A recombinant humanized monoclonal antibody of the IgG₁ subclass (IgG) with a molecular weight of 149 kDa, an isoelectric region between 9 and 10 and an aggregation temperature (Tagg = the onset of aggregation monitored by the optical density at 350 nm during a thermal scan) of about 81°C (17) was used for the study. The IgG was formulated at a concentration of 1.0 mg/ml in 100 mM sodium phosphate buffer pH 7.2 without or with different concentrations of polysorbate 20 and polysorbate 80 (Fluka, Sigma-Aldrich, Buchs Switzerland). The formulations were filtered using 0.22 µm PES low binding syringe driven filter units (Millex^TM GP, Millipore, Ireland). Sucrose, mannitol, sodium citrate, sodium citrate dehydrate and citric acid monohydrate were from Fluka (Fluka, Sigma-Aldrich Steinheim, Germany), glycerol from Merck (Merck KGaA, Darmstadt, Germany), arginine hydrochloride from SAFC (SAFC, Sigma-Aldrich Steinheim, Germany), ethanol and methanol from Biosolve (Biosolve B.V., Valkenswaard, The Netherlands), NaH₂PO₄ dihydrate, Na₂HPO₄ dihydrate and NaCl were from Sigma (Sigma, Sigma-Aldrich Steinheim, Germany).

4,4´-Dianilino-1,1´-binaphthyl-5,5´-disulfonic acid dipotassium salt, Bis-ANS, 8-anilino-1-naphthalenesulfonic acid, ANS, 9-(2-carboxy-2-cyanovinyl)julolidine, CCVJ (Sigma, Sigma-Aldrich Steinheim, Germany), 9-(2,2-dicyanovinyl)julolidine, DCVJ and Nile Red (Fluka, Sigma-Aldrich, Buchs Switzerland) were used. Stock solutions of the dyes were prepared in 99.9% ethanol (Biosolve B.V., Valkenswaard, The Netherlands). The dye content was determined by UV spectroscopy after dilution with deionized water (ANS, Bis-ANS, and CCVJ), ethanol (DCVJ), or DMSO (Nile Red) to an absorbance between 0.1 and 1.0 at the absorption maximum of the respective dye.

Commercial products Humira®40 mg (Expiry date 05-2010, lot nr. 66781VA), Enbrel®50 mg (Expiry date 03-2010, lot nr. 35561) and MabThera®100 mg (Expiry date 08-2011, lot nr. H0012) were obtained from local hospitals. Enbrel®50 mg contains 50 mg/ml etanercept and is polysorbate-free, MabThera®100 mg contains 10 mg/ml rituximab and 0.07% (w/v) polysorbate 80 and Humira®40 mg contains 50 mg/ml adalimumab and 0.1% (w/v) polysorbate 80.

Formulations with 1.0 mg/ml IgG in 100 mM phosphate, pH 7.2 without polysorbate and with different concentrations of polysorbate 20 and 80 (see “Results” section) were prepared. To induce aggregation, 1.5 ml of the formulation were stressed for 10 minutes at 75°C or 80°C in 1.5-ml reaction tubes (Eppendorf, Hamburg, Germany) using a thermomixer (Eppendorf, Hamburg, Germany). Both temperatures were below the aggregation temperature of the IgG, and no precipitation was observed after the thermal treatment.

Enbrel®50 mg was stressed for 10 min at 70°C, Humira®40 mg for 10 min at 60°C and 65°C, and MabThera®100 mg for 10 min at 60°C, 65°C and 70°C. For each condition, 800 μl of the commercial product was filled in 1.5-ml reaction tubes (Eppendorf, Hamburg, Germany) and incubated at above-mentioned conditions using a thermomixer (Eppendorf, Hamburg, Germany). All formulations were optically clear after the thermal treatment, and no precipitation was observed. The formulations were allowed to cool down to room temperature and subsequently analyzed by HP-SEC and steady-state fluorescence spectroscopy using CCVJ.

