Professional Documents
Culture Documents
Biotechnology
Applications of Differential Scanning Calorimetry for Thermal
Stability Analysis of Proteins: Qualification of DSC
JIE WEN,1 KELLY ARTHUR,2 LETHA CHEMMALIL,3 SALMAN MUZAMMIL,4 JOHN GABRIELSON,2 YIJIA JIANG1
1
Product Attribute Sciences, Amgen Inc., Thousand Oaks, California 91320
2
Analytical Sciences, Amgen Inc., Longmont, Colorado 80503
3
Process Development, Amgen Inc., West Greenwich, Rhode Island 02817
4
Biophysics & Developability, Antibody Drug Discovery, Centocor Inc. Radnor, Pennsylvania 19406
ABSTRACT: Differential scanning calorimetry (DSC) has been used to characterize protein
thermal stability, overall conformation, and domain folding integrity by the biopharmaceutical
industry. Recently, there have been increased requests from regulatory agencies for the qual-
ification of characterization methods including DSC. Understanding the method precision can
help determine what differences between samples are significant and also establish the accep-
tance criteria for comparability and other characterization studies. In this study, we identify the
parameters for the qualification of DSC for thermal stability analysis of proteins. We use these
parameters to assess the precision and sensitivity of DSC and demonstrate that DSC is suitable
for protein thermal stability analysis for these purposes. Several molecules from different struc-
tural families were studied. The experiments and data analyses were performed by different
analysts using different instruments at different sites. The results show that the (apparent)
thermal transition midpoint (Tm ) values obtained for the same protein by same and different
instruments and/or analysts are quite reproducible, and the profile similarity values obtained
for the same protein from the same instrument are also high. DSC is an appropriate method
for assessing protein thermal stability and conformational changes. 2011 Wiley Periodicals,
Inc. and the American Pharmacists Association J Pharm Sci 101:955964, 2012
Keywords: calorimetry (DSC); thermal analysis; proteins; biotechnology; method
qualification; protein structure
industry.47 However, until recently no formal evalu- Table 1. Proteins Used for the Studies
ation of the method precision has been carried out. Protein Structural Family
Understanding the method precision can help deter-
Protein 1 IgG 2
mine what differences between samples are signif-
Protein 2 Fc conjugatea
icant and also establish the acceptance criteria for Protein 3 Fc fusion proteinb
comparability and other characterization studies. Protein 4 Globular protein containing primarily beta
To assess the precision of DSC for thermal sta- sheet structure (MW < 50 kDa)
bility analysis, five proteins from different struc- Protein 5 Four-helical-bundle cytokine
Protein 6 IgG 1
tural families were studied using three different
Protein 7 Fc conjugate
instrument models: the MicroCal VP-DSC, the Micro- Protein 8 Helical bundle
Cal VP-Capillary DSC (MicroCal LLC, Northampton, Protein 9 IgG 2
Massachusetts), and the TA Instruments Q2000 DSC Protein 10 Globular protein (MW < 50 kDa)
(TA Instruments, New Castle, Delaware). Replicate Protein 11 Fc conjugate
Protein 12 Fc conjugate
measurements using the same experimental parame-
ters and procedures were carried out for each protein a A molecule formed when a peptide is fused to the Fc is referred to as
an Fc conjugate here.
at the specified concentrations in the corresponding b A molecule formed when a protein is fused to the Fc is referred to as
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012 DOI 10.1002/jps
APPLICATIONS OF DSC FOR THERMAL STABILITY ANALYSIS OF PROTEINS 957
Five proteins from different structural families erence profile. A value of 100% indicates an identical
were studied. Replicate measurements using the match. This numerical analysis approach was used
same experimental parameters and procedure were for the DSC profile similarity analysis of all studies
executed for each protein in the native state at speci- described in this paper.
