Addressing amino acid‐derived inhibitory metabolites and enhancing CHO cell culture performance through DOE‐guided media modifications

Previously, we identified six inhibitory metabolites (IMs) accumulating in Chinese hamster ovary (CHO) cultures using AMBIC 1.0 community reference medium that negatively impacted culture performance. The goal of the current study was to modify the medium to control IM accumulation through design of experiments (DOE). Initial over‐supplementation of precursor amino acids (AAs) by 100% to 200% in the culture medium revealed positive correlations between initial AA concentrations and IM levels. A screening design identified 5 AA targets, Lys, Ile, Trp, Leu, Arg, as key contributors to IMs. Response surface design analysis was used to reduce initial AA levels between 13% and 33%, and these were then evaluated in batch and fed‐batch cultures. Lowering AAs in basal and feed medium and reducing feed rate from 10% to 5% reduced inhibitory metabolites HICA and NAP by up to 50%, MSA by 30%, and CMP by 15%. These reductions were accompanied by a 13% to 40% improvement in peak viable cell densities and 7% to 50% enhancement in IgG production in batch and fed‐batch processes, respectively. This study demonstrates the value of tuning specific AA levels in reference basal and feed media using statistical design methodologies to lower problematic IMs.

biopharmaceutical industry due to their capacity to produce humancompatible monoclonal antibodies, bispecific antibodies, and other biotherapeutics. Their ability to grow in suspension cultures and ease of scale-up, capability to perform complex posttranslational modifications including human-like glycosylation, and lower susceptibility to viral contaminations are some of the major factors leading to the choice of CHO cells for bioproduction application (Dhara et al., 2018).
One of the critical aspects of bioprocess development for CHO cell cultures is the design and implementation of highly efficient biomanufacturing processes in which the nutrient requirements of the mammalian cells are identified and satisfied to ensure efficient growth and production of these valuable biologics.
CHO cells in culture grow and produce therapeutic protein of interest by relying on the nutrients provided which are processed within their metabolic networks. These nutrients include sugars, amino acids, and other less abundant nutrients and cofactors utilized by the cells for growth, maintenance, and protein production. Robust cell growth, elevated viable cell densities, and high product titers can be achieved through proper utilization of glucose and amino acids and their efficient/effective conversion into energy, biomass, and bioproducts. Unfortunately, the disposition of nutrients to energy or useful metabolites does not always occur in cell culture processes (Coulet et al., 2022;Ritacco et al., 2018;Young, 2013). Alternatively, metabolic inefficiencies are manifested in part as the formation of pathway intermediates and their accumulation and secretion as byproducts in cell culture over time (Pereira et al., 2018). Further, recent studies indicate that some of these metabolic by-products can be toxic to the cells, impacting either cell growth or protein productivity or both (Mulukutla et al., 2017). The toxicity of these secreted metabolites can be compounded when processes demand high cell densities present in fed-batch cultures. A potential consequence of this accumulation of inhibitory metabolites is the presence of a cell growth ceiling which in turn can limit the production capacity of the fed-batch process. This limitation makes growth inhibitory metabolite identification and characterization in CHO cell cultures an important step in process development.
These metabolites can emerge not just from central metabolism contributing to energy generation in CHO cells but also from inefficient utilization of amino acids and other nutrient inputs.
Incomplete amino acid catabolic pathways have been shown previously to lead to the buildup of the pathway intermediates or dead-end pathway metabolites in the culture broth surrounding the cells (Mulukutla et al., 2017). However, if the levels of these metabolites secreted into the medium can be controlled, a cleaner and more robust bioprocessing environment is possible with a minimum of waste inhibitory metabolites both intracellularly and extracellularly. Efforts to control toxic CHO cell metabolites during bioprocessing have focused primarily on two approaches: either modifying the basal and/or feed medium or genetic manipulation to alter intracellular metabolism. For example, controlling glucose concentrations and maintaining it at a reduced level throughout the fed-batch culture using HiPDOG or other methods can successfully lower the amount of lactate accumulating into the culture medium in comparison to a traditional fed-batch process (Gagnon et al., 2011).
This technique also resulted in higher growth and enhanced productivity of the cells as well. Additionally, the catabolic byproducts of tyrosine, phenylalanine, and tryptophan, which can be highly toxic to CHO cell fed-batch cultures, were reduced by lowering the source amino acid concentration by half in the basal and feed medium (Mulukutla et al., 2017). Such efforts to manipulate the basal and feed medium components can be especially beneficial in optimizing media design for high-density cultures as part of process intensification. Furthermore, cellular metabolism can be engineered so that substrates such as glucose and amino acids are directed away from problematic metabolites that lead to dead-ends, off pathways metabolites, or accumulating intermediates and into pathways that yield more effective conversion into useful downstream metabolites. For instance, genetic intervention has been implemented using siRNA to knock down or CRISPR-Cas9 to knockout upstream genes such as BCAT1 in the branched-chain amino acid (BCAA) catabolic pathway to reduce the accumulation of toxic metabolites (Mulukutla et al., 2019).
Recently, our group applied LC-MS based metabolomics on fedbatch cell culture supernatant to identify a series of metabolites secreted by CHO cells namely, CMP, NAP, ACA, TRI, HICA, and MSA (Table 1). Furthermore, several of these metabolites, found to be inhibitory to CHO cell growth and protein production, are linked to the catabolism of amino acids (Kuang et al., 2021). Controlling these inhibitory metabolites by single amino acid precursors was hypothesized to ignore the combinatorial effects arising from intertwined network of amino acid metabolism. Thus, in the present study, medium modification strategies were implemented with the help of design of experiments (DOE) that allows us to adjust multiple amino acid precursors to reduce or at least limit these inhibitory metabolites. Such a rational medium design helps determine if a more effective fed-batch protein production process could be developed in consideration of these relationships. Design of experiments has often been used in cases where multiple factors need to be screened and further optimized for achieving the desired output (Bezerra et al., 2008). The availability of multiple options in a DOE framework offers the advantage of screening from a larger precursor set with a lesser number of experiments (González-Leal et al., 2011). In the context of medium/feed optimization for cell culture applications, a T A B L E 1 List of inhibitory metabolites identified previously and considered for control strategy in the study. cell growth and product titer focused DOE has been employed successfully in previous studies (Castro et al., 1992;Rafigh et al., 2014;Torkashvand et al., 2015). Torkashvand