The corresponding placebos contained 10 mg/ml sucrose, 5.8 mg/ml NaCl, 5.3 mg/ml arginine*HCl and 3.9 mg/ml Na₂HPO₄*2H₂0 (pH 6.3) for Enbrel®50 mg (26), 6.2 mg/ml NaCl, 1.52 mg/ml Na₂HPO₄*2H₂0, 0.3 mg/ml sodium citrate, 1.3 mg/ml citric acid monohydrate, 12.0 mg/ml mannitol and 1.0 mg/ml polysorbate 80 (pH 5.2) for Humira®40 mg (27) and 9 mg/ml NaCl, 7.3 mg/ml sodium citrate dehydrate and 0.7 mg/ml polysorbate 80 (pH 6.5) for MabThera®100 mg (28).

An Agilent 8453 UV–Vis spectrometer (Agilent, Waldbronn, Germany) was used to determine the IgG content in the formulations using an extinction coefficient of 1.49 for a 1.0 mg/ml solution for the absorption at 280 nm. Samples of 1.0 ml were measured in half-micro quartz cuvettes (Hellma, Kruibeke, Belgium) with a path length of 10 mm. For IgG formulations, the UV absorbance was recorded from 240 nm to 360 nm using an integration time of 15 s and steps of 1 nm.

To determine the concentration of the dye stock solution, the following molar extinction coefficients were used: ANS (4,900 M⁻¹ cm⁻¹ at 350 nm in water), Bis-ANS (16,790 M⁻¹ cm⁻¹ at 385 nm in water) (29), Nile Red (19,600 M⁻¹ cm⁻¹ at 552 nm in DMSO) (30), DCVJ (65,900 M⁻¹ cm⁻¹ at 453 nm in ethanol) (23) and CCVJ (25,404 M⁻¹ cm⁻¹ at 440 nm in water) (25). The spectra were recorded from 240 to 600 nm using an integration time of 5 s and steps of 1 nm. All spectra were corrected for the absorbance of the particular solvent.

A Tecan Infinite M1000 platereader (Tecan Benelux BVBA, Giessen, The Netherlands) was used to record the emission spectra of the extrinsic fluorescent dyes by top reading in black polypropylene 96-well plates (Greiner Bio-One B.V., Alphen a/d Rijn, The Netherlands). 5 µl of a 100 µM dye stock solution was added to 95 µl protein solution to achieve a dye concentration of 5 µM, which is suitable for all the selected fluorescent dyes. 100 µl of the IgG/dye mixture per well (n = 3) were measured within 30 min after dye addition. Preliminary studies had shown that there were no significant changes in fluorescence intensity and emission maximum for about 2 h.

ANS was excited at 350 nm (emission scanned from: 370 to 650 nm), Bis-ANS at 385 nm (emission: 400 to 650 nm), CCVJ at 435 nm (emission: 450 to 650 nm, DCVJ at 452 nm (emission: 470 to 650 nm) and Nile Red at 550 nm (emission: 570 to 700 nm). The fluorescence measurements were performed with slits of 5 nm, steps of 2 nm and flashes of 4.8 Joule. Under these experimental settings, no inner filter effect and no disturbance by scattered excitation light was observed. By subtracting the background spectra of the respective dye-free controls from the emission spectra of the dye-containing samples, the contribution of light scattering at the solvent manifested as Raman peak was removed from the spectra.

Time-resolved fluorescence measurements were performed on a LifeSpec-ps fluorimeter (Edinburgh Instruments Ltd., Livingston, UK) with a PMT detector module. As excitation source, PDL 800-B picosecond pulsed diode lasers (Picoquant, Berling, Germany) operating at a frequency of 10 MHz of 375 nm (Bis-ANS) and 450 nm (DCVJ) were used. On the emission side, a cut-on filter of 495 nm was used for both dyes. The temperature was kept constant at 25°C. Samples (900 µl) were measured in half-micro quartz fluorescence cuvettes (Hellma, Kruibeke, Belgium). The fluorescence decays were measured over 100 ns for Bis-ANS and 50 ns for DCVJ up to a peak count of 10,000 using 2,048 channels. The dynamic instrumental response function (IRF) was recorded by measuring the ‘decay’ of a diluted LUDOX^TM LS colloidal silica solution in water. To deconvolute the IRF and to fit the intensity decay, the FAST software package (Edinburgh Instruments Ltd., Livingston, UK) was used. Calculations therein are based on a combination of global least square analysis minimization and non-negative singular value decomposition. Depending on the resulting decays, between 1 and 4 exponentials (i.e., discrete lifetimes) were required for a suitable fit of the data. The goodness of fit was evaluated by the reduced Chi² criterion and the residuals of the fit. The average lifetime was calculated as: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ average\;lifetime = \sum\limits_{n = 1}^n {ta{u_n}*\frac{{{f_n}}}{{100}}} . $$\end{document}where tau is a discrete lifetime component, n is the number of lifetime components, and f_n is the fractional contribution (in %) of lifetime n.

