Application of Raman Spectroscopy in Downstream Purification

Abstract

An in-situ Raman spectroscopy method and system are provided for characterizing or quantifying protein purification intermediates and / or final concentrate pools during preparation or manufacture. In one embodiment, in-situ Raman spectroscopy is used to characterize or quantify critical quality attributes of protein purification intermediates during downstream processing (i.e., after harvesting the protein purification intermediates). For example, the disclosed in-situ Raman spectroscopy method and system can be used to characterize and quantify the protein purification intermediates when the protein purification intermediates are purified, concentrated, or otherwise formulated into a final drug product for sale or administration.

Description

[0001] Cross - Reference to Related Applications

[0002] This application claims the benefit and priority of U.S. Provisional Patent Application No. 62 / 723,188, filed on August 27, 2018, and incorporates it herein by reference in its entirety if permitted. Technical Field

[0003] The present invention generally relates to systems and methods for monitoring and controlling one or more critical quality attributes (CQAs) or parameters in a downstream protein purification process. Background Art

[0004] In the past decade, the number of monoclonal antibodies (mAbs) approved for therapeutic use has increased significantly. This is due in part to improvements in large - scale manufacturing processes, which facilitate the preparation of large quantities of mAbs. In addition, initiatives such as the Process Analytical Technology (PAT) framework of the U.S. Food and Drug Administration (FDA) have provided innovative solutions for process development, process analysis, and process control to better understand the process and control product quality. The efficient recovery and purification of monoclonal antibodies from cell culture is a critical part of the preparation process. The purification process should reliably produce monoclonal antibodies that are safe for use in humans. This includes monitoring critical quality attributes (CQAs), which include protein attributes and impurities such as host cell proteins, DNA, viruses, endotoxins, aggregates, polymers, excipients, and other substances that may affect patient safety, efficacy, or potency. Protein concentration is also typically a CQA in the purified substance, and appropriate protein concentration in process intermediates may be a key process parameter for unit operation performance. These CQAs need to be monitored throughout the preparation process and the entire program life cycle.

[0005] To ensure that the final formulation of the mAb does not contain more than a determined level of impurities, the mAb product is tested at various stages of downstream processing. Quality control in the manufacture of biological products such as mAbs is typically accomplished by analyzing purified intermediate and formulated drug substance samples using offline methods for each batch preparation. Samples are removed from the processing equipment (e.g., UF / DF skid) and subjected to offline testing to measure product CQAs such as protein concentration (g / L), buffer excipients, and size variants. Real - time monitoring and analysis cannot be performed during the manufacturing process, resulting in increased processing time and a higher risk of batch rejection due to failure to meet CQAs. Therefore, there is a need for rapid in - line methods for real - time quality control monitoring of mAbs.

[0006] Accordingly, it is an object of the present invention to provide systems and methods for real - time monitoring of critical quality attributes during downstream purification processes. Summary of the Invention

[0007] An in-situ Raman spectroscopy method and system are provided for characterizing or quantifying protein purification intermediates during preparation or manufacturing. In one embodiment, in-situ Raman spectroscopy is used to characterize or quantify critical quality attributes of a protein drug during downstream processing (i.e., after harvesting the protein purification intermediate from cell culture fluid). For example, the disclosed in-situ Raman spectroscopy method and system can be used to characterize and quantify critical quality attributes of a protein purification intermediate when purifying the protein purification intermediate (before formulating it into the final drug product to be sold or administered). Critical quality attributes include, but are not limited to, protein concentration, excipients, high molecular weight (HMW) substances, antibody titer, and drug / antibody ratio.

[0008] One embodiment provides a method for preparing a concentrated protein purification intermediate by: using in-situ Raman spectroscopy to simultaneously measure the concentration of the protein purification intermediate in real-time and adjust the parameters of the concentration step in real-time while concentrating / diafiltering the protein purification intermediate to obtain a predetermined concentration target and excipient level required for drug substance formulation. The protein purification intermediate can have a concentration of 5 mg / mL to 300 mg / mL, preferably 50 mg / mL to 300 mg / mL, for subsequent formulation steps. In one embodiment, ultrafiltration is used to concentrate the protein purification intermediate to the desired concentration target during primary or final concentration. Diafiltration is used as a means of buffer exchange during the processing after primary concentration to achieve the desired final formulation components. The protein purification intermediate can be harvested from a bioreactor, fed-batch culture, or continuous culture. In another embodiment, the measurement of the concentration of the protein purification intermediate can be performed continuously or intermittently in real-time. The quantification of the protein concentration can be performed at intervals of about 5 seconds to 10 minutes, once per hour, or once per day. The protein purification intermediate can be an antibody or an antigen-binding fragment thereof, a fusion protein, or a recombinant protein. Spectral data can be collected in one or more wavenumber ranges selected from 977 - 1027 cm -1 、1408 - 1485 cm -1 、1621 - 1711 cm -1 、2823 - 3046 cm -1 and combinations thereof.

[0009] Another embodiment provides a method for preparing a protein purification intermediate, which is carried out as follows: Raman spectroscopic analysis is independently performed on a plurality of protein purification intermediates to generate a general model capable of quantifying any one of the plurality of protein purification intermediates. The general model can be used to determine the concentration of the protein purification intermediate using in-situ Raman spectroscopy from the start to the end of concentrating the protein purification intermediate during the concentration of the protein purification intermediate. Another embodiment provides a method for preparing a protein purification intermediate, which is used to generate a protein-specific model capable of quantifying protein concentration, and this model will be used for commercially viable protein preparation.

[0010] This model can be generated using partial least squares regression analysis of the original spectral data and applying orthogonal methods to the offline protein concentration data. Pretreatment techniques such as standard normal variate (SNV) and / or point smoothing techniques can be first-order derivatives, where the Raman spectral data can be smoothed at 21 cm -1 to reduce model variability and prediction error. Further model refinement can be carried out to isolate the spectral regions related to CQA prediction (such as protein concentration). In one embodiment, the model has an error tolerance of ≤5%, preferably ≤3%.

[0011] Another embodiment provides a method for monitoring and controlling the excipient level in the harvested cell culture fluid and / or protein purification intermediate during downstream purification, which is carried out as follows: The concentration of the excipient is measured in real time using in-situ Raman spectroscopy while purifying the cell culture fluid or protein purification intermediate, and the parameters of the purification step are adjusted in real time to obtain or maintain a predetermined amount of excipient in the harvested cell culture fluid and / or protein purification intermediate. The excipient can be acetate, citrate, histidine, succinate, phosphate, Tris (hydroxymethylaminomethane), proline, arginine, sucrose, or a combination thereof. The excipient can be a surface excipient such as polysorbate 80, polysorbate 20, and poloxamer 188. Brief Description of the Drawings

[0012] Figure 1 is a flowchart showing an exemplary protein purification process.

[0013] Figure 2 is a representative spectrogram showing the initial model development of in-line Raman spectroscopy. The X-axis represents Raman shift. The Y-axis represents intensity. The legend on the right represents protein concentration. The spectral regions used in the initial model development include 977 - 1027 cm from left to right -1 (ring structure), 1408 - 1485 cm -1 (arginine), 1621 - 1711 cm-1 (Secondary structure) and 2823 - 3046 cm -1 (C-H stretching).

[0014] Figure 3 is a bar chart showing the protein concentration (g / L) of mAb1 during standard ultrafiltration / diafiltration unit operations. The blank bars are the concentrations measured using the SoloVPE system (C-technologies) with a UV-Vis based offline method, where the error bars are ±5%, and ±5% is the target for in-line Raman prediction. The bars with wide hatched lines represent the general model that does not include mAb1 in what is referred to herein as Raman prediction, and the bars with narrow hatched lines represent the general model with mAb1. The bars with cross-hatched lines correspond to the prediction results from the initial general model in the range of 0 - 120 g / L. The bars with shaded lines correspond to the prediction results from the initial general model > 120 g / L.

[0015] Figure 4 is a bar chart showing the absolute Raman model error for various mAbs developed from the initial general model. The bars with hatched lines represent the initial general model 0 - 120 g / L (primary concentration and diafiltration), and the blank bars represent from the initial general model > 120 g / L. (final concentration). The horizontal line represents the Raman model target with an error ≤ 5%.

[0016] Figure 5 is a bar chart showing the absolute Raman model error for various mAbs. The bars with wide hatched lines represent 0 - 120 g / L (primary concentration and diafiltration), and the bars with narrow hatched lines represent > 120 g / L (final concentration). The horizontal line represents the Raman model target with an error ≤ 5%. Two forms of the general Raman model are shown. The bars with hatched lines represent the initial general model, and the bars with cross-hatched lines represent the updated general model.

[0017] Figure 6 is a schematic diagram of an ultrafiltration / diafiltration system, including the positions for in-line Raman probe placement.

[0018] Figure 7 is a bar chart showing the protein concentration of mAb10 when using the general model or the mAb10-specific model for final concentration pool (FCP) measurements. The protein concentrations of in-line real-time Raman prediction (bars with wide hatched lines), updated model (bars with narrow hatched lines), and SoloVPE (blank bars) are shown. The X-axis represents the experimental groups, and the Y-axis represents the protein concentration.

