Available online at www.sciencedirect.

com

Protein Expression and Purification 59 (2008) 55–63

www.elsevier.com/locate/yprep

A step-wise approach significantly enhances protein

yield of a rationally-designed agonist antibody fragment in E. coli
Bing Lin, Mark W. Renshaw, Kathleen Autote, Lynette M. Smith, Peter Calveley,
Katherine S. Bowdish, Shana Frederickson *
Alexion Antibody Technologies, Inc., 3985 Sorrento Valley Boulevard, Suite A, San Diego, CA 92121, USA

Received 7 August 2007, and in revised form 21 December 2007

Available online 17 January 2008

Abstract

Fab59 is a rationally-designed antibody fragment (Fab) that mimics the activity of the cytokine thrombopoietin (TPO). Fab59 activity
was initially detected directly from bacterial supernatants in a cell-based assay and was subsequently estimated to be equipotent to TPO
using purified material. However, the expression of Fab59 was insufficient to support in vivo characterization of the Fab due to extremely
low expression levels from its initial phage display expression vector. To boost expression, a new expression vector was designed and
constructed, and Fab59 light chain codons were optimized for bacterial expression. However, from this a new challenge arose, in that
the codon-optimized Fab59 was more toxic to Escherichia coli cells than parental Fab59. Co-expression of the bacterial chaperon protein
Skp alleviated this toxicity. A two-step purification method was used to isolate monomeric Fab59 from the periplasm. Although Fab59
was prone to form aggregates during the purification process, buffer modification efficiently eliminated this problem. Overall, optimiza-
tion of Fab59 expression and purification achieved a 100-fold increase in Fab59 production in E. coli relative to the starting yield. The
yield of purified monomeric Fab59 from a shake flask reached up to 3.5 mg/L, which was sufficient to support testing of the agonist
activity of purified monomeric Fab59 in vivo. Even higher yields may be achieved by purification of Fab present in the culture media,
as Skp most significantly increased accumulation of Fab59 in that location.
Ó 2008 Elsevier Inc. All rights reserved.

Keywords: Antibody; Agonist

Fab59 is a rationally-designed human antibody frag-
ment that was created to mimic thrombopoietin (TPO)1
and provide a possible mode of therapeutic intervention
for thrombocytopenia, while avoiding the risk of generat-
ing neutralizing antibodies to endogenous TPO. Fab59
was engineered by grafting a peptide with cMpl receptor
(cMpl-R) binding capability [1] into the CDR regions of
both the heavy and light chains of a well-characterized fully
human anti-tetanus toxoid Fab scaffold (Fab TT) [2]. Ini-
tial cell-based assays using unpurified Fab in bacterial
*
Corresponding author.
E-mail address: sfrederickson@alxnsd.com (S. Frederickson).
1
Abbreviations used: TPO, thrombopoietin; cMpl-R, cMpl receptor;
Fab TT, anti-tetanus toxoid Fab; RBS, ribosomal binding site; BCA,
bicinchoninic acid; TBS, tris-buffered saline.
supernatants or Fab purified in small scale demonstrated
that Fab59 had activity equivalent to TPO in vitro. How-
ever, early attempts to produce and purify Fab59 in larger
scale were hampered by poor expression levels in Esche-
richia coli. Initial yields of purified Fab59 protein were
insufficient to support in vivo modeling of Fab59 activity,
thereby hindering both further characterization and its
potential as a commercial drug candidate. In order to meet
the protein level requirements for in vivo modeling and
increase viability as a potential commercial therapeutic
candidate, production levels of Fab59 needed to be dra-
matically increased. Efficient expression of well-folded,
functional antibody fragments can be a challenge in com-
parison to other recombinant proteins expressed in
E. coli [3]. Disulfide bond formation and assembly of Fab
light and heavy chains generally require the oxidized envi-

1046-5928/$ - see front matter Ó 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.pep.2008.01.002
56 B. Lin et al. / Protein Expression and Purification 59 (2008) 55–63

ronment found in the periplasmic space of E. coli, although PCR amplification. The HA-epitope tag and the His puri-
some Fabs can be expressed in the cytoplasm of specialized fication tag were arranged so that they would be expressed
E. coli strains [4]. Furthermore, use of a rational design on the C-terminal end of the Fab’s heavy chain.
strategy, increasingly used for new therapeutic approaches, For sub-cloning of the Fab genes into the new vectors,
adds a new layer of complexity. While CDR regions nor- 1.5 kb cDNA encoding the Fab light chain and heavy chain
mally encompass a higher degree of diversity than the was released from the vector pRL4 by double digestion
remaining antibody sequence, and thus may be a more with EcoRI/NheI, and then ligated into the expression vec-
accepting region for peptide grafts, expression and proper tor pAEV1 pre-digested with EcoRI/NheI. For construc-
folding of the engineered protein may be compromised. tion of pAEV1-HCG, a 670 bp light chain fragment was
Nonetheless, expression in E. coli still provides a conve- recovered from pRL4-TT by double digestion with SacI/
nient, fast and cost-efficient way to generate material for XbaI, and then ligated into the pAEV1-59 double digested
research and therapeutic purposes. with the same restriction enzymes. For the construction
Like other recombinant proteins, which are difficult to pAEV1-LCG, an 870 bp heavy chain fragment was recov-
express, the poor expression of Fab59 could be due to ered from pRL4-TT by double digestion with XhoI/NheI,
improper folding, insolubility, intracellular degradation and then ligated into pAEV1-59 vector double digested
or toxicity to the E. coli cells [5,6]. We therefore initiated with the same restriction enzymes. These constructs are
a step-wise process to first identify which factors were listed in Fig. 1.
rate-limiting to Fab59 production in E. coli and then deter-
mine how to modify them accordingly to effect an increase Codon optimization
in protein accumulation and yield. We found that poor
expression of Fab59 was specific to the light chain, due The codon-optimized synthetic light chain gene was
to both extremely low LC expression levels and insolubility designed based on the codon preference of highly expressed
of the light chain. Therefore, our expression optimization genes in E. coli [7]. A synthetic light chain was assembled
of Fab59 was mainly focused on increasing the light chain from 52 oligonucleotides (35–45 bases each) with 15–20
expression level and improving solubility. This was system- overlapping complementarily bases (Operon Inc.). Oligos
atically accomplished by first codon optimization, second, were phosphorylated, annealed and ligated into a full-
using a tightly controlled strong promoter, third low tem- length fragment. The full-length synthetic light chain was
perature and low levels of IPTG induction, and fourth then PCR amplified. The PCR products were digested by
co-expression of a bacterial chaperon protein Skp. Further- XbaI/NotI and ligated back to pAEV1-59 pre-digested
more, the purification process was optimized to efficiently with XbaI/NotI.
avoid protein aggregation. After optimization of expres-
sion and purification, the yield of Fab59 was increased Bacterial chaperon protein Skp cloning
approximately 100-fold. This level was sufficient to allow
for the successful purification and isolation of enough The PCR primer pairs used for PCR amplifying the Skp
Fab59 to support preclinical development. Highly pure gene from bacterial genomic DNA were designed based on
monomeric and correctly folded Fab59 with intact disulfide
bonds was isolated from bacterial periplasmic extracts TPO TPO
pRL4-59 P lac Omp LC Pel HC
using a two-step purification method. The agonist activity
TPO TPO
of the purified monomeric Fab59 was then demonstrated Omp LC Pel HC
pAEV1-59 P T7
to be equipotent to rhTPO in vitro in cMpl-R signaling,
luciferase reporter-based assays, and furthermore stimu- P T7 Omp LC Pel HC
pAEV1-TT
TPO
lated increased platelet levels in vivo [2]. Thus, this broadly Omp LC Pel HC
pAEV1-LCG P T7
applicable systematic approach allowed us to overcome the TPO
existing bottlenecks in Fab59 protein production in E. coli pAEV1-HCG P T7
Omp LC Pel HC
TPO TPO
that occurred in our initial expression system, as well as to Omp LC Pel HC
pAEV1-O59 P T7
solve new problems that developed as a result of increased TPO TPO
accumulated protein levels. pAEV1-O59Skp P T7 Omp LC Pel HC Skp

