Many therapeutic monoclonal antibodies (MAbs) withvarious indications have been approved by the U.S. Food andDrug Administration and hundreds of MAbs are currently invarious stages of development. Following the approval of thefirst humanized therapeutic antibody in 1997, the humanizedantibodies trastuzumab (Herceptin) and bevacizumab(Avastin) have gone on to achieve blockbuster status.1)

Aggregation compromises the safety and efficacy of thera-peutic proteins. For example, the intravenous (i.v.) injectionof aggregated proteins can cause shock in patients, and ag-gregated antibodies are known to raise anti-idiotypic antibod-ies that lower the activity of therapeutic antibodies.2—4) Dur-ing their production, antibody solutions that contain a largepercentage of aggregates cause the fouling of antiseptic fil-ters and result in poor productivity.

Subcutaneous (s.c.) administration is the preferred routebecause this method requires a shorter time than i.v. infusion,and self-administration by patients at home is possible. Fors.c. administration, a high concentration solution of MAbs isnecessary because a large dose (100—200 mg) is required fortreatment; however, the injection volume is limited (�1.5ml).5) The development of a high-concentration MAb formu-lation is very challenging because proteins in highly concen-trated solutions tend to self-associate.6)

During their production, purification, and storage, thera-peutic proteins are subject to aggregation.7) Especially duringthe purification process, acid-induced conformationalchanges of MAbs are inevitable during their elution fromprotein-A resin at a low pH.8,9) To prevent this problem, vari-

ous approaches have been developed, e.g., milder elutionconditions,10) low affinity matrices,11,12) and the addition ofanti-aggregating reagents such as arginine.13—16)

One of the major analytical methods for the quality evalu-ation of MAbs is size-exclusion chromatography (SEC); thismethod is able to detect MAb aggregates and degradates. Inrecent years, new analytical methods have been reported thatuse markers of amyloid-like properties, and several commer-cially available biopharmaceutical products were found tocontain misfolded proteins with amyloid-like properties.17) Insome cases, the level of misfolded protein was found to in-crease upon storage under the conditions prescribed by themanufacturer.

On the other hand, the mechanisms of aggregate formationare largely unknown. Accumulating experimental data showthat proteins, especially ones with amyloidogenic properties,often contain specific sequences or structural motifs that tendto promote aggregation. Wang et al. studied the sequencesand fragment antigen-binding (Fab) structures of antibodiesfor their vulnerability toward aggregation by using sequence-based computational tools to identify potential aggregation-prone motifs. Some of these motifs are located in variabledomains, primarily in complementarity-determining regions(CDRs).18) The implication is that one possible mechanismfor MAb aggregation may be through the formation of cross-b structures and fibrils.

We incidentally obtained an anti-dinitrophenol (DNP) anti-body that was highly prone to aggregation. We reasoned thatthis antibody would be useful to develop various approaches

August 2011 1273Regular Article

Evaluation of the Aggregation States of Monoclonal Antibodies by Diverse and Complementary Methods

Tetsuya YOSHINO,*,a,b Tomoyoshi ISHIKAWA,a Takashi ISHIHARA,a Yoshimi TAKEUCHI,a

Hitomi YOSHIKAWA,c Hideaki YOSHIDA,c Hitoshi YOSHIDA,d and Kaori WAKAMATSUb

a Bio Process Research and Development Laboratories, Production Division, Kyowa Hakko Kirin Co., Ltd.; 100–1Hagiwara-machi, Takasaki, Gunma 370–0013, Japan: b Department of Chemistry and Chemical Biology, Graduate Schoolof Engineering, Gunma University; 1–5–1 Tenjin-cho, Kiryu, Gunma 376–8511, Japan: c Antibody Research Laboratories,Kyowa Hakko Kirin Co., Ltd.; and d Innovative Drug Research Laboratories, Research Division, Kyowa Hakko Kirin Co.,Ltd.; 3–6–6 Asahi-machi, Machida, Tokyo 194–8533, Japan.Received March 3, 2011; accepted May 16, 2011; published online June 6, 2011