[0019] Figures 8A to 8B is a bar chart showing the Raman model error of laboratory-scale DoE modeling of protein concentration. Figure 8AShows the Raman model errors predicted in real time during the various stages of UF / DF processing (primary concentration, diafiltration, and final concentration tank). Figure 8B Shows the Raman model errors of the final DoE model during the various stages of UF / DF processing (primary concentration, diafiltration, and final concentration tank).

[0020] Figure 9A Is a bar graph showing the percentage error in model prediction ability when the model is scaled up to a pilot-scale processing device for mAb11. Laboratory-scale model data is compared with laboratory-scale data incorporating pilot-scale data. The X-axis represents the experimental group, and the Y-axis represents the percentage error in model prediction ability (%). Figure 9B Is a bar graph showing the model prediction ability for pilot-scale processing of various monoclonal antibodies. The X-axis represents the experimental group, and the Y-axis represents the percentage error in model prediction ability (%).

[0021] Figure 10A Is a schematic diagram of an exemplary automated batch UF / DF with Raman feedback. Figure 10B Is a schematic diagram of an exemplary automated single-pass TFF with Raman feedback.

[0022] Figure 11 Is a graph showing the concentration of mAb14 during the UF / DF process. The X-axis represents the throughput (L / m 2 ) and the Y-axis represents the concentration of mAb 14 (g / L).

[0023] Figure 12A Is a bar graph showing the percentage of high molecular weight (HMW) substances in mAb2 during each processing step (primary concentration, diafiltration, final concentration) as determined by SE-UPLC or Raman modeling. The X-axis represents the experimental group, and the Y-axis represents the percentage of mAb2 HMW (%). Figure 12B Is a bar graph showing the percentage of high molecular weight (HMW) substances predicted for mAb 15 using various scan times (10 seconds, 20 seconds, 30 seconds). The X-axis represents the experimental group, and the Y-axis represents the percentage of HMW (%).

[0024] Figure 13 Is a dot plot showing the actual titer (g / L) of monoclonal antibody samples from mAb 14 against the Raman-predicted titer. The X-axis represents the Raman-predicted titer, and the Y-axis represents the actual titer (g / L).

[0025] Figure 14A Is a scatter plot showing the Raman histidine prediction in various monoclonal antibodies. The X-axis represents the histidine concentration predicted by Raman modeling, and the Y-axis represents the actual histidine concentration determined by amino acid analysis. Figure 14B Is a dot plot showing the actual histidine concentration of monoclonal antibody samples against the Raman-predicted histidine concentration.Figure 14C is a dot plot showing the actual arginine concentration of monoclonal antibody samples against the Raman-predicted arginine concentration.

[0026] Figure 15A is a scatter plot showing the actual drug / antibody ratio (DAR) of monoclonal antibody samples from mAb 3 against the Raman-predicted DAR. The X-axis represents the Raman-predicted DAR, and the Y-axis represents the actual DAR determined by ultraviolet spectroscopy. Figure 15B is a scatter plot showing the actual drug / antibody ratio (DAR) of monoclonal antibody samples from mAb 1 against the Raman-predicted DAR. The X-axis represents the Raman-predicted DAR, and the Y-axis represents the actual DAR. Detailed Description

[0027] I. Definitions

[0028] It should be understood that the present disclosure is not limited to the compositions and methods described herein and the experimental conditions described, as they may vary. It should also be understood that the terms used herein are for the purpose of describing certain embodiments only and are not intended to be limiting, since the scope of the present disclosure will be limited only by the appended claims.

[0029] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. Although any compositions, methods, and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention. All publications mentioned are incorporated herein by reference in their entirety.

[0030] In the context of describing the presently claimed invention (particularly in the context of the claims), the terms "a," "an," "the," and similar referents should be construed to cover both the singular and the plural, unless otherwise indicated herein or clearly contradicted by context.

[0031] Unless otherwise indicated herein, the description of a numerical range herein is merely intended as a shorthand method of referring individually to each separate value falling within the range, and each separate value is incorporated into the specification as if it were individually recited herein.

[0032] The use of the term "about" is intended to describe values that are above or below the stated value within a range of approximately ±10%; in other embodiments, the numerical range of values can be above or below the stated value within a range of approximately ±5%; in other embodiments, the numerical range of values can be above or below the stated value within a range of approximately ±2%; in other embodiments, the numerical range of values can be above or below the stated value within a range of approximately ±1%. The foregoing ranges are intended to be made clear by context and do not imply further limitation. Unless otherwise stated herein or the context clearly contradicts, all methods described herein can be performed in any suitable order. Unless otherwise indicated, the use of any and all examples or exemplary language (such as "such as") provided herein is only intended to better illustrate the invention and not to limit the scope of the invention. No language in the specification should be construed as indicating that any non-claimed element is essential for the practice of the invention.

[0033] As used herein, "protein" refers to a molecule comprising two or more amino acid residues linked to one another by peptide bonds. Proteins include polypeptides and peptides and may also include modifications such as glycosylation, lipid linkage, sulfation, γ-carboxylation of glutamate residues, alkylation, hydroxylation, and ADP-ribosylation. Proteins can have scientific or commercial value, including protein-based drugs, and proteins particularly include enzymes, ligands, receptors, antibodies, and chimeric or fusion proteins. Proteins are produced from various types of recombinant cells using well-known cell culture methods and are typically introduced into the cells by transfection of a genetically engineered nucleotide vector (e.g., such as a sequence encoding a chimeric protein, or a codon-optimized sequence, an intronless sequence, etc.), where the vector can exist as an episome or integrate into the cell genome.

[0034] "Antibody" refers to an immunoglobulin molecule composed of four polypeptide chains (two heavy chains (H) and two light chains (L)) linked to each other by disulfide bonds. Each heavy chain has a heavy chain variable region (HCVR or VH) and a heavy chain constant region. The heavy chain constant region contains three domains, namely CH1, CH2, and CH3. Each light chain has a light chain variable region and a light chain constant region. The light chain constant region consists of one domain (CL). The VH region and the VL region can be further subdivided into hypervariable regions, called complementarity-determining regions (CDRs), interspersed with more conserved regions called framework regions (FRs). Each VH and VL consists of three CDRs and four FRs, arranged in the following order from the amino terminus to the carboxyl terminus: FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4. The term "antibody" includes glycosylated and non-glycosylated immunoglobulins of any isotype or subtype. The term "antibody" includes antibody molecules prepared, expressed, produced, or isolated by recombinant means, such as antibodies isolated from host cells transfected to express the antibody. The term antibody also includes bispecific antibodies, which include heterotetrameric immunoglobulins that can bind to more than one different epitope. Bispecific antibodies are generally described in U.S. Patent Application Publication No. 2010 / 0331527, which is incorporated herein by reference.

[0035] "Secondary structure" refers to the local folded structure formed within a polypeptide due to interactions between backbone atoms. The most common types of secondary structure are the α-helix and the β-sheet. Both structures maintain their shape through hydrogen bonds formed between the carbonyl O of one amino acid and the amino H of another amino acid.

[0036] As used herein, "excipient" refers to a pharmacologically inactive substance that is used as a stabilizer for the long-term storage of formulated drug substances. Typically, additional excipients are added to the final concentrate pool to prepare the formulated drug substance. However, during the UF / DF process, the level of excipients is monitored to ensure that the excipient level will not affect the desired formulation strategy. Excipients provide bulk volume to a pharmaceutical formulation, facilitate drug absorption or dissolution, and provide stability and prevent denaturation. Common pharmaceutical excipients include, but are not limited to, amino acids, fillers, binders, disintegrants, coating agents, adsorbents, buffers, chelating agents, lubricants, glidants, preservatives, antioxidants, flavoring agents, sweetening agents, coloring agents, solvents and co-solvents, and thickening agents. In one embodiment, the excipient is polyethylene glycol, including but not limited to PEG-3550.

[0037] "Polyethylene glycol" or "PEG" is a polyether polymer of ethylene oxide commonly used in food, pharmaceuticals, and cosmetics. It is a non-ionic macromolecule and can be used as a molecule to reduce the solubility of biomolecules. PEG is commercially available with different molecular weights ranging from 300 g / mol to 10,000,000 g / mol. Exemplary types of PEG include, but are not limited to, PEG 20000, PEG 8000, and PEG 3350. PEG has different geometries, including linear, branched (3 to 10 chains attached to a central core), star-shaped (10 to 100 chains attached to a central core), and comb-shaped (multiple chains attached to a polymer backbone).

[0038] "Raman spectroscopy" is a spectroscopic technique used to measure the wavelength and intensity of inelastically scattered light from molecules. It is based on the principle that monochromatic incident radiation on a material will be reflected, absorbed, or scattered in a specific manner depending on the specific molecule or protein receiving the radiation. Most of the energy is scattered at the same wavelength, called elastic scattering or Rayleigh scattering. A small amount (<0.001%) is scattered at different wavelengths, called inelastic scattering or Raman scattering. Raman scattering is associated with rotational, vibrational, and electronic energy level transitions. Raman spectroscopy can reveal the chemical and structural composition of a sample.

[0039] As used herein, "ultrafiltration" refers to a membrane method of protein concentration widely used in the downstream processing of protein therapeutics during recombinant protein purification. Ultrafiltration is a size-based separation where substances larger than the membrane pores are retained while smaller substances can pass freely. During the process, the protein solution is pumped tangentially across the surface of a semipermeable parallel plate membrane. The membrane is permeable to buffer and buffer salts but is typically not permeable to monoclonal antibodies. The driving force for permeation is the transmembrane pressure (TMP) imposed by a flow restriction at the outlet of the membrane flow channel (TMP = ((P 进料 +P 渗余物 ) / 2 - P 渗透物 ..).