Fig. 1. Schematic diagram of the expression cassettes and constructs

Materials and methods
Expression vector construction
T7-based expression vector pAEV1 was developed by
modifying the pET28a vector (Novagen) so that lacI was
replaced by lacIq. In addition a ribosomal binding site
(RBS) and poly-linker in pET28a were replaced by the
OmpA RBS and an EcoRI/NheI/HA-His/SacII linker by
evaluated in our studies. Omp: OmpA leader sequence. Pel: PelB leader
sequence. LC: light chain (h). HC: heavy chain ( ). TPO :TPO mimetic
peptide with selected flanking residues (j). Skp: E. coli 17 KDa
periplasmic chaperon protein Skp ( ). Plac: lac promoter. PT7: T7/lac
promoter. pAEV1-O59: codon-optimized light chain ( ) Fab59 con-
structs. pAEV1-TT: fully human Fab TT. In all cases where TPO is shown
the TPO mimetic peptide cassette was grafted into either the CDR3 region
in the heavy chain (pAEV1-HCG) or into the CDR2 region in the light
chain (pAEV1-LCG) or both in the heavy chain and in the light chain
(Fab59 constructs).
 B. Lin et al. / Protein Expression and Purification 59 (2008) 55–63
the bacterial Skp DNA sequence in GenBank (GenBank
Accession No. M21118). The 50 primer was flanked with
NdeI: gaaacatatgaaaaagtggttattagctgcag, and the 30 primer
was flanked with XhoI: gaaactcgagttatttaacctgtttcagatcgtc.
To PCR amplify Skp, a freshly grown Top10F’ cell colony
on a LB plate was inoculated into a 50 ll PCR mixture (1
TaqPlus buffer, 0.1 mM dNTP, 1 pmol primers, 2.5 U Taq-
Plus, Stratagene). The PCR reaction was performed on a
Perkin Elmer PCR machine with 95 °C 5000 –62 °C 5000 –
72 °C 4000 for 25 cycles. PCR amplified products were
digested with NdeI/XhoI and ligated into the pAEV1-59
vector digested with the same enzymes. Skp DNA sequence
was confirmed by sequencing.
Small-scale Fab expression
Overnight starter cultures were made by inoculating sin-
gle colonies from a LB plate into 2 ml cultures of SB med-
ium with the appropriate antibiotics and 2% glucose in a
37 °C shaker. One hundred microliters of the overnight
starter cultures was inoculated into 2 ml of SB medium
(antibiotics, 2% glucose). The bacterial cultures were incu-
bated in a 37 °C shaker until OD600 reached 0.5–0.9. Then
IPTG was added for induction and the shaker temperature
was changed to 30 °C. Four hours after induction, the bac-
terial culture medium supernatants and pellets were col-
lected for Western blot. Bacterial pellets were suspended
in 1 PBS buffer and sonicated gently on ice using a power
setting of 35%, with pauses at 10 s intervals (for three inter-
vals) in order to release most of the proteins from the bac-
terial periplasm while avoiding breaking open the cells as
monitored by release of DNA into the buffer. The condi-
tions were established in pilot experiments using side-by-
side comparisons with the conventional osmotic shock
method of releasing periplasmic proteins [8]. To isolate
total soluble proteins (cytoplasmic and periplasmic), bacte-
rial pellets in PBS buffer were sonicated heavily on ice using
a power setting of 45% and pauses at 20 s intervals for six
intervals, followed by recovery of the supernate after spin-
ning down the sonicated mixture. To isolate insoluble pro-
teins, the bacterial pellets from overnight growth were
heavily sonicated as above. After spinning down the soni-
cated mixture, the insoluble proteins were recovered from
the pellet following three PBS washes.
Overnight induction at room temperature
A single colony was inoculated into 2 ml of SB medium,
and cultured in a 37 °C shaker for OD600 to reach 0.7.
IPTG was added to a final concentration of 0.02 mM,
and the culture temperature was changed to 22 °C in a
120 rpm shaker. The induction was continued overnight
(16–20 h).
To make 1 L of bacterial culture for Fab purification,
one colony was inoculated into 100 ml SB medium and
then cultured overnight in a 30 °C shaker. The next day,
100 ml of the overnight culture was inoculated into 1-L
57
SB medium in a 3-L shaker flask. When bacterial culture
OD600 reached 0.7, IPTG was added to a final concentra-
tion of 0.02 mM and temperature was changed to 22 °C
in a 120 rpm shaker. After 16–20 h of induction, the bacte-
rial pellet was collected for Fab purification.
Fab59 purification
Bacterial pellets were suspended in 50 ml binding buffer
(1 PBS buffer, or HEPES buffer (100 mM HEPES, 0.5 M
NaCl, 10% glycerol, 0.02% Tween-80)). After gentle soni-
cation, samples were cleared by centrifugation at
15,000 rpm at 4 °C for 15 min. The supernatant was col-
lected and filtered through 0.8 lm and then 0.2 lm filtra-
tion units (NALGENEÒ, Apogene Technologies). Fab59
contained in the supernatant was then bound to an anti-
human Fab affinity column pre-equilibrated with binding
buffer on a FPLC (ÃKTA, Amersham Biosciences).
Fab59 was then eluted with elution buffer (0.2 M glycine
buffer, pH 2.2, or a buffer with 0.2 M glycine–HCl, 10%
glycerol, 0.02% Tween-80, pH 2.7), and immediately neu-
tralized to pH 7.0 with neutralization buffer (1 M Tris–
HCl, 4 M NaCl, pH 9.0). The neutralized elute was dia-
lyzed against TBS buffer (20 mM Tris–HCl, pH 7.4,
0.15 M NaCl) or HEPES buffer overnight. After dialysis,
the sample was concentrated down to approximately 1 ml
and loaded onto a Superdex75 (Amersham Biosciences)
SEC column (10 cm  110 cm) pre-equilibrated with the
dialysis buffer. Size exclusion chromatography was per-
formed on the FPLC with a flow rate of 0.4 ml/min. Mono-
meric Fab59 fractions were collected. Bacterial endotoxins
were removed from Fab59 by passing through a Sarto-
bindÒ Membrane Adsorber Q15 (Sartorius). The Fab59
sample was then concentrated using a Centriprep YM-30
(Millipore Corp.) and filter-sterilized by MinisartÒ-Plus
(Sartorius) syringe tip filter.
The Ni-NTA agarose resin used for His-tag purification
was purchased from Invitrogen, and the process for the
His-tag purification was based on the protocol provided
with the kit. Protein concentration was determined by
using the bicinchoninic acid (BCA) protein assay reagent
(Bio-Rad).
SDS–PAGE and Western blot
Protein samples were combined with SDS–PAGE load-
ing buffer (with or without DTT) and heated at 100 °C for
10 min. For analysis of protein in the culture media or the
periplasmic space, the samples were normalized to 10 ll of
the culture volume (for Figs. 2, 3A, 5 and 6). For analysis
of total cellular protein (either the soluble or insoluble frac-
tion from sonication of the bacterial pellet), the samples
were normalized by loading 10 lg of protein (Figs. 3B, C
and 4). The samples were resolved on a precast 4%–15%
Gradient Ready Gel (Bio-Rad) following manufacturer’s
protocol. Proteins were then transferred onto a nitrocellu-
lose membrane (Bio-Rad) using a Trans-Blot SD Semi-Dry
58 B. Lin et al. / Protein Expression and Purification 59 (2008) 55–63

