Patent Application
20130331554
1. Field of the Invention
The present invention relates to improved methods in the separation of biological molecules from complex mixtures. More specifically, the invention relates to improved methods for selectively increasing the homogeneity of biological molecules obtained from a heterogeneous mixture of molecules, particularly wherein such methods are used for large scale preparation and manufacturing processes.
2. Background Art
The present invention provides improved methods in the purification of biological molecules compared to other previously disclosed methods. Some examples of such other previously disclosed methods may be found in: Brown et al., WO 2006/099308 (PCT/US2006/008919), “A Method of Weak Partitioning Chromatography,” published Sep. 21, 2006; Pliura, et al., U.S. Pat. No. 5,439,591, “Displacement Chromatography Process,” issued Aug. 8, 1995; Pliura, et al., U.S. Pat. No. 5,545,328, “Purification of Hemoglobin by Displacement Chromatography,” issued Aug. 13, 1996; Kelley, et al., “Weak partitioning chromatography for anion exchange purification of monoclonal antibodies,” Biotechnol. Bioeng. 101(3):553-66 (2008); Brown, et al., “Overloading ion-exchange membranes as a purification step for monoclonal antibodies,” Biotechnol. Appl. Biochem. 56:59-70 (2010).
The present invention differs from previously described methods, at least in part, because previous methods describe a flow through mode of operation wherein impurities and undesirable product forms (or “isoforms”) bind to an adsorbent and the desired product is collected in the column flow through.
The present invention is an improvement over previously described methods, at least in part, because the present method utilizes an “overload bind and elute” mode of operation wherein a biological product is allowed to contact a chromatography medium (or other matrix) at a concentration or in an amount which exceeds the static or the dynamic binding capacity of the chromatography medium (or other matrix). This constitutes the “overload and bind” portion of the method of the present invention. During the overload and bind step, biological product having a selected characteristic (such as a high overall net-negative charge or a high sialic acid content) preferentially binds to the chromatography medium (or other matrix) while biological product (as well as other impurities) not having the selected characteristic, or having less of the selected characteristic (such as biological product having a lower overall net negative charge or a lower sialic acid content) is excluded or separated from the medium (or matrix). Subsequent to the overload and bind step, the bound target product is eluted (or otherwise dissociated or separated) from the chromatography medium (or other matrix) and recovered. Hence, the biological product mixture obtained has been enriched with a higher concentration of product having the selected (target) characteristic compared to the product mixture prior to application of the overload bind and elute purification step.
Methods of the present invention can be adapted and applied to the separation/purification of biological products based on any number of physical, biological, and/or chemical characteristics. For example, product isoforms may be selectively separated on the basis of charge and/or hydrophobicity by using appropriate adsorbents (such as, for example, strong or weak anion or cation exchange resins for charge based separations and hydrophobic adsorbents for separations based on hydrophobicity). Additionally, methods of the invention may be applied using mixed-mode chromatography (mixed-mode media) for separations based on two orthogonal product attributes (e.g., charge and hydrophobicity).
In one embodiment, product may be selectively enriched/separated, for example, based on peak product pI values wherein, for example, higher pI isoforms may be separated from lower pI, deamidated product isoforms on a cation exchange adsorbent.
In one embodiment, the present invention is useful for selectively enriching biological product isoforms (or “glycoforms”) wherein the selected or desired product characteristic is that of having increased or enhanced overall (total) levels of sialic acid content. In one embodiment, biological product with increased, total sialic acid content is obtained by overloading an anion exchange chromatography medium (e.g., TMAE HiCap) with a mixture of the biological product such that the concentration of total product exceeds the binding capacity, or the dynamic binding capacity, of the chromatography medium. Overloaded and undesired product (e.g., product with a lower sialic acid content, and other impurities) are allowed to flow-through the chromatography medium, then the selected, bound product (with high sialic acid content) is eluted (or otherwise separated or dissociated from the chromatography medium).
Embodiments of the invention are useful for obtaining highly homogeneous mixtures of a wide variety of biological products. Some examples of such biological products include, without limitation, proteins and protein fragments (i.e., full-length and partial length polypeptides/peptides), antibodies (immunoglobulins), heterologous fusion proteins, etc.
In one embodiment methods of the invention may be used for separation/purification of non-immunoglobulin proteins (or fragments thereof) fused with immunoglobulins (or domains, regions, of fragments thereof). In one embodiment, for example, methods of the invention are used for separation/purification of a fusion protein comprising an extracellular receptor ligand-binding domain linked (i.e., “fused”) with the Fc-region of an immunoglobulin (such as the Fc region of an IgG molecule). Antibody fusion proteins (e.g., Fc-fusion proteins) to which methods of the present invention may be applied are well known in the art, see for example, “Antibody Fusion Proteins,” edited by Steven M. Charnow & Avi Ashkenazi, Wiley-Liss, Inc., USA (1999) (ISBN 0-471-18358-X); and “Soluble Fc Fusion Proteins for Biomedical Research” by Meg. L. Flanagan et al. in “Methods in Molecular Biology: Monoclonal Antibodies: Methods and Protocols,” 378: 33-52 (2007), edited by M. Albitar, Humana Press Inc. USA (DOI: 10.1007/978-1-59745-323-3—3). Some specific examples, without limitation, of Fc-fusion proteins to which methods of the invention may be applied include those such as described for example in Fung et al., U.S. Pat. No. 7,294,481 (issued Nov. 13, 2007); Drapeau et al., U.S. Pat. No. 7,300,773 (issued Nov. 27, 2007); and, Ryll et al., U.S. Pat. No. 6,528,286 (issued Mar. 4, 2003).
