Protein Expression and Purication 91 (2013) 192206

Contents lists available at ScienceDirect

Protein Expression and Purication

journal homepage: www.elsevier.com/locate/yprep

Review

Optimization of protein purication and characterization using

Thermouor screens
Stephane Boivin , Sandra Kozak, Rob Meijers
European Molecular Biology Laboratory (EMBL), Hamburg Outstation, Notkestrasse 85, 22603 Hamburg, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The efcient large scale production of recombinant proteins depends on the careful conditioning of the
Received 5 June 2013 protein as it is isolated and puried to homogeneity. Low protein stability leads to low purication yields
and in revised form 29 July 2013 as a result of protein degradation, precipitation and folding instability. It is often necessary to go through
Available online 12 August 2013
several iterations of trial-and-error to optimize the homogeneity, stability and solubility of the protein
sample. We have set up Thermouor assays to identify customized protocols for the preparation and
Keywords: characterization of individual protein constructs. We apply a two-step approach: we rst screen for glo-
Thermouor
bal parameters, followed by a search for protein-specic additives. The rst screen has been designed in
Differential scanning uorimetry
Fluorescence thermal shift assay
such a way, that it is possible to discern global stability trends according to pH, salt concentration, buffer
Thermodenaturation type and concentration. The second screen contains small molecules that can affect the folding, aggrega-
Protein stabilization tion state and solubility of the protein construct and also includes small molecules that specically bind
Buffer optimization and stabilize proteins. The screens are designed to evaluate purication and storage protocols, and aim to
Additive screen provide hints to optimize these protocols. The home-made screens have been tested on more than 200
Thermostability different protein constructs at the Sample Preparation and Characterization (SPC) facility at EMBL Ham-
Melting point burg. We describe which RT-PCR machines can be adapted to perform Thermouor assays, what are the
Crystallization
necessary experimental conditions to set up a screen, some leads on how to interpret the data and we
give several examples of Thermouor applications beyond stability screens.
2013 Elsevier Inc. All rights reserved.

Contents

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Thermofluor principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
RT-PCR machines compatible with Thermofluor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Choosing the fluorophore . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Preparing your sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Setting up the assay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Performing the Thermofluor assay. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
High-throughput Thermofluor at the EMBL Hamburg facility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Buffer Screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Additive Screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Data analysis and melting curve profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Plotting the melting curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Single curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Complex curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Troubleshooting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Featureless curve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
High initial fluorescence signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Optimizing the assay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

Abbreviations: ANS, Anilinonaphthalene-sulfonate; DSF, Differential Scanning Fluorimetry; DSC, Differential Scanning Calorimetry; TF, Thermouor; RT-PCR, Reverse
transcription polymerase chain reaction; Tm, Melting temperature.
Corresponding author. Fax: +49 4089902149.
E-mail address: s.boivin@embl-hamburg.de (S. Boivin).

1046-5928/$ - see front matter 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.pep.2013.08.002
 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206 193

Thermofluor applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

Screening the effect of mutations on stability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Measuring kinetics of ligand binding by Thermofluor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Thermofluor and the crystallizability of protein samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Thermofluor online resources. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Introduction There are a number of techniques that can be used to monitor

