An automated high inoculation density fed‐batch bioreactor, enabled through N‐1 perfusion, accommodates clonal diversity and doubles titers

An important consideration for biopharmaceutical processes is the cost of goods (CoGs) of biotherapeutics manufacturing. CoGs can be reduced by dramatically increasing the productivity of the bioreactor process. In this study, we demonstrate that an intensified process which couples a perfused N‐1 seed reactor and a fully automated high inoculation density (HID) N stage reactor substantially increases the bioreactor productivity as compared to a low inoculation density (LID) control fed‐batch process. A panel of six CHOK1SV GS‐KO® CHO cell lines expressing three different monoclonal antibodies was evaluated in this intensified process, achieving an average 85% titer increase and 132% space–time yield (STY) increase was demonstrated when comparing the 12‐day HID process to a 15‐day LID control process. These productivity increases were enabled by automated nutrient feeding in both the N‐1 and N stage bioreactors using in‐line process analytical technologies (PAT) and feedback control. The N‐1 bioreactor utilized in‐line capacitance to automatically feed the bioreactor based on a capacitance‐specific perfusion rate (CapSPR). The N‐stage bioreactor utilized in‐line Raman spectroscopy to estimate real‐time concentrations of glucose, phenylalanine, and methionine, which are held to target set points using automatic feed additions. These automated feeding methodologies were shown to be generalizable across six cell lines with diverse feed requirements. We show this new process can accommodate clonal diversity and reproducibly achieve substantial titer uplifts compared to traditional cell culture processes, thereby establishing a baseline technology platform upon which further increases bioreactor productivity and CoGs reduction can be achieved.


| INTRODUCTION
The monoclonal antibody (mAb) market has grown exponentially since the first mAb was approved by the United States Food and Drug Administration (US FDA) in 1986. 1 Since then, monoclonal antibodies have become a dominant presence in the biopharmaceutical market, making up 53% of the 122 new product approvals from 2014 to 2018, a nearly 100% increase over the previous 4 year period. 2Additionally, the global monoclonal antibody market valuation is ballooning rapidly, with an increase from $115.2 billion in 2018 to an expected $300 billion by 2025. 1 As the commercial pipeline of these proteins expands, a large challenge manufacturers face is dramatically reducing the cost of goods (CoGs) to remain competitive in the growing manufacturing space while also promoting democratization of therapies through reduction of cost to the customer.One effective method to reduce CoGs for biotherapeutics manufacturing is increasing the productivity in the production bioreactor, which increases total product yield without an increase in bioreactor resources.
Advances in antibody production technologies have made giant strides over the last few decades.In 1985, average fed-batch titers were approximately 0.20 g/L, which since then have grown over 1600%, achieving projected average fed-batch titers of 3.20 g/L. 3 However, some literature sources have reported fed-batch titers as high as 10 g/L in 14-18 day processes, indicating that the pace of fed-batch productivity advancements is picking up speed. 4,5These rising yields can be attributed to incremental advances in cell line engineering, 6 improved process control, 4 and advanced media development within the biopharmaceutical industry. 7Although reported yields are steadily growing, with the mAb market greatly expanding, the appetite for further bioreactor productivity advancements is higher than ever.To meet this demand, a paradigm shift toward intensified processing is required to make these substantial productivity gains rather than relying on the incremental technological step changes of the past.
One appealing and highly productive platform process that has gained momentum in recent years is the intensified or high inoculation density (HID) fed-batch bioreactor.In a HID production bioreactor, the inoculation density is as much as 20Â higher than traditional fedbatch bioreactors, which is achieved using a perfusion culture in the N-1 stage to generate the dense inoculum source. 8This process shifts the burden of the cell growth phase to the N-1 stage, while enabling the N stage bioreactor to maintain high cell concentration.Since there is a direct correlation between total mAb production and the total cell number and longevity of a culture, 6 the increased cell concentration translates to large gains in total bioreactor productivity.Many examples of this type of manufacturing process have emerged in the last decade, which report titer increases as high as 100% in bioreactors inoculated at 10 Â 10 6 cells/mL as compared to low density bioreactors. 8,9ny challenges remain with these high inoculation density cultures.Primarily, unique nutrient feeding challenges arise due to the high cell densities in HID processes.One method to monitor and control the metabolites in the culture is to use process analytical technology (PAT), such as in-line Raman spectroscopy.Raman spectra data can be used to train models that estimate metabolite concentrations. 10Others have reported the use of Raman spectroscopy to monitor glucose, lactate, product titer, and multiple different amino acids. 11,12Further, in two separate case studies, nutrient feed control was demonstrated with the use of in-line Raman metabolite estimations, including one example using glucose and phenylalanine 13 and another example glucose and arginine 14 as the controlling metabolites.
These case studies were performed on low inoculation density fedbatch bioreactors and demonstrate the ability of automated nutrient feeding using Raman estimations to maintain robust control of nutrient concentration during a cell culture.
In this work, we demonstrate an intensified process consisting of a perfused N-1 and HID process controlled using a completely automated nutrient feed strategy.The perfused N-1 step utilized a capacitance-based feeding platform, which delivered perfusion media on a per capacitance basis.The HID production bioreactor utilized Raman spectroscopy to monitor glucose, phenylalanine, and methionine to control three separate nutrient feeds.Although previous studies have demonstrated the productivity benefits of perfused N-1 and HID fed-batch cultures, this study seeks to demonstrate for the first time the generalizability of a platform process incorporating completely automated and on-demand feeding across a diverse panel of clones and products.The study evaluates the culture performance of a diverse panel of six separate CHOK1SV GS-KO ® cell lines expressing three different mAbs cultured in the HID process.This panel of cell lines was chosen to test the ability of this platform process to handle representative clonal diversity without the need for additional process optimization.Here we report the performance of these six clones in the perfused N-1 and HID bioreactor process incorporating completely automated feeding.

