A novel system for continuous protein refolding and
`on-line capture by expanded bed adsorption
`
`HENRIK FERRE´ ,1,3,4 EMMANUEL RUFFET,1 LISE-LOTTE B. NIELSEN,1
`MOGENS HOLST NISSEN,2 TIMOTHY J. HOBLEY,3 OWEN R.T. THOMAS,3,5
`AND SØREN BUUS1
`1Institute of Medical Microbiology and Immunology and 2Institute of Medical Anatomy, University of
`Copenhagen, The Panum Institute, DK-2200, Copenhagen N, Denmark
`3Center for Microbial Biotechnology, BioCentrum-DTU, Technical University of Denmark, DK-2800,
`Kgs. Lyngby, Denmark
`
`(RECEIVED February 4, 2005; FINAL REVISION May 19, 2005; ACCEPTED May 23, 2005)
`
`Abstract
`
`A novel two-step protein refolding strategy has been developed, where continuous renaturation-by-
`dilution is followed by direct capture on an expanded bed adsorption (EBA) column. The perfor-
`mance of the overall process was tested on a N-terminally tagged version of human b2-microglobulin
`(HAT-hb2m) both at analytical, small, and preparative scale. In a single scalable operation, extracted
`and denatured inclusion body proteins from Escherichia coli were continuously diluted into refolding
`buffer, using a short pipe reactor, allowing for a defined retention and refolding time, and then fed
`directly to an EBA column, where the protein was captured, washed, and finally eluted as soluble
`folded protein. Not only was the eluted protein in a correctly folded state, the purity of the HAT-
`hb2m was increased from 34% to 94%, and the product was concentrated sevenfold. The yield of the
`overall process was 45%, and the product loss was primarily a consequence of the refolding reaction
`rather than the EBA step. Full biological activity of HAT-hb2m was demonstrated after removal of
`the HAT-tag. In contrast to batch refolding, a continuous refolding strategy allows the conditions to
`be controlled and maintained throughout the process, irrespective of the batch size; i.e., it is readily
`scalable. Furthermore, the procedure is fast and tolerant toward aggregate formation, a common
`complication of in vitro protein refolding. In conclusion, this system represents a novel approach to
`small and preparative scale protein refolding, which should be applicable to many other proteins.
`
`Keywords: protein refolding; expanded bed absorption (EBA); recombinant proteins; inclusion bodies
`
`Heterologous protein production in bacteria has the
`potential to supply virtually unlimited amounts of high-
`value products. Unfortunately, the outcome is frequently
`unsatisfactory due to the deposition of recombinant
`
`Present addresses:4Alpharma ApS, DK-2300 Copenhagen S, Denmark;
`5Department of Chemical Engineering, School of Engineering, The
`University of Birmingham, Edgbaston, Birmingham B15 2TT, Great
`Britain.
`Reprint requests to: Søren Buus, Institute of Medical Microbiology
`and Immunology, University of Copenhagen, The Panum Institute,
`Building 18.3, Blegdamsvej 3C, DK-2200, Copenhagen N, Denmark;
`e-mail: sb@immi.ku.dk; fax: +45-3532-7696.
`Article and publication are at http://www.proteinscience.org/cgi/
`doi/10.1110/ps.051396105.
`
`protein as inactive insoluble inclusion bodies, particularly
`when expression levels are high (Marston 1986). To re-
`cover the biological activity of the protein, extraction into
`denaturing buffer followed by in vitro refolding is neces-
`sary. Methods available for in vitro protein refolding are
`usually of low efficiency, and therefore, the high levels of
`bacterial expression do not readily translate into high
`yields of functional product.
`During refolding, soluble and insoluble byproducts
`form as a result of inappropriate rearrangements within
`the protein itself and/or through unfavorable interactions
`with neighboring proteins. These interactions are typically
`mediated through exposed hydrophobic surfaces and/or
`interdisulfide cross-linking (Speed et al. 1996). Such inter-
`
`Protein Science (2005), 14:2141–2153. Published by Cold Spring Harbor Laboratory Press. Copyright ª 2005 The Protein Society
`
`2141
`
`ps0513961
`
`Ferre et al.
`
`Article RA
`
`Page 1
`
`KASHIV EXHIBIT 1004
`IPR2019-00791
`
`