9
9

5
-5

-59
-5

-O
1

L4
EV

1
C

EV

EV
pR
pA

pA

pA
72K
A 43K
43K 25K

M
M

M
m

1m
M

1m

0m
05
1m

0 .5

0 .0
0.

0.

Fig. 4. Immunoblotting analysis demonstrating insoluble light chain

72K
B 43K
Fig. 2. Immunoblotting analysis of Fab59 expression in E. coli periplasm.
Periplasmic samples were prepared by small-scale Fab expression as
described in Materials and methods. Both gels were run under non-
reducing SDS–PAGE conditions and blots were probed with anti-human
light chain monoclonal antibody. (A) Comparing expression levels of
Fab59 from two different expression vectors in two different hosts. Lane 1:
pAEV1-59 was expressed in BL21(DE)StarTM cells and induced with 1 mM
IPTG. Lane 2: pRL4-59 was expressed in Top10F0 cells and induced with
1 mM IPTG. Lane 3: is a control (C) of pAEV1-59 in BL21(DE)StarTM
with no IPTG induction. (B) Evaluating expression levels of the construct
pAEV1-59 in E. coli periplasm under different IPTG induction concen-
trations in BL21(DE)StarTM cells.
G
CG
accumulation in E. coli total cell lysis. Insoluble protein samples were
prepared by small-scale Fab expression as described in Materials and
methods. pAEV1-O59 contains a peptide grafted synthetic codon
optimized light chain. pAEV1-59 contains a peptide grafted wild-type
light chain. SDS–PAGE was performed under reducing conditions. The
Western blot was probed with an anti-human kappa light chain
monoclonal antibody.
Transfer Cell (Bio-Rad) according to manufacture’s proto-
col. The membranes were blocked overnight at 4 °C in 5%
skim milk powder in Tris-buffered saline (TBS). Anti-HA
monoclonal antibody (Sigma, 1:2000 dilution) was used
to detect the heavy chain and anti-human light chain poly-
clonal antibody (Pierce, 1:5000) was used for light chain
detection. After washing in TBS/0.1% Tween-20 (TBST)
three times (10 min each), the blots were probed with the
alkaline-phosphate conjugated goat anti-mouse IgG

