Leveraging machine learning to dissect role of combinations of amino acids in modulating the effect of zinc on mammalian cell growth.

Although the contributions of individual components of cell culture media are largely known, their combinatorial effects are far less understood. Experiments varying one component at a time cannot identify combinatorial effects, and analysis of the large number of experiments required to decipher such effects is challenging. Machine learning algorithms can help in the analysis of such datasets to identify multi-component interactions. Zinc toxicity in vitro is known to change depending on amino acid concentration in the extracellular medium. Multiple amino acids are known to be involved in this protection. Thirty-two amino acid compositions were formulated to evaluate their effect on the growth of CHO cells under high zinc conditions. A sequential machine learning analysis methodology was used, which led to the identification of a set of amino acids (threonine, proline, glutamate, aspartate, asparagine, and tryptophan) contributing to protection from zinc. Our results suggest that a decrease in availability of these set of amino acids due to consumption may affect cell growth in media formulated with high zinc concentrations, and in contrast, normal levels of these amino acids are associated with better tolerance to high zinc concentration. Our sequential analysis method may be similarly employed for high throughput medium design and optimization experiments to identify interactions among a large number of cell culture medium components.

Amino acid abundance and composition in cell culture medium affects trace metal tolerance and cholesterol synthesis.

Amino acid compositions of cell culture media are empirically designed to enhance cell growth and productivity and vary both across media formulations and over the course of culture due to imbalance in supply and consumption. The interconnected nature of the amino acid transporters and metabolism suggests that changes in amino acid composition can affect cell physiology. In this study, we explore the effect of a step change in amino acid composition from a DMEM: F12-based medium to a formulation varying in relative abundances of all amino acids, evaluated at two amino acid concentrations (lean LAA vs. rich HAA). Cell growth was inhibited in LAA but not HAA. In addition to the expected effects on expression of the cell cycle, amino acid response and mTOR pathway genes in LAA, we observed an unanticipated effect on zinc uptake and efflux genes. This was accompanied by a lower tolerance to zinc supplementation in LAA but not in the other formulations. Histidine was sufficient but not necessary to prevent such zinc toxicity. Additionally, an unanticipated downregulation of genes in the cholesterol synthesis pathway was observed in HAA, accompanied by an increase in cellular cholesterol content, which may depend on the relative abundances of glutamine and other amino acids. This study shows that changes in the amino acid composition without any evident effect on growth may have profound effects on metabolism. Such analyses can help rationalize the designing of medium and feed formulations for bioprocess applications beyond replenishment of consumed components.

Engineering nucleotide sugar synthesis pathways for independent and simultaneous modulation of N-glycan galactosylation and fucosylation in CHO cells.

Glycosylation of recombinant therapeutics like monoclonal antibodies (mAbs) is a critical quality attribute. N-glycans in mAbs are known to affect various effector functions, and thereby therapeutic use of such glycoproteins can depend on a particular glycoform profile to achieve desired efficacy. However, there are currently limited options for modulating the glycoform profile, which depend mainly on over-expression or knock-out of glycosyltransferase enzymes that can introduce or eliminate specific glycans but do not allow predictable glycoform modulation over a range of values. In this study, we demonstrate the ability to predictably modulate the glycoform profile of recombinant IgG. Using CRISPR/Cas9, we have engineered nucleotide sugar synthesis pathways in CHO cells expressing recombinant IgG for combinatorial modulation of galactosylation and fucosylation. Knocking out the enzymes UDP-galactose 4'-epimerase (Gale) and GDP-L-fucose synthase (Fx) resulted in ablation of de novo synthesis of UDP-Gal and GDP-Fuc. With Gale knock-out, the array of N-glycans on recombinantly expressed IgG is narrowed to agalactosylated glycans, mainly A2F glycan (89%). In the Gale and Fx double knock-out cell line, agalactosylated and afucosylated A2 glycan is predominant (88%). In the double knock-out cell line, galactosylation and fucosylation was entirely dependent on the salvage pathway, which allowed for modulation of UDP-Gal and GDP-Fuc synthesis and intracellular nucleotide sugar availability by controlling the availability of extracellular galactose and fucose. We demonstrate that the glycoform profile of recombinant IgG can be modulated from containing predominantly agalactosylated and afucosylated glycans to up to 42% and 96% galactosylation and fucosylation, respectively, by extracellular feeding of sugars in a dose-dependent manner. By simply varying the availability of extracellular galactose and/or fucose, galactosylation and fucosylation levels can be simultaneously and independently modulated. In addition to achieving the production of tailored glycoforms, this engineered CHO host platform can cater to the rapid synthesis of variably glycoengineered proteins for evaluation of biological activity.

