Final Report Summary - PM-CELL (Development of a rapid cellular characterisation technology <br/>for use in the biopharmaceutical Industry)
Executive Summary:
Identifying cell lines that stably produce high protein titres is a critical part of biopharmaceutical development – this is traditionally a time-consuming, labour intensive process because their productivity and stability can vary enormously and large numbers of cell lines must be screened. Cell-line development is unsurprisingly seen as a major bottleneck in biopharmaceutical development. As cultivated mammalian cells have become the dominant system for the production of biopharmaceuticals, the characterisation and selection of highly productive cell lines is a critical step for efficient industrial manufacturing processes.
The PMCELL project was conceived to focus on the development of rapid, cost saving, regulatory focused, screening assays for the Biopharmaceutical industry with an initial focus on the development of an innovative consumable product for the rapid determination of CHO Clone Bioprocess performance and we believe we have made significant progress in relation to this. We have also encountered a novel application of our findings towards CHO Clone stability, tracking and profiling.
Prediction of fed-batch growth – using multiple linear modeling, a predictive mode was built using the chemical specific growth response as an input and to output the predicted fed batch IVCD50. This model was validated via 3-fold internal cross validation and was found to predict whether clones were good or bad groers with 85% accuracy.
Prediction of fed-batch titre – using multiple linear modeling, a predictive mode was built using the chemical specific growth response as an input and to output the predicted fed batch maximum titre. This model was validated with 12-fold cross validation and was found to predict maximum fed batch titre with an R2 value of 0.8.
Growth fingerprint for cell type ID/stability – it was observed that cell growth response fingerprints were highly characteristic for particular cell lines. It was demonstrated that initially clones could be identified by their fingerprints with 100% accuracy, 2 passages later with 100% accuracy and 4 passages later with 89% accuracy.
The PMCELL project has been successful in the generation of a technology capable of enabling the rapid and facile generation of a chemical specific growth and productivity fingerprint. Using this fingerprint we have been able to produce accurate predictions of fed batch cell growth and maximum titre. This illustrates the potential use of our product as an alternative to traditional, expensive, small-scale bioreactors. These plates have a minimal footprint thus facilitating the screening of much larger numbers of clones than traditional SS bioreactors. Additionally, as added value, we have demonstrated the use of our product to identify and verify the stability of cell lines – this is a comparatively rapid and in expensive alternative to the existing method employed i.e. genetic identification.
Project Context and Objectives:
Identifying cell lines that stably produce high protein titres is a critical part of biopharmaceutical development – this is traditionally a time-consuming, labour intensive process because their productivity and stability can vary enormously and large numbers of cell lines must be screened. Cell-line development is unsurprisingly seen as a major bottleneck in biopharmaceutical development. As cultivated mammalian cells have become the dominant system for the production of biopharmaceuticals, the characterisation and selection of highly productive cell lines is a critical step for efficient industrial manufacturing processes.
The primary scientific rationale underpinning the project was that in order to achieve high growth and productivity a clone must be able to proliferate / produce product while under intrinsic and extrinsic stress for extended culture periods in a bioreactor - can stress response fingerprinting predict clone bioprocess performance? The experimental concept is to create a high-throughout platform [Microtitre plate] which individually and collectively simulates the stressor environment [through a plurality of inducers of intrinsic and extrinsic stress in individual wells of the Microtitre plate] of a bioreactor and can thus be used to rapidly generate a ‘Stress Response fingerprint’ for any cell line or clone with the hypothesis that elements of that ‘Stress Response fingerprint’ will be capable of predicting, individually & collectively, elements of the Bioprocess performance/Identity of any clone.
Since 1982 the biopharmaceutical industry has been concentrated among relatively large Western companies. Today, 88% of global biomanufacturing capacity (as measured by total liters of installed cell culture capacity) is located in North America, Western Europe, Ireland, and the United Kingdom. High-cost, high-margin biologics are made in those locations primarily in massive, hard-piped stainless steel and glass production systems that have defined the state of the industry for decades. In recent years, the first generation of biomanufacturing has approached maturity, with new products flowing through established pipelines and revenue growth leading to overall industry profitability following decades of large capital investments.
New technologies, especially single-use production systems, are dramatically lowering building costs (and in many cases also operating costs) of biomanufacturing plants. Modular systems make it easier to produce multiple products in a single site, adding flexibility by significantly reducing lead times needed to modify a process for each new product.