HC

(Pierce, 1:30,000 dilution). The blots were visualized by

T
59

-L

-T
1-
1-

using NBT–BCIP reagents (Pierce) following the provided

1
EV
EV
EV

EV
pA
pA

pA
pA

72K
A
43K
43K
B 25K
43K
C 25K
Fig. 3. Comparison of soluble Fab, light chain and heavy chain
accumulation for Fab59 versus singly grafted Fabs pAEV1-LCG and
pAEV1-HCG, and parental TT in E. coli. Protein samples were prepared
by small-scale Fab expression as described in Materials and methods. (A)
Fab expression for different constructs in E. coli periplasm. A non-
reducing gel was probed with an anti-human light chain monoclonal
antibody. (B) Soluble light chain expression in different constructs in
E. coli. A reducing gel was probed with an anti-human light chain
monoclonal antibody. (C) Soluble heavy chain expression in different
constructs in E. coli. A reducing gel was probed with an anti-HA
monoclonal antibody.
protocol.
Luciferase assays
The method for detecting agonist activity of Fab59
using a luciferase assay was described previously [2].
Briefly, NIH3T3 cells were co-transfected with either a con-
trol vector (pcDNA3.1, Invitrogen, Carlsbad, CA) or the
cMpl-R expressed in pcDNA3.1 and the Fos promoter/
luciferase reporter construct [2]. Co-transfections of 3T3
cells were performed by plating NIH 3T3 cells at 3  105
cells per 6 cm dish and then transfecting the following
day. NIH 3T3 cells were transfected using the Effectene lip-
ofection reagent (Qiagen), transfecting each plate with
0.1 lg pEGFP (Clonetech), 0.2 lg of the Fos promoter/
luciferase construct, 0.2 lg of pAdVAntage, and 0.5 lg of
either the empty control vector pcDNA3.1 (Invitrogen)
or the plasmid expressing the cMpl-R. 3T3 cells were
placed in 0.5% serum 24 h post transfection and incubated
for an additional 24 h in this low serum media to serum
starve the cells in order to reduce the background activa-
tion of the Fos promoter. Fab59 samples and rhTPO
(inhouse purified) [2] were then applied to these cells for
6 h. Cells were harvested and luciferase assays performed
using 50 lg of cell lysate using a Turner Designs TD20/
20 luminometer and beetle luciferin (Promega) according
to manufacturer’s protocol.
 B. Lin et al. / Protein Expression and Purification 59 (2008) 55–63
Results
Fab59 expression is increased with the T7 promoter and
affected by IPTG induction conditions
The original expression vector containing Fab59, pRL4-
59, is a modified phage display vector with a lac promoter
(Fig. 1). To enhance transcription of Fab mRNA, we trans-
ferred the Fab genes to expression vector pAEV1, that car-
ries the T7 promoter, designated pAEV1-59. In order to
minimize the basal level leakage from the T7 promoter,
we added glucose to a final concentration of 2% in the bac-
terial culture medium. Induction, where indicated, used
1 mM IPTG. As shown in Fig. 2A, Fab59 was expressed
well in BL21(DE)StarTM cells in the construct pAEV1-59,
while Fab59 expression in the original construct pRL4-59
in Top10F’ cells was barely detected.
As demonstrated in Fig. 2B, IPTG greatly affected the
bacterial growth and recombinant protein expression in
E. coli. Low IPTG (0.1–0.5 mM) induction has been used
to improve Fab expression in several previously reported
cases [9,10]. Here, we evaluated the expression level of
pAEV1-59 with differing IPTG induction concentrations,
using 0.01, 0.05, 0.1, 0.5, and 1 mM IPTG. We found that
IPTG induction concentrations higher than 0.1 mM inhib-
ited the bacterial growth and resulted in lower expression
levels while lower concentrations of IPTG had little affect
on bacterial growth, and induction at 0.05 mM resulted
in a higher expression level of Fab59 as detected by Wes-
tern blot (Fig. 2B). At a 0.05 mM IPTG induction, we
obtained the highest expression level both in shaking cul-
tures and in bioreactor fermentation (data not shown),
demonstrating that IPTG induction concentration greatly
affected Fab59 expression.
Reduced expression of Fab59 is light chain-specific
Fab59 is a rationally-designed human antibody frag-
ment that was engineered by grafting a peptide with cMpl
receptor (cMpl-R) binding capability [1] into the CDR
regions of both the heavy and light chains of a well-charac-
terized fully human anti-tetanus toxoid Fab scaffold (Fab
TT). Since expression of the original Fab scaffold, Fab
TT, was strong compared to Fab59, we investigated
whether the drop in expression was due to grafting the pep-
tide into the CDR of the light chain, the heavy chain, or
both. To accomplish this, we switched light chains between
Fab59 and parental Fab TT which resulted in two addi-
tional constructs in which only one chain contained a
grafted peptide, pAEV1-LCG (light chain grafted) and
pAEV1-HCG (heavy chain grafted). The expression levels
of the constructs, pAEV1-TT, pAEV1-HCG, pAEV1-
LCG, and pAEV1-59, were compared side-by-side. As
shown by Western blot in Fig. 3A, both human Fab TT
and heavy chain grafted Fab HCG expressed well in
E. coli. However, light chain grafted Fab LCG and
Fab59 expression levels were reduced substantially. A Wes-
59
tern blot (Fig. 3) of the reducing SDS–PAGE gel was used
to detect the total soluble light chain (Fig. 3B) and heavy
chain (Fig. 3C) expression levels in E. coli. As shown in
Fig. 3C, there were no significant differences in heavy chain
expression levels in all the constructs, while the expression
levels of grafted light chain were very low in both con-
structs pAEV1-LCG and pAEV1-59, demonstrating that
poor expression of Fab59 resulted from reduced produc-
tion of the grafted light chain in E. coli.
Codon optimization of the Fab59 light chain increased
insoluble light chain accumulation in E. coli
Codon optimization has been widely used to achieve
higher expression levels in heterologous recombinant pro-
tein expression systems [11–13]. To determine if the optimi-
zation of codon usage to remove codons that are rarely
used in E. coli would further improve Fab59 light chain
production, we optimized the grafted light chain gene
based on the codon usage of highly expressed genes in
E. coli. The expression level of the construct with the light
chain codon optimized, pAEV1-O59, was tested by Wes-
tern blot. Surprisingly, Western blot analysis demonstrated
that the construct pAEV1-O59 did not have significantly
improved Fab expression levels (data not shown). Impor-
tantly, we noticed that the growth rate of bacteria contain-
ing pAEV1-O59 was retarded and slowed significantly after
addition of IPTG, suggesting that codon optimization of
the light chain had a growth inhibitory or toxic effect on
the E. coli cells not observed with the parental pAEV1-
Fab59. As shown in Fig. 4, pAEV1-O59 contains much
more insoluble light chain accumulation than the wild-type
pAEV1-59. These results suggested that though we success-
fully increased expression of the light chain, this led to
increased accumulation of an insoluble form of grafted
light chain which was toxic to E. coli, and created addi-
tional issues to address.
Co-expression of the bacterial chaperon protein Skp reduced
pAEV1-O59 toxicity and increased Fab59 production and
secretion into the culture medium
In order to attempt to improve solubility of Fab59 we
co-expressed the bacterial chaperon protein Skp. The Skp
gene was cloned into pAEV1-O59 as part of the same
operon as Fab59 under control of the same T7 promoter
(Fig. 1). In comparison to pAEV1-O59 without Skp, the
bacteria transformed with Skp (pAEV1-O59Skp) grew fas-
ter both before and after IPTG addition. To determine if
the growth recovery of the bacteria containing pAEV1-
O59Skp was associated with reduced accumulation of the
insoluble form of the grafted light chain, a Western blot
was performed using the insoluble portion of an induced
bacterial culture. Western blots were also performed exam-
ining Fab59 levels in the periplasm isolated from bacterial
pellets, and Fab59 secreted into culture medium superna-
tant after a 4-h induction at 30 °C. The results showed that
60 B. Lin et al. / Protein Expression and Purification 59 (2008) 55–63