Mathematical model of the multi-amino acid multi-transporter system predicts uptake flux in CHO cells.

Supply and uptake of amino acids is of great importance to mammalian cell culture processes. Mammalian cells such as Chinese hamster ovary (CHO) cells express several amino acid (AA) transporters including uniporters and exchangers. Each transporter transports multiple AAs, making prediction of the effect of changed medium composition or transporter levels on individual AA transport rate challenging. A general kinetic model for such combinatorial amino acid transport, and a simplified analytical expression for the uptake rate as a function of amino acid concentrations and transporter levels is presented. From this general model, a CHO cell-specific AA transport model, to our knowledge the first such network model for any cell type, is constructed. The model is validated by its prediction of reported uptake flux and dependencies from experiments that were not used in model construction or parameter estimation. The model defines theoretical conditions for synergistic/repressive effect on the uptake rates of other AAs upon external addition of one AA. The ability of the CHO-specific model to predict amino acid interdependencies experimentally observed in other mammalian cell types suggests its robustness. This model will help formulate testable hypotheses of the effect of process changes on AA initial uptake, and serve as the AA transport component of kinetic models for cellular metabolism.

Superfluous glutamine synthetase activity in Chinese Hamster Ovary cells selected under glutamine limitation is growth limiting in glutamine-replete conditions and can be inhibited by serine.

Passaging and expansion of animal cells in lean maintenance medium could result in periods of limitation of some nutrients. Over time, such stresses could possibly result in selection of cells with metabolic changes and contribute to heterogeneity. Here, we investigate whether selection of Chinese Hamster Ovary (CHO) cells under glutamine limitation results in changes in growth under glutamine-replete conditions. In glutamine-limiting medium, compared to control cells passaged in glutamine-rich medium, the selected cells showed higher glutamine synthetase (GS) activity and attained a higher peak viable cell density (PVCD). Surprisingly, in glutamine-replete conditions, selected cells still showed a higher GS activity but a lower PVCD. We show that in glutamine-replete medium, PVCD of selected cells was restored on (a) inhibition of GS activity with methionine sulfoximine, (b) supplementation of aspartate-without affecting GS activity, and (c) supplementation of serine, which is reported to inhibit GS in vitro. Consistent with the reported effect of serine, inhibition of GS activity was observed upon serine supplementation along with reduced growth of cells under glutamine-limiting conditions. The latter observation is important for the design of glutamine-free culture medium and feed used for GS-CHO and GS-NS0. In summary, we show that CHO cells selected under glutamine limitation have superfluous GS activity in glutamine-replete medium, which negatively affects their PVCD. This may be due to its effect on availability of aspartate which was the limiting nutrient for the growth of selected cells in glutamine-replete conditions.

Nickel and cobalt affect galactosylation of recombinant IgG expressed in CHO cells.

