OR WAIT 15 SECS
FTIR can successfully measure key characteristics of therapeutic proteins in a single step.
Rapid commercialization of therapeutic proteins remains a challenge notably due to their chemical and physical instabilities. Yet, the inherent complexity of proteins requires the development of new analytical strategies to characterize and ensure the quality and safety of these products.
Harnessing the strengths of Fourier transform infrared spectroscopy (FTIR) and recent improvements in chemometrics, new analytical methods have been developed to study the stability and perform comparability studies of therapeutic proteins. In this article, the authors demonstrate the feasibility of simultaneously obtaining information regarding four key characteristics of therapeutic proteins: structural integrity, overall protein concentration, quantification of glycosylations, and quantification of phosphorylations. Depending on specific situations, gathering these data would typically require leveraging three to four distinct techniques and protocols. Moreover, infrared spectroscopy allows researchers to analyze these parameters in a precise, quick, and direct manner (i.e., giving results in a couple of minutes), with limited sample volume (<50µL) and without the need for extensive sample pre-treatment or the requirement to recalibrate on each measurement day.
The aim of this work was to develop an innovative methodology to analyze the stability of therapeutic proteins based on FTIR spectroscopy. With this technique, the attenuation of an infrared light beam is measured when it passes through a sample. This attenuation is due to the interaction between the light and vibrational transitions in the covalent bonds of the molecules present in the sample. FTIR spectroscopy has already been widely used to analyze protein structure and was shown to be very sensitive to tiny structural changes (1–6). This technique confers a series of other advantages, which are described in subsequent paragraphs.
Figure 1: Harnessing the strengths of Fourier transform infrared spectroscopy, one single measurement-which takes only a few minutes-allows researchers to obtain information concerning four key attributes of therapeutic proteins: (i) structural integrity, (ii) overall protein concentration, (iii) quantification of glycosylations, and (iv) quantification of phosphorylations. The procedure does not require any labeling or separation.As illustrated in Figure 1, the authors used FTIR to investigate four key characteristics of therapeutic proteins: structural integrity (the absence of denaturation and aggregation), overall protein concentration, quantification of glycosylations, and quantification of phosphorylations. All these parameters are a major concern because they may significantly affect both biological activity and immunogenicity. Aggregation is a common source of protein instability; this phenomenon and its triggers are not fully understood. Moreover, misfolding or alterations in the three-dimensional structure of proteins can be responsible for loss of activity and elicit immune responses (7–10).
Post-translational modifications such as glycosylations and phosphorylations are also important issues. Phosphorylation is the reversible addition of a phosphate group on a polypeptide chain. It is often involved in signalling pathways and generally influences structural properties, dynamic, and binding (11). Glycosylation is defined as the attachment of polysaccharides at specific sites of the amino acid chain in proteins. This modification is involved in product solubility; stability; half-life; pharmacokinetics and pharmacodynamics (PK/PD); and bioactivity and safety (e.g., immunogenicity). In the context of protein production, protein glycosylation is subject to a high degree of heterogeneity, depending on the manufacturing conditions (cell lines, culture parameters, protein purification, etc.). It is, therefore, important to evaluate the proportion of sugar that is present in a commercial protein product (12–15).
As these four characteristics significantly affect the efficacy and safety of protein products, they all must be carefully and systematically monitored. Regulatory agencies worldwide require robust, information-rich, reproducible, and preferably orthogonal methods to analyze the aforementioned parameters and ensure accuracy and consistency of the final drug product (16). Therefore, a single analysis for these four parameters with a short measurement time and a low sample volume is of great interest.
Albumin from human serum (A3782, Sigma-Aldrich, Bornem, Belgium) was used to demonstrate the possibility to assess the four parameters by FTIR spectroscopy. The protein was purchased powdered and used without further purification.
