ScienceDirect

Biotechnology Advances

journal homepage: www.elsevier.com/locate/biotechadv

Research review paper

Very-large-scale production of antibodies in plants: The biologization

of manufacturing
J.F. Buyel a,b,⁎, R.M. Twyman c, R. Fischer a,b,1
a
Institute for Molecular Biotechnology, Worringerweg 1, RWTH Aachen University, 52074 Aachen, Germany
b
Fraunhofer Institute for Molecular Biology and Applied Ecology IME, Forckenbeckstraße 6, 52074 Aachen, Germany
c
TRM Ltd, PO Box 463, York YO11 9FJ, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: Gene technology has facilitated the biologization of manufacturing, i.e. the use and production of complex biolog-
Received 1 February 2017 ical molecules and systems at an industrial scale. Monoclonal antibodies (mAbs) are currently the major class of
Received in revised form 20 March 2017 biopharmaceutical products, but they are typically used to treat specific diseases which individually have compa-
Accepted 23 March 2017
rably low incidences. The therapeutic potential of mAbs could also be used for more prevalent diseases, but this
Available online 25 March 2017
would require a massive increase in production capacity that could not be met by traditional fermenter systems.
Keywords:
Here we outline the potential of plants to be used for the very-large-scale (VLS) production of biopharmaceutical
Biologization proteins such as mAbs. We discuss the potential market sizes and their corresponding production capacities. We
Molecular farming then consider available process technologies and scale-down models and how these can be used to develop VLS
Monoclonal antibodies processes. Finally, we discuss which adaptations will likely be required for VLS production, lessons learned from
Plant-based production existing cell culture-based processes and the food industry, and practical requirements for the implementation of
Process economy a VLS process.
Very-large-scale production © 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
2. Potential future applications and market size for monoclonal antibodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
3. Plant cultivation strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
4. Process schemes for monoclonal antibody production in plants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
4.1. Clarification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
4.2. Purification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
5. Scalable process models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
6. Process adaptations for VLS requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
7. Translation into VLS applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
8. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464

Abbreviations: ATPS, aqueous two-phase extraction; cGMP, current good manufacturing practice; CHO, Chinese hamster ovary; HCP, host cell protein; HIV/AIDS, human
immunodeficiency virus/acquired immunodeficiency syndrome; mAb, monoclonal antibody; SMB, simulated moving bed; UF/DF, ultrafiltration/diafiltration; VFU, vertical farming
unit; VLS, very large scale.
⁎ Corresponding author at: Institute for Molecular Biotechnology, Worringerweg 1, RWTH Aachen University, 52074 Aachen, Germany.
E-mail addresses: johannes.buyel@rwth-aachen.de (J.F. Buyel), richard@twymanrm.com (R.M. Twyman), fischer@molbiotech.rwth-aachen.de (R. Fischer).
1
Current address: Indiana Biosciences Research Institute, 1345 W. 16th St. Suite 300, Indianapolis, IN 46202, USA.

http://dx.doi.org/10.1016/j.biotechadv.2017.03.011
0734-9750/© 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 J.F. Buyel et al. / Biotechnology Advances 35 (2017) 458–465
1. Introduction
The earliest manufacturing by humans was biological, e.g. animal
skins as protection and tools made from wood. As technology developed
so did the use of non-biological materials, beginning with stone tools
and progressing to metals and eventually petrochemical derivatives
such as plastics, even though the latter have an ultimate biological ori-
gin (Kolb and Kolb, 1979). This trend reversed during the last century
and there is a renewed focus on biological manufacturing, mainly be-
cause biological products are often too complex for cost-effective total
chemical synthesis. However, even simple biological products are be-
ginning to replace synthetic counterparts because they are considered
more sustainable, i.e. they originate from renewable sources such as
plants. Arguably, the first complex biological products to have a major
impact on society were the earliest antibiotics (Fleming, 1929). Nowa-
days, the pharmaceutical industry is dominated by complex small mol-
ecules as well as protein-based drugs, and we are increasingly reliant on
biotechnology for manufacturing, which was foreseen almost 25 years
ago (della Valle and Gambardella, 1993). The scope of biological
manufacturing has recently expanded beyond the familiar territory of
structurally-complex drugs and has begun to embrace spheres more
traditionally dominated by chemical synthesis, particularly the replace-
ment of petrochemical materials with ‘biological’ counterparts such as
biodegradable plastics, biogas for energy and biofuels based on ethanol.
Fermentation is another staple of biotechnology that has been used
since ancient times as a form of food processing, but since the early
1970s it has evolved into a mainstream manufacturing technology for
biological products (Jackson et al., 1972). Without fermentation, the de-
velopment of recombinant DNA and gene cloning technology would
have taken much longer, delaying the benefits brought about by gene
transfer to living organisms including protein manufacturing and gene
therapy (MedCap Advisors, 2016).
Today's protein-based biopharmaceuticals include vaccines, en-
zymes and especially monoclonal antibodies (mAbs) (PhRMA, 2015).
Since the first therapeutic mAb was commercialized in 1986, the bio-
logics industry has expanded to include diverse protein-based drugs,
vaccines and technical reagents, and the market is now worth more
than $US 70 billion (Ecker et al., 2015). Despite the intense competition
for market share and a certain degree of bureaucratic inertia (della Valle
and Gambardella, 1993), the major players have more recently begun to
cooperate in areas of mutual benefit, such as regulatory compliance
(The CMC Biotech Working Group, 2009) and clinical data analysis
(Oo and Kalbag, 2016). The approval of the first plant-derived biophar-
maceutical protein in 2012 ushered in an era where biological
manufacturing has ‘returned to its roots’, in some cases quite literally,
with plants now emerging as an alternative to production systems
based on microbes and mammalian cells (Fischer et al., 2013; Stoger
et al., 2014; Tschofen et al., 2016). The potential benefits of plants for
the manufacture of biologicals include highly scalable and rapid produc-
tion, a low burden of human pathogens, and the ability to carry out
many of the post-translational modifications required for functional
human proteins (Buyel, 2015). Here we discuss recent developments
that presage a new era of biological manufacturing in which plants
could be used to achieve an economy of scale that would have been un-
imaginable even 10 years ago.