Glycosylation is an important product quality attribute of antibody biopharmaceuticals. It involves enzymatic addition of oligosaccharides on proteins by sequential action of glycosyltransferases and glycosidases in the endoplasmic reticulum and golgi. Some of these enzymes like galactosyltransferase and N-acetylglucosaminyltransferase-I require trace metal cofactors. Variations in trace metal availability during production can thus affect glycosylation of recombinant glycoproteins such as monoclonal antibodies. Variability in trace metal concentrations can be introduced at multiple stages during production such as due to impurities in raw materials for culture medium and leachables from bioreactors. Knowledge of the effect of various trace metals on glycosylation can help in root-cause analysis of unintended variability in glycosylation. In this study, we investigated the effect of nickel and cobalt on glycosylation of recombinant IgG expressed in Chinese hamster ovary cells. Nickel concentrations below 500 µM did not affect glycosylation, but above 500 µM it significantly decreases galactosylation of IgG. Cobalt at 50 µM concentration causes slight increase in G1F glycans (mono galactosylated) as previously reported. However, higher concentrations result in a small increase in G0F (non galactosylated) glycans. This effect of nickel and cobalt on galactosylation of recombinant IgG can be reversed by supplementation of uridine and galactose which are precursors to UDP-Galactose, a substrate for the enzymatic galactosylation reaction.

Zinc supplementation decreases galactosylation of recombinant IgG in CHO cells.

Trace element composition of culture medium can be altered to modulate glycoform of recombinant glycoproteins. In this study, we show that Zn2+ supplementation at or above 100 µM decreases galactosylation of recombinant IgG expressed in Chinese Hamster Ovary cells. This decrease in galactosylation is not due to reduced galactosyltransferase expression. This effect persists upon supplementation of galactose and uridine to the culture, indicating that it may not be due to reduced UDP-Gal availability. Measurements of galactosyltransferase activity in the cell lysate show that activity decreases with increasing Zn2+/Mn2+ ratio. This suggests that one possible explanation of the effect of Zn2+ may be reduced intracellular galactosyltransferase activity due to increase in Zn2+/Mn2+ ratio. Consistent with this, the decrease in galactosylation of IgG could be reversed by supplementation of Mn2+ (a cofactor of galactosyltransferase) which increases intracellular Mn2+ content. Measurement of total intracellular Zn2+ content, however, indicates no significant upregulation of total intracellular Zn2+ content and no significant downregulation of intracellular Mn2+ content with Zn2+ supplementation. One possible explanation could be that cellular detoxification response to higher extracellular Zn2+ concentration might lead to changes in intracellular distribution of Mn2+. In this case, Zn2+ supplementation would be expected to interfere with other known effects of Mn2+. Indeed, the previously reported increase in high mannose glycans upon Mn2+ supplementation in the absence of glucose is reversed by Zn2+ supplementation. This study also suggests the use of Mn2+ supplementation as a strategy to overcome the effect of lot-to-lot variability in trace element concentrations on galactosylation.

CHO Cells adapted to inorganic phosphate limitation show higher growth and higher pyruvate carboxylase flux in phosphate replete conditions.

Inorganic phosphate (Pi ) is an essential ion involved in diverse cellular processes including metabolism. Changes in cellular metabolism upon long term adaptation to Pi limitation have been reported in E. coli. Given the essential role of Pi , adaptation to Pi limitation may also result in metabolic changes in animal cells. In this study, we have adapted CHO cells producing recombinant IgG to limiting Pi conditions for 75 days. Not surprisingly, adapted cells showed better survival under Pi limitation. Here, we report the finding that such cells also showed better growth characteristics compared to control in batch culture replete with Pi (higher peak density and integral viable cell density), accompanied by a lower specific oxygen uptake rate and cytochrome oxidase activity towards the end of exponential phase. Surprisingly, the adapted cells grew to a lower peak density under glucose limitation. This suggests long term Pi limitation may lead to selection for an altered metabolism with higher dependence on glucose availability for biomass assimilation compared to control. Steady state U-13 C glucose labeling experiments suggest that adapted cells have a higher pyruvate carboxylase flux. Consistent with this observation, supplementation with aspartate abolished the peak density difference whereas supplementation with serine did not abolish the difference. This supports the hypothesis that cell growth in the adapted culture might be higher due to a higher pyruvate carboxylase flux. Decreased fitness under carbon limitation and mutations in the sucABCD operon has been previously reported in E. coli upon long term adaptation to Pi limitation, suggestive of a similarity in cellular response among such diverse species. © 2017 American Institute of Chemical Engineers Biotechnol. Prog., 33:749-758, 2017.