Therapeutic monoclonal antibodies (MAbs) with high specificity and fewer adverse effects are becomingwidely used for the treatment of various diseases. MAbs need to be stored and administered at high concentra-tions in solution, the conditions under which MAbs may aggregate. As aggregated MAbs compromise theirsafety and efficacy, aggregation should be prevented; thus, it is important to analyze the aggregation states ofMAbs in detail. We obtained 2 MAbs against dinitrophenol (DNP) that exhibited different aggregation proper-ties: anti-DNP1 exhibited a much higher aggregate content (dimer or trimer) than anti-DNP2 when analyzed bysize-exclusion chromatography (SEC). As anti-DNP1 had a longer complementarity-determining region 3(CDR3) light chain than anti-DNP2 by 2 amino acid residues, we hypothesized that the increased aggregation ofDNP1 was due to these extra residues; therefore, we prepared mutant antibodies with shorter CDR3s to comparetheir aggregation properties. Anti-DNP1-DDEI, with the same CDR3 length as anti-DNP2, exhibited no aggregatesas expected. Anti-DNP1-DDI, with 1 additional residue, exhibited a smaller peak than wild-type (WT) in SEC,whereas this mutant exhibited stronger thioflavin T fluorescence than WT, which is indicative of amyloid forma-tion. In addition, the anti-DNP1-DDI solution (but not others) became opalescent at 4°C and exhibited large visibleparticles, which are undetectable by SEC. The fragment antigen-binding region of this mutant was found to havelower thermal stability than the others by differential scanning calorimetry. These data suggest that diverse ana-lytical methods should be applied to evaluate MAb aggregation, in addition to the commonly used SEC.

Key words monoclonal antibody; aggregation; evaluation; amyloid; differential scanning calorimetry; dynamic light scattering

Biol. Pharm. Bull. 34(8) 1273—1278 (2011)

© 2011 Pharmaceutical Society of Japan∗ To whom correspondence should be addressed. e-mail: tetsuya.yoshino@kyowa-kirin.co.jp

to prevent the MAb aggregation. As sequence analyses ofthis MAb indicated that it had a longer CDR3 than a non-ag-gregating MAb, we constructed mutant antibodies with ashorter CDR3 to compare their aggregation characteristics.In the present paper, diverse analytical techniques were ex-tensively employed to evaluate the aggregation of anti-DNPantibodies.

MATERIALS AND METHODS

Materials The immunoglobulin G1 (IgG1) antibodies(anti-DNP1 (wild-type; WT), anti-DNP1-DE, anti-DNP1-DI,anti-DNP1-DEI, and anti-DNP2) were expressed by Chinesehamster ovary cells and purified with protein A (AmershamBiosciences Corp., Piscataway, NJ, U.S.A.) as described pre-viously.19) Antibodies were diluted in phosphate-bufferedsaline (PBS; Gibco Industries Inc., Langley, OK, U.S.A.) atpH 7.2, unless otherwise specified. Thioflavin T (ThT) wasfrom Calbiochem (San Diego, CA, U.S.A.). The otherreagents were of analytical grade.

Size-Exclusion Chromatography The amounts of smallsoluble aggregates were analyzed by SEC-HPLC (LC-7A,Shimadzu Corp., Kyoto, Japan) with a G3000SWXL column(7.8 mm i.d.�30 cm; Tosoh Corp., Tokyo, Japan) essentiallyas described previously.20) The mobile phase contained20 mM sodium phosphate (pH 7.0) and 500 mM sodium chlo-ride. The experimental conditions were as follows: injectedprotein, 20 mg; flow rate, 0.5 ml/min; detection wavelength,215 nm; and analysis time, 30 min.