Process improvements have increased productivity significantly more than industrial experts had predicted just 10 years ago, making the 2,000-L scale the workhorse for most biologics, including antibodies. The desire of many governments to have local vaccine and biologics manufacturing capacity to serve their populations is driving demand. Such evolutionary pressures will expand biomanufacturing from its current concentration in the West to countries across Asia, Eastern Europe, South America, and Africa — specifically, China, Russia, Korea, Brazil, and Malaysia in the near term. Bioprocessing will migrate from large-scale single-product installations to smaller and more flexible multiproduct systems. Looking ahead five years and considering projects that are now in various stages of planning or development, Asia and Eastern Europe will likely see the fastest growth in new biomanufacturing capacity. There is also significant interest in South America for manufacturing a range of biotherapeutics and in Africa for local vaccine production. Scanning the landscape of the evolving global biomanufacturing industry, seven general categories of enterprises that are either considering or already moving to build biomanufacturing capacity in new geographies exist:
1. Large, incumbent biopharmaceutical and vaccine companies;
2. Established midsized biotechnology companies;
3. New contract manufacturers;
4. New vaccine manufacturers;
5. Government-backed initiatives focused on biodefense and local populations;
6. Emerging biotechnology companies working on new drugs and biosimilar development;
7. Small-molecule companies seeking better margins by transitioning to biologics.
Established mid-sized biotechnology companies that have experience with both traditional and flexible single-use systems are well-positioned to move into many emerging markets, and many are planning to do so. With biosimilar-licensure regulations in place in Europe and on a path to being established in the United States, the number of new manufacturers seeking to capitalize on those pathways will grow.
In countries with nationalized healthcare, governments are the principal therapeutics buyers. Many such governments are getting involved with international partners to establish or incentivize construction of biomanufacturing capacity within national borders. MAbs now dominate the list of top-selling drugs worldwide. Industry reports indicate that nearly half of all drugs now in development are biologics. There is resurgence in the vaccine sector for biodefense and for life-saving medicines to treat local populations. The advent of biosimilar licensure for drugs coming off patent protection adds a new and powerful segment. Success in this new, globally distributed biomanufacturing industry will be a function of how well companies can leverage new technologies and manage the challenges of “in market, for market” manufacturing.
The principle objective of the PM-CELL project was to develop an innovative consumable product for rapid determination of CHO Clone Bioprocess performance [Cell Characterisation Technology] which will result in time and cost savings for the biopharmaceutical industry [including CMO] customers of the participating SMEs. As the project evolved and data began to be generated it became clear to the consortium that there were a number of additional potential applications of the ‘Stress response fingerprint’ being generated by the aforementioned consumable product which can be grouped into two distinct groups:
Application No. 1 - An innovative and rapid cell line selection technology to identify the most industrially relevant cell line applicable to a particular product through a relative ranking of Bioprocess performance prediction.
The annual global biopharmaceutical market is currently valued at $50 billion (€37.4 billion), projected to reach $70 billion (€52.4 billion) by the end of the decade. The demand for cell-derived biopharmaceutical products is increasing. This demand is caused not only by the fact that there are an estimated 800 biopharmaceuticals currently in the therapeutic pipeline but also because, patent expiry on many big-selling popular branded products is giving way to a biosimilar market. It is estimated that for every 2-months a product is not on the market, the average revenue loss is $2 million.
Application No. 2 - An innovative and rapid cell line identification and quality control technology which can be used to specifically identify and track the stability of any cell line at any stage of the development/production process.
Cell line authentication is now required by certain journals prior to publication and in some cases is mandatory before receiving funding from small granting agencies. The FDA has also instituted a requirement for the authentication of cell lines used to produce pharmaceuticals in their General Requirements for Laboratory Controls and the General Standards for Biological Products (21 CFR 211.160 (b) and 21 CFR 610.18 (b)). There are methods in place for authenticating human cell lines using multiplex PCR assays that target short tandem repeat (STR) markers in the human genome Although there are successful methods in place for human cell line authentication, methods for nonhuman cell lines are not well established.