Shear stress increases cytotoxicity and reduces transfection efficiency of liposomal gene delivery to CHO-S cells.

Animal cells in suspension experience shear stress in different situations such as in vivo due to hemodynamics, or in vitro due to agitation in large-scale bioreactors. Shear stress is known to affect cell physiology, including binding and uptake of extracellular cargo. In adherent cells the effects of exposure to shear stress on particle binding kinetics and uptake have been studied. There are however no reports on the effect of shear stress on extracellular cargo delivery to suspension cells. In this study, we have evaluated the effect of shear stress on transfection of CHO-S cells using Lipofectamine 2000 in a simple flow apparatus. Our results show decreased cell growth and transfection efficiency upon lipoplex assisted transfection of CHO-S while being subjected to shear stress. This effect is not seen to the same extent when cells are exposed to shear stress in absence of the lipoplex complex and subsequently transfected, or if the lipoplex is subjected to shear stress and subsequently used to transfect the cells. It is also not seen to the same extent when cells are exposed to shear stress in presence of liposome alone, suggesting that the observed effect is dependent on interaction of the lipoplex with cells in the presence of shear stress. These results suggest that studies involving liposomal DNA delivery in presence of shear stress such as large scale transient protein expression should account for the effect of shear during lipoplex assisted DNA delivery.

Cell line development for biomanufacturing processes: recent advances and an outlook.

At the core of a biomanufacturing process for recombinant proteins is the production cell line. It influences the productivity and product quality. Its characteristics also dictate process development, as the process is optimized to complement the producing cell to achieve the target productivity and quality. Advances in the past decade, from vector design to cell line screening, have greatly expanded our capability to attain producing cell lines with certain desired traits. Increasing availability of genomic and transcriptomic resources for industrially important cell lines coupled with advances in genome editing technology have opened new avenues for cell line development. These developments are poised to help biosimilar manufacturing, which requires targeting pre-defined product quality attributes, e.g., glycoform, to match the innovator's range. This review summarizes recent advances and discusses future possibilities in this area.

Manganese increases high mannose glycoform on monoclonal antibody expressed in CHO when glucose is absent or limiting: Implications for use of alternate sugars.

Alternate sugars such as galactose and fructose are metabolized at a slower rate than glucose and result in lower accumulation of lactate. While low lactate accumulation is desirable, we report that complete substitution of glucose with these sugars results in an increase in M5 high mannose glycans. Surprisingly, this increase is much higher when the culture is supplemented with manganese: for example, when cells are cultured with galactose, M5 high mannose glycan content increased from 5% at 1 nM Mn(2+) in the basal medium to 32% with 16 µM Mn(2+) supplementation. When galactose is supplemented with glucose maintained at low concentrations, a small reduction in high mannose glycans is seen. In control cultures with glucose, the high mannose content was however <2% in this range of Mn(2+) concentration. By varying Mn(2+) and glucose supplementation levels, with or without galactose, we systematically demonstrate that Mn(2+) concentration and glucose availability, together, significantly affect the high mannose glycan content. To our knowledge, this is the first report that shows that the effect of Mn(2+) on high mannose glycan content depends on glucose availability. At each Mn(2+) supplementation level evaluated, galactosylation percentages were highest for cultures where galactose was supplemented with glucose at non-limiting concentration.

In situ pH management for microbial culture in shake flasks and its application to increase plasmid yield.

Shake flasks are widely used to culture microorganisms, but they do not allow for pH control without additional infrastructure. In the presence of a carbon source like glucose, culture pH typically decreases due to overflow metabolism and can limit the growth of microorganisms in shake flasks. In this study, we demonstrate the use of magnesium hydroxide-loaded pH managing hydrogels (m-pHmH) for in situ base release to counter the decrease in culture pH in shake flasks using Escherichia coli as a model organism, in both complex and mineral salts medium. Base release from m-pHmH is shown to increase with decreasing pH (22-fold increase in release rate from pH 8 to 5), thus providing feedback from culture pH. The addition of m-pHmH resulted in better pH maintenance and higher biomass yields of E. coli K12 in media containing glucose as a carbon source. The use of m-pHmH with additional buffer resulted in pH being maintained above 6.9 while pH decreases below 5 without m-pHmH. We demonstrate one application of such in situ pH management to increase the volumetric plasmid yield from E. coli in shake flask culture. In situ glucose release through a hydrogel to mimic fed-batch culture along with the addition of m-pHmH resulted in a 395 % increase in volumetric plasmid yield to 38 µg/ml in shake flask culture.