Thioflavin T Fluorescence Measurement To determinepossible amyloid formation by the antibodies, fluorescenceemission spectra of ThT (450—600 nm) were measured on aJasco FP-6500 Spectrofluorometer with an excitation wave-length at 440 nm at room temperature (ca. 25 °C). Antibodysolutions were diluted in 3 ml PBS supplemented with 10 mMThT and 20 mM Tris–HCl buffer (pH 7.4), and incubated for60 min at 37 °C.

Dynamic Light Scattering (DLS) The size distributionof the microparticles in the antibody solutions were deter-mined more precisely using a DLS photometer (Zeta SizerNano-ZS equipped with a He-Ne laser; Malvern InstrumentsLtd., Worcestershire, U.K.). Scattered light intensity was de-tected at an angle of 173°. The samples were analyzed at4 °C or 37 °C without prior filtration.

Differential Scanning Calorimetry (DSC) The thermalstability of the antibodies was analyzed using a VP-DSC dif-ferential scanning calorimeter (MicroCal, Piscataway, NJ,U.S.A.). Antibody samples (1 mg/ml) were analyzed at 5—95 °C with a heating rate of 1 °C/min.

RESULTS

Characterization of MAbs by SEC We expressed 2 dif-ferent anti-DNP MAbs found in our laboratory at high levelsin Chinese hamster ovary cells. One antibody (hereafter des-ignated as anti-DNP1) exhibited a high content (ca. 8%) ofsmall aggregates (dimer or trimer), whereas the other (anti-DNP2) exhibited negligible aggregates when analyzed bySEC at ambient temperatures. To identify the particularamino acid sequence that lead to the high aggregation ten-dency of anti-DNP1, we compared the sequences of the

heavy and light chains from both antibodies.Sequence alignment (Fig. 1) revealed that these 2 antibod-

ies differed from each other in the length of the CDR3 lightchain, but not in the length of CDR1 or CDR2. There were 2extra amino acids in the anti-DNP1 CDR3, and we hypothe-sized that its length affects the aggregation propensities ofantibodies. Hence, we generated 3 different mutant anti-DNP1 antibodies: anti-DNP1-DI, in which the hydrophobicisoleucine residue (which is expected to enhance protein ag-gregation) was deleted; anti-DNP1-DE, in which the gluta-mate residue flanking the isoleucine residue was deleted; andanti-DNP1-DEI in which these isoleucine and glutamateresidues were deleted.

As expected, anti-DNP1-DEI, with the same CDR3 lengthas anti-DNP2, exhibited no (�0.4%) aggregates (Fig. 2).Anti-DNP1-DI, with 1 extra amino acid in the CDR3 com-pared to anti-DNP2, exhibited intermediate aggregate con-tent (3.9%) between WT (7.0%) and anti-DNP1-DEI. How-ever, deletion of glutamate (anti-DNP1-DE) increased aggre-gate content to as much as 14.2%, even though this mutanthad the same CDR3 length as anti-DNP1-DI.

Evaluation of Antibody Aggregation with an Amyloid-Reactive Reagent In addition to SEC, we evaluated anti-body aggregation by fluorescence spectroscopy, which can beconducted more efficiently than SEC. ThT exhibits strongfluorescence on specific binding to amyloid fibrils, such asthose of amyloid b-peptides and b2-microglobulin, and canbe used to quantify amyloid levels.21,22)

Anti-DNP1 (WT) exhibited fluorescence at ca. 480 nm,which is characteristic of amyloid (Fig. 3). This observationindicates that the anti-DNP1 (WT) aggregates have someamyloid characteristics. Anti-DNP1-DEI exhibited little en-hancement compared with buffer (negative control), as ex-pected from the negligible amounts of aggregates observedby SEC. Anti-DNP1-DE exhibited fluorescence at a slightlylower level than WT, although the former exhibited higheraggregate content than the latter when analyzed by SEC.These observations suggest that the WT oligomers had asomewhat stronger tendency to elongate into larger amyloidfibrils than anti-DNP1-DE. On the other hand, anti-DNP1-DIexhibited much larger fluorescence (ca. 2-fold) than WT inspite of its much lower aggregate content (approximatelyhalf) than WT when analyzed by SEC. These observationssuggest that anti-DNP1-DI contains large particles that are

1274 Vol. 34, No. 8

Fig. 1. Sequence Alignment of the CDR3 Light Chains for the Anti-dinitrophenol (DNP) Antibodies

The CDR3s are in bold. Panel A, anti-DNP1 and anti-DNP2. Panel B, anti-DNP1wild-type (WT) and its mutants.

not detectable with SEC, and that these particles were sensi-tively detected by ThT fluorescence.