| Cell lines and seed cultures
Six different CHOK1SV GS-KO ® cell lines, including one GS Xceed ® and five GS piggyBac ® , producing three different IgG mAbs were evaluated in this study, as shown in Table 1.
Cells were thawed according to Lonza standard protocol.Cells were then cultivated in a shaking incubator (Climo-Shaker, Kuhner) at 36.5 C, 5% CO 2 , and 80% humidity.Cells were passaged every 3-4 days for a maximum of 5 passages.

| Perfused N-1 expansion bioreactors
Perfused N-1 stage cultures were generated in 3 L glass Applikon bioreactors, with a 2 L working volume.Perfused N-1 cultures were inoculated at a cell density of 0.5 Â 10 6 cells/mL.After inoculation, cultures entered a 72-h batch phase prior to the start of perfusion.In perfusion mode, fresh perfusion media was continuously added to the bioreactor, while spent media was removed via an alternating tangential flow (XCell™ ATF2, Repligen) cell retention device equipped with a 0.2 μm hollow fiber filter.Perfusion media addition was completely automated using in-line capacitance measurements from a 220 mm Aber capacitance probe at 1000 kHz.The automated feed strategy, also known as the Capacitance-Specific Perfusion Rate (CapSPR), delivered a constant rate of nutrients per unit of capacitance.
Agitation was maintained at a power-to-volume ratio of 60 W/m 3 using a pitch blade impeller.Temperature was controlled at 36.5 C using an in-line probe and heating jacket.Dissolved oxygen (DO) and pH were controlled using in-line probes from Hamilton.PH was controlled at a setpoint of 6.90 and maintained through the addition of CO 2 gas or alkaline solution.Glucose concentration was monitored using an in-line Raman probe and maintained at constant setpoint via an automated feedback control strategy.
The culture was transferred to the N stage bioreactor once the concentration in the N-1 stage reached 50.0 ± 15.0 Â 10 6 cells/mL.