Comparison of methods for identifying periodically varying genes.

Several methods have been reported for identifying periodically varying genes from gene expression datasets. We compare the performance of five existing methods and a combination of G-statistic and autocovariance (called GVAR) using simulated sine-function-based and cell-cycle-based datasets. Based on this analysis we recommend appropriate methods for different experimental situations (length of the time series, sampling interval and noise level). No single method performs the best under all tested conditions. None of the evaluated methods perform well at high noise levels for short time series data. At lower noise levels, GVAR performed the best.

Meta analysis of gene expression changes upon treatment of A549 cells with anti-cancer drugs to identify universal responses.

A meta-analysis of publicly available gene expression changes in A549 cells upon treatment with anti-cancer drugs is reported. To reduce false positives, both fold-change and significance level cutoffs were used. Simulated datasets and permutation analysis were used to guide choice of ratio cutoff. Of the genes identified, FDXR is the only gene differentially expressed in six of the seven drug treatments. Though FDXR has been reported to be differentially expressed upon treatment with 5-fluorouracil and its expression correlated to long term disease survival, to our knowledge this is a first study implicating a wide effect of anti-cancer drug treatment on FDXR expression. The other genes identified which are differentially expressed in four out of the seven drug treatments are CDKN1A and PARVB which are upregulated and MYC, HBP1, LDLR, SIM2, ALX1 and GPHN which are downregulated.

In situ pH maintenance for mammalian cell cultures in shake flasks and tissue culture flasks.

pH in animal cell cultures decreases due to production of metabolites like lactate. pH control via measurement and base addition is not easily possible in small-scale culture formats like tissue-culture flasks and shake flasks. A hydrogel-based system is reported for in situ pH maintenance without pH measurement in such formats, and is demonstrated to maintain pH between 6.8 and 7.2 for a suspension CHO cell line in CD CHO medium and between 7.3 and 7.5 for adherent A549 cells in DMEM:F12 containing 10% FBS. This system for pH maintenance, along with our previous report of hydrogels for controlled nutrient delivery in shake flasks can allow shake flasks to better mimic bioreactor-based fed batch operation for initial screening during cell line and process development for recombinant protein production in mammalian cells.

Regulation of BACE1 by miR-29a/b in a cellular model of Spinocerebellar Ataxia 17.

Polyglutamine diseases are a class of neurodegenerative disorders characterized by expansion of polyglutamine repeats, protein aggregation and neuronal cell death in specific regions of the brain. The expansion of a polyglutamine repeat in the TATA binding protein (TBP) causes a neurodegenerative disease, Spinocerebellar Ataxia 17 (SCA17). This disease is characterized by intranuclear protein aggregates and selective loss of cerebellar neurons, including Purkinje cells. MicroRNAs are small, endogenous, regulatory non-coding RNA molecules that bind to messenger RNAs with partial complementarity and interfere in their expression. Here, we used a cellular model of SCA17 where we expressed TBP with 16 (normal) or 59 (pathogenic) polyglutamines and found differential expression of several microRNAs. Specifically, we found two microRNAs, miR-29a/b, were down-regulated. With miR-29a/b down regulation, we found an increased expression of targets of miR-29a/b -beta-site amyloid precursor protein cleaving enzyme 1 (BACE1), p53 upregulated modulator of apoptosis (PUMA) and BAK, increased cytochrome c release and apoptosis. Restoration of miR-29a/b in the pathogenic polyglutamine background reduced the BACE1expression. While, antagomiRs against miR-29a/b resulted in an increase in BACE1 levels and neuronal apoptosis. In spite of the elevation of BACE1 in Alzhemiers disease, its role in neuronal cell death has not been established. Here, we show that increased BACE1 expression is not sufficient to cause apoptosis. However restoring level of BACE1 to normal in polyglutamine cells partially reduced neuronal apoptosis. We show a role for the miR-29a/b-BACE1 regulatory interaction in SCA17, suggesting that this microRNA could be part of a common molecular mechanism leading to neuronal cell death in multiple neurodegenerative disorders. The identification of a common mechanism of microRNA mediated neurodegeneration not only improves our understanding of the process, but also provides promising and novel therapeutic targets.

