Manufacturing

formation in near real-time and can be

placed in-line in process flow streams.
In-line spectroscopic sensors using
infrared, near-infrared, or Raman
spectroscopy are one such new class
of tools finding multiple PAT appli-
cations in bioprocessing (8). Spectro-
scopic PAT applications typically con-
sist of in-line sensors combined with

Near Infrared statistical or modelling methodology

to monitor, control, and/or predict

Spectroscopy as
biotechnology processes. The key ad-
vantage of spectroscopy is the ability
to gain information about the pres-

a Versatile PAT ence and quantity of multiple analytes

in the process mixture simultaneously
in a few seconds or less, overcoming

Tool for Continuous a major analytical bottleneck and fa-

cilitating real-time control decisions
(9). Spectroscopic techniques are also

DownstreamBioprocessing non-degradative and provide a wealth

of spectral information that can be
used in tracking process trajectories
and flagging deviations (10,11). Thus,
Garima Thakur and Anurag S. Rathore spectroscopic probes and f low cells
are well-suited to PAT applications,
particularly in tandem with multi-
The need for real-time monitoring and control has variate data analytics (MVDA) tools,
which are critical for reducing the di-
spurred the development of new analytical tools. mensionality of large spectral datasets
and extracting statistically significant

P
rocess analytical technology than “testing quality into the product” quantitative information (12).
(PAT) is an increasingly import- (2). The importance of PAT tools is Table I summarizes the recent litera-
ant aspect of biopharmaceutical amplified in the case of continuous ture of the past five years on implement-
manufacturing processes. FDA in 2004 manufacturing processes, where in- ing spectroscopy-based analytical tools
described a regulatory framework for dividual batches are not well-defined in downstream bioprocessing (13–29).
PAT and has been urging the industry and critical quality attributes (CQAs) The use of spectroscopic immersion
to voluntarily develop innovative tools of the therapeutic must be consistently probes as PAT tools in upstream mi-
and techniques for quality control maintained, monitored, and controlled crobial and mammalian bioreactors
and assurance in biopharmaceutical over months of operation, rather than has been extensively reported in the
manufacturing unit operations (1). The simply tested at the end of each unit recent literature for identification and
quality-by-design (QbD) framework operation or each batch (3–4). The quantification of proteins, by-products,
encourages the use of PAT tools to need for real-time monitoring and and substrates (30–32). However, their
“build quality into the process” rather control of CQAs and critical process use in downstream processing is a more
parameters (CPPs) in continuous man- recent development. Near infrared
Garima Thakur is a graduate student in
the Department of Chemical Engineering
ufacturing has led to the adaptation spectroscopy (NIRS), in particular, has
at the Indian Institute of Technology of traditional end-of-batch quality been demonstrated to have a wide vari-
New Delhi, and Anurag S. Rathore *, testing techniques, such as analyti- ety of potential applications in different
ANCHALEE - STOCK.ADOBE.COM

[email protected], is a cal high-performance liquid chroma- downstream unit operations, including

professor in the Department of Chemical tography (HPLC) and ultraviolet-280 capture chromatography (24), protein
Engineering at the Indian Institute of
Technology New Delhi and a member
(UV-280) spectroscopy, into in-line PEGylation reactions (25), and tangen-
of BioPharm International’s Editorial at-line modalities with periodic sam- tial flow ultrafiltration (26). These three
Advisory Board. pling from continuous unit operations case studies demonstrate the versatility
To whom all correspondence should
* (5–7). It has also spurred development of NIRS as a PAT tool and showcase the
be addressed. of new analytical tools that provide in- common underlying framework of the
1 Pharmaceutical Technology MARCH 2021 P h a r mTe c h . c o m
Manufacturing
Table I. Summary of spectroscopy-based downstream process analytical technology (PAT) applications in recent literature.

Type of spectroscopy PAT monitoring/control application Reference

Chromatography pooling decisions for separation of Brestrich et al., 2015 (13)