| Intensified N high inoculation density (HID) fed-batch cultures
Intensified N fed-batch cultures were generated in 2 L glass Applikon bioreactors, with a 1 L working volume for clones A, B, E and F. Clones C, and D were cultured in 7 L glass Applikon bioreactors with a 5 L working volume.Pitch blade impellers were used for agitation at a power to volume ratio of 60 W/m 3 .Temperature was maintained at an initial setpoint of 36.5 C then incorporated a shift to a lower temperature for the remainder of the 12-day culture duration.
Dissolved oxygen and pH were controlled using in-line probes from Hamilton.DO was controlled at 40%.The pH was controlled at an initial setpoint of 6.90 then incorporated a pH shift to a more alkaline setpoint for the remainder of the 12-day culture duration.The pH was maintained using CO 2 sparging and the addition of an alkaline solution.
Immediately after inoculation, three separate nutrient feeds were initiated: one glucose-containing feed and two complex nutrient feeds referred to as Feed X and Feed Y.Each feed was added to the bioreactor via an automated controller to maintain setpoint concentrations of glucose, phenylalanine, and methionine.Phenylalanine is a major component of Feed X, while methionine is a major component of Feed Y, allowing these two amino acids to act as indicators of the relative amounts of each feed needed in the bioreactor.The real time concentrations of glucose, phenylalanine, and methionine were reported from model estimates from an in-line Raman probe.The bioreactors were operated for a 12-day culture duration and were sampled daily for measurements of gasses, metabolites, electrolytes, osmolality, cell density, titer concentration, and residual glucose and amino acids.Product quality samples were taken at harvest on Day 12.

| Low inoculation density (LID) fed-batch cultures
The low inoculation density (LID) bioreactors were inoculated at 0.5 Â 10 6 cells/mL in 2 L glass Applikon bioreactors, with a 1 L working volume for clones A, B, E, and F. Clones C, and D were cultured in 7 L glass Applikon bioreactors with a 5 L working volume.Agitation was set at a power to volume ratio of 60 W/m 3 and was controlled via a single pitch-blade impeller.Temperature was maintained at an initial setpoint of 36.5 C then incorporated a shift to a lower temperature for the remainder of the 15-day culture duration.Dissolved oxygen and pH were controlled using in-line probes from Hamilton.DO was controlled at 40%.The pH was controlled at an initial setpoint of 6.90 then incorporated a pH shift to a more alkaline setpoint for the remainder of the 15-day culture duration.The pH was maintained using CO 2 sparging and the addition of an alkaline solution.The bioreactors were run for a 15-day culture duration and were sampled daily for measurements of gasses, metabolites, electrolytes, osmolality, cell density, titer concentration, and residual glucose and amino acids.
Product quality samples were taken on Day 12 and Day 15.

| In-line Raman spectra
In-line Raman spectroscopy was used for glucose feeding in the N-1 stage bioreactor, and for both glucose and nutrient feeding in the N stage bioreactor.The Raman spectra were generated using a RAMAN/RXN2™ from Endress + Hauser.The probes were bio-optics 220 mm or 420 mm with a 785 nm laser.Each spectra was generated T A B L E 1 Clone ID, expression technology, and product.
from 150 scans at a 5 s exposure.Spectra were collected approximately once every hour.A predictive projection on latent structures (PLS) model was generated for each metabolite using SIMCA v16, by matching processed spectra data with offline reference metabolite data.The model used for glucose estimation was first developed for a low density fed-batch process and was applied to the HID process.
Models for phenylalanine and methionine were developed specifically for intensified HID processes.

