1. Introduction
Monoclonal antibodies are protein molecules that are produced by recombinant DNA technology and used as treatment for a wide variety of severe diseases. The use of monoclonal antibodies in the treatment of cancer, chronic inflammatory diseases, autoimmune disorders, asthma, multiple sclerosis, osteoporosis, age-related macular degeneration and infectious diseases has been increasing lately [
1,
2,
3].
A humanized monoclonal antibody used in many cases for treating different types of cancer is bevacizumab, commercially known as Avastin
®. The action of bevacizumab is based on its binding to the Vascular Endothelial Growth Factor A (VEGF-A). This binding leads to the prevention of the interaction of VEGF-A with its receptor (VEGFR) and the inhibition of the VEGF signalling pathway. Hence, neovascularization and vessel growth are inhibited, and angiogenesis of tumour cells is reduced [
4,
5]. Thus, bevacizumab is considered an anti-angiogenic monoclonal antibody and is used for the treatment of types of cancer that are characterised by significant angiogenesis. Metastatic colorectal cancer, metastatic breast cancer, non-small-cell lung carcinoma, ovarian cancer, renal cell carcinoma and glioblastoma multiforme are some of the indications for the use of bevacizumab, usually in combination with chemotherapy [
6,
7].
As far as the dose of bevacizumab is concerned, individualized doses of the monoclonal antibody are prepared according to the prescription for each patient [
8]. Moreover, some adverse events of bevacizumab, such as hypertension and proteinuria, are found to be dose-dependent [
7,
9]. Thus, the quantitation of bevacizumab in the prepared solutions by the healthcare units is of high importance, so that the efficacy and the safety of the anti-angiogenic treatment can be guaranteed. High-performance liquid chromatography (HPLC) is mainly used for the identification and quantification of biological anticancer preparations because of the high specificity and sensitivity and low detection and quantitation limits offered by the method [
10,
11]. Regarding bevacizumab’s quantitation in Avastin
®, an HPLC analytical method coupled with mass spectrometry (MS) has been developed and validated [
12], while an LC-MS/MS analytical method was recently validated for the quantitation of the monoclonal antibody in human serum [
13]. Although the described benefits of the chromatographic techniques are important, they suffer from large time requirements for sample preparations and analysis, requiring specialized personnel, and large amounts of solvents are consumed during the analysis.
Another approach that is gaining ground for the identification and quantification of pharmaceuticals and biomedicals is Raman spectroscopy. Raman spectroscopy is associated with a change in the polarizability of molecules, which is caused by inelastic scattering of the initial monochromatic light with which the substance is irradiated [
14,
15]. It has been proposed as an alternative to HPLC for the qualitative and quantitative analysis of anticancer drugs and monoclonal antibodies, offering similar accuracy, precision and trueness to the respective analytical validation criteria measured using HPLC [
16]. Raman spectroscopy could also serve as a fast, reliable, reproducible, non-invasive, non-destructive, cost-effective and environmentally friendly technique, requiring only a few seconds or minutes for sample preparation, in contrast to HPLC [
17]. Thus, the applications of Raman spectroscopy in pharmaceutical and biomedical sciences are impressively increasing [
18,
19,
20,
21,
22].
A specific application of Raman spectroscopy regarding biopharmaceutical analysis is the identification and quantification of monoclonal antibodies. Only a few studies have been published recently for the characterization and discrimination of monoclonal antibodies using Raman spectroscopy [
23,
24,
25,
26,
27,
28], and the quantification of monoclonal antibodies through Raman spectroscopy has been performed only in very few cases [
17,
25,
29]. In the study of Rayyad et al., three monoclonal antibodies (bevacizumab, trastuzumab and atezolizumab) were quantified in a quartz cuvette and in perfusion bags using Raman spectroscopy [
17]. The results of the analytical method developed for the quantification of cetuximab, rituximab, trastuzumab and bevacizumab via Raman spectroscopy using multivariate analysis in another study were found to be significantly affected by the excipients’ concentrations, leading to reliability and reproducibility concerns [
25]. Finally, in the study of Le, et al. a machine learning approach for the quantitation of infliximab, bevacizumab, rituximab and ramucirumab in vials has been described and compared to a conventional linear regression approach [
29].