15-30% of all cell lines are cross-contaminated with other cell lines or are misidentified. This means that billions of dollars have been wasted over the past 45 years on producing misleading or false data. There have been retractions of journal articles due to this problem – one example is the 2005 discovery that stem cells could seed cancer (Garcia-Castro, J. et al, Can Res, 2005. In 2011, Bayer conducted a study that led to the cessation of two-thirds of its drug target validation projects because their in-house experimental data failed to reproduce the published academic literature claims in 65% of projects examined.
The PMCELL project was conceived to focus on the development of rapid, cost saving, regulatory focused, screening assays for the Biopharmaceutical industry.
Workpackages 1 & 2 are primarily concerned with project management and Cell model establishment. The primary product concept and development workpackages are outlined as follows:
WP 3 – Characterise Microenvironments
WP3 is focused on characterizing the microenvironments which will provide a platform for future R&D activities of PM-CELL. WP3 has 3 No. objectives and 5 No. Deliverables.
Objectives
1. Identify and simulate the environmental conditions that influence cell culture in industrial bioprocess.
2. Comparatively quantify detection methods for the evaluation of cellular metabolic phenotypes in microplate format.
3. Create database to manage and support software application development.
4. Profile variant clones in culture environments.
WP 4 – Develop Microplate
WP4 is focused on market analysis, microplate design and software interface development and has 4 No. Objectives and 6 No. Deliverables.
Objectives
1. Perform a market survey to position the product in the biopharmaceutical industry.
2. Design the microplate for maximum efficacy.
3. Identify key performance indicators (microenvironments) that most effectively assess cell line performance i.e. cell performance index.
4. Create and develop the software interface.
WP 5 – Validate in Working Models
WP5 is focused on validating the core scientific principles in working models and has 3 No. Objectives and 5 No. Deliverables.
Objectives
1. Manufacture the prototype plate.
2. Validate the microplate in industrial producing cell models.
3. To use the designed microarray plate to differentiate industrial cell lines with known differences in growth and production.
WP 6 – Prototype testing
WP6 is focused on consortium training and obtaining as much user feedback on the prototype as possible and has 3 No. Objectives and 2 No. Deliverables.
Objectives
1. Provide scientific training and transfer knowledge on the product to technical SME staff.
2. To demonstrate the prototype product to industry users and enable on-site testing in industry environments.
3. Obtain feedback on product performance in end-user sites.
WP 7 – Dissemination
WP7 is focused on Dissemination and has 6 No. Objectives and 8 No. Deliverables.
Objectives
1. To ensure expedient exchange of results between the RTD Performers and SMEs in the consortium.
2. To prepare identify, capture and protect Intellectual Property for the SMEs in the consortium.
3. To prepare exploitation plans for the new technology.
4. To prepare market penetration strategies based on the Market Survey.
5. To disseminate the potential for the microplate based cell analysis platform to the end-users.
6. To plan for scale up from prototype to manufacture and commercialisation of the product.
Project Results:
The author has considered the PMCELL project into 2 distinct phases i.e. Project establishment, Cell model establishment & project management and Product development, Proof of concept testing, Prototype development, Validation and Dissemination. The first phase conveniently encompasses Period 1 and Phase 2 encompasses Period 2.
As Period 2 is concerned with the main S & T results, the author has chosen to go into a lot more detail on Phase 2/Period 2 than Phase 1/Period 1.
It is however important to note that PMCELL made progress during Phase1/Period 1 and ultimately produced an industrially relevant cell model which could be used for the remainder of the project – this was the crucial practical outcome of this phase as without it the project could not have progressed. The consortium was not able to develop the Cell model as submitted in the original application and thus we obtained an industrially relevant cell line from Cobra Biologics [Keele, UK]. Progress was also made on the assay format; a panel of analytes were selected and underwent development and optimisation according to a specific set of criteria devised to ensure the bioprocess, manufacturing and commercial focus of the final PMCELL product. A number of potential detection methods were identified and undergone preliminary evaluation. Whilst there were some delays and challenges, the PMCELL ‘Cell model’ was established for Phase 2/Period 2.