| Experimental plan
A two-stage experimental plan was implemented to design a control strategy for inhibitory metabolites. First, a precursor amino acids matrix contributing to the accumulation of these metabolites was constructed (Precursor study). This was done initially by understanding the amino acid consumption profiles. Then, metabolite accumulation levels were evaluated by the over-supplementation of potential precursor amino acids. Subsequently, the precursor amino acids identified in the "Precursor study" were optimized using a design of experiments methodology towards reduced metabolite accumulation (Design study). The resulting optimal amino acid levels were finally tested against a control process where these metabolites were observed to be accumulating at high concentrations.

| Batch and fed-batch process
For a control batch process, cells were inoculated in medium A supplemented with 8 mM Glutamine at a seeding cell density of 0.5 × 10 6 cells/mL in a working volume of 30 mL. This study was conducted in biological duplicates and daily cell culture samples were collected for growth measurement and extracellular metabolite analysis. Cultures were harvested when the cell viability dropped below 70%. An amino acid depleted version of medium A was kindly provided by Millipore Sigma (medium A−) and was used for DOE studies described in Section 3.3.
For a control fed-batch process, cells were inoculated in medium A supplemented with 8 mM Glutamine at a seeding cell density of 0.5 × 10 6 cells/mL in a working volume of 30 mL. Starting Day 3, a nutrient-rich feed medium B (AMBIC 1.0 reference community basal medium) was added daily to the CHO cell cultures at a rate of 10% (v/v) (Cordova et al., 2023). This study was conducted in biological duplicates and daily cell culture samples were collected for growth measurement and extracellular metabolite analysis. Glucose was supplemented up to 5 g/L to the culture when its concentration dropped below 2 g/L. Cultures were harvested when the cell viability dropped below 70%. An amino acid depleted version of feed B was kindly provided by Millipore Sigma (feed B−) and was used for DOE studies described in Section 3.3.

| Cell growth, viability, and productivity determination
Cell culture samples were collected every 24 h for estimation of the culture performance at various stages of the medium design process.
Cell growth was determined through viable cell density (VCD) measurement using a hemocytometer. Also, culture viability was measured daily using the trypan blue dye exclusion method. Cell growth rate was calculated by fitting the VCD data to an exponential function. Titer analysis was also performed via HPLC using a protein A column (Poros 2 µm, 2.1 × 30 mm; Thermofisher). To compare and evaluate the productivity of the cells in optimized media and feed, specific titer productivity was calculated which is defined as the overall measured titer (mg/L) normalized against the respective integral viable cell density (IVCD).