| In-process cell culture monitoring and analytical methods
The Bioprofile ® Flex2 from Nova Biomedical was used for daily measurement of gases, metabolites, electrolytes, osmolality, and cell density.The pH was measured through offline sampling on the SevenCompact™ S220 probe from Mettler Toledo.Product titer was measured from supernatant retains by Protein A method on HPLC.
Amino acids were measured from supernatant retains on an Agilent 1290 UHPLC.
Product quality was measured for Clones A, C, D, E, and F (Clone B was not assessed for product quality).The product quality attributes measured were charge variants, size variants/aggregates, reduced/ non-reduced purity, and released N-glycans.Clarified supernatant was filtered using a 0.2 μm filter and then frozen to À65 C prior to protein A purification.Clones C, D, E, and F were purified using a MabSelect SuRe™ column on an Äkta™ Avant 25 system from Cytiva, whereas Clone A was purified with an MEA Protein A method.Charge variants were analyzed using capillary isoelectric focusing (icIEF).Size variants and aggregates were measured using Gel Permeation High Performance Liquid Chromatography (GP-HPLC).Purity was assessed using the Perkin Elmer LabChip ® GXII (Caliper) method, where samples were treated under non-reduced and reduced conditions.
N-glycans were measured using an Agilent 1260 UHPLC for Clones C, D, E, and F.

| Platform N-1 expansion
The platform N-1 process in this study utilized the capacitancespecific perfusion rate (CapSPR) of 21 Â 10 À3 mL media/(mL working vol.*(pF/cm)*day) to feed six clones expressing three products.This single perfusion rate allowed for the expansion of all six clones from 0.5 Â 10 6 cells/mL to 50.0 ± 15.0 Â 10 6 cells/mL in 7-9 days, with an average expansion time of 8.2 days, as shown in Figure 1a.
Although each culture grew at different rates, as observed in the raw capacitance traces (Figure 1b), all clones were maintained within 10% of the CapSPR setpoint of 21 Â 10 À3 VV/(pF/cm)/day for the duration of the perfused N-1 cultures (Figure 1c).Other studies have reported perfusion cultures controlled to a constant cell specific perfusion rates (CSPRs). 8,15A difference between these two methods of perfusion feed control is the potential impact of varying cell diameter on the perfusion rate.To illustrate the difference in perfusion rate between these two perfusion control methods for the cultures reported here, equivalent CSPRs were back-calculated from the culture perfusion feed rates, and VCC measurements (Figure 1d).
Cultures fed at a constant CapSPR were in fact fed at different different calculated CSPRs, ranging from 29 to 67 nL/cell/day.Every clone also showed a gradual decrease in the calculated CSPR over time.The variability and decrease in CSPR over time is likely explained by diversity in average cell diameter during the cultures between clones, which ranged from 16.6 to 18.9 μm.Cell diameter is known to impact the measured capacitance of a culture by influencing the capacitance per cell. 16Practically this would mean that a single CSPR applied to multiple clones would likely have led to feeding discrepancies during perfused N-1 culture.This potential feeding discrepancy due to varying cell diameter is avoided with the application of the CapSPR approach outlined in this study.
The total perfusion media consumed during the expansion step was below 4 vessel volumes (VV) for all clones, averaging 2.8 VV with a standard deviation of 0.49 VV (Figure 1e).Of interest, this media usage was lower than what has been reported in literature for similar expansion processes.Others report media usage as low as 3 total 8][19] Keeping the total perfusion media volume under 4 VV reduces the cost and media storage footprint associated with the N-1 stage improving the overall COGs of the process.More details of the clones cultured in the N-1 stage are available in Table 2.
Ultimately, the perfused N-1 stage of this process allowed consistent expansion of cell lines with different growth rates, to 50.0 ± 15.0 Â 10 6 cells/mL, using a simplified, single CapSPR feed strategy and utilized less than 4 vessel volumes of perfusion media.
This high VCC inoculum source then enabled the inoculation of the HID production bioreactor at high concentrations around 10.0 Â 10 6 cells/mL.