Microenvironment selection
In order to determine if the PM-CELL micro-environments selected for evaluation accurately predict CHO cell performance in bioreactor conditions it is necessary to generate two distinct clone sets - a training set and the validation set. The training set will assess if each selected micro-environment is able to differentiate between clones with varying cellular performances in simulated bioprocess conditions (fed-batch culture). Thus if a specific microenvironment is observed to exert the same effect upon each clone tested regardless of bioprocess performance it would therefore not be deemed suitable for further use. Micro-environments that appear to exert differential outputs will be incorporated for further analysis. Data from the training set clones subjected to all potential microenvironments will be sent to EuFor and AHVLA for statistical analysis to select the most effective micro-environments with the highest statistical power. Once selected, the validation set will be used to determine if these micro-environments can actually predict clonal cellular bioreactor performance.
Analysis was performed on the data collected from the fed-batch - obvious anomalous data points were removed from each data series. These artefacts arose from clumping issues that occurred during fed-batch studies resulting in erroneous readings generated by the Vi-cell. This is a common issue observed with CHO-S cells especially at high viable cell concentrations which renders the vi-cell unable to determine the existence of single cells. Any anomalous results were replaced by points extrapolated from the readings before and after the anomalous result. Various performance characteristics were considered to differentiate ‘good from bad’ clones in terms of growth and performance in fed batch culture. These include maximal viable cell density, culture duration to 90% and 50% viability and integral viable cell density (IVCD) to both 90% and 50% viability. These particular viability percentages were chosen as 90% highlights the start of the decline phase of culture. USFD was advised by industry experts that 50% is the standard point of harvest from bioreactor runs for the majority of easy to fold recombinant proteins. Bioreactor productivity is a function of both IVCD and cell specific productivity. Therefore IVCD was identified as a key culture parameter.
One-way ANOVAs were performed on all data described. Prior to this, post-hoc tests were performed to assess significance of individual differences. For this, Tukeys multiple comparisons test was performed (which takes into account increasing type I error potential with multiple comparisons). From this, “honestly significant difference” confidence limits (alpha at 0.05) were determined for ease of visualisation of significant differences. Review of all data revealed that the most promising parameter to differentiate the clones was IVCD at 50% culture viability. Statistical analysis of this parameter revealed 14 (5 low, 5 high, 2 very high IVCD 50%) significantly different clones from the 35 Cobra clones generated. This number is insufficient for a complete training set to be able to accurately test the predictive power of the microenvironments. USFD has access to 29 alternative CHO clones and 11 CHO-S non-producing clones, all of which possess a high probability of clonality. Fed-batch studies are currently in progress to enable statistical analysis.
We previously demonstrated that clones within the training set would be cultured in mini-parallel bioreactors to monitor performance in simulated bioprocess culture processes. Although fed-batch shake flask culture is a good representation of how cells behave within a bioreactor process there are certain differences between the processes, which can alter cellular behaviour. For example cells in fed-batch shake culture are maintained in suspension by rotation of the incubator plate that the culture flasks attach to. However, in bioreactors suspension is achieved by stirring the cells with an impeller or air-lift reactors. These processes can cause an additional stress to the cells. Also gassing mechanisms are different between both processes. In bioreactors the cells are sparged whilst fed-batch cultures are maintained in a CO2 environment. Sparging has been well characterised to exert detrimental effects upon cell growth and recombinant protein production.
USFD negotiated access to trial the Hexascreen automated multiple mini-bioreactor system (Hexascreen, Spain; www.hexascreen.com). This system allows for 6 mini-bioreactors to be performed in parallel (10-15 mL total volume). Culture information such as temperature, dissolved oxygen, optical density and media pH can be monitored and recorded. Discussions with Hexacsreen confirmed that although automatic feeds / sampling could not be set up it was possible to manually add feeds to the individual bioreactors and remove samples removed for subsequent analysis. The Hexascreen system was installed into USFD on 3rd July for testing to determine its suitability for this task. During the four weeks trial period USFD experienced a number of problems including persistent erroneous readings from one of the optical density sensors and issues with pausing the system to collect samples and add feeds (Fig 1).
Alternative small-scale bioreactor models exist however USFD was unable to negotiate rental of these systems. Following the failure of the Hexascreen system for the purposes intended for this project USFD entered discussions with Cobra and Life technologies regarding the expected similarity between fed-batch clone data and scale down model systems, such as the Hexascreen system. Life technologies indicated that ranking of clones based on their performance in shake flask fed-batch culture should be mirrored in scale down models when the same processes are imposed (e.g feeding regimes). Furthermore Cobra Biologics, who generated the parental cell model employed in this project, stated that they have not observed any significant differences in clonal performance between shake flask fed-batch and bioreactor systems.