| Implementation study
The optimal concentrations of the significant AAs determined by the response surface design were used to reconstruct the basal and feed medium to create "reduced medium A" and "reduced feed B" for CHO cell cultures. This was done to characterize the improvements that the optimal basal and feed medium provided to the cell culture performance in batch and fed-batch mode. These improved batch and fed-batch processes were compared to the control processes (described in Section 2.3). The design study and the derived optimal amino acid concentrations were considered to be validated if these improved cultures provided cell growth, culture viability and/or product titer enhancements and most importantly, reduced the IM accumulation level in comparison to control cultures.

| Study overview
This rational medium design study was constructed in two parts such that the output of the first part (precursor study) served as the input for the second part (design study) as shown in Figure 1a. Within the precursor study, a set of potentially contributing substrate amino acids were first identified. Next, the design study served the task of screening these amino acids to shortlist the significant ones that led to higher accumulations of IMs. Once screened, the levels of this subset of amino acids were modified to lower concentrations in the basal and feed medium formulations. Finally, fed-batch experiments were conducted to finalize the basal and feed conditions that resulted in the highest growth and protein production.

| Precursor study
The first step was to determine which amino acid substrates act as precursors into metabolic pathways that result in the accumulation of inhibitory metabolites. Previously, we determined how some of these amino acids can map to their respective by-products through analysis of CHO metabolic pathways available in the KEGG database (Kanehisa, 2000). However, cellular metabolism is a complex network of multiple interwoven reactions wherein each metabolite can emerge from multiple substrates. This makes it difficult to pinpoint a specific source leading to the accumulation of a particular metabolite. This motivated us to evaluate the effects that result from a combination of multiple substrates leading to accumulation of different inhibitory metabolites. As a result, a precursor study was undertaken to identify which individual, and combination of substrates should be considered and manipulated in the culture medium for improved growth and productivity. A simplified decision tree for the precursor study described in Figure 1b summarizes the various tasks leading up to the selection of precursor amino acids for the design study.

| Cell growth and amino acid consumption in batch culture
To evaluate the amino acid consumption rates, a control batch growth study was performed (Supporting Information: Figure S1).
The batch process ran for 7 days during which samples were collected every 24 h for measuring the viable cell density (VCD) and monitoring the metabolite levels. As shown in Supporting Information: Figure S1a,b, cell viability remained above 95% percent until Day 5 during which the cells attained a maximum cell density of 12.7 × 10 6 cells/mL. Cell growth stopped after day 5 together with a sharp decline in cell viability to less than 80% on Day 6 and 0% on were reduced to less than 30% or more of the starting levels while three others (Glu, Ala, and Gly) accumulated over the culture period.
Glutamate is derived from glutamine metabolism (for most nonglutamine synthetase mammalian cell lines as used here). Glutamine, typically the second-most preferred carbon source in the CHO-K1 cell line, was completely exhausted in this batch study (Nicolae et al., 2014). Glutamine also, directly and indirectly, contributes to the biosynthesis of multiple other NEAAs including Glu, Asn, Ala, Asp, Pro, and Ser (Grinde et al., 2019;Kalhan & Hanson, 2012;Yoo et al., 2020). Asparagine was completely exhausted during the culture as well, while more than 75% aspartate in the media was also consumed. The high levels of serine consumption may be related to its interconversion into glycine inside the cells. Alternatively, the accumulation of Gly and Ala may be due to their formation as byproducts of glycolysis while Ala may also form as a result of transamination reactions. Lastly, 60% of initially present tyrosine was gradually consumed by Day 6. Previous studies have shown that tyrosine plays a fundamental role in CHO cell biomass production F I G U R E 1 (a) Medium design guided by statistical design of experiments, (b) process flowchart for study of the amino acid precursors that contribute to the accumulation of inhibitory metabolites followed by the design of experiments approach to determine the optimal amino acid levels to enhance CHO cell performance. Previous studies have shown that catabolism of essential amino acids can lead to the formation of growth inhibitory metabolic byproducts (Ley et al., 2019;Mulukutla et al., 2017). These by-products can also be secreted by the cells into the extracellular environment and then exert an inhibiting impact on cell performance. Therefore, as a next step in the precursor study, we sought to elucidate amino acids linked to the accumulating inhibitory metabolites we identified in a previous study (Kuang et al., 2021).