[0040] As used herein, "primary concentration" refers to the initial step where transmembrane pressure drives water and salts across a permeable membrane, which can reduce the liquid volume and thus increase the protein concentration. The degree of concentration in primary concentration can be optimized to balance throughput, protein stability, processing time, and buffer consumption.

[0041] As used herein, "diafiltration" refers to a technique for exchanging a product of interest from one liquid medium to another using a semipermeable membrane. Buffer exchange and desalting are typically performed using a diafiltration mode where small amounts of impurities and buffer components are effectively washed out of the product by continuous addition of a buffer designed to condition the protein to a stable pH and an excipient concentration that allows for high product concentration. This can be performed in continuous or discontinuous mode based on the processing technique.

[0042] Diafiltration is typically used in combination with ultrafiltration to achieve the desired volume reduction while also removing impurities and salts. UF / DF is the final unit operation in downstream purification, which conditions the mAb to achieve the pH, excipient content, and protein concentration that are favorable for long-term storage and the addition of stabilizing excipients to generate the formulated drug substance (FDS).

[0043] As used herein, "final concentration" refers to the final step in which transmembrane pressure drives water and salts across a permeable membrane, which can reduce the liquid volume and thus increase the protein concentration to the desired target for storage and / or formulation. The resulting pool from the final concentration step is the final concentration pool (FCP). This final concentration step can be carried out in a continuous or discontinuous processing mode.

[0044] The terms "biological product" and "protein purification intermediate" are used interchangeably and refer to any antibody, antibody fragment, modified antibody, protein, glycoprotein, or fusion protein purified from a bioreactor process, as well as the final drug substance.

[0045] The terms "control" and "perform control" refer to adjusting the amount or concentration level of a critical quality attribute in the harvested cell culture fluid to a predetermined set point.

[0046] As used herein, the term "upstream processing" refers to the first step in which an antibody or therapeutic protein is typically produced in a bioreactor through bacterial or mammalian cell lines. Upstream processing includes media preparation, cell culture, and cell separation and harvest. When the cells reach the desired density, the cells are harvested and transferred to the downstream section of the bioprocess. The term "downstream processing" refers to the separation and purification that occurs after harvesting an antibody or therapeutic protein from a bioreactor. Typically, this means recovering the product from an aqueous solution via several different forms. The harvested product is processed to meet the purity and critical quality attribute requirements during downstream processing.

[0047] As used herein, the term "protein purification intermediate" refers to a protein that has been harvested from a bioreactor and to any intermediate during downstream processing.

[0048] The term "concentrated protein purification intermediate" refers to a protein purification intermediate with a concentration greater than 5 mg / mL. More preferably, the concentration is between 50 mg / mL and 300 mg / mL.

[0049] The terms "monitor" and "perform monitoring" refer to periodically checking the amount or concentration level of a critical quality attribute in a cell culture or harvested cell culture fluid.

[0050] The term "harvested cell culture fluid" refers to the fluid removed from a bioreactor containing cells engineered to secrete a protein of interest. The "harvested cell culture fluid" preferably contains the secreted protein of interest, such as a monoclonal antibody.

[0051] As used herein, "Critical Quality Attribute (CQA)" refers to a physical, chemical, biological, or microbiological property or characteristic that should be within an appropriate limit, range, or distribution to ensure the desired product quality of a biotherapeutic drug product. These attributes may affect safety, efficacy, and / or potency. Critical Quality Attributes include, but are not limited to, protein concentration, high molecular weight species, buffer excipients, and pH.

[0052] As used herein, "formulated API" refers to the active ingredient intended to provide pharmacological activity, excluding intermediates used in the synthesis of such ingredient.

[0053] As used herein, "generic model" refers to a mathematical correlation used to predict the spectral characteristics of different recombinant proteins for critical quality attributes.

[0054] As used herein, "mAb-specific model" refers to a mathematical correlation used to predict the spectral characteristics of a specific protein for critical quality attributes.

[0055] As used herein, "titer" refers to the amount of antibody or protein molecules in a solution.

[0056] II. Systems and Methods for Characterizing Downstream Protein Purification Products

[0057] Systems and methods are provided for monitoring and controlling protein concentration during protein manufacturing. Due to the high relative solution viscosity (> 10 cP), concentrated protein solutions are difficult to accurately measure. Accurate quantification requires dedicated offline equipment and typically involves diluting the solution. In high-concentration UF / DF, due to the Gibbs-Donnan effect, the final excipient level is a function of protein concentration. The inability to perform real-time monitoring and analysis during the manufacturing process increases processing time and may result in batch rejection due to non-compliance with CQAs. The systems and methods disclosed herein can be used for in-line monitoring of protein concentration and other critical quality attributes.

[0058] In one embodiment, the Raman spectroscopy system is an in-line or in-situ Raman spectroscopy system used during the preparation of a final concentration pool (which can be highly concentrated (≥ 150 g / L)). Typically, the Raman spectroscopy system is employed downstream of the preparation of protein purification intermediates, such as during the post-harvest processing of protein purification intermediates from a bioreactor or fed-batch culture system and subsequent purification. Figure 1An exemplary protein purification process is shown. Generally, a protein purification intermediate is harvested from a cell culture (100) and processed through multiple purification steps such as affinity capture (110), virus inactivation (120), polishing chromatography (130 and 140), virus retention filtration (150), and ultrafiltration / diafiltration (160) to produce a final concentrate, which is then formulated into a drug substance. In one embodiment, the protein concentration in the harvested cell culture fluid is monitored by in-situ Raman spectroscopy.

[0059] A. Raman Spectroscopy

[0060] In one embodiment, the protein concentration in the harvested cell culture fluid is monitored and controlled by Raman spectroscopy. Raman spectroscopy is a form of vibrational spectroscopy that provides information about molecular vibrations, which can be used for sample identification and quantification of critical quality attributes. In-situ Raman analysis is a method of analyzing a sample in its original location without extracting a portion of the sample for analysis in a Raman spectrometer. The advantage of in-situ Raman analysis is that the Raman spectrometer is non-invasive and non-destructive, which can reduce the risk of contamination and protein quality loss. In-line Raman analysis can be implemented to enable continuous processing while monitoring the protein concentration in the harvested cell culture fluid, protein purification intermediate, and / or final concentrate.

[0061] In-situ Raman analysis can provide a real-time assessment of the protein concentration in the protein purification intermediate. For example, the raw spectral data provided by in-situ Raman spectroscopy can be used to obtain and monitor the current protein concentration in the protein purification intermediate. In this regard, to ensure that the raw spectral data is continuously updated, spectral data should be acquired from the Raman spectroscopy approximately every 5 seconds to 10 hours. In another embodiment, spectral data should be acquired approximately every 15 minutes to 1 hour. In another embodiment, spectral data should be acquired approximately every 20 minutes to 30 minutes.

[0062] Monitoring of the protein concentration in the protein purification intermediate can be analyzed by any commercially available Raman spectrometer capable of performing in-situ Raman analysis. The in-situ Raman analyzer should be able to obtain raw spectral data within the protein purification intermediate. For example, the Raman analyzer should be equipped with a probe that can be inserted into the in-line fluid circuit. Suitable Raman analyzers include, but are not limited to, the RamanRXN2 and RamanRXN4 analyzers (Kaiser Optical Systems, Inc., Ann Arbor, MI, USA).

[0063] The raw spectral data obtained by in-situ Raman spectroscopy can be compared with the results of off-line protein concentration measurements in order to correlate the peaks in the spectral data with the protein concentration. The results of the off-line protein concentration measurements can be used to determine which spectral regions exhibit protein signals. The off-line measurement data can be collected by any suitable analytical method. For example, for protein concentration, the off-line measurements can be collected using SoloVPE (C Technologies). Additionally, any type of multivariate software package, such as SIMCA 13 (MKS Data Analytic Solutions, Umea, Sweden), can be used to correlate the peaks within the raw spectral data with the off-line measurement results of the protein concentration. However, in some embodiments, it may be necessary to preprocess the raw spectral data with a spectral filter to remove any varying baselines. For example, the raw spectral data can be preprocessed with any type of point smoothing technique or normalization technique. Normalization may be required to correct for any probe, optics, laser power variations, and exposure time of the Raman analyzer. In one embodiment, the raw spectral data can be processed with point smoothing (such as the first derivative of the combined 21 cm -1 point smoothing) and normalization (such as standard normal variate (SNV) normalization). These preprocessing techniques can be combined for certain spectral regions to improve model prediction.

[0064] Chemometric modeling can also be performed on the obtained spectral data. In this regard, one or more multivariate methods can be applied to the spectral data, including but not limited to partial least squares (PLS), principal component analysis (PCA), orthogonal partial least squares (OPLS), multiple regression, canonical correlation, factor analysis, cluster analysis, graphical procedures, and the like. In one embodiment, the obtained spectral data is used to create a PLS regression model. The PLS regression model can be created by projecting the predicted variables and the observed variables into a new space. In this regard, the measured values obtained from the Raman analysis and the off-line measurements can be used to create the PLS regression model. The PLS regression model provides predicted process values, such as predicted protein concentration values. In one embodiment, the model provides predicted protein concentration values with an error of ≤5% compared to the off-line protein concentration values. In a preferred embodiment, the model provides predicted protein concentration values with an error of ≤3% compared to the off-line protein concentration values.