USFD and ITT generated a detailed database listing potential culture microenvironments (i.e. chemicals/biochemical), replicating a wide variety of stresses encountered during a typical bioreactor run, for inclusion on the final PM-CELL plate (Table 1). However, this range of chemicals was substantially reduced due to limited solubility in water/media. Following unsuccessful preliminary efforts to solubilise the compounds, it was deemed that a major criterion for compound selection was solubility. Additional criteria were also considered whilst selecting which compounds detailed in the database to test. Overall, the aforementioned criteria can be considered as primary and secondary considerations depending on importance.
Primary considerations are as follows:
(1) Solubility: Originally thought that all compounds must either dissolve in water or media. This attribute would greatly facilitate the manufacture, distribution and storage of the plate as no ‘solubilising agents’ such as organics would be required;
2) Manufacturing compatibility: Manufacturing compatibility is an issue of high significance, as compounds may potentially have toxic effects in the PM-CELL plate manufacturing process and/or for the end user. Compounds with a higher category toxicity rating of grade III were eliminated;
Secondary conditions are as follows:
(1) Stability: Compounds ideally should be stable in light to increase the shelf life of the product. For example, Actinomycin D was immediately ruled out as in powder format, it is sensitive to light and must be sealed and protected from light and moisture at 2-8OC. Dilute solutions are also very sensitive to light and must be discarded and not stored long term;
(2) Cost: This is a major consideration in relation to the economics of final product manufacture and retail price. Estimated price per well was calculated from reported effective concentrations in the literature and price of compounds from the Sigma Aldrich catalogue. Compounds with an estimated cost greater than 10p per well were not usually considered unless alternative compounds simulating the same effect could not be identified.
A screen of compounds to test were ordered and tested on the Cobra parental cell line. Initial investigations were performed in small-scale shake flask culture, cultiflasks (Satrorius Stadium) total volume 10ml, whilst the microplate detection method was being optimised. Cultiflasks were seeded at standard concentrations (0.2x106 cells mL-1) with a range of concentrations per compound. Appropriate concentration ranges were selected following a detailed search of the literature. Cell counts were collected daily for 96 hrs (Figure 2). From the data collected an IC50 concentration (concentration required to inhibit 50% of response, growth in this case) was estimated for each micro-environment (Table 1). From this 26 compatible microenvironments have been identified simulating a wide range of bioreactor stresses which permitted the elucidation of an IC50 value.
Following compound selection and screening in ‘cultiflask’ format, for plate design, the compounds were to be tested in 96-well plate format with the finalised detection method. This would enable determination of compound specific microwell compatibility and microwell specific IC50 in addition to evaluation of detection method compatibility.
Detection method selection
Clone Select Imager (CSI)
To evaluate the CSI as a potential detection method it was initially optimised to detect CHO cell growth over 96 hrs. During optimisation of the seeding density it became apparent that there was a large variation in estimated confluency between wells, in some cases a 2-fold difference. This variation is due to a phenonemon deemed the ‘edge effect’ (figure 4), whereby the cells appear to settle around the edge of a well. We speculate that the variation in confluency between an ‘edge effect’ well and one where the cells are evenly distributed across a well is due to the fact that once cells are settled at the edge of a well there is limited room for growth. Therefore the CSI is unable to differentiate between cells in mutliple layers and underestimates the well confluency. To investigate the cause of this effect and potentially reduce well-well varation within a plate a number of experiments were performed, testing various hypotheses.
(1) Cell line specific effect
We hypothesized from this work (figure 5) that the ‘edge effect’ arises from the cells proprensity to clump together, an artefact exhibited with CHO-S cells.
(2) Anti-clumping agents
We tested 2 anti-clumping agents at various dilutions to investigate their effects upon the edge effect. Suspension flasks (total volume 30ml) were seeded to monitor the agents’ effects upon Cobra cell line growth (figure 6) and plates seeded for analysis on the CSI. All dilutions tested for the anti-clumping agent from Invitrogen and PF-68 (0.1%) increased final viable cell density compared to control flasks. PF-68 at 1% appeared to reduce cell proliferation rate and was therefore not tested in the plate format. Despite the positive effects upon viable cell density no differences in the edge effect were observed with no change in well-well variation.