`

`Ferre´ et al.
`
`molecular aggregation processes are second or higher
`order reactions (Kiefhaber et al. 1991; Middelberg 2002).
`At high-protein concentrations, they dominate over pro-
`ductive folding, leading to decreased yields (Rudolph and
`Lilie 1996). To minimize adverse intermolecular inter-
`actions, the protein concentration should be kept low
`throughout the folding reaction, i.e., one is required to
`work with large reaction volumes. Even so, the formation
`of aggregates is difficult to avoid completely and the net
`result of an in vitro folding process is therefore likely to
`be a large reaction volume containing the diluted protein
`in a turbid suspension, a situation that severely hampers
`downstream processing, as several clarification steps are
`needed prior to fractionation by packed bed chromato-
`graphy. Furthermore, the handling of large volumes
`increases the cost of industrial processes significantly.
`Recent developments have therefore been aimed at
`establishing processes where refolding can be performed
`at high-protein concentrations (Zardeneta and Horowitz
`1994a; Clark 1998). This includes batch and/or on-
`column refolding in the presence of enzymatic folding
`catalysts immobilized onto chromatographic supports
`(Stempfer et al. 1996; Altamirano et al. 1997; Dong
`et al. 2000). Possible folding catalysts include chaper-
`onins, which bind to intermediate protein structures,
`thereby preventing aggregation; protein disulfide isom-
`erases; and peptidyl-prolyl cis-trans isomerases, which
`promote disulfide bond formation and cis-trans isomer-
`ization, respectively, events that have often been identi-
`fied as rate-limiting steps of the folding pathway (Jager
`and Pluckthun 1997; Rothwarf et al. 1998). Although
`these processes have facilitated the refolding of difficult-
`to-fold proteins (Altamirano et al. 1999, 2001), a number
`of problems persist that have prevented exploitation at
`large scale. These problems include (1) availability and
`cost of catalytic enzymes, (2) difficulty in regenerating
`the chromatographic support, as harsh conditions can
`lead to inactivation of the immobilized enzymes, and (3)
`scalability of the process. Furthermore, such assisted
`refolding might be limited to a narrow set of substrates,
`and conditions for optimal
`in vitro refolding might
`not be compatible with the activity of the catalytic
`enzymes.
`On-column renaturation of protein is generally diffi-
`cult to handle, as most in vitro refolding reactions are
`highly inefficient, resulting in both soluble and insoluble
`misfolded species, which could potentially foul the chro-
`matographic adsorbents. Batch operations are not ideal
`either, as the conditions for folding-by-dilution are dif-
`ficult to control at larger scales due to the problem of
`rapidly mixing large suspension volumes. The latter is
`particularly evident for folding reactions that proceed at
`rates in which the majority of the events (i.e., productive
`folding or aggregation) are completed within milli-
`
`2142
`
`Protein Science, vol. 14
`
`seconds to seconds (Goldberg et al. 1991; Dobson and
`Karplus 1999) i.e., faster than effective mixing can be
`achieved. Thus, there is a need to develop systems for
`protein folding, in which the conditions can be con-
`trolled and maintained throughout the reaction at both
`small and large scales of operation. In particular, the
`refolding reaction should be defined with respect to time
`in order to achieve high reproducibility and predictabil-
`ity of the protein yield from small- to large-scale process-
`ing. Moreover, the system should be robust toward
`insoluble protein aggregates, which might form during
`the course of the folding reaction.
`Here, a novel process is presented, in which refolding
`can be carried out in continuous mode under controlled
`and specified conditions. This is obtained by diluting the
`denatured protein within a small flowthrough mixing
`chamber, in which the folding conditions can be care-
`fully controlled and maintained throughout the entire
`process. From this mixing chamber, the proteins pass
`through a folding pipe reactor with sufficient retention
`time to allow folding. Finally, the nascently folded pro-
`tein is directly captured by expanded bed adsorption
`(EBA)—a special type of fluidized bed chromatography
`that can handle crude suspensions and large reac-
`tion volumes (Draeger and Chase 1990, 1991; Tho¨ mmes
`1997). Our approach uncouples the events of protein
`refolding and capture, thereby allowing each event to be
`optimized individually. The performance of the present
`process was tested with a crude inclusion body extract of
`N-terminally tagged human b2-microglobulin (HAT-
`hb2m), the light chain of the major histocompatibility
`complex class I (MHC-I) molecule.
`
`Results
`
`Production of denatured and correctly oxidized
`HAT-h