[0065] After chemometric modeling, signal processing techniques can be applied to the predicted protein concentration values. In one embodiment, the signal processing techniques will mitigate model variability and prediction error. In this regard, one or more preprocessing techniques can be applied to the predicted protein concentration values. Any preprocessing technique known to those skilled in the art can be utilized. For example, noise reduction techniques can include data smoothing and / or signal rejection. Smoothing is achieved through a series of smoothing algorithms and filters, while signal rejection uses signal characteristics to identify data that should not be included in the analyzed spectral data. In one embodiment, the predicted protein concentration values are denoised by a noise reduction filter. The noise reduction filter provides the final predicted protein concentration values. In this regard, the noise reduction technique combines the raw measurement with a model-based estimate to arrive at what the measurement should be according to the model. In one embodiment, the noise reduction technique combines the current predicted protein concentration value with its uncertainty. The uncertainty can be determined by the reproducibility of the predicted protein concentration value and the current protein concentration value. Once the next predicted protein concentration value is observed, the estimate of the predicted protein concentration value is updated using a weighted average, where estimates with higher certainty are given more weight. Using an iterative method, the final protein concentration value can be updated based on previous measurements and current measurements. In this regard, the algorithm should be recursive and capable of running in real time in order to utilize the current predicted protein concentration value, previous values, and experimentally determined constants. The noise reduction technique can improve the robustness of the measurements received from Raman analysis and PLS prediction by reducing the noise on which the automatic feedback controller will act.

[0066] B. Method of Use

[0067] The disclosed method can be used to monitor and control the protein concentration in harvested cell cultures and / or protein purification intermediate fluids during downstream protein purification processes. Common downstream purification processes include, but are not limited to, centrifugation, direct depth filtration, Protein A affinity purification, virus inactivation steps, ion exchange chromatography, hydrophobic interaction chromatography, size exclusion chromatography, ultrafiltration / diafiltration, virus retention filtration, and combinations thereof. These unit operations are used in a defined sequence to isolate the protein of interest and ensure the monitoring of impurities and / or critical quality attributes prior to the preparation of the formulated drug substance. In one embodiment, the disclosed method includes a Raman probe within a fluid return line. In another embodiment, the disclosed method can be used to prepare a concentrated protein purification intermediate or a final concentrate pool.

[0068] 1. Antibody Titer and Protein Concentration

[0069] Both antibody titer and protein concentration are important factors in the purification of biological products. The antibody titer measured after initially harvesting the cell culture fluid is important for determining the column load and ensuring a reliable purification process for removing impurities. Monitoring the protein concentration throughout the purification steps is important for ensuring the appropriate concentration of the final product and the appropriate performance of the purification unit operations performed. Improper protein concentration can result in an ineffective drug product or the preparation of a formulated drug substance.

[0070] In one embodiment, the harvested cell culture fluid is immediately subjected to Raman spectroscopy after harvesting but before any other purification begins. Raman spectroscopy data can be used after harvesting to quantify the antibody titer in the harvested cell culture fluid. The disclosed method can be used to measure the protein concentration during the protein purification process, for example, during multiple steps such as affinity capture, polishing chromatography, virus retention filtration, or ultrafiltration / diafiltration. In-line Raman probes can detect Raman scattering in the harvested cell culture fluid and / or protein purification intermediates within the fluid circuit without removing the sample from the system, thereby providing analytical characterization that is typically determined in an off-line manner.

[0071] In one embodiment, if the protein concentration is not within a predetermined concentration during the ultrafiltration / diafiltration process, the system is notified and the protein purification intermediate is changed accordingly. For example, if the protein concentration in the protein purification intermediate is below the predetermined protein concentration, the protein purification intermediate can be further concentrated by performing ultrafiltration / diafiltration.

[0072] In one embodiment, the concentration step is performed by protein A affinity chromatography.

[0073] 2. Drug / antibody ratio

[0074] In another embodiment, the disclosed method can be used to monitor and control the drug / antibody ratio (DAR). DAR is a quality attribute that is monitored during the development of antibody-drug conjugates (ADCs), antibody-radionuclide conjugates (ARCs), and common protein conjugates (active steroids, non-cytotoxic payloads, etc.) to ensure consistent product quality and to facilitate subsequent labeling with payloads. DAR is the average number of drugs or other therapeutic molecules conjugated to an antibody and is an important quality attribute in the preparation of therapeutic conjugates. The DAR value can affect the efficacy of the conjugated drug, as low drug loading reduces drug potency while high drug loading negatively impacts pharmacokinetics and safety.

[0075] In one embodiment, the antibody-radionuclide conjugate is immediately subjected to Raman spectroscopy after conjugation but before any other purification is performed. The DAR can be determined using Raman spectroscopy data after conjugation.

[0076] In one embodiment, if the DAR is not within a predetermined concentration during processing, the system is notified and the ADC intermediate is changed accordingly. For example, if the DAR in the ADC intermediate is lower than the predetermined DAR, the conjugation reaction components can be changed, such as optimizing the reactant concentration, changing the type of linker, optimizing the temperature, or other manufacturing variables.

[0077] 3. Buffer excipients

[0078] The disclosed methods can be used to monitor and control the levels of buffer excipients in harvested cell culture fluids and / or protein purification intermediates during downstream purification. Buffer excipients commonly used in monoclonal antibody preparation include, but are not limited to, acetate, citrate, histidine, succinate, phosphate, and tris(hydroxymethyl)aminomethane (Tris), proline, and arginine. Surfactant excipients include, but are not limited to, polysorbate 80 (Tween 80), polysorbate 20 (Tween 20), and poloxamer 188. Polyol / disaccharide / polysaccharide excipients include, but are not limited to, mannitol, sorbitol, sucrose, and dextran 40. Antioxidant excipients include, but are not limited to, ascorbic acid, methionine, and ethylenediaminetetraacetic acid (EDTA). Two commonly used amino acid excipients are histidine and arginine. In a preferred embodiment, the excipients monitored and controlled are histidine and arginine.

[0079] Due to the combination of excluded volume and the Donnan effect, the final concentrate pool excipient concentration is different from the composition of the diafiltration buffer. The Donnan effect is a phenomenon caused by the retention of positively charged proteins by the membrane during UF / DF, combined with the requirement for electrical neutrality in both the retentate and the permeate. To balance the positively charged proteins, negatively charged buffer components are enriched in the retentate relative to the diafiltration buffer, while positively charged buffer components are excluded. This effect can result in significant differences in the FCP pH and buffer excipient concentration compared to the diafiltration buffer composition (Stoner et al., J Pharm Sci, 93:2332-2342 (2004)).

[0080] Volume exclusion describes the behavior of high-concentration samples where the protein occupies a large portion of the solution volume. The buffer is excluded from the volume occupied by the protein, resulting in a decrease in buffer solute concentration as the protein concentration increases, expressed as moles (or mass) of solute per solution volume. The buffer is excluded from the volume occupied by the protein, resulting in a decrease in buffer solute concentration as the protein concentration increases when expressed as moles (or mass) of solute per solution volume.

[0081] Based on these principles and the fact that buffer excipient levels are critical quality attributes, in-line Raman probes will minimize off-line analytical characterization and provide further process understanding to ensure that excipient levels are adequate prior to formulation.

[0082] 4. High Molecular Weight Impurities

[0083] In the preparation of monoclonal antibodies, even after extensive purification steps, low levels of product-related impurities remain. High molecular weight (HMW) species, such as antibody dimer species, are product-related impurities that contribute to the size heterogeneity of mAb products. The formation of HMW species due to protein aggregation in therapeutic mAb drug products may potentially compromise drug efficacy and safety. HMW species are considered a CQA for routine monitoring during drug development and are part of the release testing of purified protein drug products during manufacturing.

[0084] In one embodiment, the disclosed method can be used to identify protein drug products containing HMW species. HMW species can be detected by Raman spectroscopy at multiple steps during the purification process, including but not limited to during affinity capture, during virus inactivation, during polishing chromatography, during virus retention filtration, during ultrafiltration / diafiltration, or combinations thereof.

[0085] In one embodiment, the disclosed method detects HMW species in harvested cell culture fluid and further processes the fluid to remove HMW species. Methods for removing HMW species from cell culture fluid include additional polishing steps, including but not limited to cation exchange chromatography and anion exchange chromatography.