fied concentrations. The experiments were conducted
by different analysts at different sites on different
RESULTS AND DISCUSSION
dates. Intraday versus interday precision and concen-
tration effect were also evaluated. Qualification of DSC: Precision Assessment
Sensitivity Study Replicate Measurements of the Same Protein Using the
Same Instrument
Selected proteins were analyzed in their stable stor-
age buffers, C3N buffer, and in the presence of 6 M The precision of DSC was first assessed by replicate
Gdn. The profiles of these proteins in low pH buffer measurements of the same protein using the same
or 6 M Gdn were compared with that of the control instrument. Proteins from different structural fami-
(folded protein in stable storage buffer). The sensitiv- lies were selected for analysis. Specifically, protein 1
ity of the DSC method could be assessed by spiking [immunoglobulin G (IgG) 2], protein 2 (Fc conjugate),
the denatured species into a solution of native protein protein 3 (Fc fusion protein), protein 4 (small glob-
and correlating the profile similarity with the per- ular protein containing primarily beta sheet struc-
centage of denatured species present in the sample. ture), and protein 5 (four-helical-bundle cytokine)
The challenge for the experimental mixing approach were evaluated. The Tm values were obtained by Ori-
is that the unfolding of a protein by low pH or Gdn is gin and TA universal analysis software. The similar-
often reversible or partially reversible. Thus, a stan- ity values were obtained by OMNIC softwareQC
dard curve that correlates the changes in DSC profile compare function (Thermo Electron).14,15 The overlay
with the percentage of a denatured protein at various of the DSC scans of replicate measurements of pro-
mass ratios cannot be generated empirically because tein 2 from the same instrument, shown in Figure 1,
after mixing the samples, it is virtually impossible to is representative of the data obtained. There are two
determine what percentage of native protein remains apparent endothermic thermal transitions at approx-
as well as the state of the unfolded species in the sam- imately 57.4 C and 77.9 C corresponding to the un-
ples. To address this issue, we carried out a blending folding of the CH 2 and CH 3 domains. The DSC profiles
study using two folded proteins of different structures of protein 2 are typical of an Fc conjugate, with the two
as surrogates for the native and unfolded forms of transitions demonstrating that the protein is properly
the same protein. The DSC profiles of the two pro- folded into the distinct domains. The standard devia-
teins mixed in defined buffers at different mass ratios tions (SDs) of this set of replicate measurements for
were obtained empirically. Computer-simulated pro- each transition are 0.1 C and 0.02 C, respectively, in-
files were obtained by mathematically blending the dicating that the DSC method is quite reproducible.
profiles of the individual proteins at the same mass Protein 2 is an Fc conjugate. The unfolding of the
ratios. The profiles obtained empirically were then CH 2 and CH 3 domains of an Fc conjugate is often well
compared with the mathematically obtained profiles resolved. However, for monoclonal antibodies (mAbs),
at the same mass ratios. This ensures that the em- the unfolding of the Fab domain may overlap (or even
pirical mixing and computer mixing approaches are distort) that of the CH 2 and/or CH 3 domain. Therefore,
equivalent and verifies that the computer simulation the DSC profiles of mAbs usually show overlapped
of mixing can be used to assess the sensitivity of DSC. peaks. The typical unfolding DSC profiles of mAbs
can be found in Ref. 7.
Data Analysis
The summary of Tm values of replicate measure-
The values of Tm were obtained by Origin (Origin- ments of the same protein using the same instrument
Lab Corp, Northampton, Massachusetts) and TA uni- is shown in Table 2, and the summary of profile sim-
versal analysis software (Thermo Electron, Madison, ilarity values of replicate measurements of the same
Wisconsin).1013 The similarity of the DSC profiles protein using the same instrument is shown in Ta-
of the replicate measurements was analyzed using a ble 3. There are two apparent endothermic thermal
quantitative comparison method from Thermo Elec- transitions observed for proteins 1, 2, and 4, three ap-
tron (Madison, Wisconsin) (OMNIC softwareQC parent endothermic thermal transitions for protein 3,
compare function).14,15 The method correlates the and one for protein 5. The Tm values are shown in sep-
overall profile information in a specified region of both arate columns according to the individual transitions.
sample and reference profiles to determine the sim- The similarity values are listed for the comparison of
ilarity between the two. The result of the analysis the overall profile. The corresponding SDs of the repli-
is reported as a value between 0% and 100%, which cate measurements are also listed in Tables 2 (indi-
indicates how well the sample profile matches the ref- vidual transitions) and 3 (overall profile). The SDs of