Embodiments of the invention include, without limitation, a method for enhancing or increasing the concentration of biological product in a final mixture, wherein said biological product has one or more selected characteristics, wherein said method comprises:
(a) allowing an initial mixture of biological products with and without said selected characteristics to contact a chromatography medium wherein the quantity of biological products in said initial mixture exceeds the binding capacity or the dynamic binding capacity of said chromatography medium;
(b) allowing biological product not having said one or more selected characteristics to dissociate or be separated from said chromatography medium; and
(c) recovering a final mixture of biological products from said chromatography medium wherein said final mixture comprises an enhanced or increased concentration of biological product with one or more selected characteristics, compared to the concentration of biological product in said initial mixture.
Embodiments of the invention include use of any known, or subsequently disclosed or developed, chromatography media or matrix. Examples, without limitation, of such media comprise: ion exchange media; anion exchange media; cation exchange media; hydroxyapatite media; hydrophobic interaction chromatography media; antibody-affinity media (e.g., Protein-A or variants thereof); immunoglobulin Fc-region affinity media (e.g., Fc-receptor affinity media); and, ligand-affinity media; receptor-affinity media; and mixed-mode media.
Embodiments of the invention include methods wherein the binding capacity or dynamic binding capacity of a chromatography medium is exceeded by: 10% or more; 20% or more; 30% or more; 40% or more; 50% or more; 100% or more; 200% or more; 500% or more; and 1000% or more.
Embodiments of the invention include methods wherein the binding capacity or dynamic binding capacity of said chromatography medium is exceeded by: 1.5-fold or more; 2-fold or more; 3-fold or more; 4-fold or more; 5-fold or more; 6-fold or more; 7-fold or more; 8-fold or more; 9-fold or more; 10-fold or more; 20-fold or more; 30-fold or more; 40-fold or more; 50-fold or more; 100-fold or more; and 500-fold or more.
Embodiments of the invention include methods wherein the amount of biological products recovered in the final mixture, compared to the amount of biological products in the initial mixture, is: about 10% to about 80% recovered; about 20% to about 60% recovered; about 30% to about 60% recovered; about 30% to about 50% recovered; about 35% to about 50% recovered; about 35% to about 45% recovered; about 40% to about 45% recovered; about 40% to about 50% recovered; about 45% to about 50% recovered; about 10% recovered; about 15% recovered; about 20% recovered; about 25% recovered; about 30% recovered; about 35% recovered; about 40% recovered; about 45% recovered; about 50% recovered; about 55% recovered; about 60% recovered; about 65% recovered; about 70% recovered; about 75% recovered; and about 80% recovered.
Embodiments of the invention include methods wherein the concentration of biological product with one or more selected characteristics is increased or enhanced, compared to the initial mixture of biological products, by: at least about 5%; at least about 10%; at least about 20%; at least about 30%; at least about 40%; at least about 50%; at least about 60%; at least about 70%; at least about 80%; and at least about 90%.
Embodiments of the invention include methods wherein the selected characteristic (or characteristics) comprises any one or more of: degree of net negative charge at a set pH value; degree of net positive charge at a set pH value; degree of hydrophobicity; degree of hydrophilicity; quantity and/or type of carbohydrate content; quantity and/or type of N-linked glycosylation content; quantity and/or type of O-linked glycosylation content; total sialic acid content.
Embodiments of the invention include methods wherein the biological product comprises: a protein; an antibody or fragment thereof; a polypeptide comprising an extracellular receptor ligand-binding domain; a receptor ligand; a heterologous fusion protein; a fusion protein comprising an immunoglobulin Fc-region; a fusion protein comprising an extracellular receptor ligand-binding domain and an immunoglobulin Fc-region.
Embodiments of the invention include methods for recovering a selected biological product at a manufacturing scale; including wherein the selected biological product is a therapeutically useful or beneficial compound.
Dynamic Binding Capacity (DBC):
The dynamic binding capacity of a chromatography media is the amount of target product the media will bind under actual flow conditions before significant breakthrough of unbound target product occurs. As this parameter reflects the impact of mass transfer limitations that may occur as flow rate is increased, it is much more useful in predicting real process performance than a simple determination of saturated or static binding capacity. See, Millipore Corp., Technical Brief, Lit. No. TB1175EN00 (2005). In general the lower the flow rates, the higher the dynamic capacity. Dynamic binding capacities are routinely determined by those of ordinary skill in the art. For example, DBC can be determined by loading a sample containing a known concentration of the target product, and monitoring for the product in the column flow-through while applying the sample. See e.g., Bioseparation and Bioprocessing, Vol. 1, Sect. 1.4.3 (edited by Ganapathy Subramanian, published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany) 2nd Ed.; 2007.
Binding Capacity/Static Binding Capacity:
Amount of target product the media will bind under static (non-flow through) conditions.
Overload/Overloading/Overloaded:
Exceeding the binding capacity or dynamic binding capacity of an adorbent or other affinity or product capturing medium or resin.
By applying knowledge within the skill of those in the art, embodiments of the invention may be modified and/or adapted for various applications, without undue experimentation, without departing from the general concept of the present invention. Therefore, such adaptations and modifications are intended to be within the meaning and range of equivalents of the disclosed embodiments, based on the teaching and guidance presented herein.