[0086] C. UF / DF System

[0087] Figure 6An ultrafiltration / diafiltration processing system is shown, including multiple locations (circled numbers 1 to 4) for in-line Raman probes. The protein purification intermediate is pumped into the retentate container 205 by the diafiltration pump 200, which can be a peristaltic pump, a rotary vane pump, a pressure transfer pump, or a diaphragm pump. The fluid from the retentate container 205 flows to the feed pump 210 (which can be a rotary vane pump, a peristaltic pump, or a diaphragm pump), passes through the feed pressure valve 215, and enters the tangential flow filtration module (TFFM) 220. In the TFFM 220, the protein purification intermediate undergoes transmembrane ultrafiltration. The biological product of interest is retained in the fluid (retentate), while water and low molecular weight solutes (including buffer excipients) pass through the membrane in the permeate (filtrate), and the permeate leaves the system by passing through the permeate pressure valve 225 and entering the waste tank 230. The retentate leaves the TFFM 220 and passes through the retentate pressure valve 235, the transmembrane pressure control valve 240, and the retentate return channel 245, where the retentate flows back into the retentate container 205. This process can be repeated as needed to concentrate the biological product, remove impurities, and ensure that the CQAs are within acceptable limits. During diafiltration, the same flow path as described above is followed, where permeable solutes are replaced when fresh buffer is washed into the product stream. When fresh buffer is added at the same rate as the permeate is removed from the system, the sum of the retentate tank and skid hold-up volume will define the system volume. One turnover volume (TOV) is defined as the amount of diafiltration buffer added to the UF / DF process equal to the system volume. Typically, replacing 8 times the system volume (8 TOV) ensures >99.9% buffer exchange (Schwarts, L., Scientific and Technical Report, PN 33289)

[0088] In addition, during the UF / DF process, it is necessary to mix the protein solution in the retentate container 205. The density differences between the diafiltration buffer, the retentate return, and the bulk retentate during diafiltration require that the agitation in the tank be sufficient to ensure adequate buffer exchange but moderate enough to avoid shear, as it has been observed that this can lead to protein aggregation and sub-visible particle (SVP) formation in some products. Additionally, it is important to ensure adequate mixing of the retentate return during the concentration phase to prevent protein concentration polarization in the retentate tank, which can result in a higher protein concentration being delivered to the UF / DF membrane.

[0089] In one embodiment, the Raman probe is disposed at position 1 in the retentate container 205 downstream of the peristaltic pump 200. As an alternative, the Raman probe can also be disposed at position 2 in-line downstream of the retentate container 205 and before the feed pump 210. In another embodiment, the Raman probe is disposed in-line at position 3 between the feed pressure pump 215 and the tangential flow filtration module 220. In another embodiment, the Raman probe is disposed in-line in the retentate return channel 245. The position of the Raman probe is crucial for ensuring accurate in-line measurements in a complex system with engineering and processing constraints.

[0090] D. Cell Culture

[0091] The harvested cell culture fluid can be harvested from a bioreactor containing cells engineered to produce monoclonal antibodies. The term "cell" includes any cell suitable for expressing a recombinant nucleic acid sequence. Cells include prokaryotic and eukaryotic cells, such as bacterial cells, mammalian cells, human cells, non-human animal cells, avian cells, insect cells, yeast cells, or cell fusions, such as, for example, hybridomas or quadromas. In certain embodiments, the cells are human, monkey, ape, hamster, rat, or mouse cells. In other embodiments, the cells are selected from the following cells: Chinese hamster ovary (CHO) (e.g., CHO K1, DXB-11 CHO, Veggie-CHO), COS (e.g., COS-7), retinal cells, Vero, CV1, kidney cells (e.g., HEK293, 293EBNA, MSR293, MDCK, HaK, BHK21), HeLa, HepG2, WI38, MRC 5, Colo25, HB8065, HL-60, lymphocytes such as Jurkat (T lymphocytes) or Daudi (B lymphocytes), A431 (epidermal cells), U937, 3T3, L cells, C127 cells, SP2 / 0, NS-0, MMT cells, stem cells, tumor cells, and cell lines derived from the above cells. In some embodiments, the cells contain one or more viral genes, such as retinal cells expressing viral genes (e.g., PER. cells). In some embodiments, the cells are CHO cells. In other embodiments, the cells are CHO K1 cells.

[0092] In protein production, "fed-batch cell culture" or "fed-batch culture" refers to a batch culture in which cells and a culture medium are initially supplied to a culture vessel, and additional culture nutrients are slowly fed to the culture in discrete increments during the culture, with or without periodic cell and / or product harvest before termination of the culture. Fed-batch culture includes "semicontinuous fed-batch culture" in which the entire culture (possibly including cells and medium) is periodically removed and replaced with fresh medium. Fed-batch culture is different from simple "batch culture" in which all components for cell culture (including animal cells and all culture nutrients) are provided to the culture vessel at the beginning of the culture process. Fed-batch culture can be different from "perfusion culture" in that supernatant is not removed from the culture vessel in a standard fed-batch process, while in perfusion culture, cells are retained in the culture, e.g., by filtration, and medium is introduced into and removed from the culture vessel continuously or intermittently. However, it is contemplated that samples may be removed during fed-batch cell culture for testing purposes. The fed-batch process continues until it is determined that a maximum working volume and / or protein production has been reached, and then the protein is harvested.

[0093] The phrase "continuous cell culture" relates to a technique for continuously culturing cells (usually in a specific growth phase). For example, if a continuous supply of cells is needed, or if a specific protein of interest needs to be produced, the cell culture may need to be maintained in a specific growth phase. Therefore, the conditions must be continuously monitored and adjusted accordingly to maintain the cells in that specific growth phase.

[0094] The terms "cell culture medium" and "culture medium" refer to a nutrient solution for culturing mammalian cells, which typically provides the necessary nutrients to enhance cell growth, such as carbohydrate energy sources, essential amino acids (such as phenylalanine, valine, threonine, tryptophan, methionine, leucine, isoleucine, lysine, and histidine) and non-essential amino acids (such as alanine, asparagine, aspartic acid, cysteine, glutamic acid, glutamine, glycine, proline, serine, and tyrosine), trace elements, energy sources, lipids, vitamins, etc. The cell culture medium may contain extracts, such as serum or peptone (hydrolysate), which can provide the raw materials to support cell growth. The culture medium may contain extracts from yeast or soy instead of extracts from animals. Chemically defined medium refers to a cell culture medium in which all chemical components are known (i.e., have a known chemical structure). Chemically defined medium is completely free of components derived from animals, such as serum or animal-derived peptone. In one embodiment, the medium is a chemically defined medium.

[0095] The solution can also contain components that enhance growth and / or survival rate above the minimum rate, including hormones and growth factors. The solution can be formulated to have an optimal pH and salt concentration for the survival and proliferation of the specific cells being cultured.

[0096] E. Protein of interest

[0097] Any protein of interest suitable for expression in prokaryotic or eukaryotic cells can be monitored using the disclosed methods. For example, proteins of interest include (but are not limited to) antibodies or antigen-binding fragments thereof, chimeric antibodies or antigen-binding fragments thereof, ScFv or fragments thereof, Fc fusion proteins or fragments thereof, growth factors or fragments thereof, cytokines or fragments thereof, or extracellular domains of cell surface receptors or fragments thereof. The protein of interest can be a simple polypeptide consisting of a single subunit or a complex multi-subunit protein consisting of two or more subunits. The protein of interest can be a biopharmaceutical product, a food additive or preservative, or any protein product subject to purification and quality standards.

[0098] In some embodiments, the antibody is selected from anti-programmed cell death 1 antibody (e.g., anti-PD1 antibody as described in U.S. Patent No. 9,987,500), anti-programmed cell death ligand 1 (e.g., anti-PD-L1 antibody as described in U.S. Patent No. 9,938,345), anti-D114 antibody, anti-angiopoietin 2 antibody (e.g., anti-ANG2 antibody as described in U.S. Patent No. 9,402,898), anti-angiopoietin-like 3 antibody (e.g., anti-AngPt13 antibody as described in U.S. Patent No. 9,018,356), anti-platelet-derived growth factor receptor antibody (e.g., anti-PDGFR antibody as described in U.S. Patent No. 9,265,827), anti-Erb3 antibody, anti-prolactin receptor antibody (e.g., anti-PRLR antibody as described in U.S. Patent No. 9,302,015), anti-complement 5 antibody (e.g., anti-C5 antibody as described in U.S. Patent No. 9,795,121), anti-TNF antibody, anti-epidermal growth factor receptor antibody (e.g., anti-EGFR antibody as described in U.S. Patent No. 9,132,192 or anti-EGFRvIII antibody as described in U.S. Patent No. 9,475,875), anti-proprotein convertase subtilisin / kexin type 9 antibody (e.g., anti-PCSK9 antibody as described in U.S. Patent No. 8,062,640 or U.S. Patent No. 9,540,449), anti-growth and differentiation factor 8 antibody (e.g., anti-GDF8 antibody, also known as anti-myostatin antibody, as described in U.S. Patent No. 8,871,209 or No. 9,260,515), anti-glucagon receptor (e.g., anti-GCGR antibody as described in U.S. Patent No. 9,587,029 or No. 9,657,099), anti-VEGF antibody, anti-IL1R antibody, interleukin 4 receptor antibody (e.g., anti-IL4R antibody as described in U.S. Patent Application Publication No. US2014 / 0271681A1 or U.S. Patent No. 8,735,095 or No. 8,945,559), anti-interleukin 6 receptor antibody (e.g., anti-IL6R antibody as described in U.S. Patent No. 7,582,298, No. 8,043,617 or No. 9,173,880), anti-IL1 antibody, anti-IL2 antibody, anti-IL3 antibody, anti-IL4 antibody, anti-IL5 antibody, anti-IL6 antibody, anti-IL7 antibody, anti-interleukin 33 (e.g., anti-IL33 antibody as described in U.S. Patent No. 9,453,072 or No. 9,637,535), anti-respiratory syncytial virus antibody (e.g., anti-RSV antibody as described in U.S. Patent Application Publication No. No. 9,447,173), anti-cluster of differentiation 3 (e.g., as described in U.S. Patent No. 9,447,173 and No. 9,447,173 and in U.S. Application No.anti-CD3 antibodies as described in 62 / 222,605), anti-cluster of differentiation 20 (e.g., anti-CD20 as described in U.S. Patent No. 9,657,102 and No. US20150266966A1 and in U.S. Patent No. 7,879,984), anti-CD19 antibodies, anti-CD28 antibodies, anti-cluster of differentiation 48 (e.g., anti-CD48 antibodies as described in U.S. Patent No. 9,228,014), anti-Fel d1 antibodies (e.g., as described in U.S. Patent No. 9,079,948), anti-Middle East Respiratory Syndrome virus (e.g., anti-MERS antibodies as described in U.S. Patent No. 9,718,872), anti-Ebola virus antibodies (e.g., as described in U.S. Patent No. 9,771,414), anti-Zika virus antibodies, anti-lymphocyte activation gene 3 antibodies (e.g., anti-LAG3 antibodies or anti-CD223 antibodies), anti-nerve growth factor antibodies (e.g., anti-NGF antibodies as described in U.S. Patent Application Publication No. US2016 / 0017029 and in U.S. Patents No. 8,309,088 and No. 9,353,176), and anti-activin A antibodies.

[0099] In some embodiments, the bispecific antibody is selected from anti-CD3×anti-CD20 bispecific antibodies (such as those described in U.S. Patent No. 9,657,102 and U.S. Patent Application Publication No. US20150266966A1), anti-CD3×anti-mucin 16 bispecific antibodies (e.g., anti-CD3×anti-Muc16 bispecific antibodies), and anti-CD3×anti-prostate specific membrane antigen bispecific antibodies (e.g., anti-CD3×anti-PSMA bispecific antibodies). In some embodiments, the protein of interest is selected from abciximab, adalimumab, adalimumab-atto, ado-trastuzumab, alemtuzumab, alirocumab, atezolizumab, avelumab, basiliximab, belimumab, benralizumab, bevacizumab, bezlotoxumab, blinatumomab, brentuximab vedotin, brodalumab, canakinumab, capromab pendefide, PEGylated certolizumabpegol), cemiplimab, cetuximab, denosumab, dinutuximab, dupilumab, durvalumab, eculizumab, elotuzumab, emicizumab-kxwh, emtansinealirocumab, evinacumab, evolocumab, fasinumab, golimumab, guselkumab, ibritumomab tiuxetan, idarucizumab, infliximab, infliximab-abda, infliximab-dyyb, ipilimumab, ixekizumab, mepolizumab, necitumumab, nesvacumab, nivolumab, obiltoxaximab, obinutuzumab, ocrelizumab, ofatumumab, olaratumab, omalizumab, panitumumab, pembrolizumab, pertuzumab, ramucirumab, ranibizumab, raxibacumab, reslizumab, rinucumab, rituximab, sarilumab, secukinumab, siltuximab, tocilizumab, tocilizumab, trastuzumab, trevogrumab, ustekinumab, and vedolizumab.

[0100] Examples

[0101] Example 1: General in-line protein concentration model for UF / DF applications

[0102] Materials and Methods

[0103] Data collection for this model included spectral data obtained from Raman Rxn2 and Rxn 4 analyzers (Kaiser Optical Systems, Inc., Ann Arbor, MI, USA) using an MR-Probe-785 and RamanRxn Probehead-758 (Kaiser Optical Systems, Inc., Ann Arbor, MI, USA). In addition, several different optical elements were used throughout the development, depending on availability. The operating parameters of the Raman analyzer were set to a 10-second scan time, with 6 accumulations, repeated 5 times. SIMCA 13 (MKS Data Analytic Solutions, Umeå, Sweden) was used to correlate the peaks within the spectral data to off-line protein concentration measurements. In-line measurements were made at all different points of the UF / DF unit operation, including primary concentration, diafiltration, and final concentration. Off-line protein concentration was determined using SoloVPE (C Technologies, Inc.). SoloVPE measurements were made in triplicate.

[0104] Figure 2 The spectral regions used to build the chemometric model are shown. These regions include Region 1 - 977 - 1027 cm -1 (ring structure), Region 2 - 1408 - 1485 cm -1 (arginine), Region 3 - 1621 - 1711 cm -1 (secondary structure), and Region 4 - 2823 - 3046 cm -1 (C-H stretch). The following spectral filtering was performed on the raw spectral data: Combining a 21 cm -1 point smoothed first derivative to remove the varying baseline

[0105] Results

[0106] To determine the feasibility of the general in-line protein concentration model for ultrafiltration / diafiltration (UF / DF) applications, mAb1 was analyzed using Raman spectroscopy. Protein concentrations were measured before diafiltration (primary concentration), during diafiltration (diafiltration), and after diafiltration (final concentration). The concentration calculated by the model was compared to the protein concentration determined by SoloVPE ( Figure 3 ). When including the training set data (mAb 1 spectra incorporated into the PLS model), the model error was 3.1% for 0 - 120 g / L (primary concentration and diafiltration) and 1.8% for >120 g / L (final concentration).

[0107] Processing errors can be detected by Raman spectroscopy. Figure 3 Shows detection of air entrainment in the system during final recycle through the UF / DF system by Raman spectroscopy data. The bars enclosed by the rectangle show that the mAb1 concentration predicted by SoloVPE > 200 g / L, while the Raman prediction is about 65 g / L.

[0108] Figure 4 Shows the absolute Raman model errors for ten representative mAbs. This data shows successful development of models for the shown mAbs, which include different mAb isotypes (IgG1 and IgG4) as well as bispecific molecules. Fourteen out of seventeen model predictions meet the error ≤ 5%. However, specific models (0 - 120 g / L and > 120 g / L) were created for each probe, which were used as probes during development to probe the variability that increases the error predicted by the PLS model.

[0109] To optimize the model, different probes and lasers were tested with multiple mAbs. Model refinement was carried out, where only one spectral region was the focus of the updated general model: 2823 - 3046 cm -1 (C-H stretch), spectral filtering using standard normal variate (SNV) was performed to correct for laser power variations and probe variability as baseline correction. The comparison of the two developed model components is summarized in Table 1. Using the corresponding offline SoloVPE measurements (performed in triplicate), a partial least squares (PLS) regression model was created. The details of the partial least squares regression model are shown in Table 2. The updated dataset predicted by the optimized laser / probe general model shows that fifteen out of seventeen model predictions meet the error ≤ 5%, compared to only fourteen out of seventeen model predictions previously ([ Figure 5 ).

[0110] Table 1. Comparison of general model components

[0111]

[0112] Table 2. Details of the partial least squares regression general model for protein concentration (version 2)

[0113] Final Model 0 - 120 g / L > 120 g / L Sample Size 1412 879 <![CDATA[R 2 X]]> 0.993 0.987 <![CDATA[Q 2 > 0.984 0.958 RMSECV 3.34 7.77

[0114] R 2 Percentage of variance explained by the X - model, target: R 2 > 0.9

[0115] Q 2 - Percentage of variance of model predictions during cross - validation, target: Q 2 > 0.8

[0116] RMSECV: Root Mean Square Error of Cross-Validation

[0117] Example 2: Scale-up Performance of Protein Concentration Model

[0118] Materials and Methods

[0119] The optimized general model (version 2) (see Table 1) was tested with mAb10 using a scaled-up 1 / 2-inch single-use tangential flow filtration system (Pall Corporation) for experimental use. The mAb10 feed material was formulated API, rather than the typical processed feed material. The FDS material was diluted to a representative UF / DF feed source (including protein concentration and buffer excipients). However, due to the presence of additional excipients in the feed source that were not tested during the development of the general model, a mAb 10-specific model was created. For methods regarding Raman data collection and scan length information, see Example 1. The mAb-specific model for >120 g / L uses the same spectral regions and preprocessing techniques as the general model (version 2). However, the 0-120 g / L mAb-specific model uses four spectral regions as shown in Figure 2 and standard normal variate preprocessing. The differences in the 0-120 g / L model can be attributed to the additional excipients in the feed source.

[0120] Results

[0121] When the training set was included in the model prediction, 2 out of 4 updated models met the model goal of error ≤ 5%, as shown in Table 3. Based on significant differences during the preliminary experiments at IOPS such as feed source, laser, and scale (laboratory scale vs. scaled-up scale), a second run was conducted. Before the second experiment, changes were made to the 0-120 g / L mAb 10-specific model. All four regions were included using SNV preprocessing, but three regions: 977-1027 cm -1 , 1408-1485 cm -1 and 1621-1711 cm -1 also additionally used the first derivative combined with 21 cm -1 point smoothing. The summarized experimental results are presented in Table 3. In the second experiment, when the training set was included in the model prediction, 3 out of 4 updated models met the model goal of error ≤ 5%, as shown in Table 3. Another observation during data analysis was that the feed source was the main contributing factor to increased error. If the feed sample was removed, the general model error would be reduced from 8.6% to 5.7%. In Figure 7In [Figure 0], the in-line prediction results (real-time, bar with bandwidth shaded lines) of the final concentration tank and the results of the updated model (bar with narrow shaded lines) are shown, and the protein concentration is compared with the off-line measurement results of SoloVPE (blank bar). For the final concentration tank, the in-line of the general model and the updated model have an error of 2.7%, while the in-line of the mAb 10 model and the updated model are 12.0% and 4.2% respectively. The increased error observed in the mAb 10 model can be attributed to the limited data in the range of about 250 g / L, while the general model has a larger data set. Another factor contributing to the increased error is the inability of the model to extrapolate outside the characterization range (i.e., >250 g / L in the mAb 10 model).

[0122] Table 3. Average model error of protein concentration prediction in scale-up experiments

[0123]

[0124] Example 3: Scaling up the protein concentration model to a pilot-scale processing device.

[0125] Materials and Methods

[0126] During process development in the commercialization process, UF / DF is characterized by unit operations following the "Quality by Design" principle to understand key process parameters and key quality attributes. Raman and model development were included during the development of mAb11 to enhance process understanding and provide a streamlined approach to model development. For information on the method of Raman data collection, see Example 1, but the scan duration was adjusted from 10 seconds to 5 seconds. The developed mAb 11 model uses SNV preprocessing for all four spectral regions, but for three regions: 977 - 1027 cm -1 、1408 - 1485 cm -1 and 1621 - 1711 cm -1 ,an additional first derivative combined with 21 cm -1 point smoothing is used.

[0127] Results :

[0128] A laboratory-scale model was generated using four DoE experiments and 4 spectral regions. Figure 8A Shows the Raman model error of real-time prediction from four DoE experiments. Figure 8B Shows the Raman model error of another 15 experiments using the laboratory-scale model. mAb-specific protein concentration models of 0 - 120 g / L and >120 g / L were generated with an error ≤5%.

[0129] Using a laboratory-scale model (n = 15), the prediction error for the pilot-scale run (n = 3) for mAb11 was 0.6% - 10.6%( Figure 9A ). When pilot-scale data was incorporated into the laboratory-scale model (n = 18), the pilot-scale prediction error decreased to 0.6 - 2.2%. The increased laboratory-scale model error may be due to the effect of temperature on Raman spectral shifts caused by heat dissipation associated with the scaled-up equipment. Temperature will be a consideration in future Raman development and model validation claims.

[0130] In seven pilot-scale experiments using three different monoclonal antibodies (mAb J, mAb K, and mAb L), the final model prediction error was 0.6 - 2.2%; well within the 5% target( Figure 9B ).

[0131] Example 4: Using the Raman model for real-time concentration determination to enable process decisions.

[0132] Materials and Methods :

[0133] For further information, see Example 3, as the same procedures for Raman spectroscopy collection and modeling are used for Raman automation. An automated control strategy was developed to use Raman spectroscopy data to achieve the final protein concentration target. Using the generated prediction model, the data was filtered and used to provide input to the instrument on the UF / DF to terminate the unit operation when the protein concentration target was achieved. SoloVPE measurements were performed in triplicate.

[0134] Results :

[0135] Figure 10A and 10B shows an exemplary screen setup for the automated monitoring of batch UF / DF and single-pass TFF protein concentration. Figure 11 The prediction model was shown to be useful for monitoring the real-time concentration of mAb14 in each processing step and triggering the stop of the concentration unit operation when the desired concentration target was reached. The Raman prediction for the final concentrate pool was 260 g / L, compared to an offline SoloVPE measurement result of 262 g / L, resulting in an error of 0.8%, meeting the error target of ≤5%. Raman is a suitable application for making automated process decisions to verify that the desired protein concentration target is met.

[0136] Example 5: Proof of concept for high molecular weight (HMW) substance modeling in UF / DF.

[0137] Materials and Methods :

[0138] Data collection for the HMW substance model included spectral data from a Raman Rxn2 analyzer (Kaiser Optical Systems, Inc., Ann Arbor, Michigan, USA) and a RamanRxn Probehead-758 (Kaiser Optical Systems, Inc., Ann Arbor, Michigan, USA). Additionally, several different optical elements were used throughout the development process, depending on availability. The operating parameters of the Raman analyzer were set to a scan time of 72 seconds, 1 accumulation, and repeated 25 times. In-line measurements were taken at all different points of UF / DF unit operation, including primary concentration, diafiltration, and final concentration. The spectral range was 110 - 3415 cm -1 -1. The raw spectral data was preprocessed using SNV, and additional filtering was performed using a first derivative combined with 21 cm -1 -1 smoothing. Size-exclusion ultra-high performance liquid chromatography was used to determine the offline HMW substance measurements.

[0139] Results :

[0140] The HMW substance is another attribute that is considered a preliminary critical quality attribute in protein purification. Current technology cannot monitor the HMW substance in real time during processing. Figure 12A It has been shown that the disclosed Raman modeling method can be used to monitor the HMW substance during protein purification. Throughout the purification process (primary concentration, diafiltration, and final concentration), the predicted results of the HMW substance obtained by Raman modeling were compared with the measurement results collected using SE-UPLC. Raman modeling effectively predicted the percentage of high molecular weight substances in real time during protein processing. A model with an average error of 3.4% was generated.

[0141] Example 6. Proof-of-concept for modeling high molecular weight (HMW) substances in polishing chromatography.

[0142] Materials and Methods :

[0143] Data acquisition for the HMW substance model included spectral data obtained from a Raman Rxn2 analyzer (Kaiser Optical Systems, Inc., Ann Arbor, Michigan, USA) using an MR-Probe-785. The operating parameters of the Raman analyzer were set to a scan time of 10 seconds, 30 seconds, or 60 seconds, 1 accumulation, and repeated measurements 5 times. Offline measurements were performed using anion exchange chromatography (AEX) pools with a total HMW of 6.2% - 76.2%. The spectral range used for modeling and the preprocessing techniques employed are described in Table 4. Size-exclusion ultra-high performance liquid chromatography was used to determine the offline HMW substance measurements.

[0144] Results :

[0145] The disclosed Raman modeling method can be used to monitor HMW substances during purified chromatographic protein purification. The predicted results of HMW substances obtained by Raman modeling are compared with the measurement results collected from the generated AEX pool using SE-UPLC. As summarized in Table 4, the RMSEP of the evaluated method is 3.2 - 7.6%. In Figure 12B , a compressed data set of 6.2% - 19.7% HMW was used to evaluate the model generated using the spectral region of 350 - 3100 cm -1 and the SNV preprocessing technique. By reducing the HMW range, for 10 seconds, 30 seconds, and 60 seconds, the RMSEP was reduced to 1.2%, 1.4%, and 2.1% respectively. Based on these results, Raman can be used to determine the HMW content in the AEX pool.

[0146] Table 4. Error (RMSEP) of HMW model prediction when the HMW content is 6.2% - 76.2%.

[0147]

[0148]

[0149] Example 7. Proof of concept for titer modeling.

[0150] Materials and Methods :

[0151] The training set model was for 35 protein A flow-through samples doped with FCP (265 g / L) to achieve titers of 0.36 - 9.8 g / L. The model was evaluated using diluted depth filtration filtrate samples with titers of 1.3 - 8.8 g / L. Data collection for the titer model included spectral data obtained using an MR-Probe-785 from Raman Rxn2 (Kaiser Optical Systems, Inc., Ann Arbor, Michigan, USA). An immersion probe was used to generate spectral data offline, with the operating parameters set to a 20-second scan time, 1 accumulation, and repeated 5 times. The spectral ranges were 977 - 1027, 1408 - 1485, 1621 - 1711, and 2823 - 3046 cm -1 . The raw spectral data was preprocessed using SNV and additional filtering was performed using a first derivative combined with 21 cm -1 smoothing. The model characteristics are described in Table 5.

[0152] Table 5. Antibody titer model characteristics.

[0153]

[0154] Results :

[0155] Antibody titer is a process attribute in protein purification, and this attribute is required to inform subsequent downstream purification unit operations, including affinity column loading, production consistency, and intermediate volume limits during the process. Inaccurate column loading will affect subsequent preliminary critical quality attributes, so monitoring techniques such as Raman spectroscopy are needed. Figure 13 Shows the actual antibody titer of a monoclonal antibody versus the Raman-predicted antibody titer. In this experiment, the model error was 26%, higher than the desired target of ≤5%. Increasing the scan duration and developing the model using diluted and undiluted depth filtration filtrates will reduce the model error.

[0156] Example 8: Raman model for buffer excipient measurement that meets the current orthogonal assay error of approximately 10%.

[0157] Materials and Methods :

[0158] Data was collected from previous concentration model development runs for various antibodies. For information on the method of Raman data collection, see Example 1. Table 6 shows the model components for detecting histidine and arginine in the samples. The spectral regions are based on known histidine / arginine peaks (Zhu et al., Spectrochim Acta A Mol Biomol Spectrosc, 78(3): 1187 - 1195 (2011)). After the initial model development for histidine and arginine, further model characterization was performed with mAb 14. Using a non-contact optical probe, the operating parameters of the Raman analyzer were set to a 20-second scan time with 5 accumulations. The spectral range for histidine was 1200 - 1480 cm -1 , and for arginine it was 860 - 1470 cm -1 . For both buffer excipients, the raw spectral data was preprocessed using SNV, and additional filtering was performed using a first derivative combined with 21 cm -1 smoothing. Offline histidine and arginine substance measurements were determined using a method based on ultra-performance liquid chromatography (UPLC) amino acid quantification.