| Automated feed control strategy in HID bioreactors
The HID fed-batch process employed an automated Raman-based feeding method.Specifically, an in-line Raman probe provided model estimations of three key metabolites: glucose, methionine, and phenylalanine.These metabolites were maintained at target setpoints using an automated controller, which added each of the nutrient feeds semi-continuously.Feed X and Feed Y contain a high concentration of phenylalanine and methionine respectively, which allowed the amino acids to act as indicators of the relative amounts of each feed needed by the culture.This feed strategy was first developed on a model cell line, referred to as Clone G, which expresses mAb2.Figure 2 demonstrates the controller's ability to maintain each metabolite setpoint within the acceptable range.Each metabolite was well-maintained to the setpoint, as reported by both the Raman model and an offline amino acid measurement.This indicates that the Raman model is estimating accurately and that the feed controller is successfully able to deliver the nutrient demands in the bioreactor.Similar setpoint control within acceptable ranges was observed for each of the Raman-based feeding cultures and clones shown in Figure 4 (data not shown).In the example of the methionine/Feed Y controller in Figure 2b, there was a small spike in methionine concentration between Day 9 and 10.This brief excursion is hypothesized to be due to an overly aggressive controller, which caused a large addition of feed to the bioreactor.A quick response by the controller allowed for the re-stabilization of the methionine concentration back to the desired setpoint within a 24-h window.This feed error was later mitigated through better controller tuning, after which, more stable methionine/Feed Y control was observed.T A B L E 2 N-1 culture summary.

| Repeatability of intensified process across six cell lines with high titer uplift
The intensified HID fed-batch process was demonstrated using a panel of six clones expressing three different mAb products.For every clone that was evaluated, the growth curves and peak viable cell densities of 35.0-40.0Â 10 6 cells/mL were comparable in the HID bioreactor, despite the growth diversity displayed in the N-1 stage.
Further, five out of the six clones maintained >90% viability for the duration of the culture.Clone C displayed a viability decline to 72% on the harvest day; however, this was due to an unintentional alkali addition at the end of the culture duration (data not shown).The osmolality was comparable for each of the cultures and the Day 12 osmolality remained below 400 mOsm/kg, suggesting that the Raman automated feeding strategy did not overfeed any of the clones.To further support this claim, the peak lactate concentration, which is indicative of waste accumulation, was comparable across all clones and once re-consumed remained below 2 g/L during HID cultures.
The final titers measured in the HID process ranges between 6.6 g/L and 9.3 g/L.Additionally, this titer was compared to a 15-day LID control process where titer was measured on both Day 12 and on Day 15.When comparing the HID process to the 12-day LID process the titer uplift averages 170%, and when compared to the 15-day LID process the titer uplift averages 85% (Figure 3f).An alternative comparison which normalizes the HID and LID cultures by the culture duration, reveals a space-time yield increase of 132%.

| Product quality in for HID vs. LID fed-batch process
Product quality attributes of product harvested from HID and LID cul- Of the cell lines tested for product quality, comparable aggregate levels and purity between the LID and the HID control process were observed (Figure 4b-d).Some differences in charge variant profiles were seen (Figure 4a).When comparing product harvested at the end of the LID culture (Day 15) to product harvested at the end of the G1F, and G2F making up >96% of the total n-glycan species across all cultures.Additionally, high mannose species accounted for less than 5% of the glycan species for all samples.It is desirable to keep high mannose species at a low concentration to avoid negative effects on pharmacokinetics, stability, and efficacy on the final mAb product. 20

