Impacts of intentional mycoplasma contamination on CHO cell bioreactor cultures

Abstract Mycoplasma contamination events in biomanufacturing facilities can result in loss of production and costly cleanups. Mycoplasma may survive in mammalian cell cultures with only subtle changes to the culture and may penetrate the 0.2 µm filters often used in the primary clarification of harvested cell culture fluid. Culture cell‐based and indicator cell‐based assays that are used to detect mycoplasma are highly sensitive but can take up to 28 days to complete and cannot be used for real‐time decision making during the biomanufacturing process. To support real‐time measurements of mycoplasma contamination, there is a push to explore nucleic acid testing. However, cell‐based methods measure growth or colony forming units and nucleic acid testing measures genome copy number; this has led to ambiguity regarding how to compare the sensitivity of the methods. In addition, the high risk of conducting experiments wherein one deliberately spikes mycoplasma into bioreactors has dissuaded commercial groups from performing studies to explore the multiple variables associated with the upstream effects of a mycoplasma contamination in a manufacturing setting. Here we studied the ability of Mycoplasma arginini to persist in a single‐use, perfusion rocking bioreactor system containing a Chinese hamster ovary (CHO) DG44 cell line expressing a model monoclonal immunoglobulin G1 (IgG1) antibody. We examined M. arginini growth and detection by culture methods, as well as the effects of M. arginini on mammalian cell health, metabolism, and productivity. We compared process parameters and controls normally measured in bioreactors including dissolved oxygen, gas mix, and base addition to maintain pH, to examine parameter changes as potential indicators of contamination. Our work showed that M. arginini affects CHO cell growth profile, viability, nutrient consumption, oxygen use, and waste production at varying timepoints after M. arginini introduction to the culture. Importantly, how the M. arginini contamination impacts the CHO cells is influenced by the concentration of CHO cells and rate of perfusion at the time of M. arginini spike. Careful evaluation of dissolved oxygen, pH control parameters, ammonia, and arginine over time may be used to indicate mycoplasma contamination in CHO cell cultures in a bioreactor before a read‐out from a traditional method.

mycoplasma contamination in CHO cell cultures in a bioreactor before a read-out from a traditional method.

K E Y W O R D S
Chinese hamster ovary (CHO) cell culture, monoclonal antibody (mAb), mycoplasma

| INTRODUCTION
Mollicutes, commonly referred to as Mycoplasma, are a genus of bacteria that are challenging to detect and remove from mammalian cell culture.
Mycoplasma are the smallest, mostly self-replicating bacteria (0.1-0.3 µm in diameter); the organism's pleomorphic nature due to the lack of a cell wall allows these bacteria to penetrate the typical 0.1-0.22 µm sterilizinggrade filters used in biomanufacturing. Mycoplasma can potentially contaminate a bioprocessing scheme through raw materials, such as cell culture media components (Drexler & Uphoff, 2002;Kljavin, 2011), and during steps that require manual manipulation of cell lines such as cell banking and cell line development (Nikfarjam & Farzaneh, 2012).
Mycoplasma have small genomes which limits metabolic and replication options and typically require a host cell for survival. Some species of mycoplasma have caused occult contaminations of cell cultures, especially in basic research settings, with minimal visible changes to cell health or cell culture performance. However, mycoplasma can alter the culture performance and product quality in more subtle ways through several potential mechanisms including competition for culture nutrients, induction of abnormal cell growth and cytopathic effects by invading or fusing with the host cells, and alteration of the host cell expression profile (Rottem, 2003).
The ability of mycoplasma to evade detection and grow in mammalian cell cultures creates risk to patients that receive injectable biopharmaceuticals. Yet, there is sparse literature on the details of kinetics and process impacts by mycoplasma contamination in commercial upstream biomanufacturing. Usually when firms find mycoplasma in their cultures, they immediately decontaminate after taking a limited number of culture samples for specification and test raw material samples to trace the source of the contamination. Some published studies regarding mycoplasma species contaminating mammalian cell lines have been conducted in cell culture flasks, shake flasks, spinner flasks, and other standard cell culture vessels (Dabrazhynetskaya et al., 2011;David, Volokhov, Ye, & Chizhikov, 2010;Faison et al., 2018;Laborde et al., 2010;Wang et al., 2017), but not bioreactors; GMPs preclude introducing mycoplasma to a cell culture facility. However, unlike culture vessels, bioreactors can both monitor and control process parameters as well as perfuse fresh media and remove waste materials at a controlled rate. Thus, bioreactors are a much better process model for commercial upstream operations than shake flasks. Further, many of the above-described studies were performed using serum-containing media, which is also not realistic in a commercial setting. While it has been shown that mycoplasma are killed and cleared in a typical downstream bioprocessing scheme (Wang et al., 2017) which reduces the risk for contamination carryover to drug substance, the risk and impact of mycoplasma contamination in relevant models of upstream bioprocessing is less understood in terms of culture performance and product quality effects.
Conventional assays commonly employed for mycoplasma testing take 14-28 days to complete and require labor-intensive manipulation (FDA, 1993(FDA, , 2010. To facilitate more rapid detection, some biomanufacturers have begun to adopt nucleic acid test (NAT) methods as well as other rapid microbial techniques. Thus far, challenges in demonstrating comparability and method validation have hampered adoption. Here we developed models of early-stage process and late-stage process bioreactor culture contamination events using M. arginini and Chinese hamster ovary (CHO) cells grown in serum-free medium expressing a model immunoglobulin 1 (IgG1) product to examine the growth kinetics of M. arginini in a controlled bioreactor environment and the effects on CHO cell culture performance and process parameters. Our data support the understanding of what a mycoplasma contamination event may look like in a typical biomanufacturing scheme and provide knowledge of how process monitoring may identify a mycoplasma contamination event in the manufacturing environment. Additionally, this study provides data for the feasibility and benefits for development and implementation of rapid mycoplasma testing, such as NAT methods, in upstream biomanufacturing.