p
p
sk

sk
sk
59

T
59
59

59

-T
59
1-O

-0
1-O

1-O

1
1-O

EV
B

EV
EV
EV

EV
EV

pA
pA
pA
pA

pA
pA
A 72K
72K
43K
25K
43K 43K

Fig. 6. Immunoblotting analysis of Fab expression level in E. coli

p
sk

59

periplasm. The bacteria were induced at 0.02 mM IPTG at room

59

1-O
1-O

temperature for 16 h. A Western blot from a non-reducing SDS–PAGE

EV
EV

was probed with anti-human light chain monoclonal antibody.

pA
pA

72K
lower in overnight room temperature induction, as detected
C 43K by Western blot analysis and shown in Fig. 6.

Fig. 5. Immunoblotting analysis of Fab expression enhanced by Skp.
Protein samples were prepared by small-scale Fab expression as described
in Materials and methods. (A) Soluble Fab59 expression in the periplasm
of E. coli. A Western blot from a non-reducing SDS–PAGE was probed
with anti-human light chain monoclonal antibody. (B) Insoluble light
chain accumulation in E. coli from total cell lysis. A Western blot from a
reducing SDS–PAGE gel was probed with anti-human light chain
monoclonal antibody. (C) Fab59 accumulation in culture medium. A
Western blot from a non-reducing SDS–PAGE gel was probed with anti-
human light chain monoclonal antibody.
the soluble Fab59 expression level was further increased
slightly in the periplasm by co-expression of Skp protein
(Fig. 5A). As shown in Fig. 5B, the bacteria containing
pAEV1-O59Skp accumulated somewhat less insoluble
grafted light chains than the bacteria containing pAEV1-
O59, which correlated with a reduction in Fab59 toxicity
to E. coli. Interestingly, a much more dramatic difference
was observed by Western blot of culture medium superna-
tant. As shown in Fig. 5C, significantly more fully assem-
bled and secreted Fab59 was detected from pAEV1-
O59Skp than from pAEV1-O59. The increase in Fab59
accumulation in the culture medium of pAEV1-O59Skp
may be the result of increased secretion of properly folded
Fab59 from the periplasm.
Modification of induction conditions significantly increased
Fab production in the modified expression system
Although poor expression of Fab59 was partially resolved
by all of the changes described thus far, including codon-
optimized LC expression and the presence of the Skp chap-
eron protein, further enhancements to protein expression
were investigated by re-evaluating IPTG induction condi-
tions as well as the effects of temperature in the modified
expression system. We found that 0.02 mM IPTG induction
at room temperature overnight (16–20 h) gave the highest
yield of Fab59. In comparison to wild-type pAEV1-TT,
the pAEV1-O59Skp expression level was only 1- to 2-fold
Fab purification
Despite successful efforts to increase yield through codon
optimization, introduction of bacterial Skp, and modifica-
tion of induction conditions, we still noted that nearly 40%
of the purified protein precipitated out during dialysis, and
more than 20% of the soluble Fab59 formed soluble
dimers/aggregates as shown by SEC profile (Fig. 7A),
decreasing the yield and purity of the final product.
In initial purification attempts, we used 1 PBS as a
binding buffer, 0.2 M glycine buffer, pH 2.7, as an elution
buffer, and TBS buffer as a dialysis and SEC running buffer
in a two-step purification. However, we noted that Fab59
was sensitive to the low pH elution buffer in our process.
To address this, a subsequent attempt using a His-tag Ni-
NTA-based purification protocol with neutral elution con-
ditions was evaluated. Unfortunately the neutral elution
conditions did not prevent Fab59 from aggregating.
Since it has previously been demonstrated that glycerol,
detergents and high ion strength can increase protein solu-
bility and prevent protein aggregation in some cases
[14,15], we switched to a HEPES loading buffer that con-
tained 10% glycerol, 0.02% Tween-80 and 0.5 M NaCl.
For elution, we again used 0.2 M glycine elution buffer,
pH 2.7, but importantly added 10% glycerol and 0.02%
Tween-80 in the elution buffer. The elute from the column
was immediately neutralized and the NaCl concentration
brought to 0.5 M. This strategy worked extremely well in
the purification of Fab59. No aggregation was observed
in the protein preparation during purification and dialysis,
and no detectable dimers/aggregates were observed in the
SEC profile (Fig. 7A). Importantly, no aggregation was
observed to occur during 1 month of storage at 4 °C based
on SEC analysis, and no loss in agonist activity was
observed based on our in vitro luciferase reporter assay
(data not shown).
After purification optimization, Fab59 was purified to a
monomeric form with extremely high purity. Non-reducing
SDS–PAGE gel showed that the purified monomeric Fab
 B. Lin et al. / Protein Expression and Purification 59 (2008) 55–63
A O.D.
Monomer
500
400
300
200
Dimer/aggregate
100
20 40 60 Vol (ml)
1 2 3 4
B an effective agonist capable of stimulating the cMpl -recep-
50K Fab59
40K LC dimer
25K
20K
Fig. 7. (A) SEC profiles of Fab59 on Superdex75. Solid line: Fab59
purified in regular binding and elution buffers. Dash line: Fab59 purified
in modified binding and elution buffers. (B) SDS–PAGE analysis of
monomeric Fab59 purified from bacterial pellets (periplasmic preps). (1)
Monomeric Fab59 from induction at room temperature (reducing
condition). (2) Standard molecular weight marker. (3) Monomeric
Fab59 from induction at room temperature (non-reducing condition).
(4) Monomeric Fab59 from induction at 30 oC (non-reducing condition).
LC dimer, light chain dimer.
was very clean and free of detectable contaminates
(Fig. 7B). Equal amounts of light chains and heavy chains
could be seen on a reducing SDS–PAGE gel, demonstrat-
ing that Fab59 was properly folded and assembled and that
the disulfide bonds were well formed. No detectable light
