Viral Contamination in Biologic Manufacture and Implications for Emerging Therapies

Perspective https://doi.org/10.1038/s41587-020-0507-2

Viral contamination in biologic manufacture and implications for emerging therapies

Paul W. Barone1, Michael E. Wiebe1, James C. Leung1, Islam T. M. Hussein1, Flora J. Keumurian1, James Bouressa2,25, Audrey Brussel3,26, Dayue Chen4,27, Ming Chong5, Houman Dehghani6,28, Lionel Gerentes7, James Gilbert8,29, Dan Gold9, Robert Kiss10,30, Thomas R. Kreil11, René Labatut3, Yuling Li12,31, Jürgen Müllberg13, Laurent Mallet7,32, Christian Menzel14, Mark Moody15,33, Serge Monpoeho16, Marie Murphy4, Mark Plavsic17,34, Nathan J. Roth18, David Roush19, Michael Ruffing20, Richard Schicho21,35, Richard Snyder22, Daniel Stark23, Chun Zhang24,36, Jacqueline Wolfrum1, Anthony J. Sinskey1 and Stacy L. Springs 1 ✉

Recombinant protein therapeutics, vaccines, and plasma products have a long record of safety. However, the use of cell culture to produce recombinant proteins is still susceptible to contamination with viruses. These contaminations cost millions of dollars to recover from, can lead to patients not receiving therapies, and are very rare, which makes learning from past events difficult. A consortium of biotech companies, together with the Massachusetts Institute of Technology, has convened to collect data on these events. This industry-wide study provides insights into the most common viral contaminants, the source of those contaminants, the cell lines affected, corrective actions, as well as the impact of such events. These results have implications for the safe and effective production of not just current products, but also emerging cell and gene therapies which have shown much therapeutic promise.

n the twentieth century, several vaccine products were uninten- human pathogens. The lessons of the past have led to the current tionally contaminated with unwanted viruses during their pro- best practice, which relies on three pillars: the selection of appropri- duction1–3. This included the contamination of poliovirus vaccine ate starting and raw materials with a low risk of containing adventi- I 3 with simian virus 40 (SV40) , for which the health impacts were tious virus; testing of cell banks and in-process materials to ensure not fully known for many decades4. In the early 1980s, unknowingly they are free from detectable viruses; and finally, the incorporation contaminated therapeutic proteins from human plasma caused of steps to remove and inactivate potential undetected adventitious widespread transmission of viruses such as human immunodefi- and endogenous viral contaminants during purification of the prod- ciency virus (HIV) to people with hemophilia who received these uct9,13,14. Because of this approach, these products have been safe for treatments5,6. As a result, public trust in the plasma industry’s abil- over 35 years, and, to our knowledge, there has been no transmis- ity to safely make these therapies declined7,8. To ensure that current sion of a contaminating virus to a patient from a therapeutic protein plasma-derived, vaccine, and recombinant biotherapeutics are safe, produced using recombinant DNA technology. complementary safety strategies to reduce the risk of virus contami- Despite this excellent safety record, viral infection of mam- nation were developed and implemented9–11. malian cell culture is a real risk with severe consequences. Even Since that time, the production of therapeutic proteins has if no contaminated lots are released, patients who require treat- largely shifted to the use of recombinant DNA technology in pro- ment can be affected by drug shortages and public confidence in karyotic and eukaryotic cells12. However, culturing of these cells is the biotech industry can be severely damaged. These events can susceptible to contamination from adventitious agents (primarily cost tens of millions of dollars in investigation, cleanup, correc- bacteria and viruses). Viruses are of particular concern as they are tive actions, lost sales and manufacturing plant downtime15. They often more difficult to detect than other microbial contaminants1 also divert company leadership, encourage the competition, and and in the case of mammlian cell culture can potentially replicate can decrease company value. Finally, they expose the company

1Center for Biomedical Innovation, Massachusetts Institute of Technology, Cambridge, MA, USA. 2Pfizer Inc., Andover, MA, USA. 3Sanofi, Paris, France. 4Eli Lilly, Indianapolis, IN, USA. 5CSL Behring, Melbourne, Victoria, Australia. 6Amgen, Thousand Oaks, CA, USA. 7Sanofi Pasteur, Marcy L’Étoile, France. 8Biogen Inc., Cambridge, MA, USA. 9BioMarin, San Raphael, CA, USA. 10Genentech, South San Francisco, CA, USA. 11Baxter/Baxalta/Shire, now part of Takeda, Vienna, Austria. 12MedImmune, Gaithersburg, MD, USA. 13Bristol-Myers Squibb, Devens, MA, USA. 14Merck Group, Geneva, Switzerland. 15Merrimack Pharmaceuticals, Cambridge, MA, USA. 16Regeneron, Rensselaer, NY, USA. 17Sanofi Genzyme, Cambridge, MA, USA. 18CSL Behring, Bern, Switzerland. 19Merck & Co., Inc., Kenilworth, NJ, USA. 20Boehringer Ingelheim, Biberach, Germany. 21Bristol-Myers Squibb, New York, NY, USA. 22Brammer Bio, Alachua, FL, USA. 23Novartis, Basel, Switzerland. 24Shire, Lexington, MA, USA. 25Present address: Retired, Andover, MA, USA. 26Present address: LFB, Les Ulis, France. 27Present address: Genentech, South San Francisco, CA, USA. 28Present address: Allogene Therapeutics, South San Francisco, CA, USA. 29Present address: Independent consultant, Westcliffe, CO, USA. 30Present address: Sutro Biopharma Inc., South San Francisco, CA, USA. 31Present address: Innoforce, Hangzhou, Zhejiang, China. 32Present address: EDQM, Strasbourg, France. 33Present address: Immunova, Cambridge, MA, USA. 34Present address: Lysogene, Cambridge, MA, USA. 35Present address: Shire, Atlanta, GA, USA. 36Present address: Evelo Biosciences, Cambridge, MA, USA. ✉e-mail: [email protected]

Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology 563 Perspective NATuRe BIoTecHnology