protein size, charge, and type variants Hansen et al., 2017 (14)
Multi-wavelength ultraviolet Loading of capture chromatography Rudt et al., 2017 (15)
spectroscopy Column failure detection in capture chromatography Brestrich et al., 2016 (16)
Ghodbane et al., 2019 (17)
Monitoring of ultrafiltration and diafiltration
Rolinger et al., 2020 (18)
Grobhans et al., 2018 (19)
Chromatography pooling decisions for separation of
Fourier transform infrared Sanden et al., 2019 (20)
protein size, charge, and type variants
spectroscopy (FTIRS) Sauer et al., 2019 (21)
Monitoring of protein refolding Walther et al., 2014 (22)
Chromatography pooling decisions for separation of
Mid infrared spectroscopy (MIRS) Walch et al., 2019 (23)
protein size, charge, and type variants
Loading of capture chromatography Thakur et al., 2019 (24)
Near infrared spectroscopy (NIRS) Monitoring and control of PEGylation reactions Hebbi et al., 2020 (25)
Control of ultrafiltration and diafiltration Thakur et al., 2020 (26)
Loading of capture chromatography Feidl et al., 2019 (27)
Chromatography pooling decisions for separation of
Goldrick et al., 2020 (28)
Raman spectroscopy protein size, charge, and type variants
Titer estimation post perfusion processes for down-
Yilmaz et al., 2019 (29)
stream process adjustment