A chromatography process was developed for the isolation/purification/enrichment of biological products with enhanced or increased levels of sialylation compared to biological products with decreased or lower levels of sialylation present in the same initial mixture.
The biological product utilized in the present example was an Fc-fusion protein comprising an extracellular receptor ligand-binding domain linked to the Fc domain of human a human immunoglobulin (hereinafter “ExcR-Fc”). The biological product has an apparent molecular weight of 130-150 kDa and a theoretical pI of 7.15. It was produced by recombinant DNA technology in a Chinese hamster ovary (CHO) cell mammalian expression system. The product has a complex carbohydrate profile with multiple potential N-glycans in the extracellular domain and Fc region. Additionally, the product has multiple potential O-glycans in the extracellular domain. These multiple potential sites for post-translational modification result in a heterogeneous product pool containing molecules with varying degrees of N- and O-glycan occupancy and varying levels of sialylation.
Highly sialylated forms of biological products can represent a highly desirable class of therapeutically advantageous protein isoforms. The present example describes development of a robust process capable of enriching the content of recovered/isolated, highly sialylated forms of biological product while maintaining acceptable product recovery and yield. In this particular example, a process was developed wherein TMAE HiCap chromatography was used to recover/isolate highly sialylated forms of an Fc-fusion protein.
Initially, a number of different adsorbents and chromatographic modes (i.e., CEX, AEX, HIC, and cHT) were screened for their ability to separate protein isoforms on the basis of the differences in the degree of sialylation. TMAE HiCap was selected because it provided improved resolution of product glycoforms and higher product yield as compared to the other tested adsorbents. Breakthrough experiments were performed on TMAE HiCap to measure the dynamic binding capacity of ExcR-Fc and the partitioning of glycoforms in the flow through mode. Based on these results, an “overloaded” bind and elute process was developed wherein more highly sialylated glycoforms were preferentially adsorbed and concentrated on the TMAE HiCap adsorbent, while the lower sialylated and non-sialylated glycoforms of ExcR-Fc flowed through the column. Subsequently, the column was eluted to recover the enriched higher sialylated glycoforms in the product pool.
Bind and elute experiments were performed on the TMAE HiCap column to optimize loading conditions to maximize enhancement of product sialylation and recovery. A Design of Experiments (DoE) approach was employed to evaluate process robustness and to identify the effect of the key process parameters on product quality and yield. Additionally, purification runs were performed under exemplary “worst-case” conditions to test the ability of the process to consistently meet desired product quality conditions.
Chromatography Adsorbent
Columns
Equipment
The following sections provide a detailed description of experiments performed to develop and optimize a chromatography process for enrichment of the overall level of sialylation of ExcR-Fc product. Early development experiments (Sections 4.1-4.3) employed enzyme-based desialylation followed by fluorescent labeling of the released sialic acid and fluorimetric analysis for the quantitation of the TSA content (i.e., DMB method). However, this assay had low precision and sensitivity. Therefore, a new, more precise TSA assay that employed acid hydrolysis for desialylation followed by UV-HPLC analysis of the released sialic acids was developed and employed for subsequent development studies (Sections 4.4 and 4.5). In each analysis, a control sample was run as an internal assay control which allowed for the determination of relative TSA enhancement provided by the process steps and/or experimental conditions being evaluated in the study.
At the onset of development, linear gradient experiments were performed on cation exchange (SE HiCap), anion exchange (TMAE HiCap and Capto DEAE), and hydroxyapatite (cHT Type I) chromatography to identify chromatographic modes and adsorbents that provided good resolution for product glycoforms on the basis of the level of sialylation. These experiments used Phenyl Sepharose 6FF column eluate obtained from an early pilot-scale run (PS4) diafiltered into neutralized Prosep Ultra Plus elution buffer (i.e., 75 mM Acetate, pH 3.3 neutralized to pH 5.0 using 2.4 M Tris base).
Additionally, hydrophobic interaction chromatography on Phenyl Sepharose 6FF (high sub) was also evaluated for its ability to enrich the level of sialylation of ExcR-Fc. Hence, eluate samples from experiments performed during the development of a Phenyl Sepharose chromatography process were also assessed for Total Sialic Acid (TSA) analysis.
Once TMAE HiCap was selected as the adsorbent for further process development, additional linear gradient experiments were performed to evaluate the effect of pH (i.e., pH 5.5, 6.5, 7.5, and 8.5) on the resolution of various product glycoforms. These experiments used biological product from a pilot-scale run (PS10) diafiltered into 50 mM Acetate, pH 5.0 solution as the starting material. The following method was employed with the column effluent monitored at 280, 300 and 313 nm. Experiments were performed at a linear flow velocity of 150 cm/hr and all steps were operated in downflow.
Experiments were performed to evaluate the partitioning of glycoforms as well as the dynamic binding capacity (DBC) on TMAE HiCap as a function of load pH i.e., pH 5.5, 6.5, 7.5, and 8.5. These experiments employed biological product from a pilot-scale run (PS10) that was diafiltered into 50 mM Acetate, pH 5.0, diluted to approximately 5 mg/mL with DF buffer, and adjusted to the appropriate pH with 2.4 M Tris base. For each load pH condition, the TMAE HiCap column was loaded to 150 mg/mL and flow through fractions were collected in 10 mg/mL increments and analyzed for TSA levels. The following method was employed with the column effluent monitored at 280, 300 and 313 nm. Experiments were performed at a linear flow velocity of 150 cm/hr and all steps were operated in downflow.
Based on the results of the partitioning experiments, preliminary bind and elute experiments were performed to determine the feasibility of using TMAE HiCap in overloaded mode for the purpose of enriching sialylated product glycoforms. These experiments employed PS10 biological product diafiltered into 50 mM Acetate, pH 5.0, diluted to approximately 5 mg/mL with DF buffer, and adjusted to pH 5.5 with 2.4 M Tris base. Column loadings of 100 and 150 mg/mL were examined and the corresponding eluate pools were analyzed by the TSA and SEC assays. The following method was employed with the column effluent monitored at 280, 300 and 313 nm. Experiments were performed at a linear flow velocity of 150 cm/hr and all steps were operated in downflow.
To evaluate the option of employing diluted biological product as the column load material without the need for a prior UF/DF step, overloaded bind and elute experiments were carried out with diluted biological product to evaluate the effect of load composition on column performance. To this end, PS10 biological product adjusted to pH 5.5 was diluted (i) 5× with WFI, (ii) 10× with WFI, and (iii) 5× with WFI followed by 2× dilution with 50 mM Acetate, pH 5.5. A bind and elute run at 100 mg/mL loading was performed for each load preparation using the method described above.
A Design of Experiments (DoE) study was performed to identify suitable operating conditions for an overloaded bind and elute TMAE HiCap step to provide the desired level of enrichment of product sialylation while maintaining acceptable product yield. The key process parameters and their ranges investigated were: Column loading (90-160 mg/mL), load pH (5.2-5.8), and load conductivity (2.4-4.4 mS/cm). A central composite design comprising of 18 runs (including 4 center point experiments) was created using the Stat-Ease Design Expert v8.0 (Stat-Ease, Inc., Minneapolis, Minn.) software. For each DOE run, operating parameters were varied as per the DoE experimental design table (Table 1).
Load material employed for the DoE experiments was generated from biological product obtained from representative a prototype run (PS 14) diafiltered into 50 mM Acetate, pH 5.5 and diluted to approximately 5 mg/mL with DF buffer. For each DoE run, the load was adjusted to the desired pH by titrating with 1M Bis Tris or 2N acetic acid, while the load conductivity was adjusted using WFI or a 500 mM acetate, pH 5.5 solution.
The two key process outputs—TSA and % HMW—were measured for each eluate pool and the software was employed to analyze the data and build models for each response. In addition, process recovery was also analyzed and modeled. The chromatography method described above was employed for these DoE experiments.
Additional single-point bind and elute experiments (Table 2) were performed on TMAE HiCap in order to:
PS10 and PS16 biological product separately diafiltered into 50 mM Acetate, pH 5.5 was employed for these experiments. For each run, the load was adjusted to the desired pH by titrating with 1M Bis Tris or 2N acetic acid, while the concentration and load conductivity were adjusted using WFI and/or 500 mM acetate, pH 5.5 solution. The following method was employed with the column effluent monitored at 280, 300 and 313 nm. Experiments were performed at a linear flow velocity of 150 cm/hr and all steps were operated in downflow.
Subsequently, additional single-point experiments were performed to further optimize the load concentration of the TMAE HiCap process step. Load concentrations between 2-30 mg/mL were evaluated. PS16 biological product diafiltered into 55 mM Acetate, pH 5.3 was employed as the starting material for these experiments. The diafiltered pool was adjusted to worst-case load pH and conductivity conditions for HMW (i.e., pH 5.1 and 3.9 mS/cm) and diluted to the appropriate target load concentration using a 55 mM Acetate adjusted to pH 5.1 and 3.9 mS/cm. Experiments were performed on a 1.66 mL (0.46×10 cm) screening column at a linear flow velocity of 100 cm/hr using the chromatography method described above. Note: The velocity for these runs was reduced in order to match the residence time on the 15 cm bed height development column.
Linear gradient experiments were performed to screen various chromatographic modes and adsorbents in order to identify a suitable chromatography step for the enrichment of product sialylation. To this end, cation exchange (SE HiCap), anion exchange (TMAE HiCap and Capto DEAE), hydroxyapatite (cHT Type I), and HIC (Phenyl Sepharose 6FF) adsorbents were evaluated. Results of the TSA analysis performed for the eluate fractions for each run are shown below. These results clearly showed that TMAE HiCap provided the highest amount of SA enrichment at relatively higher process yield as compared to the other adsorbents evaluated in this study.
4.1.1 Fractogel SE HiCap
4.1.2 Fractogel TMAE HiCap
The linear gradient chromatogram on TMAE HiCap is shown in
4.1.3 Capto DEAE
4.1.4 cHT Type I
The linear gradient chromatogram for the experiment performed on cHT Type I is shown in
4.1.5 Phenyl Sepharose 6FF (high-sub)
Eluate pools generated using different elution conditions between 400-700 mM ammonium sulfate during the course of development of the Phenyl Sepharose column as the third step in the ExcR-Fc purification process were analyzed by the TSA assay. As shown in
4.1.6 Effect of pH on Glycoform Separation
Additional linear gradient experiments were performed to evaluate the effect of pH on the separation of the glycoforms of ExcR-Fc on TMAE HiCap. To this end, experiments were performed at pH 5.5, 6.5, 7.5, and 8.5 and the eluate fractions analyzed by TSA analysis.