Effect of addition of 'carrier' DNA during transient protein expression in suspension CHO culture.

Transient protein expression using polyethyleneimine as a transfection agent is useful for the rapid production of small amounts of recombinant proteins. It is known that an increase in extracellular DNA concentration during transfection can lead to a nonlinear increase in intracellular DNA concentration. We present an approach that hypothesizes that this nonlinearity can be used to decrease the amount of plasmid required for productive transfections. Through addition of non coding 'carrier' DNA to increase total DNA concentration during transfection, we report a statistically significant increase in protein (IgG) expression per unit plasmid used for transfection. This approach could be useful to increase protein yields for large scale transfections under conditions where plasmid availability is limited.

Controlled release of nutrients to mammalian cells cultured in shake flasks.

Though cell culture-based protein production processes are rarely carried out under batch mode of operation, cell line and initial process development operations are usually carried out in batch mode due to simplicity of operation in widely used scale down platforms like shake flasks. Nutrient feeding, if performed, is achieved by bolus addition of concentrated feed solution at different intervals, which leads to large transient increases in nutrient concentrations. One negative consequence is increased waste metabolite production. We have developed a hydrogel-based nutrient delivery system for continuous feeding of nutrients in scale down models like shake flasks without the need for manual feed additions or any additional infrastructure. Continuous delivery also enables maintaining nutrient concentrations at low levels, if desired. The authors demonstrate the use of these systems for continuous feeding of glucose and protein hydrolysate to a suspension Chinese Hamster Ovary (CHO) culture in a shake flask. Glucose feeding achieved using the glucose-loaded hydrogel resulted in a 23% higher integral viable cell density and an 89% lower lactate concentration at the end of the culture when compared with a bolus-feed of glucose.

Increase in efficiency of media utilization for recombinant protein production in Chinese hamster ovary culture through dilution.

Animal cells are extensively used for the large-scale production of recombinant proteins. Processes and genetically engineered cell lines have been developed to enhance longevity of the culture and increase protein productivity. In this study, we tested the effect of diluting a culture of Chinese hamster ovary (CHO) cells with phosphate-buffered saline (PBS) on cell growth and efficiency of media utilization. An immunoglobulin G-expressing CHO cell line was cultured in CD CHO media followed by dilution of the culture with PBS after the end of the exponential phase. A 28% and 61% increase in protein yield per milliliter of media was observed in the diluted culture in the batch and fed-batch mode with glucose and protein hydrolysate feeding, respectively. To aid in analyzing the potential causes of this observed increase, an unstructured mathematical model was constructed using previously reported kinetics to simulate cell growth, nutrient utilization, and protein production. The model predicts an increase in recombinant protein yield per milliliter of media in PBS diluted cultures under both batch and fed-batch conditions, and suggests that this observed increase could at least partly be due to a decrease in inhibitor concentration in the diluted culture.

Comparison of feature selection and classification combinations for cancer classification using microarray data.

High throughput gene expression data can be used to identify biomarker profiles for classification. The accuracy of microarray based sample classification depends on the algorithm employed for selecting the features (genes) used for classification, and the classification algorithm. We have evaluated the performance of over 2000 combinations of feature selection and classification algorithms in classifying cancer datasets. One of these combinations (SVM for ranking genes + SMO) shows excellent classification accuracy using a small number of genes across three cancer datasets tested. Notably, classification using 15 selected genes yields 96% accuracy for a dataset obtained on an independent microarray platform.