different applications, namely the col- mance (35). Over-loading the Protein the percentage breakthrough from the
lection of real-time spectra followed by A column with high-titer feed can lead column, and these were used to make
comparison with spectral calibration to the loss of expensive mAb product, control decisions to start and pause
libraries using multivariate statistical while under-loading decreases the resin the loading in a three-column periodic
techniques. NIRS is demonstrated to utilization of the Protein A resin, one counter current Protein A process on
be a reliable tool for acquiring rapid of the most expensive consumables that a continuous chromatography system
quantitative information in a range of has been reported to contribute to up (Cadence BioSMB, Pall). The normal
downstream bioprocesses, and a key to 60% of the costs of the downstream operating range of the control system
enabler for real time control. processing train (36). allowed for concentration variations in
The NIRS-based PAT strateg y the harvest between 3–8 g/L, achiev-
NIRS as a downstream PAT tool shown in Figure 1 has been used to ing optimal resin utilization as well as
Case study 1: NIRS for controlled loading. address this challenge of handling process scheduling within this range.
Protein A capture chromatography is potential titer variability in contin- The NIRS-based PAT control strat-
typically the first downstream step in uous capture chromatography while egy was tested by inducing a range of
manufacturing of monoclonal antibod- maintaining resin utilization and pre- linear and non-linear deviations in the
ies (mAb). One of the key difficulties in venting both under- and over-loading load titer in real time. The deviations
continuous Protein A chromatography (24). NIRS flowthrough cells placed were designed to closely simulate those
is handling variability in the titer of the in the loading stream and the outlet which would potentially occur during
upstream material. The product titer is of the load column were used to col- a perfusion process. The NIR flow cells
expected to change over time in the case lect spectra of the harvest and column provided the advantage of monitoring
of upstream perfusion cell culture sys- f lowthrough every three seconds. not only loading but also changes in
tems (33). Even in the case of fed-batch These spectra were passed to online the column binding capacity in real
processes, batch-to-batch variability MVDA models calibrated with ref- time, which is useful for providing early
necessitates f lexibility in the down- erence spectra to determine the con- warning of resin degradation or column
stream capture step (34). An increase centration of mAb in the harvest and quality issues and facilitating optimal
or decrease in the protein concentration flowthrough to within ± 0.05 mg/mL. loading without relying on adsorp-
in the load stream affects the dynamic The real-time concentration data were tion isotherm models, which may not
binding capacity of the Protein A resin, used to calculate both the total mg of be valid after multiple cleaning cycles.
leading to changes in process perfor- mAb loaded on the column as well as The system allowed resin utilization to
2 Pharmaceutical Technology MARCH 2021 P h a r mTe c h . c o m
 Figure 1. Near infrared spectroscopy (NIRS) as a process analytical monoPEGylated and multiPEGylated
technology tool for control of loading in continuous Protein A forms on a time scale of a few seconds
chromatography by monitoring load titer and column breakthrough (25). The NIRS spectra were acquired
percentage in real time. MVDA is multivariate data analytics. every three seconds and compared
against a calibrated MVDA regression
model, built using control runs of the
reaction with orthogonal HPLC-based
quantification of the PEGylated vari-
ants. The spectra acquired during the
control runs were also used to develop
a multivariate batch evolution model
of the ideal reaction trajectory. A flow
chart of the NIRS-based PAT tool is
shown in Figure 2.
The PAT tool was demonstrated in
various control and deviated reaction
runs in which online spectra were ac-
quired by the NIRS immersion probe
and used not only for quantification
of the PEGylated variant, but also for
comparison against the ideal reaction
trajectory. This allowed process devi-
ations to be identified and flagged in
case a statistically significant differ-
ence was found between the trajecto-
be maximized, lowering consumable stimulating factor (rh-GCSF) was con- ries of the current process versus the
costs by ensuring that the total protein sidered. The challenge was to monitor ideal one. A range of deviations were
processed per mg of resin was consis- the PEGylation process trajectory and found to be identifiable by the PAT
tently maintained. The NIRS flow cells reaction kinetics to flag potential devi- tool, including incorrect concentration
enabled online measurement of concen- ations in real time, as well as to control of chemical additives, incorrect ratio
tration and facilitated real-time control the reaction quenching to optimize the of PEG to rhGCSF, incorrect order of
decisions for increased efficiency, flex- production of the desired monoPE- addition of the reactants, and incorrect
ibility, and agility of the continuous Gylated variant before its further con- quenching time. The impact of each
chromatography process in the face of version into undesired diPEGylated and deviation on CQAs of the PEGylation
unexpected deviations. multiPEGylated forms. reaction product was characterized,
Case study 2: NIRS for monitoring/control. A few different PAT tools have been and each resulted in lower product pu-
Protein modification with biocompati- explored in the literature for monitor- rity and process yield. The NIRS probe
ble polymers, such as polyethylene gly- ing and control of PEGylation. Some was demonstrated to provide critical