(3) Trypsin
We investigated the possibility of using trypsin to detach the cells, back into suspension mode, to reduce the well-well variation. A wide variation was still apparent the post trypsin incubation which was reduced following the 30 min settling period.
Nephlometer
NephloSTAR (BMG Labtech, UK) is a laser-based microplate nephlometer that measures suspsended particles in liquid by light scattering. This system fulfils the requirements to be considered as a potential detection method as it is very easy to implement following successful generation of a standard curve to extrapolate unknown concentrations. It is a relatively common system in many cell line development companies especially those working in the high-throughput microplate format. A standard curve was generated using the Cobra cell line with limited variability between replicates (figure 7a). However these measurements were taken immediately post seeding prior to settling and occurrence of the edge effect. However, once the cells start to adhere huge variations between replicates develop (figure 7b).
Presto blue
Presto blue (Life Technologies) is a cell permeable resazurin-based solution that functions as a cell viability indicator by utilizing the reducing power of living cells to quantitatively measure the cellular proliferation. Resazurin is reduced by cellular NADH to give fluorescent resorufin which can be detected using a standard fluorescent plate reader (Figure 6). Compared to alternative viability assays such as MTT, Presto blue is very easy to use with a single addition of the solution and no cell lysis required. USFD evaluated Presto blue as a potential detection method as it fulfills all necessary requirements. Firstly, the assay is easy to implement. Secondly, the fluorescence change is detected by a standard fluorescent plate reader, a common piece of equipment in all biopharmaceutical companies. Finally the output, metabolic activity, is directly linked to bioreactor performance. Following optimization of the protocol it was found to give accurate determination of cellular density with very low variability (figure 8).
Presto blue has been implemented to initiate microenvironment IC50s determination in microplates using optimized seeding density (0.2x106 cells mL-1). Plates were seeded and Presto blue measurements performed every 24 h. Various concentrations were tested, selected using the cultiflask suspension data (Deliverable 3.1). Comparable replicate variation was observed (figure 9).
A total of 5 detection methods have been evaluated by USFD and ITT. Two potential detection methods have been identified which fulfil all necessary requirements (USFD). Flow Cytometer has been proven unsuitable for use with the PM-CELL end product. Fluorometric measurement with Prestoblue® is the detection method of choice as it is easy to implement, detectable by a standard plate reader and metabolic activity as the output is linked to bioreactor performance.
Microenvironment characterisation
Following this decision, all the above mentioned compounds were screened in microwell format using the ‘PrestoBlue’ (PB) assay. Ranges of compound concentrations were screened based on those used in the 10ml format. Due to the kinetics of the cell growth in the 96-well format, i.e. there is no substantial cell growth after day 3, it was decided that 72 hours would be the optimal time of incubation of cells in the microenvironments since this maximises control cell growth (this signal to noise ratio in model) while minimising the time period.
An envisaged issue with using a chemical based detection method is that the microenvironments may directly interfere with the assay chemical which may compromise the accuracy of cell growth determination in certain microenvironments. To investigate this, a plate containing cells and microenvironments at relevant concentrations were assayed at 0 hours post seeding. This would allow a measure of direct influence of compound on the absorbance value from the PB compound. From this it is evident that none of the microenvironments tested directly affect the PB assay to any substantial degree.
IC50s for 72 hours were determined for each microenvironment in 96-well format utilising the PB assay. From this data it is evident that a clear dose response curve can be obtained for each microenvironment in 96-well format.
Following this work, it is now necessary to construct a prototype clone screening plate to test for repeatable IC50s measured in high throughput format, plate-plate variation and within plate variation.
A schematic overview [figure 10] of creation and utilisation of microplates (1) 45 ul of 2x IC50 chemical (in appropriate media) was added to the plates en masse, (2) 45ul of clonal cells were added to the plate prior to incubation at 37C for 3 days, (3) 20ul of ‘Presto-Blue’ (diluted 1:1 in CD-CHO) was added to the plates. Plates were mixed by shaking and incubated at 37°C for 30 minutes prior to reading absorbance.
From each clone, a spread-sheet of growth in chemical inhibitors was generated. These data were the average of 9 technical replicates for each well (9 readings were taken to cover the whole well area). Subsequently the mean of the triplicates was taken to give a specific growth for each chemical microenvironment.