2. Potential future applications and market size for monoclonal
antibodies
Nearly 10 t of diagnostic and therapeutic mAbs were manufactured
in 2013, and the demand could double by 2020 assuming the annual
growth rate remains at the current level of ~ 10% (Ecker et al., 2015).
Much of the current demand centers on mAbs used for the diagnosis
and treatment of comparably rare diseases (Kelley, 2007). Even the big-
gest selling therapeutic mAbs, with millions of patients each requiring
gram quantities of product (such as rituximab for the treatment of
459
lymphomas), can be provided by todays large-scale manufacturing pro-
cesses based on fermenter capacities of up to 250,000 L (Ecker et al.,
2015). However, if mAbs are to be used for the prevention or treatment
of even more common diseases, e.g. Alzheimer's disease, HIV/AIDS or
dental caries (Ma et al., 2015), then the demand could escalate dramat-
ically, as observed when rituximab was approved for the treatment of
rheumatoid arthritis a (Ecker et al., 2015; Singh et al., 2009). For the pre-
vention of HIV infection, a typical dose of 30 mg mAb would be applied
twice weekly by susceptible women in the form of a vaginal gel, equat-
ing to ~ 100 doses per year. Therefore, N3 t of the mAb would be re-
quired per year per million women. The HIV+ population in sub-
Saharan Africa is currently 26 million, and the at-risk population is
therefore even higher. An effective prophylactic campaign scaled to en-
tire countries or regions would therefore require 50–100 t of mAb per
year, far outstripping the capacity of even the largest current processes
and defining a new era of very-large-scale (VLS) production more akin
to the food processing industry (Lokhorst et al., 2015).
Such demands are currently impossible to meet using mammalian
cell cultures in stirred tank bioreactors, which are today's gold standard
for production. Chinese hamster ovary (CHO) cells under optimal condi-
tions produce titers of 5–10 g L−1, and with the largest available fer-
menter volumes of 10,000–25,000 L and fermentation lasting
~12 days, the maximum output of a single fermenter would be 250 kg
per campaign or ~7.5 t per year assuming continuous and perfect oper-
ation (Kelley, 2007). In reality, the yield would be much lower. Further-
more, such VLS reactor setups would require stainless steel equipment,
thus attracting high investment costs but lacking the advantages of sin-
gle-use technologies that have become popular for both upstream pro-
duction and downstream processing in the pharmaceutical industry
over the last decade (Shukla and Gottschalk, 2013).
Unlike fermentation platforms, plants are effectively living single-
use bioreactors that can be scaled indefinitely simply by sowing more
seeds. Biomass yields of 10,000 t km−2 y− 1 have been reported for
closely cropped tobacco in the open field (Stoger et al., 2002), and a ver-
tical farming unit (VFU) has recently been built with a footprint of
~ 6500 m2 including ~2000 m2 used for plant growth, with an annual
output of 182 t of biomass (Holtz et al., 2015). Accordingly, the annual
space–biomass yield of the VFU was more than nine times higher than
open field production (91,000 t km−2 y−1). The yields of recombinant
protein achieved in the VFU reached 0.65 g kg− 1, but titers of up to
2 g kg−1 have been reported for mAbs produced in plants after several
rounds of optimization (Zischewski et al., 2015). Given these numbers
and an average recovery of 70% for plant-derived mAbs in a process
that handles 250 kg of biomass (our unpublished data), an annual
demand of 50 t of pure mAb would require ~72 t y−1 of bulk mAb pro-
duced using 3.5–11.0 km2 of open fields or 0.4–1.2 km2 of VFU area. Im-
portantly, VFUs can be designed as fully-automated plant-handling
facilities that comply with current good manufacturing practice
(cGMP) requirements (Wirz et al., 2012) whereas these requirements
are never likely to be met byopen fields. Although VFUs offer the most
promising approach, the production of mAbs in tonne amounts will be
accompanied by significant investment costs. However, similar or even
higher investments are necessary for CHO-based processes because the
same output would require a total bioreactor volume of ~250,000 L. Nev-
ertheless, the investments required for VFUs are likely to pay off due to
the large quantities of product that will be manufactured, meaning that
mAbs will become a “bulk product”. However, this may require a para-
digm shift, turning mAbs from “small quantity – high margin” to “large
quantity – low margin” products.
The cost of CHO cell medium is $US 55–90 L−1, corresponding to $US
14–22 million per batch for medium alone, which can be regarded as a
high risk investment given recent issues with contamination (Lolas,
2013). In contrast, media costs for a plant-based expression system
would be $US ~4.5 million, assuming recently reported biomass yields
and fertilizer consumption (Buyel and Fischer, 2012). The VLS produc-
tion of mAbs in plants therefore appears to have several advantages
460 J.F. Buyel et al. / Biotechnology Advances 35 (2017) 458–465
over cell culture platforms for upstream production, whereas the down-
stream capture and polishing steps would be similar to those already
used for the purification of mAbs from CHO cells, with similar cost impli-
cations (Ma et al., 2015; The CMC Biotech Working Group, 2009). Howev-
er, the extraction and clarification steps between upstream production
and downstream capture/polishing differ substantially between plant-
based and fermentation-based processes. The major difference is that
mAbs are generally secreted by cultured cells and recovered from the me-
dium, whereas they must be released from intact plants by some form of
mechanical tissue and cell disruption which also releases a large number
of dispersed particles and soluble impurities that must be removed prior
to chromatography (Buyel et al., 2015b). These issues require cost-effec-
tive and scalable solutions before VLS mAb production in plants becomes
commercially viable, as discussed in more detail below.