| DISCUSSION
To the best of the authors knowledge, this study is the first account of a perfused N-1 and intensified HID N bioreactor process which is completely automated using in-line capacitance and Raman based feeding strategies.Further, the utility and robustness of this process was confirmed with a panel of six clones expressing three different products, which represents the largest sample size of different cell lines tested compared to other published studies of intensified processes to date (Table 3).
A key hallmark of the N-1 perfusion stage of this process is the use of a capacitance-specific perfusion rate, which enables a "onesize-fits-all" perfusion feed approach.This study demonstrated that a single perfusion rate of 21 Â 10 3 CapSPR successfully expanded six different cell lines, to 50.0 Â 10 6 ± 15.0 Â 10 6 cells/mL in 7-9 days.
The CapSPR feeding strategy was chosen specifically to create a more consistent nutrient feeding platform that can accommodate diverse growth and metabolic demands better than other commonly used perfusion feeding strategies.Most publications report the use of a cell-specific perfusion rate (CSPR), which is a method that relies on a correlation to relate the in-line capacitance value to a predicted VCC. 8,15,18Capacitance is an indirect measurement of biomass, 16 however, the CSPR feed method delivers nutrients on a per-cell basis rather than on a per-biovolume basis.This means that cells which display variable average cell diameters (and therefore variable biovolume per cell) will require a different VCC-capacitance correlation and a different CSPR.Additionally, due to the reliance on variable cell counting instruments to generate a VCC-capacitance correlation, the use of a CSPR can cause inconsistencies in perfusion feeding.There is known variability in the available methods for cell counting due to sample heterogeneity, operating procedure, measurement device, environment, or data analysis procedure. 24This variability suggests that a VCC-capacitance correlation cannot be easily generalized across differing cell counting methods or across a broad range of cell types or cell densities.Therefore, to reduce the potential for perfusion feeding inconstancies, a CapSPR feed strategy was utilized in this N-1 process.
Despite the advantages of the CapSPR feed strategy to reduce feeding variability, one potential limitation is the generalizability across clones with extremely different metabolic demands.Unlike the Raman controlled feeding, capacitance is not a direct measure of the nutrient demands of the cell, but rather delivers feed simply based on the capacitance within the bioreactor.Because of this, cell lines with very different metabolic characteristics, excessive nutrient consumption rates, or very slow growth may not be successful in this perfusion process.It is possible to redevelop perfusion feeds and alter the CapSPR setpoint to accommodate potential outlier cell lines; however, more work must be done to further explore the limits of capacitance-based feeding.
In the HID N-stage bioreactor, the nutrient feeds are controlled using Raman spectroscopy to estimate and maintain setpoint concentration of key metabolites.The adaptability of Raman feeding combined with the ability to completely automate the feeds in the HID process support the goal of achieving a platform process which requires no or minimal additional optimization to support a diverse array of clones and products.While others have previously published on the control of nutrient feeds based on amino acid estimations from Raman Spectroscopy, these case studies were limited to low density fed-batch cultures. 13,14This study presents for the first time a fully automated Raman-based feed control strategy that maintains glucose and complex nutrient feed through online measurements in a high inoculation density fed-batch platform process capable of culturing a diverse panel of clones.
