Enhancing Recombinant Protein Quality and
`Yield by Protein Stability Profiling
`
`TARA M. MEZZASALMA, JAMES K. KRANZ, WINNIE CHAN, GEOFFREY
`T. STRUBLE, CÉLINE SCHALK-HIHI, INGRID C. DECKMAN, BARRY A.
`SPRINGER, and MATTHEW J. TODD
`
`The reliable production of large amounts of stable, high-quality proteins is a major challenge facing pharmaceutical protein
`biochemists, necessary for fulfilling demands from structural biology, for high-throughput screening, and for assay purposes
`throughout early discovery. One strategy for bypassing purification challenges in problematic systems is to engineer multi-
`ple forms of a particular protein to optimize expression, purification, and stability, often resulting in a nonphysiological sub-
`domain. An alternative strategy is to alter process conditions to maximize wild-type construct stability, based on a specific
`protein stability profile (PSP). ThermoFluor®, a miniaturized 384-well thermal stability assay, has been implemented as a
`means of monitoring solution-dependent changes in protein stability, complementing the protein engineering and purifica-
`tion processes. A systematic analysis of pH, buffer or salt identity and concentration, biological metals, surfactants, and com-
`mon excipients in terms of an effect on protein stability rapidly identifies conditions that might be used (or avoided) during
`protein production. Two PSPs are presented for the kinase catalytic domains of Akt-3 and cFMS, in which information
`derived from a ThermoFluor® PSP led to an altered purification strategy, improving the yield and quality of the protein using
`the primary sequences of the catalytic domains. (Journal of Biomolecular Screening 2007:418-428)
`
`Key words: Akt-3, cFMS, protein stability, ThermoFluor®, assay development
`
`INTRODUCTION
`
`AGGREGATION IS A COMMON OBSTACLE both in protein bio-
`
`chemistry and in protein therapeutics. For more than 50
`years, scientists have been struggling to understand the relationship
`between reversible protein denaturation and competing irreversible
`processes.1,2 Protein aggregation is highly correlated with certain
`diseases, such as Alzheimer’s, Parkinson’s, and Huntington’s, and
`prion-mediated amyloidogenic diseases.3,4 In addition, aggregation
`is an impediment toward formulation of biologicals during devel-
`opment5,6 and in protein overexpression and purification during
`discovery.
`Protein stability measurements, and finding means of increas-
`ing stability, have similarly spurred widespread interest in the
`scientific community. A thermodynamic description of protein
`stability requires quantitation of the equilibrium between folded
`and unfolded states, either by denaturant-induced7,8 or thermally
`induced unfolding.9,10 In addition to denaturants and temperature,
`
`Johnson & Johnson Pharmaceutical Research & Development, LLC, Exton,
`Pennsylvania.
`
`Received Sep 18, 2006, and in revised form Oct 25, 2006. Accepted for publi-
`cation Nov 21, 2006.
`
`Journal of Biomolecular Screening 12(3); 2007
`DOI:10.1177/1087057106297984
`
`variables known to influence protein stability include pH and
`proton linkage,8,11 salt type and concentration,12 cosolvents and
`osmolytes,12,13 preservatives,5,14 and surfactants.15 Variations in
`any of these can alter ligand binding and enzymatic activity, may
`influence unfolding and aggregation,5,8 and can drive crystalliza-
`tion in structural biology programs.12
`Solution effects on protein stability are generally discovered
`through serial observations. A common approach to augmenting
`protein stability is to engineer constructs, potentially also improv-
`ing expression, purification, or activity. Alternatively, investiga-
`tion of the protein stability landscape as a function of its
`environment may be used to overcome problems of stability with-
`out the need to vary the primary protein sequence through muta-
`genesis. A thorough investigation of intrinsic protein stability in
`the context of environmental factors may achieve the same
`improvements in protein stability without the need for protein
`engineering. Over time, a detailed picture of a protein’s unique
`stability may be fully explored.
`The unique challenges associated with understanding in
`detail each new pharmacologically important protein are most
`efficiently addressed by a systematic approach. ThermoFluor®1
`
`1 The ThermoFluor® assay was developed by 3-Dimensional Pharmaceuticals,
`Inc, which has been merged into Johnson & Johnson Pharmaceutical Research
`& Development, LLC “ThermoFluor” is a trademark registered in the United
`States and certain other countries.
`
`418 www.sbsonline.org
`
`© 2007 Society for Biomolecular Sciences
`
`KASHIV EXHIBIT 1073
`IPR2019-00791
`
`Page 1
`
`