Low-Temperature-Induced Reversible Aggregates ofAnti-DNP1-DDI During the preparation of the 4 antibodysamples, we noticed that the anti-DNP1-DI solution becameopalescent immediately after purification at 4 °C, and thatthis opalescence was clearly reversible at ambient tempera-tures. The other antibody solutions remained clear, irrespec-tive of temperature. To analyze these phenomena in more de-tail, we measured the turbidity of the antibody solutions. Theanti-DNP1 (WT), anti-DNP1-DE, and anti-DNP1-DEI (1mg/ml) solutions exhibited null turbidity at 600 nm on ice.

On the other hand, the anti-DNP1-DI solution exhibited aturbidity of 1.22 under the same condition. Immediately afterincubation at 37 °C, the turbidity of the anti-DNP1-DI solu-tion started to decrease with a half-life of ca. 1 min; the tur-bidity was 0.04 at 5 min (Fig. 4). The turbidity of the solutionreturned to 1.18 after incubation on ice for 15 min. Thus,anti-DNP1-DI formed reversible aggregates at low tempera-tures.

Size Distribution of Antibody Aggregates as Analyzedby DLS SEC is an authentic analytical method for deter-mining aggregate content in immunoglobulin G (IgG) solu-tions; monomers with a diameter of ca. 10 nm elute well sep-

August 2011 1275

Fig. 2. Size-Exclusion Chromatography Chromatograms of the Anti-dinitrophenol (DNP) 1 Antibody and Its Mutants at Ambient Temperatures

(A) Wild-type; (B) anti-DNP1-DE; (C) anti-DNP1-DI; (D) anti-DNP1-DEI. Peaks for the dimer/trimer are annotated with arrows.

Fig. 3. Thioflavin T Fluorescence Spectra of the Anti-dinitrophenol(DNP) 1 Antibody and Its Mutants

Each trace indicates (from top to bottom): anti-DNP1-DI, anti-DNP1-wild-type(WT), anti-DNP1-DE, anti-DNP1-DEI, and phosphate-buffered saline (PBS) (negativecontrol).

Fig. 4. Turbidity Decrease of Anti-dinitrophenol (DNP)1-DI Solution onWarming

Ice-cold solution of anti-DNP1-DI (1 mg/ml in phosphate-buffered saline) waswarmed at 37 °C and its turbidity at 600 nm was measured at the indicated times. Theright-most data point indicates the turbidity of the solution that was re-chilled for15 min on ice.

arated from dimers and trimers.6,20) However, SEC is not ap-plicable to the detection of much larger aggregates. The ag-gregates of anti-DNP antibodies were investigated by DLSmeasurements, which give information on the size distribu-tion from monomers (10 nm) to large aggregates (1000 nm).Since we observed that the turbidity of the solution (anti-DNP1-DI) changed by incubation at 37 °C, we analyzed thediameter distributions of the 4 antibodies by DLS at 4 °C and37 °C.

Anti-DNP1 (WT) exhibited 2 peaks at 37 °C: 1 peak at ca.14 nm, as expected from the hydrodynamic diameter of IgGmonomers, and the other at ca. 78 nm, corresponding to ag-gregates much larger than dimers or trimers (Fig. 5A). Thescattering intensity of the peak at 78 nm was higher than thatof the monomeric peak. At 4 °C, the positions of the 2 peaksremained unchanged, but the relative intensity of the aggre-gates increased. It should be noted, in DLS measurements,that dimers and trimers cannot be observed as a single peakseparate from monomers and that they contribute to the in-tensity of the peak corresponding to the monomers.

