Method for the comparison of complex matrix assisted laser desorption ionization-time of flight mass spectra. Stability of therapeutical monoclonal antibodies

Highlights

• Straightforward andeasy-to-implement method for comparing mass spectra by multivariate chemometrictechniques.

• A home-made MATLABfunction called Protiago has been developed.

• A new similarityindex named ’nearness index’ which assess the spatial proximity of two vectorswas proposed.

• The methodology wasapplied to the mass spectra of two therapeutical mAbs to state their chemicalstability over time.

Abstract

This paper describes a straightforward and easy-to-implement method for comparing mass spectra using multivariate chemometric techniques in order to detect differences and obtain representative similarity metrics. For this purpose, we have programmed own MATLAB function called Protiago. Protiago successfully transforms complex vectors (i.e. intensities and m/z values from mass spectra) with different lengths in binary vectors with the same number of elements. In addition, Protiago is able to read and properly process a set of spectra vectors in order to carry out a proper similarity analysis using four similarity metrics (i.e. the coefficient of determination, the cosine of the angle, the Bray-Curtis index and the nearness index). The latter is a new similarity index proposed by the authors and applied for the first time in this study. It calculates a standardized Euclidean distance between two vectors in the space in order to obtain a numerical value, ranged between 0 and 1, of the proximity of both vectors. To supplement the similarity analysis information, two multivariate exploratory methods were applied, i.e. principal component analysis (PCA) and multivariate analysis of variance (MANOVA). As an example of the proposed method, the comparison of peptide mass fingerprints obtained using MALDI-TOF mass spectrometry from two therapeutical monoclonal antibodies, infliximab (INF) and rituximab (RTX), was carried out. By using this method it was possible detect changes in the primary structure of the two proteins in order to study their chemical stability for 7 days under two storage conditions (refrigerated at 4 °C, and frozen at −20 °C).

Introduction

The use of therapeutic monoclonal antibodies (mAbs) in clinical practice is well established nowadays. They are widely used in hospitals for the treatment of highly prevalent diseases such as cancer and auto-immune diseases. These biotechnological drugs have importantly contributed to the pharmaceutical world in the recent decades and promise to have more relevance in the future, therefore their study and characterization represent an important task. [1].

mAbs are large glycoproteins (≈150 kDa) composed of four peptide chains, two light (L) and two heavy (H) ones, stabilized by inter-chain disulphide bonds. Both H-chains and L-chains contain variable (V_L,V_C) and constant regions (C_L,C_C). The variable regions, which are the antigen-binding site of the antibody, display different specificities and differ in the amino acid sequence, whilst the constant regions do not. The H-chains have a flexible hinge region which contains several proline residues. In addition, mAbs have glycans linked to the H-chains via an N-linked glycosidic bond [2]. Fig. 1 shows the general structure of a mAbs.

RTX (empirical formula: C₆₄₁₆H₉₈₇₄N₁₆₈₈O₁₉₈₇S₄₄; average molecular mass: 143859.7 Da [3]) is a chimeric murine/human IgG1 mAb that binds to CD20, a transmembrane protein, located on pre-B and mature B-lymphocytes. It is intended for use in the treatment of non-Hodgkin's lymphoma, rheumatoid polyarthritis, chronic lymphoid leukemia and granulomatosis with polyangiitis and microscopic polyangiitis [4]. INF (empirical formula: C₆₄₂₈H₉₉₁₂N₁₆₉₄O₁₉₈₇S₄₆; average molecular mass: 144190.3 Da [5]) is also a chimeric murine/human IgG1 mAb, which binds to the tumour necrosis factor (TNF-alpha). INF is an anti-inflammatory medicine, and it is usually used when other medicines or treatments have failed in adults with the following diseases: rheumatoid arthritis, Crohn's disease, ulcerative colitis, ankylosing spondylitis, psoriatic arthritis and psoriasis [6].

Due to their protein nature, these biopharmaceutics are complex and unstable; slight changes in the environment may produce significant changes in their structure. For those that are approved, the stability is unknown after the expiry date specified by the manufacturer (in most cases, this does not exceed 24 h). This fact means that large amounts of these expensive drugs are daily discarded in hospitals with the subsequent economic loss. There is a need for additional stability data covering the practical uses of mAbs based, for example, on physicochemical characterizations over time.