The aim of our study was to develop a novel simple, rapid and environmentally friendly analytical method for the quantitative analysis of bevacizumab monoclonal antibody in the final formulation using Raman spectroscopy. The method proposed in this work was also capable of providing automatization for the on-the-spot quantification of bevacizumab in hospitals in the concentration range that is used in the final formulations. The sample preparation method included the drying of a sample droplet on a highly reflective carrier. The acquisition of the Raman spectra was performed by applying rotation on the sample carrier with the dried droplet of the monoclonal antibody solution using a home-made apparatus. A conventional linear regression model was applied for the quantification of the monoclonal antibody in the formulation. It is the first time in the literature that a rotary apparatus has been employed for the quantitation of biopharmaceuticals via Raman spectroscopy, and dehydration of the monoclonal antibody formulation has not been used previously in such an application. A further objective of our study was the validation of the developed quantitative analytical method regarding its specificity, working range, accuracy, precision and sensitivity, as no other study has described a validated Raman spectroscopic method for the quantification of monoclonal antibodies. Finally, our study, also, aimed at the investigation of the optimal position of the dried droplet for Raman spectra acquisition, at which the lowest detection limits could be achieved.
3. Discussion
In the present study, a novel analytical method was developed for the quantitation of bevacizumab via Raman spectroscopy. This technique has been recently adopted as a quick, simple and environmentally friendly alternative to the well-established chromatographic techniques, such as high-performance liquid chromatography (HPLC) and electrochemical assays, such as ELISA [
41,
42,
43], for the quantitation of various monoclonal antibodies in their formulations [
17,
25,
29]. This approach involved the initial dehydration of a droplet of the solution on a highly reflective carrier.
The drying of the droplet on a surface is a complicated process, and the final morphology of the dried drop depends on many factors. During this process, the capillary flow causes the transfer of solutes from the main core of the droplet to its contact line, which refers to the edge of the droplet every moment it is drying. This phenomenon triggers the formation of a coffee ring [
44,
45,
46,
47]. The dehydration rate is the crucial parameter influencing the shape of the dried droplet, as fast drying leads to homogenous deposition of the solutes, while slower dehydrations possibly result in the formation of a coffee ring deposition. The high surface tension of the solvent, in combination with the surface hydrophobicity, leads to the formation of the coffee ring. Moreover, the formation of the coffee ring is the result of the competition of capillary flow and Marangoni flow [
48], which refers to the recirculation of the solutes from the edge to the main core of the droplet. A final parameter that affects the formation of the coffee ring is the viscosity of the solution, promoting the migration of solutes in the droplet. The most viscous solutions hinder capillary flow and promote Marangoni flow, suppressing the coffee-ring effect [
44].
The developed method involved the employment of a home-made apparatus for rotating the sample on the gold-coated carrier during Raman spectra acquisition. This method was selected in order to minimize any source of errors in the quantitative analysis due to subsampling when a Raman spectrum from only one point of a solid sample is acquired (point irradiation). It has been widely found through studies that the small size of the laser spot leads to small volumes of sampling and, thus, errors in quantification occur because of the small sampling area and unequal distribution of the sample [
36]. Rotation of the sample was proposed as a solution to the errors caused by the unequal deposition of the solute from a dried droplet, as signal from a large number of points (located on the circumference of a circle) is acquired simultaneously. Thus, any issues because of the inhomogeneity of the monoclonal antibody on the dried droplet were reduced.
Three different circular circumferences were tested for obtaining the Raman spectra; the first one was set near the edge of the dried drop, where a coffee ring was formed, the second one in the middle of the dried drop and the third one close to the centre. All three calibration curves possessed a satisfactory linear response in the working range of 3.75–25.00 mg/mL, while the ones corresponding to the middle or central circular circumferences provided R
2 values higher than 0.990 (
Figure 3). When the Raman spectra were acquired under rotation from the centre of the dried droplet, the highest coefficient of determination was found (R
2 = 0.999), while the standard errors of the slope and intercept had the lowest values (
Figure 3) and the DL was the lowest determined (1.06 mg/mL) (
Table 4). Low enough also were the DL (2.48 mg/mL) and the standard errors of the calibration curve determined for the middle circle of rotation. These satisfactory linear fits, ensuring low DLs and satisfactory selectivity, accuracy, precision, sensitivity and trueness in the working range of 3.75–25.00 mg/mL, implied an equal distribution of bevacizumab in the main core of the dried droplet, which was independent of the bevacizumab concentration and the exact position of the Raman spectra acquisition.