Five peptides ordered from Eurogentec (AS-61329, AS-61334, AS-61332, AS-20292 and AS-24537, Liège, Belgium) were also used in this work. The peptide AS-61329 is derived from the human mucin MUC5AC gene sequence. AS-61334 and AS 61332 are glycopeptides with the same sequence as AS-61329, with respectively one site (Thr3) or two sites (Thr3 and Thr13) labeled with a N-Acetyl galactosamine. AS-20292 and AS-24537 are derived from the insulin receptor tyrosine kinase. AS-24537 is the unphosphorylated form of the peptide and AS-20292 has one phosphorylation.
All samples were purchased powdered and diluted in purified water at a concentration of 5 mg/mL. Powders were conserved at 4 °C and solutions at -20 °C.
All measurements were carried out on a Bruker Tensor 27 FTIR spectrometer (Bruker, Karlsruhe, Germany) equipped with a liquid N2-refrigerated mercury cadmium telluride detector. All spectra were recorded by attenuated total reflection (ATR)-for a review, see (17). A diamond internal reflection element was used on a Golden Gate Micro-ATR from Specac (Orpington, UK). The angle of incidence was 45 degrees. A 0.5-µL amount of the proteins was deposited on the diamond crystal. The sample was quickly evaporated in N2 flux to obtain a homogenous film of proteins. The FTIR measurements were recorded between 4000 and 600 cm-1. Each spectrum was obtained by averaging 128 scans recorded at a resolution of 2 cm-1.
All the spectra were pre-processed as follows. The water vapor contribution was subtracted as described previously (18, 19) with 1956–1935 cm-1 as reference peak. The spectra were also baseline-corrected. Straight lines were interpolated between frequencies corresponding to local minimum and were then subtracted from the spectrum. Normalization for equal area was finally applied between 1724 and 1480 cm-1. In the case of the overall protein concentration, the normalization for equal area was applied between 2080 and 2000 cm-1.
The results presented describe the development and the validation of the methodology for each parameter separately. However, it should be noted that all these data are extracted from the same spectrum.
Albumin at 5 mg/mL was exposed to 16 successive temperatures from room temperature to 90 °C. After the heating and before the measurements, all samples were cooled at room temperature. For each temperature, between three and six independent samples were prepared. For each sample, between three and five FTIR spectra were recorded.
Figure 2: (A): ATR–FTIR spectra of progressively heated albumin spectra are gradually colored from blue to red according to increasing temperature. Blue corresponds to the lowest temperature and red to the highest. Spectra were processed as described in the methodology. (B) Stability index of albumin obtained by FTIR spectroscopy according to temperature. Each point corresponds to the mean of all the spectra recorded at one temperature. Error bars indicate the standard deviation. ATR–FTIR=attenuated total reflectance–Fourier transform infrared spectroscopy.Figure 2A displays all the pre-processed spectra recorded for albumin. Proteins were progressively heated from room temperature (blue color) to 90 °C (red color). Significant variations can be observed with the naked eye, especially for the highest temperature (red color). In the present case, the absorbance at the maximum of this band (1622 cm-1) was monitored. The mean of this absorbance and the standard deviation is calculated for each temperature and presented in Figure 2B. A typical denaturation curve is observed on this figure and underlines the possibility to use attenuated total reflection (ATR)–FTIR spectroscopy as a relevant tool to monitor the structural stability of protein.
It must be noted that FTIR spectroscopy is an interesting orthogonal technique for assessing aggregation, which is almost always associated with a structural change. Whereas size-exclusion chromatography characterizes aggregates according to their size, FTIR spectroscopy provides structural information characteristic of intermolecular interactions. Moreover, this technique is especially attractive in the case of high protein concentration, as no dilution is required to perform the analysis.
With ATR–FTIR spectroscopy, determining absolute concentrations remains a challenge because the film thickness is not known accurately. To measure the quantity of protein in a sample, the authors have added a molecule that serves as an internal reference. An internal reference should be soluble, not volatile, and it should absorb in a spectral range different from that of biological molecules (1800-900 cm-1). This molecule is added at the same concentration in all samples. The normalization step is then realized on a specific absorption peak of this reference. The molecule used as an internal reference is ferrocyanide.