Previous studies have indicated that the production of mAbs in
plants is not competitive at low expression levels in the 0.02–
0.05 g kg− 1 range because the costs per gram of purified product
would reach €10,000–20,000 (Buyel and Fischer, 2012, 2014c) com-
pared to ~€200 g−1 for mAbs produced in CHO cells (Lim et al., 2010).
However, feeding the same cost models with data from recent reports
in which mAb yields in plants reached 2.0 mg kg−1 (Zischewski et al.,
2015) and in which filter capacities for the processing of plant extracts
reached 100–1000 L m−2 (Buyel et al., 2014b) showed that production
costs for plant-derived mAbs can fall substantially below €100 g−1 (our
unpublished data). Therefore, the production of mAbs in plants per se
can be economically competitive with CHO cells. Even if the investment
costs for the production facilities are similar, the operating costs for
plant-based processes will be lower especially due to the simple cultiva-
tion of intact plants compared to the cost-intensive sterile requirements
for cell suspension cultures.
3. Plant cultivation strategies
There are several types of plant-based expression systems that re-
quire cultivation techniques similar to those used with mammalian cell
cultures, e.g. aquatic plants, plant cell suspension cultures, and plant tis-
sue cultures such as hairy roots (Xu et al., 2012). Even though these sys-
tems can benefit from simple nutrient requirements, the absence of
animal components and pathogens, they will be subject to the same
scale-up limitations as CHO cells as they require sterilizable fermenters.
Thus, they are ultimately unsuitable for the VLS production of mAbs. In
contrast, there are two major expression strategies using intact plants
that have the potential to reduce the operating costs for upstream pro-
duction substantially compared to cell cultures. In the transient expres-
sion approach, wild-type plants are vacuum infiltrated with
Agrobacterium tumefaciens carrying a vector encoding the mAb heavy
and light chains, allowing large quantities of mAb to accumulate within
5–8 days (Shoji et al., 2012). Although this method is rapid, its potential
drawback for VLS production is that each plant in every batch must be in-
filtrated with bacteria, which merely transfers the fermentation costs to
the production of bacteria (Houdelet et al., 2017). This is not the case for
transgenic plants, where well-defined master and working seed banks
can be established and used for consecutive batches without the need
for additional manipulation (Sack et al., 2015). Therefore, transgenic
plants appear to be the most suitable plant-based format for VLS produc-
tion. There are three options to cultivate such plants: under open field
conditions, in conventional greenhouses or in VFUs. Open fields offer vir-
tually unlimited scaling potential but no containment, so it is unlikely
that this approach will be used for pharmaceutical products (Fischer et
al., 2012). In contrast, VFUs provide a high degree of containment and
can designed to comply with cGMP as described above. However, this
benefit comes at the cost of greater capital investment for the vertical
cultivation racks and automation equipment. Greenhouses offer a com-
promise between capital investment and ease of scale-up. The plants
we used to produce the first cGMP-compliant batch of mAb used in clin-
ical trials were cultivated in a greenhouse (Ma et al., 2015). Greenhouses
or VFUs therefore appear to be the most likely scenarios for the VLS pro-
duction of mAbs in plants.
4. Process schemes for monoclonal antibody production in plants
One key characteristic of plant-based manufacturing processes is the
broad spectrum of techniques initially used by academic and industrial
development teams to facilitate extraction, clarification and product re-
covery (Fig. 1). These techniques have recently been reviewed (Buyel et
al., 2015b) and continuous operations, such as screw-press extraction,
are likely to be the most beneficial for VLS applications (Buyel and
Fischer, 2014c). Typically, large numbers of dispersed particles and
large quantities of host cell protein (HCP) are released during product
extraction, which can be addressed by using flocculants or other filter
aids, as well as various conditioning steps such as a temperature or pH
shift. These measures can increase the capacity of filters to
~1000 L m−2 and remove N90% of the HCP (Buyel and Fischer, 2014b;
Menzel et al., 2016).
4.1. Clarification
The most common unit operations for clarification are centrifugation
and/or filtration. The former can be difficult to scale up due to the changes
in the apparatus geometry, and it may be necessary to reduce the flow
rate by N50% to maintain clarification capacity or settling area (Roush
and Lu, 2008). Combinations of filters are therefore preferred for the clar-
ification of plant extracts because they can also remove pigments, HCP
and DNA (Kandula et al., 2009). However, the filter must be chosen care-
fully because protein products can adsorb to the diatomite contained in
most filters (Buyel et al., 2015a). The optimal combination of filters in
terms of capacity, clarification and product recovery must be determined
for each process depending on the particle size distribution, e.g. using bag
filters, depth filters and/or pre-coat filters.
4.2. Purification
The purification of mAbs typically involves an initial capture step
using a Protein A resin (The CMC Biotech Working Group, 2009). There-
fore, despite the different upstream production and extraction/clarifica-
tion procedures associated with fermenters and plants, most mAb
production processes converge at this process step. The pH and temper-
ature shifts discussed above may also facilitate mAb purification be-
cause both conditioning steps may inactive proteases that could harm
the product and/or the Protein A resin. This is particularly important
in the context of the resin life cycle because Protein A resins are one of
the largest capital investment costs, i.e. ~$US 11 million for the produc-
tion of 10 t of mAb using CHO cells (Kelley, 2007). In contrast to the pro-
teases mentioned above, mAbs typically remain stable at pH ~3.5 and
~70 °C (Rouet et al., 2014; Woodard et al., 2009).
Regardless of the conditioning, the concentrations of mAbs in plant
extracts are generally in the 0.1–0.2 g L−1 range and thus 50–100-fold
lower compared to processes based on optimized CHO cells, although
yields of N 2 g kg−1 have also been reported for plants, corresponding
to ~ 0.5 g L−1 (Zischewski et al., 2015). Accordingly, instead of the
high binding capacities (~ 50 g L− 1) of packed-bed columns
(Thommes and Etzel, 2007), the faster flow rates of membrane ad-
sorbers (up to 18 m h−1) help to reduce the process time (Varadaraju
et al., 2011). Alternatively, an ultrafiltration/diafiltration (UF/DF) step
can be integrated prior to Protein A column chromatography to reduce
the process volume and accelerate product recovery (Lightfoot and
Cockrem, 2013) particularly if the mAb concentrations are higher (e.g.