On the contrary, at the edge’s circular circumference, a lower coefficient of determination (R
2 = 0.96) was calculated, the standard errors of the slope and intercept were higher than the respective errors of the other two calibration curves and the DL (5.82 mg/mL) (
Table S1) was much higher than the DL estimated visually from the respective Raman spectra. Because of the rotation of the sample, it was not possible to focus exclusively on the coffee ring. Hence, from the edge’s circular circumference, the Raman signal was recorded not only from the coffee ring but also from a small area outside the coffee ring and a small area inside the core of the droplet.
Bevacizumab could be detected at all three circular circumferences of the dried droplet. However, the intensity of the monoclonal antibody’s characteristic Raman peak was higher at the edge’s circumference compared to the main core of the dried drop. This behaviour implied that bevacizumab is deposited in the whole area of the dried droplet, but it is highly distributed in the coffee ring. The deposition of the monoclonal antibody in the coffee ring was unequal at different distances and for different concentrations.
Concerning the deposition of the compounds in the coffee ring, their concentrations, their molecular weights and sizes, their solubility and the solvent’s volatility are the most important factors. An unequal distribution of bevacizumab was observed not only among the different areas of the dried droplet, but also in the coffee ring. Hence, the intensity of the most prominent Raman peak of bevacizumab varied significantly according to the distance from the edge of the dried droplet and the concentration of the solute. As a result, the highest intensities of the bevacizumab peak at 1674 cm
−1 for the 25.00 mg/mL and 18.75 mg/mL bevacizumab solutions were found close to the external edge of the coffee ring, while for the middle and lowest concentrations, they were determined at distances between 25 μm and 100 μm from the outer side of the coffee ring (
Figure 6). Thus, bevacizumab was found to be highly distributed between the edge and the middle of the coffee ring (0–100 μm) depending on its concentration.
Although the calibration curves of bevacizumab intensity plotted against the monoclonal antibody concentration offered sufficient coefficients of determination at most of the distances from the edge of the dried droplet (black dashed lines in
Figure S1), it was observed that even better linear responses could be achieved if two working ranges were employed [
40]; one for the higher (6.25–25.00 mg/mL) (blue dashed lines in
Figure S1) and one for the lower bevacizumab concentrations (0.25–6.25 mg/mL) (red dashed lines in
Figure S1). In this case, most R
2 values of both working ranges at all distances had a value over 0.95, implying a satisfactory linearity. Especially at distances between 25 μm and 125 μm, the highest coefficients of determination were observed for both working ranges (
Figure S1). For the optimal results, the acquisition of Raman spectra at a distance of 75 μm from the external side of the coffee ring was suggested. At that distance, the highest R
2 values were determined, and the calibration curves were sensitive enough, as implied by the highest values of slopes, the smallest standard errors of slope and intercept and the low calculated DL (0.53 mg/mL) (
Table S2).
According to the summary of product characteristics of Avastin
®, the initial solution should be diluted with NaCl 0.9%
w/
v solution for injection. The final concentration of bevacizumab in the diluted solution should be regulated in a range between 1.4 mg/mL and 16.5 mg/mL according to the patient [
7]. The quantification method developed in this study using rotation of the sample offers a linear response in this working range, and the calculated DL for the central circular circumference is lower than 1.4 mg/mL (
Table 4). Thus, the bevacizumab concentration in the final solution could be determined successfully by applying rotation on a dried drop. For achieving optimal results and achieving an even lower DL, focusing on the coffee ring at a distance between 25 μm and 125 μm from the external edge is suggested, approximately in the middle of the coffee ring. At 75 μm from the outer side of the coffee ring, the lowest DL was determined, almost 0.5 mg/mL, and the QL was found to be approximately 1.5 mg/mL (
Table 6), very close to the lowest concentration used for treatment.