Size exclusion chromatography was performed using a TSKgel4000SWXL column (Tosoh Biosep, Stuttgart, Germany) on an isocratic HPLC system with a Waters 515 pump, a Waters 717 plus autosampler, a Waters 474 fluorescence detector (Waters, Milford Massachusetts, USA) and a Shimadzu SPD UV/Vis detector (Shimadzu, Tokyo, Japan) at a flow rate of 0.5 ml/min. For the standard IgG formulations of 1.0 mg/ml 50 μl sample was injected. For the commercial products, 5 μl were injected for Enbrel®50 mg and Humira®40 mg and 10 μl for MabThera®100 mg.

The mobile phase was composed of 50 mM sodium phosphate, 150 mM arginine and 0.025% NaN₃, pH 7.0. To quantify aggregation in the IgG formulations, UV absorption at 280 nm was used. For fluorescence dye detection 5 µM CCVJ or 2.5 µM Bis-ANS were added to the mobile phase. CCVJ fluorescence was detected using excitation at 435 nm (bandwidth 18 nm) and monitoring emission at 500 nm (bandwidth 18 nm). For Bis-ANS excitation was at 385 nm (bandwidth 18 nm) and emission at 485 nm (bandwidth 40 nm).

In order to identify suitable dyes for the analysis of polysorbate-containing formulations, a number of dyes were compared for their sensitivity to polarity and viscosity. To do so, the fluorescence emission of the dyes was measured in 100 mM phosphate pH 7.2 (dielectric constant ε = 80.1, viscosity at 20°C η  = 1 mPa s), 10% (w/v) sucrose (ε = 80.1, η  = 1.2 mPa s), ethanol (ε = 25, η  = 1.2 mPa s), methanol (ε = 33.6, η  = 0.59 mPa s) and glycerol (ε = 42.5, η  = 945 mPa s) (Fig. 1). A considerable intensity increase combined with a blue shift of at most 32 nm was measured for Bis-ANS in ethanol and methanol. In glycerol, Bis-ANS fluorescence intensity did not increase as much, and the blue shift was only 10 nm. DCVJ and CCVJ exhibited significantly higher fluorescence intensities in glycerol, but unchanged (CCVJ) or slightly increased (DCVJ) intensities in ethanol and methanol. The blue shift in organic solvents was higher than the one in glycerol. Fig. 1

Representative steady state fluorescence spectra of 5 µM DCVJ (A), 5 μM CCVJ (B) and 5 μM Bis-ANS (C) in 100 mM phosphate pH 7.2 (buffer), 10% (w/v) sucrose, methanol, ethanol and glycerol.

A formulation with 1.0 mg/ml IgG in 100 mM phosphate pH 7.2 (non-stressed, NS, and stressed for 10 min at 80°C, HT), as well as the corresponding formulation placebo, were spiked with polysorbate 20 to final concentrations between 0 and 0.1% (w/v). Aggregates created by the thermal stress interact with non-covalent extrinsic dyes, resulting in an increase in fluorescence intensity, as described in detail in our previous publications (15,17). The rationale behind spiking with polysorbate after the application of thermal stress was to obtain comparable levels and types of aggregates in the formulations but varying the polysorbate 20 concentration. Analysis of the spiked IgG preparations confirmed that the addition of polysorbate 20 did not alter the properties of the samples with respect to Z-average and polydispersity index in dynamic light scattering, as well as aggregate content and relative recovery in HP-SEC in the time frame of the experiment (data not shown).