The listed methodology was followed:
Validation of the model to predict protein concentration using Fourier transform infrared spectra of the samples. Predicted values are represented as a function of the true values. Each point corresponds to the mean of all the spectra recorded at one concentration. The error bars indicate the standard deviation.
presents the results of the validation of the model. It reports the predicted protein concentration as a function of the true protein concentration. Each point corresponds to the mean of all the spectra recorded at one concentration. The error bars indicate the standard deviation associated with this concentration. With a correlation coefficient of 0.97, the predicted values are highly correlated to the true values underlining the performance of this model. This correlation can be up to 0.99 using other internal reference (data not shown). The results demonstrate that ATR–FTIR spectroscopy can successfully be used to measure protein concentration.
It must be underlined that for most of the techniques currently used to assess protein concentration (UV absorption spectroscopy, Bradford or Bicinchoninic assay, etc.), the measurement depends on the presence of specific amino acids. For example, in the case of UV absorption at 280 nm, the measure relies upon the presence of aromatic amino acids. Conversely, FTIR spectroscopy relies on the presence of the amide bond, which is always present in a protein.
The goal was to develop a method to quickly predict the carbohydrate content, which is defined as the ratio between the mass of carbohydrate and the total mass of the sample (carbohydrate and polypeptide chain).
Glycosylated peptides were mixed to albumin to obtain a range of carbohydrate content from 0.43% to 7.03% (w/w). Ferrocyanide was added in all samples as an internal reference at a concentration of 0.1mg/mL to use the model built previously to predict protein concentration. It must also be noted that all samples were prepared six times on a different day with new albumin solution and new ferrocyanide solution. The FTIR spectra of the first three preparations were used to build the quantification models based on PLS regression for glycosylations. The FTIR spectra of the other three were used to validate these models. For the validation samples, new vials of peptides were ordered and new solutions of peptides were prepared.
The calibration (and validation) samples contain both glycosylated and phosphorylated peptides. As carbohydrate and phosphate absorb in the same zone of FTIR spectra, the absorption bands associated with these post-translational modifications overlap. Due to this overlapping, the prediction of the carbohydrate content in the presence of phosphate was challenging and unexpected (20).
Figure 4: Validation of the model to predict (A) the carbohydrate content (the mass percentage of carbohydrate) and (B) the phosphate content (the mass percentage of phosphate) using Fourier transform infrared spectra of the samples. Predicted values are represented as a function of the true values. Each point corresponds to the mean of all the spectra recorded at one concentration. The error bars indicate the standard deviation.Figure 4A presents the results of the validation of these models and reports the predicted carbohydrate content as a function of the true carbohydrate content. Each point corresponds to the mean of all the spectra recorded for one carbohydrate content. The error bars indicate the standard deviation associated with this ratio. The correlation coefficient is larger than 0.99; the predicted values are thus highly correlated with the true values, underlining the performance of these models.
Most of the current procedures to quantify carbohydrate in a protein sample are based on mass spectroscopy (MS) or fluorescence detection, coupled with chromatography techniques. These procedures generally involve several preliminary steps of sample preparation such as cleavage from the protein backbone, labeling, etc. Moreover, to obtain quantitative information, internal standards such as stable isotopes should be used (12, 14).
The goal was to develop a method to quickly predict the ratio between the mass of phosphate and the total mass of the sample (carbohydrate and polypeptide chain). It is called below the phosphate content.
The same samples prepared to calibrate and validate the model to quantify glycosylation were used. The range of the phosphate content obtained was from 0.1% to 2.63% (w/w). Figure 4B reports the predicted phosphate content as a function of the true phosphate content. The predicted values are highly correlated with the true values (correlation coefficient superior to 0.99), underlining the performance of these models. The results underline the feasibility of quantifying simultaneously glycosylation and phosphorylation in protein samples containing both phosphate and carbohydrate. It was particularly challenging to obtain such measurements, as carbohydrate and phosphate absorb in the same zone of FTIR spectra (20).