N0.5 g L−1). Cost-effective replacements for the Protein A ligand, e.g.
mixed-mode resins such as MEP HyperCel that are designed specifically
to bind mAbs, will also be valuable for VLS mAb manufacturing
(Jackewitz, 2012), especially if the nonspecific binding of HCPs is
negligible because they have been removed by pre-conditioning
 J.F. Buyel et al. / Biotechnology Advances 35 (2017) 458–465 461

Fig. 1. Comparison of process schemes for the large-scale production of mAbs. Production in CHO cells (A; (Kelley, 2007)) is compared to a reported tobacco-based platform (B; (Ma et al.,
2015)) and its recent improvement (C; Fraunhofer IME unpublished data) as well as a hypothetical continuous very-large-scale process (D). API – active pharmaceutical ingredient.
462 J.F. Buyel et al. / Biotechnology Advances 35 (2017) 458–465
steps such as low-pH incubation or heat treatment (Buyel et al.,
2014a; Hassan et al., 2008).
The polishing steps that follow the Protein A capture step are sim-
ilar for all mAb manufacturing platforms, typically involving ion ex-
change or hydrophobic interaction chromatography (Ma et al., 2015;
The CMC Biotech Working Group, 2009). A design-of-experiments
approach is often recommended for the optimization of these steps
(Buyel and Fischer, 2013) because this exploits the individual prop-
erties of each mAb product and complies with quality-by-design
guidelines (Rathore, 2009).
Virus filtration is another cost driver for mAb production, accounting
for up to 12% of the cost of goods for processes based on CHO cells
(Kelley, 2007). This step may not be necessary for plant-based produc-
tion processes because plants do not support the growth of human
viral pathogens (Fischer et al., 2012). To comply with industry standards
(Shukla and Gottschalk, 2013), a virus-filtration step was incorporated
as part of the manufacturing process for the HIV-neutralizing mAb
2G12 produced in tobacco (Ma et al., 2015). However, this accounted
for b5% of the consumables costs and b 3% of the predicted cost of
goods (our unpublished data).
Selecting the sequence of unit operations to achieve the most cost-
effective purification becomes more challenging when using plants for
VLS mAb production. It is prohibitive to perform VLS process optimiza-
tion ‘at scale’ due to the high costs of consumables and labor. Instead,
the rapid testing of different process setups (including cost estimates
for scaled-up production) is facilitated by a rational model-based ap-
proach. This requires the availability of models that adequately describe
the different unit operations, as discussed in more detail below.
5. Scalable process models
The choice of unit operations and devices for a given mAb produc-
tion process should be based on the recovery, yield, purity and per-
formance that can be achieved at production scale. Small-scale
models can be used to assess these quality indicators in a rapid and
cost-effective manner if they provide scalable information about
the various devices, facilitating cost estimates for different process
setups (Buyel et al., 2015a; Chhatre et al., 2011). Two constraints
apply to the sample size used in these approaches. Smaller sample
or aliquot sizes, e.g. 50–2000 μL in 96- or 384-well plate formats,
are typically associated with lower costs and higher throughput,
and thus facilitate the screening of a broader set of process condi-
tions to increase the likelihood of identifying optimal parameter
combinations, e.g. achieving the highest yield, lowest costs or fastest
processing. However, sample volumes must be large enough to en-
sure that equipment with properties comparable to process-scale
devices can be used, which may require N0.2 L per run (Buyel and
Fischer, 2014d; Roush and Lu, 2008). Recent developments have re-
duced the sample volumes required for flocculation testing to
b1.0 mL, and authentic process-scale geometries have been introduced
into small filtration devices (Buyel et al., 2015a; Espuny Garcia Del Real
et al., 2014). Screw-press and blender-based extraction have been
shown to scale effectively from 150 g to N 200 kg biomass (Buyel and
Fischer, 2014b), and rapid, scalable screening formats are also available
for chromatography-based purification steps (Chollangi et al., 2014;
Keller et al., 2015). If scalable equipment is not available for screening,
a two-tier approach can be applied, in which the first dataset is gathered
in a screening format that yields qualitative results, and the most prom-
ising steps are then verified using scalable equipment that provides
quantitative data and estimates the production costs. Combining the
data for different process steps allows the overall performance of the
process to be evaluated, so that different setups can be compared in
small-scale experiments.
Several functions have been developed to optimize dependent vari-
ables (also known as “responses”) simultaneously, thus maximizing the
desirability of a process outcome (Anderson and Whitcomb, 2005).
More than one process condition may achieve the same overall desir-
ability, e.g. as a tradeoff between productivity and resin capacity utiliza-
tion, a situation described as Pareto optimality, hence the final choice of
process may depend on manufacturer preferences. A set of the most de-
sirable combinations of operations can be selected, e.g. based on mAb
yield. However, before this presumably optimal process is scaled up,
its potential benefits must be compared with the cost of implementa-
tion at the production scale.
6. Process adaptations for VLS requirements
The performance of optimal processes (e.g. the highest mAb
yield) in small-scale experiments must be confirmed at the pilot or
process scale to ensure that the selected operation conditions
match with the corresponding VLS equipment. However, process
scale-up must be preceded by the detailed cost–benefit analysis of
each step, taking into account all aspects that may be negligible during
screening or small-scale evaluation, but that can become cost drivers in
VLS processes. For example, flocculation was found to increase depth
filter capacity during the clarification of plant extracts by a factor of
five, reducing the cost of filters accordingly. However, implementing
flocculation at the process scale also introduces additional costs that
are not evident at the laboratory scale (e.g. an additional hold tank to
accommodate the flocculation process, which as a consequence be-
comes discontinuous or semi-continuous at best if the process cycles
among several hold tanks). Furthermore, pumps are required to add
the flocculant, which must be pharmaceutical grade in terms of purity
and must be held with a sufficient inventory. Finally, the flocculant
may interfere with subsequent chromatography steps (Buyel and
Fischer, 2014a) and adequate quality controls may be required to dem-
onstrate that the flocculant is not present in the final mAb formulation.