the effect of external factors on the thermostability of the sample.
The characterization of individual proteins often requires that
they are produced in quantities that go far beyond their abundance
in the cell. They are isolated in an environment that is devoid of
potentially stabilizing factors. It is not surprising that many protein
samples show loss of function and reduced stability in standard
sample buffer conditions. It is therefore useful to identify the com-
ponents that are essential to recover the integrity and activity of
the protein, and this identication process itself can contribute to
an understanding of the proteins function. The necessary activa-
tion ingredients can often be inferred from prior knowledge of
the protein species, but some may be unknown. Here, we present
an approach to screen common environmental factors such as
pH, buffer type, ionic strength and a range of crowding agents
using Thermouor.
Recombinant protein expression often results in large quanti-
ties of the protein of choice that will yield sufcient amounts for
characterization through brute force purication. The resulting
protein sample may seem pure at face value, but it is important
to check its quality in terms of protein activity, integrity, dispersity
and stability. Structural studies are especially demanding, because
the protein sample has to remain stable under excruciating exper-
imental conditions. Crystallization requires highly concentrated
samples and can take any time from several hours to weeks. Stan-
dard protein crystallization is done at room temperature, and
many proteins are not stable over long time periods under these
conditions. Thermostability is therefore an important feature that
needs to be tested, and protein constructs may have to be opti-
mized with stability in mind.
A common approach is to use protein homologues from
extremophiles, i.e., species that live in circumstances of extreme
temperature and pressure. These proteins have evolved their struc-
ture so they can still function under extreme conditions [1], and
they are more stable under standard crystallization conditions.
Classical examples for the crystallization of proteins from thermo-
philic organisms are the 70S ribosome [2] and the potassium ion
channels [3]. When switching species is not an option, it is possible
to optimize the protein construct though truncations, mutations or
Fig. 1. Typical thermal denaturation assay using Thermouor. (A) Melting curve of
fusions with stabilizing factors. Proteins can be co-expressed with glucose isomerase in 100 mM TrisHCl pH 8.5. The protein solution is heated in the
other partners that help to stabilize the protein fold. In the end, the presence of a hydrophobic dye (SYPRO Orange). Upon denaturation, the dye binds
protein sample can be further stabilized by adjusting pH, salt con- to the internal hydrophobic protein core increasing signicantly the uorescence
centration, buffer system or by adding additives. The identication (right Y axis). Maximal uorescence intensity is obtained when the protein unfolds
completely, then SYPRO Orange signal decreases corresponding to dye-protein
of a stabilizing solution increases the ability to purify, concentrate
dissociation. Residual signal from SYPRO Orange is explained by the interaction
and crystallize the protein [4,5]. Unfortunately, no clear correlation from the dye with aggregated protein. Usually the uorescence signal is plotted as a
has been observed between stability and intrinsic properties of a function of temperature to get a sigmoidal curve that shows the fraction of the
protein such as the isoelectric point, molecular weight and per- unfolded protein (left Y axis). The inection point corresponds to the melting
centage of charged residues. No empirical rules can be drawn; temperature (Tm), at which 50% of the protein is unfolded. A Tm of 81.3 C has been
calculated. (B) Alternative representation of the melting curve using the rst
therefore we strongly recommend measuring the thermal stability derivative (dRFU)/dT of the raw data: the Tm corresponds to the apex. (For
of each new recombinant protein and systematically screen for interpretation of the references to color in this gure legend, the reader is referred
optimum conditions. to the web version of this article.)
194 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206
Circular dichroism coupled to a temperature gradient can be used
to monitor the loss of secondary structure as a sign of the unfolding
of the protein at elevated temperatures [6]. Protein aggregation
coupled to a temperature gradient can be monitored by light scat-
tering based methods [7,8] or by measuring the optical density of
the sample [9]. Differential scanning calorimetry (DSC)1 is another
non-evasive technique to study protein thermostability [10], but it is
very time consuming. Here, we discuss the adaption of a technique
that was developed by the pharmaceutical industry to screen for li-
gands that thermally stabilize protein targets. Thermouor has been
shown to be useful to assess thermostability in a systematic way,
and it enables many conditions to be tested simultaneously. The
technique requires relatively small amounts of sample and is inex-
pensive. Many structural genomic consortia use Thermouor as a
standard technique for quality control and to optimize sample con-
ditioning. However, anyone who has access to a real time PCR ma-
chine can modify it to screen protein thermostability.
Thermouor principle
The uorescence-based thermal stability assay or differential
scanning uorimetry (DSF) has been developed by Pantoliano
and coworkers in 2001 [11]. Registered as a trademark of 3-Dimen-
sional Pharmaceuticals (3DP), the term Thermouor subsequently
passed into common usage [12]. The methodology takes advantage
of the fact that the uorescence of many nonspecic protein-bind-
ing dyes increases with increasing hydrophobicity of their environ-
ment [13,14]. In the procedure, a dye interacts with exposed
hydrophobic regions generated by partial or full unfolding of pro-
teins. It is based on the detection of changes in the exposure of a
proteins hydrophobic core upon heat denaturation. The dye is
quenched in aqueous solutions, but when the aromatic moieties
of the dye intercalates into a hydrophobic pocket, it regains its
uorescence. For many proteins, the gradual increase of tempera-
ture has little effect on the protein fold, until a temperature is
reached where it will quickly unfold. At this point, the unfolded
protein will expose its hydrophobic core and the dye will become
uorescent (Fig. 1A). In an ideal case, the sudden unfolding of all
the proteins in the sample will lead to a sharp increase of the uo-
rescent signal over a short temperature range. A sharp sigmoidal
curve allows for the calculation of a melting temperature (Tm),
which correspond to the temperature where the protein is 50% un-
folded. The melting temperature can also be derived by calculating
the peak of the rst derivative (Fig. 1B). It has been shown that a
sharp melting curve is highly reproducible, and this opens the pos-
sibility to compare melting temperatures between samples that
contain the same protein but different buffer compositions. A posi-
tive shift in the melting temperature DTm can be coupled to an in-
crease in structural order and a reduced conformational exibility,
whereas a negative DTm, indicates that the buffer induces protein
structural changes towards a more disordered conformation.
Although this process can in principle be observed with any uo-
rimeter with a thermoblock, the most ideal conguration is ob-
tained with a real-time PCR machine. These thermocyclers are
equipped to detect uorescence in multi-well plates so that many
conditions can be tested simultaneously.
RT-PCR machines compatible with Thermouor
The adaption of commercially available RT-PCR machines for
Thermouor has made this technique easily accessible [15]. In
1
Abbreviations used: ANS, Anilinonaphthalene-sulfonate; DSF, Differential Scanning
Fluorimetry; DSC, Differential Scanning Calorimetry; TF, Thermouor; RT-PCR,
Reverse transcription polymerase chain reaction; Tm, Melting temperature.
principle, any RT-PCR instrument can be used for a Thermouor
assay as long as the optical system of the instrument is compatible
with the uorescent properties of the dye used in the assay. Instru-
ments reported to have been used for thermal shift assays include
the Mx3005P (Stratagene) and the iCycler (Bio-Rad) for a 96-well
format assay, the LightCycler 480 (Roche) and the FluoDia T70
(PTI) for a 384-well format assay, while the ABI 7900 (Applied
Biosystems) provides both 96- and 384-well format assays. The
choice of the instrument depends mainly on the required capacity,
the type of dye, the uorescent lters available, and the tempera-
ture range required. An RT-PCR machine equipped with a cooling
system provides the advantage to analyze protein stability from
4 C upwards, which can be crucial for temperature sensitive pro-
teins. At the SPC facility, we use an iCycleMyIQ RT-PCR Detection
System (Bio-Rad), equipped with a charge-coupled device (CCD)
detector for imaging of the uorescence. This RT-PCR is designed
to support the use of a single lter pair optimized for excitation
and emission of green uorescent dyes, resulting in excellent sen-
sitivity for the detection of uorophores such as FAM and SYBR
Green I (with excitation maxima at 485 and emission spectra at
530 nm, respectively). These lters have shown to be compatible
and useful to perform thermal denaturation of proteins using SY-
PRO Orange which, upon binding protein, has a uorescence exci-
tation/emission maximum spectra at 470 and 569 nm. However,
lters can also be replaced manually and adaptations can be made
for dyes with different uorescent properties. If required, custom-
ized lters can be purchased from CHROMA technology (http://
www.chroma.com).
Choosing the uorophore
Various uorescent dyes have been tested to probe changes in
protein conformation [16]. Nile red is a uorescent probe for intra-
cellular lipids and hydrophobic protein domains with a maximum
absorption at 553 nm. The 1,8-ANS dye makes hydrophobic and
electrostatic interactions with proteins with a maximum absorp-
tion at 350 nm. The bis-ANS dye binds stronger to the hydrophobic
cavities of the protein than 1,8-ANS, and has a maximum absorp-
tion at 385 nm. However, the most popular dye for thermal shift
assays is SYPRO Orange [1719].
Fig. 2. Effect of a His6-tag on the thermal stability of a viral protein. Thermal
stability of the pure protein was measured using Thermouor before and after
cleavage of the purication tag using TEV protease. The untagged protein shows a
thermal shift of 4.1 C compared to the tagged protein. This illustrates that it is
important to verify the effect of protein purication tags on protein stability.
 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206
The SYPRO Orange dye is relatively sensitive with a high signal
to noise ratio for the uorescence upon protein binding [17,20].
The uorescent properties of the dye-protein complex with a max-
imal absorption at 470 nm and maximal emission at 569 nm [17],
minimize the interference with background uorescence from
small molecules. For all environmentally sensitive dyes including
SYPRO Orange, the Thermouor effect depends on the partitioning
between the high-dielectric environment of water and the hydro-
phobic, low-dielectric environment that occurs when a protein is
melted. In some cases, interaction with native proteins, e.g., via
hydrophobic pockets on the protein surface, as well as interaction
with detergent leads to a high initial signal background. For this
reason, systematic efforts to understand membrane protein stabil-
ity with Thermouor assays is challenging. There are alternative
dyes such as the thiol-specic uorochrome N-[4(7-diethyl-
amino-4-methyl-3-coumarinyl]maleimide (CPM) which reacts
with cysteines that become accessible when a protein unfolds
[21]. To overcome this problem Enzo Life Sciences developed the
Proteostat dye, which is compatible with detergents and which
monitors protein aggregation rather than protein unfolding
[22,23]. If the protein is not compatible with any of the uorescent
dyes available, Warne and coworkers reported the development of
a dye-free microscale assay to measure the thermal stability of a
membrane protein depending on 3H labeled ligand binding [24].
Preparing your sample
In order to obtain a reliable Tm value, it is recommended to use
highly puried protein for the assay. It is occasionally possible to
obtain a sharp melting curve with proteins that are 60% pure, but
we generally recommend a purity of 75% or higher. Special caution
should be taken with recombinant and over-expressed proteins as
they often contain modications, such as the addition of afnity or
solubility tags. A solubility tag can add features to the melting
curve that are hard to distinguish from the protein of interest. In
the worst case, it can give the completely misleading impression
that the protein is folded, whereas the tag is the only part of the
recombinant protein that is folded and contributing to the melting
curve. On the other hand, an afnity tag such as a 6His-tag or
streptavidin-tag can decrease the protein thermal stability. We
show an example in (Fig. 2), where the cleavage of a C-terminal
6xHis-tag with TEV protease leads to an increase in the melting
temperature of 4.1 C. It should be kept in mind that the peptide
afnity tags have a highly dynamic structure and can initiate the
destabilization of the tagged protein. It is generally desirable to re-
move tags, and ensure that other modications have not inu-
enced protein folding or integrity.
The initial protein sample should not contain high concentra-
tions of buffers or elution agents such as imidazole, because these
may obviously interfere with the nal pH of the assay. Since the
sample in our described setup is diluted in the Thermouor buffer
by a factor of 10, we recommend to prepare the sample in a low
ionic strength environment; with salt and buffer concentrations
no higher than 200 mM at a neutral pH. It is recommended to
use a sample buffer free of stabilizing reagents such as glycerol,
reducing reagents such as DTT or TCEP and detergents. These
chemicals can inuence protein stability or may interfere at high
concentration with the uorescence measurement. If the use of
stabilizing agents cannot be avoided, it is advised to modify their
concentrations in separate assays to assess their inuence.
Setting up the assay
A typical thermal stability assay will require for each condition
tested an aliquot of 2 ll of puried-protein at an initial concentra-
tion of 20 lM. A typical assay volume is 25 ll per well so the nal
195
protein concentration will be 1.6 lM. For a 96-well assay, 115 lg
of a protein with a molecular weight of 30 kDa will be required.
The dye signal is concentration dependent; a higher protein con-
centration will result in a higher signal to noise ratio but may affect
the oligomerization state of the protein and cause a higher back-
ground signal. Under favorable circumstances, it is possible to
measure a melting curve with a protein concentration lower than
5 lM. However, it should be kept in mind that small proteins
(<15 kDa) tend to generate a blurry transition shape that may be
hard to extract from a weak signal. In case a protein cannot be con-
centrated to 20 lM, we suggest using a larger volume of the pro-
tein sample for the assay in order to reach a nal concentration
of 1.6 lM. The mixing protocol should then be modied to keep
a nal volume of 25 ll. The use of a larger volume of the protein
sample solution can inuence the nal chemical environment of
the assay, and this should be taken into consideration.
SYPRO Orange (Invitrogen) is delivered as a 5000 concen-
trated solution in DMSO. For a 96-well assay, we prepare a solution
of SYPRO Orange by diluting 3 ll of the 5000 concentrated dye
into 237 ll of distilled water to have a working solution that is
62 concentrated. For each condition, 2 ll of 62 SYPRO Orange
is needed per well to reach a nal concentration of 5, which is
the typical concentration used in most experiments. At this dye
concentration, the uorescent response is practically independent
of the size of the protein.
The mixing protocol we use is in the following order: (1) Suf-
cient distilled water to reach a total volume of 25 ll, (2) 5 ll of 5
buffer, (3) 5 ll of 5 salt [optional], (4) 5 ll of 5 additive [op-
tional], (5) 2 ll of 20 lM protein and then 2 ll of SYPRO Orange
solution. We advise against pre-mixing the protein and the dye, be-
cause the dye contains DMSO. This solvent can damage the protein
in higher concentrations, or it can interact with the protein affecting
the initial background signal during the Thermouor experiment.
The PCR-plate should be kept on ice during the preparation in order
to prevent protein denaturation and to equilibrate the sample for
the starting temperature of the assay.
Performing the Thermouor assay
Once the microplate has been lled with samples and buffers,
the plate is sealed with highly transparent optical-clear quality
sealing tape (e.g., Greiner Bio-one, catalogue number 676070).
The plate is centrifuged at 4 C at 2500g for 30 s immediately be-
fore the start of the assay to remove possible air bubbles. The
RT-PCR machine can be programmed rst with a 5 min equilibra-
tion time at 5 C to allow SYPRO Orange to diffuse. We have ob-
served that temperature equilibration of the assay lowers the
initial background uorescence. Subsequently, the plate is heated
from 5 to 95 C (20 to 95 C if no cooling system is available) with
initial stepwise increments of 1 C per minute, followed by the
uorescence reading optimized for SYPRO Orange at 485/20 nm
(Ex) and 530/30 nm (Em). When it is needed to record the melting
temperature at higher precision, smaller temperature increments
can be used. It has been reported that the heating gradient can af-
fect the precision of the Tm within the experimental error (Tm -
 0.2 C), but that it will not affect the DTm measured between
two experimental conditions [15]. The entire procedure requires
less than two hours. We present a summary of the procedure in or-
der to set up a high-throughput thermal stability assay (Fig. 3).
High-throughput Thermouor at the EMBL Hamburg facility
Samples that enter the SPC facility for optimization are rst
going through a quality control protocol that includes a single
Thermouor experiment. In this single experiment the buffer
196 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206