To give a ‘clone specific response’ to a chemical microenvironment, growth in a ‘chemical well’ was normalised to the control well growth (i.e. control well growth was taken as 100% and growth in chemicals was relative to this). Subsequently, for each clonal group (either PM cobra or P2) to the chemicals were normalised, i.e. the mean of the cell specific response for a chemical for the cell line in question was subtracted and divided by the standard deviation of the response to the chemical.
In addition, cell lines were given a rank according to their performance in the fed batch study. This ‘ranking’ enabled comparisons between the cell lines (PM and P2). Subsequently the data files from the 2 cell lines were merged. This merging allowed statistical analysis to be performed on the entire data set.
Statistical analysis was performed in R. The initial approach of the analysis was to model the ‘standardised cell specific microenvironment response’ (CSCR) (explanatory variable) to the standardised rank IVCD50 (SRI) (response variable).
Following analysis of individual correlations between ‘CSCR’ and ‘SRI’ it was evident that (a) there were few ‘significant’ individual correlations and (b) from a view of scatter plots of CSCR and SRI, it was clear that there were no ‘non-linear’ relationships present.
From this, it was decided to build a multiple linear model between the various CSCRs and the SRI. The overall approach to building this model is illustrated in Figure 11.
From this modelling approach, a model was elucidated which is in the Table below. From initial analysis the model built is highly significant. i.e. there is a 0.00000008249% chance that this model could occur by chance. The model R2 was around 0.9 indicating that approximately 90% of the variance in the data can be explained by the model. The adjusted R2 was also considered very good (this is a version of R2 which is penalised for the number of regression terms used).
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 1.24442 0.01765 70.509 < 2e-16 ***
BCH -0.14454 0.03969 -3.642 0.00298 **
Sodium Butyrate -0.14459 0.03858 -3.748 0.00244 **
Cadmium Acetate 0.09827 0.03190 3.080 0.00877 **
Citric Acid 0.14028 0.03348 4.190 0.00106 **
Sodium Lactate -0.11869 0.03096 -3.834 0.00207 **
3 Bromopyruvate 0.13289 0.03244 4.097 0.00126 **
---Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.07893 on 13 degrees of freedom
Multiple R-squared: 0.9003,Adjusted R-squared: 0.8543 F-statistic: 19.57 on 6 and 13 DF, p-value: 8.249e-06
Microplate development
Optimal concentrations of chemical inhibitors were determined. Proof of concept was produced and consequently the status of the project is “GO”. Following discussions with the end plate manufacturer, Microcoat, the final plate format was agreed [figure 12] and designs for pre-beta testing plates were finalised.
End-user protocol
To define the end user protocol, two main process variables were investigated. These were the time post-subculture [see Figure 13, 14 & 15] for plate assay and also, passage of the cells. Information from these will allow assessment of the impact of these process parameters to accordingly design the end user protocol. Other aspects of the user protocol such as the set up of the plates and the assay of cell growth in the plates has been established. The following end user protocol is proposed:
(1) Use clones at the same passage where possible;
(2) Set up plates using cells at day 3 post subculture (or a fixed time post subculture where day 3 is not appropriate e.g. when this is not ‘mid exponential’;
(3) Dilute cells to a density of 0.2x106 cells/ml in growth media;
(4) Add 90ul of cells to the chemical plate;
(5) Incubate plate for 3 days in a static, humidified incubator with conditions optimised for CHO cell growth (such as 37C, 5% CO2, humidified);
(6) After 3 days, dilute presto blue 1:1 in growth media, allow to equilibrate to room temperature. Ensure this is thoroughly mixed. Add 20ul of the presto mix to each well;
(7) Shake for 10 seconds (orbital, 1200rmp, 1mm diameter);
(8) Incubate to 30 minutes (although this may be optimised by the end user and may depend on the rate of growth of the cells in question);
(9) Shake as above, and read fluorescence (Abs 535, Emi 620) using the supplied BMG plate reader.