The different operations available for extraction, clarification, HCP
removal and purification are summarized in Table 1 in terms of their
benefits and drawbacks for the VLS production of mAbs, and these
must be balanced according to their effects on the cost of goods. Here,
the overall effect on process or product costs must be considered rather
than the impact on a particular process step. For example, if UF/DF is
used to reduce the process volume before Protein A capture, the isolated
evaluation of this step would indicate an increase in costs. However, if
the entire process is taken into account, the benefits of process volume
reduction achieved by UF/DF would include the use of a smaller Protein
A column and smaller downstream storage/hold tanks, thus reducing
both the capital and consumables costs overall. These savings should
then be offset against the increased costs of implementing the UF/DF
step to determine whether there is an overall benefit.
One trend that is apparent for the large-scale manufacturing of mAbs
using mammalian cells is the increasing number of continuous process-
es, either in part (such as cultivation and chromatography) or even
completely integrated continuous production (Klutz et al., 2015;
Xenopoulos, 2015). The benefits of continuous production include the
higher space–time yield of manufacturing facilities, smaller footprints
and lower investment costs. Therefore, plant-based production facilities
should follow this lead if possible, adjusting for the discrete growth
phases of plants by staggered cultivation and automated harvesting
and extraction. In this context, regulatory issues may arise from contin-
uous manufacturing in terms of batch definition (Hertrampf et al.,
2015), and technical issues may arise due to the need for some equip-
ment redundancy to insure against device failure.
7. Translation into VLS applications
To our knowledge, the largest cGMP-compliant plant-based produc-
tion process for mAbs has been established by the Fraunhofer IME (Aa-
chen, Germany) for the manufacture of 2G12. This facility can process
250 kg of tobacco leaf biomass per batch, corresponding to approxi-
mately 900 L of bulk extract (Ma et al., 2015). The original downstream
 J.F. Buyel et al. / Biotechnology Advances 35 (2017) 458–465 463

Table 1
Benefits and drawbacks of downstream processing operations for the very-large-scale (VLS) production of mAbs.

Process stage Operation Benefits Drawbacksa Comments

Extraction Blade-based Applicable to all types of proteins Discontinuous Alternating operation of several devices
homogenizer Well established at kg scale Limited scalability due to can facilitate semi-continuous
volume/area ratio and power input processing
Infiltration centrifugation Reduced HCP extraction Discontinuous
Reduced particle burden Limited scalability due to g-force
requirements
Limited to secreted proteins
Technically demanding
Rhizosecretion Continuous Dilute target protein Hydroponic cultivation is required
Reduced hcp extraction Large process volumes
Reduced particle burden Limited to secreted proteins
Screw-press Applicable to all types of proteins May interfere with target proteins Extraction in the absence of buffer
Continuous Sensitive to low pH
Reduced particle burden
Well established at tonne scale
Clarification Centrifugation Low consumables costs Discontinuous
Well established at L scale High energy demand
Limited scalability due to g-force
requirements
Filtration Continuous High consumables costs
Established at m3 scale May bind target protein
Linear scalability
DNA/HCPb reduction
Filter aids Established at m3 scale Increased investment and
Increased filter capacity Consumables costs
May bind target protein
Flocculation/filter aids Established at m3 scale Increased investment and Special assays confirming the absence
Increased filter capacity consumables costs of flocculants in the final product can
May bind target protein be required
May interfere with subsequent
process steps
Conditioning/HCP ATPSb Increased purity Increased investment and Continuous operation has been
removal Process volume reduction consumables costs reported
Reduced particle burden Increased disposal costs
Heat treatment Continuous High energy demand
Established at m3 scale Increased investment costs
Increased purity May interfere with target protein
Low-pH precipitation Increased purity Discontinuous Continuous operation may be possible
Increased investment and via in-line mixing
consumables costs
May interfere with target protein
UF/DFb Process volume reduction Increased investment costs Continuous operation possible
Established at m3 scale Increased purity for certain target
proteins
Purification Membrane adsorbers High flow rates Discontinuous Supply may be difficult
Low binding capacity
Packed-bed High binding capacity Discontinuous
chromatography Low flow rates
SMBb chromatography Continuous Sophisticated control system Applicable to adsorbers and
Reduced investment and consumables required packed-bed columns
costs
a
Discontinuous operations may be converted to semi-continuous operations by a combination of numbering-up and alternating use.
b
Abbreviations: ATPS - aqueous two-phase extraction, HCP - host cell protein, SMB - simulated moving bed, UF/DF - ultrafiltration/diafiltration.

process incorporated alternating blade-based homogenizers, five filtra-
tion steps for extract clarification, a Protein A membrane adsorber for
mAb capture and initial purification, a subsequent ceramic hydroxyap-
atite polishing step followed by nanofiltration, UF/DF, and a final 0.2-
μm sterile filtration step (Fig. 1). Since 2009, this process has been mod-
ified to accommodate a packed-bed Protein A column to replace the
membrane adsorbers that are no longer available from the manufactur-
er. Therefore, an additional UF/DF step was introduced before the cap-
ture chromatography step to concentrate the extract by a factor of 15,
reducing the time required for loading the Protein A column by the
same factor but otherwise maintaining the overall process time. Fur-
thermore, the small-scale screening of new depth filters resulted in an
optimized clarification procedure, reducing the number of steps from
five to three and the cost of goods by 40% (our unpublished data).
