Results and Discussion
Surfactant structure influences recovery yield of IgG. In an
attempt to reveal any significant relationship between IgG purity/yield
and surfactant structure, we studied several different non-ionic
surfactant families (Figure 1, B) i.e., Tween, Brij, Triton and
Pluronic (triblock-copolymer) detergents. While a hydrophobic anchor in
the form of a hydrocarbon chain of varying length is present in Tween,
Brij and Triton surfactants it is not present in the same form in
Pluronic F-127 triblock copolymer surfactants. Rather the polypropylene
central block forms the micellar core, while the polyethylene oxide
blocks form the hydrophilic shell (Figure 1, B). The number of ethylene
glycol moieties per detergent monomer did not affect the purity of
recovered IgG’s (Figure 2, C-D lanes 2-10) but did influence overall
yield. The most efficient detergents in each family, from the point of
view of overall yield, were found to be: Tween-20 (hIgG- 75%, mouse
IgG- 74%); Brij-O20 (hIgG-69%, mouse IgG- 70%); Brij-100 (hIgG-73%,
mouse IgG- 70%) and Triton X-100 (hIgG-72%, mouse IgG -69%). Pluronic
F-127 was the least efficient (hIgG -42%, mouse IgG -53%) (Figure 2,
C-D). This trend was also observed when analyzing pellet composition
immediately following the IgG capturing step, though the presence of
PEG-6000 in the system, led to diffused migration of the heavy and light
chains that made accurate quantification unreliable (Figure 2, A-B lanes
2-10). Here, Tween-20 with the shortest alkyl tail (12 carbons) produced
higher overall yields in comparison to other Tween family members with
longer alkyl tails (16-18 carbons) (Figure 2, C-D, lanes 2-5). Moreover,
results from the Brij family clearly indicated that overall yields
increase with increasing number of oxyethylene units. Accordingly,
Brij-10 with ~10 oxyethylene units was less efficient
than Brij-O20 or Brij-100 containing ~20 or
~100 oxyethylene units, respectively (Figure 2, C-D,
lanes 6-8). It seems therefore that process efficiency is determined by
several parameters that include both the length of the detergent
hydrocarbon tail and the number of oxyethylene groups.
Influence of aggregate size. An additional parameter that may
influence total yield is aggregate size. We recently reported that
aggregates with size ranging between 100-200 nm do not efficiently
capture IgG. 22 It was
therefore surprising to find that DLS characterized Pluronic F-127
aggregates as having the largest size (1312 nm) of all the detergents we
studied: i.e., Tween (355 - 685 nm), Brij (635 - 683 nm) , Triton
X-100 (913 nm) (Figure 3, A-D) . Thus, the larger size of Pluronic F-127
aggregates apparently could not compensate for the block copolymer
core/shell micelle structure.
Filtration. In order to be considered for industrial-scale,
continuous production flows of therapeutic
mAbs12, IgG capturing
and extraction from the aggregates would preferably be accomplishedvia filtration rather than centrifugation. We therefore replaced
centrifugation steps with filtration at three critical stages in our IgG
purification protocol involving Tween-20, Brij-O20 and Triton X-100
surfactants (Figure 4): (i) aggregate incubation with the IgG/BSA
mixture in hybridoma serum free media; (ii) removing residual unbound
IgG and/or BSA from the aggregates via 20 mM cold NaCl wash;
(iii) recovering IgG with extraction buffer: 50 mM Gly in 30 mM NaCl at
pH 3.8 during 5 minutes at room temperature. The results clearly
indicated that filtration can indeed replace centrifugation. The purity
of recovered hIgG (Figure 4, A lanes 2-7) was not different from that
obtained with centrifugation (Figure 2, C-D lanes 2, 7, 9) while the
overall yields improved and reached values ranging between 80-90% after
two consecutive extraction steps (Figure 4, A). Of the three detergents
tested, Triton X-100 repeatedly exhibited superiority over Tween-20 and
Brij-O20 (Figure 4, A). Thus, a relatively pure Ab preparation can be
conveniently and efficiently obtained via filtration using
detergent aggregates belonging to three surfactant families. This
finding emphasizes the wide scope of working conditions that can be
employed with our purification platform.
Since pharmaceutical use of therapeutic mAbs requires these to be
monomeric, 31 recovered
hIgG obtained with the filtration protocol was analyzed by dynamic light
scattering (DLS). The results indicated that purified hIgG is indeed
monomeric, as no significant difference in particle size was observed
when compared to the control, pure hIgG which had not been subjected to
the surfactant aggregate protocol (Figure 4, B). Particle sizes were in
agreement with our previous findings32 and ranged between:
10.4-11.0 nm regardless of the detergent used (i.e. Tween-20,
Brij-O20, Triton X-100).
Ab specificity. Though IgG purity, overall recovery yields and
monomeric state are factors of paramount importance in the development
of new Ab purification methodologies, preservation of Ab specificity
must also be validated. This question was answered for a specific Ab (an
anti-BSA IgG) using the detergent aggregate platform and the filtration
protocol. Accordingly, the recovered anti-BSA was subjected to an ELISA
assay in order to assess its capability in binding target (BSA).