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012
958 WEN ET AL.
the Tm values range from 0.01 C to 1.13 C and the The thermal transition temperatures are relatively
SDs of profile similarity values range from 0.04% to constant and reproducible regardless of site, instru-
3.76%, suggesting that the precision of replicate DSC ment, or analyst. The profile similarity values ob-
measurements from the same instrument is very good tained from the same instrument are also good. How-
regardless of the differences in protein structure fam- ever, similarity analysis of the DSC profiles of the
ilies. same protein using different instruments has large
SDs. The major sources of poor precision using pro-
Replicate Measurements of the Same Protein file similarity between instruments could be due to
Using Different Instruments the differences in the melting profile shapes and the
The precision of the DSC analysis was also assessed operation environments. Even with proper DSC cali-
by replicate measurements of the same protein using brations, only temperature and H (enthalpy change)
different instruments. The summary of Tm and sim- are calibrated. The unfolding profile (or peak shape)
ilarity values of the replicate measurements of the is not calibrated. The same peak area may have quite
same protein using different instruments is listed in different profiles. Therefore, it is possible that with
Table 4. different instruments, some factors such as slight dif-
The SDs of the Tm values for the same protein range ferences in sample cell configurations and so on may
from 0.38 C to 1.49 C, which is only slightly greater affect how the protein unfolds and cause DSC pro-
than those observed using the same instrument. This files to be different. Therefore, in most cases, Tm is
suggests that the precision of the Tm values of the a preferred parameter to determine protein thermal
same protein using different instruments is also good. stability changes.
Compared with the profile similarity values mea- There is a significant difference in using MicroCal
sured using the same instrument, those using dif- VP-DSC and MicroCal VP-Capillary DSC (MicroCal
ferent instruments vary dramatically (Table 4). The LLC) with regard to the detection of protein precipi-
profiles from the TA instrument are significantly dif- tation. A MicroCal VP-DSC (MicroCal LLC) typically
ferent from those of the MicroCal instruments (Micro- shows an exothermic signal after a protein precip-
Cal LLC). Even comparing the profiles from the same itates, but a MicroCal VP-Capillary DSC (MicroCal
manufacturer, MicroCal, the similarity values and LLC) does not show such an exothermic signal when
SDs are much poorer than those obtained on the same the same protein is studied. In conditions where pro-
instrument. The similarity values range from 44.2% tein precipitation are detected using MicroCal VP-
to 88.9% and the SDs range from 4.4% to 22.8%. Re- DSC (MicroCal LLC), we typically saw minimal im-
sults suggest that comparison of thermal melting pro- pact for Tm comparison, but more significant impact
files from different DSC instruments/manufacturers for profile similarity comparison. This is consistent
should not be used to determine protein confirmation with our observations that precipitation often occurs
and stability changes. around or after the middle point of a thermal transi-
tion; thus, it has less impact on Tm and more impact
on the overall profile.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012 DOI 10.1002/jps
APPLICATIONS OF DSC FOR THERMAL STABILITY ANALYSIS OF PROTEINS 959
Table 2. Summary of Tm Values and SDs of Replicate Measurements of the Same Protein Using the Same Instrument
Replicate Measurements of the Same Protein Sample at can be protein concentration dependent, affecting the
Different Concentrations Using the Same Instrument or unfolding and apparent Tm . Therefore, proteins with
Different Instruments different thermodynamic properties may show differ-
ent Tm results when measured at significantly differ-
The protein concentrations used for the TA in-
ent protein concentrations. In addition, some proteins
strument are quite different from the concentra-
may show significantly different interactions at differ-
tions used for the MicroCal instruments (25 vs.
ent concentrations (0.5 vs. 25 mg/mL). In such cases,
0.51.0 mg/mL). Even though the five proteins stud-
the Tm values measured at different concentrations
ied here showed similar Tm values for the TA instru-
can also be different.
ment, wherein proteins were measured at 25 mg/mL,
and the MicroCal instruments (MicroCal LLC), Qualification of DSC: Sensitivity Assessment
wherein the same proteins were measured at
Comparison of Native Proteins with their pH 3 and 6 M
0.51.0 mg/mL, the Tm values of a protein at differ-
Gdn Denatured Forms
ent concentrations could still be different depend-
ing on specific proteins. This is due to the fact that The DSC scans of four selected proteins (protein 6,
proteins may irreversibly aggregate after thermally IgG1; protein 1, IgG2; protein 7, Fc conjugate; protein
induced unfolding, and this irreversible aggregation 8, helical bundle) in different stable storage buffers
Table 3. Summary of Similarity Values and SDs of Replicate Measurements of the Same Protein Using the Same Instrument
Table 4. Summary of Tm and Similarity Values of Replicate Measurements of the Same Protein Using Different Instruments
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012
960 WEN ET AL.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012 DOI 10.1002/jps
APPLICATIONS OF DSC FOR THERMAL STABILITY ANALYSIS OF PROTEINS 961
Figure 3. DSC scans of four selected proteins(a) protein 6, (b) protein 1, (c) protein 7, and
(d) protein 8in three different buffers: formulation (black), C3N (red), and 6 M Gdn (blue).