Anti-DNP1-DEI also exhibited 2 peaks at 37 °C; however,the intensity of the aggregates (with a diameter of 185 nm)was much lower than for WT (Fig. 5D), in agreement withthe low aggregate content of the mutant as observed by SECand ThT fluorescence measurements. The size distributionremained unchanged at 4 °C. Anti-DNP1-DE exhibited asimilar size distribution as anti-DNP1-DEI, except that thediameter of the aggregates was larger (380 nm; Fig. 5B).

Anti-DNP1-DI exhibited no aggregate peak in the 70—1000 nm range at 37 °C (Fig. 5C), in contrast to the other antibodies. However, this mutant generated huge particleswhose size could not be evaluated accurately by DLS meas-urements. Interestingly, at 4 °C, this mutant exhibited a singlepeak at ca. 1100 nm, but no peak corresponding to themonomer. This observation was in agreement with the high

turbidity of the mutant at a low temperature and suggests thatanti-DNP1-DI has different aggregation mechanisms fromthe other antibodies.

Stability Analysis by DSC Anti-DNP1-DI is expectedto have lower temperature stability because this mutantgreatly changes its aggregation state depending on the tem-perature. We analyzed the temperature stability of our anti-bodies by DSC, which provides information on the stabilityof each domain in a multi-domain protein, such as antibod-ies. The effects of the length of CDR3 on the stability of Fabwere also of interest.

Each thermogram of anti-DNP1 (WT), anti-DNP1-DE,and anti-DNP1-DEI consisted of 2 transitions, centered at ca.70 °C and 83 °C (Fig. 6). For these antibodies, the first transi-tion had a larger amplitude than the second one. These ther-mograms are similar to those for the humanized monoclonalantibody bevacizumab against vascular endothelial growthfactor (Avastin)23) and a murine IgG2a antibody against HIV-

1276 Vol. 34, No. 8

Fig. 5. Particle Size Distribution of the Anti-dinitrophenol (DNP) 1 Antibody and Its Mutants at 4 °C (Red Trace) and 37 °C (Green Trace)

(A) Wild-type (WT); (B) anti-DNP1-DE; (C) anti-DNP1-DI; (D) anti-DNP1-DEI.

Fig. 6. Differential Scanning Calorimetry Thermograms of the Anti-di-nitrophenol (DNP) 1 Antibody and Its Mutants

Black trace, wild-type (WT); magenta, anti-DNP1-DE; green, anti-DNP1-DI; cyan,anti-DNP1-DEI.

1 capsid protein p24 (CB4-1).8) The first and second transi-tions were assigned, respectively, to the unfolding of the Faband fragment crystallizable (Fc) regions by DSC analyses ofeach fragment.23)

Although the Fab region was the same among the 4 anti-bodies, except for the CDR3 length, their melting points (Tm)differed in the following order: anti-DNP1-DEI (74.63°C)�anti-DNP1 (WT, 71.57 °C)�anti-DNP1-DE (70.56°C)�anti-DNP1-DI (70.28 °C). These observations suggestthat the stability of the Fab region is affected by the length ofthe CDR3 light chain.

For anti-DNP1-DEI, the first transition was increased by3 °C compared with the WT, which indicates the increasedstability of Fab (and CH2) by the deletion of the 2 residues.This observation is in good agreement with the low aggre-gate content of this antibody.

For anti-DNP1-DI, the thermogram exhibited an extrapeak at 58 °C compared with the other antibodies. The de-crease in the intensity of this peak at 70 °C and its concomi-tant emergence at 58 °C indicate that the latter peak is as-cribed to Fab and the former one to CH2. The lower unfold-ing temperature of Fab in anti-DNP1-DI indicates that thedeletion of the isoleucine residue destabilized the Fab. Theseobservations confirm that the first large transition of anti-DNP1 (WT), anti-DNP1-DE, and anti-DNP1-DEI corre-sponds to the Fab and CH2 fragments. On the other hand, thesecond transition (CH3) remained unchanged for all antibod-ies, suggesting that the stability of Fab (and CH2) does notaffect that of CH3.