`

`ThermoFluor®®-based Protein Stability Profiling
`
`provides a high-throughput measure of stability perturbations
`due to effects of ligand binding and solution conditions.16,17.
`ThermoFluor® was originally developed for high-throughput
`screening (HTS), yet it has other applications. Practically speak-
`ing, assay development in advance of ThermoFluor®-based HTS
`is a function of maximizing the fluorescent signal corresponding
`to protein unfolding, minimizing the concentration of protein per
`well, and simultaneously maximizing protein longevity and plate
`reproducibility over the course of robotic manipulation (up to
`24 h). In addition, applications of ThermoFluor® protein stabil-
`ity profiling (PSP) have had a surprising impact on protein pro-
`duction, notably for the 2 example proteins presented here: the
`kinase catalytic domains from Akt-3 and cFMS.
`The demand for finding therapeutic agents active against
`kinases is significant.18,19 One example is the c-fms proto-
`oncogene, a receptor protein tyrosine kinase, which is involved in
`regulating differentiation and maturation of most macrophages.20,21
`A 2nd example is the soluble Ser/Thr kinase Akt-3, a member
`of the AGC kinase family and a critical component of intracel-
`lular signaling controlling the response to insulin and inflamma-
`tory agents.22 Specifically, Akt-3 is overexpressed and amplified
`in different tumors22,23 while not being highly expressed in liver
`or skeletal muscle (where Akt-1 and Akt-2 expression is rele-
`vant), suggesting Akt-3 may be a promising target for the dis-
`covery of novel chemotherapeutics that do not interfere with
`insulin signaling.22
`Eukaryotic protein kinases are composed of a 250-amino-
`acid catalytic domain, under control of 1 or more regulatory
`domains.24 During initial characterization of cFMS and Akt-3
`in preparation for separate HTS campaigns, progress in both
`systems suffered from aggregation during protein purification.
`In lieu of a protein engineering effort, we pursued a systematic
`approach to generate a PSP using ThermoFluor® technology.
`From each, a different set of solution conditions was identified
`as being stabilizing and/or destabilizing to the core catalytic
`domains. Following changes to protein purification procedures,
`the aggregation problem was alleviated; purification quality
`and yield was increased for both systems.
`
`MATERIALS AND METHODS
`
`ANS (1-anilino-8-naphthanlenesulfonate) was from Molecular
`Probes (Carlsbad, CA); buffers, salts, and additives were from
`Sigma-Aldrich (St Louis, MO). PCR plates with 384 wells
`(Abgene, Epsom, UK) were used for ThermoFluor® experiments.
`Thrombin was from Enzyme Research Labs (South Bend, IN).
`
`Protein expression and purification
`
`cFMS purification. Cloning of c-fms and protein overexpression
`is described elsewhere.25 Briefly, the cFMS catalytic domain was
`expressed in Sf9 cells; cell pellets were thawed in lysis/column
`
`wash buffer (25 mM HEPES pH 7.5, 400 mM NaCl, 10%
`glycerol, 1 mM glutathione, 20 mM imidazole, 0.1mM PMSF,
`(1×) complete EDTA-free protease inhibitors [Roche, Basel,
`Switzerland]), followed by dounce homogenization and clarifica-
`tion by centrifugation (40,000g, 40 min). Initially, cFMS was
`batch purified on an Ni-NTA Superflow resin (Qiagen, Venlo, the
`Netherlands) at 4 °C and was eluted using a linear gradient
`up to 100 mM imidazole. Following ThermoFluor®-based PSP
`(described below), alterations were made to the above cFMS
`purification protocol: TALON metal affinity resin (Clontech
`Laboratories, Inc, Mountain View, CA) was substituted for Ni-
`NTA resin, the column buffer was changed to 25 mM KH2PO4,
`in lieu of HEPES buffer (pH unchanged at 7.5), and glycerol
`was reduced from 10% to 5%. The column gradient was also
`modified using a lower initial imidazole concentration (reduced
`from 20 mM to 5 mM) and eluting with a gradient to 200 mM
`imidazole.
`
`Akt-3 purification. A description of Akt-3 cloning and overex-
`pression is described elsewhere.22 Cell pellets were resus-
`pended in cell lysis buffer (20 mM Tris pH 8.0, 100 mM NaCl,
`1 mM EDTA, 0.5% NP-40, 2 mM DTT, (1×) Roche protease
`inhibitor tablets, lysed (Avestin Emulsiflex-C5), and clarified
`by centrifugation (56,000g, 50 min) to remove insoluble mate-
`rial. The Akt-3 was batch purified on Glutathione Sepharose 4
`Fast Flow resin (Pharmacia, New York, NY) overnight at 4° C
`on an AKTA Explorer System (GE Healthcare, Piscataway,
`NJ). The protein-bound resin was washed with cell lysis buffer
`and then with thrombin cleavage buffer (50 mM Tris pH 8.0,
`150 mM NaCl, 1 mM EDTA, and 1 mM DTT). The GST-Akt-
`3–bound resin was recovered, followed by batch thrombin cleav-
`age (∼9 NIH units thrombin/mg fusion protein, 25 °C ∼2 h).
`Thrombin was removed by batch binding to benzamidine
`sepharose. Following ThermoFluor®-based PSP (described
`below), alterations were made: the thrombin cleavage buffer
`was changed (50 mM PIPES pH 6.5, 100 mM NaCl, 10% glyc-
`erol), and the fusion protein cleavage conditions were reopti-
`mized for ∼18 h.
`The oligomeric state of both Akt-3 and cFMS was assessed
`using a Superdex 200 column pre-equilibrated in the appropri-
`ate buffer; elution was monitored at 280 nm. Standards were
`200, 150, 66, 29, and 12 kD (Sigma).
`
`Assembly of 384-condition plates
`
`Three 384-well plates were developed, each designed to test
`a different set of common biochemical conditions: a pH/salt
`plate, a crystallography buffer plate, and an excipient/additive
`plate. All conditions appear in left-right duplicates (layouts are
`supplied as supplemental materials), prepared as (2×) stock
`plates. Each solution was individually pH adjusted to the des-
`ignated pH (pH 7.0 for the excipient plate).
`
`Journal of Biomolecular Screening 12(3); 2007
`
`www.sbsonline.org
`
`419
`
`Page 2
`
`