| Identification of precursor amino acids
Previously, we identified 6 toxic metabolites, namely CMP, NAP, ACA, TRI, HICA, and MSA, listed in Table 1 that are inhibitory to growth of CHO cells in culture (Kuang et al., 2021) The accumulation of metabolic by-products is governed not just by their primary amino acid source, but also by other potentially interacting amino acids that feed into particular pathways. A list of primary and secondary substrates for these six toxic metabolites was compiled on the basis of whether or not they feed into a specific inhibitory metabolite ( Figure 3a). In addition to these specific amino acid sources, other amino acids were added to the list of contributing amino acids as secondary substrates based on their potential to contribute into the pathway though other metabolites. For example, CMP and ACA are derived from the purine and pyrimidine metabolic pathway which predominantly has glucose as the contributing substrate with glutamine being the next dominant source. We added arginine, alanine, and asparagine to the substrate set since they can indirectly contribute to the accumulation of CMP and ACA (Brosnan & Brosnan, 2007;Hoang et al., 2022;Kim et al., 2011;Kuang et al., 2022;Saas et al., 2000). In addition to these precursors, KEGG Scale shown in the heat map refers to 1% of amino acid level in culture from left end (orange) to 1000% of amino acid level toward the right end (blue). n = 2 (refer to Supporting Information: Table S1A and S1B for numerical data).

| Design study
Following the initial evaluation of the amino acid precursors, a Design of Experiments (DOE) approach was implemented in three stages. The first stage involved a screening design to rank the amino acid precursors in order of their contribution toward buildup of the inhibitory metabolites. This would allow us to narrow down the list of amino acids and focus on the most significant contributors. These most significant contributors were then taken forward for a second stage of DOE analysis that utilized a response surface design (RSD) approach to find the potential optimal levels of the most relevant contributors. Finally, stage 3 of the study involved the validation of our optimized AA concentrations from the previous DOE studies with the end goal of evaluating its impact on cell growth and product titers. Figure 1b summarizes the overall methodology followed for the design study to determine the optimal AA concentrations.