Characterization of Anti-DNP Antibodies by Light Mi-croscopy To further characterize the large particles foundin the anti-DNP1-DI solution at low temperatures, antibodysolutions were analyzed by light microscopy. Anti-DNP1-DIexhibited many particles as large as 50 mm in diameter (Fig.7A). Although we expected fibrillar structures from the highThT fluorescence of this mutant, the particles were amor-phous. These particles disappeared almost completely afterincubation at 37 °C for 10 min (Fig. 7B). No particles wereobserved for anti-DNP1 (WT), anti-DNP1-DE, and anti-DNP1-DEI, irrespective of temperature (data not shown), asexpected from their turbidity and DLS measurements.

DISCUSSION

The high aggregation property of the anti-DNP1 antibodywas eliminated by deleting the 2 extra residues (EI) in itsCDR3 light chain compared with that of anti-DNP2, whichexhibited low levels of aggregation. Adjusting the length ofthe CDR3 light chain may therefore represent an effectivetechnique for preventing antibody aggregation.

However, the length of CDR3 may not be the sole determi-nant of the aggregation property because anti-DNP1-DE andanti-DNP1-DI, with the same CDR3 length (1 extra residuecompared to anti-DNP2), exhibited quite different aggrega-tion behavior: anti-DNP1-DE and anti-DNP1-DI, respec-tively, exhibited a larger and smaller dimer/trimer peak thanWT in SEC, whereas they had lower and higher amyloidproperties than WT. Anti-DNP1-DI, but not anti-DNP1-DE,formed large visible particles at low temperatures. These dif-ferent behaviors may be related to the lower thermal stabilityof anti-DNP1-DI as revealed by DSC.

The different aggregation properties of the anti-DNP anti-bodies and their mutants will serve as good models for theanalysis of the aggregation mechanisms of MAbs and the de-velopment of effective methods/reagents that prevent anti-body aggregation.

In addition to the authentic SEC method, we evaluated theaggregation of anti-DNP antibodies by diverse analyticalmethods: SEC, DLS, DSC, turbidity, light microscopy, andThT fluorescence. Although SEC is quite useful for deter-mining dimer/trimer content, this method cannot detect ag-gregates larger than 100 Å. For the evaluation of large aggre-gates, DLS can be used to obtain information on their sizedistribution, although not in a high-throughput fashion. Analternative means to analyze large aggregates would be ThTfluorescence measurement, which can be conducted in ahigh-throughput manner by using microplate readers.24) Infact, anti-DNP1-DI, which showed a lower dimer/trimer con-tent than WT and anti-DNP1-DE, had higher ThT fluores-cence levels than these antibodies, and this discrepancy inantibody behavior prompted us to analyze their size distribu-tion by DLS. DSC was another useful method that can alsobe conducted in a high-throughput manner. This method candetect antibodies with low thermal stability that have poten-tial aggregation properties. It should also be noted that analy-ses at multiple temperatures (e.g., 4, 37 °C) provides impor-tant information that cannot be obtained at a single tempera-ture, e.g., the high aggregation properties of anti-DNP1-DIwere not observed at 37 °C. We evaluated the aggregationproperties of anti-DNP antibodies with different methods andfound that all of the DLS, ThT fluorescence, and DSC analy-ses consistently suggest the high aggregation property ofanti-DNP1-DI, in contrast to the SEC analysis. Thus, at leastThT fluorescence analysis should be executed concurrentlywith SEC one, taking advantage of its ability to be done effi-ciently with microplate readers. Nevertheless, we recom-mend that the exhaustive analyses as carried out in this studyshould be conducted, especially during the early stages oftherapeutic MAb development.

Acknowledgment We thank Mr. Kazumasa Hasegawa,Mr. Kaname Kimura, Mr. Takafumi Tomura, and Mr. TakafumiIwura for valuable discussions and experimental suggestions.