The offsets of some scans are adjusted for clarity. The experiment was performed on a MicroCal
VP-Capillary DSC.
mass ratios cannot be generated empirically due to the same mixing ratios. This suggests that the im-
the changes in the unfolded intermediate (structure pact of irreversible reactions is negligible in this case.
and/or amount) after mixing. To address this issue, we This result demonstrates that the computer simula-
blended two stable native proteins of different struc- tion method may be used for evaluating the sensi-
tures as surrogates for the native and unfolded forms tivity of DSC. From the results of the protein 9 and
of the same protein. protein 10 mixing study, we can detect the presence
The DSC profiles of two proteins (protein 9 and of as low as 10% of denatured protein using the DSC
protein 10) mixed in the same buffer empirically and method. In both cases, the similarity values in Ta-
mathematically at different mass ratios are shown in ble 7 are below 95% (precision of the method). There
Figure 4. The profile similarity results are presented are some assumptions and limitations in this method:
in Table 7. The results show that the computer mix- The impact of irreversible reactions should be negli-
ing profiles and similarity values agree very well with gible, the proteins should unfold separately and inde-
the experimental profiles and similarity values ob- pendently, and the same buffer should be used for both
tained empirically, with less than 5% difference at proteins.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012
962 WEN ET AL.
APPLICATIONS
To further demonstrate that DSC is suitable for com-
paring the thermal stability of different proteins in
the same condition and the same protein in different
conditions, several examples are shown below.
pH Stability of Candidates
pH stability is one of the important aspects
that should be evaluated during protein product
development.6 The DSC profiles of protein 11 at dif-
ferent pHs in citrate buffer are shown in Figure 5.
Figure 4. Comparison of the computer mixing (a) versus The changes in Tm are greater than the instrument
experimental mixing DSC profiles (b) of protein 9 and pro-
variability, demonstrating that the DSC method is
tein 10 samples. Black, 100% protein 9; green, 5% protein
able to detect the changes in thermal stability and
9 + 95% protein 10; cyan, 10% protein 9 + 90% protein 10;
blue, 50% protein 9 + 50% protein 10; dark yellow, 75% overall conformation of the protein. The DSC profiles
protein 9 + 25% protein 10; red, 100% protein 10. The ex- in Figure 5 indicate that the order of pH-dependent
periment was performed on a MicroCal VP-Capillary DSC. thermal stability for this protein is pH 7 pH 6 >
pH 5 > pH 4. A dramatic decrease in stability in cit-
rate buffer occurred at pH 5 and below. The unfolded
protein became very soluble at pH 3 and pH 2.6
Table 7. Comparison of the Profile Similarity of the Computer Mixing Versus Experimental Mixing Samples of the Protein 9
and Protein 10
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012 DOI 10.1002/jps
APPLICATIONS OF DSC FOR THERMAL STABILITY ANALYSIS OF PROTEINS 963
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012
964 WEN ET AL.
instrument on different days show no significant dif- Reese E, Spotts S, Eds. Northampton, Massachusetts: Micro-
ferences. Cal LLC, pp 93108.
Our sensitivity studies demonstrate that DSC is 6. Wen J, Jiang Y, Narhi L. 2007. Applications of DSC for anti-
bodies and Fc-conjugated proteins. Am Pharm Rev 10:1015.
able to detect pH- and denaturant-induced changes 7. Wen J, Jiang Y, Narhi L. 2008. Effect of carbohydrate on ther-
in protein thermal stability, and it is appropriate to mal stability of antibodies. Am Pharm Rev 11:98104.
use DSC when assessing the thermal stability of a 8. Hahn R, Schlegel R, Jungbauer A. 2003. Comparison of protein
protein under these denatured conditions. We may be A affinity sorbents. J Chromatogr B 790:3551.