2m
`
`HAT-hb2m was produced as insoluble inclusion bodies
`by Escherichia coli fermentations, and Figure 1A shows
`an SDS-PAGE analysis of samples collected before and
`after induction with IPTG. Expression of HAT-hb2m
`was detected 1 h after induction, and the level continued
`to increase until a maximum plateau was reached after
`2 h (Fig. 1A). Three hours after induction, the inclu-
`sion bodies were released by either enzymatic or
`mechanical disruption of the cells. The released inclu-
`sion bodies were washed and solubilized in 8 M urea
`under nonreducing conditions, yielding denatured and
`oxidized HAT-hb2m (Fig. 1B,C).
`Correctly oxidized monomeric hb2m molecules contain
`two cysteine residues, forming a single disulphide bond,
`which can be demonstrated by comparing the mobility of
`the hb2m protein band under reducing and nonreducing
`
`Page 2
`
`

`

`Continuous refolding and on-line EBA capture
`
`spectrophotometry at three different urea concentrations
`(Fig. 2). Aggregation could not be detected at OD450nm
`when refolding was conducted in the protein concentra-
`tion range of 10–100 mg/mL, irrespective of the urea con-
`centrations. However, the Psol/Ptot ratio in this particular
`region decreased slightly (2%–3%) (Fig. 2), indicating
`that microaggregates had formed, and that these could
`be removed by centrifugation at 20,000g. Visible particu-
`late aggregates started to appear when the protein con-
`centration was raised to 200 mg/mL during refolding at
`100 mM urea. However, the refolding reaction could be
`completed without further aggregation at two- to three-
`fold higher protein concentrations when the level of urea
`was increased to 224 mM and 630 mM, respectively. Thus,
`aggregation of HAT-hb2m is highly dependent on the
`level of urea after folding by dilution. In all cases, the
`recovered soluble HAT-hb2m was
`highest
`level of
`observed at the lowest protein concentration, i.e., 10 mg/
`mL (Fig. 2).
`
`EBA-based system for continuous protein refolding
`and purification
`
`A diagram of the system used for continuous protein
`refolding and on-line EBA capture is depicted in Figure 3.
`Protein refolding is initiated by dilution of the denatured
`protein suspension in a flowthrough mixing chamber.
`Refolding conditions (i.e., total protein concentration,
`urea concentration, and refolding time) can be set and
`maintained throughout the process by adjusting the flow
`rates of buffer and denatured protein suspension into the
`
`Figure 2. Aggregation of HAT-hb2m as a function of the protein and
`urea concentration. Refolding was initiated by batch dilution in 20 mM
`Tris-HCl (pH 8.0), and the formation of particulate aggregates was
`measured by spectrophotometry at 450 nm after 30 min incubation at