`

`Mezzasalma et al.
`
`Table 1. Variation in Tm and ∆
`UHapp(Tm) for cFMS and
`Akt-3 in Different Buffers (Figure 1)
`
`cFMS
`
`Akt-3
`
`∆U
`
`Happ(Tm)
`(cal/mol)
`
`Tm (°C)
`
`∆U
`
`Happ(Tm)
`(cal/mol)
`
`Tm (°C)
`
`50.2
`
`52.6
`
`55.3
`
`53.6
`
`54.8
`
`95,500
`
`95,000
`
`88,000
`
`92,000
`
`99,000
`
`57.0
`
`55.2
`
`55.3
`
`56.9
`
`55.5
`
`104,000
`
`95,000
`
`90,000
`
`64,000
`
`81,000
`
`Buffer Condition
`
`25 mM PIPES,
`pH 6.0, 100 mM NaCl
`25 mM PIPES,
`pH 7.0, 100 mM NaCl
`25 mM PIPES,
`pH 7.0, 500 mM NaCl
`25 mM Pi, pH 7.0,
`100 mM NaCl
`100 mM Pi, pH 7.0
`(no additional NaCl)
`
`Here, effects of solution conditions on protein stability are
`measured using a matrix-based approach in a of 384-well char-
`acterization plates (For supplemental materials, go to http://
`jbx.sagepub.com/cgi/content/full/12/3/418/DC1.) containing vari-
`able buffer composition by well within prearranged plates. These
`are combined with cFMS or Akt-3 with ANS into assay plates and
`then are subjected to ThermoFluor® protein unfolding analysis.
`Representative normalized ThermoFluor®-derived thermal denatu-
`ration data depict a subset of results from plate-based exploration of
`cFMS and Akt-3 stability profiling (Fig. 1, Table 1). For graphical
`comparison of ∆Tm, relative fluorescence intensity is normalized to
`the fitted baselines for native and denatured states for each unfold-
`ing transition to express fraction unfolded protein versus tempera-
`ture.16,17 The Tm of cFMS is increased +2.4 °C from pH 6.0 to 7.0,
`suggesting a link between stability and a pH-dependent protonation
`event on the protein (Fig. 1A). Stability is also dependent on ionic
`= +2.7 °C)
`strength, based on an observed increase in stability (∆Tm
`when [NaCl] is raised from 100 to 500 mM at pH 7.0. In contrast
`= –1.8 °C) in
`to cFMS, Akt-3 stability (Fig. 1B) is decreased (∆Tm
`response to elevating solution pH from 6.0 to 7.0, whereas chang-
`ing NaCl has little effect on stability.
`The effect of NaPO4 on cFMS and Akt-3 stability is evaluated
`at 2 concentrations of the buffer (Fig. 1C, D). For cFMS (Fig.
`= +0.8 °C
`1C), phosphate induces an increase in stability, ∆Tm
`from 25 mM to 100 mM NaPO4, an effect above an ionic strength
`effect alone. Conversely, Akt-3 thermal stability is diminished,
`= –1.4 °C, in response to increasing phosphate concentration
`∆Tm
`(Fig. 1D); there is an additional effect of phosphate concentration
`on the apparent unfolding enthalpy (Table 1), decreasing from 81
`kcal/mol to 64 kcal/mol in 100 mM and 25 mM NaPO4, respec-
`tively. This effect is visually confirmed by a change in the slope
`of the unfolding transition around the fitted Tm of Akt-3.
`Relatively minor changes in solution conditions can have a
`sizeable effect on protein stability. An approach to address many
`
`ThermoFluor®® assay plate assembly
`ThermoFluor® assays are generally ≤ 4 µL, ∼1 to 5 µM protein
`(generally 20-100 µg/mL). For each experiment, 2 µL of the rele-
`vant (2×) stock solutions were dispensed into assay plates, followed
`by addition of 2 µL protein solution, containing both (2×) protein
`and ANS in a dilute solution (eg, 5 mM PIPES, pH 7.5, 10 mM
`NaCl). One microliter of silicone oil overlay prevents loss from
`evaporation during sample heating. Assay plates were centrifuged
`at ∼1000g, ∼1 min (Eppendorf plate centrifuge). For both Akt-3 and
`cFMS, 50 µM ANS was included in the final protein solution, and
`the protein concentration was 0.1 mg/mL.
`Assay plates were analyzed using ThermoFluor® instrumenta-
`tion17 manufactured within Johnson & Johnson Pharmaceutical
`Research and Development, LLC. ThermoFluor® instruments
`were programmed to ramp temperature from 25 °C to 85 °C at
`∼1 °C/min; fluorescence was measured at 1 °C increments, imag-
`ing plate fluorescence via CCD camera. Fluorescence is generated
`by Hg-Xe arc lamp (Hamamatsu, Bridgewater, NJ), filtered
`through custom interference excitation (385 ± 20 nm) and emis-
`sion (500 ± 25 nm) filters (Omega Optical, Battleboro, VT). The
`primary data (relative fluorescence intensity v. temperature) are fit
`to standard equations describing protein thermal stability as previ-
`ously described,16 giving 6 fit parameters: Tm, the midpoint in the
`transition between native and nonnative protein; ∆
`UH(T), the van’t
`Hoff enthalpy for reversibly unfolding reactions; and linear base-
`lines for native and unfolded protein (referenced to the protein Tm).
`The precision in Tm determination is generally ±0.1 °C to 0.2 °C,
`although it is somewhat dependent on the type and concentration
`of both protein and dye.
`
`RESULTS
`
`Protein Stability Profiling
`ThermoFluor® is a protein stability assay that uses an envi-
`ronmentally sensitive dye to monitor the amount of unfolded
`protein in solution as a function of temperature.16,17 As the tem-
`perature increases, the fraction of nonnative protein increases,
`producing a cooperative unfolding transition. The midpoint
`temperature, Tm, of such a transition is defined as the tempera-
`ture at which the concentration of native and nonnative protein
`is equivalent ([N] = [U]), and the equilibrium between native
`= [U]/[N] = 1; thus, the free energy
`and nonnative species is KU
`is zero, as ∆
`= –nRT ln KU . Under conditions in which
`UG(Tm)
`unfolding is reversible, fits also provide information on the
`enthalpy of unfolding, ∆
`UH(Tm), and heat capacity of unfolding,
`∆
`UCp, providing a complete description of the temperature
`dependence of protein stability.26 Owing to the simple relation-
`ship between Tm and ∆
`UG(T) for the free energy of protein
`unfolding, the effects of solution composition on Tm can be
`directly related to changes in protein stability.
`
`420 www.sbsonline.org
`
`Journal of Biomolecular Screening 12(3); 2007
`
`Page 3
`
`