These modifications highlight how cost–benefit analysis can be affected
by the changing consumables market, i.e. in this case discontinued
membrane adsorbers and the availability of new filters. Processes for
the VLS production of mAbs should therefore be designed with the
long-term stability of raw material supplies in mind.
Using the expression levels and product recoveries discussed above,
the 250-kg Fraunhofer IME process would yield 0.0002–0.0005 t of mAb
per batch, which corresponds to an annual output of 0.005–0.016 t as-
suming 45 batches per year. This falls short of the anticipated top-end
demand of 50 t per year by a factor of at least 3000. Accordingly, the cur-
rent “large-scale” plant-based mAb production process can be regarded
as “small scale” in the context of VLS mAb manufacturing. The current
process equipment is therefore unlikely to be suitable if the process is
scaled up by three orders of magnitude.
Instead of trying to develop entirely new solutions for this scale-up
problem, the plant biotechnology community can learn and benefit
from the expertise developed during the optimization of CHO cells, par-
ticularly with regard to chromatography (Fig. 2). Examples include the
potential integration of simulated moving bed chromatography (Baur
et al., 2015) and continuous processing operations to reduce the size
464 J.F. Buyel et al. / Biotechnology Advances 35 (2017) 458–465
Fig. 2. Comparison of devices and equipment currently used in pilot-scale production and potential alternatives for VLS production.
and footprint of the equipment, increasing the space–time yield as de-
scribed for the single-use continuous manufacturing of mAbs (Klutz et
al., 2015). Another benefit of smaller equipment is the reduction in cap-
ital costs, which is particularly striking for the Protein A resin, resulting
in cost reductions of 10–30% (Angarita et al., 2015).
The benefits of VLS processes are likely to be even greater if we
also embrace the technologies, experiences and procedures that
have been gathered during almost 200 years of industrialized food
processing (Fig. 2) (Goody, 1997). The food-processing industry
often handles multi-tonne quantities of plant biomass, which is
processed by extraction, blanching, filtration and separation, under
highly-regulated hygienic conditions designed to minimize process-
ing errors (Kress-Rogers and Brimelow, 2001; Vasconcellos, 2003).
In this context, the VFU concept described above could be beneficial
because it brings upstream production and downstream processing
into direct spatial proximity, and facilitates the continuous process-
ing of plant biomass from seeding, through cultivation and harvest-
ing, to extraction and product purification. For downstream
processing, one could envisage a fully-integrated continuous process
beginning with screw-press extraction followed by filtration for clarifi-
cation and single-pass UF/DF for concentration and a final sequence of
simulated moving bed chromatography steps for mAb purification
(Fig. 1). Once such a fully automated facility completes the proof-of-
concept phase under cGMP conditions, it could feasibly be scaled up
to achieve an annual multi-tonne output for plant-derived mAbs. This
would revolutionize the biologics market by making therapeutic mAbs
affordable in both developed and developing economies and thus acces-
sible to many more patients than currently possible, providing an eco-
nomically feasible strategy for the prevention and treatment of
prevalent diseases.
8. Conclusions
In the future, mAbs could be used for the prevention or treatment of
many common diseases, including global health challenges such as ma-
laria, HIV/AIDS and tuberculosis. Facilitating access to the corresponding
medicines will be challenging given current market prices and the glob-
al capacities of production platforms based on mammalian cells. How-
ever, advances in mAb production using mammalian cells (e.g.
continuous manufacturing) could be combined with the established
large-scale methods used in the food processing industry and recent im-
provements in plant molecular farming to allow the VLS production of
mAbs in plants, i.e. production at the multi-tonne scale, including ge-
nerics and biobetters. The establishment of such an industry will require
substantial investment, and it remains to be seen which incentives and
motivations will be required to encourage public and private stake-
holders to tackle this challenge.
Acknowledgements
This work was funded by the Fraunhofer-Gesellschaft Internal Pro-
grams under Grant No. Attract 125-600164 and the European Research
Council Advanced Grant “Future-Pharma”, proposlal number 269110.
The authors have no conflict of interest to declare.
References
Anderson, M.J., Whitcomb, P.J., 2005. Response Surface Methods Simplified. Productivity
Press.
Angarita, M., Muller-Spath, T., Baur, D., Lievrouw, R., Lissens, G., Morbidelli, M., 2015.
Twin-column CaptureSMB: a novel cyclic process for protein A affinity chromatogra-
phy. J. Chromatogr. A 1389, 85–95.
Baur, D., Angarita, M., Muller-Spath, T., Morbidelli, M., 2015. Pareto-optimal design of the
twin-column CaptureSMB process improves capacity utilization and productivity in
protein A affinity capture. Biotechnol. J. 11, 135–145.
Buyel, J.F., 2015. Process development strategies in plant molecular farming. Curr. Pharm.
Biotechnol. 16, 966–982.
Buyel, J.F., Fischer, R., 2012. Predictive models for transient protein expression in tobacco
(Nicotiana tabacum L.) can optimize process time, yield, and downstream costs.
Biotechnol. Bioeng. 109, 2575–2588.
Buyel, J.F., Fischer, R., 2013. Processing heterogeneous biomass: overcoming the hurdles
in model building. Bioengineered 4, 21–24.
Buyel, J.F., Fischer, R., 2014a. Downstream processing of biopharmaceutical proteins pro-
duced in plants: the pros and cons of flocculants. Bioengineered 5, 138–142.
Buyel, J.F., Fischer, R., 2014b. Flocculation increases the efficacy of depth filtration during
the downstream processing of recombinant pharmaceutical proteins produced in to-
bacco. Plant Biotechnol. J. 12, 240–252.
Buyel, J.F., Fischer, R., 2014c. A juice extractor can simplify the downstream processing of
plant-derived biopharmaceutical proteins compared to blade-based homogenizers.
Process Biochem. 50, 859–866.
Buyel, J.F., Fischer, R., 2014d. Scale-down models to optimize a filter train for the down-
stream purification of recombinant pharmaceutical proteins produced in tobacco
leaves. Biotechnol. J. 9, 415–425.