chain dimer contamination was observed in the monomeric
Fab59 purified from bacterial pellets induced at room tem-
perature. In contrast, some light chain dimer contamina-
tion in monomeric Fab59 purified from bacterial pellets
induced at 30 °C (Fig. 7B) was noted, suggesting that
Fab59 had a better folding and light/heavy chain paring
at room temperature induction rather than at 30 °C.
Because the molecular weight (MW) of light chain dimers
is similar to the MW of Fab59, making SEC separation dif-
ficult, expression at room temperature was best for optimal
purity. Although SEC separation is useful at this scale, it
may be less feasible for large-scale manufacturing proto-
cols therefore requiring additional optimization. However,
61
since conditions that efficiently prevent Fab59 aggregation
have been found using an optimized buffer, this should not
present a substantial difficulty.
The activity of purified Fab59 was equivalent to recombinant
human TPO in cell-based assay
Purified monomeric Fab59 was tested in comparison to
recombinant human TPO in vitro for its ability to stimulate
cMpl-R present on cells in a luciferase reporter assay [2].
Dose dependent receptor activation was observed with
addition of increasing amounts of Fab59, with comparable
activities from Fab59 preparations purified in two different
buffers (Fig. 8). No activity was observed in the control
vector transfected cells treated in an equivalent manner
(data not shown). The results indicate that Fab59, recov-
ered following a series of step-wise expression and purifica-
tion improvements that significantly increased its yield, was
tor in a specific manner.
Discussion
The rationally-designed agonist antibody fragment
Fab59 may provide therapeutic benefit in patients suffering
from low platelet counts, or thrombocytopenia, and is
therefore a potential candidate for clinical development
[2]. However, large-scale bacterial production of the origi-
nal Fab59 was not feasible since its expression level was
extremely low and barely detected by Western blot.
Poor expression of Fabs in E. coli can often be related to
improper protein folding, intracellular degradation, aggre-
gation in the periplasm and toxicity of the antibody frag-
ment to the E. coli host [4,5]. Several different strategies
have been shown to improve expression levels efficiently
5
rhTPO
50 Kd Fab 59 monomer prep F1
Relative Fold Luciferase Activity
50 Kd Fab 59 monomer prep F2
4
3
2
1
1000
100
10
Concentration (pM)
Fig. 8. Fab59 agonist activity demonstrated in cMpl receptor expressing
BaF3 cells using a luciferase reporter assay. rhTPO: purified recombinant
human TPO protein (R&D Systems). Fifty kiloDalton Fab59 monomer
prep F1: Fab59 monomer purified in PBS buffers. Fifty kiloDalton Fab59
monomer prep F2: Fab59 monomer purified in HEPES buffers.
62 B. Lin et al. / Protein Expression and Purification 59 (2008) 55–63
for various Fabs. Toxicity of some Fabs can be greatly alle-
viated by modulating transcriptional and translational
rates, using methods such as low induction temperature,
low level IPTG induction, and modification of culture med-
ium [8–10,16–18], or by using more tightly controlled pro-
moters (tet and ara) [9, 19–21]. Co-expression of bacterial
chaperons or heterologous disulfide isomerases [4,22] has
successfully improved Fab folding and disulfide bond for-
mation. In addition optimizing secretion signal sequences
could efficiently promote protein translocation to the bac-
terial periplasm where proteins can be properly folded
and form the correct disulfide bonds [23]. Other strategies
have been reported to enhance Fab expression levels such
as expression in different hosts other than in E. coli
[4,24], vector engineering to balance light chain and heavy
chain production [10,25], and even more sophisticated pro-
tein point mutations [26] to remove critical surface exposed
hydrophobic amino acid residues thereby increasing Fab
solubility. However, each particular strategy may not be
applicable to every difficult to express protein since the rea-
sons in each case will be different. Furthermore, multiple
factors are likely at play in most cases. As a result, optimi-
zation of expression can be laborious and time-consuming.
In an effort to boost the Fab59 expression in E. coli in
this study, we first determined what factors were contribut-
ing to the low expression level of Fab59. In comparison to
the parental scaffold Fab TT or Fab with a peptide grafted
heavy chain (Fab HCG), expression of Fabs containing a
grafted light chain (Fab LCG and Fab59) was dramatically
decreased as noted for both Fab expression (non-reducing
gels) and light chain expression (reducing gel). We system-
atically accomplished improvements by first codon opti-
mizing the light chain, second, using a tightly controlled
strong promoter, and third inducing at low temperature
with low levels of IPTG.
Additionally, a significant decrease in solubility of light
chain grafted Fabs was noted, both in E. coli cells and in
buffers used for purification. The solubility features of
Fab59 may be related to an increase of surface exposed
hydrophobic residues residing within the grafted peptide
which was selected for optimal presentation within the
Fab CDRs in order to achieve binding with target recep-
tors, and therefore required for function. Poor expression
of Fab59 protein also appeared to be linked to grafted light
chain aggregation as a result of limited solubility, and sub-
sequent toxicity to E. coli. Other cases of poor light chain
expression, and related toxicity to the cell, have been
reported [5,27]. In our case, the parental LC expressed well
on its own, but insertion of the TPO mimetic peptide into
CDR2 resulted in decreased LC expression and concomi-