`

`ThermoFluor®®-based Protein Stability Profiling
`
`C
`
`fU
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`40
`
`45
`
`55
`50
`Temperature (˚C)
`
`60
`
`65
`
`40
`
`45
`
`55
`50
`Temperature (˚C)
`
`60
`
`65
`
`D
`
`fU
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`A
`
`fU
`
`B
`
`fU
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`40
`
`45
`
`55
`50
`Temperature (˚C)
`
`60
`
`65
`
`40
`
`45
`
`55
`50
`Temperature (˚C)
`
`60
`
`65
`
`FIG. 1. Solution dependence on protein stability. Normalized ThermoFluor® data for cFMS (A, C) and Akt-3 (B, D) from 5 wells of an ∼1000-
`well protein stability profile (PSP). (A, B) Stability measured in 25 mM PIPES, pH 6.0 (blue) or at pH 7.0 (pink and maroon), with 100 mM
`(open) or 500 mM (filled) NaCl. (C, D) Stability measured in 25 mM NaPO4, 100 mM NaCl (open circles), or 100 mM NaPO4 (filled circles),
`both at pH 7.0. Lines represent fits to unfolding equations.16
`
`variables simultaneously is a matrix of conditions in which
`individual variables might be systematically varied, testing sta-
`bility versus a large set of conditions in parallel. Three arrays of
`conditions (supplemental materials) were assayed for cFMS and
`Akt-3. The 1st explores pH, buffer identity, concentration, salt
`concentration, and the presence of the common biological metal
`magnesium. The 2nd assays common biochemical excipients.
`The 3rd assesses protein stability in buffers commonly used in
`crystallography. These matrices of conditions are routinely
`examined for systematic effects on Tm, on protein (or dye) fluo-
`rescence, and on protein unfolding enthalpy, ∆
`U,Happ(Tm).
`Stability surfaces are readily generated from matrix-based
`protein characterization studies; the effect of pH and NaCl on
`
`cFMS (Fig. 2A) or Akt-3 (Fig. 2B) thermal stability are gener-
`ated from a 9 × 7 subset of conditions, composed of 63 indi-
`vidual measurements of protein unfolding from within the
`∼1000-well characterization profile. These stability surfaces
`are useful in graphically representing global differences across
`protein samples. It is apparent that cFMS and Akt-3 differ in
`the dependence of stability on pH and ionic strength (stabiliz-
`ing for cFMS and destabilizing for Akt-3).
`Significantly more information can be discerned by compar-
`ing individual combinations of buffer effects, such as the effect
`of systematic variation of pH and salt on the stability of cFMS
`and Akt-3 (Fig. 3). The pH effect identified for cFMS stability in
`Figure 1A is expanded in Figure 3A, showing fluctuations in
`
`Journal of Biomolecular Screening 12(3); 2007
`
`www.sbsonline.org
`
`421
`
`Page 4
`
`

`

`Mezzasalma et al.
`
`The effect of NaCl concentration on the stability of both pro-
`teins was further defined by comparing stability over a gradient
`of NaCl, using different buffers within the condition profiling
`plates. Comparing 4 different buffers at pH 7.0, cFMS is stabilized
`(∆Tm
`∼ 4 °C) as a function of increasing [NaCl] concentration
`(Fig. 3C). The same conditions show a moderate destabilization of
`Akt-3 as a function of ionic strength (Fig. 3D). A further compar-
`ison of 4 different buffers shows a buffer effect on cFMS stability
`when compared across any individual [NaCl]; cFMS stability is
`highest in phosphate buffer, followed by PIPES and then HEPES
`and MOPS. The opposite is seen to be true for Akt-3. From these
`data alone, it is difficult to discern a mechanism for this rank order,
`although the trend is consistent across varying concentrations of
`buffer components. Figures 3C and 3D also define how protein