`RT. The protein content of the samples was determined after centrifu-
`gation and the ratio of soluble (Psol) and total protein (Ptot) was calcu-
`lated. Filled symbols indicate OD450nm measurements, and open
`symbols indicate Psol/Ptot values at urea concentrations of 100 mM
`(squares), 224 mM (circles), and 630 mM (triangles), respectively. The
`data shown are the average of three independent folding reactions, and
`standard deviations are included at each point.
`
`www.proteinscience.org
`
`2143
`
`Figure 1. SDS-PAGE analysis of expression levels in E. coli and
`solubilized inclusion body preparations of HAT-hb2m. (A) Expression
`level of HAT-hb2m before and after induction. (Lane 1) Protein marker;
`(lane 2) before induction; (lanes 3–5) expression levels after 1, 2, and 3 h
`of induction, respectively. (B,C) Reducing and nonreducing SDS-PAGE
`analysis of feedstock A. (Lane M) Protein marker. DTT, dithiothreitol;
`2-ME, 2-mercaptoethanol. Molecular weights of standard protein are
`indicated, and the positions of reduced (0) and oxidized (1) HAT-hb2m
`monomers are shown with arrows.
`
`conditions in SDS-PAGE gels (Fig. 1B,C). The mobility
`shift of the reduced (Fig. 1B,C, band 0) and oxidized (Fig.
`1B,C, band 1) HAT-hb2m bands is most clearly seen with
`DTT as the reductant (Fig. 1B). The kink in the HAT-
`hb2m protein band seen in Figure 1C is the result of
`diffusion of 2-ME from the reducing lane into the border-
`ing nonreducing lane. This partial mobility shift of the
`HAT-hb2m band in the nonreducing lane confirms that
`the oxidized protein band (1) shifts into the reduced band
`(0) upon reduction. Monomeric HAT-hb2m was found to
`be fully oxidized in the feedstock preparations used in the
`following refolding experiments. The advantage of refold-
`ing correctly oxidized species is that the soluble target
`molecule can be generated in a simple and fast folding-
`by-dilution reaction without adding expensive redox pairs
`such as glutathione and oxidized glutathione to promote
`disulphide bond formation.
`The purity of feedstocks A (enzymatic cell disruption)
`and B (mechanical cell disruption) with respect to oxi-
`dized HAT-hb2m monomers was 34.0% (62.6%) and
`56.0% (61.3%), respectively, as determined by SDS-
`PAGE and densitometry. The total inclusion body pro-
`tein recovery was 0.7 g/L and 3.0 g/L bacterial culture
`for feedstock preparations A and B, respectively.
`
`Batch refolding studies
`
`The impact of the total protein and urea concentration on
`the refolding yields of HAT-hb2m was investigated in
`analytical scale batch reactions. The formation of insolu-
`ble aggregates of HAT-hb2m in the protein concentration
`range of 10–1000 mg total protein/mL was measured by
`
`Page 3
`
`