| Screening design
A two-level screening experiment was performed to identify the amino acids that contributed the most towards accumulating inhibitory metabolites. The precursor matrix with 10 different amino acids shown in Figure 3 was used as input for constructing the screening design. An additional 11th amino acid (Ala) was added as a dummy variable to estimate possible error in the design experiment.
Unlike the preliminary study, where the amino acids were increased by two and three times of the basal medium level, this two-level screening design matrix included a high (+1) and a low (−1) value for each amino acid substrate. Since the goal of the screening design was to reduce amino acid concentrations, basal medium levels were designated as the high (+1) level for each amino acid in the design space. For determining the low (−1) level for the amino acids, the consumption profile of each amino acid (Figure 2) was used as input to estimate how much an amino acid can be reduced with minimal impact on growth. The concentration of the dummy variable (11th factor) was kept constant for both cases to help quantify any variability and eliminate any bias in the design. The two levels chosen for each amino acid in this screening design are listed in Figure 4.
The nonessential amino acids were in general tested at a range between 25% and 75% of the nominal level for their low-level value.
In choosing the low (−1) levels of the essential amino acids, the maximum reduction was 50% to avoid any detrimental effects on growth due to potential exhaustion in the basal medium. However, leucine and arginine were only reduced by 25% to minimize the AAs were grouped together and considered for the first stage Box-Behnken DOE (herein referred to as BB1). This simplification of the design space reduced the conditions for testing to only 13, allowing ample capacity to conduct these experiments in replicate. The optimized levels of AAs levels from the first BB1 statistical design were then taken forward to optimize the levels of Arg and Trp in the basal medium using a second stage Box-Behnken design, BB2.
The cell culture studies for both stages of the Box-Behnken optimization were conducted in CHO-K1 batch cultures. The design matrix for the first step (Stage 1) of this study is shown in Figure 6a.
Two additional control conditions, "complete (basal) medium" (with "+1" levels of all AAs) and "negative control" (with "−1" levels of all AAs), were added to the 13-case design matrix to facilitate the comparison of test cases with extreme scenarios. The amino acids being tested for optimization were added back to Medium A− at appropriate concentrations (−1, 0, +1; with specific levels noted in Information: Table S5. The results from the 3 concentration levels (+1, 0, and −1) of Leu, Ile and Lys tested for the BB1 test are shown with the predicted "semi-optimal" levels in Figure 6b. Since the results from BB1 study were used as inputs for a second level of optimization in the BB2 study, the criteria for choosing the first stage "semi-optimal" levels of Leu, Ile, and Lys (from the BB1 study) were not kept highly stringent to eliminate any potential detrimental effects due to nutrient deprivation. For example, the desirable value of Leu arising from BB1 DOE optimization for each metabolite appearing within the "−1" and "0" level of concentrations tested for the BB1 study were averaged, giving us 86% of the "+1" value as the first stage "semi-optimal" prediction ( Figure 6b). For Ile, a value of 87% (Table 2) was chosen as the first stage "semi-optimal" concentration ( Figure 6b) since this represented the lowest critical value (none of the critical values fell within the "−1" and "0" range of concentrations tested). Lastly, for Lys, the second lowest critical value (76% of the original concentration, Figure 6b) was selected as first stage "semi-optimal" level. The lowest critical value of 54% was disregarded because the AA consumption profiles for Medium A ( Figure 2) indicated that half of the initial Lys is consumed during the batch culture and thus reducing Lysine concentration to such a low value (54%) could have resulted in complete exhaustion of this AA.
Therefore, the criteria for selection of initial optimal concentrations (a) (b) F I G U R E 6 (a) Box-Behnken design matrix 1 with different amino acids, 15 conditions were tested with varying levels (+1: high, 0: medium and −1: low) of Leu, Ile, and Lys (B) Concentration levels of Leu, Ile, and Lys tested for Box-Behnken basal medium optimization 1 (BB1) and the "semi-optimal" values chosen for stage 2.
T A B L E 2 Optimal levels of amino acids derived from the two levels of Box-Behnken optimization study (BB1 and BB2). Note: These optimal values were used for subsequent implementation study.
of the AA factors were kept generous to avoid any detrimental effects due to potential nutrient depletion. It should be noted that any method including the Box-Behnken design can exhibit limitations in mapping out the appropriate response surface. An alternative to Box-Behnken in the space of response surface methodology is central composite design which is a traditional fractional factorial design. This approach could also be beneficial in relatively unknown processes (Kazemian et al., 2021). Future studies should consider multiple complementary approaches in determining the ideal concentrations of target amino acids or other substrates that limit IM accumulation.
The first stage "semi-optimal" concentrations of Leu, Ile, and Lys from the BB1 study were then used when reconstituting the culture media Medium A− for the BB2 study. Shown in Figure 7a,b are the design matrix and the three concentration levels of Trp and Arg used for the BB2 optimization. The additional "complete (basal) medium" (with "+1" levels of all AAs) was added to serve as a positive control for this second stage. A statistical approach similar to the BB1 study was followed for determination of optimal levels of Trp and Arg in the basal medium. The "optimal" values of Trp and Arg were then selected based on the lowest critical values (Supporting Information: Table S6; see Materials and Methods) within the range of concentrations tested arising out of the BB2 optimization study. These results suggested that the levels of Trp and Arg in Medium A could be reduced to 67% and 75% of their original levels to lower the accumulation of inhibitory metabolites ( Figure 7b). Following completion of this second stage, the optimal values of AAs from the BB1 study were revisited as a final step to finalize the target levels which would subsequently be used in the following validation study. In the case of Leu, the optimal value was retained at its lowest acceptable critical value, 80% of the original concentration, as indicated in Table 2. The optimal levels of Ile and Lys were kept at 87% and 76%, respectively based on their solutions from BB1.
Overall, the design concentrations of significant AA substrates in basal Medium A were lowered to a level predicted based on the response surface methodology resulting from their impact on the inhibitory metabolites.