Fig. 3. Overview of the protocol for a high-throughput Thermouor assay. First, all stock solutions need to be prepared in a 96 deep well block; buffer (5 concentrated), salt
(5 concentrated) and water. A stock solution of additives (5 concentrated) can be also prepared when screening for further stabilization. These deep well blocks with stock
solutions can be stored at 20 C, if its chemical contents allow. Afterwards, to prepare the 96-microplate, 5 ll of each component is added to the plates, and then the volume
adjusted to 21 ll with distilled water. When preparing the 96-microplates in batch, we recommended to use a liquid handling robot system or a multichannel pipette to
optimize time and assure reproducibility between experiments. If the plates are not used the same day, they need to be sealed immediately to prevent evaporation and stored
at 20 C. When retrieved from storage, 96-microplates need to be thawed at room temperature and centrifuged (4 C, 30 s, 2500g). A working solution of SYPRO Orange 62
need to be prepared freshly and added to each well of the 96-microplate to reach a nal concentration 5 per well; by diluting 3 ll of SYPRO Orange 5000 to 237 ll of
water. Then 2 ll of protein and 2 ll of dye-working solution is added with a repeater pipette to each of the 96 wells. Final volume per well is 25 ll. The PCR plate is sealed
with a optical clear lid, centrifuged (4 C, 30 s, 2500g) and is being analysed on a real-time PCR machine using a temperature gradient of 1 C/min from 5 to 95 C. The whole
assay lasts approximately 2 h. The melting curves from each individual well have to be processed with appropriate software. (For interpretation of the references to color in
this gure legend, the reader is referred to the web version of this article.)

consists of 100 mM Hepes pH 7.4 and 100 mM NaCl. This buffer is
in general suitable for a preliminary test for a large number of pro-
teins. Based on the initial assessment of the Tm value and prole of
the melting curve, we advise our facility users to follow a stepwise
approach to nd the most optimal sample conditions. We rst try
to assess the effect of global parameters including buffer type and
concentration, pH and salt concentration in a systematic way with
a Buffer Screen using 96 different conditions (Table 1). This screen
is suitable for proteins for which globally optimal conditions have
to be identied. The screen is set up so that better handling condi-
tions can be designed for the protein, for instance during protein
purication. The identication of optimal conditions is useful for
proteins that cannot be frozen without loss of activity and need
to be stored at 4 C for a long time. The screen has also been useful
to optimize the conditions for the study of proteinprotein interac-
tions where the proteins are rst screened individually and then
mixed together to assess if this changes the behavior of the protein
complex in the Buffer Screen assay. Once the most stabilizing
global parameters have been established, the protein sample is
conditioned accordingly. In a second iteration process, the Addi-
tive Screen can be used to nd additional factors that stabilize
the protein, favor crystallizability, or may hint at potential protein
ligands and even functional aspects (Table 2). We have found that
for almost all proteins investigated in our facility with thermal
shift methodology, conditions were found that were more stabiliz-
ing than the starting buffers condition. Below, we present the
screens we developed in detail and the rationale for its compo-
nents. When the screens are run in the absence of protein sample,
the components themselves do not have a uorescent ngerprint.
In the presence of protein sample, some components increase the
uorescent background, without affecting the overall interpreta-
tion of the thermal shifts.
Table 1
Layout of the Thermouor Buffer Screen.

1 2 3 4 5 6 7 8 9 10 11 12
A Water Citric Acid Na Acetate Citric Acid MES KH2PO4 Citric Acid Bis-Tris Na Cacodylate NaH2PO4 KH2PO4 HEPES
pH 4.0 pH 4.5 pH 5.0 pH 6.0 pH 6.0 pH 6.0 pH 6.5 pH 6.5 pH 7.0 pH 7.0 pH 7.0

B MOPS Am Acetate TrisHCl NaH2PO4 Imidazol HEPES TrisHCl Tricine Bicine Bicine TrisHCl Bicine
pH 7.0 pH 7.3 pH 7.5 pH 7.5 pH 8.0 pH 8.0 pH 8.0 pH 8.0 pH 8.0 pH 8.5 pH 8.5 pH 9.0

C Water Citric Acid Na Acetate Citric Acid MES KH2PO4 Citric Acid Bis-Tris Na Cacodylate NaH2PO4 KH2PO4 HEPES
NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl
pH 4.0 pH 4.5 pH 5.0 pH 6.0 pH 6.0 pH 6.0 pH 6.5 pH 6.5 pH 7.0 pH 7.0 pH 7.0

D MOPS Am Acetate TrisHCl NaH2PO4 Imidazol HEPES TrisHCl Tricine Bicine Bicine TrisHCl Bicine
NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl NaCl
pH 7.0 pH 7.3 pH 7.5 pH 7.6 pH 8.0 pH 8.0 pH 8.0 pH 8.0 pH 8.0 pH 8.5 pH 8.5 pH 9.0

E Buffer A Buffer A Buffer A Buffer A Buffer A Buffer A Buffer A Buffer A Buffer A Buffer A Buffer A Buffer A
pH 4.0 pH 4.78 pH 5.21 pH 5.62 pH 5.95 pH 6.23 pH 6.53 pH 6.81 pH 7.16 pH 7.80 pH 9.0 pH 10.0

F Buffer B Buffer B Buffer B Buffer B Buffer B Buffer B Buffer B Buffer B Buffer B Buffer B Buffer B Buffer B
pH 4.0 pH 4.22 pH 4.62 pH 5.06 pH 5.62 pH 6.49 pH 7.25 pH 7.76 pH 8.20 pH 8.75 pH 9.0 pH 10.0

S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206

G 10 mM HEPES 50 mM HEPES 100 mM HEPES 250 mM HEPES 10 mM 50 mM 100 mM NaPO4 200 mM NaPO4 10 mM TrisHCl 50 mM TrisHCl 100 mM TrisHCl 250 mM TrisHCl
pH 7.5 pH 7.5 pH 7.5 pH 7.5 NaPO4 NaPO4 pH 7.5 pH 7.5 pH 8.0 pH 8.0 pH 8.0 pH 8.0
pH 7.5 pH 7.5

H 50 mM 50 mM 50 mM 50 mM 50 mM 50 mM 50 mM 50 mM 50 mM 50 mM 50 mM 50 mM
HEPES HEPES HEPES HEPES HEPES HEPES TrisHCl TrisHCl TrisHCl TrisHCl TrisHCl TrisHCl
50 mM NaCl 125 mM NaCl 250 mM NaCl 500 mM NaCl 750 mM NaCl 1 M NaCl 50 mM NaCl 125 mM NaCl 25 mM NaCl 500 mM NaCl 750 mM NaCl 1 M NaCl
pH 7.5 pH 7.5 pH 7.5 pH 7.5 pH 7.5 pH 7.5 pH 8.0 pH 8.0 pH 8.0 pH 8.0 pH 8.0 pH 8.0

Buffers were used at concentration of 100 mM unless otherwise indicated. Sodium chloride was used at concentration of 250 mM, unless otherwise indicated. Buffer A composition: Succinic Acid/NaPO4/Glycine [2:7:7]; Buffer B composition: Citric Acid/
CHES/HEPES [2:4:3].

Table 2
Layout of the Thermouor Additive Screen.

1 2 3 4 5 6 7 8 9 10 11 12
A 0.1 M 0.5 M 1M 2M 4M 6M 150 mM 500 mM 3% (v/v) 100 mM 100 mM 100 mM
Urea Urea Urea Urea Urea Urea GdnHCl GdnHCl DMSO NaHCO2 KHCO2 NH4HCO2

B 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM
NaC2H3O2 Ca(C2H3O2)2 CH3CO2K CH3COONH Na2SO4 Mg2SO4 K2SO4 (NH4)2SO4 Na2HPO4 NaH2PO4 K2HPO4 KH2PO4

C 100 mM 100 mM 100 mM 100 mM 10 mM 1 mM 5 mM 20 mM 100 mM 1 mM 1 mM 5 mM

Na2C4H4O Na3C6H5O7 Na2C3H2O4 NaNO3 DTT TCEP TCEP TCEP TMA-HCl Spermidine Spermine-HCl EDTA

D 10 mM 50 mM 100 mM 250 mM 500 mM 10 mM 10 mM 1 mM 1 mM 1 mM 1 mM 1 mM

Betaine Imidazole Imidazole Imidazole Imidazole MgCl2 CaCl2 MnCl2 NiCl2 FeCl3 ZnCl2 CoCl2

E 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 100 mM 2 mM 1 mM
NaF KF NH4F LiCl NaCl KCl NH4Cl NaI KI NaBr CHAPS OG

F 3% (v/v) 1% (v/v) 5% (v/v) 10% (v/v) 20% (v/v) 5% (v/v) 5% (w/v) 5% (w/v) 10 mM 50 mM 50 mM 25 mM
EG Glycerol Glycerol Glycerol Glycerol PEG 400 PEG 1000 PEG 4000 Pro Taurine Gly His

G 50 mM 50 mM (each) 500 mM (each) 50 mM 25 mM 50 mM 100 mM 100 mM 100 mM 100 mM 2 mM 2 mM

Arg Arg + Glu Arg + Glu Glu Gln Lys D-Glucose Sucrose Maltose D-Sorbitol NADH ATP
(5 mM MgCl2) (5 mM MgCl2)

H 2 mM 2 mM 2 mM 2 mM 2 mM Buffer Custom Custom Custom Custom Custom Custom

ADP cAMP GTP GDP cGMP
(5 mM MgCl2) (5 mM MgCl2) (5 mM MgCl2) (5 mM MgCl2) (5 mM MgCl2)

Abbreviations: guanidine hydr Chloride (GdnHCl), trimethylamine (TMA), Tris(2-carboxyethyl)phosphine (TCEP), ethylenediaminetetraacetic acid (EDTA), ethylene glycol (EG), Ctyl-b-D-glucoside (OG). The additive screen contains: chaotropic/dissociation
reagents [A1:A9], salts [A10:C4], reducing reagents [C5:C8], polyamines [C9:C11], chelating agent [C12], linker [D1], imidazole [D2:D5], multivalent ions [D6:D12], monovalent ions [E1:E10], detergents [E11:E12], polyols [F1:F8], amino acids [F9:G6],
carbohydrates [G7:G10], co-factors [G11:H5].