Positive signals in the ELISA assay were observed with all three
aggregates (Figure 4, C) and reflected recovery yields similar to those
obtained with human or mouse IgG’s using identical detergent aggregates
(Figure 4, A). Preservation of IgG specificity is not surprising when
considering the presence of at least ≥14 disulfide bonds in all IgG
isotypes. 33 The fact
that Tween-20, Brij-O20, Brij-100 and Triton X-100 repeatedly
demonstrated their utility with polyclonal human and mouse IgG’s
suggests that the IgG capturing step is independent of antibody amino
acid sequence. This is an important finding since it implies that a
single nonionic detergent may provide a purification platform for many
different antibodies, regardless of their biological origins. Such a
capability is ideal for industrial downstream processing of therapeutic
grade mAbs.
Increasing IgG concentration. The fact that mAb production in CHO
cells has improved significantly and currently can reach Ab titers
greater than 1 mg/mL 34led us to test our filtration protocol at IgG concentrations up to 5
mg/ml (Figure 5). As before, we focused on the three detergents that led
to the highest recovery yields via the centrifugation protocol.
The results presented in Figure 5 indicate that process efficiency at 3
mg/mL and 5 mg/mL decreases by 5-10% relative to that obtained at 1
mg/mL. Clearly, working at higher IgG concentrations requires a larger
number of aggregated micelles. Indeed, good overall yields
(~80%), necessitated an increase by a factor of 1.6 and
2.3 (wt/vol) for 3 mg/ml and 5 mg/ml hIgG, respectively. These results
are encouraging as they are obtained at current industrial IgG
expression levels. 34
Chelator recycling. Recycling of raw material represents an
important environmental and economic concern for large-scale
pharmaceutical production. We therefore studied the possibility of
regenerating the most costly component in our platform, i.e. the
chelator (batho). Chelator recycling via crystallization appears
to be both economical and realistic due to the inherent physiochemical
properties of batho. These include: water-insolubility;35 stability under both
acidic 36 and basic
conditions, and at elevated temperatures;37 and the ability of
batho to crystallize rapidly in aqueous media due to its relatively
rigid and extended aromatic system which is capable of formingπ-π interactions and thereby , aromatic stacking between batho
molecules.
Tween-20 was the surfactant of choice for this recycling experiment.
Initial removal of Tween-20 from the aggregates was essential for
optimal batho recycling, as the presence of any detergent would likely
improve batho’s water-solubility and hence, suppress its crystallization
in aqueous media. However, Tween-20 was found to be soluble even in 1 M
NaCl (data not shown). Therefore, Tween-20 aggregates that were used for
IgG capture and extraction were thoroughly washed with 0.25 M NaCl to
remove the detergent as well as the IgG’s that might not have been
extracted from the aggregates. Under these conditions, the extremely
hydrophobic [(batho)3:Fe2+] red
complex is obviously water-insoluble (Figure 5, Step I). The pellet size
decreased after NaCl washing and the resulting, excluded aqueous
solution was colorless, indicating no significant loss of the
[(batho)3:Fe2+] red complex (data
not shown). Light microscopy analysis of washed aggregates showed that
these are composed of numerous red crystals (~30-70 µm)
(Figure 7, A) containing the
[(batho)3:Fe2+] complex as
determined by mass spectrometry (MS) (Supplementary S1). In order to
dissociate the red complex, crystals were briefly (2-3 min.) heated at
95 °C in DDW/MeOH (1:1 vol/vol) containing 10 mM EDTA. Upon cooling, the
red color was no longer observed, being replaced by a colorless,
transparent solution (not shown). Thus, the loss of the red color served
as a convenient internal indicator for monitoring the rate and
efficiency at which the
[(batho)3:Fe2+] red complex
disintegrates in the presence of the competing EDTA chelator. It should
be emphasized that the presence of 50% MeOH in the aqueous medium is
essential as it led to the complete dissolution of the red crystals and
thus, physical access of EDTA to the chelated iron ions in the
[(batho)3:Fe2+] red complex. In
order to then completely remove MeOH, which suppresses batho
crystallization , the solution was heated to 95 °C for 30 minutes and
the resulting precipitate was thoroughly washed with DDW to remove EDTA,
the [EDTA-Fe2+] colorless complex and free
Fe2+ ions if present (Figure 5, Step IV). Analysis of
the precipitate by light microscopy showed that it contains colorless,
plate-like crystals (Figure 7, B) while liquid chromatography
mass-spectrometry (LC-MS) analysis confirmed that the crystals are
indeed >98% pure batho (Supplementary S2, C) identical to
, or perhaps even purer than commercial batho crystals prior to use
(Supplementary S2, D) (Figure 7, C). The fact that the MS spectrum of
recycled batho (Supplementary S2, C) was essentially identical to that
of the pure, commercial batho (Supplementary S2, A) implies that, no
chemical modifications occurred during chelator regeneration. To further
rule out possible chemical modifications of the chelator during the
recycling procedure, regenerated batho crystals were dissolved in MeOH
and Fe2+ ions were added to generate the
[(batho)3:Fe2+] red complex. The
characteristic absorbance of the latter (λmax = 530 nm)
was compared to the absorption spectrum of a freshly prepared
[(batho)3:Fe2+] red complex
generated from the commercial (unused) chelator and no marked
differences were observed (not shown). This finding provided additional
supporting evidence for the preservation of the chelating ability of
recycled crystalline batho. The characteristic absorption of the complex
at 530 nm served as a means of quantitating chelator regeneration
efficiency which was found to be, on average, 95% when tested in10 independent experiments (not shown). It should be emphasized
that this value represents the recycling yield relative to the total
amount of batho present initially in the system. These findings agree
with the physiochemical properties of batho described above and show
that near quantitative recycling can be achieved.