| Seed train expansion and inoculum preparation
This experiment used a previously described recombinant CHO DG44 cell line that expresses a model chimeric IgG1 (Velugula-Yellela, Williams, et al., 2018). Frozen cell stocks (2 × 10 7 cells/ml) were thawed and seeded into multiple 1 L spinner flasks containing 300 ml CD OptiCHO (Life Technologies; A11222) media supplemented with 8 mM L-glutamine (Corning; 25-005-CV). Spinner flasks were incubated at 37°C and 8% CO 2 at an agitation speed of 73 rpm. Fresh media was added to double the volume after the viable cells reached ≥2 × 10 6 cells/ml. The next day (day before inoculation), the total volume of each spinner flask was brought up to 1 L using fresh media.
On the day of inoculation, cells were pelleted by centrifugation at 300g for 10 min at 23°C and resuspended in fresh CD OptiCHO media supplemented with 8 mM L-glutamine and 1× Soy Hydrolysate (Sigma-Aldrich; 58903C or S1674). Soy hydrolysate was added to the media to support growth of M. arginini in serum-free media based on previous studies (Wang et al., 2017). Inoculum cell density was measured, and each bioreactor was inoculated with identical volumes to reach the target seeding density of 1 × 10 6 cells/ml.  Three 14-19 day runs of two bioreactors (n = 6) were completed with identical setpoints and strategies as listed in Table 1. Six bioreactors were run in total, but for the third run Day 12-High was treated as a control for Day 9-Low until Day 12, and from Day 12 onward was used for a latestage contamination event bioreactor and served as an internal control to itself (Table 2). With the exceptions of rocking speed and rocking angle, culture parameters were automatically controlled using the UNICORN system control software. Bioreactors were run in batch mode until glutamine reached ≤ 1 mM, except for Day 2-High and Day 2-Control in which a glutamine feed of 2 mM was added 1 day before the start of perfusion. Cell bleeds were generally performed when cells measured above 55 × 10 6 cells/ml to bring cell density to 20-40 × 10 6 cells/ml, depending on the batch age and desired cell density reduction.

| Mycoplasma preparation and spiking
The mycoplasma spike for the bioreactor was prepared from a frozen stock of M. arginini in 45% glycerol thawed at room temperature in a laminar flow hood using a sterile syringe and disposable pipette basin. The mycoplasma stock was mixed with 3 ml media (OptiCHO/ L-Glutamine + soy hydrolysate) warmed to 37°C to seed a final target concentration of 10 1 (low spike) or 10 2 -10 3 (high spike) CFU/ml in 1 L bioreactor volume. A negative control containing an identical volume of media and supplements only was also prepared for the uncontaminated bioreactor. Syringes were attached to the WAVE bioreactor sample ports, the contents were dispensed, and the attached syringes were used to flush the sample line with bioreactor media before disconnecting. The initial mycoplasma titration samples for time 0 were taken approximately 1-2 hr post-spike to allow for contaminant and clump dispersal. As mycoplasma doubling times are about 6 hr or more, this lag is unlikely to lead to titer overestimates (Waites & Talkington, 2004).