[0159] Results :

[0160] To determine whether the disclosed Raman modeling method and system can be used to measure buffer excipients in processed antibody samples, data collected from previous concentration model development runs was analyzed for histidine and arginine. The values predicted using Raman modeling were compared with those calculated according to the UPLC-based amino acid method, as Figure 14A shown. The predicted values and average model errors for this primary histidine / arginine Raman modeling are given in Table 6. In Figure 14BIn it, a dot plot of the predicted histidine of mAb 14 against the actual histidine model is shown, with an average model error of 8.2%, meeting the target of the buffer excipient ≤ 10%. The target of ≤ 10% is based on the assay variability of the current UPLC orthogonal method. In Figure 14C In it, a dot plot of the predicted arginine of mAb 14 against the actual arginine model is shown, with an average model error of 2.9%, meeting the target of ≤ 10%.

[0161] This data indicates that Raman modeling can be used to predict the buffer excipient levels in UF / DF and FCP materials during the process. The successful quantification of these excipients can ensure that UF / DF provides a final concentrate pool that will enable subsequent formulation.

[0162] Table 6. Model components and histidine / arginine data collected from the processing of the general concentration model (Example 1).

[0163] Histidine Arginine <![CDATA[Spectral region (cm -1 )]]> 1200-1480 970-1100，1300-1500 Pretreatment Technology First Derivative and SNV First Derivative and SNV <![CDATA[R 2 Y]]> 0.940 0.964 <![CDATA[Q 2 > 0.938 0.963 RMSEP 1.20 mM 6.19 mM Average Model Error 10.4％ 7.39％

[0164] Histidine range: 0 - 25 mM; Arginine range: 0 - 81 mM

[0165] SNV - Standard Normal Variate - Mean centered and normalized

[0166] R 2 - Percentage of variance in the training set explained by the model, R 2 >0.9

[0167] Q 2 - Percentage of variance in the training set predicted by the model during cross - validation,

[0168] Q 2 >0.8

[0169] (RMSEP) Root Mean Square Error of Prediction

[0170] Example 9: Raman Model for Drug / antibody Ratio Measurement.

[0171] Materials and Methods :

[0172] DAR is a quality attribute that is monitored during the development of antibody-drug conjugates (ADCs), antibody-radionuclide conjugates (ARCs), and common protein conjugates (potent steroids, non-cytotoxic payloads, etc.) to ensure consistent product quality and to facilitate subsequent labeling with the payload. Raman was evaluated as a technique for monitoring DAR levels, which could then be used as a control strategy for the reaction. The DAR assay feasibility of two different mAbs under development (mAb 1 and mAb 3) was evaluated using Raman. Using a non-contact optical probe, the operating parameters of the Raman analyzer were set to a 10-second scan time with 10 accumulations. Using a spectral range of 350 - 3100 cm -1 the raw spectral data was preprocessed with SNV, and additionally filtered using a second derivative combined with 21 cm -1 smoothing. Table 7 shows the model components used to determine the DAR in the sample. Offline DAR measurements were determined using a UV spectroscopy-based method.

[0173] Table 7. Model components for drug / antibody ratio measurement.

[0174]

[0175] Results :

[0176] Figures 15A to 15B It was shown that Raman modeling could be used to measure the drug / antibody ratio of the iPET drug conjugates of mAb 1 and mAb 3. For both models, the root mean square error of cross-validation was 0.6 DAR. The current UV-based orthogonal assay has a variability (one standard deviation) associated with 0.3 DAR. This initial Raman prediction was within two standard deviations and indicated that DAR could be successfully predicted by Raman with some further refinement.

[0177] Although the present invention has been described in connection with certain embodiments thereof in the foregoing specification and numerous details have been set forth for purposes of illustration, it will be apparent to those skilled in the art that the present invention is susceptible to additional embodiments and that certain of the details described herein can be considerably altered without departing from the basic principles of the invention.

[0178] All references cited herein are hereby incorporated by reference in their entirety. The present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof, and accordingly, reference should be made to the appended claims rather than to the foregoing specification to indicate the scope of the present invention.

Claims

1. A method for preparing a concentrated protein purification intermediate, comprising: Providing a general model generated from a plurality of protein purification intermediates; During concentration of the protein purification intermediate, using in-situ Raman spectroscopy to measure the concentration of the protein purification intermediate in real time using the general model; And Adjusting the parameters of the concentration step in real time to obtain a concentrated protein purification intermediate and / or a final concentrate pool, and Wherein the concentration of the protein purification intermediate is at least 150 mg / mL, Wherein the general model is generated by independently performing Raman spectroscopy analysis on a plurality of protein purification intermediates and using partial least squares regression analysis of the original spectral data and offline protein concentration data, Among them, spectral data is collected in the wavenumber range selected from 977 - 1027 cm -1 , 1408 - 1485 cm -1 , 1621 - 1711 cm -1 , 2823 - 3046 cm -1 , and the spectral data is subjected to first derivative processing with 21 cm -1 point smoothing, or Among them, spectral data is collected in the wavenumber range selected from 2823 - 3046 cm -1 and the standard normal variate processing is adopted for the spectral data.

2. The method according to claim 1, wherein the protein purification intermediate has a concentration of 150 mg / mL to 300 mg / mL.

3. The method according to claim 1 or 2, wherein the protein purification intermediate is concentrated using ultrafiltration, buffer exchange, or both.

4. The method according to claim 1 or 2, wherein the protein purification intermediate is harvested from a bioreactor, fed-batch culture, or continuous culture.

5. The method according to claim 1 or 2, wherein the concentration of the protein purification intermediate is measured continuously or intermittently in real time.

6. The method according to claim 1 or 2, wherein the protein concentration is quantified at intervals of 30 seconds to 10 minutes, once per hour, or once per day.

7. The method according to claim 1 or 2, wherein the protein purification intermediate is an antibody or an antigen-binding fragment thereof, a fusion protein, or a recombinant protein.

8. The method according to claim 1, further comprising: Independently performing Raman spectroscopy analysis on a plurality of protein purification intermediates to generate the general model, wherein the general model is capable of quantifying primary concentration and / or permeation of any one of the plurality of protein purification intermediates at a concentration of 0 - 120 g / L in an ultrafiltration / diafiltration (UF / DF) system, and / or final concentration at a concentration greater than 200 g / L.

9. The method according to claim 1, further comprising: During purification of the harvested cell culture fluid or protein purification intermediate, using in-situ Raman spectroscopy to measure the concentration of excipients in real time; And Adjusting the parameters of the purification step in real time to obtain or maintain a predetermined amount of excipients in the harvested cell culture fluid and / or protein purification intermediate.

10. The method according to claim 1, further comprising: Quantifying one or more critical quality attributes in the protein purification intermediate using in-situ Raman spectroscopy; And Adjusting the one or more critical quality attributes in the protein purification intermediate to match a predetermined critical quality attribute level, Wherein the critical quality attributes are selected from the group consisting of antibody titer, protein concentration, high molecular weight substances, drug / antibody ratio, and buffer excipients.

11. A method for preparing a protein purification intermediate, comprising: Raman spectroscopy analysis is independently performed on multiple protein purification intermediates to generate a general model capable of quantifying any one of the multiple protein purification intermediates; During concentration of the protein purification intermediate, in-situ Raman spectroscopy is used to determine the concentration of the protein purification intermediate using the general model; And When the concentration of the protein purification intermediate or the final concentrated pool reaches a predetermined concentration, the concentrated protein purification intermediate is prepared, and wherein the predetermined concentration of the protein purification intermediate is at least 150 mg / mL, wherein the model is generated by partial least squares regression analysis using raw spectral data and offline protein concentration data, Among them, spectral data is collected in the wavenumber range selected from 977 - 1027 cm -1 , 1408 - 1485 cm -1 , 1621 - 1711 cm -1 , 2823 - 3046 cm -1 , and the spectral data is subjected to first derivative processing with 21 cm -1 point smoothing, or Among them, spectral data is collected in the wavenumber range selected from 2823 - 3046 cm -1 and the standard normal variate transformation is applied to the spectral data.

12. The method according to claim 11, wherein compared with the offline protein concentration value, the model provides a predicted protein concentration value with an error of ≤5%.

13. The method according to claim 11, wherein compared with the offline protein concentration value, the model provides a predicted protein concentration value with an error of ≤3%.

14. The method according to claim 11, wherein the concentrated protein purification intermediate has a concentration of 150 mg / mL to 300 mg / mL.

15. The method according to claim 11, wherein the concentrated protein purification intermediate is concentrated using ultrafiltration, diafiltration, or both.

16. The method according to claim 11, wherein the concentrated protein purification intermediate is harvested from a bioreactor, fed-batch culture, or continuous culture.

17. The method according to claim 11, wherein determination of the concentration of the protein purification intermediate is performed continuously or intermittently in real time.

18. The method according to claim 11, wherein the protein concentration quantification is performed at intervals of 30 seconds to 10 minutes, once per hour, or once per day.

19. The method according to claim 11, wherein the protein purification intermediate is an antibody or an antigen-binding fragment thereof, a fusion protein, or a recombinant protein.

20. The method according to claim 1 or 11, wherein the protein purification intermediate comprises an extracellular domain of a cell surface receptor or a fragment thereof.

21. The method according to claim 1 or 11, wherein the protein purification intermediate comprises an Fc fusion protein or a fragment thereof.

22. The method according to claim 1 or 11, wherein the protein purification intermediate comprises an anti-VEGF antibody or a fragment thereof.

Images