Buyel, J.F., Gruchow, H.M., Boes, A., Fischer, R., 2014a. Rational design of a host cell protein
heat precipitation step simplifies the subsequent purification of recombinant pro-
teins from tobacco. Biochem. Eng. J. 88, 162–170.
Buyel, J.F., Opdensteinen, P., Fischer, R., 2014b. Cellulose-based filter aids increase the ca-
pacity of depth filters during the downstream processing of plant-derived biophar-
maceutical proteins. Biotechnol. J. 10, 584–591.
Buyel, J.F., Gruchow, H.M., Fischer, R., 2015a. Depth filters containing diatomite achieve
more efficient particle retention than filters solely containing cellulose fibers. Front.
Plant Sci. 6, 1–11.
Buyel, J.F., Twyman, R.M., Fischer, R., 2015b. Extraction and downstream processing of
plant-derived recombinant proteins. Biotechnol. Adv. 33, 902–913.
Chhatre, S., Pampel, L., Titchener-Hooker, N.J., 2011. Integrated use of ultra scale-down
and financial modeling to identify optimal conditions for the precipitation and cen-
trifugal recovery of milk proteins. Biotechnol. Prog. 27, 998–1008.
 J.F. Buyel et al. / Biotechnology Advances 35 (2017) 458–465
Chollangi, S., Jaffe, N.E., Cai, H., 2014. Accelerating purification process development of an
early phase mAb with high-throughput automation part 2. BioProcess International.
12, pp. 48–52.
Ecker, D.M., Jones, S.D., Levine, H.L., 2015. The therapeutic monoclonal antibody market.
MAbs 7, 9–14.
Espuny Garcia Del Real, G., Davies, J., Bracewell, D.G., 2014. Scale-down characterization of
post-centrifuge flocculation processes for high-throughput process development.
Biotechnol. Bioeng. 111, 2486–2498.
Fischer, R., Schillberg, S., Hellwig, S., Twyman, R.M., Drossard, J., 2012. GMP issues for re-
combinant plant-derived pharmaceutical proteins. Biotechnol. Adv. 30, 434–439.
Fischer, R., Schillberg, S., Buyel, J.F., Twyman, R.M., 2013. Commercial aspects of pharma-
ceutical protein production in plants. Curr. Pharm. Des. 19, 5471–5477.
Fleming, A., 1929. On the antibacterial action of cultures of a penicillium, with special ref-
erence to their use in the isolation of B. influenzæ. Br. J. Exp. Pathol. 10, 226–236.
Goody, J., 1997. Industrial food. Towards the development of a world cuisine. In: Counihan,
C., van Esterik, P. (Eds.), Food and Culture. Routledge, London, pp. 338–356.
Hassan, S., van Dolleweerd, C.J., Ioakeimidis, F., Keshavarz-Moore, E., Ma, J.K., 2008. Con-
siderations for extraction of monoclonal antibodies targeted to different subcellular
compartments in transgenic tobacco plants. Plant Biotechnol. J. 6, 733–748.
Hertrampf, A., Muller, H., Menezes, J.C., Herdling, T., 2015. A PAT-based qualification of
pharmaceutical excipients produced by batch or continuous processing. J. Pharm.
Biomed. Anal. 114, 208–215.
Holtz, B.R., Berquist, B.R., Bennett, L.D., Kommineni, V.J., Munigunti, R.K., White, E.L., et al.,
2015. Commercial-scale biotherapeutics manufacturing facility for plant-made phar-
maceuticals. Plant Biotechnol. J. 13, 1180–1190.
Houdelet, M., Galinski, A., Holland, T., Wenzel, K., Schillberg, S., Buyel, J.F., 2017. Animal
component-free Agrobacterium tumefaciens cultivation media for better GMP-compli-
ance increases biomass yield and pharmaceutical protein expression in Nicotiana
benthamiana. Biotechnol. J. (In press).
Jackewitz, A., 2012. Development of a high capacity, mixed-mode resin for high conduc-
tivity mAb feedstocks. Biopharm Int. 10, 82–84.
Jackson, D.A., Symons, R.H., Berg, P., 1972. Biochemical method for inserting new genetic
information into DNA of Simian Virus 40: circular SV40 DNA molecules containing
lambda phage genes and the galactose operon of Escherichia coli. Proc. Natl. Acad.
Sci. U. S. A. 69, 2904–2909.
Kandula, S., Babu, S., Jin, M., Shukla, A.A., 2009. Design of a filter train for precipitate re-
moval in monoclonal antibody downstream processing. Biotechnol. Appl. Biochem.
54, 149–155.
Keller, W.R., Evans, S.T., Ferreira, G., Robbins, D., Cramer, S.M., 2015. Use of MiniColumns
for linear isotherm parameter estimation and prediction of benchtop column perfor-
mance. J. Chromatogr. A 1418, 94–102.
Kelley, B., 2007. Very large scale monoclonal antibody purification: the case for conven-
tional unit operations. Biotechnol. Prog. 23, 995–1008.
Klutz, S., Magnus, J., Lobedann, M., Schwan, P., Maiser, B., Niklas, J., et al., 2015. Developing
the biofacility of the future based on continuous processing and single-use technolo-
gy. J. Biotechnol. 213, 120–130.
Kolb, D., Kolb, K.E., 1979. Petroleum chemistry. J. Chem. Educ. 56, 465.
Kress-Rogers, E., Brimelow, C.J.B., 2001. Instrumentation and Sensors for the Food Indus-
try. Woodhead, Cambridge, UK.
Lightfoot, E.N., Cockrem, M.C.M., 2013. Complex fitness diagrams: downstream process-
ing of biologicals. Sep. Sci. Technol. 48, 1753–1757.
Lim, J.A.C., Patkar, A., McDonagh, G., Sinclair, A., Lucy, P., 2010. Modeling bioprocess cost.