| Perfusion of single-use bioreactors
As previously noted, bioreactors were run in batch mode until glutamine reached ≤ 1 mM, apart from Day 2-High and Day 2-Control in which a glutamine feed of 2 mM was added 1 day before the start of perfusion. Perfusion was started for each run at a rate of 1 L/day, based on the results of previous experiments (data not shown). In general, to keep glutamine ≥ 0.5 mM and glucose ≥ 0.5 g/L, perfusion was increased to 2 L/day when cells were > 10 × 10 6 cells/ml, 3 L/day when cells were > 20 × 10 6 cells/ mL, and the maximum rate used of 3.5 L/day when cells were measured > 40 × 10 6 cells/day. Freshly prepared media bottles were stored no more than 7 days at 4°C protected from light and were manually attached each day at roughly the same time and perfusate collection bottles were exchanged out for empty bottles. Raw samples from each day's perfusate were retained for mycoplasma testing and titer measurements. Perfusion of contaminated bags was discontinued when CHO cell viability had dropped below~30%.

| Mycoplasma quantification
Mycoplasma arginini strain 23243 (ATCC, Manassas, VA) was cultivated with SP4 + arginine medium agar and broth (Hardy Diagnostics). Inprocess samples from both control and spiked bioreactors were plated, and colonies were counted (Wang et al., 2017). Briefly, mycoplasma titers in bioreactor test articles were determined by performing serial 10-fold dilutions using phosphate-buffered saline (PBS) (Gibco, Carlsbad, CA) and 100 µL of each dilution was plated on SP4 + arginine medium agar.
Colonies were counted 5 days after plating and all samples were plated in duplicate.

| Titer measurements using protein A bio-layer interferometry (BLI) sensors
Samples were thawed at room temperature and titer was measured using Protein A dip-and-read BLI sensors (FortéBio, 18-5010) on the Octet RED96 system (FortéBio) as described previously (Velugula-Yellela, Kohnhorst, et al., 2018). Cell-specific IgG1 production rates (Q p, pg·cell −1 ·day −1 ) for each perfusion day were calculated using the following adapted formula (Clarke et al., 2011) arginini grew for 3 days, plateaued at a peak density of~10 7 CFU/ml which held steady for 5 days, and then plummeted in 1 day from >10 7 to 10 1 CFU/ml (Figure 1). However, live M. arginini residuals persisted nearly 10 days after the spike. Interestingly, the low concentration spike (15 CFU/ml) showed a rapid increase in the growth phase to reach the same peak density (10 7 CFU/ml) as the higher density spike culture (300 CFU/ml). Based on these kinetics, we conclude that the It is important to note that glucose would be rapidly consumed as a carbon source by common bacterial contaminants from skin or soil such as Staphylococcus sp. and Bacillus sp. (Baumstummler et al., 2010;Strasters & Winkler, 1963). In commercial settings, rapid drops in glucose, pH, and DO are often relied on as signs of contamination.
Because M. arginini does not metabolize glucose (Sugimura, Ohno, Azuma, & Yamamoto, 1993), glucose exhaustion is not likely an indicator of an M. arginini contamination. Glucose in the bioreactors rose to 5-6 g/L approximately 4-5 days after contamination of all cultures studied (Figure 3), indicating little or no net consumption of F I G U R E 1 Mycoplasma arginini growth profiles in four CHO cell bioreactor cocultures. Bioreactors Day 2-High and Day 3-High were spiked with M. arginini when CHO VCD reached 2 × 10 6 cells/ml before the perfusion start day (early-stage contamination events). M. arginini was spiked into Day 9-Low when CHO VCD reached 10 × 10 6 cells/ml and in Day 12-High after a third cell bleed to bring the CHO VCD to 10-15 × 10 6 cells/ml (late-stage contamination events). CHO, Chinese hamster ovary; VCD, viable cell density In comparison, glutamine only rose to a steady-state of approximately half the concentration of the perfusion media, indicating that the mycoplasma were consuming glutamine as an energy source and producing ammonia and glutamate as end products (Figure 3). This is expected given well-understood metabolic pathways (Smith, 1955(Smith, , 1957a(Smith, , 1957b. The metabolites most likely to be useful as sentinels for mycoplasma contamination were ammonia which, in general, steadily increased in the culture once M. arginini was present (Figure 3c-f), and arginine, which rapidly declined by >90% 3-4 days into M.
arginini presence (Figure 4b,c). Of the 15 amino acids we were able to accurately measure within the in-process samples, only arginine decreased drastically after M. arginini was spiked into the bioreactors.
The main pathways for ammonia are arginine and glutamine metabolism. Arginine degradation is a significant metabolic process for most mycoplasma species and is carried out through a set of three reactions that ultimately result in production of ornithine, ATP, ammonia, and carbon dioxide (Schimke & Barile, 1963). The high concentrations of ammonia by-product in the cultures, reaching 25-30 mM despite perfusion rates of up to 3.5 CV/day, could alone be the cause for CHO cell death in the bioreactors, as ammonia has been shown to have negative effects on other mammalian cells (hybridomas) in the range of 2-10 mM (Ozturk, Riley, & Palsson, 1992). CHO cells are susceptible at concentrations as low as 4 mM (Kurano, Leist, Messi, Kurano, & Fiechter, 1990). Ammonia production by one species of mycoplasma, M. salivarium, has been postulated to be a virulence mechanism in patients with oral infections (Matsuura, Seto, & Watanabe, 1990). It is also important to note that ammonia and glutamate concentrations in the prepared starting media and perfused media were higher in Bioreactors Day 9-Low and Day 12-High than in Bioreactors Day 2-High, Day 2-Control, Day 3-High, and Day 3-Control (Figure 3). This was confirmed to be due to a change in soy hydrolysate source, which appeared to result in both a slower entrance into exponential growth as well as a decrease in exponential growth rate before spiking.

| Effects of M. arginini on culture conditions and process controls
In commercial settings, critical process parameters are consistently collected in real-time with automatic feedback to control the process.
Thus, we closely monitored any changes in pH and dissolved oxygen (DO) after M. arginini contamination to investigate if they could serve as sentinels for contamination. The DO setpoint was 50% of air saturation and was controlled automatically with O 2 overlay, and manually with rocking speed and rocking angle (Table 1). In control bioreactors, DO gradually equilibrates to the 50% setpoint over 4-5 days of culture and remains at or above 50% for the duration of the runs (Figure 5a). To maintain 50% air saturation, the control system automatically adds O 2 into the gas mix to increase DO. When possible, the rocking speed and/or angle can be manually increased when O 2 % reaches 30-40% because if the O 2 % reaches its maximum of 50%, the system loses its ability to control for increasing oxygen demand arginini spike (Figure 5c). Overall, we found that timing of the DO accumulation was directly related to the culture conditions, namely CHO VCD and perfusion rate, as the M. arginini took over the bioreactor culture and overall oxygen consumption ceased.
In control bioreactors, pH control via CO 2 overlay and 0.5 M NaOH addition followed similar trends. In the first few culture days,  (Figure 6a). In early stage contamination event models, CO 2 percentage began to rise 1-2 days after mycoplasma was spiked into the bioreactor, which corresponded to the approximate time when the CHO cell growth rate slowed, and the addition of base was never required (Figure 6b).
In late-stage contamination event models, this process lagged as CO 2 increased 4 days after the cultures were spiked with mycoplasma and base additions were still required (Figure 6c).

| CHO cell IgG1 production after contamination with M. arginini
IgG1 production from the uncontaminated CHO DG44 cell averaged 1.49 ± 0.41 pg·cell −1 ·day −1 , with variability likely influenced by multiple factors including cell density, growth rate, and perfusion rate.
This variability made it challenging to determine if M. arginini contamination specifically caused any changes to productivity.
However, based on the comparison of parameters and measuring Q p after M. arginini contamination relative to uninfected controls, the CHO cells maintained some productivity even as the health of the culture declined (Figure 7).
In the early-stage contamination model, the CHO cell density was as low as 3.7 × 10 5 cells/ml and the reduced IgG concentration resulted in a less precise Q p calculation. However, it was evident that the viable cells maintained a close to average productivity for some duration after the M. arginini spike. In Day 2-High, Q p of the spiked culture was significantly lower 2 days after spiking but appears to have recovered 5 days after the spike. One week after the spike on Day 9, the Q p of the spiked culture was significantly lower than the uncontaminated control because after Day 9 there were too few viable CHO cells left in the bioreactor to make accurate calculations (less than 100,000 CHO cells) (Figure 7a). In Day 3-High, Q p between the spiked and uninfected cultures were significantly different. The In the late-stage contamination event models, the higher CHO cell densities and higher volumetric titer resulted in more precise Q p measurements. In Day 9-Low, Q p decreased from 3 to 6 days after contamination, but remained within 3 standard deviations of the average of the control runs until 7 days after M. arginini contamination when Q p dropped to 10% of the average of the uninfected cultures ( Figure 7c). In Day 12-High, the Q p of the infected culture was slightly below the average after mycoplasma contamination but was not significantly lower until 6 days after spiking with M. arginini