9. Zhou JX, Tressel T, Hong T, Li F, Yang X, Lee B. (2007) Op-
able to detect as low as 10% denatured protein using
timization of antibody processing: Upstream and downstream
the DSC method. in advances in large scale biomanufacturing and scale-up pro-
The applications of DSC in pH screening, buffer duction, Chapter 20. 2nded. ASM Press (American Society for
screening, and candidate screening for formulation, Microbiology) and BioPlan Associates, Rockville, Maryland.
storage conditions, and downstream processing have 10. MicroCal VP-Capillary DSC system users manual. Northamp-
ton, Massachusetts: MicroCal LLC. 2005.
further demonstrated that information from DSC
11. MicroCal VP-DSC MicroCalorimeter users manual.
thermal stability analysis can be applied to compara- Northampton, Massachusetts: MicroCal LLC. 2005.
bility and characterization studies as well as to eval- 12. MicroCal DSC Data Analysis in Origin, version 7.0. Northamp-
uate conformational stability of proteins. ton, Massachusetts: MicroCal LLC. 2004.
13. TA Instruments Q Series Manuals. New Castle, Delaware: TA
Instruments. 2006.
ACKNOWLEDGMENTS 14. Cover, TM, Hart PE. 1967. Nearest neighbor pattern classifi-
cation. IEEE Trans Inf Theory 13:2127.
The authors thank Linda Narhi, Brent Kendrick, 15. Thermo OMNIC software manual. Waltham, Massachusetts:
Vladimir Razinkov, and David Brems for their con- Thermo Fisher Scientific Inc.
tributions and helpful discussions to this study and 16. Welfle K, Misselwitz R, Hausdorf G, Hohne W, Welfle H.
1999. Conformation, pH-induced conformational changes, and
manuscript.
thermal unfolding of anti-p24 (HIV-1) monoclonal antibody
CB4-1 and its Fab and Fc fragments. Biochim Biophys Acta
REFERENCES 1431:120131.
17. Tischenko VM, Abramov VM, Zavyalov VP. 1998. Investiga-
1. Cooper A, Nutley MA, Wadood A. 2000. Differential scan- tion of the cooperative structure of Fc fragments from myeloma
ning microcalorimetry. In Proteinligand interactions: Hydro- immunoglobulin G. Biochemistry 37:55765581.
dynamics and calorimetry; Harding SE, Chowdhry BZ, Eds. 18. Vermeer AW, Norde W, van Amerongen A. 2000. The unfold-
Oxford, New York: Oxford University Press, pp 287318. ing/denaturation of immunogammaglobulin of isotype 2b and
2. Privalov PL. 1979. Stability of proteins: Small globular pro- its F(ab) and F(c) fragments. Biophys J 79:21502154.
teins. In Advances in protein chemistry; Anfinsen CB, Edsall 19. Vermeer AW, Norde W. 2000. The thermal stability of im-
JT, Richards FM, Eds. Vol. 33. New York: Academic Press, munoglobulin: Unfolding and aggregation of a multi-domain
pp 167241. protein. Biophys J 78:394404.
3. Privalov PL. 1982. Stability of proteins: Proteins which do not 20. Garber E, Demarest SJ. 2007. A broad range of Fab stabili-
present a single cooperative system. In Advances in protein ties within a host of therapeutic IgGs. Biochem Biophys Res
chemistry; Anfinsen CB, Edsall JT, Richards FM, Eds. Vol. 35. Commun 355:751757.
New York: Academic Press, pp 1101. 21. Jefferis R, Lund J, Pound J. 1998. IgGFc mediated effector
4. Remmele RL, Nightlinger NS, Srinivasan S, Gombotz, WR. functions: Molecular definition of interaction sites for effector
1998. Interleukin-1 receptor (IL-1R) liquid formulation devel- ligands and the role of glycosylation. Immunol Rev 163:50
opment using differential scanning calorimetry. Pharm Res 76.
15:200208. 22. Mimura Y, Church S, Ghirlando R, Ashton PR, Dong S, Goodall
5. Wen J, Jiang Y, Narhi L, Hymes K, Gong K. 2007. Correlation M, Lund J, Jefferis R. 2000. The influence of glycosylation
between thermal stability and protein stability. In Proceedings on the thermal stability and effector function expression of
of the 2007 Current Trends in Microcalorimetry Conference; human IgG1Fc. Mol Immunol 37:697706.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 3, MARCH 2012 DOI 10.1002/jps