197
198 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206
Fig. 4. Effect of buffer and pH on glucose isomerase thermal stability. Representative melting curve from different buffers over pH 69. The results obtained in absence of
buffer are indicated as a black line and those in presence of buffers in colored lines. The different melting points extracted from the melting curves for the more stabilizing and
destabilizing conditions are given in the insert. Compared to water (Tm of 76.1 C), TrisHCl pH 8.0 (blue) is shown to be the most stabilizing condition increasing the Tm by
D + 5.9 C (82.0 C) whereas MES buffer at pH 6.0 is shown to be the most destabilizing condition with a Tm of 71.5 C. (For interpretation of the references to color in this
gure legend, the reader is referred to the web version of this article.)
Fig. 5. Effect of pH on glucose isomerase thermal stability. Representative melting
curve at pH 410 using an extended-range buffer system composed of succinic acid/
NaH2PO4/glycine (2:7:7). The aim of such a buffer system is to deconvolute the
effect of pH from the buffer himself. In this example, pHs below 6.2 show a strong
destabilizing effect while the optimal pH range is 6.59.0. The melting temperature
for the three component buffer system is lower for the whole pH range than what is
obtained for Tris buffer. The pH effect shows globally that it is preferred to work
with neutral or basic pH, but the choice of the actual buffer system is crucial.
Buffer Screen
Choosing the right buffer system for a macromolecular sample
is vital from the rst extraction through purication and character-
ization of the protein. In some cases, buffers such as Tris have been
reported as a source of sample heterogeneity [25]. In contrast,
buffer systems can have a positive effect on the protein stability
by modulating the pH of the protein solution and by directly inter-
acting with the protein itself. It is common to nd ordered buffer
molecules within protein crystal structures that form intrinsic
interactions within the protein. To date, there are more than 600
structures in the Protein Data Bank [26] that contain ordered
MES, about 500 that contain ordered Hepes, and more than 600
that contain ordered Tris, showing that it is not unusual to have
specic interactions between buffers and protein molecules. For
example, one of the structures of glucose isomerase reports the
presence of a Tris molecule (PDB ID 1MNZ). It is therefore essential
to test the effect of different buffer systems on protein stability.
This section describes our Buffer Screen which contains a set of
23 common buffer systems (16 different chemicals) each at a
concentration of 100 mM covering a pH range from 4.0 to 9.0 in
absence or in presence of 250 mM NaCl to favor protein solubility
(Table 1, conditions A1B12 and C1D12, respectively). We
included a condition with water only as a reference control to
evaluate the effect of the absence of a buffer system on the melting
temperature. Our aim is to assess whether a protein shows
increased stability for a certain buffer system over a certain pH
range. As a case study, we present the outcome of the Buffer
Screen on glucose isomerase (Fig. 4). An issue in optimization is
to nd factors which are independent, so as to simplify the analysis
of the results of the optimization trials. This can be a challenge
with buffers, as a buffer and its pH range tend to be highly corre-
lated. Moreover, some buffer properties are affected by other
factors. For instance, the buffer capacity of Tris is temperature-
dependent. Our Buffer Screen was designed in such a way that
these trends can be deconvoluted (Fig. 5). Since most single buffer
systems have a narrow pH range, we suggest to test at least one
extended-range buffer system as these systems ensure a smooth
gradient in pH and a consistent buffering capacity without neces-
sitating a change in the buffer components. A multi-buffer system
for pH modulation is useful to identify the best stabilizing pH using
a ne grid by titrating surface groups. The ratio of the components
within the buffer system are selected so as to produce a reasonably
linear response to pH [27]. We included in our Buffer Screen the
succinic acid/phosphate/glycine buffer as well as the citric acid/
Hepes/Ches buffer system to cover a pH from 4 to 10 (Table 1, con-
ditions E1F12). In most cases, we obtained a very good correlation
between results from these two buffers systems within 0.3 pH
units and we can identify specic buffer effects. A similar strategy
has been used successfully to assess the variation in stability of the
CorA Mg2+ transport channel with pH [28]. In addition, some afn-
ity purication protocols involve a pH shock to elute the protein
which may not be suitable for the folded protein as it results in
protein denaturation, precipitation or aggregation [29,30]. Based
on the trends observed in the Buffer Screen, it can be useful to
modify the sample purication protocol. The nature of the buffer
and the pH are not the only important parameters to optimize with
the aim to increase protein stability. Given the potential interac-
tions between certain buffers and proteins, it is also important to
gauge the effect of buffer concentration on protein stability. The
Buffer Screen samples the effect of the buffer concentration
(10 mM up to 250 mM) for three buffers widely used for protein
purication and characterization; Hepes, phosphate and Tris
 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206 199

Fig. 6. Effect of ionic strength on a proteins thermal stability. Representative melting curves of an (A) adaptor protein at different concentrations of NaH2PO4 from 10 to
200 mM and (B) DNA binding domain in presence of NaCl from 0.05 M to 1 M in 50 mM TrisHCl pH 7.5. In both cases, the calculated Tm value increases with the ionic
strength of the solution. In the case of DNA binding domain, the addition of 1 M NaCl increases the Tm value by +7.8 C. This data has been used to titrate the optimal NaCl
concentration for NMR studies, and it was shown that 225 mM NaCl promotes protein stability without having an impact on DNA binding.