| Implementation study
The amino acid levels predicted from this rational medium design were used to formulate a "reduced medium A" from Medium A−. Reduced medium A was then used in a batch cell culture experiment to compare the performance of cells in the reduced medium against the control medium (Medium A, or complete medium A). Clear differences were observed in the growth and protein titers ( Figure 8a,b), as the cells cultured in the reduced A medium attained 13% improvement in peak viable cell density and 7% increase in IgG titer. The cells in the reduced medium A grew at the same rate as the complete medium A until Day 3 as indicated in Figure 8c. However, the growth rate of cells started increasing in the reduced medium A condition starting Day 4 and was 7% higher from Day 4 to Day 5. This could be due to the reduced availability of amino acids that are catabolized into dead end pathway intermediates with potential toxic effects on the cellular performance.
Next, we tested the performance of reducing medium components in a fed-batch process (all conditions listed in Supporting Information: Table S7). The control fed-batch (Condition 1) was reproduced as described in Section 2.3. To test the effect of reduced medium, the basal medium of the fed-batch process (complete medium A) was replaced with reduced medium A (Condition 2) with lowered amino acid levels ( Figure 9a). CHO cells cultured in the reduced basal medium (condition 2) maintained higher VCD a from Day 5 onward until the final day of culture (15), attaining an 10% higher peak VCD (Figure 9a (i)) and an 8% increase in titer as compared to condition 1 (Figure 9a (ii)).  Table S8). MSA in the spent medium for condition 1, 2, 3, and 4. Spent medium analysis was done for supernatants collected on Day 9 and 13 of the cell culture. Mean ± SD; n = 2, *p < 0.1. Statistics by unpaired two-tailed student t test against condition 1 (refer Supporting Information: Table S9). CHO, Chinese hamster ovary.
Furthermore, we wanted to examine if these concepts of reduced amino acids could also be applied to the feed medium (feed B). The concentrations of amino acids in standard feed B were designed to be five times higher than those of basal medium A. We in turn implemented a "reduced feed B" in which the five target amino acids, leucine, isoleucine, lysine, tryptophan, and arginine, were added at 5x levels present in "reduced medium A." Hence, condition 3 tested this effect by replacing feed B with reduced feed B while keeping the same basal medium (complete medium A). We observed that this resulted in an 18% increase in peak VCD (Figure 9b (i)).
However, this improvement in cell growth was not sustained, as the VCD for condition 3 dropped to the same level as condition 1 on Day 10, indicating the possibility of nutrient depletion due to a reduction in amino acids in feed B. Nonetheless, condition 3 achieved an 11% titer increase as compared to condition 1 (Figure 9b (ii)). To test the combination effect of reducing the AA levels in both basal and feed media, condition 4 replaced both the complete basal A and feed B with their reduced versions. For this condition, the cells did not attain higher peak VCD but did maintain an extended stationary phase from Day 8 to Day 12 as compared to condition 1. The 17% increase in the IgG titer for condition 4 could potentially reflect this enhanced stability since CHO cells are often more productive during stationary phase Rish et al., 2022;Sellick et al., 2015). In fact, lower nutrient availability may have altered cell metabolism, shifting from the growth phase starting on Day 8 and causing the sustained production phase for IgGs in condition 4.
To examine the inhibitory metabolite accumulation in the fedbatch culture conditions tested above, the secreted target metabolite concentrations were measured on Day 9 and Day 13 of the process and plotted as metabolite accumulation on a per cell basis (qMetabolite) as shown in Figure 9c. We observed a reduction in HICA and NAP accumulation in all three fed-batch culture conditions with a 10%-50% decline on Day 9 and Day 13 compared to condition 1 (control). MSA levels were also reduced by 30% and 15% in condition 2 and 3 on Day 13 while they remained relatively unchanged in condition 4. In addition, CMP levels were only slightly reduced between 5% and 15% in conditions 2 and 3, at Day 13, perhaps due to the importance of CMP in many metabolic processes.
Overall, the rational medium design approach adopted in this study was successful in achieving reduction in the accumulation of 4 out of 6 inhibitory metabolites under specific media or feed conditions by modulating the amino acid levels. One metabolite, ACA, a TCA cycle related metabolite, previously suggested for tracking the TCA cycle activity, correspondingly remained constant or increased slightly in conditions 2, 3, and 4, perhaps due to enhanced TCA cycle activity.
Unlike the other metabolites, TRI was not controlled well with the amino acid optimization strategy likely because the source of TRI also comes from vitamins. Some metabolites will not be controlled well if their sources are not solely due to amino acid catabolic pathways.
Lastly, the observed extension in stationary phase as a result of lowered nutrient availability in condition 4 motivated us to test an alternative feed rate strategy wherein the amount of feed B added to the cultures was reduced from a daily rate of 10% (v/v) to 5% (v/v).
Hence, conditions 5 and 6 were tested with the reduced feeding strategy but mimicking the same medium and feed combinations as conditions 2 and 4, respectively. Interestingly, conditions 5 and 6 led to the greatest enhancement in cell growth (Figure 10a). Reducing the feeding volume prolonged the stationary phases of the cultures until Day 12 as opposed to Day 10 in condition 1 (control). Both the conditions (5 and 6) achieved more than 35% and 40% increments in peak VCDs as well as almost 50% enhancement in IgG titers ( Figure 10b). Furthermore, these conditions rendered an earlier shift F I G U R E 10 Cell growth characteristics, protein yields, lactate profile and inhibitory metabolite accumulation in the fed-batch implementation study for condition 1, 5, and 6. (a) Viable cell density of CHO cells, [B] end of culture IgG titers, (c) concentration of lactate in the spent medium, and (d) specific inhibitory metabolite concentration (per cell accumulation) of (i) HICA, (ii) NAP, (iii) CMP, and (iv) MSA in the spent medium. Spent medium analysis was done for supernatants collected on Day 9 and 13 of the cell culture. Mean ± SD; n = 2, *p < 0.1. Statistics by unpaired two-tailed student t test against condition 1 (refer Supporting Information: Table S9). CHO, Chinese hamster ovary; CMP, cytidine monophosphate; NAP, N-acetyl putrescine.
| 2555 in lactate metabolism from production to consumption as compared to condition 1 (even as the cells were still growing) to provide evidence of changes in the cell metabolism due to differences in nutrient availability (Figure 10c), which aligns with previous studies associating the lactate metabolic shift to higher TCA cycle activity and higher protein producing fed-batch processes (Mulukutla et al., 2015;Templeton et al., 2013). This improved cell growth and protein titer phenotype also translated to reduction in accumulation of select inhibitory metabolite levels ( Figure 10d). Similar to other modified conditions, HICA and NAP had significant (p < 0.1) 80% and 60% declines in their accumulation levels on Day 13, respectively.