col (PEG), is often used to improve the researchers used at-line size exclusion real-time information and form the
pharmacological properties and stability chromatography to monitor the prog- basis of a robust PAT tool for monitor-
of biotherapeutics (37). Manufacturing ress of the PEGylation reaction, though ing and control of the PEGylation reac-
of PEGylated drugs requires a PEG con- this was not suited for real-time con- tion of rhGCSF. The overall approach
jugation reaction to be carried out on the trol due to the long method run times is generalizable to other downstream
purified drug substance at the end of the of >30 minutes for an overall reaction reaction steps, such as enzyme reac-
downstream train (38). Proper control time of one to two hours (40). Other tions or esterification.
of the PEGylation reaction is critical to researchers used at-line matrix assisted Case study 3: NIRS for control of retentate
maximize PEGylation efficiency while laser desorption ionization–time of concentration. Tangential f low ultra-
minimizing the presence of over-PE- flight (MALDI–TOF) mass spectrom- filtration (UF) is a key unit operation
Gylated variants. Various critical pro- etry for tracking the kinetics of PEGyla- in downstream processing of biother-
cess parameters affect PEGylation qual- tion, though the time scale of analysis apeutics, used for concentrating and
ity, including pH, reaction time, and the was again too long at over one hour (41). buffer exchanging the in-process drug
concentration and order of addition of The use of NIRS as a PAT tool in the substance into the desired target con-
the reactants (39). In the present case present case study overcame the bottle- centration and formulation of the final
study, the PEGylation reaction of re- neck of analysis time and was able to drug product (42). This is typically the
combinant human granulocyte colony track the conversion of rhGCSF into its last unit operation prior to fill/finish
Pharmaceutical Technology MARCH 2021 3
Manufacturing
and packaging. The concentration of
Figure 2. Near infrared spectroscopy (NIRS) as a process analytical technology the biotherapeutic in the final drug
tool in PEGylation reaction of recombinant human granulocyte colony product is a CQA determined solely by
stimulating factor (rhGCSF) by monitoring concentrations of mono- and di- this unit operation, necessitating robust
PEGylated species over time and tracking deviations in the process trajectory. control. In batch mode, concentration is
achieved by recirculating the drug sub-
stance held in a large tank through the
UF membrane until the desired volume
reduction is achieved (43). However,
batch-mode recirculation is not possi-
ble in the case of constantly incoming
flow streams in continuous processing.
The solution is to use single-pass UF
in which the membrane module has
a larger area and a long flow path, fa-
cilitating volume reduction of the feed
stream in a single pass without the need
for recirculation (44).
There are several operational chal-
lenges that must be overcome, how-
ever, to ensure that the retentate stream
emerges from the single-pass tangential
flow filtration (SPTFF) module consis-
tently at the fixed target concentration.
Reversible and irreversible membrane
fouling as well as deviations in the con-
Figure 3. Near infrared spectroscopy (NIRS) as a process analytical centration of the incoming flow stream
technology tool for control of retentate concentration in single pass from prior unit operations, such as pol-
tangential flow ultrafiltration in continuous formulation of monoclonal ishing chromatography, lead to changes
antibodies (mAbs). UF is ultrafiltration. in the concentration factor achieved
in a single pass (45,46). Therefore, a
PAT strategy is needed to monitor the
concentration of the incoming feed
stream and control the flux across the
membrane, to ensure that the reten-
tate concentration does not vary over
the course of the process. Researchers
demonstrated a strategy leveraging in-
line NIRS flow cells in the feed and re-
tentate streams of an SPTFF module to
make flux-based control decisions and
ensure consistency in the retentate con-
centration, as shown in the schematic
in Figure 3 (26).
The NIRS flow cells were able to mea-
sure the concentration of mAb in the
range of 0.5–200 g/L, a significant im-
provement over UV-based quantifica-
tion methods, using suitably calibrated
spectral libraries. The control decisions
were made on the basis of the real-time
NIRS data as well as a pre-charac-
terized design space for the limits of
maximum f lowrates and concentra-
tion factors achievable for a given feed
4 Pharmaceutical Technology MARCH 2021 P h a r mTe c h . c o m
concentration and flow rate within the or years. More fundamental studies are 19. S. Großhans, et al., Journal of Chromatog-
raphy A 1547, 37–44 (2018).
pressure limits of the membrane mod- also needed to draw clear relationships
20.A. Sanden, et al., Journal of Chromatogra-
ule. The NIRS data were analyzed, and between biotherapeutic CQAs and their phy A 1608, 460410 (2019).
closed-loop control established with effect on spectral features, as has been 21. D.G. Sauer, et al., Biotechnology and Bioen-
a permeate pump and retentate valve, done in the case of generic pharmaceu- gineering 116 (8) 1999–2009 (2019).
allowing flux control on the time scale ticals. Finally, the ability of spectroscopic 22.C. Walther, Process Biochemistry 49 (7)
1113–1121 (2014).
of <1 second. Control and scheduling sensors to collect time-stamped spectral 23. N. Walch, et al., Biotechnology Journal 14,
decisions were made for integrating process signatures at high frequency over 1800521 (2019).
the SPTFF step with the rest of the months or years is also a potential advan- 24.G. Thakur, V. Hebbi, and A. S. Rathore,