BioProcess International. 8, pp. 62–70.
Lokhorst, H.M., Plesner, T., Laubach, J.P., Nahi, H., Gimsing, P., Hansson, M., et al., 2015.
Targeting CD38 with daratumumab monotherapy in multiple myeloma. N. Engl.
J. Med. 373, 1207–1219.
Lolas, A., 2013. Microbial Control Strategies in Bioprocessing Falling Short of Assuring
Product Quality and Satisfying Regulatory Expectations. American Pharmaceutical
Review. Fishers: CompareNetworks.
Ma, J.K., Drossard, J., Lewis, D., Altmann, F., Boyle, J., Christou, P., et al., 2015. Regulatory
approval and a first-in-human phase I clinical trial of a monoclonal antibody pro-
duced in transgenic tobacco plants. Plant Biotechnol. J. 13, 1106–1120.
MedCap Advisors, 2016. Biologics Revolution. MedCap Viewpoint. MedCap Advisors,
Cambridge, MA, p. 5.
465
Menzel, S., Holland, T., Boes, A., Spiegel, H., Bolzenius, J., Fischer, R., et al., 2016. Optimized
blanching reduces the host cell protein content and substantially enhances the recov-
ery and stability of two plant-derived malaria vaccine candidates. Front. Plant Sci. 7,
1–15.
Oo, C., Kalbag, S.S., 2016. Leveraging the attributes of biologics and small molecules, and
releasing the bottlenecks: a new wave of revolution in drug development. Expert.
Rev. Clin. Pharmacol. 9, 747–749.
PhRMA, 2015. In: PhRMA (Ed.), Biopharmaceutical Research Industry Profile, 1 ed. Phar-
maceutical Research and Manufacturers of America, Washington, D.C., USA, pp. 1–76
Profile.
Rathore, A.S., 2009. Roadmap for implementation of quality by design (QbD) for biotech-
nology products. Trends Biotechnol. 27, 546–553.
Rouet, R., Lowe, D., Christ, D., 2014. Stability engineering of the human antibody reper-
toire. FEBS Lett. 588, 269–277.
Roush, D.J., Lu, Y.F., 2008. Advances in primary recovery: centrifugation and membrane
technology. Biotechnol. Prog. 24, 488–495.
Sack, M., Rademacher, T., Spiegel, H., Boes, A., Hellwig, S., Drossard, J., et al., 2015. From
gene to harvest: insights into upstream process development for the GMP production
of a monoclonal antibody in transgenic tobacco plants. Plant Biotechnol. J. 13,
1094–1105.
Shoji, Y., Farrance, C.E., Bautista, J., Bi, H., Musiychuk, K., Horsey, A., et al., 2012. A plant-
based system for rapid production of influenza vaccine antigens. Influenza Other
Respir. Viruses 6, 204–210.
Shukla, A.A., Gottschalk, U., 2013. Single-use disposable technologies for biopharmaceuti-
cal manufacturing. Trends Biotechnol. 31, 147–154.
Singh, J.A., Christensen, R., Wells, G.A., Suarez-Almazor, M.E., Buchbinder, R., Lopez-Olivo,
M.A., et al., 2009. Biologics for rheumatoid arthritis: an overview of cochrane reviews.
Cochrane Database Syst. Rev., CD007848
Stoger, E., Sack, M., Perrin, Y., Vaquero, C., Torres, E., Twyman, R.M., et al., 2002. Practical
considerations for pharmaceutical antibody production in different crop systems.
Mol. Breed. 9, 149–158.
Stoger, E., Fischer, R., Moloney, M., Ma, J.K.C., 2014. Plant molecular pharming for the
treatment of chronic and infectious diseases. Annu. Rev. Plant Biol. 65, 743–768.
The CMC Biotech Working Group, 2009. A-Mab: a case study in bioprocess development.
A-Mab: A Case Study in Bioprocess Development, 2.1 ed. CASSS - An International
Separation Science Society, Emeryville, CA, USA, pp. 1–278.
Thommes, J., Etzel, M., 2007. Alternatives to chromatographic separations. Biotechnol.
Prog. 23, 42–45.
Tschofen, M., Knopp, D., Hood, E., Stoger, E., 2016. Plant molecular farming: much more
than medicines. Annu. Rev. Anal. Chem. 9, 271–294.
della Valle, F., Gambardella, A., 1993. ‘Biological’ revolution and strategies for innovation
in pharmaceutical companies. R&D Manag. 23, 287–302.
Varadaraju, H., Schneiderman, S., Zhang, L.F., Fong, H., Menkhaus, T.J., 2011. Process and
economic evaluation for monoclonal antibody purification using a membrane-only
process. Biotechnol. Prog. 27, 1297–1305.
Vasconcellos, J.A., 2003. Quality Assurance for the Food Industry: A Practical Approach.
CRC Press, New York, NY.
Wirz, H., Sauer-Budge, A.F., Briggs, J., Sharpe, A., Shu, S.D., Sharon, A., 2012. Automated
production of plant-based vaccines and pharmaceuticals. J. Lab. Autom. 17, 449–457.
Woodard, S.L., Wilken, L.R., Barros, G.O.F., White, S.G., Nikolov, Z.L., 2009. Evaluation of
monoclonal antibody and phenolic extraction from transgenic Lemna for purification
process development. Biotechnol. Bioeng. 104, 562–571.
Xenopoulos, A., 2015. A new, integrated, continuous purification process template for
monoclonal antibodies: process modeling and cost of goods studies. J. Biotechnol.
213, 42–53.
Xu, J.F., Dolan, M.C., Medrano, G., Cramer, C.L., Weathers, P.J., 2012. Green factory: plants
as bioproduction platforms for recombinant proteins. Biotechnol. Adv. 30,
1171–1184.
Zischewski, J., Sack, M., Fischer, R., 2015. Overcoming low yields of plant-made antibodies
by a protein engineering approach. Biotechnol. J. 11, 107–116.