A raman spectroscopy method for monitoring and feedback control of perfusion cell culture

Abstract

The present disclosure provides a method of controlling a process attribute of a perfusion cell culture, the method comprising: generating an inline Raman spectra of a process attribute by taking multiple measurements of the concentration of the attribute, wherein the concentration is measured in a sample downstream of a perfusion bioreactor prior to purification of the cells in the culture, correlating the Raman spectra measurements with offline reference measurements of the concentration of the attribute, wherein the offline reference measurements were measured in one or more samples taken from the perfusion bioreactor, building a chemometrics model using the inline Rama spectra and the offline reference measurements, monitoring the concentration of the attribute downstream of the perfusion bioreactor; and using the chemometrics model to control the perfusion cell culture by maintaining the concentration of the process attribute.

Description

A RAMAN SPECTROSCOPY METHOD FOR MONITORING AND FEEDBACK CONTROLOF PERFUSION CELL CULTURECROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application Serial No. 63 / 582,885 filed September 15, 2023, the entire contents of which is incorporated by reference herein.FIELD

[0002] The present disclosure relates to perfusion cell culture systems using Raman spectroscopy for measurement in the permeate for perfusion monitoring and feedback control.BACKGROUND

[0003] Process analytical technology (PAT) is a mechanism to design, analyze, and control pharmaceutical manufacturing processes through the measurement of critical process parameters which affect product quality attributes including critical quality attributes (CQAs) that ensure final product quality. PAT allows for understanding of the manufacturing process by defining critical processing parameters (CPPs) and accordingly monitoring them in a timely manner (inline, at-line, or on-line) and adjusting process parameters if necessary, allowing for greater efficiency in testing, reducing over-processing, enhancing consistency, and minimizing rejected batches.

[0004] The objective for PAT implementation includes better process understanding, improved yield because of prevention of the reprocessing, reduction in the production cycle time by using on-line / at-line or in-line measurements and control, improved efficiency from conversion of the batch process into continuous processing, cost reduction, and real-time release of the batches.

[0005] At-line process analytical technology, including cell counters and biochemistry analyzers, is physically located next to the process and requires manual sampling. On-line process analytical technology, which includes aseptic sampling or on-line liquid chromatography, are physically connected directly to the process, and include automatic sampling where the sampling never returns to the process due to potential contamination risk to the purification batch.

[0006] In-line process analytical technology is physically in contact with or in the flow of the process. In-line process analytical technology requires no sampling and provides continuous data.Such examples of in-line process analytical technology includes capacitance, spectroscopy, light scattering, and sensors.

[0007] Perfusion cell culture has been gaining increasing popularity for biologies manufacturing due to benefits such as smaller footprint, increased productivity, consistent product quality and manufacturing flexibility, cost savings, etc. PAT tools are highly desirable for effective monitoring and control of long-running perfusion processes.

[0008] Perfusion cell cultures enable a more sustained and efficient cell grow th and protein production compared to traditional batch or fed-batch cultures. Perfusion cell cultures offer several advantages over traditional batch cultures, including higher cell densities, longer culture durations, and improved product yields. Unlike batch and fed-batch cultures, perfusion cell cultures can be run in steady state or pseudo-steady state mode where fluctuations in process parameters and nutrient profiles are minimized reducing cell stress and improving product quality7and yield.

[0009] There are some inherent challenges associated with the perfusion cell culture operation including a higher likelihood of process perturbation in long-term perfusion, particularly in intensified cultures w here the kinetics of metabolic changes are amplified due to high cell densities. It is, therefore, critical to monitor and control critical process parameters and critical quality attributes to ensure the process operates under the desired operating conditions. Process analytical technology also plays a critical role in perfusion cell culture by providing real-time measurement of certain CPPs and CQAs to enable efficient perfusion process monitoring and control.

[0010] Raman spectroscopy is an optical spectroscopy technique that provides a molecular fingerprint of a sample and is recognized as a process analytical technology' tool. Raman spectroscopy' has been w idely investigated for monitoring and control of traditional fed batch cell culture process. How ever, implementation of Raman spectroscopy for perfusion cell culture has been extremely limited due to challenges with high-cell density and long running time during perfusion which cause extremely7high fluorescence interference to Raman spectra and consequently, it is exceedingly difficult to develop robust chemometrics models.

[0011] Raman spectroscopy has been widely investigated for implementations in cell culture processes. Typically. Raman spectra are collected from inside bioreactor by placing a probe directly inside bioreactor. Raman data collected this way can be correlated with offline reference measurement to build chemometrics models for analytes of interest, such as glucose, lactate, and titer etc. This methodology7has worked well for traditional fed batch processes, which often last approximately tw o w eeks with ty pical peak cell density below730 million cells / ml. There havebeen numerous publications reporting successful implementation of inline Raman for monitoring and control of fed batch cell culture. However, it could be challenging when the same methodology is applied to perfusion cell culture process which typically operates at much higher cell density and runs much longer compared with fed batch processes.

[0012] Most cell culture bioprocesses use CHO cells fed by glucose. Glucose is a critical process parameter, as it affects the cell’s metabolic profile, production of waste products, and post-translational non-enzymatic glycosylation of proteins. CHO cells are typically fed in a batch, known as fed-batch where glucose is delivered into the bioreactor as a large bolus at set time points, and the time points are based on prior process knowledge and off-line analysis. However, this off-line analysis is not ideal as it increases the risk of contamination and does not adequately control glucose and lactate in a cell culture bioreactor.SUMMARY

[0013] The present disclosure provides a method of measuring process attribute(s) of a perfusion cell culture. In one aspect, the present disclosure provides a method for collecting inline Raman spectra downstream of a bioreactor, correlating the Raman spectra measurements with corresponding offline reference measurements to build a chemometrics model and, using the chemometrics model to monitor and / or control the perfusion cell culture.

[0014] In another aspect, the present disclosure provides consistent Raman spectra from a cell- free sample matrix enabling a Raman / chemometrics model for measurements of several analytes or process attributes including glucose, lactate, glutamine, glutamate, and permeate titer.

[0015] The summary of the technology’ described above is non-limiting and other features and advantages of the technology will be apparent from the following detailed description, and from the claims.BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG. 1(A) and FIG. 1(B) show a set-up of perfusion process using alternative tangential flow filtration (ATF) and tangential flow filtration (TFF).

[0017] FIG. 2(A) shows a set-up of an inline Raman data collection for the perfusion process with the inline Raman data collected from the Raman probe located in the bioreactor.

[0018] FIG. 2(B) shows a set-up of an inline Raman data collection for the perfusion process with the inline Raman data collected from the Raman probe that is in the flowcell in the permeate line of the bioreactor.