August 2011 1277

Fig. 7. Light Microscopy Images of the Anti-dinitrophenol (DNP) 1-DIAntibody Solution

(A) 4 °C; (B) 37 °C (scale bar: 20 mm).

REFERENCES

1) Reichert J. M., Rosensweig C. J., Faden L. B., Dewitz M. C., Nat.Biotechnol., 23, 1073—1078 (2005).

2) Fradkin A. H., Carpenter J. F., Randolph T. W., J. Pharm. Sci., 98,3247—3264 (2009).

3) Carpenter J. F., Randolph T. W., Jiskoot W., Crommelin D. J., Mid-daugh C. R., Winter G., Fan Y. X., Kirshner S., Verthelyi D., Ko-zlowski S., Clouse K. A., Swann P. G., Rosenberg A., Cherney B., J.Pharm. Sci., 98, 1201—1205 (2009).

4) Hermeling S., Schellekens H., Maas C., Gebbink M. F., Crommelin D.J., Jiskoot W., J. Pharm. Sci., 95, 1084—1096 (2006).

5) Shire S. J., Shahrokh Z., Liu J., J. Pharm. Sci., 93, 1390—1402 (2004).6) Nishi H., Miyajima M., Nakagami H., Noda M., Uchiyama S., Fukui

K., Pharm. Res., 27, 1348—1360 (2010).7) Cromwell M. E., Hilario E., Jacobson F., AAPS J., 8, E572—E579

(2006).8) Welfle K., Misselwitz R., Hausdorf G., Höhne W., Welfle H., Biochim.

Biophys. Acta, 1431, 120—131 (1999).9) Thies M. J., Kammermeier R., Richter K., Buchner J., J. Mol. Biol.,

309, 1077—1085 (2001).10) Varady L., Kalghatgi K., Horvath C., J. Chromatogr., 458, 207—215

(1988).11) Gülich S., Uhlén M., Hober S., J. Biotechnol., 76, 233—244 (2000).12) Li R., Dowd V., Stewart D. J., Burton S. J., Lowe C. R., Nat. Biotech-

nol., 16, 190—195 (1998).13) Arakawa T., Philo J. S., Tsumoto K., Yumioka R., Ejima D., Protein

Expr. Purif., 36, 244—248 (2004).14) Tsumoto K., Ejima D., Kita Y., Arakawa T., Protein Pept. Lett., 12,

613—619 (2005).15) Arakawa T., Tsumoto K., Nagase K., Ejima D., Protein Expr. Purif.,

54, 110—116 (2007).16) Ejima D., Yumioka R., Tsumoto K., Arakawa T., Anal. Biochem., 345,

250—257 (2005).17) Maas C., Hermeling S., Bouma B., Jiskoot W., Gebbink M. F., J. Biol.

Chem., 282, 2229—2236 (2007).18) Wang X., Das T. K., Singh S. K., Kumar S., mAbs, 1, 254—267

(2009).19) Nakagawa T., Ishihara T., Yoshida H., Yoneya T., Wakamatsu K.,

Kadoya T., J. Sep. Sci., 33, 2045—2051 (2010).20) Ishikawa T., Kobayashi N., Osawa C., Sawa E., Wakamatsu K., Biol.

Pharm. Bull., 33, 1043—1046 (2010).21) LeVine H. 3rd, Protein Sci., 2, 404—410 (1993).22) Ohhashi Y., Kihara M., Naiki H., Goto Y., J. Biol. Chem., 280,

32843—32848 (2005).23) Ionescu R. M., Vlasak J., Price C., Kirchmeier M., J. Pharm. Sci., 97,

1414—1426 (2008).24) Lockhart A., Ye L., Judd D. B., Merritt A. T., Lowe P. N., Morgenstern

J. L., Hong G., Gee A. D., Brown J., J. Biol. Chem., 280, 7677—7684(2005).

1278 Vol. 34, No. 8