`

`Ferre´ et al.
`
`Figure 3. Schematic representation of the system for continuous protein
`refolding and on-line EBA capture. All EBA operations—i.e., equilibra-
`tion,
`loading, washing, elution, and cleaning—were performed in
`expanded mode in Fastline10 and 50 columns. The current system allows
`the refolding buffer to be recycled through the system. Closed arrows
`indicate the direction of liquid flow during the folding/capture step, and
`open arrows indicate manual valves.
`
`flowthrough mixing chamber and by changing the length
`of the folding-pipe reactor. Initial experiments were per-
`formed with feedstock A at 10 mg/mL and 224 mM urea to
`achieve the highest possible recovery of HAT-hb2m and
`the least amount of insoluble aggregates (Fig. 2). The
`length of the folding-pipe reactor was adjusted to allow
`the protein 14 sec to fold. The intermediate urea concentra-
`tion was selected to minimize the viscosity of the feed
`stream entering the EBA column.
`The elution profile from the EBA column in expanded
`mode using a linear 0–1 M NaCl gradient together with a
`SDS-PAGE analysis of selected fractions are shown in
`Figure 4, A and B. It can be seen that soluble monomeric
`HAT-hb2m was eluted in a broad peak with an extensive
`tail (0.1–0.6 M NaCl). The majority of the soluble con-
`taminants were eluted between 0.2 M and 0.4 M NaCl.
`The tailing of the HAT-hb2m peak is most likely the
`result of elution being performed in expanded mode. A
`second and sharper peak was detected later in the gradi-
`ent, which could not be explained by the protein deter-
`mination data or the SDS-PAGE analysis (Fig. 4, cf. A
`and B). However, fractions collected in this region also
`gave rise to an absorbance at 254 nm, indicating the
`presence of DNA and/or RNA in the samples. Agarose
`gel electrophoresis analysis followed by CYBR Gold
`staining confirmed that fractions 30 through 36 con-
`tained trace amounts of contaminating DNA and/or
`RNA (data not shown).
`
`2144
`
`Protein Science, vol. 14
`
`HAT-hb2m was not detected by SDS-PAGE in the
`flowthrough fractions (Fig. 4B, lane 4), demonstrating
`that all of the folded HAT-hb2m had adsorbed to the
`STREAMLINE DEAE medium. Purities of fractions
`enriched in monomeric HAT-hb2m were estimated by den-
`sitometric analysis and ranged from > 95% for the first
`fraction (Fig. 4B, lane 6) down to 60% for fractions 27–
`30 (Fig. 4B, lanes 8–10). Although no visible aggregation
`(i.e., OD450nm = 0) could be detected under the refolding
`conditions of this experiment (Fig. 2), > 50% of the total
`amount of monomeric HAT-hb2m applied to the system
`could not be accounted for in the linear elution with NaCl.
`In order to increase the purity of the product peak,
`and to improve the control of the process at larger
`scales, the gradient profile was changed from linear to
`stepwise elution using a concentration of 0.15 M NaCl
`in the first step, followed by a second step with 1 M
`NaCl. To account for the protein that could not be
`desorbed from the EBA medium using 1 M NaCl, two
`additional elution steps under denaturing conditions
`(i.e., one without reducing agent, followed by one with
`reducing agent) were included.
`
`Figure 4. Small-scale continuous refolding and EBA capture of HAT-
`hb2m using a 1-cm diameter EBA column with STREAMLINE DEAE
`operated in expanded mode. (A) Elution profile using a linear gradient
`of 0–1 M NaCl. Also shown is the protein concentration in the eluted
`fractions (&). (B) SDS-PAGE analysis of selected fractions from A.
`(Lane 1) Protein marker; (lane 2) reduced feedstock A; (lane 3) non-
`reduced feedstock A; (lane 4) flowthrough; (lanes 5–13) fractions 4, 7,
`10, 13, 16, 19, 22, 24, and 26, respectively. (C) Elution profile using a
`stepwise gradient consisting of the following steps: (1) 0.15 M NaCl, (2)
`1 M NaCl, (3) 1 M NaCl in 8 M Urea, and finally, (4) 1 M NaCl in 8 M
`Urea and 5 mM 2-ME as indicated above the chromatogram. (D) SDS-
`PAGE analysis of selected fractions from C. (Lane 1) Protein marker;
`(lane 2) reduced feedstock A; (lane 3) nonreduced feedstock A; (lane 4)
`flowthrough; (lanes 5–8) peaks 1, 2, 3, and 4 from C, respectively.
`Molecular weights of standard proteins are shown and the position of
`monomeric HAT-hb2m is indicated with arrows.
`
`Page 4
`
`