Different experimental strategies could be applied in order to carry out a proper physicochemical characterization: (i) a top-down approach, which is referred to intact mAb analysis; (ii) a bottom-up approach, which is based on a prior enzymatic digestion of the mAb and a subsequent analysis of the digested peptides; and (iii) a middle approach, in which a partial digestion of the mAb is made [7], [8]. The physicochemical characterization is achieved using a wide range of analytical techniques, including reversed-phase liquid chromatography (RPLC), size-exclusion chromatography (SEC), ion-exchange chromatography (IEX), sodium-dodecyl sulphate polyacrylamide-gel electrophoresis (SDS-PAGE), capillary isoelectric focusing (CIEF), capillary zone electrophoresis (CZE), circular dichroism spectrometry (CD), Fourier-transform infrared spectroscopy (FT-IR), and fluorescence spectrophotometry (FL) [9]. In addition, mass spectrometry (MS) represents the most important analytical technique regarding mAb structural investigation [10]. Nevertheless, very few papers which focus on long-term stability studies of marketed therapeutic mAbs based on MS measures have been published [11].

The MS spectrum obtained from the resulting bottom-up approach digested fraction depends on the particular peptide sequence of each protein and it may be used for protein identification, characterization or comparison [7]. For this reason, such a spectrum is also labelled as a peptide mass fingerprint (PMF). Two mass spectrometric ionization methods are habitually applied in order to acquire reliable PMFs: electrospray ionization (ESI) and matrix-assisted laser desorption and ionization (MALDI). MALDI has some important advantages, such as its ability to generate singly charged ions and its relative robustness in the presence of salts and buffers [11], however it shows an intrinsic lack of reproducibility which should be kept under control.

In order to perform suitable stability studies, it is necessary to collect a spectra set at different times from the same protein sample and compare them to detect possible changes. However, the direct visual comparison is difficult due to both the complexity of PMF and the low MALDI spectra reproducibility; therefore the use of mathematical-statistical methods is required. Surprisingly, there are very few papers which propose specific tools in for carrying out such comparisons. For instance, Holton et al. use a log transformation, Pearson correlation and principal component analysis (PCA) for the comparison of mass spectrometric peptide profiles [12], Máté et al. apply a topological and geometric approach to measure protein similarity [13], Bazsó et al. [14] and Teska et al. [15] use different similarity indexes for the comparison of mass spectra, Fourier transform infrared spectra and near ultraviolet circular dichroism spectra. We have also recently developed a cluster-based comparison tool for PMF obtained from MALDI time of flight (TOF) mass spectrometry [11].

A valid and straightforward way for monitoring the possible changes of PMF is the use of similarity metrics, so the larger the value is, the more similar they are. There are many such metrics that can be applied. The similarity analysis is based on comparing one by one the figures from two data vectors with the same number of elements. This means that the acquired spectra have to be previously transformed into data vectors; nevertheless this is a routine option into the most of the current analytical instrument software. The data vectors are then assembled to build a matrix data containing all the spectra. In this way, the similarity between each pair of rows (samples) in the data matrix can be measured. The similarity metrics usually assess the distance or the degree of mathematical correlation between the elements from the data vectors which are being compared. Several multipurpose similarity metrics have been proposed for comparing vectors [16]; only for the particular case of binary vectors (all the elements are 0 or 1), it is possible to find more than 70 similarity metrics [17], [18]. However, the similarity indexes most frequently used for quantitative measurement data sets are the cosine, the correlation coefficient and indexes based on the distance; the values of most of them vary from 0 to 1.

In this paper, a robust mathematical comparison method is proposed to be applied in complex mass spectrometric data, particularly in PMF obtained by MALDI-TOF-MS. Each single PMF was extracted in a data vector which denotes intensities vs m/z. Next, the development of an ad hoc MATLAB function which firstly transforms the intensities of the PMF in binary data was carried out; the value “1” indicates the presence of a fragment to a certain m/z value, and the “0” the absence of it. In a second step, an averaged no-binary vector is obtained for each sample from the replicates of the previous PMF binary vectors. The similarity analysis was then applied on the averaged vectors by developing, a new multipurpose distance similarity metrics which has been called the 'nearness similarity index'. Subsequently, the new developed index was validated by comparing with other well-known similarity metrics. In addition, two multivariate exploratory methods were applied to the binary vectors, i.e. PCA and multivariate analysis of variance (MANOVA), in order to detect patterns in the PMFs. As an application example, the method has been applied on the long-term stability study of two marketed therapeutics, mAbs, widely used in hospital practice, i.e. rituximab (RTX) and infliximab (INF), in order to contribute robust data to evaluate re-utilization of the daily surplus.