In addition, the eluate samples were also analyzed for HMW by SEC (
Breakthrough experiments were performed at different load pH conditions to evaluate the partitioning of differently sialylated product glycoforms on TMAE HiCap adsorbent.
TSA results for the column flow through fractions for the runs at pH 5.5 and 6.5 showed that higher sialylated product forms, being more negatively charged, had a greater binding affinity to the TMAE HiCap adsorbent compared to the lower sialylated and non-sialylated species. Based on the product concentrations and TSA content of the column load and flow through fractions, a mass balance was performed to determine the mass (i.e., recovery) and TSA of the adsorbed product as a function of column loading (
Initial proof-of-concept experiments were performed to evaluate the ability of the TMAE HiCap column step to increase product sialylation by operating in the overloaded bind and elute modes under the loading conditions identified in the above breakthrough/partitioning studies i.e., pH 5.5/˜3.2 mS/cm. The results of the single point experiments performed at 100 and 150 mg/mL column loadings are summarized in Table 4. The results showed that when operated in this mode, the TMAE HiCap provided significant enrichment in sialylation, with the TSA levels being increased from 11.3 mol SA/mol ExcR-Fc in the load to 14.3-15.4 mol/mol in the eluate pools. In addition, eluate TSA levels increased with increasing loading with a concomitant drop in recovery due to higher product losses in the column flow through. Thus, the ability of the TMAE HiCap step to match the level of control sample sialylation (i.e., 14.6-15.2 mol/mol)) was successfully demonstrated. SEC analysis revealed that HMW levels were slightly increased from 2.2% in the load to 2.9-3.1% in the eluate pools. However, HMW levels in the column eluate were below the action limit of ≦3.5% for the release of biological product.
All previous TMAE HiCap development experiments employed a diafiltration step performed in a UF/DF system in order to prepare the TMAE HiCap load material from ExcR-Fc biological product generated in pilot-scale. In order to reduce the number of unit operations required for the reprocessing of the biological product to enhance sialylation, experiments were performed to evaluate the feasibility of employing diluted biological product as the column load material. As shown in
The experiment performed with biological product diluted 5× with WFI followed by a 2× dilution with buffer (50 mM acetate, pH 5.5) showed more typical breakthrough behavior (chromatogram not shown, 6). In addition, the TSA enrichment (relative to the load TSA) obtained in this case was comparable to that obtained with diafiltered biological product at comparable column loading and load pH and conductivity (i.e., 3.1 mol/mol enrichment, Table 5 vs. 3.0 mol/mol enrichment, Table 4). However, greater product binding and consequently lower yield was obtained in this case relative to the run performed using diafiltered biological product (i.e., 56%, Table 5 vs. 65%, Table 4). Thus, the above results clearly demonstrate the need for the UF/DF step for diafiltering the ExcR-Fc biological product into the appropriate load buffer to ensure better process control and higher product recovery from the TMAE HiCap step.
4.4 Design of Experiments (DoE) Study
Table 6 shows the results of the DoE experiments performed to determine suitable conditions for the operation of the TMAE HiCap overloaded bind and elute chromatography process. TSA, HMW, and recovery were measured and modeled as the key process outputs. Host cell protein (HCP) levels and product purity by reduced and non-reduced LC90/GXII were also monitored to ensure that all other purity targets were adequately controlled by the process. As shown in the table, the eluate pools of all DoE runs had comparable and high purity as well as very low levels of HCP i.e., at or near the assay LOQ.
Results from experimental runs were imported into Stat-Ease Design Expert v8.0 software and analyzed. Analysis of each output involved the selection of a mathematical model that described the effect of process parameters on that output. The data were fit by the best model equation and an analysis of variance (ANOVA) was performed to remove any insignificant parameters. This is outlined for each output in subsequent sections. The following methodology was utilized for analysis of each response:
The fit summary program was exploited to perform regression calculations to fit all of the polynomial models to the selected response. The fit summary program calculated the effects for all model terms and produced P-values, lack of fit, and R2 statistics in order to compare and select the best model(s). The model selected was the highest order polynomial where the additional terms were significant, the model was not aliased, lack of fit was minimized and adjusted and predicted R2 values were maximized.
Significant parameters were determined using analysis of variance for each response by removing input parameters one at a time via backward selection until only process variables remained if there was less than a 5% chance that their effect on a response could be due to noise alone (probability of a larger F-value [Prob>F]<0.05; p-value<0.05). These factors were deemed to be significant factors affecting the response. Factors were included in the model when P-values≧0.05 to support hierarchy.
Additional analysis of variance was examined to ensure that the model was significant and lack of fit was insignificant (p-values≧0.05) and to examine adjusted and predicted R2 values. A predicted R2 value of ≧0.5 was utilized to determine if a model was adequate for process modeling. The predicted and adjusted R2 values should be within 0.2 of each other.
Diagnostic tools were utilized to examine if the data contained outliers, non-normality or heteroskedasticity of residuals or required a transformation. After transformations were applied and outliers investigated, significant factors were identified and models were generated.
Table 7 summarizes results from the analysis of variance which identifies significant factors that affect each response (p-value<0.05) and examines model fit and prediction capabilities through sum of squared error calculations and lack of fit significance. No significant model or significant parameters were identified that affected HCP and % purity. The analysis of the individual responses can be found in Sections 5.4.1-5.4.6.
A model was produced for each response in the form of the following equation. Coefficients for each response are listed in Table 8.