This platform process, developed to boost bioreactor productivity, achieved an average 85% titer uplift when comparing a 12-day Although the unit of measure differs between published accounts, the average titer and spacetime yield increases reported range between 75% and 100% respectively (Table 3).Both Yongky et al. 19 and Mahé et al. 9 report sizable titer increases up to 200% as compared with low inoculation processes.This agrees with the highest titer increases achieved in this study (average of 170% increase) when comparing the 12-day titers of the HID and LID processes.Further, the highest titers achieved using the intensified platform presented in this study are between 9 and 10 g/L in a 12-day process, which are aligned with the highest titers reported in literature using a fed-batch paradigm. 5,9though intensification of the bioreactor process can increase titers and therefore decrease the cost of goods of the bioreactor step, the cost of goods of the drug substance is driven by the entire bioprocess, including the downstream product purification process.Two potential challenges for the downstream process associated with this intensified upstream process are increased cell densities and increased harvest titers.Notably, in the process presented in this study, the peak viable cell concentrations were similar in the LID and HID processes for the same clone (data not shown), thus did not create an increased challenge for cell removal.The total integrated viable cell concentration is elevated in the HID process; therefore, one may expect an increase in host cell protein upon harvest.The elevated titers from the HID process can create a bottleneck for the capture step.Connecting the HID process with intensified capture technologies, such as higher binding capacity resins, and simulated moving bed chromatography, have been used to increase the capture step capacity to handle increased bioreactor titers. 25Overall, although downstream bottlenecks can arise when intensifying upstream bioreactor processes, some of these can be addressed by adoption of new technologies for limiting unit operations.Continuing innovation in downstream technologies is needed to continue to keep pace with upstream titer increases.
All clones and products accessed for product quality achieved comparable aggregate levels and purity in both the HID and the control LID process indicating that the HID process does not detrimentally impact these attributes.Some differences in the charge variant profiles were observed and differences in culture duration may have an impact.Previous studies have shown that longer culture durations can lend to a higher proportion of acidic species and a lower proportion of main peak species. 26This influence of culture duration on charge variant species is in alignment with what was observed for some of the clones in this study; LID cultures, which are harvested on Day 15, showed a higher proportion of acidic charge variants and a lower proportion of main peak charge variants as compared to product harvested on Day 12 in the HID cultures.This may be driven by differences in asparagine deamidation as a result of culture conditions, but further characterization of the mAbs is necessary to confirm.
While the N-glycan profiles indicated subtle differences, with higher percentage of G0 and G0F species and a lower percentage of galactosylated species G1Fa, G1Fb, and G2F for the HID culture, this may be explained by differences in cell state on the same culture day between the HID and LID cultures.Others have reported that N-glycan species can change over time in fed-batch cultures, with a reduction in more complex N-glycans (galactosylated and sialylated species) and an increase in less complex glycans (agalactosylated and afucosylated species) at later culture durations as the cells enter stationary/death phase. 27,28Since the HID cultures are inoculated at a higher cell density, it is hypothesized that the cells reach stationary/ death phase sooner than in the LID process.Therefore, it is not surprising to have some differences in N-glycans when directly compar- Ultimately, this study demonstrates an effective perfused N-1 and HID process that substantially increases bioreactor productivity.
For the past four decades, relatively modest improvements have been made in titer production that increased fed-batch yields from >0.5 g/L in the 1980s to titers averaging 3.20 g/L in 2022. 3The proposed perfused N-1 processes coupled with a high inoculation density bioreactor presents an important leap forward in productivity advancement in an effort to simplify biopharmaceutical manufacturing, increase bioreactor productivity, and reduce COGS.