The results of the method developed in our study are in accordance with the results of the studies of Rayyad A. et al. [
17] and Makki A.A. et al. [
25]. In the first study, bevacizumab was quantified in solution via Raman spectroscopy using a specially designed cuvette with a spherical mirror behind the sample compartment, as well as directly through perfusion bags. The values of the coefficients of determination determined for the middle (R
2 = 0.993) and the central (R
2 = 0.999) circular circumferences are close to the respective values found when a quartz cuvette (R
2 = 0.999) was employed or when bevacizumab was analysed through the perfusion bag wall (R
2 = 0.999). Moreover, the E
r of the predicted concentrations to the expected concentrations and the DLs were similar to the respective values found when the analysis was performed in the quartz cuvette or through the perfusion bag wall [
17]. In the latter study, ultrafiltration was applied to the monoclonal antibody solutions, so that bevacizumab would be separated from the excipients and, subsequently, it was identified on the filter. However, bevacizumab was significantly concentrated on the filter, and this concentration is not proportional to the stock solution concentration. Thus, quantitation in this study was performed in the initial monoclonal antibody solutions. The calculated coefficient of determination (R
2 = 0.995) [
25] was comparable to the ones determined in our study and those found in the study of Rayyad A. et al. [
17]. In both studies, chemometrics were applied for the quantitative analysis of bevacizumab. These multivariate approaches [
17,
25], though, did not offer more satisfactory results than the univariate method developed in our study. Moreover, this is the first time in the literature that a combination of the dehydration of a droplet and a rotary apparatus has been applied for the quantitation of a monoclonal antibody.
In another study, various monoclonal antibodies were quantified as solutions in vials via Raman spectroscopy by using a machine learning approach and a conventional linear approach. No significant differences were found between the two methods, although machine learning had an advantage against the conventional linear approach in the prediction of the low concentrations, as the relative errors were significantly reduced. However, the results of machine learning could be biased if the training concentrations are over- or under-sampled [
29]. Hence, the linear approaches suggested by the combination of the method employing the home-made rotary apparatus and the method involving focus on the coffee ring developed in our study could be adopted for the quantitative analysis of bevacizumab in the final solutions of Avastin
® formulations, as low DLs were predicted and the trueness of the method was satisfactory enough.
4. Materials and Methods
4.1. Samples
An Avastin® 100 mg/4 mL vial (Roche, Basel, Switzerland) was donated by the Ophthalmology Department of the School of Medicine of Patras University. D-(+)-Trehalose dihydrate, 99% was purchased from Thermo Scientific (Thermo Fisher (Kandel) GmbH, Kandel, Germany) for recording the Raman spectrum of trehalose dihydrate. Sodium chloride (NaCl) 0.9% w/v 1000 mL solution (Vioser S.A., Taxiarches, Trikala, Greece) was used for the dilutions of the initial Avastin® solution and the preparation of the calibration standards.
The calibration standard solutions were prepared by mixing the appropriate volume of the initial Avastin® 25 mg/mL solution with the respective volume of NaCl 0.9% w/v solution. Hence, Avastin® solutions with bevacizumab concentrations in a range of 0.25 mg/mL to 25.00 mg/mL were created. As a result, seven standard solutions (0.25 mg/mL, 1.25 mg/mL, 3.75 mg/mL, 6.25 mg/mL, 12.50 mg/mL, 18.75 mg/mL and 25.00 mg/mL) were prepared and used for the construction of the calibration curves. A 10 μL automated pipette (Pipetman Neo®, Gilson Inc, Middleton, WI, USA) and a 1 mL automated pipette (BioPette® Autoclavable Pipettes, Labnet International Inc, Edison, NJ, USA) were used for the preparation of the standards. For the homogenization of the standard solutions, shaking in a vortex (MS2 Minishaker, IKA®-Werke GmbH & CO., KG, Staufen im Breisgau, Germany) was applied. After the production of the calibration standards, they were analysed immediately and then stored in a refrigerator at 2–8 °C until next use.
4.2. Stereoscope
For the observation of the coffee ring of the dried Avastin® droplet, a Stereoscope Leica M80 (Leica Microsystems Ltd., Heerbrugg, Switzerland) with a digital camera (Leica DFC295, Leica Microsystems Ltd., Heerbrugg, Switzerland) was employed. The stereoscope was also equipped with a set of 10×/23B widefield adjustable eyepiece lenses (Leica part number: 10450023, Leica Microsystems Ltd., Heerbrugg, Switzerland) and a 1.0× achromat objective (Leica part number: 10450159, Leica Microsystems Ltd., Heerbrugg, Switzerland). The zoom was set at 1.25×, providing a total magnification of 12.5. The software LAS© V4.13 (Leica Microsystems Ltd., Heerbrugg, Switzerland) was used for capturing the images of the coffee ring of Avastin® droplet dried on the gold-coated carrier.
4.3. Raman Spectroscopy
An InVia Raman spectrometer (Renishaw, Wotton-under-Edge, UK) coupled with optical microscope (DM Leica, Leica Microsystems, Wetzlar, Germany) was used for the acquisition of the Raman spectra of Avastin® solutions, trehalose dihydrate and bevacizumab calibration standards. The spectrometer consisted of a 785 nm diode laser for the excitation of the samples, a 1200 gr/mm diffraction grating and a charge-coupled device (CCD) detector. The resolution of the laser was 2 cm−1 and its nominal power was 250 mW. For the measurements of the calibration standards, Avastin® solutions and trehalose dihydrate, 90% of the nominal laser power was selected. The time of each Raman spectrum acquisition was set at 10 s, while each spectrum was the result of the accumulation of 10 scans. A 20× (0.4 NA) objective lens (model 566026, Leica Microsystems, Wetzlar, Germany) was employed for the quantitative analysis of the samples. For each Raman spectrum, the spectral area of 100–2000 cm−1 was recorded. For the acquisition of the Raman spectra, the Windows-based software WiRE© 2.0 was used.
For the daily calibration of the Raman spectrometer, the Raman spectrum of a silicon (Si) reference standard was recorded before the first measurement of the day. The Raman spectrum of the reference standard was acquired using 1 s as scan time and 5 × 10−8% of the nominal laser power, and it was the result of 2 accumulated scans. The calibration of the instrument was validated by the shift of the Si peak at 520 cm−1 and the intensity of this characteristic peak.
As sample carrier, a highly reflective gold-coated microscope slide (EMF Corporation, Ithaca, NY, USA) was employed. This sample carrier was a classic microscope slide consisting of two layers; one external bare gold layer (1000 Å) and a second layer of titanium (Ti) (50 Å) binding the external gold layer to the glass of the microscope slide. The slides’ dimensions were 2.6 cm × 7.6 cm and their thickness was 1.0 mm [
49]. For the developed quantitative method, each slide was cut into three pieces, so that each slide would be rectangular with a size of approximately 2.6 cm. A droplet (5–10 μL) of each Avastin
® solution or calibration standard was placed in the centre of the gold-coated carrier using a 1 mL syringe and left to dry at RT overnight. The rectangular gold-coated carriers were placed on the home-made rotary apparatus and the apparatus was transferred under the microscope of the Raman spectrometer. The point of focus was changed to predetermined distances using the high-precision levers of the microscope stage, and manual focusing correction was applied before the acquisition of each Raman spectrum.
4.4. Calibration Curves
After the acquisition of the Raman spectra, the most characteristic peak of bevacizumab at 1674 cm
−1 was integrated using OriginPro 8
® (Originlab Corporation, Northampton, MA, USA). The intensity of the peak was measured after baseline correction in the spectral region of the specific peak [
50]. This process was performed for all recorded calibration standards for Raman spectra through both developed quantification methods.
In the developed quantitative method, the average intensity of bevacizumab Raman peaks at 1674 cm−1 from three sets of dried droplets was determined at each different circular circumference of rotation. The average intensity of bevacizumab Raman peaks at 1674 cm−1 was plotted against the bevacizumab concentration expressed in mg/mL, in order to prepare the calibration curve at each different circumference of the dried droplet (edge, middle, centre). A linear fit was applied to all three calibration curves in a concentration working range of 3.75–25.00 mg/mL.
Concerning the investigation of the optimal position of the dried drop for Raman acquisition, the intensities of the bevacizumab characteristic peak at 1674 cm−1 determined from the four sides/quarters at each different distance were averaged. The average intensity of this peak was plotted against the bevacizumab concentration expressed in mg/mL, in order to produce the calibration curve at a distance of 0 μm from the edge of the dried droplet. The same procedure was followed to create the calibration curves at all different distances (0 μm, 25 μm, 50 μm, 75 μm, 100 μm, 125 μm, 150 μm, 175 μm and 200 μm). A linear fit was applied to all calibration curves in a concentration range of 0.25–25.00 mg/mL and two different working ranges; one low-concentration range (0.25–6.25 mg/mL) and another high-concentration range (6.25–25.00 mg/mL).
4.5. Validation of the Developed Quantitative Analytical Method
The developed method for quantitating bevacizumab in Avastin
® solutions using rotation of the sample carrier was validated for its specificity/selectivity, working range, accuracy, precision, sensitivity and trueness according to the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) guidelines for validation of analytical procedures Q2 (R2) [
37].
The specificity of the quantification method was validated by the absence of interference of any trehalose peaks or any other peaks due to matrix effects from the sample carrier or the NaCl 0.9% w/v solution with the bevacizumab peak at 1674 cm−1 in the Raman spectra of Avastin® solutions. In case that interference of peaks was observed at 1674 cm−1, the selectivity of the method was evaluated by its ability to distinguish between the bevacizumab peak at 1674 cm−1 and the interfering peak of the matrix.
Concerning the working range, it was determined by the concentrations of the calibration standard solutions used for the preparation of the calibration curves. A linear relationship between the average intensity of the bevacizumab peak at 1674 cm−1 and the bevacizumab concentration was evaluated for the specific working range (3.75–25.00 mg/mL) by the value of the calculated coefficient of determination (R2) of the respective regression line. The higher the R2 value, the more satisfactory was the linear regression.
The accuracy of this quantitative analytical method was determined by the relative error (E
r) of the average intensity from three sets of dried droplets of bevacizumab 25.00 mg/mL calibration-standard Raman peaks at 1674 cm
−1 and the expected values of intensity calculated from the calibration curve for this monoclonal antibody concentration. For a satisfactory accuracy, the E
r should not exceed the limit of 15% for the higher concentrations and 20% for the lower concentrations approaching the QL [
38].
Regarding the precision of the developed quantitative method, the repeatability using three sets of dried droplets at three different bevacizumab concentrations was determined. For this purpose, the RSD (%) values of the bevacizumab intensity at 1674 cm
−1 from three sets of dried droplets were calculated for the bevacizumab 3.75 mg/mL, 12.50 mg/mL and 25.00 mg/mL calibration-standard solutions. The acceptance criteria for satisfactory repeatability involve RSD values lower than 15% for the higher concentrations and lower than 20% for concentrations close to the QL [
38]. In the case of biopharmaceutical samples, it is also acceptable even if two out of the three RSD values of the three calibration standards are inside the acceptance criteria [
39].
As far as sensitivity is concerned, the detection (DL) and quantitation limit (QL) of the method were determined. Two different approaches were employed for the calculation of the DL and QL. The first method involved the visual evaluation of the DL, i.e., the minimum concentration at which the monoclonal antibody peak at 1674 cm
−1 could be distinguished visually from the noise level in the Raman spectra. The second approach involved the determination of the DL and QL based on the calibration curve using the following equations:
where σ is the standard deviation of the response (standard error of y-intercept) and m is the slope of the respective calibration curve [
37].
Finally, the trueness of the quantification method was estimated by the prediction of known bevacizumab solution concentrations used as unknown samples in the developed calibration curves. For this purpose, Avastin
® solutions with bevacizumab concentrations of 3.75 mg/mL, 12.50 mg/mL and 25.00 mg/mL were prepared, their Raman spectra were recorded from three sets of dried droplets and bevacizumab intensities at 1674 cm
−1 were measured. Finally, the average intensity of the three measurements was calculated and the concentration of bevacizumab was found from the respective calibration curve. The E
r of the determined bevacizumab concentration to the expected bevacizumab concentration was calculated and it should not exceed the value of 15% or 20% for the lowest concentrations [
38].