Response=B0+A(Loading)+B(Load pH)+C(Load Conductivity)+D(Loading*Load pH)+E(Loading*Load Conductivity)+F(Load pH*Load Conductivity)+G(Loading)2
For the TSA, HMW, and Recovery responses, 99% prediction intervals were generated for a given combination of input parameters within the operating space to predict the corresponding reponse levels. For the reponses for which no model was identified i.e., HCP, Highest Single Impurity and Purity, three standard deviations around the mean value were utilized to predict response levels (Table 9).
4.4.1 Total Sialic Acid (TSA)
The enhancement of TSA is the key function of the TMAE HiCap chromatography step in the ExcR-Fc downstream purification process. Thus, the TSA level in the eluate pool was the key response for the design and modeling of this step. The Fit Summary of these data is shown in Tables 10-12. The sequential model sum of squares table (Table 10) shows the cumulative improvement in the model fit as higher order model terms are added. For example, Linear vs. Block shows the significance of the linear terms after accounting for the linear and block terms. Similarly, 2FI vs. Linear gives the significance of adding two-factor interaction terms to the model; Quadratic vs. 2FI indicates the significance of adding the quadratic terms to the linear, block, and the 2FI terms; and so forth. Each row in the table contains the statistics for additional terms only and not for the complete model.
The significance associated with the addition of model terms is calculated using the sum of squares (SS) for the model as well as the residual error. The sum of squares for all effects which were not included in the model were pooled together and used as an estimate of the residual error. The mean square error was calculated by dividing the sum of squares by the degrees of freedom (SS/df). The ratio of the mean squares (MSModel/MSResidual) was used to determine the F-value for the model, which was then used to compare the variance of the model with the variance of the residual error. The larger the F-value, the greater the likelihood of the model being significant. A quantitative measure of model significance can be obtained by comparing the F-value to known F-distribution tables for a given percentage risk. F-values greater than the critical value given on these tables are considered significant within the stated percentage risk. In addition, the Probability>F value (i.e., p-value) was calculated. This value is equal to the fractional area under the curve of the F-distribution that lies beyond the observed F-value. For example, a p-value of <0.05 indicates that there is 95% confidence that the model is significant and not due to experimental noise. In general, model terms having a p-value of <0.05 are deemed to be significant. The highest order model where the additional terms were significant and where aliasing was not present was indicated in the table as the “suggested” model. Analysis of the TSA data suggested that a quadratic model best described this response (Table 10).
The second table compiled in the fit summary is the lack of fit tests table (Table 11). This table compares how well each model fits the data. The lack of fit test compares the residual error to the pure error from replicated design points. A lack of fit error significantly larger than the pure error indicates that there are experimental points that differ significantly from the values predicted by the model. Accordingly, a more appropriate model should be used fit to the data. The p-value is used to measure the lack of fit. The higher the p-value, the greater the likelihood that the difference between the experimental and model points is due to experimental noise. In general, a p-value greater than 0.1 is desired. For the TSA response, the quadratic model had the highest p-value for the non-aliased models, suggesting it had the best fit to the data.
Table 12 contains the model summary statistics for the TSA response. The R-squared, adjusted R-squared, predicted R-squared, and the PRESS statistic for each complete model type are shown. As shown in the table, the quadratic model had the highest R-squared and adjusted R-squared values, which indicated that it best fit the experimental data.
The quadratic model for TSA was then reduced to remove any model terms that were not significant. This was done by using a backward selection process, involving the calculation of the ANOVA (analysis of variance) for the model, removing the least significant model term, calculating the ANOVA for the model with the term removed, and again removing the least significant term. This process was repeated until all model terms having a p-value>0.05 were removed from the model. The ANOVA results for the eluate TSA model is shown in Table 13.
The ANOVA results showed that the model for TSA had high R2 values (>0.9) indicating that the model captured more than 90% of the variance in the data. In addition, the Prediction R2 value was in close agreement with the Adjusted R2 value, suggesting that this model had good predictive ability. The Adequate Precision, which measures the signal to noise ratio, had a value of 44.28 indicating that the model had adequate signal and may be used to navigate the design space.
In order to validate the assumptions of the ANOVA (normal distribution of residuals, independence of errors, and constant variance) a number of diagnostic plots were evaluated. All residuals in the diagnostic plots were studentized (i.e., residuals were divided by their estimated standard deviation) for the purpose of comparison. A normal distribution plot (
The lack of trending in the TSA residuals vs. run number plot (
The graph of the predicted eluate TSA levels vs. the actual values enables the (
4.4.2 HMW
Previous development results showed that the TMAE HiCap eluate had higher HMW levels than the corresponding load material. This was expected due to the higher binding affinity of HMW relative to monomer, resulting in the accumulation and concentration of HMW on the adsorbent and correspondingly higher levels in the eluate. This behavior is analogous to that shown by the higher sialylated glycoforms and leads to the co-enrichment of HMW species. Thus, eluate HMW is another key process output that must be modeled and characterized and controlled (if required).
Analysis of HMW results for the TMAE HiCap eluate pools within the Design Expert software suggested that a linear model best described this response (tables not shown, 7). After applying the backward selection with alpha=0.05 to identify the significant factors, the ANOVA table for the model identified A, B, and C as significant factors (Table 14).
The lack of fit for the above model was insignificant and the model had high R2 and adjusted R2 values. Also, the predicted R2 value was in good agreement with the adjusted R2 indicating that the model had good predictive ability. Diagnostic plots revealed that the distribution of residuals was normal, the variance was constant, and there was no influence by lurking variables on the response, which confirmed that the ANOVA was valid (plots not shown, 7). Analysis of the model plots (not shown, 7) showed that HMW levels in the TMAE eluate increased with increasing column loading, decreasing load pH, and/or increasing load conductivity. Since the load material employed for these experiments had ˜0.9% HMW, the eluate HMW levels for all runs were well below the action limit of >3.5% HMW (Table 6). Thus there were additional single-point experiments performed to ensure that the eluate HMW was acceptable at the highest load HMW level and under worst-case operating conditions for the TMAE HiCap step (Sections 4.5 and 5.5).
4.4.3 Recovery
Analysis of the recovery data showed that this response was best described by a 2 factor-interaction (2FI) model (data not shown). After applying the backward selection with alpha=0.05 to identify the significant factors, the ANOVA table for the model identified B, C, AB, and AC as significant factors (Table 15). Additional details can be found elsewhere.
Overall, the model had high R2 and adjusted R2 values and the predicted R2 value was in good agreement with the adjusted R2 indicating that the model had good predictive ability. Although the lack of fit was not significant in the above model, the p-value was only slightly >0.05 i.e., there was a 5.16% probability that a lack of fit value this large could occur due to noise. Typically, a p-value>0.1 is desirable for insignificant lack of fit of the model. However, the relatively large lack of fit value might have been an artifact of the small pure error calculated from the replicates and not a true indication of any lack of fit between the experimental data and the model predictions. Diagnostic plots revealed that the distribution of residuals was normal, the variance was constant, and there was no influence by lurking variables on the response, which confirmed that the ANOVA was valid (data not shown). Analysis of the model plots (data not shown) showed that process recovery increased with decreasing loading, increasing load pH, and/or decreasing load conductivity.
4.4.4 Purity by Reduced and Non-Reduced LC90/GXII and Highest Single Impurity
A review of the results of the reduced and non-reduced LC90/GXII assays showed that the differences between the % Purity and % Highest Single Impurity (HIS) values for the DoE runs were within the range of variability of the respective assays. Therefore, the mean±3 standard deviations were determined to be the best predictor of these responses and no model was developed.
4.4.5 HCP
Analysis of HCP levels in the TMAE HiCap eluate pools showed that all values were very close to or below the limit of quantitation (LOQ) of the assay. Thus, no model was developed for this response.
4.4.6 DoE Study Summary
The predictive models developed for TSA and % Recovery were employed to identify suitable ranges for column loading and load pH and conductivity in order to achieve the desired level of TSA enhancement while maintaining acceptable product yield. Since lower SA enrichment is obtained at lower column loading, this condition represents the worst-case for product quality. Accordingly, as shown in the TSA model plot in
On the other hand, as aforementioned (Section 5.4.3), process recovery decreases with increasing column loading. Thus, there was a need to specify an upper range for column loading in Manufacturing to maintain reasonable process yields. Based on Manufacturing column sizing considerations as well as accounting for variability in the mass of starting material (i.e., ExcR-Fc biological product) and step yield of the upstream UF/DF step, a maximum loading of 180 mg/mL was determined for TMAE HiCap. As shown in the contour plot generated from the % Recovery model at the maximum loading condition, a minimum step yield of ˜36% was predicted. Overall, step yields between 36-59% can be expected for the TMAE HiCap process within the specified operating ranges for column loading and load pH and conductivity. Note: The load pH range was subsequently changed to 5.1-5.5 in order to provide a symmetrical operating range for Manufacturing i.e., 5.3±0.2. Based on the prediction plots below, this change is not expected to adversely impact process performance.
4.5 Additional Process Characterization
Additional single-point experiments were performed on the TMAE HiCap step within the operating ranges determined from the DoE study in order to evaluate the effects of other input parameters (e.g., load concentration and flow velocity) that were not examined in the earlier DoE study as well as to further assess process robustness. Results of these experiments are summarized in Table 16.
aSetpoint conditions for column loading, load pH and load conductivity
bWorst-case column loading, load pH and load conductivity for TSA, best-case for Recovery and HMW
cWorst-case column loading, load pH and load conductivity for Recovery and HMW, best-case for TSA
The DoE experiments discussed in Section 5.4 employed diafiltered biological product from PS14 as the load material for the TMAE HiCap column. In order to evaluate the robustness of the TMAE HiCap step in the context of variability in the load material, experiments (Runs 2, 5, 8) were performed using a different biological product lot (i.e., PS16). The results of these runs were used as test points for verifying the predictive ability of the DoE models. As shown in Table 17, the actual values for % Recovery and TSA were within the 99% prediction intervals of the corresponding models. Thus, these DoE models were successfully verified. However, the measured HMW values were higher than the predicted ranges. This was due to the higher HMW levels in the load material employed for the verification runs (1.5% HMW in PS16 load) compared to the material used for the DoE experiments (0.9% HMW in PS14 load). Overall, DoE models generated using one lot of load material were able to successfiffly predict TMAE HiCap process performance for a different lot of column load material. Thus, the robustness of the TMAE HiCap process against lot-to-lot variability in the load material can be inferred from these results. In addition, the adsorbent lot employed for these runs was shown to have comparable performance as the lot used for the process development. Accordingly, this adsorbent lot was deemed to be suitable for use in the 2011 2K GMP Manufacturing campaign.
aRepresents the 99% prediction intervals obtained from the corresponding models using the input parameters for each run
bThe column loading for this run was outside the design space of the DoE study. Thus, the predicted values were obtained by extrapolation of the models.
ExcR-Fc biological product produced during the 2010 2K GMP Manufacturing campaign had lower TSA levels than that observed for representative prototype runs (including PS 14 and 16). Therefore, an experiment was performed with PS10 material having TSA levels comparable to that observed for the GMP batches in order to determine the product quality obtained under worst-case conditions for TSA enrichment (Run 12). Analysis of the eluate pool for Run 10 showed that the TSA level was 16.0 mol/mol which was comparable to the TSA level measured for the control sample control i.e., 16.0-16.5 mol/mol. Thus, the TMAE HiCap process was shown to be capable of enriching the SA content of ExcR-Fc biological product to match the control sample even under these worst-case conditions.
In addition, HMW levels in representative prototype and GMP biological product lots were ≦1.6%. In order to evaluate the highest HMW level that can be expected in the TMAE HiCap eluate, Run 12 employed PS 10 load that had 2.0% HMW and was performed under worst-case operating conditions for HMW. As shown in Table 16, the eluate HMW level for Run 10 was 3.3%, which was less than the action limit of >3.5% specified for biological product. The use of GMP material with ≦1.6% HMW is expected to provide adequate safety margin for eluate HMW levels. Accordingly, the HMW increase observed over the TMAE HiCap step was deemed to be acceptable and no additional process steps were required in order to further reduce HMW levels in biological product.
The flow velocity of the TMAE HiCap process was varied between 75-250 cm/hr to evaluate the effect on product quality (Runs 1-3 and 11-13). The results for these runs showed that HMW levels in the eluate dropped with increasing flow velocity, while the product yield and TSA levels were comparable within the limits of assay variability. The observed HMW increase was likely due to the shorter residence time available for the larger HMW species to bind onto the column with increasing flow velocity. Thus, it is advantageous to operate at higher load velocity to minimize the HMW levels in the TMAE HiCap column eluate. Accordingly a flow velocity range of 150-250 cm/hr for the column load step was recommended for Manufacturing.
Runs 4-6 and 7-9 were performed to study the effect of load concentration on process performance. As shown in Table 16, HMW levels were observed to decrease as the load concentration increased from 3 mg/mL to 7 mg/mL. In addition, the decrease in eluate HMW levels was observed to be more significant as load concentration increased from 3 mg/mL to 5 mg/mL and compared to that observed for the increase in load concentration from 5 mg/mL to 7 mg/mL. At the same time, the process recovery and eluate TSA levels were comparable at the different load concentrations. Based on these results, additional single-point experiments were performed to determine the optimum load concentration to minimize the eluate HMW levels. As shown in
In order to further investigate this behavior, batch experiments were performed to measure the isotherms for monomer and HMW with the same load material that was used for the above experiments (9). The input load concentration for these experiments was varied between 2-45 mg/mL (Load HMW ˜1.6%), which represents the entire feasible operating range for the TMAE HiCap column.
A Phase I TMAE HiCap chromatography process was successfully developed for the enrichment of sialic acid (SA) levels in ExcR-Fc biological product. Initially, experiments were performed to evaluate the ability of different adsorbents and modes of chromatography to separate various ExcR-Fc isoforms on the basis of the differences in the degree of sialylation. To this end, cation exchange (SE HiCap), anion exchange (TMAE HiCap and Capto DEAE), hydrophobic interaction (Phenyl Sepharose), and hydroxyapatite (cHT Type I) chromatography were evaluated by linear gradient or step elution experiments. TMAE HiCap was selected for further development since it provided the highest degree of SA enhancement with the potential to match and even exceed the total sialic acid (TSA) levels of the control sample lots.
Subsequently, a novel “overloaded” bind and elute process was developed wherein the higher sialylated glycoforms having higher net negative charge and binding affinity to TMAE HiCap competed effectively for binding sites with the lower affinity lesser sialylated and non-sialylated glycoforms, thus displacing these lower affinity species. This resulted in the build-up of higher sialylated ExcR-Fc molecules on the adsorbent surface with increasing column loading. Subsequently, the column was eluted to recover the enriched higher sialylated glycoforms in the product pool, A DoE study was performed to optimize operating ranges for the key process parameters i.e., loading, load pH, and load conductivity. The results of these studies showed that the TSA levels in the TMAE HiCap eluate increased with increasing loading, decreasing load pH, and/or increasing load conductivity within the design space examined. On the other hand, product yield and eluate HMW levels showed the reverse trends within the same operating space. Accordingly, operating ranges were established for column loading and load pH and conductivity to ensure that product quality targets were achieved while maintaining reasonable process yields. In addition, single point experiments were also performed on TMAE HiCap to optimize the load concentration and flow velocity to improve product quality (specifically HMW levels) in the TMAE HiCap eluate.
Finally, single point experiments were performed to test the robustness of the TMAE HiCap process under worst-case operating conditions. These runs employed pilot-scale material having low TSA and high HMW levels (i.e., worst-case load PQ) to test the overall process capability. The results of these experiments clearly demonstrated that the TMAE HiCap process performed consistently, delivering product that met all pre-determined quality targets, even under worst-case operating conditions.
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/US11/60821 | 11/15/2011 | WO | 00 | 8/30/2013 |
| Number | Date | Country | |
|---|---|---|---|
| 61413908 | Nov 2010 | US |