Most of the current procedures to quantify phosphate in a protein sample are also based on mass spectroscopy and generally involve several preliminary steps of sample preparation (11).
For all the four parameters, the information obtained is extracted from the same FTIR spectrum. The methodologies described previously can be combined in a data interpretation automated process. Moreover, only a few minutes are required to record the spectra. The full process (measurement and data interpretation) is thus very quick.
Furthermore, two other practical advantages of FTIR spectroscopy need to be underlined: There is no (or there is limited) sample preparation (no dilution, no probe, no label) required; and very few materials are required to perform the analysis (maximum 50 microliters per condition).
Another advantage of FTIR should be mentioned-FTIR spectra analyses of proteins, as a global fingerprint of all the molecules present, also allow the detection of a series of unexpected deviations or problems in the production and purification process.
In conclusion, considering the benefits of FTIR spectroscopy and its high sensitivity to tiny changes in a sample, a multi-attribute technique is an attractive and innovative tool for stability testing (including stress studies such as heating, shaking, pH, freeze-thaw cycles, etc.), long-term stability studies, and the comparison of formulations.
1. E. Goormaghtigh, V. Cabiaux, and J. Ruysschaert, Subcell Biochem., 23363–23403 (1994).
2. E. Goormaghtigh et al., Biochim. Biophys. Acta 1794 (9), 1332–1343 (2009).
3. A. Barth, Biochim. Biophys. Acta 1767 (9), 1073–1101 (2007).
4. A. Barth and C. Zscherp, Q. Rev. Biophys. 35 (4), 369–430 (2002).
5. K. Oberg, J. Ruysschaert, and E. Goormaghtigh, Eur. J. Biochem. 271 (14), 2937–2948 (2004).
6. V. Grimard et al., J. Biol. Chem. 279 (7), 5528–5536 (2004).
7. T. Arakawa et al., BioProcess Intl. 4 (10), 32–42 (2006).
8. T. Arakawa et al., BioProcess Intl. 5 (4), 36–47 (2007).
9. T. Arakawa et al., BioProcess Intl. 5 (10), 52–70 (2007).
10. H. Mahler et al., J. Pharm. Sci. 98 (9), 2909–2934 (2009).
11. H. Kuyama et al., Rapid Commun. Mass Spectrom. 17 (13), 1493–1496 (2003).
12. A. Planinc et al., Anal. Chim. Acta 921, 13–27 (2016).
13. R. Jefferis, Trends Pharmacol. Sci. 30 (7), 356–362 (2009).
14. L. Zhang, S. Luo, and B. Zhang, MAbs 8 (2), 205–215 (2016).
15. P. Hossle, S. Khattak, and Z. Li, Glycobiology 19 (9), 936–949 (2009).
16. L. Hajba, E. Csanky, and A. Guttman, Anal. Chim. Acta 943, 8–16 (2016).
17. E. Goormaghtigh, V. Raussens, and J. Ruysschaert, Biochim. Biophys. Acta 1422 (2), 105–185 (1999).
18. E. Goormaghtigh, “FTIR Data Processing and Analysis Tools,” in Biological and Biomedical Infrared Spectroscopy, A. Barth, P.I. Haris, Eds (IOS Press ed, 2009), pp. 104–128.
19. E. Goormaghtigh and J. Ruysschaert, Spectrochim. Acta 50 (12), 2137–2144 (1994).
20. E. Goormaghtigh and A. Derenne, “Method for determining the degree of phosphorylation and the degree of glycosylation of a protein in a protein sample,” patent filed in March 2015, (Ref. : PCT/EP2016/054619).
About the Author
Allison Derenne, PhD, is spin-off developer at Spectralys Biotech, Université Libre de Bruxelles. Erik Goormaghtigh, PhD, is research director at the National Fund for Scientific Research, Belgium, and professor at the Université Libre de Bruxelles.
Vol. 30, No. 4
When referring to this article, please cite it as A. Derenne and E. Goormaghtigh, "FTIR Spectroscopy as a Multi-Parameter Analytical Tool,"