tant low Fab expression. To resolve this limitation, addi-
tional analysis was undertaken.
The bacterial periplasmic protein Skp is a chaperon pro-
tein that has been shown to improve solubility and reduce
toxicity of recombinant proteins in E. coli [4,28–31]. Skp’s
native role is to act on newly synthesized bacterial outer-
membrane proteins, and protect them from misfolding
and aggregation during passage through the bacterial per-
iplasm. Skp does this by recognizing unfolded or misfolded
structures and interacting with them [32]. Aggregating pro-
teins can be toxic to bacterial cells by acting as a sink for
other folding host proteins causing a stress response and
draining cell resources [33]. Therefore it is believed that
co-expression of Skp alleviates the toxicity of aggregation
prone recombinant proteins in E. coli by promoting correct
polypeptide folding. Skp has been used successfully with
antibody fragments, including single chain antibodies
(scFv) [34,35] and Fabs [4]. In this study, we showed that
Skp protein co-expression significantly reduced the toxicity
of Fab59 and reversed the growth retardation of E. coli
cells bearing pAEV-O59. Skp mediated reduction of
Fab59 toxicity may result from the observed decrease in
the accumulation of insoluble light chain.
Interestingly, our results also demonstrated that co-
expression of the Skp protein significantly increased assem-
bled Fab59 ‘‘secretion” into the culture medium. The
increased Fab59 appearance in the medium is not from cell
lysis since we did not see significant free chain accumula-
tion in the medium, even though there was significant sol-
uble free chain accumulation in the periplasmic fraction.
Skp mediated increase in light chain solubility correlated
with a reduction of Fab59 toxicity and with an increase
of Fab59 secretion into the medium. Increased soluble
LC likely allows for more Fab assembly within the peri-
plasm, leading to more efficient release into the culture
supernate, though direct effects of Skp protein in promot-
ing Fab secretion cannot be ruled out at this time.
In a final step, the purification process from periplasmic
extracts was optimized to efficiently avoid protein aggrega-
tion. The use of glycerol, Tween-80 and increased ion
strength in the purification buffers efficiently prevented
Fab59 from forming insoluble and soluble aggregates,
and did not interfere with downstream purification or func-
tional assays. Purification of monomeric Fab59 from bac-
terial periplasmic extracts was accomplished in a two-step
protocol using SEC to remove antibody related contami-
nates, such as degradation products, free light and heavy
chains, and dimers/aggregates. This two-step purification
provided a convenient, fast and straightforward method
to obtain highly pure monomeric and correctly folded
Fab59 with intact disulfide bonds sufficient to support fur-
ther preclinical development. After expression and purifica-
tion optimization, the final yields of purified monomeric
Fab59 reached 3.5 mg/L bacterial culture in a shake flask
and represented a 100-fold increase from our starting yield.
Although this was a significant improvement, even higher
yields may be obtained in a larger-scale bioreactor, thereby
increasing Fab590 s potential for practical and economical
production.
In summary, Fab59 expression and purification from
E. coli has been successfully enhanced 100-fold by combining
an alternate vector, codon optimization, optimized induc-
tion conditions, co-expression of a chaperon protein and
improved purification processes. Following these efforts,
 B. Lin et al. / Protein Expression and Purification 59 (2008) 55–63
the agonist activity of the purified monomeric Fab59 was
demonstrated to be equipotent to rhTPO in in vitro cMpl-
R signaling, luciferase reporter-based assays, and further-
more stimulated increased platelet levels in vivo [2].
References
[1] S.E. Cwirla, P. Balasubramanian, D.J. Duffin, C.R. Wagstrom, C.M.
Gates, S.C. Singer, A.M. Davis, R.L. Tansik, L.C. Mattheakis, C.M.
Boytos, P.J. Schatz, D.P. Baccanari, N.C. Wrighton, R.W. Barrett,
W.J. Dower, Peptide agonist of the thrombopoietin receptor as the
natural cytokine, Science 276 (1997) 1696–1699.
[2] S. Frederickson, M.W. Renshaw, B. Lin, L.M. Smith, P. Calveley,
J.P. Springhorn, K. Johnson, Y. Wang, X. Su, Y. Shen, K.S.
Bowdish, A novel rationally designed agonist antibody fragment that
functionally mimics thrombopoietin, Proc. Natl. Acad. Sci. USA 103
(2006) 14307–14312.
[3] R. Verma, E. Boleti, A.J. George, Antibody engineering: comparison
of bacterial, yeast, insect and mammalian expression systems, J.
Immunol. Methods 216 (1998) 165–181.
[4] R. Levy, R. Weiss, G. Chen, B.L. Iverson, G. Georgiou, Production
of correctly folded Fab antibody fragment in the cytoplasm of
Escherichia coli trxB gor mutants via the coexpression of molecular
chaperones, Protein Express. Purif. 23 (2001) 338–347.
[5] S. Kumar, J. Kalsi, D.S. Latchman, L.H. Pearl, D.A. Isenberg,
Expression of the Fabs of human auto-antibodies in Escherichia coli:
Optimization and determination of their fine binding characteristics
and cross-reactivity, J. Mol. Biol. 308 (2001) 527–539.
[6] B.D. Stollar, Bacterial expression of anti-DNA antibody domains,
Methods 11 (1997) 12–19.
[7] P.M. Sharp, W.H. Li, The codon adaptation index—a measure of
directional synonymous codon usage bias, and its potential applica-
tions, Nucleic Acids Res. 15 (1987) 1281–1295.
[8] P.M. O’Brien, G. Maxwell, M.S. Campo, Bacterial expression and
purification of recombinant bovine Fab fragments, Protein Express.
Purif. 24 (2002) 43–50.
[9] B. Seliko, G. Cosio, C. Garcia, B. Becerril, L.D. Possani, E. Horjales,
Bacterial expression, purification and functional characterization of a