continuous downstream train and for tage that can be leveraged to create large- Biotechnology and Bioengineering, 117 (3)
673-686 (2020).
ensuring timely cleaning. The overall scale manufacturing datasets providing 25.V. Hebbi, G. Thakur, and A. S. Rathore,
PAT strategy was demonstrated over deep process history information, which Journal of Biotechnology 325 (10) 303–311
two 12-hour case studies. The NIRS can be of use to regulatory bodies. (2021).
monitoring sensors enabled the devel- 26.G. Thakur, S. Thori, and A. S. Rathore,
Journal of Membrane Science, 613 (1)
opment of a robust PAT control strategy References 118492 (2020).
to ensure that retentate concentration 1. FDA, Guidance for Industry, Q8 (R2) Phar-
27. F. Feidl, et al., Biotechnol. Prog. 35 (5) e2847
maceutical Development (CDER, CBER,
targets were consistently met over long November 2009). (2019).
continuous campaigns. 2. A.S. Rathore and H. Winkle, Nat Biotech- 28. S. Goldrick, et al., Processes 8 (9) 1179 (2020).
nol 27 (1) 26–34 (2009). 29. D. Yilmaz, et al., Biotechnol Progress 36 (3)
3. J.Glassey, et al., Biotechnology Journal 6 (4) e2947 (2020).
Summary 369–377 (2011). 30. S.M. Mercier, et al., Eng. Life Sci. 16 (1)
The need for rapid, non-degenerative, 4. A.S. Rathore, R. Bhambure, and V. Ghare, 25–35 (2015).
and robust analytical tools will con- Anal Bioanal Chem 398 (1) 137–154 (2010). 31. N.R. Abu-Absi, et al., Pharmaceutical Bio-
5. A. Tiwari, et al., Anal Chem 90 (13) 7824– processing 2 (3) 267–284 (2014).
tinue to grow as the biotherapeutic
7829 (2018). 32. J. Claßen, et al., Anal. Bioanal. Chem. 409
industry embraces the PAT and QbD 6. S. Huffman, K. Soni, and J. Ferraiolo, Bio- (3) 651–666 (2017).
paradigms. Spectroscopic sensors are process Int 12 (8) 66–72 (2014). 33. L. Chemmalil, et al., Biotechnology and Bio-
uniquely suited to fulfill this need as 7. M. Jiang, et al., Biotechnol Bioeng 114 (11) engineering 17 (12) 3757–3765 (2020).
2445–2456 (2017). 34.F. Li, A. Shen, and A. Amanullah, Pharma-
spectral information can be collected in 8. K. A. Bakeev, Process Analytical Technol- ceutical Sciences Encyclopedia: Drug Dis-
a few seconds and analyzed in millisec- ogy: Spectroscopic Tools and Implementa- covery, Development, and Manufacturing
onds using multivariate techniques. The tion Strategies for the Chemical and Phar- 2, 1–38 (2010).
maceutical Industries (John Wiley & Sons,
spectra can yield a wealth of information West Sussex, UK, 2nd Ed. 2010).
35. S. Hober, K. Nord, M. Linhult, Journal of
about the in-process sample, including Chromatography B 848 (1) 40–47 (2007).
9. M. Rüdt, T. Briskot, and J. Hubbuch, Jour-
nal of Chromatography A 1490, 2–9 (2017). 36.A. Rathore, et al., “Re-use of Protein A
quantification of multiple analytes and Resin: Fouling and Economics,” BioPharm
10. R.W Kessler, W. Kessler, and E. Zikul-
the overall health of the process com- nig-Rusch, Chem Ing Tech 88 (6) 710–721 International 28 (3) 2015.
pared to an ideal trajectory. Spectro- (2016). 37. D. Pfister and M. Morbidelli, Journal of
scopic applications in upstream micro- 11. J. ClaSSen, et al., Anal Bioanal Chem 409 Controlled Release 180, 134–149 (2014).
(3) 651–666 (2017). 38. S. Jevsevar, M. Kunstelj, and V.G. Porekar,
bial and mammalian cell cultures have Biotechnol. J. 5 (1) 113–128 (2010).
12. L. Eriksson, et al., Multi- and Megavariate
grown exponentially in the past 10 years, Data Analysis: Basic Principles and Applica- 39. R. Shekhawat, et al., PLoS ONE 14 (3)
and this can be expected to happen in tions (MKS Umetrics AB, Malmö, Sweden, e0212622 (2019).
the case of downstream unit operations 2013). 40. J. Behi, et al., International Journal of Bio-
13. N. Brestrich, et al., Biotechnology and Bio- logical Macromolecules 109, 888–895 (2018).
in the near future. The rapidly growing engineering 112 (7) 1406–1416 (2015). 41. D.H. Na, Y.S. Youn, and K.C. Lee, Rapid
interest in continuous downstream pro- 14. S. Hansen, et al., “Mid-UV Protein Ab- Commun. Mass Spectrom. 17 (19) 2241–
cesses in both academia and industry sorption Spectra and Partial Least Squares 2244 (2003).
Regression as Screening and PAT Tool,” in 42. H. Lutz, Ultrafiltration for Bioprocessing,
will amplify this effect due to the unique Preparative Chromatography for Separation
real-time PAT control challenges that (Woodhead Publishing, Cambridge, UK,
of Proteins, A. Staby, A.S. Rathore, and S. 2015).
arise when batch-mode product quality Ahuja, Eds. (John Wiley & Sons, 2017), pp.
43. L.J. Zeman, and A.L. Zydney, Microfiltra-
testing is no longer an option. To truly 501-536.
15. M. Rüdt, et al., Biotechnology and Bioengi- tion and Ultrafiltration: Principles and
bring spectroscopic applications into neering 114 (2) 368–373 (2017). Applications (CRC Press, Boca Raton, FL,
a manufacturing setting, however, the 16. N. Brestrich, T. Hahn, and J. Hubbuch, 2017).
Journal of Chromatography A 1437, 158–167 44. E. Thomas, et al., “Application of Single
industry needs to gain confidence in the Pass TFF to Enable Intensified and Con-
(2016).
use of multivariate calibration models, 17. M. Ghodbane, et al., “Analytical and Data tinuous Biological Manufacturing,” in In-
which can only be done by having robust Strategy for Continuous Downstream tegrated Continuous Biomanufacturing III
guidelines and frameworks for spectro- Manufacturing,” in Integrated Continu- (ECI Symposium Series, 2017) pp. 1.
ous Biomanufacturing IV, V. Warikoo, et 45. C. Casey, et al., Journal of Membrane Sci-
scopic model calibration, validation, and al., Eds, ECI Symposium Series, 2019. ence 384, 82–88 (2011).
periodic checking to account for shifts in 18. L. Rolinger, Analytical and Bioanalytical 46.A. Arunkumar, et al., Journal of Membrane
spectroscopic sensor data over months Chemistry 412 (9) 2123–2136 (2020). Science 524, 20–32 (2017). PT

Pharmaceutical Technology MARCH 2021 5