The polysorbate 20-spiked samples were analyzed by steady-state fluorescence spectroscopy after the addition of 5 μM DCVJ, CCVJ, Bis-ANS or Nile Red (Fig. 3). Comparable spectra with respect to emission maximum and intensity were obtained for the dyes when added to placebo and the corresponding non-stressed IgG preparation, pointing at negligible interactions of the dyes with non-stressed IgG. Heat-stressed placebos showed similar dye fluorescence as non-stressed placebos (results not shown). Fig. 3

Average maximum fluorescence intensity of 5 µM DCVJ (A), 5 µM CCVJ (B), 5 µM Bis-ANS (C) and 5 µM Nile Red (D) for placebo, non-stressed IgG (NS) and 10 min 80°C heat-stressed IgG (HT) spiked with 0 to 0.1% polysorbate 20 (PS20). Error bars represent standard deviations of 3 independent experiments.

To gain deeper insight into the underlying photo-physical behavior of the dyes upon interaction with protein and polysorbate, fluorescence lifetime measurements were performed for DCVJ and Bis-ANS. The average lifetimes are shown Table I, whereas the fluorescence decays, as well as detailed results and description of the data fitting are found in the “Supplementary Material.” As shown in Table I, the average lifetimes of DCVJ were short, which is in agreement with the literature (16). Slightly shorter average lifetimes were found for the polysorbate 20-containing formulations, as compared to those without polysorbate, which confirms the observation of lower fluorescence intensity at higher polysorbate 20 concentrations in steady-state fluorescence (Fig. 3B). Table II

HP-SEC Results Calculated from UV Absorbance at 280 nm for the IgG Formulations of 1.0 mg/ml IgG in 100 mM Phosphate pH 7.2 Without Polysorbate (PS) or With 0.02% Polysorbate 20 (PS20) or 80 (PS80)^a

The fluorescence decays of Bis-ANS were clearly more influenced by polysorbate 20. Longer average lifetimes of about 5.6 ns were determined for Bis-ANS added to polysorbate 20-containing placebo and non-stressed IgG, compared to Bis-ANS added to polysorbate-free placebo (0.49 ns) or non-stressed IgG (1.49 ns). For the interaction of Bis-ANS with polysorbate, a characteristic lifetime of about 5.5 ns could be identified, which was also dominating the decays of the polysorbate 20-containing IgG formulations. For Bis-ANS interacting with heat-stressed IgG, two characteristic lifetimes, one at about 6 ns and a longer component >10 ns, were measured (see “Supplementary Material”). Whereas the lifetime of 6 ns found for heat-stressed polysorbate-containing IgG formulations can be attributed to a combination of Bis-ANS interacting with polysorbate and with stressed IgG, the longer lifetime of >10 ns is specifically related to the interaction of Bis-ANS with heat-stressed IgG.

Time-resolved fluorescence confirmed the results from steady-state fluorescence and was particular useful to distinguish between Bis-ANS interacting with polysorbate and with heat-stressed IgG.

From the spiking experiments, it was obvious that DCVJ and CCVJ are capable of detecting aggregates in the presence of polysorbates. To be more in step with actual formulation work practice, we evaluated the advantage of DCVJ and CCVJ over polarity-sensitive dyes ANS, Bis-ANS and Nile Red for the characterization of formulations that contain polysorbate already during the stress testing. For this experiment, 1.0 mg/ml IgG was formulated in 100 mM phosphate pH 7.2 without and with 0.02% polysorbate 20 or 80 and heat-stressed for 10 min at 75°C and 80°C. Fig. 4 shows the fluorescence emission spectra for DCVJ, CCVJ and Bis-ANS added at a concentration of 5 µM to IgG formulations without and with 0.02% polysorbate 20. For the polysorbate-free formulations, an increase in fluorescence intensity combined with a blue shift of less than 20 nm was measured for DCVJ and CCVJ in heat-stressed formulations (Fig. 4A,C). Bis-ANS appeared to be the most sensitive dye to detect heat-induced changes in polysorbate-free formulations, indicated by the stronger intensity increase (9.8-fold versus 3.8-fold for DCVJ and 1.3-fold for CCVJ, respectively) and blue shift of about 15 nm for the formulation stressed for 10 minutes at 75°C (Fig. 4E), as compared to the molecular rotors (Fig. 4A,C). Fig. 4