recombinant chimeric Fab derived from murine mAb BCF2 that
neutralizes the venom of the scorpion Centruroides noxius Hoffmann,
Toxicon 43 (2004) 43–51.
[10] S. Corisdeo, B. Wang, Functional expression and display of an
antibody Fab fragment in Escherichia coli : study of vector designs
and culture conditions, Protein Express. Purif. 34 (2004) 270–279.
[11] M. Nalezkova, A. de Groot, M. Graf, P. Gans, L. Blanchard,
Overexpression and purification of Pyrococcus abyssi phosphopant-
etheine adenylytransferase from an optimized synthetic gene for
NMR studies, Protein Express. Purif. 39 (2005) 296–306.
[12] M.W. Laird, G.C. Sampey, K. Johnson, D. Zukauskas, J. Pierre, J.S.
Hong, B.A. Cooksey, Y. Li, O. Galperina, J.D. Karwoski, R.N.
Burke, Optimization of BLyS production and purification from
Escherichia coli, Protein Express. Purif. 39 (2005) 237–246.
[13] E.P. Rocha, Codon usage bias from tRNA’s point of view: redun-
dancy, specialization, and efficient decoding for translation optimi-
zation, Genome Res. 14 (2004) 2279–2286.
[14] A. Neagu, M. Neagu, A. Der, Fluctuations and the Hofmeister effect,
Biophys. J. 81 (2001) 1285–1294.
[15] B.A. Kerwin, M.C. Heller, S.H. Levin, T.W. Randolph, Effects of
Tween 80 and sucrose on acute short-term stability and long-time
storage at 20 °C of a recombinant hemoglobin, J. Pharm. Sci. 87
(1998) 1062–1068.
[16] H. Wlad, A. Ballagi, L. Bouakaz, Z. Gu, J.-C. Janson, Rapid two-
step purification of a recombinant mouse Fab fragment expressed in
Escherichia coli, Protein Express. Purif. 22 (2001) 325–329.
[17] J.G. Wall, A. Pluckthum, Effectors of overexpression folding mod-
ulators on the in vivo folding of heterologous proteins in Escherichia
coli, Curr. Opin. Biotechnol. 6 (1995) 507–516.
63
[18] D. Humphreys, A.P. Chapman, D.G. Reeks, V. Lang, P.E. Stephens,
Formation of dimeric Fabs in Escherichia coli: effort of hinge size and
isotype, presence of interchain disulphide bond, Fab’ expression
levels, tail piece sequences and growth conditions, J. Immunol.
Methods 209 (1997) 193–202.
[19] K. Alfthan, T. Kristiina, S. Dorothea, S. Ilkka, I. Tiina, V. Liisa,
K. Sirkka, K. Matti, K.C. Jonathan, T.T. Tuula, Efficient
secretion of murine Fab fragments by Escherichia coli is deter-
mined by the first constant domain of the heavy chain, Gene 128
(1993) 203–209.
[20] F. Maeda, Y. Nagatsuka, S. Ihara, S. Aotsuka, Y. Ono, H. Inoko, M.
Takekoshi, Bacterial expression of a human recombinant monoclonal
antibody Fab fragment against Hepatitis B surface antigen, J. Med.
Virol. 58 (1999) 338–345.
[21] D. Rau, K. Kramer, B. Hock, Cloning, functional expression and
kinetic characterization of pesticide-selective Fab fragment variants
derived by molecular evolution of variable antibody genes, Anal.
Bioanal. Chem. 372 (2002) 261–267.
[22] A. Ideno, F. Masahiro, I. Yoshitaka, K. Yoshikazu, M. Tadashi,
FK506 binding protein from the hyperthermophilic archaeon Pyro-
coccus horikoshii suppresses the aggregation of protein in Escherichia
coli, Appl. Environ. Microbiol. 68 (2002) 464–469.
[23] D. Humphreys, M. Sehdev, A.P. Chapman, R. Ganesh, B.J. Smith,
L.M. King, D.J. Glover, D.G. Reeks, P.E. Stephens, High-level
periplasmic expression in Escherichia coli using a eukaryotic signal
peptide: Importance of codon usage at the 50 end of the coding
sequence, Protein Express. Purif. 20 (2000) 252–264.
[24] M. Venturi, C. Seifert, C. Hunte, High level production of functional
antibody Fab fragment in an oxidizing bacterial cytoplasm, J. Mol.
Biol. 315 (2002) 1–8.
[25] D.P. Humphreys, B. Carrington, L.C. Bowering, R. Ganesh, M.
Sehdev, B.J. Smith, L.M. King, D.G. Reeks, A. Lawson, A.G.
Popplewell, A plasmid system for optimization of Fab0 production in
Escherichia coli: importance of balance of heavy chain and light chain
synthesis, Protein Express. Purif. 26 (2002) 309–320.
[26] J.G. Wall, A. Pluckthun, The hierarchy of mutations influencing the
folding of antibody domains in Escherichia coli, Protein Eng. 12
(1999) 605–611.
[27] G. Forsberg, M. Forsgren, M. Jaki, M. Norin, C. Sterky, A.
Enhorning, K. Larsson, M. Ericsson, P. Bjork, Identification of
framework residues in a secreted recombinant antibody fragment that
control production level and localization in Escherichia coli, J. Bio.
Chem. 272 (1997) 12430–12436.
[28] J. Maynard, E.J. Adams, M. Krogsgaard, K. Petersson, C.W. Liu,
K.C. Garcia, High-level bacterial secretion of single-chain ab T-cell
receptors, J. Immunol. Methods 306 (2005) 51–67.
[29] D.K. Chatterjee, D. Esposito, Enhanced soluble protein expres-
sion using two new fusion tags, Protein Expr. Purif. 46 (2005)
122–129.
[30] T.A. Walton, M.C. Sousa, Crystal structure of Skp, a profoldin-like
chaperone that protects soluble and membrane proteins from
aggregation, Mol. Cell. 15 (2004) 367–374.
[31] H. Boothmann, A. Pluckthun, Selection for a periplasmic factor
improving phage display and functional periplasmic expression,
Nature Biotechnol. 16 (1998) 376–380.
[32] J. Qu, C. Mayer, S. Behrens, O. Holst, J.H. Kleischmidt, The trimeric
periplasmic chaperone Skp of Escherichia coli forms 1:1 complexes
with outer membrane proteins via hydrophobic and electrostatic
interactions, J. Mol. Biol. 374 (2007) 91–105.
[33] D. Missiakas, S. Raina, Protein folding in the bacterial periplasm, J.
Bacteriol 179 (1997) 2465–2471.
[34] A. Hayhurst, W.J. Harris, Escherichia coli Skp chaperone coexpres-
sion improves solubility and phage display of single-chain antibody
fragments, Protein Express. Purif. 15 (1999) 336–343.
[35] A. Hayhurst, S. Happe, R. Mabry, Z. Koch, B.L. Iverson, G.
Georgiou, Isolation and expression of recombinant antibody frag-
ments to the biological warfare pathogen Brucella melitensis, J.
Immunol. Methods 276 (2003) 185–196.