`

`Continuous refolding and on-line EBA capture
`
`Table 1. Comparison of small and large scale processing
`
`Scales of operation/Samples
`
`Volume
`(mL)
`
`Total protein
`(mg)
`
`Total
`HAT-hb2m (mg)
`
`Concentration
`factor (fold)
`
`Total recovery
`(%)
`
`Purity
`(%)
`
`Small scale
`Folded suspension
`Peak 1
`Peak 2
`
`Large scale
`Folded suspension
`Peak 1
`Peak 2
`
`735a
`45
`15
`
`18700a
`1530
`925
`
`10.5
`1.6
`0.5
`
`262.0
`44.0
`33.0
`
`3.60
`1.50
`0.07
`
`89.00
`37.00
`1.30
`
`1.0
`7.0
`0.9
`
`1.0
`5.0
`0.4
`
`100.0b
`43.0
`2.0
`
`100.0b
`41.0
`1.5
`
`34.0 (6 2.6)c
`94.0
`14.0
`
`34.0 (6 2.6)c
`83.0
`4.0
`
`No monomeric HAT-hb2m was detected in the unbound fraction (see Figs. 4B, 6B).
`a Calculated total volume after refolding at ~10 mg total protein/mL.
`b Assuming a 100% refolding efficiency.
`c Purities of folded feedstock A represent an average of four independent measurements.
`
`The elution profile using the stepwise gradient and the
`corresponding SDS-PAGE analysis of the peaks col-
`lected are shown in Figure 4, C and D. Highly purified
`soluble HAT-hb2m monomer (94% purity) eluted in
`the first peak (Fig. 4C, step 1), whereas soluble contami-
`nants together with minor amounts of HAT-hb2m (2%
`of the total denatured amount applied to the folding reac-
`tion) were eluted in the second peak (Fig. 4C, step 2).
`A summary of the recovery data is presented in Table 1.
`Approximately 43% of the total amount of denatured
`monomeric HAT-hb2m offered to the refolding reaction
`could be recovered as monomeric HAT-hb2m under native
`elution conditions (Table 1; Fig. 4C, peaks 1,2), and the
`target protein was concentrated sevenfold in the first peak.
`The remaining HAT-hb2m protein could only be eluted
`under denaturing and reducing conditions (Fig. 4C,D,
`peaks 3,4). Furthermore, the SDS-PAGE analysis of
`peaks 3 and 4 demonstrate that a large amount of contami-
`nating proteins, which apparently remained soluble during
`the refolding reaction, had adsorbed to the medium in a
`state that precluded their elution under native conditions
`(Fig. 4D, lanes 7,8). Note that the high degree of aggrega-
`tion of the target molecule and other proteins did not, at
`any stage, affect the performance and stability of the
`expanded bed, i.e., no channels or stagnant zones were
`observed, and the bed height remained constant through-
`out the loading/refolding step. After elution with urea and
`re-equilibration with binding buffer (i.e., 20 mM Tris-Hcl
`[pH 8.0]), the bed expanded to the initial height, indicating
`that the medium could be regenerated.
`
`Investigation of refolding time, efficiency,
`and system robustness
`
`Although the refolding reaction in the previous experiments
`was conducted under conditions that did not lead to ag-
`gregation as measured by OD450nm (Fig. 2), a significant
`
`amount of the presumably folded HAT-hb2m was ad-
`sorbed to the EBA medium in a form that could only be
`released by renewed denaturation (Fig. 4; Table 1). This loss
`might be the result of insufficient refolding time (i.e., 14 sec)
`in the pipe reactor, which produced an immature protein
`that was prone to aggregation or unfolding on the chromato-
`graphy medium. To further investigate the importance of
`the refolding time on the total recovery of soluble mono-
`meric HAT-hb2m, a pipe reactor was inserted between the
`flowthrough mixing chamber and the EBA column, allow-
`ing the protein to fold for 10 min before EBA capture.
`Moreover, in a batch-refolding experiment conducted un-
`der the same conditions, the protein was allowed 30 min to
`fold before starting to load the suspension onto the EBA
`column. Due to limited availability of feedstock A, these
`experiments were conducted with feedstock B, and in all
`experiments, the EBA column was washed and eluted with
`a step gradient of 1 M NaCl in a similar way to that shown in
`Figure 4C.
`Despite the increase in refolding time, the resulting yield
`of soluble monomeric HAT-hb2m was only marginally im-
`proved (53% using 14-sec refolding time vs. 57.6% and
`56.7%, using 10- and 30-min refolding times, respectively).
`This indicates that the refolding reaction for HAT-hb2m
`had essentially run to completion within 14 sec and had left
`a significant fraction of soluble monomer forms that could
`not be recovered under native elution conditions in the
`subsequent EBA chromatography.
`In light of the above results, the possibility that some
`folded and mature HAT-hb2m became irreversibly bound
`to the EBA support due to, for example, unfolding or
`aggregation, was examined. Refolded monomeric HAT-
`hb2m, which had already gone through one complete
`refolding–EBA purification cycle as shown in Figure 3,
`was desalted by Sephadex G-25 chromatography to
`remove the NaCl used during elution and then reapplied
`to the EBA column. To mimic the refolding conditions,
`
`www.proteinscience.org
`
`2145
`
`Page 5
`
`