`stability is affected by 100 mM of each buffer at no added NaCl
`(filled symbols), relative to Tm values measured in 25 mM buffer
`and varying [NaCl] (open symbols). These data are consistent with
`a buffer-specific stabilizing effect of phosphate on cFMS relative
`to a general ionic strength effect of HEPES and MOPS (note
`Tm increases from 51 °C to 55 °C, resulting from increased phos-
`phate from 25 to 100 mM). A different trend is observed for Akt-
`3, in which high buffer concentration destabilizes protein by an
`amount equivalent to the ionic strength change. These same pH
`and salt effects are mirrored in other combinations of primary solu-
`tion conditions comparing cFMS and Akt-3 stability (not shown).
`A more extensive profile of buffer effect on protein stability
`(at pH 7.5) was undertaken using a set of buffers commonly used
`in crystallography, tested at either 50 mM or 100 mM, as a func-
`tion of varying pH. Selected data collected using 9 buffers, each
`at pH 7.5, is shown for cFMS and Akt-3 (Fig. 4A, B, respectively).
`cFMS has the highest stability in phosphate, citrate, and succinate
`buffers, relative to a water reference Tm; buffers that show signifi-
`cant destabilization for cFMS include HEPES, imidazole-maleate,
`and bis-Tris-propane but, interestingly, not Tris. These data are
`consistent with that in Figure 3C and suggest that not only phos-
`phate but also citrate and perhaps succinate may bind to the pro-
`tein. It should be noted that many proteins are stabilized by
`citrate and succinate buffers (Tara M. Mezzasalma, unpublished
`data), either because of the high ionic strength accompanying
`these buffers or because of specific binding to the protein. These
`components of the citric acid cycle are at moderate concentra-
`tions in vivo and may have a relevant effect on protein stability
`and potentially enzyme activity. Akt-3 exhibits a different stabil-
`ity pattern against these same 9 buffers (Fig. 4B); bis-Tris-
`propane was destabilizing and HEPES was slightly stabilizing,
`whereas all other buffers had no statistically significant effect on
`stability. Notably, the citrate- and succinate-mediated stabiliza-
`tion observed for cFMS is absent for Akt-3, highlighting the pro-
`tein-specific nature of PSP.
`An excipient/additive plate is also assayed (at 2 concentrations
`per component); the general composition includes variations in
`salts, the chloride salts of common divalent cations (such as CaCl2
`
`FIG. 2. Protein stability surfaces. A subset of ThermoFluor®-derived
`Tm values from the pH-salt profile plotted as a function of NaCl and pH,
`generating a stability surface for (A) cFMS and (B) Akt-3. Stability sur-
`faces represent 9 buffers (acetate, pH 4 and 5; MES, pH 6 and 6.5,
`HEPES, pH 7, 7.5, and 8; borate, pH 8.5; each at 25 mM) and 7 [NaCl]
`(25, 50, 100, 200, 300, 400, and 500 mM).
`
`cFMS stability from pH 5.5 to 8.0 in each of 4 different solu-
`tions: a PIPES buffer plus 0 or 100 mM NaCl and 0 or 5 mM
`MgCl2. The maximum in cFMS stability is achieved at a pH
`greater than 7.0, regardless of NaCl and MgCl2 concentration.
`NaCl at 100 mM causes a uniform, pH-independent increase in
`∼ +2° C, consistent with a general stabilizing effect
`stability, ∆Tm
`of NaCl.8 MgCl2 has limited effect on cFMS stability under any
`of these conditions. Figure 3B demonstrates a significantly dif-
`ferent pH effect on Akt-3 as compared to cFMS, indicating a pH
`of maximum stability between 6 and 6.5. For Akt-3, NaCl and/or
`MgCl2 caused a uniformly modest pH-independent decrease in
`stability.
`
`422 www.sbsonline.org
`
`Journal of Biomolecular Screening 12(3); 2007
`
`Page 5
`
`

`

`ThermoFluor®®-based Protein Stability Profiling
`
`5.5
`
`6.0
`
`6.5
`
`7.0
`
`7.5
`
`8.0
`
`pH
`
`A
`
`Tm
`
`B
`
`Tm
`
`56
`
`54
`
`52
`
`50
`
`48
`
`46
`
`60
`
`58
`
`56
`
`54
`
`52
`
`50
`
`0
`
`100
`
`200
`
`300
`[NaCl]
`
`400
`
`500
`
`C
`
`Tm
`
`D
`
`Tm
`
`56
`
`54
`
`52
`
`50
`
`48
`
`46
`
`60
`
`58
`
`56
`
`54
`
`52
`
`50
`
`5.5
`
`6.0
`
`6.5
`
`7.0
`
`7.5
`
`8.0
`
`0
`
`100
`
`200
`
`pH
`
`300
`[NaCl]
`
`400
`
`500
`
`FIG. 3. Stability versus pH, ionic strength, and buffer. ThermoFluor®-derived Tm for cFMS (A, C) and Akt-3 (B, D) showing different depen-
`dencies on pH and NaCl. (A, B) Conditions were 25 mM PIPES at varying pH, with 0 mM (open) or 5 mM (filled) MgCl2 and 0 mM (squares)
`or 100 mM (triangles) NaCl. (C, D) Stability versus ionic strength determined using 25 mM (open) or 100 mM (closed) of NaPO4 (diamonds),
`PIPES (squares), HEPES (triangles), or MOPS (circles) at pH 7.0, with increasing [NaCl].
`
`or NiCl2), as well as common detergents and additives (such as
`Tween and glycerol). Most salts again show dose-dependent sta-
`bilization of cFMS (Fig. 4C) and destabilization of Akt-3 (Fig.
`4D). Zinc showed a substantial dose-dependent destabilization of
`cFMS but has no effect on Akt-3. Calcium and nickel showed
`little effect on the stability of cFMS and Akt-3, yet these cations
`have significantly altered the stability of some affinity-tagged
`proteins (data not shown).
`
`Optimizing protein purification. The initial purification proce-
`dures for cFMS and Akt-3 catalytic kinase domains were derived
`from either previously published protocols25 or by using a stan-
`
`dard kit-based protocol typically employed for purification of
`affinity-tagged proteins (details are available in the Materials and
`Methods section). Size-exclusion chromatography (SEC) and
`sodium dodecyl sulfate (SDS) polyacrylamide gels were used to
`gauge quality, purity, and tendency to aggregate protein preps
`following affinity chromatography. Initially, both proteins exhib-
`ited a strong tendency to aggregate and generally low purity (Fig.
`5A, B); SEC revealed high-molecular-weight species eluting
`prior to the monomeric peaks. Moreover, SDS polyacrylamide
`gel electrophoresis indicated that the monomeric fractions of
`both cFMS and Akt-3 are not homogenous (< 90% purity) and
`were unsuitable for structural studies.
`
`Journal of Biomolecular Screening 12(3); 2007
`
`www.sbsonline.org
`
`423
`
`Page 6
`
`

`

`Mezzasalma et al.
`
`Ni
`
`Zn
`
`Ca
`
`Mn
`
`Mg
`
`NH4
`
`K Li
`
`Na
`
`C
`
`55
`
`Tm
`
`50
`
`45
`
`40
`
`D
`
`60
`
`Tm
`
`55
`
`50
`
`45
`
`Tris
`
`Na-suc
`Na-K-Pi
`ES
`
`PIP
`
`al
`
`Imid-m
`
`Hepes
`
`Na-Citr
`
`Na-Cac
`
`p
`
`B-T-Pro
`
`A
`
`55
`
`Tm
`
`50
`
`45
`
`40
`
`B
`
`60
`
`Tm
`
`55
`
`50
`
`45
`
`Ni
`
`Zn
`
`Ca
`
`Mn
`
`Mg
`
`NH4
`
`K Li
`
`Na
`
`Tris
`
`Na-suc
`
`Na-K-Pi
`ES
`
`PIP
`
`al
`
`Imid-m
`
`Hepes
`
`Na-Citr
`
`Na-Cac
`
`p
`
`B-T-Pro
`
`FIG. 4. Protein stability versus buffer identity, salt identity, and common biological metals. Variation in Tm versus solution composition for cFMS
`(A, C) or AKT3 (B, D) proteins. (A, B) Stability was measured at 50 mM (tan) or 100 mM (blue) buffers; bis-Tris-propane (B-T-Prop), NaCacodylate
`(Na-Cac), NaCitrate (Na-Citr), HEPES, imidazole-malate (Imid-mal), PIPES, NaPO4 and KPO4 (Na-K-Pi), NaSuccinate (Na-suc), and Tris-HCl.
`(C, D) Protein Tm versus chloride salts at low (tan) or high (blue) concentration (25 mM PIPES, pH 7.0). Concentrations: NaCl, KCl, LiCl, and
`= 51 °C (A and C) or
`= 25, 100 µM. Reference lines: cFMS, Tm,ref
`(NH4)Cl = 100, 300 mM; MgCl2 and MnCl2
`= 1, 5 mM; CaCl2, NiCl2, and ZnCl2
`= 55 °C (B and D), each in water.
`Akt-3, Tm,ref
`
`The ThermoFluor®-derived PSPs established for cFMS and
`Akt-3 provided information on conditions that could stabilize
`(or destabilize) these proteins. Stabilizing conditions were
`incorporated into protein purification protocols to limit aggre-
`gation, increase protein purity, and improve yield. For cFMS,
`stability was enhanced at high pH, high salt, and in phosphate
`buffer, whereas stability was decreased by Zn2+ and imidazole
`∼ –1.2 at 100 mM; not shown). Based on these observa-
`(∆Tm
`tions, new conditions for affinity chromatography were chosen.
`First, the buffer was changed from HEPES to NaPO4 (each
`
`were pH 7.5) to exploit the stabilizing effect of phosphate
`on cFMS and to minimize the pH increase associated with
`HEPES buffer at low temperatures. Second, to reduce the
`potential negative impact of imidazole and divalent metal on
`protein aggregation and purity, the Ni-NTA resin was replaced
`with a cobalt-based TALON resin, which limits the amount of
`imidazole required to elute His6-cFMS (data not shown), consis-
`tent with the performance expectations of the TALON resin.27
`Following column elution, the imidazole was removed by imme-
`diate dialysis. Incorporating these changes in cFMS purification
`
`424 www.sbsonline.org
`
`Journal of Biomolecular Screening 12(3); 2007
`
`Page 7
`
`

`

`ThermoFluor®®-based Protein Stability Profiling
`
`FIG. 5. Protein aggregation and purity before and after protein stability profiling (PSP)–optimized conditions. Size exclusion chromatography
`and sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (insets) for cFMS (A, C) or AKT3 (B, D) from kinase purifications
`before (A, B) and after (C, D) optimization; the asterisk denotes monomer fractions analyzed by SDS-PAGE. (A) cFMS protein purified from
`Ni-NTA column, showing aggregation; purity was ∼90% by SDS-PAGE. (B) Akt-3 protein purified by Ni-NTA column showing contamination and
`aggregation; monomer purity was < 90% by gel analysis. (C) Optimized cFMS protein; purity was >98% by SDS-PAGE. (D) Optimized Akt-3;
`purity was >98% by SDS-PAGE.
`
`collectively reduced aggregation and significantly increased the
`purity (>98%) of the monomeric fraction when compared to the
`original purification (Fig. 5A vs 5C).
`By a similar process, several changes were also made to the
`GST-Akt-3 purification, based on the ThermoFluor®-derived
`PSP. The buffer used during purification was changed from
`Tris to PIPES, which has a smaller heat of buffer ionization
`relative to Tris28 (minimizing the increase in pH at low tem-
`peratures, a destabilizing effect). The pH was decreased from
`8.0 to 6.8, where Akt-3 showed better stability (Fig. 3B).
`Based on the sensitivity of Akt-3 to [NaCl] the salt concentra-
`tion was decreased from 150 to 100 mM during purification.
`Finally, 10% glycerol was added (a stabilizing effect, data not
`shown). As a result, Akt-3 aggregation was reduced, purity
`was increased (>98%), and yields improved compared to the
`original Akt-3 purification.
`
`To demonstrate the performance benefits of aggregate free pro-
`tein during in vitro stability assays, non-normalized ThermoFluor®
`stability data are shown for cFMS and Akt-3 (Fig. 6) comparing
`data from initial and optimized protein purification procedures.
`One measure of the amount of (initially) native protein is the ampli-
`tude of ThermoFluor® fluorescence intensity change between
`folded and unfolded forms, ∆y(T) = yU(T) – yF(T), the difference in
`absolute change in fluorescence between pretransition and post-
`transition baseline fluorescence. For cFMS (Fig. 6A), unfolding
`transitions are shown for initial and optimized purification; the
`4-fold difference in signal amplitude is readily apparent (∆y ∼3000
`RFU for aggregate-prone protein compared with ∆y ∼11,000 RFU
`postoptimization). Likewise for Akt-3 (Fig. 6B), the aggregate-
`prone protein has a roughly 2-fold lower signal amplitude asso-
`ciated with unfolding (∆y ∼13,000 RFU initially compared with
`∆y ∼30,000 RFU postoptimization). Although signal amplitude
`
`Journal of Biomolecular Screening 12(3); 2007
`
`www.sbsonline.org
`
`425
`
`Page 8
`
`

`

`Mezzasalma et al.
`
`above functions and is supplied using the same protein expression/
`purification resource. Aggregation is a problem that may become
`a critical bottleneck and is typically addressed by protein engi-
`neering through a combination of random and designed alter-
`ations in the expressed protein sequence. An alternative approach
`is to find conditions that stabilize existing protein constructs.
`Here, a ThermoFluor®-derived PSP facilitated the transition of 2
`targets into drug discovery programs, each of which had issues
`with both aggregation and purity.
`In addition to the primary amino acid sequence and tertiary
`fold, protein stability and the tendency of proteins to aggregate
`are influenced by a number of factors. Ligand binding is well
`documented as a general effector of protein stability; for ligands
`that bind the native state, protein stability is increased in propor-
`tion to the affinity and concentration of ligand.29 Solution com-
`position has also been well established as influencing protein
`stability, either directly through binding or indirectly through sol-
`vent effects that differentially modulate the thermodynamics of
`the native and unfolded protein.8 Finally, as the pH of a system is
`varied, the charged state of the protein will also vary, fundamen-
`tally changing the effect of solution composition on protein sta-
`bility.5 All of these can unpredictably influence protein stability
`and the propensity of a protein to aggregate.
`
`Buffer ionization effects. One caveat associated with experiments
`designed to discriminate effects of solvent composition is the
`contribution of buffer ionization enthalpy (∆
`ionH) and the result-
`ing temperature-dependent pH.28 Buffer solutions are commonly
`prepared at room temperature. Protein purification is routinely
`performed at 4 °C; thus, some modulation in pH as a function of
`temperature can be expected. The pH change associated with this
`temperature difference may unpredictably affect protein stability.
`Likewise, during the course of ThermoFluor® experiments, pro-
`tein solutions are heated well in excess of physiological temper