The CAACB virus contamination in biomanufacturing study Table 1 | Virus contaminations of mammalian cell culture To date, the CAACB has collected a comprehensive set of data on to produce proteins and vaccines, segregated by year, both virus contamination experience, as well as controls in place to pre- publicly reported and contained in the CAACB study vent contaminations, from 20 major biopharmaceutical manufac- Year of Contaminations (virus / host cell) Total turers. A 166-question survey of the CAACB members was used to contamination conduct the study (see Supplementary Note). To ensure a manage- 1985–1989 Blue tongue / CHO 2 able dataset for comparable processes, the scope of the project was EHDV / CHO18,19 limited to virus contaminations in mammalian cell culture manufacturing. The project did not include bacterial or yeast fermen- 1990–1994 Herpesvirus / primary monkey 7 tation, plasma fractionation or egg-based production of vaccines Herpesvirus / Vero and covered manufacturing from the pilot to commercial scales, MVM / CHO (×2)20–22 Parainfluenza 3 / MRC5 including both current Good Manufacturing Practice (cGMP) and Reo3 / MRC5 non-cGMP operations. Unless otherwise noted, all data and discus- Simian adenovirus / primary monkey sion here relates to information reported directly to the CAACB and does not include information from other published reports. 1995–1999 CVV / CHO 3 Nine out of the 20 responding companies (or almost half of the Reovirus / human primary kidney23 Vesivirus 2117 / CHO24 involved industry respondents) reported having one or more adventitious virus contaminations in mammalian cell culture operations 25 2000–2004 CVV / unknown (×2) 3 since 1985, totaling 18 virus contamination events reported to the 26 Human adenovirus / HEK293 CAACB. All of these reported contamination events occurred at 2005–2010 CVV / CHO 6 manufacturing sites in North America and Europe, but there is MVM / CHO (×2) insufficient data to determine whether one geographic location has Vesivirus 2117 / CHO (×3)27–29 a disproportionately increased risk of contamination over another. 2010–present MVM / CHO30 3 MVM / BHK-2135 Different host cells, different risks. The contaminated cell type, PCV-1 / Vero31,32 contaminating virus and suspected source of contamination for the Unknown MVM / BHK-2133 2 18 events reported to the CAACB are shown in Table 2. In 67% of Reovirus / Unknown34 reported events, the manufacturing platform was Chinese hamster ovary (CHO) cells, whereas the other 33% of events involved human Contaminating virus and host cell line are indicated where known. Note: some contamination events were reported publicly and in more detail to the CAACB. CVV, Cache Valley virus; EHDV, or primate cell lines. This result is not unexpected as CHO cells are epizootic hemorrhagic disease virus; MVM, minute virus of mice; PCV-1, porcine circovirus type 1; the most commonly used host cells by the recombinant-biologic Reo3, reovirus type 3. industry, with published reports indicating that approximately 70% of approved biotech products are manufactured using CHO cells12. The reported virus contaminations occurred at all stages to intense regulatory scrutiny and can result in a delay in the of the product life cycle, with 3 events occurring during preclini- approval of new products or the accelerated approval of a com- cal non-cGMP manufacture, 2 during clinical cGMP manufacture, petitor’s product16,17. and the remaining 13 occurring during commercial manufacture. Despite the above damaging consequences from virus con- Considering the strict controls in place for clinical and commer- tamination, effectively learning from previous contamination cial manufacturing, the fact that most contaminations reported to events is a challenge. These events are rare; we are aware of only the CAACB occurred under cGMP production may be surprising. 26 virus contaminations over the past 36 years (Table 1)18–35, 18 of One possible explanation is that the scale of cGMP production and which were reported directly as a part of this study. Furthermore, a volumes of media used were much greater than those of non-cGMP comprehensive dataset and evaluation of virus contaminations in production, allowing a greater opportunity to introduce a low-level biomanufacturing has not previously been published. Therefore, contaminant. Additionally, viral testing is not mandatory for the Consortium on Adventitious Agent Contamination in non-cGMP manufacturing and some contamination events may Biomanufacturing (CAACB)—a biopharmaceutical industry con- not have been recognized. All of the commonly used cell culture sortium including more than 20 biotechnology companies housed processes in the industry, such as roller bottle (one event), batch at the Massachusetts Institute of Technology’s Center for Biomedical culture (three events), fed-batch culture (four events) and perfusion Innovation—collected comprehensive data on virus contamina- culture (four events), were implicated. In six events, the type of cell tions in cell culture operations from CAACB member companies36. culture process was not identified by respondents. On the basis of These data were consolidated with information from published the available data, it is not possible to determine whether one opera- reports of virus contamination events. To our knowledge, this tional mode has a higher risk of contamination than another. is the only comprehensive dataset available on adventitious virus Nine viral contaminants have been identified as responsible for contaminations of mammalian cell culture in the biotech industry. the 18 virus contamination events reported to the CAACB (Table 2). This industry-wide study is the first of its kind and provides insights No overlap exists between the four viruses found to contaminate into the most common viral contaminants, the source of those con- CHO cell culture and the five viruses found to contaminate human taminants, the cell lines affected, corrective actions taken, and the or primate cells. This highlights the fact that the contamination and impacts of such events. safety risks are distinct for CHO cells versus human or primate cells. In this Perspective, we describe the work to date and discuss In 11 of the 12 reported contaminations in CHO cell culture, a raw the implications of our findings for manufacturers of recombinant material or medium component was identified or suspected to be protein therapies. We then use these insights to outline viral con- the source. In comparison, for the human and primate cell lines, the tamination considerations for developers of emerging gene and cell manufacturing operators or cell line itself were suspected to be the therapies. Our aim in this paper is to facilitate the industry’s mis- source. The fact that operators are only indicated as a source of the sion of producing safe and effective biologic products. We note that contaminant in human or primate cell culture and not in CHO cell this is a living project and that we expect to continually collect and culture is likely due to the ‘species barrier’ for viral infection between analyze data in the future. human or primate cells on the one hand and rodent cells on the other.

564 Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology NATuRe BIoTecHnology Perspective

Table 2 | Viruses reported to have contaminated mammalian cell culture operations for the production of vaccines or recombinant proteins, the number of events where the virus source was identified, and the source of the contaminant Contaminated Contaminating virus Pathogenic Suspected and confrmed sources of contamination cell line to humans Serum Recombinant Undetermined Operator Host Not medium component medium component cell line found Viruses found to contaminate CHO cell culture CHO Blue tongue virus No 1 CHO Cache Valley virus Yes 2 CHO Minute virus of mice No 1 3 1 CHO Vesivirus 2117 No 4 Viruses found to contaminate human or primate cell lines Primary monkey, Herpesvirus Yes 1 1 Vero HEK293 Human adenovirus Yes 1 type 1 MRC5 Parainfluenza virus Yes 1 type 3 MRC5 Reovirus type 3 Yes 1 Primary monkey Simian adenovirus No 1

Simply put, viruses that infect humans are more likely to be able to 12 Non-cGMP positive material replicate in human cells than in non-human mammalian cells. 10 Non-cGMP negative material cGMP positive material 8 Safety implications. Four (herpesvirus, human adenovirus type 1, cGMP negative material parainfluenza virus type 3 and reovirus type 3) of the five viruses 6 ents where tested found to contaminate human and primate cell lines are known to be 4 pathogenic in humans, whereas only one (Cache Valley virus) of the viruses found to contaminate CHO cell culture has been reported 2 37,38 to cause disease in humans . These data highlight that the viral Number of ev 0 ials contamination of protein products produced in human or primate aces ater W A filters cell lines pose a higher safety risk to patients and the manufacturing ug product w mater HEP ug substanceDr Ra king cell banks process due to human cell line susceptibility to infection by viruses r Dr wo that are pathogenic in humans. vironment, surf Concurrent seed trains En ification process samples Animal-derived raw materials (ADRMs), especially serum, Preceding cell culture units Master or Pur carry a higher risk of being contaminated with virus and are thus Concurrent cell culture processes being replaced where possible throughout the industry1,9,13. This is further corroborated by our data: three (blue tongue virus, Cache Fig. 1 | Virus testing of contaminated processes. Virus tests on samples Valley virus and vesivirus 2117) of the four viruses that contami- from different process steps of the affected runs during investigation of nated CHO cell culture were suspected or definitively identified to the contamination events reported to the CAACB. Data reported to the have come from serum. However, the removal of ADRMs does not CAACB included samples from cGMP operations that tested positive (dark eliminate the risk of contamination. In one contamination with the orange) or that were below the limit of detection of the assay and assumed minute virus of mice (MVM), the process contained no ADRMs. negative (dark blue) and samples from non-cGMP operations that tested Minute virus of mice is especially challenging as a potential con- positive (light orange) or that were below the limit of detection of the assay taminant. It is shed from ever-present wild mouse populations, may and assumed negative (light blue). Note: not all materials were tested in not be detectable even with established rodent control, and can per- each contamination event. sist in the environment and in raw materials long after being shed. Despite the fact that raw materials were determined to be the most likely source of the contamination in 11 events, testing those that their contamination of mammalian cell culture would lead raw materials did not necessarily detect the contaminating virus. In to obvious changes in culture performance parameters (for exam- only 3 events was the viral contaminant directly detected in the sus- ple, viable cell density). For 11 of the 18 contamination events pect raw material (Fig. 1). In all 3 cases, it was necessary to increase reported to the CAACB, a change in cell culture parameters was the viral load to a level detectable by PCR through either amplifi- the leading indicator of a contamination (5 of 18 events occurred cation by virus replication in cell culture or concentration of the sufficiently long ago that it is not known whether there was a raw material. In the other 8 contamination events, virus testing of change in cell culture parameters). However, in 2 events, there raw materials was negative and the source of the contamination was was no apparent change in cell culture performance and the con- only identified using indirect evidence. taminating virus was only detected by a virus-specific PCR assay. This suggests that cell culture performance alone may not pro- Virus detection methods vide sufficient warning of contamination. Additionally, changes As viruses are obligate molecular parasites that co-opt the cellu- to cell culture performance can be due to many factors other than lar machinery of the host cell they infect, it might be expected virus contamination.

Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology 565 Perspective NATuRe BIoTecHnology

Table 3 | Methods used for the detection (both initial detection and confirmation of a contamination) and identification of the viral contaminant of a virus contamination in cell culture operations Methods Detection Virus Used for QC lot release assay identifcation Used for QC test QC test positive QC test negative PCR 8 11 1a 1 0 IVV test 7 14 11b 4b Electron microscopy 4 3 Viral genome sequencing 3 Immunofluorescence 2 2 Mass spectrometry protein sequencing 2 RNA fingerprinting 1 1 Serology 6 Massively parallel sequencing 1 1

The number of contamination events where each test was used is listed in each column; multiple tests may have been used in each event. Tests used for QC lot release, and whether they were positive or negative, are indicated. aOne company used both PCR and IVV tests for QC lot release; in one event, PCR was positive, but the IVV test was negative. bOne company found some bioreactors to be positive by PCR but negative by IVV test, whereas other bioreactors were positive by both tests. This is counted in both columns.

The in vitro virus (IVV) assay is a cell-based assay used to test cell controls currently in place. Although outside the scope of this work, culture harvest samples for potential viral contaminants39. It is able a comprehensive discussion of biomanufacturing controls to pre- to detect a wide range of viruses and was used as a quality control vent cross-contamination can be found in the ISPE Baseline Guide (QC) lot release assay in all 15 events that were carried out under Volume 6: Biopharmaceutical Manufacturing Facilities40. It should cGMP manufacturing. Despite this, the IVV assay was negative be noted that, in one contamination event, high-efficiency par- in 4 events and the contamination was detected by using another ticle absorbing (HEPA) filters tested positive for the contaminat- method (Table 3). These data imply that the safety of biologic prod- ing virus. Whereas some may consider the likelihood of virus being ucts should not rely on testing alone (including orthogonal meth- aerosolized in a manufacturing setting to be low, this highlights the ods) but be assured by multiple controls (including prevention, fact that it is not impossible—the 0.2-µm vent filters on bioreactors detection and viral clearance) throughout the process. are not designed to retain virus—and manufacturing facility design Whereas the data in Table 3 highlight the deficiencies of a range and decontamination activities in the event of a contamination of widely used detection assays, the use of rapid virus detection should take this into account (for example, using a decontamina- assays has prevented the spread of a viral contaminant throughout tion approach proven to be virucidal and capable of reaching areas a production facility. Of the 18 contamination events reported to potentially exposed to aerosols). the CAACB, 7 were contained in cell culture bioreactors (Fig. 2). Data reported to the CAACB also support the effectiveness of the Noteworthy is the fact that in 3 of the events, virus-specific PCR virus removal and inactivation capacity of downstream purification tests performed before bioreactor harvest detected and identified operations, which has been documented elsewhere41,42. As an assur- a viral contaminant in the bioreactor and prevented the spread of ance of safety, the unit operations of the downstream purification the virus to downstream purification processes and other parts of process, such as chromatography, are evaluated at small scale for the manufacturing facility. This greatly reduced the time, effort and their ability to separate potential viral contaminants from the final cost of both investigating the event and getting the manufacturing product. Dedicated steps to inactivate virus (for example, a low-pH facility back up and running. Conversely, no rapid PCR assays were hold or solvent or detergent treatment for large, enveloped viruses) in place in the 6 events in which contaminated cell culture fluid and remove virus (for example, the use of nanofiltration) are also was processed downstream. As PCR assays are designed for a spe- designed into downstream purification. These processes are evalu- cific target virus or panel of viruses, a viral contamination will only ated for their ability to clear model adventitious viruses with a range be detected if primers and probes for the contaminating virus are of biochemical and biophysical properties. As these studies are not included in the assay. However, these data highlight the capability of designed to evaluate a specific safety risk, there is no minimum rapid detection assays to reduce business risk and increase product clearance suggested in the guidance9. In four of the six contamina- safety, especially in known high-impact situations. tions that spread downstream, the capacity of the large-scale purification steps to clear virus was evaluated (Fig. 2). Whereas some Current viral safety approaches work purification process intermediates tested positive for virus (Fig. 1), Data collected as a part of the CAACB study indicate that current in all cases, the entire purification process reduced the contaminat- manufacturing controls used to prevent the spread of a potential ing virus to levels below the limit of detection of the assay. Where contaminant within manufacturing facilities are effective as no tested, all drug substance and drug product tested negative for virus. cross-contamination of other concurrent manufacturing operations Additionally, for all six contaminations that spread to downstream was reported. Figure 1 shows the results for in-process materials that processes, no virus testing (for example, PCR) of bioreactors, with were tested for virus during the post-contamination investigation. a negative result required to further process the cell culture har- For cGMP production, five of eight cell cultures preceding the reac- vest, had been implemented. It is important to note that in all cases tor were originally identified as contaminated; one of six concurrent reported to the CAACB, no drug substance produced from a con- seed trains and no concurrent cell cultures for different products taminated cell culture was ever released for human use, and all such were also found to be contaminated. In all cases, the contamina- drug substances were subsequently destroyed. tion of concurrent cell culture operations came from a shared raw Without a doubt, an adventitious virus contamination dur- material and not from cross-contamination within the manufacturing cell culture manufacture of a biologic is incredibly disruptive. ing facility. This supports the effectiveness of cross-contamination Investigating a viral contamination event costs both time and

566 Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology NATuRe BIoTecHnology Perspective

18 Total events

7 Contained in cell culture 6 Spread to downstream 5 Unknown

Virus tests not implemented as forward Cell culture terminated processing 3 due to cell death

1 Planned termination 4 Viral clearance evaluated

Virus testing as >3 LRV forward processing Virus species 1 3 prevented downstream >7 LRV contamination

>12 LRV Virus species 2 >8 LRV

Fig. 2 | Extent of contamination. Schematic showing the extent of contamination in the manufacturing process and the use of virus detection as a process forwarding criteria. For seven events, the contamination was contained in the cell culture, for six events the contamination was spread to downstream purification operations, and for five events the extent of contamination was unknown. The ability of the downstream process to remove or inactivate the viral contaminant was evaluated in four of the six contamination events and was found to remove contaminating virus below the limit of detection of the assay. For all six contaminations that spread to downstream processes, no virus testing was implemented as process forwarding criteria. LRV stands for log reduction value and is a measure of the ability of the process to remove or inactivate virus. As an example, a process that is capable of reducing the viral load by a factor of 104, such as from a viral titer of 1010 to a titer of 106, is said to have a LRV of 4. resources. Depending on the severity of the event, the investigation 6 could take several months for personnel involved. The expense of such an investigation, as reported to the CAACB, was in the $1–10 5 million range, but in the worst cases the cost of investigating a ents 4 contamination, implementing corrective actions, decontaminat- ing the facility, and other ancillary costs could be in the hundreds 3 of millions of dollars. In multiple events reported to the CAACB, the contamination resulted in a competitive disadvantage to some 2

products. Furthermore, viral contaminations have been reported to Contamination ev lead to manufacturing interruptions (with a median plant down- 1 time of 1 to 2 months; Fig. 3), loss of revenue, discarded batches, 0 lawsuits against the company, a decrease in company reputation 0 <1 1–2 2–3 >3 Unknown as reflected in stock value, and significant delays in product development. Finally, these events can affect patients. In one event, the Months manufacturing was shut down negative impact on drug supply resulted in a modified treatment regimen, with some patients not able to obtain adequate levels of Fig. 3 | Manufacturing shut down. Months that manufacturing plants were drug until manufacturing operations were restored. shut down due to virus contaminations.

Insights from the CAACB study cell lines. Our data demonstrate that virus contaminations in The CAACB study results have several implications for how bio- cell-culture-based biopharmaceutical manufacturing are rare events logic manufacturers approach viral contamination in producer when evaluated against the cumulative volume of the biotechnology

Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology 567 Perspective NATuRe BIoTecHnology

100% remove or inactivate virus in media or components thereof. The 90% Implemented most common technologies that have been explored in the indus- 80% Evaluating try include flash pasteurization (also known as high-temperature, 70% 20,44,45 46 60% short-time heat treatment or HTST) , UV-C irradiation and None 47 50% nanofiltration (Fig. 4). Although a small sample size, to date none 40% of the four manufacturers that has implemented HTST heat treat- 30% ment to inactivate potential virus in media has experienced a con- 20% tamination event after its implementation. % of study respondent s 10% Finally, our data support the effectiveness of viral clearance in 0% downstream protein purification operations (Fig. 2) for large-scale HTST UV-C Nanofiltration manufacturing processes. In all cases evaluated, downstream puri- Fig. 4 | Implementing treatment of media. The percentage of respondents fication (for example, chromatography) and dedicated downstream who have implemented (blue) or are evaluating (orange) technologies such viral inactivation or removal steps (for example, low-pH hold, a as high-temperature, short-time (HTST) treatment, UV-C irradiation or mixture of solvent and detergent known as solvent/detergent treat- nanofiltration to remove or inactivate potential viral contaminants from cell ment, or nanofiltration) inactivated or removed the contaminating culture media and medium components. virus to below the limit of detection of the assay. This highlights the value, from a viral safety perspective, of downstream virus-clearance operations for the removal of potentially undetected contaminants. industry over the past 35 years. However, our data also indicate Implications for cell and gene therapy manufacturers that, on a per-company basis (of those that completed our survey), At the end of the third quarter of 2019, >1,052 advanced therapy the experience is not that rare. Of the 20 companies completing the medicinal products (ATMPs) were being tested in phase 1–3 clinical CAACB virus contamination survey, 45% of respondents reported trials (https://alliancerm.org/publication/q3-2019-data-report/), experiencing at least one virus contamination event between 1985 and several cell48 and gene49 therapies had received regulatory and 2018, which is higher than we expected. By some estimates43, approval. Autologous cell-based products are personalized for a companies participating in the CAACB study comprise >75% of specific patient in critical need and are often last-line treatments. global mammalian cell culture manufacturing capacity and, there- A delay in administering treatment, such as due to contamination fore, the risk of experiencing a virus contamination, based on total of a production process, could negatively affect patients, possibly processed volume, may be expected to be higher for those com- leading to their death. Yet practical steps to reduce the virus con- panies. However, the number of contaminations reported to the tamination risk are a challenge, especially for organizations that CAACB per company does not correlate with total manufacturing are without current institutional practices focused on viral safety volume, implying that a combination of circumstance, manufactur- and that may have limited resources. Below, we outline some of the ing controls in place, and prior lack of virus contamination disclo- key viral safety challenges and detail how lessons from the CAACB sures may have contributed to this rate. These data also highlight Virus Contamination in Biomanufacturing Study can be leveraged that no manufacturer is immune from a contamination event. to ensure the safety of these emerging products. Second, CHO cell cultures were contaminated by viruses different from those contaminating human or primate cell lines Virus contamination risks. The three main risks for viral contami- (Table 2). The sources of the viruses contaminating CHO cell cul- nation in cell culture for therapeutic production are cell sources, ture and human or primate cell culture were also different. The materials used in cell culture, and exposure of the cell culture pro- implication is that different host cells may require the consider- cess stream to the operator or environment. We examine each risk ation and management of different virus contamination risks, with in detail below. human and primate cell lines being more susceptible to contamina- In the case of cell sources, both recombinant biopharmaceutical tion from operators. products and viral vector gene therapy products have a low risk of Our data also provide a clear demonstration of the current limits contaminated starting cell sources as both manufacturing processes of virus testing in ensuring viral safety. Testing bioreactor-harvest start with exhaustively characterized master cell banks. For alloge- samples using the IVV assay in runs contaminated with virus was neic therapies in which cells from one donor are used to create thera- negative for virus in 4 of 14, or 28.6% of, cases reported to the pies for multiple patients, the donor cells should also be characterized CAACB (Table 3). These false negatives are due either to the virus to assure they are virus free, per regulatory guidance. Conversely, not replicating in the indicator cell lines chosen for the test, to viral autologous cell therapy products originate from the collection of replication not causing measurable cytopathic effect in the chosen cells from human blood or tissues each time a production process is indicator cells, or to replication of the viral isolate occurring too initiated. Tests to assure that the derived cells are free of adventitious slowly to be detected by the end of the test. The IVV assay also takes virus generally cannot be completed before initiating cell therapy 14 to 28 days—much too long to assess the reactor for contami- manufacturing, and the process generally proceeds at risk. As previ- nation before processing downstream for purification. Therefore, ously noted, human cells are more susceptible to the replication of a several respondents implemented PCR assays as a rapid virus test variety of human viruses than are CHO cells (Table 2). before bioreactor harvest. When a viral contaminant matched a Cell culture processes employed for the manufacture of bio- PCR target, this was effective at preventing contamination of the pharmaceutical products and ATMP products all utilize a variety entire manufacturing facility. Finally, in the events reported to the of basal medium formulations comprised of a mixture of more than CAACB, testing raw materials was found to have limited value. In 50 essential nutrients (for example, amino acids, vitamins and trace the 11 contamination events when raw materials were identified as elements) and other chemicals. These are filter sterilized before use, the source of the viral contaminant, initial testing of that raw mate- typically with 0.1-µm-rated sterilizing-grade filters through which rial did not detect virus. Only after the viral load was increased, most viruses will pass. If any components of media are contami- through concentration or biological amplification, was the virus nated with virus during their manufacture or handling, they may detected in the raw material, and then only in 3 events (Fig. 1). initiate an infection during the cell culture process. Animal-derived Since testing has clear limitations, several companies have (Table 2) and human-derived components (for example, serum and focused on prevention by implementing or exploring methods to growth factors), which carry a higher risk of virus contamination

568 Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology NATuRe BIoTecHnology Perspective

When considering the application of these three approaches to Table 4 | Common viral clearance methods, their suitability for virus risk mitigation of ATMPs, virus clearance is the weakest link use with adeno-associated virus (AAV) and lentiviral vectors, in ATMP virus safety. Many of the virus clearance unit operations and potential approaches to viral clearance used during purification of therapeutic proteins described above are Clearance method AAV Lentivirus not suitable for use with, or have not been widely applied to, ATMPs. Size 20 nm; Size 80–100 nm; If the product itself is a virus or a living cell, how will potential viral non-enveloped enveloped contaminants be removed or inactivated? In the case of viral vectors for gene therapy, characteristics of commonly used vectors could be Heat Heat stable; application Heat sensitive; may not exploited to provide differential clearance from many potential con- of heat will inactivate be a suitable clearance taminating viruses53. Two examples of how common virus removal heat-sensitive viruses method approaches can be used with different viral vectors are shown in with minimal impact on AAV Table 4. These differential clearance strategies, coupled with an understanding of the most likely viral risks, potential sources of Low pH Low-pH stable; a hold pH sensitive; may not virus, and host cell line susceptibility to those viruses, could enable at low pH will inactivate be a suitable clearance the development of a virus removal strategy. pH-sensitive viruses method In the case of living cell-based therapies, viral clearance would with minimal impact be required to remove or inactivate viruses in the cell culture super- on AAV natant, as well as separating or destroying infected cells, which har- Solvent/detergent Non-enveloped virus; Enveloped virus bor virus, from any cells not infected with virus. To our knowledge, detergent can be used sensitive to detergent; there is currently no technology capable of meeting this challenge. to inactivate enveloped not a suitable clearance Additionally, none of the virus inactivation methods used for tra- virus with minimal method ditional biopharmaceutical manufacturing is compatible with the impact on AAV survival of living cells. Therefore, the viral safety of cell therapies Chromatography Differences in surface Differences in currently relies solely on contamination prevention and in-process charge can allow AAV surface charge can detection and lot rejection. to be separated from allow lentivirus to be other viruses separated from other Contamination prevention. Viral-vectored gene therapy products viruses utilize plasmids or recombinant viruses to initiate production54. Nanofiltration AAV will pass through, Lentivirus will be Plasmids are generated in prokaryotic cells and should be free of for example, 35-nm retained by nanofilters viruses that would replicate in mammalian cell cultures. For recom- filters, allowing (pore size, for example, binant viruses, master virus banks are generated and thoroughly separation of AAV from 35 or 50 nm), potentially characterized for contamination by adventitious viruses55. While larger viral contaminants allowing separation it is challenging to test for adventitious viruses in the presence of from smaller viral recombinant viral stocks, approaches to develop successful virus contaminants tests have been developed and used56. Additionally, new detection technologies, such as high-throughput sequencing (HTS), have detected adventitious viral contaminants in the presence of virus than other components50, are commonly added to media for ATMP product57 and are being explored for use in the industry58,59. production51. With the exception of some legacy products, these Our study showed that the control of raw materials by direct components are generally not added to media for protein and vac- testing had limited value (Fig. 1), either because virus concentra- cine manufacturing. tions in the raw material are below the assay detection limit or The final viral contamination risk is exposure of cell-culture because the contaminating virus was not homogeneously distrib- process streams to potential virus contamination from the envi- uted in the raw material. Therefore, as noted earlier, a number of ronment, including operators, especially during open manufac- biopharmaceutical manufacturers treat media before use to inac- turing steps (for example, vessel transfers). For both traditional tivate or remove potential viruses in raw materials (Fig. 4), and biopharmaceutical and ATMP manufacturing processes, such open this approach should be strongly considered by manufacturers of processes are performed in a well-controlled environment by expe- ATMPs where feasible. rienced, gowned and masked operators following aseptic proce- The use of animal- and human-derived raw materials during dures, yielding a so-called operationally closed process. However, ATMP production increases viral contamination risk. When use due to the scale of their manufacture, ATMPs may rely much more of these materials is unavoidable, one risk-mitigation strategy is to than recombinant proteins and vaccines on open cell culture trans- increase the viral titer of a potential contaminant to a detectable fers. The result is an increased possibility of virus contamination level in a high-risk raw material through biological amplification or from open operations for these types of products. concentration. This enabled the detection of virus in a raw material in 3 of 11 cases, or 27%, reported to the CAACB (Fig. 1). Virus risk-mitigation strategies. As noted above, virus-risk mitiga- Another option is treating a high-risk material to reduce con- tion for biopharmaceutical manufacturing uses three complemen- tamination risk. As an example, gamma irradiation of serum has tary approaches: (i) prevention of virus entry by selecting low-risk been shown to be effective against several viruses60. It is not yet stan- starting and raw materials and using manufacturing controls; (ii) dard practice for human serum, but should be strongly considered testing of in-process materials to ensure they are free of virus and to reduce the risk of these raw materials if safer alternatives are not enable lot rejection; and (iii) clearance of virus (inactivation and/ suitable. Despite the effectiveness of treating raw materials, we note or removal) from the product9,14. Of the three, virus clearance has that some animal- or human-derived materials may be sensitive to been shown to be of extraordinary importance in reducing the risk heat, radiation or UV exposure, which may ultimately affect cell of virus contamination of final product (Fig. 2)52. A key question growth and performance. then is: can the risk mitigation approaches used for traditional bio- A better strategy, also used in the recombinant protein manufac- pharmaceutical manufacturing be applied to gene therapy and cell turing industry, is to develop media that do not require animal- or therapy manufacturing? human-sourced raw materials61. This can be a challenge for some

Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology 569 Perspective NATuRe BIoTecHnology

ATMPs, especially those where nutrient requirements are not well rapid results. Despite this, appropriate virus detection assays should understood, such as primary cell cultures, or which may have vari- be added to the current safety testing done on each lot. Final test able starting cells, such as autologous cell therapies. results would likely come after an autologous cell therapy treatment In addition to media, ancillary materials used in the produc- had begun but would enable informed patient treatment decisions tion of many cell therapy products, such as monoclonal antibodies should a virus contaminant be detected or suspected. In summary, and retrovirus vectors, will each need to be assessed individually virus control efforts for organizations developing and operating for their virus risk to assure that they are free of adventitious virus ATMP manufacturing processes should focus primarily on meth- before use in the cell-therapy manufacturing process. ods to prevent virus contamination in the first place, although we Virus contamination during ATMP cell culture manufactur- would note that best practices may change as new technologies are ing processes by environmental sources must be strictly avoided. developed to meet current challenges in testing and viral clearance This can be done by the use of functionally closed systems, which for ATMPs. often employ single-use, disposable equipment. If cell culture transfers cannot use closed transfer systems, they must be conducted in Conclusions hoods supplied with HEPA-filtered air by appropriately gowned The biotechnology industry has a long history of supplying safe operators using aseptic technique. This is often more challenging and effective therapies to patients owing to the extensive controls in in ATMP manufacturing as there are more open manipulations that place to ensure product safety. Despite these controls, viral infection provide opportunity to introduce an environmental contaminant of cell culture is a real risk with severe consequences. Learning from and many small lots may be manufactured at once. these events has historically been a challenge; the work presented here represents a comprehensive collection and analysis of previ- In-process testing and lot rejection. Testing for adventitious ously unpublished industry-wide viral contamination information. virus contamination at relevant points in the cell culture manu- The CAACB study has identified five viruses that have been shown facturing process, usually just before production cell culture har- to contaminate CHO cell culture and four viruses that have con- vest, has been conducted in recombinant protein manufacture taminated cell culture of human or primate cells. Importantly, the for many years. The current gold standard for lot release testing viruses that have been shown to contaminate human or primate cell in recombinant protein products is the IVV assay, a cell-based lines can also infect humans. The choice of which cell line to use for assay that has a large breadth of detection for potential viral con- recombinant protein or vaccine production is a complicated deci- taminants. However, our study demonstrated that the testing of sion, of which viral contamination risks are just one consideration. bioreactor pre-harvest samples using the IVV assay in runs con- However, manufacturers that are using human or primate cells taminated with virus was unable to detect virus in about one should be aware of the difference in the potential risk to patients quarter of the cases (Table 3). An additional challenge is that the from a viral contaminant in products produced in those cells com- IVV assay takes between 14 and 28 days to complete39,56,62,63 and is pared with CHO cells. not amenable to the rapid release required of some ATMP prod- Although testing is a key component of viral safety in biotech- ucts. Nucleic acid-based assays, such as PCR, are faster than the nology products, the data presented here indicate that testing IVV assay, taking less than a day. Yet PCR assays require prior alone is not enough to ensure that a given product is free of a viral knowledge of potential contaminants and only detect viral nucleic contaminant, and that a holistic, multifaceted approach must be acids. HTS provides a greater breadth of detection than PCR and taken. This is never more true than when faced with a previously is seeing widespread interest from the vaccine and recombinant unknown emerging virus, such as SARS-CoV-2, where the capac- protein industry59. However, current HTS sample preparation ity of the virus to infect production cell lines or be detected in approaches and bioinformatic pipelines are not as rapid as PCR existing assays is not initially known. Some approaches, such as the and can take 7–10 days58. Additionally, determining if the con- implementation of rapid PCR tests for forward processing deci- taminant identified in a nucleic acid-based assay is biologically sions, have been shown to enhance containment and prevent the active may require a different method, though we would note spread of a contaminating virus to other parts of the manufactur- that HTS of viral RNA has been used to demonstrate a virus is ing facility. We believe that collective effort and shared knowledge biologically active64. Despite these challenges, testing should be can ensure the continued success of the life-saving therapies of conducted for ATMP production for samples taken before virus today and tomorrow. harvest (for viral-vectored gene therapy products) and at the end Finally, lessons from the CAACB study, applied to emerg- of the manufacturing process (for cell therapy products) so that ing biotech products, lead us to conclude that the viral safety of if contamination with an adventitious virus is detected, informed some ATMPs rely almost exclusively on preventing contamination decisions regarding product lot rejection can be made. through the use of rigorous process controls. Further, the short time Because of the limitations of viral clearance for ATMPs and the frame associated with the use of many ATMPs, relative to their short shelf lives of autologous cell therapy products, it would be manufacture, is a challenge for current viral testing paradigms and prudent for companies developing ATMP manufacturing processes offers a clear opportunity for technological advancement. and analytical methods to focus on viral detection methods, such as HTS, that are rapid, broad spectrum, and more sensitive than those Received: 4 April 2018; Accepted: 1 April 2020; traditionally used. Published online: 27 April 2020 On the basis of the lessons learned from the CAACB virus contamination project and the discussion above, it can be concluded that, at the current state of technological development, the viral References 1. Merten, O. W. Virus contaminations of cell cultures — a biotechnological safety of some ATMPs, especially autologous cell therapies, will rely view. Cytotechnology 39, 91–116 (2002). almost exclusively on preventing a contamination through the use of 2. Sawyer, W. A. et al. Jaundice in army personnel in the western region of the rigorous process barriers (for example, treatment of media, reduc- United States and its relation to vaccination against yellow fever: Part I. Am. tion in the use of high-risk materials, testing of high-risk materials J. Epidemiol. 39, 337–430 (1944). that cannot be eliminated from use or treated to reduce risk, and 3. Shah, K. & Nathanson, N. Human exposure to SV40: review and comment. Am. J. Epidemiol. 103, 1–12 (1976). closed manufacturing systems). In-process virus testing, particu- 4. Stratton, K., Alamario, D. A. & McCormick, M. C. Immunization Safety larly for autologous cell therapies, has clear limitations. Current Review: SV40 Contamination of Polio Vaccine and Cancer. (National approaches cannot provide both broad-spectrum detection and Academies Press, Washington, DC, 2003).

570 Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology NATuRe BIoTecHnology Perspective

5. Rosendaal, F. R., Smit, C. & Briët, E. Hemophilia treatment in historical 39. Gombold, J. et al. Systematic evaluation of in vitro and in vivo adventitious perspective: a review of medical and social developments. Ann. Hematol. 62, virus assays for the detection of viral contamination of cell banks and 5–15 (1991). biological products. Vaccine 32, 2916–2926 (2014). 6. Weinberg, P. D. et al. Legal, fnancial, and public health consequences of HIV 40. International Society for Pharmaceutical Engineering. Baseline Guide Volume contamination of blood and blood products in the 1980s and 1990s. 6: Biopharmaceutical Manufacturing Facilities (International Society for Ann. Intern. Med. 136, 312–319 (2002). Pharmaceutical Engineering, 2013) 7. Butler, D. Petition prompts backlash against scientists. Nature 367, 304 41. Brorson, K. et al. Current and future approaches to ensure the viral safety of (1994). biopharmaceuticals. Dev. Biol. (Basel) 118, 17–29 (2004). 8. Barker, S. & Swinbanks, D. Health ministry accepts liability in Japanese 42. Farshid, M., Tafs, R. E., Scott, D., Asher, D. M. & Brorson, K. Te clearance HIV-infected blood row. Nature 379, 663–664 (1996). of viruses and transmissible spongiform encephalopathy agents from 9. International Council for Harmonisation of Technical Requirements for biologicals. Curr. Opin. Biotechnol. 16, 561–567 (2005). Pharmaceuticals for Human Use. Q5A(R1). Viral Safety Evaluation Of 43. Anonymous. Te 2019 16th Annual Report and Survey of Biopharmaceutical Biotechnology Products Derived From Cell Lines Of Human Or Animal Origin Manufacturing Capacity and Production (BioPlan Associates, Rockville, MD, (International Council for Harmonisation of Technical Requirements for USA, 2019). Pharmaceuticals for Human Use, 1999). 44. Schleh, M. et al. Susceptibility of mouse minute virus to inactivation by heat 10. Suomela, H. Inactivation of viruses in blood and plasma products. Transfus. in two cell culture media types. Biotechnol. Prog. 25, 854–860 (2009). Med. Rev. 7, 42–57 (1993). 45. Murphy, M., Quesada, G. M. & Chen, D. Efectiveness of mouse minute virus 11. Burnouf, T. Modern plasma fractionation. Transfus. Med. Rev. 21, inactivation by high temperature short time treatment technology: a statistical 101–117 (2007). assessment. Biologicals 39, 438–443 (2011). 12. Walsh, G. Biopharmaceutical benchmarks 2014. Nat. Biotechnol. 32, 46. Meunier, S. M., Sasges, M. R. & Aucoin, M. G. Evaluating ultraviolet 992–1000 (2014). sensitivity of adventitious agents in biopharmaceutical manufacturing. 13. Grillberger, L., Kreil, T. R., Nasr, S. & Reiter, M. Emerging trends in J. Ind. Microbiol. Biotechnol. 44, 893–909 (2017). plasma-free manufacturing of recombinant protein therapeutics expressed in 47. Mundt, W. et al. Virus fltration of cell culture media. Patent mammalian cells. Biotechnol. J. 4, 186–201 (2009). WO2013192009A1 (2003). 14. Kozlowski, S. & Swann, P. Current and future issues in the manufacturing 48. Mullard, A. Second anticancer CAR T therapy receives FDA approval. and development of monoclonal antibodies. Adv. Drug Deliv. Rev. 58, Nat. Rev. Drug Discov. 16, 818 (2017). 707–722 (2006). 49. Anonymous. FDA approves hereditary blindness gene therapy. 15. Liu, S. et al. Development and qualifcation of a novel virus removal flter for Nat. Biotechnol. 36, 6 (2018). cell culture applications. Biotechnol. Prog. 16, 425–434 (2000). 50. Mahmood, A. & Ali, S. Microbial and viral contamination of animal and 16. Bethencourt, V. Virus stalls Genzyme plant. Nat. Biotechnol. 27, 681 (2009). stem cell cultures: common contaminants, detection and elimination. 17. Allison, M. As Genzyme founders, competitors and activist investors swoop J. Stem Cell Res. Ter. 2, 149–155 (2017). in. Nat. Biotechnol. 28, 3–4 (2010). 51. Brindley, D. A. et al. Peak serum: implications of serum supply for cell 18. Burstyn, D. G. Contamination of genetically engineered Chinese hamster therapy manufacturing. Regen. Med. 7, 7–13 (2012). ovary cells. Dev. Biol. Stand. 88, 199–203 (1996). 52. Shukla, A. A. & Aranha, H. Viral clearance for biopharmaceutical 19. Rabenau, H. et al. Contamination of genetically engineered CHO-cells downstream processes. Pharm. Bioprocess. 3, 127–138 (2015). by epizootic haemorrhagic disease virus (EHDV). Biologicals 21, 53. Ye, G. J. et al. Herpes simplex virus clearance during purifcation of a 207–214 (1993). recombinant adeno-associated virus serotype 1 vector. Hum. Gene Ter. Clin. 20. Garnick, R. L. Experience with viral contamination in cell culture. Dev. 25, 212–217 (2014). Dev. Biol. Stand. 88, 49–56 (1996). 54. Wirth, T., Parker, N. & Ylä-Herttuala, S. History of gene therapy. Gene 525, 21. Garnick, R. L. Raw materials as a source of contamination in large-scale cell 162–169 (2013). culture. Dev. Biol. Stand. 93, 21–29 (1998). 55. US Food and Drug Administration. Draf Guidance for Industry: Chemistry, 22. Kiss, R. D. Practicing safe cell culture: applied process designs for Manufacturing, and Control (CMC) Information for Human Gene Terapy minimizing virus contamination risk. PDA J. Pharm. Sci. Technol. 65, Investigational New Drug Applications (INDs) (US Food and Drug 715–729 (2011). Administration, 2018). 23. Ault, A. US FDA again stops urokinase sales. Lancet 354, 310 (1999). 