(Table 1, conditions G1G12). We provide an example where
protein thermostability of an adaptor protein is dependent on buf-
fer concentration, increasing the Tm up to D + 10.8 C (Fig. 6A).
These results indicate that the ionic strength might be a key factor
affecting the thermostability of proteins [31].
Protein stability is also strongly affected by salt ions incorpo-
rated in the hydration shell. Too high or too low salt concentrations
can lead to precipitation of a protein. For this reason, we imple-
mented a NaCl concentration screen (Table 1, condition H1H12)
to further assess how ionic strength affects protein stability. Titra-
tion for the optimal NaCl concentration has shown in some cases to
have a strong stabilizing effect and may contribute to nd opti-
mum conditions for structural studies. For example, a DNA binding
domain of Sfc1 (from Schizosaccharomyces pombe) was not behav-
ing well in its standard buffer during NMR measurements. The
thermostability of this protein was tested by Thermouor using
the Buffer Screen, which helped to identify a strong stabilizing -
effect induced by NaCl. Fine-tuning of the experimental conditions
allowed to determine the concentration of NaCl suitable to pro-
mote the protein stability in solution (up to 225 mM NaCl) without
impairing afnity to DNA (Fig. 6B). While ionic strength plays a key
role in the stability of several proteins, no clear trend has been
observed between the stabilizing effect of buffer and NaCl concen-
tration, suggesting that both of these parameters should be opti-
mized independently. In summary, the Buffer Screen designed
for our facility explores the effect of four basic parameters on the
protein thermal stability (buffers, pH, buffer and NaCl concentra-
tion) to identify an optimal buffer environment.
Additive Screen
Ligand-induced conformational stabilization of proteins is a
well-understood phenomenon. Substrates, inhibitors, cofactors,
and protein binding partners provide enhanced stability to pro-
teins by selective binding. A thermal denaturation assay can be
used to screen for the effect of additives while the buffer condi-
tions are kept constant. Upon ligand binding, the protein complex
denaturates at a higher temperature and the difference in the Tm
value in the presence and absence of the compound reects ligand
binding. Thus, the thermal shift assay can serve as a tool to seek for
stabilizing reagents, and to identify natural ligands that provide in-
sight into the biological function of the protein [20,32]. The Addi-
tive Screen requires highly puried and active protein, to ensure
that the additive-protein interaction is the sole cause of the in-
crease in stability. Once a ligand has been identied, it is important
to include it in the early stages of protein production, sometimes
already during protein expression to promote proper folding and
to prevent aggregation. Many proteins or domains remain unstable
or partially soluble in absence of their cofactors or stabilizing
ligands [4].
200 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206
Fig. 7. Effect of imidazole concentration on the thermal stability of a protein involved in the immune system. Imidazole is widely used as an elution buffer to remove His-
tagged recombinant proteins from a metalchelate resin. Here we show that imidazole can have a strong destabilizing effect on proteins, leading to aggregation and protein
precipitation. The concentration of imidazole should be adjusted in order to minimize the destabilizing effect on the folding of the protein which can occur at imidazole
concentrations as low as 50 mM. The standard elution protocol for this protein used 250 mM imidazole, which destabilizes this particular protein with a thermal shift of
15.3 C.
At the SPC facility, once we identied a suitable buffer using the
initial Buffer Screen, the proteins are rescreened with the new
optimal buffer system against the Additive Screen. The Additive
Screen consists of a selection of 72 different physiological and
non-physiological ligands that include amino acids, nucleotides,
sugars, cofactors, monovalent and divalent ions, and some other
additives representing 14 categories (Table 2). We typically select
410 commercially available compounds per framework. It is con-
venient to prepare stocks of 5 concentrations of all compounds
and dispense 5 ll into each well (Fig. 3). The nal concentrations
of the compounds within the screens varies between 1 mM and
100 mM, depending on their solubility limit. In principle, a com-
pound can show an effect at 10 lM concentrations when the bind-
ing afnity is expected to be between 1 nM and 1 lM. However,
compounds that bind with relatively low afnity such as nucleo-
tides have to be present in high concentrations. The Additive
Screen is not exhaustive, and it is advised to customize it based
on prior knowledge of the proteins function. We designed this
screen especially to further optimize protein purication and favor
overall protein stability. Articial and natural stabilizing com-
pounds as well as chaotropic reagents are included since they have
the potential to prevent protein aggregation.
We have shown with the Buffer Screen that the nature of the
buffer and the ionic strength can have an important effect on pro-
tein stability. The presence of ions might induce changes of protein
solubility, protein denaturation and changes in enzyme kinetics by
modulating electrostatic interactions within the protein. We in-
cluded 17 ion species in the Additive Screen, varying the combina-
tion of cations and anions (Table 2, conditions A10C4, E1E10)
and metals ions (Table 2, conditions D6E10) which are widely
used for protein purication and characterization. For example,
using this screen it is possible to compare the stability of the pro-
tein with different halogens, to see if the commonly used NaCl has
a deleterious effect when replaced with NaF. Moreover, since NaCl
is not compatible with circular dichroism measurements in the far-
UV range, NaF is an interesting alternative.
Divalent ions and cations may coordinate with the protein and
act as generic stabilizers [31]. For example, the addition of MgCl2 to
ribonuclease A [33] and the CorA Mg2+ transport channel from Met-
hanococcus jannaschii showed a positive thermal shift [28]. Dias
and coworkers performed a Thermouor assay on the PA subunit
of inuenza virus polymerase [34] and observed a substantial shift
for divalent metal ions (exceeding 10 C). Based on this assay, they
showed that this domain has intrinsic RNA and DNA endonuclease
activity that is strongly activated by manganese ions, matching
observations reported for the endonuclease activity of the intact
trimeric polymerase. We encourage to gauge metal afnity using
different concentrations of metal ions, since proteins can have
multiple metal binding sites. However, caution should be taken
since it has been reported that metal ions such as FeCl2, CuCl2,
and CoCl2 have a tendency to quench uorescence in thermal sta-
bility assays at high concentration (mM) [35]. In our Additive
Screen, this effect was minimized for 7 metals by using low con-
centrations (110 mM) (Table 2, conditions D6D12). Given the
side effects, it is advisable to further test the optimal concentration
of metal ions once a metal dependency has been observed. We also
suggest to use fresh metal preparations that are no more than a
few weeks old, since compounds such as FeCl2 tend to oxidize over
time.
The most common rst purication step for proteins prepared
for structural characterization is afnity chromatography based
on the binding of a poly-Histidine tag to beads charged with Nickel
or Cobalt. Imidazole is commonly used to elute the recombinant
proteins from the afnity column. This procedure involves a high
concentration of imidazole (250500 mM) which may affect pro-
tein stability. Many proteins precipitate immediately when the
imidazole is dialyzed out. The combination of a suboptimal buffer
condition and the presence of high concentrations of His-tagged
protein can promote protein aggregation and precipitation. For this
reason, it is useful to explore the stability of the protein in different
concentrations of imidazole to identify the highest concentration
where it does not affect protein stability. This upper concentration
can then be used in the elution buffer. We have included an imid-
azole gradient in the Additive Screen from 10 mM up to 500 mM
(Table 2, conditions D2D5), see example (Fig. 7). This example
shows a strong destabilizing effect even at low imidazole concen-
trations. In such cases, we recommend to use pH shock as an elu-
tion method instead of an imidazole gradient. If this does not work,
it may be necessary to change the purication method by using a
different afnity purication-tag. In parallel, the Additive Screen
veries the effect of Ni2+ and Co2+ ions on the protein. This may
guide the choice of a Nickel or Cobalt resin (Table 2, conditions
D9 and D12).
zIt is common to use stabilizing and destabilizing reagents
during purication that help to reduce aggregation. Metal chela-
tors such as ethylenediaminetetraacetic acid (EDTA) at a concen-
tration of 15 mM avoid metal-induced oxidation of SH groups
and help to maintain the protein in a reduced state. Moreover,
reducing reagents such as beta-mercaptoethanol, dithiothreitol
(DTT) and tris(2-carboxyethyl)phosphine (TCEP) can be added
 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206 201

Fig. 8. Different melting curve proles. (A) Melting curve of esterase shows a sharp and fast thermal denaturation transition within a small temperature interval (5563 C).
(B) Melting curve of DNA transposase shows good thermal stability, but with a weak denaturation transition (3865 C). (C) Example of a multiphase curve suggesting the
presence of a multi-domain protein or a mixture of different proteins. Two overlapping melting curves are observed and the Tm cannot be calculated without ambiguity. (D)
Protein precipitation or aggregation can result in a low melting curve intensity. This is also observed for small proteins lacking a hydrophobic core. The resulting curve is
nearly at-shaped and the uorescence signal is too weak to rely on. (E) The presence of interfering hydrophobic compounds such as detergents, protein aggregates or
solvent-exposed hydrophobic patches can cause high uorescence background at lower temperatures.

in order to prevent non-specic aggregation mediated by the
scrambling of free cysteines forming disulde bonds. This also
applies to proteins that are in vivo in the cytosol, because the
reducing environment is not present in vitro. There are protein
systems that have labile disulde bonds, and the addition of a
particular reducing agent can shift the equilibrium towards a
more stable conformation. For many proteins, it is therefore
important to add a small amount of reducing agent to prolong
the life time and monodispersity of the protein. The Additive
Screen helps to identify the concentration limit for such reagents
where they will not impair protein folding. We have included low
and high concentrations of TCEP (1 and 20 mM) and DTT at a
concentration of 10 mM for comparison, as well as 5 mM EDTA
(Table 2, conditions C5C8 and C12).
Chaotropic agents that can unfold proteins at high concentra-
tions can have a stabilizing effect at low concentrations [36].
Chaotropic species can rearrange the protein into a more stable
conformation, interfering with the intermolecular interactions that
lead to aggregation. Low concentrations of denaturing agents such
as guanidine hydrochloride (GdnHCl) or urea can help to stabilize
proteins with a tendency to aggregate, without affecting the
secondary structure of the protein [37,38]. These chaotropic
reagents can be added to purication buffers also to help to reduce
non-specic binding. Urea and GdnHCl have been used to study
202 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206

Table 3 to investigate other lipids and detergents if a thermal shift is ob-

Optimization grid for thermal stability assay using SYPRO Orange. Thermouor served for these compounds.
experimental conditions might need to be optimized for some proteins. At the SPC
facility, we used a grid varying the protein concentration versus SYPRO Orange
Proteins with a potentially unstable fold tend to be temperature
concentration. The grid can help to nd optimal conditions for the thermal sensitive, and the unfolding can be slowed down by the addition of
denaturation assay: which protein/dye ratio gives higher signal to noise ratio, charged amino acids such as L-arginine and L-glutamic acid. Protein
sharper unfolding transition and lower background signal. The conditions used as refolding studies have shown that the presence of L-arginine in the
standard conditions at the SPC facility are colored in yellow.
buffer inhibits the aggregation of the protein and increases the
Protein concentration [mg/ml]* yield of renatured and biologically active protein [37]. The
amounts used in refolding are substantial (up to 500 mM), and this
0 0.01 0.05 0.1 0.25 is not ideal for structural studies, but it can help during protein
SYPRO Orange
concentration*

1X purication and storage. We include conditions with either L-argi-

2.5X nine, L-glutamic acid, L-lysine, L-glutamine alone, or L-arginine and
L-glutamic acid mixed with concentrations from 50 up to 500 mM
5X
in our Additive Screen (Table 2, conditions G1G6).
10X The Additive Screen focuses on small molecules that can affect
20X the overall stability of the protein during all stages of the protein
*Final concentration in 25 l production process. It can help to review the purication process,
a and it provides leads to some of the most common proteinligand
Final concentration in 25 ll.
interactions occurring in a cell. Once a trend is established, it is
worthwhile to further investigate the effect of a compound with
global change in protein conformation by differential scanning customized Thermouor screens.
calorimetry [39] and to investigate the effect of ligand binding on
the thermal stability in presence or absence of GdnHCl denatur-
Data analysis and melting curve prole
ation [40]. However, the tolerance for denaturing agents varies
from protein to protein. In this Additive Screen, we have included
Plotting the melting curve
urea from low to high concentrations (0.1 up to 6 M) where most
proteins become unfolded at room temperature. For comparison,
The analysis of Thermouor data is based on a plot of the melt-
we also included two concentrations of GdnHCl (Table 2, condi-
ing curve that represents relative values of the detected uores-
tions A1A6 and A7A8).
cence intensity coming from the dye versus temperature
Crowding agents that mimic the macromolecular crowding
(Fig. 1A). In case that a constant decrease of the uorescence is ob-
occurring in a cell have been shown to have a stabilizing effect,
tained before the appearance of the melting peak, we advise to look
and they interfere with the hydration shell of proteins [41]. Glyc-
at the increase in slope (d(RFU)/dT) (Fig. 1B) and not to look at just
erol is a small polar molecule that can insert itself in the rst
the raw signal. To identify a ligand or a buffer condition that stabi-
hydration shell of proteins. It is also found in many cavities, stabi-
lizes a protein, the Tm value of the protein under each condition of
lizing the dynamic motions of the side chains surrounding the
the screen needs to be compared with the reference Tm. When test-
cavities. Addition of 520% glycerol to cell lysate and during puri-
ing a large number of conditions in parallel as in the case of the
cation often contributes to an improved solubility of proteins.
Buffer Screen and Additive Screen, we strongly recommend to
Moreover, glycerol is commonly used for protein storage as it
organize the data by categories such as pH, salt concentration,
prevents unfolding during the freezing step and can be used as
divalent ions, nucleotides, etc. Several software packages are avail-
cryoprotectant for ash frozen protein crystals. Because glycerol
able to assist in the data processing as well as in the comprehensi-
is used for a wide range of applications, we analyze the effect of
ble display of the results; we recommend the commercially
glycerol on the protein stability and identify the optimal concen-
available packages GraphPad (http://www.graphpad.com) and Ori-
tration using the Additive Screen (Table 2, conditions F2F5).
gin (http://www.originlab.com). Alternatively, Frank Niesen from
Several other crowding agents are included in Additive Screen
the Oxford Structural Genomics Consortium developed custom-
in order to cover a wide variety of conditions such as polyethylene
made calculation software based on Microsoft Excel to read in
glycols, polyamines, sugars and polyhydric alcohols such as sorbi-
Thermouor scans and to compute Tm values from a variety of
tol [33] (Table 2, conditions C9C11, F6F9 and G7G10, respec-
available RT-PCR instruments [ftp://ftp.sgc.ox.ac.uk/pub/biophys-
tively). Individual nucleotides are abundant in the cell, and they
ics/]. Data processing can be assisted also by software such as Ther-
may bind in any pocket that allows the stacking of an aromatic
moQ [44]: http://jshare.johnshopkins.edu/aherna19/thermoq/
ring. Thus we encourage to test nucleotide binding even if this is
downloads.html.
not expected to occur for the protein under investigation. We in-
The reproducibility in thermal shift determination between dif-
cluded several nucleotides in the Additive Screen in the presence
ferent RT-PCR machines for a particular condition is around <0.2 C
of MgCl2 (Table 2, conditions G11H12). It is necessary to use
which is much smaller than the thermal shift that can be observed
freshly prepared material in order to prevent the degradation of
for the binding of a ligand [4]. When optimizing a buffer, we con-
sensitive nucleotides such as ATP and ADP, and non-hydrolysable
sider shifts in Tm larger than 2 C signicant within a screen where
analogues should be considered.
an experiment has been done under identical conditions. In gen-
Low concentrations of detergents such as CHAPS or octyl-b-D-
eral, Tm values are highly reproducible and have shown to vary
glucoside may be useful to prevent protein aggregation and have
by less than 2 C between repeated experiments using separately
shown to be compatible with thermal stability assays using SYPRO
puried samples [28].
Orange (Table 2, conditions E11E12). These detergents have rela-
tively small aliphatic tails, and they are sometimes used to solubi-
lize membrane proteins. There are many small lipids present in the Single curve
cytosol of the cell, and these compounds can mimic lipid binding.
In addition, we have added spermidineHCl, which has been a Since protein unfolding is a cooperative process, the unfolding
silver bullet for the crystallization of proteins and protein of a small protein region will induce the immediate unfolding of
complexes [28,42,43] (Table 2, conditions C10C11). It is worth the remaining protein core; thus an optimal protein stabilization
 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206
buffer should result in a sharp and fast thermal denaturation tran-
sition between the folded and unfolded state, detected through
high transition slopes, in parallel with a higher Tm (Fig. 8A). How-
ever, less stable proteins can show a gradual melting curve where
the protein unfold over a longer temperature range (Fig. 8B). A sin-
gle coherently unfolding protein will produce a sigmoidal melting
curve. The midpoint or inection point of the transition curve is
calculated using simple equations such as the Boltzmann equation.
The melting curve is often followed by a decrease in uoresence
intensity at higher temperatures. Our interpretation is that a frac-
tion of the SYPRO Orange is released gradually from the unfolded
protein to the aqueous solution where the signal is quenched.
However, a residual uorescence signal remains which might cor-
respond to a fraction of SYPRO Orange bound to the denaturated
protein or to protein aggregates that form at higher temperatures.
Complex curves
Thermal shift assays unfortunately do not always generate sin-
gle sigmoidal curves. The presence of a multi-domain protein or
the presence of multiple proteins [25] can lead to a multiphasic
curve suggesting multiple melting transitions (Fig. 8C). Since sepa-
rate melting events often overlap, it may be difcult to calculate a
single Tm value. However, many multidomain proteins have do-
mains that are sufciently coupled energetically that they melt
in a cooperative manner resulting in a single melting transition
state [11]. In some cases, we have to assume a two-state unfolding
model, and more sophisticated algorithms are needed to t these
complex curves.
Troubleshooting
Although the Thermouor assay is used as a high throughput
method, it may require assay optimization for some proteins. In
a systematic case study, approximately 25% of 60 soluble proteins
examined by Thermouor using SYPRO Orange were not amenable
to such screening, partly due to a high uorescence background
[4]. Recently, the crystallization facility at EMBL-Grenoble per-
formed a large scale study on proteins submitted for crystallization
trials [45]. They measured the thermal stability of 657 soluble pro-
teins that entered their facility using SYPRO Orange. Results were
classied in three categories. In the majority of cases (66%) the
samples produced typical denaturation curves with a clear and
sharp temperature transition, allowing a straightforward estima-
tion of the Tm. For 30% of the samples, no clear temperature tran-
sition was observed, precluding the calculation of a value for the
Tm. In a few cases complex temperature transitions were observed.
Only small differences in crystallization rates were observed
between the rst and the second group indicating that failure to
obtain any a denaturation curve should not be interpreted by itself
as a sign that the protein is not feasible for crystallization [45]. Pro-
teins with intrinsically disordered regions with a complicated fold-
ing landscape are not suitable for Thermouor, because they give
complex curves that are often irreproducible. Even if certain trends
can be observed, it is near impossible to interpret them since the
state of the protein over the temperature range and its interactions
with the uorescent dye are unknown.
Featureless curve
A featureless curve can occur because of problems in the sample
preparation (Fig. 8D), the protocol or the set up of the RT-PCR
instrument. First, we recommend to visually inspect the microplate
to see if the protein completely precipitated under the conditions
tested. Second, it is advisable to repeat the assay with a higher
203
protein concentration (40100 lM): the higher the amount of pro-
tein, the cleaner the signal will be as long as the proteins are not
unfolded. Increasing the volume of the reaction up to 50 ll will
produce a larger signal, but the strength of the signal is not propor-
tional to the increase in volume [28]. If the assay fails again to gen-
erate a melting curve, the protein might not contain a hydrophobic
core: we have observed that the proteins that showed no transition
were often small proteins. In that case, it may be necessary to test
the thermostability of the protein sample by other methods such
as circular dichroism [46] or dynamic light scattering [7]. We also
observe in rare cases that the Tm of the protein is higher than the
maximum temperature reached on the instrument, so some
proteins with an intrinsically high Tm may not be amenable to this
assay. This can be veried by circular dichroism, because that will
clearly show if there is a loss in secondary protein structure within
the experimental temperature range. It can be useful to include a
positive control in thermal denaturation assays such as glucose
isomerase or citrate synthase to ensure the experimental set up
is correct and the dye is responsive and not degraded. Especially
if the RT-PCR machine is used for other purposes as well, it is
important to test whether the machine is calibrated properly for
Thermouor before the start of each individual experiment.
High initial uorescence signal
The highly hydrophobic nature of proteins substantially in-
creases the background uorescence of the assay (Fig. 8E), and in
many cases practically masks the melting transitions [47]. Proteins
displaying aberrantly high initial uorescence in the presence of
SYPRO Orange can be partially unfolded or contain hydrophobic
patches exposed in the native state which might interact with
the dye. It is important to make sure that all aggregated protein
is removed by ltration or centrifugation. It is also possible that
other ingredients in the protein sample (for instance detergents)
interact with the dye.
Optimizing the assay
Even if the thermal stability protocol presented above is suit-
able for a large number of proteins, some ne tuning might be nec-
essary for specic proteins in order to optimize the signal/noise
ratio and increase accuracy in the measurement of the Tm. Some
parameters can be optimized such as the concentration of SYPRO
Orange. Kean and coworkers have tested the effect of a range of SY-
PRO Orange concentrations and found that the amplitude of signal
increased in a linear fashion with dye concentration. However,
they also observed that SYPRO Orange has a destabilizing effect
at high concentrations (>20) [28]. We suggest optimizing 3
parameters; the incubation time for the protein/dye mixture prior
to the assay, the protein concentration and the dye concentration
(Table 3).
Thermouor applications
Screening the effect of mutations on stability
Protein-engineering techniques such as site-directed mutagen-
esis, random mutagenesis and directed-evolution techniques have
been successfully employed for various proteins. Single amino-acid
substitutions can have signicant impact on the folding and aggre-
gation properties of proteins that are both deleterious and favor-
able [4850]. Therefore, the thermal shift assay can be used to
quickly pinpoint the effects of specic point mutations on mutant
stability relative to wildtype [23,51] or on binding interactions
[52]. We present a comparative study (Boivin, unpublished,
204 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206
Fig. 9. Characterization of the effect of point mutations on the protein stability. Each histogram indicates the thermal shift of a particular point mutation from the melting
temperature of the wild-type (53.8 C). Mutations can be either stabilizing or destabilizing compared to the thermal stability of wild-type protein (depending on the amino
acid residue that has been mutated).
Fig. 9) which compares the thermal stability of single point
mutants of a viral protein. The mutants are tested under different
buffer conditions over a wide pH range to ensure that the thermal
shift for the mutant is not simply a result of protein-buffer interac-
tions. Indeed, some of the thermal shifts are buffer or pH depen-
dent. This gure also illustrates a different way to represent the
data, which is normalized against the wild-type Tm. Thermouor
is ideal to compare the stability of several mutants at the same
time, under identical experimental conditions.
Measuring kinetics of ligand binding by Thermouor
Thermouor measures the binding afnity by detecting ligand-
dependent changes in the thermal stability of a target protein. To
conrm that a compound binds and stabilizes the protein in a
concentration-dependent manner, it is possible to perform an as-
say in the presence of different concentrations of the compound.
While the thermal denaturation assay is used primarily to optimize
buffers and identify ligands, it has been shown that the stabilizing
effect of compounds upon binding is in several cases proportional
to the concentration and afnity of the ligands [4,12,53]. Thermo-
uor has been used successfully to calculate ligand association
constants and they are in good agreement with isothermal titration
calorimetry (ITC) measurements [15] as well as radioactive compe-
tition and uorescence polarization assays [11]. For example, Ved-
adi and coworkers correlated inhibition data for a protein kinase
and showed that Tm shifts larger than 4 C translate into values
for IC50 < 1 lM. For several compounds, this correlation suggests
that thermodynamic models adequately describe binding kinetics
despite the irreversible nature of the thermal protein melting
[54,55]. This kind of information can for instance be used for pro-
tein crystallization, to determine the concentration of a compound
needed to reach maximum occupancy. Nevertheless, interpreta-
tions regarding thermodynamics need to be done with caution
and conrmed by other methods such as isothermal titration calo-
rimetry since these afnities depend both on enthalpy and entro-
py, which is not properly assessed with Thermouor [56].
Thermouor and the crystallizability of protein samples
The homogeneity, stability and solubility of biological
macromolecules are key factors that have a strong effect on the
propensity to crystallize. Moreover, crystallization of macromole-
cules is a complex procedure which is strongly inuenced by di-
verse environmental factors such as pH, ionic strength, additives,
precipitants, protein concentration and temperature. Identifying
conditions that optimally stabilize proteins in solution has the
effect of reducing the structural heterogeneity of a protein and this
promotes the formation of crystals that are amenable for X-ray
structure determination, since crystallization generally favors
3-dimensional lattice assembly from structurally identical objects.
Ericsson et al., observed a 2-fold increase in the number of crystal-
lization hits when proteins were co-crystallized with stabilizing
additives identied with Thermouor [35]. This suggests that
Thermouor constitutes a valuable and efcient high-throughput
method to predict and improve the crystallizability of proteins
[4,57]. A correlation between complex multiphasic denaturation
behavior and a low likelihood of success in protein crystallization
has been recently established [45]. Dupeux and coworkers showed
in a systematic study that proteins with a Tm higher than 45 C
have a substantially higher success rate in crystallization than
proteins with lower melting temperatures and proposed to use
the results of this assay to help make rational decisions on which
constructs or samples to prioritize. Their data also indicated that
crystallization success rate is maximal when the incubation tem-
perature for the crystallization experiments is at least 25 degrees
below the Tm of the sample determined in the reference buffer con-
ditions and advised to set up crystallization trails at 4 C instead of
room temperature when the Tm is lower than 45 C [45].
Concluding remarks
The use of thermal shift assays to monitor protein stability with
differential uorescent calorimetry or Thermouor has become a
mainstream technique. Originally designed to identify protein li-
gands, it has now been developed to improve protein crystallizabil-
ity [45], to screen for proteinprotein interactions [52] and
differential modes of protein inhibition [58]. Here we describe the
use of Thermouor as a general tool to identify stabilizing
conditions for recombinant proteins, to improve purication yields,
protein storage as well as crystallizability. The two screens pre-
sented here test the effect of commonly used chemicals in protein
purication and storage. The evaluation of these screens provides
a quick survey of the overall behavior of the protein, which can be
used to improve protein purication and characterization protocols.
Thermouor online resources
http://www.embl-hamburg.de/facilities/spc.
http://Thermouor.org.
https://jshare.johnshopkins.edu/aherna19/thermoq/download.
 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206
Acknowledgments
We thank the users of the facility that shared their Thermouor
results for this manuscript. We would like to thank Sebastian Glatt
and Christoph Mller (EMBL Heidelberg) for sharing data on Sfc1.
We thank the staff of the crystallization platform of EMBL-Heidel-
berg and EMBL-Grenoble for constructive discussions. The research
leading to these results received funding from the European Com-
munitys Seventh Framework Program (FP7/2007-2013) under
grant agreement No. 227764 (P-CUBE), and under grant agreement
N283570 (Biostruct-X).
References
[1] S. Kumar, C.J. Tsai, R. Nussinov, Factors enhancing protein thermostability,
Protein Eng. 13 (2000) 179191.
[2] C. Glotz, J. Mussig, H.S. Gewitz, I. Makowski, T. Arad, A. Yonath, H.G. Wittmann,
Three-dimensional crystals of ribosomes and their subunits from eu- and
archaebacteria, Biochem. Int. 15 (1987) 953960.
[3] Y. Jiang, A. Lee, J. Chen, M. Cadene, B.T. Chait, R. MacKinnon, Crystal structure
and mechanism of a calcium-gated potassium channel, Nature 417 (2002)
515522.
[4] M. Vedadi, F.H. Niesen, A. Allali-Hassani, O.Y. Fedorov, P.J. Finerty Jr., G.A.
Wasney, R. Yeung, C. Arrowsmith, L.J. Ball, H. Berglund, R. Hui, B.D. Marsden, P.
Nordlund, M. Sundstrom, J. Weigelt, A.M. Edwards, Chemical screening
methods to identify ligands that promote protein stability, protein
crystallization, and structure determination, Proc. Natl. Acad. Sci. USA 103
(2006) 1583515840.
[5] T.M. Mezzasalma, J.K. Kranz, W. Chan, G.T. Struble, C. Schalk-Hihi, I.C.
Deckman, B.A. Springer, M.J. Todd, Enhancing recombinant protein quality
and yield by protein stability proling, J. Biomol. Screen. 12 (2007) 418428.
[6] N.J. Greeneld, Using circular dichroism collected as a function of temperature
to determine the thermodynamics of protein unfolding and binding
interactions, Nat. Protoc. 1 (2006) 25272535.
[7] J. Kopec, G. Schneider, Comparison of uorescence and light scattering based
methods to assess formation and stability of proteinprotein complexes, J.
Struct. Biol. 175 (2011) 216223.
[8] S. Benjwal, S. Verma, K.H. Rohm, O. Gursky, Monitoring protein aggregation
during thermal unfolding in circular dichroism experiments, Protein Sci. 15
(2006) 635639.
[9] L.A. Kueltzo, B. Ersoy, J.P. Ralston, C.R. Middaugh, Derivative absorbance
spectroscopy and protein phase diagrams as tools for comprehensive protein
characterization: a bGCSF case study, J. Pharm. Sci. 92 (2003) 18051820.
[10] C.N. Pace, G.R. Grimsley, S.T. Thomas, G.I. Makhatadze, Heat capacity change
for ribonuclease A folding, Protein Sci. 8 (1999) 15001504.
[11] M.W. Pantoliano, E.C. Petrella, J.D. Kwasnoski, V.S. Lobanov, J. Myslik, E. Graf, T.
Carver, E. Asel, B.A. Springer, P. Lane, F.R. Salemme, High-density miniaturized
thermal shift assays as a general strategy for drug discovery, J. Biomol. Screen.
6 (2001) 429440.
[12] D. Matulis, J.K. Kranz, F.R. Salemme, M.J. Todd, Thermodynamic stability of
carbonic anhydrase: measurements of binding afnity and stoichiometry
using ThermoFluor, Biochemistry 44 (2005) 52585266.
[13] L. Stryer, The interaction of a naphthalene dye with apomyoglobin and
apohemoglobin. A uorescent probe of non-polar binding sites, J. Mol. Biol. 13
(1965) 482495.
[14] E. Daniel, G. Weber, Cooperative effects in binding by bovine serum albumin. I.
The binding of 1-anilino-8-naphthalenesulfonate. Fluorimetric titrations,
Biochemistry 5 (1966) 18931900.
[15] M.C. Lo, A. Aulabaugh, G. Jin, R. Cowling, J. Bard, M. Malamas, G. Ellestad,
Evaluation of uorescence-based thermal shift assays for hit identication in
drug discovery, Anal. Biochem. 332 (2004) 153159.
[16] D.E. Epps, R.W. Sarver, J.M. Rogers, J.T. Herberg, P.K. Tomich, The ligand afnity
of proteins measured by isothermal denaturation kinetics, Anal. Biochem. 292
(2001) 4050.
[17] T.H. Steinberg, L.J. Jones, R.P. Haugland, V.L. Singer, SYPRO orange and SYPRO
red protein gel stains: one-step uorescent staining of denaturing gels for
detection of nanogram levels of protein, Anal. Biochem. 239 (1996) 223237.
[18] T.H. Steinberg, H.M. White, V.L. Singer, Optimal lter combinations for
photographing SYPRO orange or SYPRO red dye-stained gels, Anal. Biochem.
248 (1997) 168172.
[19] T.H. Steinberg, R.P. Haugland, V.L. Singer, Applications of SYPRO orange and
SYPRO red protein gel stains, Anal. Biochem. 239 (1996) 238245.
[20] F.H. Niesen, H. Berglund, M. Vedadi, The use of differential scanning
uorimetry to detect ligand interactions that promote protein stability, Nat.
Protoc. 2 (2007) 22122221.
[21] A.I. Alexandrov, M. Mileni, E.Y. Chien, M.A. Hanson, R.C. Stevens, Microscale
uorescent thermal stability assay for membrane proteins, Structure 16 (2008)
351359.
[22] W.M. Rabeh, F. Bossard, H. Xu, T. Okiyoneda, M. Bagdany, C.M. Mulvihill, K. Du,
S. di Bernardo, Y. Liu, L. Konermann, A. Roldan, G.L. Lukacs, Correction of both
NBD1 energetics and domain interface is required to restore DeltaF508 CFTR
folding and function, Cell 148 (2012) 150163.
205
[23] S. Bershtein, W. Mu, A.W. Serohijos, J. Zhou, E.I. Shakhnovich, Protein quality
control acts on folding intermediates to shape the effects of mutations on
organismal tness, Mol. Cell 49 (2013) 133144.
[24] T. Warne, M.J. Serrano-Vega, C.G. Tate, G.F. Schertler, Development and
crystallization of a minimal thermostabilised G protein-coupled receptor,
Protein Expr. Purif. 65 (2009) 204213.
[25] T.W. Geders, K. Gustafson, B.C. Finzel, Use of differential scanning
uorimetry to optimize the purication and crystallization of PLP-
dependent enzymes, Acta Crystallogr., Sect. F: Struct. Biol. Cryst. Commun.
68 (2012) 596600.
[26] H.M. Berman, T. Battistuz, T.N. Bhat, W.F. Bluhm, P.E. Bourne, K. Burkhardt, Z.
Feng, G.L. Gilliland, L. Iype, S. Jain, P. Fagan, J. Marvin, D. Padilla, V.
Ravichandran, B. Schneider, N. Thanki, H. Weissig, J.D. Westbrook, C.
Zardecki, The protein data bank, Acta Crystallogr. D Biol. Crystallogr. 58
(2002) 899907.
[27] J. Newman, Novel buffer systems for macromolecular crystallization, Acta
Crystallogr. D Biol. Crystallogr. 60 (2004) 610612.
[28] J. Kean, R.M. Cleverley, L. ORyan, R.C. Ford, S.M. Prince, J.P. Derrick,
Characterization of a CorA Mg2+ transport channel from Methanococcus
jannaschii using a Thermouor-based stability assay, Mol. Membr. Biol. 25
(2008) 653663.
[29] M.A. Hoffmann, P.J. van Mil, Heat-induced aggregation of beta-lactoglobulin as
a function of pH, J. Agric. Food Chem. 47 (1999) 18981905.
[30] R.N. Zuniga, A. Tolkach, U. Kulozik, J.M. Aguilera, Kinetics of formation and
physicochemical characterization of thermally-induced beta-lactoglobulin
aggregates, J. Food Sci. 75 (2010) E261E268.
[31] A. Neagu, M. Neagu, A. Der, Fluctuations and the Hofmeister effect, Biophys. J.
81 (2001) 12851294.
[32] T.E. Carver, B. Bordeau, M.D. Cummings, E.C. Petrella, M.J. Pucci, L.E. Zawadzke,
B.A. Dougherty, J.A. Tredup, J.W. Bryson, J. Yanchunas Jr., M.L. Doyle, M.R.
Witmer, M.I. Nelen, R.L. DesJarlais, E.P. Jaeger, H. Devine, E.D. Asel, B.A.
Springer, R. Bone, F.R. Salemme, M.J. Todd, Decrypting the biochemical
function of an essential gene from Streptococcus pneumoniae using
ThermoFluor technology, J. Biol. Chem. 280 (2005) 1170411712.
[33] G. Xie, S.N. Timasheff, Temperature dependence of the preferential
interactions of ribonuclease A in aqueous co-solvent systems:
thermodynamic analysis, Protein Sci. 6 (1997) 222232.
[34] A. Dias, D. Bouvier, T. Crepin, A.A. McCarthy, D.J. Hart, F. Baudin, S. Cusack, R.W.
Ruigrok, The cap-snatching endonuclease of inuenza virus polymerase
resides in the PA subunit, Nature 458 (2009) 914918.
[35] U.B. Ericsson, B.M. Hallberg, G.T. Detitta, N. Dekker, P. Nordlund, Thermouor-
based high-throughput stability optimization of proteins for structural studies,
Anal. Biochem. 357 (2006) 289298.
[36] A.B. Datta, S. Roy, P. Parrack, Disorder-order transition of lambda CII promoted
by low concentrations of guanidine hydrochloride suggests a stable core and a
exible C-terminus, FEBS J. 270 (2003) 44394446.
[37] H. Lu, H. Zhang, Q. Wang, H. Yuan, W. He, Z. Zhao, Y. Li, Purication, refolding
of hybrid hIFNgamma-kringle 5 expressed in Escherichia coli, Curr. Microbiol.
42 (2001) 211216.
[38] F. Edwin, Y.V. Sharma, M.V. Jagannadham, Stabilization of molten globule state
of papain by urea, Biochem. Biophys. Res. Commun. 290 (2002) 14411446.
[39] B. Zelent, K.A. Sharp, J.M. Vanderkooi, Differential scanning calorimetry and
uorescence study of lactoperoxidase as a function of guanidiniumHCl, urea,
and pH, Biochim. Biophys. Acta 2010 (1804) 15081515.
[40] C.N. Pace, T. McGrath, Substrate stabilization of lysozyme to thermal and
guanidine hydrochloride denaturation, J. Biol. Chem. 255 (1980) 38623865.
[41] D.K. Eggers, J.S. Valentine, Crowding and hydration effects on protein
conformation: a study with solgel encapsulated proteins, J. Mol. Biol. 314
(2001) 911922.
[42] X. Xie, S. Rao, P. Walian, V. Hatch, G.N. Phillips Jr., C. Cohen, Coiledcoil packing
in spermine-induced tropomyosin crystals. A comparative study of three
forms, J. Mol. Biol. 236 (1994) 12121226.
[43] C. Qian, L. Lagace, M.J. Massariol, C. Chabot, C. Yoakim, R. Deziel, L. Tong, A
rational approach towards successful crystallization and crystal treatment of
human cytomegalovirus protease and its inhibitor complex, Acta Crystallogr. D
Biol. Crystallogr. 56 (2000) 175180.
[44] K. Phillips, A.H. de la Pena, The Combined use of the Thermouor assay and
ThermoQ Analytical Software for the Determination of Protein Stability and
buffer Optimization as an aid in Protein Crystallization, In: Frederick M.
Ausubel (Ed.), et al., Current Protocols in Molecular Biology, 2011, Chapter 10,
Unit10 28.
[45] F. Dupeux, M. Rower, G. Seroul, D. Blot, J.A. Marquez, A thermal stability assay
can help to estimate the crystallization likelihood of biological samples, Acta
Crystallogr. D Biol. Crystallogr. 67 (2011) 915919.
[46] J.J. Lavinder, S.B. Hari, B.J. Sullivan, T.J. Magliery, High-throughput thermal
scanning: a general, rapid dye-binding thermal shift screen for protein
engineering, J. Am. Chem. Soc. 131 (2009) 37943795.
[47] A.P. Yeh, A. McMillan, M.H. Stowell, Rapid and simple protein-stability
screens: application to membrane proteins, Acta Crystallogr. D Biol.
Crystallogr. 62 (2006) 451457.
[48] D.R. Booth, M. Sunde, V. Bellotti, C.V. Robinson, W.L. Hutchinson, P.E. Fraser,
P.N. Hawkins, C.M. Dobson, S.E. Radford, C.C. Blake, M.B. Pepys, Instability,
unfolding and aggregation of human lysozyme variants underlying amyloid
brillogenesis, Nature 385 (1997) 787793.
[49] K. Yutani, K. Ogasahara, T. Tsujita, Y. Sugino, Dependence of conformational
stability on hydrophobicity of the amino acid residue in a series of variant
206 S. Boivin et al. / Protein Expression and Purication 91 (2013) 192206
proteins substituted at a unique position of tryptophan synthase alpha
subunit, Proc. Natl. Acad. Sci. USA 84 (1987) 44414444.
[50] K. Takano, Y. Yamagata, S. Fujii, K. Yutani, Contribution of the hydrophobic effect
to the stability of human lysozyme: calorimetric studies and X-ray structural
analyses of the nine valine to alanine mutants, Biochemistry 36 (1997) 688698.
[51] T. Kajander, M.J. Lehtinen, S. Hyvarinen, A. Bhattacharjee, E. Leung, D.E.
Isenman, S. Meri, A. Goldman, T.S. Jokiranta, Dual interaction of factor H with
C3d and glycosaminoglycans in hostnonhost discrimination by complement,
Proc. Natl. Acad. Sci. USA 108 (2011) 28972902.
[52] C.J. Layton, H.W. Hellinga, Quantitation of proteinprotein interactions by
thermal stability shift analysis, Protein Sci. 20 (2011) 14391450.
[53] G.A. Senisterra, E. Markin, K. Yamazaki, R. Hui, M. Vedadi, D.E. Awrey,
Screening for ligands using a generic and high-throughput light-scattering-
based assay, J. Biomol. Screen. 11 (2006) 940948.
[54] J.F. Brandts, L.N. Lin, Study of strong to ultratight protein interactions using
differential scanning calorimetry, Biochemistry 29 (1990) 69276940.
[55] F.P. Schwarz, K. Puri, A. Surolia, Thermodynamics of the binding of
galactopyranoside derivatives to the basic lectin from winged bean
(Psophocarpus tetrogonolobus), J. Biol. Chem. 266 (1991) 2434424350.
[56] G.A. Holdgate, W.H. Ward, Measurements of binding thermodynamics in drug
discovery, Drug Discovery Today 10 (2005) 15431550.
[57] S.P. Santos, T.M. Bandeiras, A.F. Pinto, M. Teixeira, M.A. Carrondo, C.V. Romao,
Thermouor-based optimization strategy for the stabilization and
crystallization of Campylobacter jejuni desulforubrerythrin, Protein Expr.
Purif. 81 (2012) 193200.
[58] W.A. Lea, A. Simeonov, Differential scanning uorometry signatures as
indicators of enzyme inhibitor mode of action: case study of glutathione S-
transferase, PLoS One 7 (2012) e36219.