`

`Ferre´ et al.
`
`the desalted HAT-hb2m pool was diluted in binding buf-
`fer (20 mM Tris-Hcl [pH 8.0] without urea) to a final
`protein concentration of 10 mg/mL prior to the EBA-
`loading step. In this case, > 99% of the applied HAT-
`hb2m was recovered in the eluted fraction using 1 M
`NaCl, and only 0.1% appeared in the 8-M urea cleaning
`steps (Fig. 5). Thus, the observed loss during continuous
`refolding and direct EBA capture can be attributed to the
`refolding reaction and not the chromatographic operation
`itself. It could be reasoned that the refolding reaction
`produces at least two soluble monomeric HAT-hb2m
`populations. One population is correctly refolded and
`fully recoverable; the other population is misfolded and
`interacts with the STREAMLINE DEAE medium in a
`way that leads to aggregation. Alternatively, a large
`amount of microaggregates are generated during the re-
`folding reaction, which adsorbs strongly to the chromato-
`graphic medium.
`Finally, the performance of the system for continuous
`refolding and direct EBA capture was tested at higher pro-
`tein concentrations during refolding. Continuous refold-
`ing was conducted at 200 and 600 mg total protein/mL
`at a urea concentration of 224 mM urea. In both cases, the
`same amount of total protein was applied to the EBA
`column as for the 10 mg/mL experiment. Importantly,
`refolding at 600 mg/mL lead to the formation of insoluble
`aggregates (see also Fig. 2). Adsorbed proteins were eluted
`with 1 M NaCl in binding buffer, and the aggregated
`
`Figure 5. Investigation of HAT-hb2m monomer recovery after contin-
`uous refolding at different protein concentrations and EBA purification.
`The total recovery following elution of previously refolded and purified
`HAT-hb2m, which was reloaded onto the EBA column, is shown in the
`first column. No soluble monomeric HAT-hb2m was detected in the
`flowthrough fraction in any of the four experiments, and equal amounts
`of total protein (10 mg) were applied to the column. (Filled bars)
`Soluble monomeric HAT-hb2m eluted with 1 M NaCl in 20 mM Tris-
`HCl (pH 8.0); (open bars) insoluble monomeric HAT-hb2m eluted under
`denaturing conditions (8 M urea, 1 M NaCl in 20 mM Tris-Hcl [pH 8.0]).
`
`2146
`
`Protein Science, vol. 14
`
`proteins were recovered using 8 M urea in binding buffer
`as described above. The recovery data are compared in
`Figure 5. In all cases, soluble monomeric HAT-hb2m
`could be eluted from the EBA column. Not surprisingly,
`the total recovery of soluble monomeric HAT-hb2m
`decreased from 53% to 16% when the protein concentra-
`tion during refolding was raised from 10 to 600 mg/mL as
`a result of the decreased refolding efficiency and forma-
`tion of insoluble aggregates (Fig. 5). Compared with the
`initial continuous refolding experiments conducted with
`feedstock A at 10 mg/mL, the total recovery of monomeric
`HAT-hb2m was increased by 10% with feedstock B (cf.
`Fig. 5 and Table 1). Feedstock B had a higher purity than
`feedstock A, raising the possibility that the presence of
`contaminating proteins affects the refolding efficiency of
`HAT-hb2m negatively.
`No bed stability problems were observed visually dur-
`ing loading of the EBA column when the continu-
`ous protein refolding reaction was performed at 10 and
`200 mg protein/mL, even though >50% of the mono-
`meric HAT-hb2m was adsorbed to the chromatographic
`medium in an insoluble state (Figs. 4, 5). Although
`insoluble aggregates were seen breaking through the
`column when refolding was conducted at 600 mg pro-
`tein/mL, some channeling and stagnant zones were
`observed at the bottom part of the bed (i.e., between 1
`and 5 cm). The channeling effects were not very severe
`and did not lead to breakthrough of soluble HAT-hb2m
`as judged by SDS-PAGE analysis of the flowthrough
`fraction. Since the instabilities in the bed disappeared
`during the elution phase with NaCl, they can most likely
`be attributed to electrostatic interactions between the
`soluble and insoluble protein aggregates formed at 600
`mg/mL and the DEAE ligands on the support surface
`causing cross-linking of neighboring beads. The bed
`stability problems observed at 600 mg/mL might be
`partly resolved by increasing the urea concentration
`during refolding from 224 to 630 mM, as the degree of
`aggregate formation was decreased by approximately
`sevenfold at the latter condition (Fig. 2).
`
`Scalability of the system
`
`The scalability of the present process was investigated
`by increasing the amount of feedstock processed 25-fold
`through the use of a 5-cm diameter EBA column. Con-
`ditions applied for the scaled-up process during refold-
`ing (i.e., total protein concentration, urea concentration,
`and retention time in the pipe reactor) and EBA opera-
`tion (i.e., linear fluid velocity, elution conditions, and
`total load of protein per volume of adsorbent) were kept
`constant. Refolding conditions were 10 mg/mL, 224 mM
`urea using a folding time of 14 sec in the pipe reactor.
`Although, as discussed above, slightly higher recoveries
`
`Page 6
`
`