| CONCLUSION
In this work, a fully automated intensified process capable of routinely achieving high titer increases as compared to LID processes was described.The perfused N-1 stage utilized an automated capacitance feeding strategy that enabled the expansion of six cell lines expressing three products from 0.5 Â 10 6 cells/mL to 50.0 Â 10 6 cells/mL in 7-9 days.The intensified and automated HID N-stage bioreactor was able to achieve an average of 85% titer uplift or 132% STY uplift when comparing the 12-day HID process to a 15-day LID process.
Moreover, this utilized a novel feeding strategy based on in-line Raman estimations of two key amino acids and glucose, which enabled nutrient feeds to be added based on a desired nutrient setpoint.The automated feeding strategy in both the N-1 and HID stages of this process increased operational ease and aided to the success of this intensified platform in achieving repeatably high titer results for six separate clones displaying variable growth characteristics.

F
I G U R E 1 Perfusion N-1 cultures for cell lines A, B, C, D, E, and F. (a) Viable cell concentration.All cell lines achieved 40.0-60.0Â 10 6 cells/mL in 7-9 days.(b) Capacitance at 1000 kHz.(c) Capacitance-specific perfusion rate (CapSPR).All the cultures were fed at 21 Â 10 À3 CapSPR.(d) Cell-specific perfusion rate (CSPR).(e) Perfusion media usage.The average media usage was 2.73 vessel volumes, and the maximum was 3.66 vessel volumes.
tures of Clones A, C, D, and E were measured to evaluate what impact, if any, the intensified platform process had on charge variants, size variants/aggregates, reduced/non-reduced purity, and released N-glycans.The LID process typically specifies product harvest on Day 15, while product is harvested on Day 12 in the HID process.For some clones, product was harvested at different time points in the LID process so that the impact of culture duration on product quality could be assessed and direct Day 12 to Day 12 comparisons could be made between the LID and HID processes.LID product quality was measured on Day 12 for Clones C, D, and E, and on Day 15 for Clones A, C, and D. Product quality data was not available for Clone F cultured in the LID process but is shown for the HID process.

2
In-line Raman model estimations (black line), the offline values (blue dots), and feed totalizer (dotted lined) for: (a) Normalized phenylalanine concentration, (b) Normalized methionine concentration, (c) Normalized glucose concentration for a completely automated HID run using Clone G. HID culture (Day 12), there was a slightly higher proportion of acidic species and lower proportion of main peak species expressed in the LID cultures for Clones C and D. These differences are lessened when comparing product harvested on Day 12 in the LID culture to product from the HID culture, suggesting that culture duration may impact charge variant species.Basic charge variant isoforms were similar between product expressed in LID and HID culture for clones C and D, across all culture days analyzed.Product from the Clone A LID culture had a slightly higher proportion of basic species compared to the HID culture, while product from the Clone E LID culture (Day 12) had a lower proportion of basic species compared to the respective HID culture.The glycan profiles, which were evaluated for Clones C, D, E, and F, showed some slight variations between the HID and LID samples (Figure4e).The mAbs expressed in the HID process were observed to have higher percentages of agalactosylated and afucosylated species (G0 and G0F species) and a lower percentage of galactosylated species (G1F and G2F species) compared to LID cultures at Day 12 and Day 15.However, the harvest day (Day 15 for the LID culture and Day 12 for the HID culture) glycan profiles were more comparable than the Day 12 to Day 12 comparison as shown with both Clones C and D. In all samples for both the HID and LID cultures, the G0F species remained the dominant N-glycan species with G0F, F I G U R E 3 (a) Viable cell concentration, (b) viability, (c) osmolality, (d) and lactate for cell lines A, B, C, D, E, and F. (e) Titer values for HID (dark bars) and LID process measured at Day 12 and Day 15.(f) Titer uplift percentage between the HID process and the LID process at 12 days (black bars) and 15 days (striped bars).

F I G U R E 4
Product quality attributes for Clones A, C, D, E, and F in the HID process (solid bars) and a 12-day and 15-day LID control process (striped bars).Day 12 LID data available for clones C, D, and E. Day 15 LID data available for Clones A, C, D, and E. (a) Charge variants.(b) Size exclusion.(c) Reduced species.(d) Non-reduced species.(e) N-Glycans.

ing
Day 12 HID cultures to Day 12 LID cultures.For Clones C and D, the comparisons of the HID culture Day 12 and LID culture Day 15, when the cells are likely in a similar stationary/death cell state, show a more comparable N-glycan profile.
Summary of intensified N bioreactor publications.process to a 15-day LID process.That titer uplift translates into an average 132% space time yield increase, when normalizing the titers to the culture duration.This is on the upper range of what other publications have reported in terms of productivity increases. HID Olin: Conceptualization; data curation; formal analysis; investigation; writingoriginal draft; writingreview and editing.Nicolas Wolnick: Data curation; investigation; writingoriginal draft; writingreview and editing.Hunter Crittenden: Data curation; investigation; writingreview and editing.Anthony Quach: Investigation; methodology; supervision.Brian Russell: Investigation.Shannon Hendrick: Investigation.Julia Armstrong: Conceptualization; investigation; supervision; writingreview and editing.Thaddaeus Webster: