Managing Cell Line Instability and Its Impact During Cell Line Development

June 2, 2009

By considering stability as part of the cell line selection and cell banking paradigm, we can ensure that instability problems are not observed during clinical or commercial manufacturing.


For a Phase 1 antibody project, the primary objective of our cell line development group is to deliver a stable, high performing, clonal cell line in an aggressive timeframe, leaving little time to react to the challenges associated with cell line instability, such as dramatically reduced product expression. Our current cell line development timeline allows for 6.5 months from transfection to the establishment of a cell bank for a lead clone, which is expected to deliver 1–3 g/L in a fed-batch process. Our cell line selection process consists of numerous rounds of successive screening, initially assessing hundreds of clones. During that time, clones are eliminated because of poor performance in scaled-down models of the fed-batch process or because of a loss of product expression over time. When such instability in product expression occurs for a lead clone at a late stage in the timeline, both upstream and downstream process groups must quickly identify alternatives so that the material needs of the project will be satisfied and deadlines are still met.

This manuscript presents examples of cell line instability that we have encountered during Phase 1 antibody projects, and the operational consequences that have ensued. It describes the different types of instability we have observed, our attempts to analyze the presumptive underlying genotypic causes of these different types of instability, and the general methods and tools that we use to investigate cell line instability. Finally, we describe the recent development of an RT-PCR–based assay that allows us to eliminate potentially unstable cell lines from consideration early in the cell line development timeline.

There are various approaches to developing a therapeutic antibody product, and many development organizations now use platform, or standard procedures, (see Figure 1 for an example) because this approach works well for achieving the objectives of delivering a stable, high performing, clonal cell line in a short timeframe. The goal of using a platform approach is to enable a greater number of molecular entities through the pipeline, and to decrease both the length of time and the level of investment in developing these molecular entities before they have proven efficacy in human clinical trials. Cell line development platforms rely primarily on multiple rounds of successive screening of large numbers of clones with the recognition that clones will be eliminated over the course of development because they have not performed sufficiently well in a scaled-down model of the fed-batch production process or they have lost expression of the protein product, even in the presence of selective pressure. During the cell line adaptation period shown in Figure 1, cell line expression characteristics can be evaluated and cell banks for each clone can be generated at some frequency. Hence, an enabling aspect of the design of the platform shown in Figure 1 is that the phenotypic stability of a clone can be established before its selection as the production cell line. By considering stability as part of the cell line selection and cell banking paradigm, we can ensure that the problems encountered and described here during the cell line development window are not observed during the clinical, or future commercial, manufacturing window.

Wyeth BioPharma

Although our cell line development group has been successful in meeting our project goal of generating Chinese hamster ovary (CHO) Phase 1 cell lines that can deliver >1 g/L in a platform production process, cell line instability occurring at a late stage in the adaptation period timeline has resulted in both upstream and downstream process groups scrambling to deal effectively with last-minute changes to meet deadlines. Hence, we have developed specific methods and approaches to enable an increased understanding of the root causes of our observed cell line instability. This knowledge potentially could be used for the early, proactive identification of unstable clones or the elimination of cell line instability altogether. This would be highly beneficial to the cell line and cell culture process development organization, thus ensuring sustainable success in meeting deliverables and timelines.

This current work describes the types of instability we observe: "acute," a precipitous loss of product expression occurring over a short time span of approximately 2–4 weeks, and "gradual," a slow decline of product expression occurring over the months of the cell line adaptation period shown in Figure 1. In addition, we describe our attempts to analyze the presumptive underlying genotypic causes of these different manifestations of instability, which include, but are not limited to, DNA rearrangements and DNA methylation. We discuss the general methods and tools that we use to investigate cell line instability, including Southern and northern blot analysis, PCR-based methods, and an assay to detect methylated DNA, and how, when, and why we use these methods. Based on results from an intensive study examining the most frequent causes of cell line instability at the molecular level in our Phase 1 antibody projects, we have developed and implemented an RT-PCR–based cell line screening assay that enables the elimination of unstable cell lines from consideration early in the platform process.

Figure 1. A prototype platform cell line development timeline for a Phase 1 antibody project. Following transfection and selection procedures, clonal transfectants are isolated and subjected to successive high-throughput screens. The top performers from this screening process then undergo a cell line adaptation period, consisting of several months. During this time, cell line stability information is collected, and the clones are screened in a series of production fed batches, first at small scale, and then in benchtop bioreactors. A production clone is selected, the final production process is defined, and a master cell bank is generated.

The Causes of Common Types of Cell Line Instability and Their Impact

Figure 2 shows the stability profiles for the 24 highest performing clones for a Phase 1 antibody project. These 24 clones were selected from hundreds of clones that were initially subjected to two rounds of high throughput clone screening. The figure illustrates not only the types of instability we observe (acute and gradual), but also shows how the performance of high-performing clones may change in even a very short amount of time. This is clearly seen for the nine clones, designated with red lines and symbols, where a 50% drop in expression occurred within two weeks, and product expression was completely lost within 50 days of continuous culturing. The ramifications of this lack of predictability in clone performance can affect the entire upstream and downstream process development teams during the critical period when they are striving to bring together the clone, the cell culture process, and the purification process to achieve the desired performance and product quality.

Figure 2. Cell line instability manifests itself in various forms. Plotted is the specific productivity (Qp, picograms/cell/day) for 24 clonal cell lines for a Phase 1 antibody project during 90 days of continuously culturing the cells. Cell lines showing a stable Qp profile are marked by black symbols, cell lines showing gradual instability are marked by blue symbols, and cell lines showing acute instability are marked by red symbols.

Variability in Instability

To further emphasize this point regarding unpredictability, Table 1 shows the frequency of clones exhibiting acute and gradual instability for seven different antibody projects for which the same platform procedures were used. For each project, a total of 24 clones were assessed. The numbers illustrate that the level of instability is quite variable from project to project. This variability in instability emphasizes our inability to consistently predict how our system will perform. We therefore attempted to determine the causes of instability we observe, and to see if this information might enable us to take corrective action to reduce or eliminate instability from our cell line development process.

Table 1. Frequency of acute and gradual instability in Phase 1 antibody projects. This table contains information on the number of clones demonstrating acute and gradual cell line instability in seven different Phase 1 antibody projects. For each project, 24 adapted clones were monitored over a similar length of time.

The results presented in Figure 3 are representative of one type of instability that was observed in several Phase 1 antibody projects. Platform procedures, as outlined in Figure 1, were followed for all activities from transfection through cell line adaptation to cell banking. As shown in Figure 3A, a high producing clone from this project demonstrated a dramatic and eventually complete loss of antibody expression (as marked by the red line and triangles in Figure 3A). RNA was prepared at the time points marked by the black triangles in Figure 3A, and was run on the northern blot shown in Figure 3B, which was hybridized with a probe specific for the selectable marker dihydrofolate reductase (DHFR). At the earlier time points, the expected heavy chain (HC)-DHFR bicistronic transcript (marked by an arrow, labeled "HC-DHFR") was seen. Over time, however, there was a detectable loss of this HC-DHFR bicistronic transcript with the appearance of a smaller transcript (also marked by an arrow, labeled "rearranged-DHFR"). To show that each lane contained equivalent amounts of RNA, the blot was subsequently hybridized with a probe encoding the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH), and similar signal intensities were observed across all the lanes. Cloning and sequencing the smaller transcript demonstrated that it lacked any HC sequence, but it still contained an intact DHFR gene. These results indicated that DHFR and HC had become uncoupled: although this clone continued to synthesize DHFR and to grow in the presence of methotrexate, it was no longer producing HC protein or intact antibody.

Figure 3. Rearranged DHFR in an unstable Phase 1 antibody cell line. A) The small red triangles mark the specific productivity (Qp, picograms/cell/day) for a Phase 1 antibody expressing cell line over 150 days in continuous culture. The large black triangles denote the time points (34, 62, 102, 119, and 147 days in culture) at which RNA was prepared for the northern blot shown in Panel B. B) Each lane was loaded with 3 μg of isolated RNA, the gel was transferred to nitrocellulose, and the nitrocellulose was hybridized with probes encoding DHFR and GAPDH. Migration of HC-DHFR, rearranged DHFR, and GAPDH (control) are as indicated.

In referring back to the timeline depicted in Figure 1, this clone had already made it through clone screening and small-scale productivity assessments, and its performance was in the process of being evaluated in benchtop bioreactors. It was considered to be a top performer and a favorite for being selected as the production clone for this project. However, because of its instability, this clone was immediately eliminated as a contender, and both the upstream and downstream groups began to focus their efforts on an alternative clone whose performance was adequate, but relatively inferior, compared to the unstable clone before it became unstable. Although deadlines and material needs were successfully met for this project, some additional bench-scale bioreactor work had to be performed at a late stage on the timeline to sufficiently assess the performance of the alternative clone.

An Alternative Vector Strategy

Following these results and additional work, alterations were made to our vector strategy. A head-to-head comparison of clones resulting from transfections with both the newly engineered vectors and the previously used ("old") vectors, with the same antibody molecule, showed no differences in recombinant protein expression or cell growth performance during clone screening before cell line adaptation (data not shown). High-producing clones from this head-to-head comparison were grown in continuous culture, and RNA was prepared at various time points and was run on a northern blot hybridized with probes encoding DHFR and GAPDH.

Rearranged DHFR

The results are shown in Figure 4A. The lane labeled "old vector clone" contained RNA prepared at a single time point from a single clone derived using the previous vector system, and showed that there was rearranged DHFR transcript present. The other lanes (labeled 1, 2, 3, 4, 5) contained RNA from five clones arising from transfection with the newly engineered vectors at several time points (these time points ranged from 11 to 113 generations). None of these clones were found to have rearranged DHFR. Information contained in Figure 4B shows that of the 12 high-producing clones derived using the new vectors, none exhibited rearranged DHFR transcripts, whereas 9 of the 12 high producing clones derived using the previous ("old") vectors demonstrated a loss of product expression during the same time period, and these latter clones were all found to have rearranged DHFR.

Figure 4. Alterations in antibody vector constructs result in improved cell line stability. A) RNA was prepared at indicated time points (in generations) from five clones derived from a transfection using newly engineered vector constructs, and from a single clone derived from a transfection using previous "old" vector constructs. Each lane was loaded with 3 μg of RNA, the gel was transferred to nitrocellulose, and the nitrocellulose was hybridized with probes encoding DHFR and GAPDH. Migration of HC-DHFR, rearranged DHFR, and GAPDH (control) are as indicated. B) Refer to text for explanation of table.

Although the alternative vector strategy significantly reduced the frequency of DHFR rearrangement, it did not completely eliminate the ability of cells to uncouple DHFR and HC. This is represented by the data shown in Figure 5. Using the new vector strategy, platform procedures were followed for all activities from transfection through cell line adaptation to cell banking. Figure 5A shows the growth and productivity profiles for a clone that demonstrated a cell-specific productivity (Qp) of >30 picograms per cell per day out to 42 generations. From 42 generations onward, the growth rate (marked by the red line and triangles) of the clone appeared relatively stable, however, a precipitous drop in antibody expression (Qp shown in the black line and triangles) was observed starting at about 77 generations, resulting in a total loss of product expression by 107 generations. RNA was prepared at several time points (marked by blue generation numbers on Figure 5A) during the time that this clone was continuously cultured, and presented in Figure 5B is the northern blot of these RNA samples hybridized with a DHFR probe.

Figure 5. DNA rearrangement in an unstable Phase 1 antibody cell line. A) A clone from a Phase 1 antibody project was continuously cultured for 107 generations. Plotted are the specific productivity (Qp, picograms/cell/day) designated by black triangles, and the growth rate (hr-1) designated by red triangles. RNA was isolated at the time points (19, 42, 63 and 100 generations) indicated on the graph. B) Each lane was loaded with 3 μg of isolated RNA, the gel was transferred to nitrocellulose, and the nitrocellulose was hybridized with a probe encoding DHFR. The migration of rearranged DHFR is indicated.

As shown, there was little change in the level of expected HC-DHFR bicistronic transcript from 19 to 63 generations. By 100 generations, however, there was a detectable loss of the expected HC-DHFR transcript concomitant with the appearance of an even smaller, rearranged DHFR transcript than was previously observed (see Figures 3B and 4A). The cloning and sequencing of this smaller transcript demonstrated that it contained an intact DHFR-coding sequence and no HC sequence. Thus, the cells were able to grow in the presence of methotrexate by continuing to synthesize DHFR. By 107 generations, however, they were clearly expressing no HC protein and therefore, no intact antibody.

Detecting DHFR Rearrangement

Further investigation of other unstable clones, by Southern and northern blot analyses, RT-PCR, and cloning and sequencing, was performed to further characterize instances of DHFR rearrangement. In all cases studied (data not shown), our results indicated that the HC gene had become deleted, or "looped out." As this phenomenon continued to be problematic, we developed the RT-PCR–based assay shown in Figure 6 to enable the early and efficient detection of rearranged DHFR transcripts in cell line candidates.

As depicted in Figure 6A, forward and reverse primers were designed to span the HC coding sequences lost as a result of rearrangement within our HC-DHFR-containing vector. Using these primers, RT-PCR performed on RNA prepared from clones, which had not undergone HC rearrangement, would be expected to generate an RT-PCR product of approximately 2,700 bp. However, a rearrangement deleting any of the HC would give rise to an RT-PCR product of less than 2,700 bp, and depending on the extent of HC sequence deletion, a product as small as ~500 bp if only intact DHFR coding sequences remained. Using this assay, referred to as the loop out detection assay (LODA), we analyzed the RNA samples prepared from the clone described in Figure 5. As shown in Figure 6B, a ~500 bp PCR product, indicative of HC rearrangement, was detected in the RNA sample from 63 generations. As this rearrangement was not detected until the 100 generation time point by the northern blot in Figure 5B, LODA was more sensitive than the northern blot in detecting this rearrangement. Additionally, LODA detected the rearrangement before the time point at which instability was marked by a noticeable decline in Qp at ~77 generations in Figure 5A.

Figure 6. Development of the loop out detection assay (LODA). A) Schematic depicting the loop out detection assay. For a more detailed explanation, refer to the text. B) Samples from the same RNA preparations run in the northern blot shown in Figure 5B were subjected to the LODA depicted in Figure 6A. LODA samples were run on an agarose gel which was subsequently stained with ethidium bromide to visualize RT-PCR products. Expected intact product of 2.7 Kb and a smaller product (~500 bp), in which the HC has been "looped out," are marked by arrows.

LODA, which is fairly high throughput, is now used at several key nodes during our cell line development platform timeline. This reduces the probability that the development team will be investing valuable time and resources in moving forward high producing clones that will eventually lose protein expression. In addition, LODA, coupled with sequencing, has enabled us to catalog the sites of the HC rearrangement. This information is being used to assess whether additional modifications to our vector strategy could result in further reductions in the frequency of HC-DHFR uncoupling.

A combination of factors

The acute instability described above is most likely the consequence of a combination of factors. One cause may be our selection scheme, and another might be that CHO cells, although clearly a workhorse of biopharma, were not naturally designed to produce large amounts of recombinant protein. Nonetheless, with the screening procedures and tools we have put in place, the threat of instability to our clone selection process has been greatly reduced.

Gradual Instability Caused by Gene Silencing

In addition to the acute instability caused by DNA rearrangements, we also observe gradual instability which we believe results from epigenetic changes resulting in gene silencing. The scientific literature contains a great deal of information on gene silencing, which is marked by changes in chromatin structure, including the methylation of cytosine residues at specific CpG dinucleotides in DNA and changes in histone protein modifications.1–3 These events inhibit the binding of transcription factors, consequently shutting down transcription. Hence, we wished to ascertain whether DNA methylation might play a role in the gradual instability in our cell lines. A conventional method for determining whether DNA is methylated is to perform a restriction digest with an enzyme whose activity is inhibited by the presence of a methyl group in its CpG-containing recognition sequence, and then to analyze the digest by Southern blot.

The phenotypic stability profile for cell line 16-6F, which was continuously cultured for more than 200 days, showed gradual declines in both titer and Qp. northern blots showed that HC and light chain (LC) transcript levels were decreased, and Southern blots of the coding region demonstrated that HC and LC gene copy number were unchanged (data not shown). Genomic DNA samples, prepared at various time points during the time the cells were cultured, were digested with AatII, which cuts the specific sequence 5'-GACGTC-3' only when the CpG dinucleotide in the sequence is unmethylated. If the sequence is methylated, AatII will not cut. The AatII digested DNA was run on a gel, transferred to nitrocellulose membranes, and hybridized with specific HC and LC probes. As shown in Figure 7, with increasing generations there was progressively less AatII digestion of the DNA, resulting in undigested DNA migrating high up in the lane. A conclusion from these results was that the decline in titer and Qp in the 16-6F cell line correlated with, and may have in part been caused by, DNA methylation.

Figure 7. Evidence for DNA methylation in an unstable cell line from a Phase 1 antibody project. Cell line 16-6F was continuously cultured for more than 211 days, and genomic DNA was prepared at various time points (45, 82, 117, 134, and 211 generations). The DNA was digested with AatII, 2 μg were run on an agarose gel, and the gel was transferred to nitrocellulose, which was hybridized with probes encoding specific HC and LC genes. The lane labeled "CHO K1 + vector" contains 50 pg AatII digested specific plasmid DNA spiked into CHO K1 genomic DNA. The migration of both nonmethylated and methylated DNA is indicated.

The drawbacks to performing this type of analysis are that it is dependent on the presence of specific enzyme restriction sites in the plasmid region of interest, it is fairly low throughput, it is labor intensive, and it requires the use of radioactivity to generate results of consistently high sensitivity and high quality.

Methylated DNA Immunoprecipitation (MeDIP)

As an alternative, we developed and now use a methylated DNA immunoprecipitation (MeDIP) procedure.4 This procedure is illustrated in Figure 8: DNA is denatured and digested, and is incubated with an antibody that specifically recognizes 5-methylcytidine. The resultant immune complexes are incubated with protein A/G beads, and following washing and centrifugation steps, the captured material is eluted and can be subjected to endpoint PCR, or, if quantitation is desired, to qPCR. The advantages of this procedure compared to a methylation-sensitive Southern blot are that it is high throughput, requires no radioactivity, allows for quantitative results using qPCR, and additionally, through the use of specific PCR primers, can enable a determination of which region(s) of DNA are methylated.

Figure 8. Schematic of the methylated DNA immunoprecipitation (MeDIP) procedure. For a more detailed explanation, refer to the text.

Figure 9 shows the results of a MeDIP experiment on the same 16-6F genomic DNA from the Southern blot experiment shown in Figure 7. Endpoint PCR made use of primers directed to the promoter region of our transfected vector construct, and as expected, the results showed no detectable PCR product from untransfected CHO cells. Also as expected, the lane labeled "+control" showed a very strong signal for in vitro methylated DNA plasmid. The lane labeled "16-6F," which contained the MeDIP PCR product from the genomic DNA sample from the phenotypically unstable 16-6F cell line prepared at 211 generations (recall from Figure 7 that this DNA was not fully digested with the methylation sensitive enzyme AatII), showed a strong signal as well. Although the size of DNA fragments generated during the digestion step of the MeDIP procedure needs to be considered, these latter results suggested that at least a portion of the methylation is in the vicinity of the promoter region of our integrated transfected vector construct. The 16-6F cell line was later re-cloned, and genomic DNA from a resultant high-expressing subclone, 16-6F A5, whose protein expression profile was stable, was subjected to the same MeDIP and PCR assay conditions. Figure 9 shows that in the lane labeled "16-6F A5," there is a much reduced methylation signal. Taken together, these data indicated that recloning resulted in the isolation of a clone having reduced levels of expression vector methylation, and that the markedly reduced methylation in this clone correlated with its increased protein expression and increased stability.

Figure 9. MeDIP assay results of 16-6F and subclone 16-6F A5. Genomic DNA from the 211 generation time point of the unstable cell line 16-6F described by Figure 7 and associated text was subjected to the MeDIP procedure illustrated in Figure 8. Forward and reverse PCR primers were specific for 5' and 3' sequences just upstream and downstream of the promoter region of the transfected vector construct. The lane labeled "+ control" is in vitro methylated plasmid also subjected to the MeDIP procedure. 16-6F was later re-cloned, and a resultant stable subclone, A5, was also subjected to the MeDIP procedure. The arrow marks the migration of the methylated PCR product.

The MeDIP assay allows us to increase our understanding of the methylation status of our integrated expression constructs. By applying it broadly to our Phase 1 antibody projects, the MeDIP assay will enable us to monitor the extent to which epigenetic changes of this nature occur in our cell lines and negatively affect the outcome of our cell line development efforts.


In conclusion, we have developed, and will continue to develop, tools to increase our understanding of, and continuously improve, our expression systems, our cell line development practices, and our outcomes. This manuscript has described how certain of these tools have enabled us to better understand, and specifically cope with, cell line instability. Using these tools has enabled the earlier identification of unstable clones, which is critical given our aggressive development timelines. In addition, given the industry-wide movement toward using more efficient practices during the development of protein therapeutics, our ability to predict cell line instability, and to invest time and resources on only those clones that are viable candidates, is beneficial. Finally, the information and learnings that result from using these tools and assays will continue to be key in eliminating instability altogether, thus enabling improved consistency and predictability of cell lines in our future projects.

ROBIN HELLER-HARRISON is the associate director and the corresponding author, KERSTIN CROWE is a research scientist II, CECILIA COOLEY is a research scientist I, MEGAN HONE is scientist II, KEVIN MCCARTHY is a principal scientist I, and MARK LEONARD is the director, all in the cell and molecular sciences group in the department of Drug Substance Development at Wyeth BioPharma, Andover, MA, 978.247.1406,


1. Herman J, Baylin S. Gene silencing in cancer in association with promoter methylation. N Engl J Med. 2003;349:2042–2054.

2. Ehrlich M. DNA Methylation and Cancer-Associated Genetic Instability. In: Nigg EA, editor. Genome Instability in Cancer Development. The Netherlands: Springer; 2005. p. 363–92.

3. Doerfler W. In pursuit of the first recognized epigenetic signal-DNA methylation: A 1976 to 2008 synopsis. Epigenetics. 2008;3:125–33.

4. Weber M, Davies J, Wittig D, Oakeley E, Haase M, Lam W, Schubeler D. Chromosome-wide and promoter-specific analyses identify sites of differential DNA methylation in normal and transformed human cells. Nat Genet. 2005;37:853–62.