| CONCLUDING REMARKS
CHO cells in culture can accumulate problematic inhibitory metabolites. In the current project, we first demonstrated that these inhibitory metabolites can be linked to specific amino acids substrates through metabolic pathway analysis and an accompanying amino acid over-supplementation study. Specifically, we observed that increasing the initial supplemented amount of precursor amino acids up to 200% results in higher accumulation of inhibitory metabolites identified in a previous study. Next, we implemented multilevel statistical DOE analysis to determine the five most critical amino acid targets linked to these inhibitory metabolites in an initial screening study, and secondly to predict how much these amino acids should be reduced to limit inhibitory metabolite accumulation in cultures. We observed that reductions in the starting concentrations of Leu, Ile, Lys, Trp, and Arg were especially impactful on the inhibitory metabolites, HICA, NAP, CMP, and MSA. Finally, we applied these predictions to modulate the levels of the five target amino acids in a DOE-driven, rational modification of culture basal and feed medium.
Implementation of these modified basal and feed media with reduced amino acid levels resulted in improved overall cell culture performance along with a reduction in inhibitory metabolites in both batch and fed batch conditions to demonstrate that "less can be better" in media design. Indeed, we saw enhancements in the viable cell densities and, in some cases, extension of the more productive stationary phase with improvements in the overall IgG titers, by reducing initial levels of select amino acids in the basal media and/or feeds by up to 1/3. Thirteen and 7% improvements in peak VCD and IgG titer, respectively, were observed for a batch CHO process.
When translated to a fed-batch process, these enhancements varied depending on the combination of basal and feed media suggesting intertwined relationships between nutrient utilization and cell metabolism during different fed-batch phases. Indeed, the inhibitor metabolites HICA and NAP were reduced by up to 50%, MSA by 30%, and CMP by 15% in cultures with modified media. Interestingly, the most significant impact on culture performance was evident when modifications in the basal and fed media were combined with a reduction in the overall feed rate, which lowered the amino acids available to the cells in culture over the extended stationary phase and increased VCD and antibody titers by 40% and 50%, respectively. In this way, the study demonstrates the danger of overfeeding and the value of tuning substrate levels in media formulation based on cellular metabolism to lower problematic byproducts while enhancing cellular performance for production of recombinant proteins. Furthermore, the advent of defined reference media formulation such as AMBIC 1.0 and its descendants will serve to create a knowledgebase around how to build a more effective basal and feed media formulation for the entire cell culture community. Studies such as this one will enable researchers and users to better understand the relationships between amino acid and other media components and problematic inhibitory metabolites and metabolism in general that will yield greater insights on how basal and feed media formulation can be adjusted to improve productivities and overall culture performance in the future. supervision; project administration; funding acquisition.