56. US Food and Drug Administration. Guidance for Industry: Characterization 24. Oehmig, A. et al. Identifcation of a calicivirus isolate of unknown origin. and Qualifcation of Cell Substrates and Other Biological Materials Used in the J. Gen. Virol. 84, 2837–2845 (2003). Production of Viral Vaccines for Infectious Disease Indications (US Food and 25. Nims, R.W. et al. Detection of Cache Valley virus in biologics manufactured Drug Administration, 2010). in CHO cells. Biopharm. Int. 21, 89–95 (2008). 57. Victoria, J. G. et al. Viral nucleic acids in live-attenuated vaccines: 26. Rubino, M. J. Experiences with HEK293: a human cell line. PDA J. Pharm. detection of minority variants and an adventitious virus. J. Virol. 84, Sci. Technol. 64, 392–395 (2010). 6033–6040 (2010). 27. Plavsic, M. et al. Caliciviridae and vesivirus 2117. Bioprocess. J. 9, 58. Khan, A. S. et al. A multicenter study to evaluate the performance of 6–12 (2011). high-throughput sequencing for virus detection. mSphere 2, 28. Plavsic, M., Shick, K., Bergmann, K. F. & Mallet, L. Vesivirus 2117: cell line e00307–17 (2017). infectivity range and efectiveness of amplifcation of a potential adventitious 59. Khan, A. S. et al. Advanced virus detection technologies interest group agent in cell culture used for biological production. Biologicals 44, (AVDTIG): Eforts on high throughput sequencing (HTS) for virus detection. 540–545 (2016). PDA J. Pharm. Sci. Technol. 70, 591–595 (2016). 29. Qiu, Y. et al. Identifcation and quantitation of vesivirus 2117 particles in 60. Gauvin, G. & Nims, R. Gamma-irradiation of serum for the inactivation of bioreactor fuids from infected Chinese hamster ovary cell cultures. adventitious contaminants. PDA J. Pharm. Sci. Technol. 64, 432–435 (2010). Biotechnol. Bioeng. 110, 1342–1353 (2013). 61. Miki, H. & Takagi, M. Design of serum-free medium for suspension culture 30. Moody, M., Alves, W., Varghese, J. & Khan, F. Mouse minute virus (MMV) of CHO cells on the basis of general commercial media. Cytotechnology 67, contamination—a case study: detection, root cause determination, and 689–697 (2015). 62. US Food and Drug Administration. Points to Consider in the Characterization corrective actions. PDA J. Pharm. Sci. Technol. 65, 580–588 (2011). of Cell Lines Used to Produce Biologicals (US Food and Drug Administration, 31. Baylis, S. A., Finsterbusch, T., Bannert, N., Blümel, J. & Mankertz, A. Analysis 1993). of porcine circovirus type 1 detected in Rotarix vaccine. Vaccine 29, 63. World Health Organization. Requirements for the use of animal cells as 690–697 (2011). in vitro substrates for the production of biologicals. in WHO Expert 32. McClenahan, S. D., Krause, P. R. & Uhlenhaut, C. Molecular and infectivity Committee on Biological Standardization, 47th Report (World Health studies of porcine circovirus in vaccines. Vaccine 29, 4745–4753 (2011). Organization, 1998). 33. Besselsen, D. G. et al. Molecular characterization of newly recognized rodent 64. Brussel, A. et al. Use of a new RNA next generation sequencing approach for parvoviruses. J. Gen. Virol. 77, 899–911 (1996). the specifc detection of virus infection in cells. Biologicals 59, 29–36 (2019). 34. Nims, R. W. Detection of adventitious viruses in biologicals—a rare occurrence. Dev. Biol. (Basel) 123, 153–164; discussion 183–197 (2006). 35. Willkommen, H. et al. Meeting report: 2013 PDA Virus & TSE Safety Forum. Acknowledgements PDA J. Pharm. Sci. Technol. 68, 193–214 (2014). The authors would like to acknowledge the support of the members of the Consortium 36. Anonymous. A united front. Nature 472, 389–390 (2011). on Adventitious Agent Contamination in Biomanufacturing (CAACB). CAACB 37. Campbell, G. L. et al. Second human case of Cache Valley virus disease. members, past and present, include Amgen, Asahi Kasei Bioprocess, AstraZeneca, Emerg. Infect. Dis. 12, 854–856 (2006). Biogen, BioMarin, Boehringer Ingelheim, Brammer Bio, Bristol-Myers Squibb, Charles 38. Sexton, D. J. et al. Life-threatening Cache Valley virus infection. N. Engl. J. River Labs, CSL Behring, Eli Lilly and Co., EMD Serono, Genentech, Gritstone Med. 336, 547–549 (1997). Oncology, Histogenics, LFB, Merck & Co., Merrimack Pharmaceuticals, MilliporeSigma,

Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology 571 Perspective NATuRe BIoTecHnology

Novartis, Pall Corporation, Pfizer, Regeneron, Sanofi, Sanofi Genzyme, Sanofi Pasteur, Additional information Takeda and Thermo Fisher. Supplementary information is available for this paper at https://doi.org/10.1038/ s41587-020-0507-2. Competing interests Correspondence should be addressed to S.L.S. P.W.B., M.E.W., J.C.L., I.T.M.H., F.J.K., J.W., A.J.S. and S.L.S. performed the research Reprints and permissions information is available at www.nature.com/reprints. at the Massachusetts Institute of Technology with funding from CAACB member companies; J.B., A.B., D.C., M.C., H.D., L.G., J.G., D.G., R.K., T.R.K., R.L., Y.L., J.M., L.M., C.M., M. Moody, S.M., M. Murphy, M.P., N.J.R., D.R., M.R., R. Schicho, R. Snyder, Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in D.S. and C.Z. were employees of CAACB member companies while engaged in the published maps and institutional affiliations. reported research. © The Author(s), under exclusive licence to Springer Nature America, Inc. 2020

572 Nature Biotechnology | VOL 38 | May 2020 | 563–572 | www.nature.com/naturebiotechnology