[0019] FIG. 3 shows a setup of perfusion cell culture with the inline Raman probe located in the permeate line of the bioreactor. This figure also shows the feedback loop for automated glucose feed control based on the Raman permeate measurement.

[0020] FIG. 4 shows a plot of Raman spectrum collected from day 3 to day 28 inside a 2 L bioreactor for mAb A. One representative Raman spectrum from each day was included.

[0021] FIG 5. shows a plot of Raman spectrum collected from day 3 to day 28 of the same perfusion run as shown in FIG. 4. FIG. 5 shows the spectrum data collected in the permeate line using a stainless steel flowcell downstream of the 2 L bioreactor for mAb A.

[0022] FIG. 6(A) and FIG. 6(B) show the performance of glucose (FIG. 6A) and lactate (FIG.6B) model on a 2 L perfusion run of mAb A. FIG. 6A and FIG. 6B are both models that were developed using Raman data collected from inside perfusion bioreactors.

[0023] FIG. 7 show s the performance of glucose and lactate measured in the permeate as compared with offline values in a 2 L perfusion run.

[0024] FIG. 8 shows a profile of measured glucose dynamics in the permeate after glucose spiking inside the perfusion bioreactor. The change in glucose level was first observed ~ 10 minutes after glucose spiking and reached a plateau about 30 minutes after initial spiking.

[0025] FIG. 9 show s a glucose profile by Raman measurement compared to an offline measurement in a 50 L perfusion bioreactor run of mAb A. The same Raman model was used in FIG. 10.

[0026] FIG. 10 show s a glucose profile by Raman measurement compared to an offline measurement in a 500 L perfusion bioreactor run of mAb A. The same Ramen model w as used in FIG. 9.

[0027] FIG. 11 shows a glucose profile by Raman measurement compared to an offline measurement in a 50 L perfusion bioreactor run of mAb B.

[0028] FIG. 12 shows a glucose profile by Raman measurement compared to an offline measurement in a 500 L perfusion bioreactor run of mAb B.

[0029] FIG. 13 is a line graph comparing the glucose measurement from a circulated permeate stream compared to a flowrate.

[0030] FIG. 14 is a line graph of glucose measurements taken from the permeate comparing two flowcells of different diameters.

[0031] FIG. 15 is a line graph of a glucose profile in a 500 L perfusion run of mAb B where glucose feed was automatically controlled based on a permeate based Raman measurement.

[0032] FIG. 16 is a line graph of a glucose profile in a 500 L perfusion run of mAb B where glucose feed was automatically controlled based on a Raman measurement starting from day 6. Note that the sharp drop around day 6 was due to a malfunction of the feed pump.

[0033] FIG. 17 is a line graph of a Raman profile of glutamine in a 500 L perfusion run of mAb B.

[0034] FIG. 18 is a line graph of a Raman profile of glutamate in a 500 L perfusion run of mAb B.

[0035] FIG. 19 is a line graph of a Raman profile of a permeate titer profile of a 500 L run of mAb A compared against offline reference values.

[0036] FIG. 20 is a line graph of a Raman profile of a permeate titer profile of a 500 L run of mAb B compared against offline reference values.DETAILED DESCRIPTION

[0037] The present disclosure is directed to an improved perfusion cell culture system using Raman spectroscopy measurements of process attributes, correlating the Raman spectra with offline reference measurements, building a chemometrics model to monitor and / or control the perfusion cell culture, and wherein the Raman spectra are measured dow nstream of a perfusion bioreactor but prior to purification.

[0038] In one aspect, the present disclosure is directed to a method of controlling a process attribute of a perfusion cell culture, the method comprising:(a) generating an inline Raman spectra of a process attribute by taking multiple measurements of the concentration of the attribute, wherein the concentration is measured in a sample downstream of a perfusion bioreactor prior to purification of the cells in the culture,(b) correlating the Raman spectra measurements with offline reference measurements of the concentration of the attribute, wherein the offline reference measurements were measured in one or more samples taken from the perfusion bioreactor.(c) building a chemometrics model using the inline Rama spectra and the offline reference measurements,(d) monitoring the concentration of the attribute downstream of the perfusion bioreactor; and(e) using the chemometrics model to control the perfusion cell culture by maintaining the concentration of the process attribute.

[0039] In one aspect, the measurement is taken from the perfusion cell culture permeate. In another aspect, the measurement is taken immediately downstream of a perfusion bioreactor. Instill another aspect, the measurement is taken in the permeate of the perfusion cell culture, downstream of the cell retention device.

[0040] In one aspect, the present disclosure provides an improved process for monitoring and / or control of a perfusion cell culture.

[0041] In one aspect, the Raman spectra are measured using a flowcell. In yet a further aspect, the Raman spectra are measured by a probe located within the flowcell. In another aspect, the Raman spectra are measured from a surge vessel. In one aspect, the collection of Raman spectra is taken with a flow cell downstream of the filter unit.

[0042] In another aspect, the location of the flow cell downstream of the filter unit avoids the interference introduced by the cells and cell debris. In yet another aspect, the Raman spectra collection downstream of the bioreactor improves feedback control as compared to immersion Raman probes inside the bioreactors allowing for improved modelling due to the cell-free matrix.

[0043] In another aspect, the present disclosure comprises measuring levels of process attributes or process analytes including glucose, lactate, glutamine, glutamate, and / or permeate titer. In one aspect, the process attribute is a metabolite. In another aspect, the process attribute is a nutrient source. In another aspect, the process attribute is a waste product. In another aspect, the process attribute is an amino acid. In another aspect, the process attribute is Glutamine. In another aspect, the process attribute is Glucose. In another aspect, the process attribute is Lactate. In another aspect, the process attribute is permeate titer. In another aspect, the process attribute is a vitamin.

[0044] In one aspect, the present disclosure comprises a method of measuring at least one process attribute of recombinant protein manufacturing by an inline measurement of the perfusion cell culture permeate. In yet a further aspect, the present disclosure comprises a method of measuring at least one process attribute of manufacturing a monoclonal antibody by an inline measurement of the perfusion cell culture permeate.

[0045] In one aspect, the present disclosure comprises monitoring and controlling the perfusion cell culture to a specific setpoint of a process attribute. In another aspect, the present disclosure comprises monitoring and controlling the perfusion cell culture to a range of a process attribute.

[0046] In one aspect, the present disclosure provides a Raman model with feedback for automated glucose feed control in perfusion bioreactors. In a further aspect, the method of measuring the perfusion cell culture collecting Raman spectra downstream of a perfusion bioreactor provides for accurate glucose control and feed into the bioreactor. In yet a further aspect, collection of the Raman spectra from the permeate line enables continuous, autonomous inline glucose control and feedback for the glucose level in the bioreactor.

[0047] In another aspect, the present disclosure provides consistent spectral data enabling a Raman model for measurements of several process attribute or process analytes. In yet a further aspect, the consistent spectral data allow for accurate chemometrics modeling to incorporate and account for the changes in perfusion scale and variations in permeate flowrate from a small lab scale to implementation in a larger scale of perfusion.

[0048] In one aspect, the present disclosure provides collection of Raman spectra through a probe in a flowcell dow nstream of the bioreactor. The use of probes to gather spectroscopic data and measurements is then modeled in software programs and the models provide information similar to that of hardware sensors. The use of the Raman models is beneficial to extract the maximum amount of value out of the measurements from the Rama spectra. These models are typically multivariate chemometrics models and can account for correlation between Raman spectra and offline measurements.

[0049] In one aspect, the collection of Raman spectra downstream of a perfusion bioreactor avoids fluorescence interference and allows for an accurate measurement.

[0050] In one aspect, the collection of Raman spectra from the permeate line utilizes a clean, consistent sample matrix. In yet a further aspect, the collection of Raman spectra taken downstream of the filter unit eliminates interference due to the auto fluorescence of the cells in high density cell cultures when measuring using Raman immersion probes inside the bioreactor.Definitions

[0051] Listed below7are definitions of various terms used herein. These definitions apply to the terms as they are used throughout this specification and claims, unless otherwise limited in specific instances, either individually or as part of a larger group.

[0052] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. Generally, the nomenclature used herein and the laboratory procedures in cell culture, molecular genetics, biologies process development and manufacturing are those well-known and commonly employed in the art.

[0053] As used herein, the articles “a” and “an” refer to one or to more than one (i.e., to at least one) of the grammatical object of the article. By way of example, “an element” means one element or more than one element. Furthermore, use of the term “including” as w ell as other forms, such as “include,” “includes,” and “included,” is not limiting.

[0054] As used herein, the term “about” in quantitative terms refers to plus or minus 10% of the value it modifies (rounded up to the nearest w hole number if the value is not sub-dividable, such as a number of molecules or nucleotides).

[0055] All ranges disclosed herein are inclusive of the recited endpoint and independently combinable (for example, the range of “from 50 mg to 500 mg’’ is inclusive of the endpoints, 50 mg and 500 mg, and all the intermediate values). The endpoints of the ranges and any values disclosed herein are not limited to the precise range or value; they are sufficiently imprecise to include values approximating these ranges and / or values.

[0056] As used herein, the term “comprising” may include the embodiments “consisting of’ and “consisting essentially of.” The terms “comprise(s),” “include(s),” “having,” “has,” “may,” “contain(s),” and variants thereof, as used herein, are intended to be open-ended transitional phrases, terms, or words that require the presence of the named ingredients / steps and permit the presence of other ingredients / steps. However, such description should be construed as also describing compositions or processes as “consisting of’ and “consisting essentially of’ the enumerated components, which allows the presence of only the named components or compounds, along with any acceptable carriers or fluids, and excludes other components or compounds.

[0057] The term “perfusion cell culture” refers to a continuous process of culturing cells. Perfusion is upstream processing which retains cells inside the bioreactor while continually removing cells, cell waste products and media depleted of nutrients by cell metabolism. Fresh media is provided to the cells at the same rate as the spent media is removed. The most common means to achieve perfusion is the use of hollow fiber filtration in the mode of alternative tangential flow filtration (ATF) or tangential flow filtration (TFF). With the aid of ATF or TFF devices, unwanted by-products like lactate, ammonium, dissolved carbon dioxide and some other byproducts, are removed while fresh nutrients are added continuously. Other means of achieving perfusion like hydrocyclones and acoustic separators have been explored in the industry.

[0058] A “bioreactor” as defined herein is a device that comprises cell culture (i.e., cells and a medium for culture). It preferably maintains a favorable environment for the cells by providing appropriate conditions such as temperature, pH, dissolved oxygen concentration, ions concentration, continuous stirring, and the like. It provides sterile environment.

[0059] In general, the bioreactor, for instance, may comprise a fermenter, a stirred-tank reactor, an adherent bioreactor, a wave-type bioreactor, a disposable bioreactor, and the like.

[0060] Raman spectroscopy measures changes in the vibrational frequency of component specific molecular bonds. Raman provides complimentary’ information to more traditional mid- IR spectroscopy, while having more utility in aqueous solutions due to its resistance to water interference, making it desirable for bioreactor applications.

[0061] Raman spectra collected from inline Raman probes, coupled with a host computer capable of multivariate analysis (MV A), can be used to monitor nutrients, metabolites, titer and product quality attributes within the bioreactor. Raman spectroscopy provides the ability7to monitor bioprocesses in real time which allows for the implementation of feedback controls for nutrient feeds leading to improved product quality and cell productivity.

[0062] A “Raman analyzer” as used herein refers to a device capable of collecting Raman spectrum in the bioreactor or permeate line through one or more Raman probes and calculating the collected Raman spectrum for biochemical parameter values, such as viable cell density7(VCD), cell diameter, pH, pCO2, glucose, glutamine, glutamate, lactate, ammonium ions and titer, osmolality, and the like.

[0063] The Raman analyzer generally comprises (i) one or more Raman probes which are immersed into the bioreactor or permeate line and are configured to collect Raman spectra in the bioreactor or permeate line, and (ii) a host computer which is configured to receive the Raman spectrum collected and transferred by the Raman probe and turn it into biochemical parameter values. The Raman analyzer further comprises (iii) a laser-emitting module, which emits a laser into the bioreactor or permeate line with excitation of certain wavelength periodically (e.g., 15 min ~1 hour). One example of a wavelength used is 785nm.

[0064] “Cell retention system” as used herein refers to a cell retention device, i.e., crossflow membrane filters, like an ATF system, through which the product harvest is obtained while the cells are kept in the bioreactor based on their sizes.

[0065] The term “permeate” refers to the harvest stream after any cell retention device (e.g., filtration based), but precedes the first purification unit operation.

[0066] The term ’online’ or ‘on-line’ includes sampling to a unit that is physically connected directly to the process and includes automatic sampling (including when instructed by an operator) where the sampling never returns to the process due to potential contamination risk to the batch.

[0067] The term ‘at-line’ or ‘atline’ includes sampling to a unit that is physically located next to the process and requires manual sampling.

[0068] The term ‘in-line’ or ‘inline’ includes sampling physically in contact with or in the flow the of the process. For example, a measurement taken from the permeate is in-line.

[0069] The term “antibody” includes an immunoglobulin molecule comprised of four polypeptide chains, two heavy (H) chains and two light (L) chains inter-connected by disulfide bonds. Each heavy chain is comprised of a heavy chain variable region (abbreviated herein as HCVR or VH) and a heavy chain constant region (CH). The heavy chain constant region iscomprised of three domains. CHI. CH2 and CH3. Each light chain is comprised of a light chain variable region (abbreviated herein as LCVR or VL) and a light chain constant region. The light chain constant region is comprised of one domain, CL. The VH and VL regions can be further subdivided into regions of hypervariability, termed complementarity determining regions (CDRs), interspersed with regions that are more conserved, termed framework regions (FR). Each VH and VL is composed of three CDRs and four FRs, arranged from amino-terminus to carboxy -terminus in the following order: FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4. The term “antibody’', as used herein, also includes alternative antibody and antibody-like structures, such as, but not limited to, dual variable domain antibodies (DVD-Ig).

[0070] The term “antigen-binding portion” of an antibody (or “antibody portion”) includes fragments of an antibody that retain the ability’ to specifically bind to an antigen (e.g., hIL-12, hTNFa, or hIL-18). It has been show n that the antigen-binding function of an antibody can be performed by fragments of a full-length antibody. Examples of binding fragments encompassed within the term “antigen-binding portion” of an antibody include (i) a Fab fragment, a monovalent fragment comprising the VL, VH, CL and CHI domains; (ii) a F(ab')2 fragment, a bivalent fragment comprising two Fab fragments linked by a disulfide bridge at the hinge region; (iii) a Fd fragment comprising the VH and CHI domains; (iv) a Fv fragment comprising the VL and VH domains of a single arm of an antibody, (v) a dAb fragment (Ward et al. ,(1989) Nature 341:544-546, the entire teaching of which is incorporated herein by reference), which comprises a VH domain; and (vi) an isolated complementarity determining region (CDR). Furthermore, although the tw o domains of the Fv fragment, VL and VH, are coded for by separate genes, they can be joined, using recombinant methods, by a synthetic linker that enables them to be made as a single protein chain in which the VL and VH regions pair to form monovalent molecules (known as single chain Fv (scFv); see, e.g., Bird et al. (1988) Science 242:423-426; and Huston et al. (1988) Proc. Natl. Acad. Sci. USA 85:5879-5883, the entire teachings of which are incorporated herein by reference). Such single chain antibodies are also intended to be encompassed within the term “antigen-binding portion” of an antibody. Other forms of single chain antibodies, such as diabodies are also encompassed. Diabodies are bivalent, bispecific antibodies in which VH and VL domains are expressed on a single polypeptide chain, but using a linker that is too short to allow for pairing between the two domains on the same chain, thereby forcing the domains to pair with complementary domains of another chain and creating two antigen binding sites (see, e.g., Holliger, P., et al. (1993) Proc. Natl. Acad. Sci. USA 90:6444- 6448; Poljak, R. J., et al. (1994) Structure 2:1121-1123, the entire teachings of which are incorporated herein by reference). Still further, an antibody may be part of a largerimmunoadhesion molecule, formed by covalent or non-covalent association of the antibody with one or more other proteins or peptides. Examples of such immunoadhesion molecules include use of the streptavidin core region to make a tetrameric scFv molecule (Kipriyanov, S. M., et al. (1995) Human Antibodies and Hybridomas 6:93-101, the entire teaching of which is incorporated herein by reference) and use of a cysteine residue, a marker peptide and a C- terminal polyhistidine tag to make bivalent and biotinylated scFv molecules (Kipriyanov, S. M., et al. (1994) Mol. Immunol. 31 : 1047-1058, the entire teaching of which is incorporated herein by reference). Antibody portions, such as Fab and F(ab')2 fragments, can be prepared from whole antibodies using conventional techniques, such as papain or pepsin digestion, respectively, of whole antibodies. Moreover, antibodies, antibody portions and immunoadhesion molecules can be obtained using standard recombinant DNA techniques, as described herein. In one aspect, the antigen binding portions are complete domains or pairs of complete domains.

[0071] The term ‘antibody’ refers to any form of antibody that exhibits the desired biological activity. Thus, it is used in the broadest sense and specifically covers, but is not limited to. monoclonal antibodies (including full length monoclonal antibodies), polyclonal antibodies, bispecific antibodies, humanized, fully human antibodies, and chimeric antibodies.

[0072] "Monoclonal antibody" or “mAb” or “Mab”, as used herein, refers to a population of substantially homogeneous antibodies, i.e., the antibody molecules comprising the population are identical in amino acid sequence except for possible naturally occurring mutations that may be present in minor amounts. In contrast, conventional (polyclonal) antibody preparations typically include a multitude of different antibodies having different amino acid sequences in their variable domains, particularly their CDRs, which are often specific for different epitopes. The modifier "monoclonal" indicates the character of the antibody as being obtained from a substantially homogeneous population of antibodies and is not to be construed as requiring production of the antibody by any particular method. For example, the monoclonal antibodies to be used in accordance with the present invention may be made by the hybridoma method first described by Kohler et al. (1975) Nature 256: 495, or may be made by recombinant DNA methods (see. e.g., U.S. Pat. No. 4,816,567). The "monoclonal antibodies" may also be isolated from phage antibody libraries using the techniques described in Clackson et al. (1991) Nature 352: 624-628 and Marks et al. (1991) J. Mol. Biol. 222: 581-597, for example. See also Presta (2005) J. Allergy Clin. Immunol. 116:731.EXAMPLES

[0073] The following examples are meant to be illustrative and should not be construed as further limiting. The contents of the figures and all references, patents, and published patent applications cited throughout this application are expressly incorporated herein by reference.Materials:ATF system: Repligen Alternating Tangential Flow (ATF2) SystemRaman system: Kaiser RXN2 / RXN4 multi-channel analyzer with 785 nm excitation laser (Kaiser Optical Systems. Inc., Ann Arbor, MI).The Bioprofile FLEX2 analyzer (Nova Biomedical BioProfile, Waltham, MA) ProA-HPLC (Alliance, Waters Corporation, Milford, MA).SIP AT: Siemens PAT Solution (Siemens AG, Munich, Germany).DeltaV Automation Systems, Emerson softwareFlowcell (1 / 4 inch ID, flow assembly CYA680. Endress & Hauser Inc.).Example la: Perfusion Cell Culture Process

[0074] Two CHO-K1 cell lines secreting mAb A and mAb B respectively were used in the study. For both antibodies, the antibody was produced in a stirred-tank bioreactor using chemically defined proprietary in-house media in a suspension cell culture, which was initiated from a Working Cell Bank (WCB) vial. Inoculum expansion was carried out sequentially in shake flasks, rocker bags, and further in the N-l perfusion bioreactor prior to final scale-up in the production bioreactor. Perfusion started on day 3 after inoculation of production bioreactor. Perfusion rate was initially 0.5 vessel volume per day (VVD) and then gradually ramped up to 2 VVD on day 5 and then maintained at 2 VVD till the end of perfusion run.

[0075] Production culture was supplemented with glucose (maintaining setpoint of approximately 2.5 g / L for mAb A and setpoint of 3 g / L for mAb B) and antifoam (> 25 parts per million, ppm) as needed. Glucose feed rate was adjusted once even’ ~24 hours based on once- daily offline measurement (unless specified otherwise when automated feed control was used based on Raman measurement). Temperature, pH, and dissolved oxygen (DO) were automatically controlled at pre-defined setpoints. For perfusion, media exchange started on Day 3 post inoculation in production reactor. Cells were retained inside the bioreactor with either 0.2 pm alternating tangential flow filtration (ATF, Repligen) or 0.2 pm tangential flow filtration (TFF, Asahi) cell retention devices. Biovolume inside the bioreactor was controlled withcapacitance (Incyte, Hamilton). Once the capacitance setpoint is reached, cell removal / bleed occurred to maintain the biovolume inside the bioreactor.

[0076] A flowcell coupled with a Raman probe was connected to the permeate line downstream of the permeate pump. A Raman probe along with the flowcell was calibrated and autoclaved before sterile welding on to the permeate line. One Raman spectrum was collected on each channel every 60 minutes starting Day 3 when perfusion started, until the end of the process. Raman spectra data was sent to SIP AT and then to DeltaV (process analytical technology software automation software). For Raman based glucose control, DeltaV receives the glucose concentration from SIP AT, performs necessary calculations and provides feedback to adjust the glucose pump rate accordingly.Offline Sampling during Perfusion:

[0077] Reference data (offline sampling during perfusion) for multivariate calibration model development were generated by running samples from perfusion bioreactor on offline analyzers. Samples were taken from bioreactors daily. The Bioprofile FLEX2 analyzer (Nova Biomedical BioProfile, Waltham, MA) was used to measure concentration of certain metabolites including glucose, lactate, glutamine, and glutamate in each sample.

[0078] The offline mAb titer in permeate was measured by ProA-HPLC (Alliance, Waters Corporation, Milford, MA). The Trypan blue exclusion-based method was used to measure viable cell density, cell viability, and cell diameter using Cedex HiRes analyzer (Roche diagnostic, Indianapolis, IN).Inline Raman Spectra Acquisition and Glucose Feed Automation:All Raman data were acquired in the Raman shift range of 100- 3425 cm-1 using Kaiser RXN2 / RXN4 multi-channel analyzer with 785 nm excitation laser (Kaiser Optical Systems, Inc., Ann Arbor, MI). RXN2 and RXN4 are equivalent analyzers except that RXN4 is enclosure-based and is more suitable for large scale process environments. Each Raman analyzer was equipped with four optical fiber channels and each individual channel was connected to a bIO-LAB-220 (Kaiser Optical Systems, Inc.) optics probe which can be immersed into a cell culture bioreactor or a flow-cell in permeate stream for inline Raman data collection. All Raman data were collected through SIP AT (SIEMENS AG, Germany) software. Raman data were collected from both inside perfusion bioreactor and / or in permeate through a stainless steel flowcell (flow assembly CYA680, Endress & Hauser Inc., Greenwood, IN) which was connected to the permeate line by sterile welding. Raman spectra collection inside bioreactor were initiated upon bioreactor inoculationwhile data collection with permeate typically started from day 3 when perfusion was enabled. Unless otherwise specified, Raman data were collected using 10 seconds exposure with 72 accumulations, taking approximately 13 minutes to complete one spectrum acquisition. For each individual probe, typically Raman data were collected at a frequency of one spectrum every 60 minutes.

[0079] For selected perfusion runs at 500L scale, Raman glucose model based on perfusion permeate measurement was tested for automated continuous glucose feed control. Under such circumstances using Raman for feed control, SIP AT was configured to interact with process automation controller (DeltaV or MFCS) to ensure glucose values predicted by Raman chemometrics model were sent to the controller. Based on Raman glucose measurement and predefined control algorithm and logics, DeltaV or MFCS then communicates with glucose feed pump to control glucose feed rate so that glucose concentration inside perfusion bioreactor can be adjusted to desired target value. FIG 1(A) and FIG 1(B) show the filtration setup for ATF and TFF.Chemometrics Model Building

[0080] Raman data were collected on perfusion runs across different scales, including lab scale at 2 L and 10 L, pilot scale at 50 L, and clinical manufacturing scale at 500 L. For mAb A, the initial model was built on Raman spectra collected from 2 L and 10 L perfusion runs. The model developed was then implemented at additional larger scales. For mAb B, data from 50 L and 500 L were evaluated.

[0081] Multivariate PLS (Partial Least Squares) calibration models were constructed using SIMCA v.13 (Sartorius Stedim, Germany). Raman spectra were first preprocessed to correct baseline and background shifts, residual Rayleigh scattering and process related intensity variations in raw Raman spectra. All preprocessing procedures were performed in SIMCA v.13 and were implemented sequentially. In the first step, the first derivative spectrum for each Raman spectrum w as calculated, using a second order Savitzky-Golay algorithm with a 15-datapoint window . Standard normal variate (SNV) w as then applied to the derivative spectra in the spectral range of 450-1800 cm-1. Following SNV, mean-centering was applied on the spectral dataset. The PLS model was then developed to correlate the processed spectral data with offline reference measurements of each analyte of interest, including glucose, lactate, glutamine, glutamate and permeate titer. During model development, for each individual analyte of interest, modeling parameters such as spectral range and number of latent variables (LVs) w ere optimized based onsystematic evaluation of cross-validation results, i.e., root mean squared error of cross-validation (RMSECV).

[0082] Inline Raman data for the perfusion process can be collected from both inside the perfusion bioreactor and on the permeate with a stainless steel flowcell connected to the permeate line (see FIG 2(A) and FIG 2(B)). FIG. 2(A) and FIG. 2(B) display a TFF setup, however, the same can apply to ATF setup.Example lb: A Comparison of Raman Spectra collected from Perfusion Bioreactors and Permeates

[0083] To evaluate the feasibility of Raman method development using spectral data collected directly from a perfusion bioreactor, spectra collected in different days across the perfusion run were initially assessed. FIG. 4 shows a Raman spectra collected from day 3 to day 28 of a 2L perfusion run for mAb A. One representative spectrum from each day is included in the plot, which shows that fluorescence background constantly increasing throughout the entire perfusion process. Offline cell count analysis demonstrated that typically viable cell density (VCC) was increasing in the first 10 days and was then controlled around 120 million cells / ml by using capacitance probe. Starting from day 15, cell viability started to decrease. Despite stable VCC in the second half of perfusion process, fluorescence background in Raman spectra increased due to the accumulation of dead cells / cell debris inside perfusion bioreactor. With such vast and constantly changing fluorescence background, it is difficult to develop a robust Raman model using data collected from perfusion bioreactor.

[0084] In comparison, Raman spectra were also collected in the permeate line for this perfusion run by using a flowcell configuration described above. In comparison to FIG. 4, FIG. 5, shows representative Raman spectra from each day (starting from day 3) in the entire process. Raman spectra collected from the permeate line are very consistent across the entire perfusion run and there is no fluorescence background variation present. Since the permeate is free of cells and cell debris, as they are retained inside the bioreactor by the perfusion filter, it can be confirmed that the accumulation of cells and cell debris was the main source of fluorescence background observed in FIG. 4 for Raman spectra collected inside the perfusion bioreactor.

[0085] With the Raman spectra data collected from the permeate being fluorescence-free and having a higher consistency compared with spectra from inside the perfusion bioreactor, the high-quality Raman spectra for measurement of certain analytes during perfusion process is beneficial for process optimization. While the perfusion filter is typically known to have mAb sieving problems which can cause higher mAb concentration inside the bioreactor compared tothat in the permeate, there is no such retention issue for highly soluble and much smaller molecules like glucose and lactate etc. Therefore, glucose or lactate in the permeate is shown to be similar to that inside perfusion bioreactor.

[0086] PLS models were developed to correlate permeate Raman spectra from a 2L perfusion run with offline glucose / lactate measurement values from inside perfusion bioreactor. As a comparison, models were also developed using Raman spectra collected from inside bioreactor of the same perfusion run. Models developed from the two different approaches were then applied to measure a separate, new 2 L perfusion run under similar process conditions. In this 2L run, glucose feed was controlled / adjusted based on once-daily offline measurement using Flex2. FIG. 6A and FIG. 6B shows the glucose and lactate measurement profiles by the Raman methods developed using probe inside bioreactor. Raman predictions showed significant deviations from offline measurement in the model with measurements taken from inside the bioreactor, especially in the second half of the run because there were enormous fluorescence background variations with change of cell density and accumulation of cell debris inside perfusion bioreactor.Additionally, variations in the offline glucose values were observed, despite glucose feed being controlled to a target glucose setpoint of 3 g / L. It appears the implemented glucose feed control strategy was unable to deliver reliable and consistent glucose control along the perfusion process. The once-daily adjustment in glucose feed rate will not be sufficient in realizing robust and consistent glucose control as perfusion dynamics as glucose consumption rate can change significantly in 24 hours.

[0087] Similarly, models developed using Raman spectra measured from the permeate w ere also applied to Raman data collected on this perfusion run as shown in FIG. 7. FIG. 7 shows the Raman model with predicted glucose and lactate concentration profiles relative to offline measurement values from inside perfusion bioreactor. The Raman predictions align with offline measurements and show a higher correlation as compared with models developed using Raman data from inside bioreactor. This demonstrated that there is good feasibility of using permeate Raman data to model the glucose and lactate (etc.) concentration values inside the perfusion bioreactor.Estimation of residence time insider perfusion filter by spiking study

[0088] Even though the Raman spectra collected from the permeate line showed a good correlation with offline glucose and lactate concentration values measured from inside perfusion bioreactor, measurements needed to be confirmed for how' the measurement on the permeate represents the true glucose and lactate concentration values inside the bioreactor.

[0089] For the perfusion process, there is a time delay for the permeate flow to pass through the perfusion filter. A spiking experiment was designed on a perfusion bioreactor using ATF2 perfusion filter to estimate the residence time of glucose inside the perfusion filter. The experiment was implemented on day 28 of a 2L perfusion run for mAh A and VCC was measured to be about -120 million / ml with viability measured around 70%. During the experiment, a high-concentration glucose bolus feed was introduced into the perfusion bioreactor using a syringe while Raman data was continuously collected on the permeate side using a probe / flowcell assembly. Raman data were collected using 10 seconds exposure with 10 accumulations so that one spectrum can be collected every 2 minutes. FIG. 8 shows the profile of glucose by Raman measurement on the permeate. The corresponding glucose concentration change in the permeate was first observed about 10 minutes after glucose spiking initiation and the measured glucose profile plateaued in approximately 30 minutes after glucose spiking. The residence time of glucose inside the perfusion filter was therefore determined in the range of 10- 30 minutes for the perfusion bioreactor. The experiment was performed at the end of the perfusion run when viability is at a low level representing the worst-case scenario for any chemical components inside the perfusion reactor to pass through perfusion filter. Therefore, under typical operation conditions, residence time for small molecules like glucose or lactate should be within 30 minutes. As perfusion cell culture is normally operated under steady-state and changes are very minimal within 30 minutes, measurement of small molecules like glucose and lactate in permeate can represent the environment inside the perfusion bioreactor at the time the measurement is taken.Evaluation of cross-scale transferability and possible impact of permeate flowrate

[0090] Changes of process dynamics inside bioreactors across different scales have prevented cross-scale transfer of Raman bioreactor measurements. Process parameters such as mixing speed / flow dynamics, cell suspension state, gas sparging, foaming behaviors etc. can all vary between bioreactor scales. These changes can result in significant variations in Raman spectra collected inside the bioreactor and therefore variation in Raman models based on different scales. To develop a robust chemometric model, Raman data collected across different scales need to be included to build a calibration dataset representative of all different scales. However, due to time, labor, and cost constraints of maintaining a bioreactor at large scale, development of a model with different scales can limit the feasibility of Raman method development and limit the transferability and implementation of Raman methods developed across scales.

[0091] Raman measurements taken on the cell-free permeate using a flowcell eliminate the process variability that is observed within the bioreactors at different scales as only the permeate flowrate will fluctuate between the scales. Therefore, Raman measurements taken on the permeate line are independent of process variability among different bioreactor scales.

[0092] To evaluate cross-scale transferability of models developed from the permeate measurements, Raman data was collected on mAb A perfusion runs of different scales using the same type of flowcell. Table 1 lists Raman data taken from the permeate line in a mAb A 2 L and a 10 L lab scale run, a 50 L pilot scale run, and a 500 L clinical facility scale run to enable glucose model development. All data were collected using same type of stainless flowcell (flow assembly CYA680, Endress & Hauser Inc.). Four different Raman spectrometers were used for data collection on these perfusion runs at different scales. Glucose and lactate models were initially developed using data from the 2 L and 10 L scale runs and then implemented to the 50 L and 500 L scale. FIG. 9 and FIG. 10 show Raman prediction profiles using the same model relative to an offline glucose measurement throughout a 50 L and a 500 L perfusion process respectively. It is noted that the permeate flowrates are much higher than 2L and 10 L runs. As shown in the plots, Raman measurements correspond to the offline reference values for both the 50L and 500 L runs, with calculated root mean squared error of prediction (RMSEP) of 0.22 g / L and 0.29 g / L, respectively. Therefore, glucose models developed from Raman measurements taken from the permeate line in small scale perfusion cell culture processes are shown to be adequately transferred to different scales.

[0093] Raman data collected from the permeate line in a 50L and 500 L perfusion run with mAb B (shown in Table 2) was used for a cross-scale model evaluation. Glucose and lactate models were initially developed using Raman data from one 50L run and the model was then applied to a separate 50L run and additionally to a 500 L run. FIG. 11 and FIG. 12 show glucose measurement profiles for the 50 L and 500 L run, respectively. In both cases, the glucose model developed correlated with the offline readings and the corresponding glucose prediction error (RMSEP) are 0.24 g / L and 0.3 g / L, respectively.Table 1Table 2Raman measurement with changed flowrate

[0094] An experiment was designed to collect Raman spectra on same permeate stream at different flowrate to assess the impacts of flowrate change during a perfusion run. A 2 L permeate sample was taking at day 24 from a 500 L perfusion run of mAh A. The permeate sample was then circulated through a ! inch flowcell coupled with a Raman probe and the permeate flowrate was controlled using a circulating pump. During this experiment, the flowrate was changed in the range from 139 ml / min to approximately 1 L / min, which corresponds to 0.5 volumes per day (VVD) to about 3 VVD for the 500 L perfusion bioreactor. Raman data were collected using 10 seconds exposure with 72 accumulations and one spectrum was collected every 15 minutes. In total, 18 spectra were collected for this study. FIG. 13 displays a profile of Raman glucose measurement with change of permeate flowrate. The calculated RSD is around 2.5%. There is no significant difference in Raman measurement despite the flowrate changes. Therefore, the Raman models developed based on flowcell permeate measurement are accurate against flowrate variations and therefore, are shown to be transferable across scales.

[0095] Previously, all experiments were performed using flowcells of the same diameter ( inch in outlet diameter for all scales investigated including 2L, 10L, 50L and 500 L). An experiment w as performed comparing tw o different diameters of a flow cell as for a manufacturing facility. 'A inch flowcell may not be sufficient to accommodate a higher permeate flowrate. Raman measurements were then taken on the same perfusion run using two flowcells of different diameters. During a 50 L run for mAb A, tw o flowcells, one that has a diameter of A and another one that has a % inch diameter, were connected sequentially to the permeate line. Each flowcell was coupled with a Raman probe, and Raman data were collected from the two flowcells with using two different channels on the same Raman spectrometer. FIG. 14 shows theRaman glucose profiles from the two flowcells. The measurement values are consistent between the two sets of flowcells indicating that flowcell of different diameters does not cause significant difference to Raman permeate measurement.Automated Glucose Feed Control based on Raman Permeate Measurements

[0096] Glucose feed control is important in ensuring robust perfusion cell culture performance as it’s desirable to control glucose concentration at a consistent level to maintain steady or pseudo-steady process conditions. Glucose control has been challenging in the perfusion process due to high glucose consumption rate from high cell density’ and dynamics in cell metabolic state across the long perfusion process. Traditionally, glucose is often controlled based on once-daily offline measurement (e.g., by Flex2), which is not only time and labor consuming but is also unreliable. As shown in FIG. 6, glucose was controlled using daily offline Flex2 measurement and there are huge variations in glucose concentration throughout the perfusion process. Such variations can significantly compromise process performance and product quality.

[0097] Given the robust performance of the glucose model on perfusion runs of different scales, a permeate measurement-based Raman model was used for automated glucose feed control in a 500 L perfusion run of mAb B in a clinical manufacturing facility . The process flow diagram for automated glucose feed control is as shown in FIG. 3. Typically, Raman measurements started on day 3 when perfusion started. Raman based feed control occurs on day 6 and allows for glucose control based on Raman measurements taken hourly. The glucose setpoint was 3 g / L for this run. Raman spectrum was collected hourly, and a glucose measurement was generated using a glucose model embedded in a SIPAT method which them communicated with DeltaV and provided the hourly measurement. With an integrated logic / algorithm involving the measurement value and glucose set point, DeltaV was able to adjust glucose feed rate of the pump to adjust and control the glucose concentration inside the perfusion bioreactor. An offline glucose measurement was performed daily with samples taken from the perfusion bioreactor. Figure 15 shows the glucose profile by Raman monitoring / control compared with an offline measurement taken throughout the 500 L run. The calculated RMSEP is about 0.22 g / L. Overall, starting from day 6, glucose concentration inside perfusion bioreactor w as well controlled in a narrow’ range of 2.6 - 3.3 g / L, demonstrating robustness of glucose feed control based on the permeate Raman measurement.

[0098] The same methodology was implemented for automated continuous glucose feed control in a 500L perfusion run of mAb A (FIG. 16) where inline Raman measurements matched offline glucose values. Overall, glucose was w ell controlled by Raman within a tight range throughoutthe whole perfusion run except on two occasions (around day 9.5 and day 25) when glucose went low due to pump malfunction.Measurement of Glutamine. Glutamate and Permeate Titer

[0099] The same flowcell set-up shown to demonstrate an accurate correlation for glucose / lactate analysis based on the Raman permeate measurement can be used to measure other analytes, such as glutamine, glutamate and permeate titer. Similar to glucose and lactate, chemometrics models were developed for glutamine and glutamate based on data from selected small-scale runs. FIG. 17 and FIG. 18 show examples of glutamine and glutamate profdes when developed models were applied to a 500 L run of mAb B. In both cases, Raman measurement was well correlated with offline reference values.

[0100] Titer measurements can be valuable for efficient process monitoring / characterization during perfusion production and subsequent downstream process control. Traditionally, a titer is often measured by offline Pro A HPLC. which can take time and resources. The flowcell located in the permeate line can be used for a permeate titer measurement with a properly developed Raman titer model.

[0101] FIG. 19 and FIG. 20 show Raman titer profiles for 500 L perfusion run of mAb A and B. In both cases, titer values correlated offline values from pro A HPLC. The calculated RMSEP was around 0.07 g / L and 0.1 1 g / L respectively.

[0102] The disclosed subject matter is not to be limited in scope by the specific embodiments and examples described herein. Indeed, various modifications of the disclosure in addition to those described will become apparent to those skilled in the art from the foregoing description and accompanying figures. Such modifications are intended to fall within the scope of the appended claims.

[0103] All references (e.g., publications or patents or patent applications) cited herein are incorporated herein by reference in their entirety and for all purposes to the same extent as if each individual reference (e.g.. publication or patent or patent application) was specifically and individually indicated to be incorporated by reference in its entirety for all purposes. Other embodiments are within the following claims.References1. Bielser, J.M., et al., Perfusion mammalian cell culture for recombinant protein manufacturing - A critical review. Biotechnol Adv, 2018. 36(4): p. 1328-1340.2. Gomez. N.. et al., Perfusion CHO cell culture applied to lower aggregation and increase volumetric productivity for a bispecific recombinant protein. J Biotechnol, 2019. 304: p. 70-77.3. Matanguihan, C. and P. Wu, Upstream continuous processing: recent advances in production of biopharmaceuticals and challenges in manufacturing. Current Opinion in Biotechnology, 2022. 78: p. 102828.4. US Food and Drug Administration,. Guidance for Industry PAT-A Framework for Innovative Pharmaceutical Development, Manufacturing and Quality Assurance. 2004; Available from: https: / / www.fda.gov / media / 71012 / dowiiload.5. Gillespie. C., et al.. Systematic assessment of process analytical technologies for biologies. Biotechnology and Bioengineering, 2022. 119(2): p. 423-434.6. US Food and Drug Administration,. Quality consideration for continuous manufacturing guidance for industry. 2019; Available from: https: / / www.gmp- compliance.org / files / guidemgr / UCM632033.pdf.7. US Food and Drug Administration,. Q13 Continuous Manufacturing of Drug Substances and Drug Products. 2023; Available from: https: / / www.fda.gov / media / 165775 / download.8. Esmonde-White, K.A., et al., Raman spectroscopy as a process analytical technology' for pharmaceutical manufacturing and bioprocessing. Analytical and Bioanalytical Chemistry. 2017. 409(3): p. 637-649.9. Jesus, J., N. A. Lobenberg R Fau - Bou-Chacra, and N. A. Bou-Chacra, Raman Spectroscopy for Quantitative Analysis in the Pharmaceutical Industry. (1482-1826 (Electronic)).10. Abu-Absi, N.R., et al., Real time monitoring of multiple parameters in mammalian cell culture bioreactors using an in-line Raman spectroscopy probe. Biotechnology and Bioengineering, 2011. 108(5): p. 1215-1221.11. Andre, S., et al., In-line and real-time prediction of recombinant antibody titer by in situ Raman spectroscopy. Analytica Chimica Acta, 2015. 892: p. 148-152.12. Czeterko. M., et al., In Situ Raman Spectroscopy Systems and Methods for Controlling Process Variables in Cell Cultures. 2019, Google Patents.13. Li, B., et al., Performance monitoring of a mammalian cell based bioprocess using Raman spectroscopy. Analytica Chimica Acta, 2013. 796: p. 84-91.14. Santos, R.M., et al., Improving reliability' of Raman spectroscopy for mAb production by upstream processes during bioprocess development stages. Taianta. 2019. 199: p. 396-406.15. Liu, Z., et al.. The application of Raman spectroscopy for monitoring product quality attributes in perfusion cell culture. Biochemical Engineering Journal, 2021. 173: p. 108064.16. Baradez, M.. et al.. Raman spectroscopy for adaptive process control of adoptive T-cell immunotherapy manufacture. Cytotherapy, 2020. 22(5, Supplement): p. SI 17.17. Craven, S., J. Whelan, and B. Glennon, Glucose concentration control of a fed-batch mammalian cell bioprocess using a nonlinear model predictive controller. Journal of Process Control, 2014. 24(4): p. 344-357.18. W. Eyster, T., et al., Tuning monoclonal antibody galactosylation using Raman spectroscopy-controlled lactic acid feeding. Biotechnology Progress, 2021. 37(1): p. e3085.19. Matthews, T.E., et al., Closed loop control of lactate concentration in mammalian cell culture by Raman spectroscopy leads to improved cell density, viability, and biopharmaceutical protein production. Biotechnology and Bioengineering, 2016. 1 13(11): p. 2416-2424.20. Matthews, T.E., et al., Glucose monitoring and adaptive feeding of mammalian cell culture in the presence of strong autofluorescence by near infrared Raman spectroscopy. Biotechnology progress, 2018. 34(6): p. 1574-1580.

Claims

WHAT IS CLAIMED IS:

1. A method of controlling a process attribute of a perfusion cell culture, the method comprising:(a) generating an inline Raman spectra of a process attribute by taking multiple measurements of the concentration of the attribute, wherein the concentration is measured in a sample downstream of a perfusion bioreactor prior to purification of the cells in the culture,(b) correlating the Raman spectra measurements with offline reference measurements of the concentration of the attribute, wherein the offline reference measurements were measured in one or more samples taken from the perfusion bioreactor,(c) building a chemometrics model using the inline Rama spectra and the offline reference measurements,(d) monitoring the concentration of the attribute downstream of the perfusion bioreactor; and(e) using the chemometrics model to control the perfusion cell culture by maintaining the concentration of the process attribute.

2. The method of claim 1, wherein the inline Raman spectra are measured from a perfusion cell culture permeate.

3. The method of claim 1 or 2, wherein the process attribute is a metabolite.

5. The method of claim 1 or 2, wherein the process attribute is a nutrient source.

6. The method of claim 1 or 2, wherein the process attribute is a waste product.

7. The method of claim 1 or 2, wherein the process attribute is an amino acid.

8. The method of any one of claims 1-3, wherein the process attribute is Glutamine.

9. The method of claim 1, wherein the process attribute is Glutamate.

10. The method of claim 1, wherein the process attribute is Glucose.

11. The method of claim 1 , wherein the process attribute is Lactate.

12. The method of claim 1 or 2, wherein the process attribute is permeate titer.

13. The method of claim 1 or 2, wherein the process attribute is a vitamin.

14. The method of any one of claims 1-13, wherein the Raman spectra are taken from a probe in a flowcell.

15. The method of any one of claims 1-13, wherein the Raman spectra are taken from a surge vessel.

Images