Steady state fluorescence spectra (average of n = 3) of fluorescent dyes with placebo, non-stressed IgG (NS), IgG heat-stressed for 10 min at 75°C and 10 min at 80°C in the absence (A,C,E) and presence of 0.02% polysorbate 20 (PS20) (B,D,F). 5 µM dye was used: DCVJ (A,B), CCVJ (C,D) and Bis-ANS (E,F). The relative standard deviations of the maximum intensity were <1% for placebo, <3% for non-stressed IgG and 10 min 75°C stressed IgG and <5% for 10 min 80°C stressed IgG.

Non-stressed, heat-stressed IgG (10 min 75°C and 80°C) and placebo without and with 0.02% polysorbate 20 or 80 were analyzed by HP-SEC using UV detection at 280 nm (Fig. 5) and fluorescent dye detection, using either 2.5 μM Bis-ANS (excitation 385 nm, bandwidth 18 nm, emission 490 nm, bandwidth 40 nm) or 5 μM CCVJ (excitation 435 nm, emission 500 nm, bandwidths 18 nm) in the mobile phase (Fig. 6). From the molecular rotors, CCVJ was chosen because the solubility in water is better than the one of DCVJ, because one of the nitrile functions is replaced by a carboxylic acid group (24) (Fig. 2). The higher sensitivity of Bis-ANS compared to CCVJ detection is in part due to the relatively higher fluorescence intensity of Bis-ANS (compare also Fig. 3 and 4) and in part due to the experimental settings. The larger Stokes’ shift of the emission maximum allows for working with a larger monochromator bandwidth for Bis-ANS. To avoid interference from scattered light during CCVJ detection, a narrower bandwidth of 18 nm had to be chosen. Moreover, the sensitivity of the PMT detector is lower at 500 nm than around 490 nm. Fig. 5

Representative HP-SEC chromatograms of UV detection at 280 nm of placebo, non-stressed (NS) IgG and heat-stressed IgG (10 min 75°C or 10 min 80°C) for polysorbate-free formulations (A), formulations with 0.02% polysorbate 20 (PS20) (B) and 0.02% polysorbate 80 (PS80) (C).

After having shown that molecular rotors are suitable to detect aggregation in polysorbate-containing IgG formulations, we aimed to use CCVJ for the analysis of commercial protein therapeutics. Possible concerns about using CCVJ for these preparations may be the higher protein concentrations of 10 to 50 mg/ml and the more complex compositions with respect to excipients. For the study, we selected Enbrel®50 mg, which is polysorbate-free but contains a high concentration of 50 mg/ml etanercept, and the polysorbate 80-containing monoclonal antibody preparations MabThera®100 mg (10 mg/ml rituximab and 0.07% (w/v) polysorbate 80), as well as Humira®40 mg (50 mg/ml adalimumab and 0.1% (w/v) polysorbate 80). The products were stressed for 10 minutes at different temperatures below the aggregation temperature (determined by UV spectroscopy—data not shown) to obtain aggregate-containing, but optically clear, solutions. HP-SEC confirmed the formation of aggregates by the thermal stressing of the commercial preparation as shown in Fig. 7. The absolute quantity of aggregates obvious from the UV detection at 280 nm in these high concentrated formulations might be impacted by the dilution effect observed during the HP-SEC analysis (38), but this was not further studied. Fig. 7

Representative HP-SEC with UV detection at 280 nm (A,C,E) and CCVJ dye detection, Exc. 435 nm, Em. 500 nm (B,D,F) of Enbrel®50 mg (50 mg/ml etanercept) (A,B), MabThera®100 mg (10 mg/ml rituximab) (C,D) and Humira®40 mg (50 mg/ml adalimumab) (E,F). Each commercial preparation was analyzed non-stressed (NS) and after heat stress for 10 minutes at the temperatures indicated in the figure legend. Insets represent a zoom into the chromatograms of the UV detection at 280 nm.