`

`Continuous refolding and on-line EBA capture
`
`Figure 6. Preparative scale continuous refolding and direct EBA capture of HAT-hb2m. (A) Elution profile in expanded mode
`from a 5-cm diameter EBA column using a step gradient of 0.15 M NaCl, followed by 1 M NaCl. (B) SDS-PAGE analysis of
`peaks collected. (Lane 1) Protein marker; (lane 2) reduced feedstock A; (lane 3) nonreduced feedstock A; (lanes 4,5) peaks 1 and
`2, respectively; (lane 6) supernatant from support sample incubated with 8 M urea, 1 M NaCl, 10 mM 2-ME in 20 mM Tris-HCl
`(pH 8.0) overnight. Molecular weights of standard proteins, and the position of monomeric HAT-hb2m is shown with an arrow.
`
`could be obtained with a longer retention time in the
`pipe reactor, 14 sec was selected to keep the system as
`simple as possible at larger scale. The elution profile
`obtained at preparative scale (Fig. 6A) was similar to
`that observed for the small scale process (Fig. 4C), and
`SDS-PAGE analysis of collected fractions showed that
`highly purified soluble monomeric HAT-hb2m was
`eluted in the first peak (Fig. 6B). At preparative scale,
`the 8-M urea elution steps were omitted, and instead,
`the medium was cleaned by pumping 1–2 CVs of 1 M
`NaOH through the system, followed by recycling of the
`cleaning agent overnight. Prior to cleaning, a sample of
`support was removed from the column and treated with
`8 M urea, 1 M NaCl, supplemented with 10 mM 2-ME
`overnight at 4C. SDS-PAGE analysis of the superna-
`tant showed that a large amount of HAT-hb2m and
`contaminating proteins had adsorbed to the medium in
`an insoluble state (Fig. 6B), similarly to the scenario
`observed at small scale processing (Fig. 4D).
`The recoveries for small and preparative scale opera-
`tion are summarized and compared in Table 1 for feed-
`stock A. Very similar total recoveries of soluble mono-
`meric HAT-hb2m were obtained, but the purity and
`concentration of the target protein was slightly lower
`at preparative scale. The latter is the result of instability
`in the expanded bed caused mainly by the rotating dis-
`tributor system and the presence of coalesced bubbles
`arising from dissolved air in the buffers, which necessi-
`tated positioning the top adaptor 10 cm above the
`expanded bed to avoid chromatographic beads being
`flushed from the column. By comparison, the top adap-
`tor was placed 4 cm above the expanded bed during
`the elution procedure at small scale. Hence, the lower
`concentration factor at preparative scale in this particu-
`lar case is a direct result of the increased liquid head-
`space above the expanded bed, leading to dilution of the
`
`product peak. In a similar refolding and capture experi-
`ment using a STREAMLINE 50 EBA column, these
`problems were not observed (data not shown), and the
`decrease in concentration factor shown in Table 1 there-
`fore does not affect the general conclusion that the
`process is scalable.
`
`Determination of the biological activity
`of HAT-h

2m and h

2m
`
`For the small-scale preparation, the recombinant hb2m
`was released from the refolded and purified HAT-hb2m
`molecule by FXa cleavage, and the resulting peptide tag
`and undigested product were removed by Ni-NTA chro-
`matography as described in Materials and Methods.
`hb2m was collected in the flowthrough fraction, concen-
`trated, and further purified by size-exclusion chromatog-
`raphy on Sephadex G-50 to remove FXa and peptides.
`The ability of the MHC-I heavy chain to bind peptide
`is completely dependent on the presence of correctly
`folded hb2m (Garboczi et al. 1992). The functionality
`of the recombinant hb2m was therefore tested in a pep-
`tide-MHC-I binding assay. Figure 7 compares the hb2m
`dose-response data of three different versions of hb2m:
`(1) hb2m purified from natural sources; (2) folded,
`released, and purified recombinant hb2m; and (3) folded
`HAT-hb2m. Peptide binding was detectable at hb2m
`concentrations < 10 nM, which is comparable to what
`has previously been reported for recombinantly pro-
`duced hb2m (Pedersen et al. 1995, 2001). The biological
`activity of the natural hb2m and the recombinant hb2m
`were essentially identical, demonstrating that correctly
`folded hb2m can be generated using the EBA-based
`system for continuous refolding presented here.
`In contrast, the HAT-hb2m was less potent. Peptide
`binding became detectable around 50 nM HAT-hb2m,
`
`www.proteinscience.org
`
`2147
`
`Page 7
`
`

`

`Ferre´ et al.
`
`Figure 7. Determination of the biological activity of refolded and
`purified hb2m (filled symbols) and HAT-hb2m (open symbols) u