Cellular-performance Index [CPI]
The CPI focussed on utilising the biological data to accurately model the optimal predictive capacity of the various environments thereby establishing a model which can be further evolved through beta testing. Two panels of industrially relevant clones were created and their fed batch performance was assessed. Here, our metric of “fed batch performance” was integral of viable cell density using a cut-off were cells dropped below 50% viability (figure 16). A wide range of chemical inhibitors of cell growth and metabolism were selected. These stress the cells in a wide “multi-dimensional” way. These were chosen to simulate relevant stresses and to generally test the resilience of the cells in a in a number of different ‘metabolic dimensions’ e.g. resistance to oxidative stress/cell cycle inhibition/inhibition of protein production/ NAD production etc. For a selected panel of clones, each was subjected to the chemicals in a 96 well plate format. This enabled a “cell specific” chemical growth profile to be generated. This can be viewed as a “metabolic fingerprint”. Using aspects of clone metabolic fingerprint, a multi-linear model was build with the power to accurately predict the relative cell performance of a clone (IVCD50 generated from fed batch studies) using information from the metabolic fingerprint. The Model was validated using an “n fold cross validation” technique. Data was randomly split into “training and “validation” sets. Models were built using training sets, and their predictive power was tested using validation sets. Essentially, this model was able to predict new data Points and classify new clones into “good” and “bad” performers with 85% accuracy (figure 17 & 18). Using our plate based “stress” plates to generate a “stress resistance profile” we were able to use aspects of this to accurately predict bioreactor performance (relative fed batch IVCD50) with 85% accuracy. Using our technology it is speculated that this will allow very high through put screening of hundreds of ‘candidate’ clones with a minimal fingerprint (all that is required is static incubator space – average incubator capacity could easily exceed 200 plates), using a process which requires little specialist knowledge or training. Moreover due to the ‘user-friendly nature of this technology is it easily incorporated into an automated system making it compatible with large scale workflows.
Software Interface
To enable correlation between the microplates phenotype fingerprinting of the antibody producing cell lines, the cell lines will be classified into high and low producers based on calculations made on the measures obtained from growth capacity in multiple chemical micro-environments. The role of the SW is to provide a user platform to
1. Import learning data from know cell lines
2. Create a classification model from the learning data
3. Import test data from unknown cell lines to be classified
4. Classify the unknown cell lines based on the learned model into high and low producers
5. Report the classification results to the user for decision making and for further integration into other laboratory management systems.
A beta-version (figure 19) of the SW has been achieved and can now be tested with users in a laboratory environment.
Plate Manufacture
Following proof of concept, the next step towards commercialisation is to make a technology transfer from wet “stress profile plates” manufactured ‘in-house’ to dry stress profile plates. We carried out an extensive evaluation of the requirements and technologies required to successfully manufacture the plates internally at Technopath and came to the conclusion that in the time frames permitted there was no possible way this could be achieved. We chose instead to partner with an expert in the field and we carried out an extensive review of the various key players operating in this space in Ireland, UK and across the EU. We eventually settled on a company called Microcoat and the TPath Technology transfer and validation team visited the Microcoat facilities in Germany. MicroCoat Biotechnologie GmbH are based at Am Neuland 3, 82347 Bernried, Germany, www.microcoat.de.
A plethora of manufacturing conditions were evaluated. Using the reconstitution conditions of “Just add cells”, these manufacturing conditions were compared. Criteria considered were sterility, inter-plate CV, and level of chemical inhibition achieved.
With respect to sterility, there was only one contamination and that was present on an 'Unradiated' plate. This is not strong evidence for a systematic problem with unradiated plates since there was only 1 well contaminated, despite many unradiated plates being examined. Nonetheless, it would be best to proceed with radiated plates as this certainly did not have an apparent detrimental effect on the performance.
With respect to all other manufacturing conditions, the plate performances were all very similar. Given this, the best manufacturing conditions for us to proceed with are those which gave a high level of chemical inhibition and a low CV. These manufacturing conditions were:
Nunc plates, 12h drying time, 30C drying temperature, 4G irradiated.
These conditions gave good levels of inhibition and a inter-plate CV value of 4.5% from a 4 plate replicate. The levels of chemical inhibition achieved with these conditions is illustrated in figure 20. The dotted lines represent 100%, 75% and 50% growth. Bars highlighted in red are chemical which exceed the 75% inhibition limit. The error bars are standard deviations from 4 separate plates.
The limit of acceptability was 75% growth inhibition (I have highlighted any chemical which goes above this limit in red). There are 5 chemicals, which exceed this limit. Overall the majority of the chemicals give adequate levels of inhibition. Also, it is clear, there is very little variation between the plates, which is very promising.
Product level assessment
An in-house panel of clones which produce IgG was sourced. These were subjected to a fed batch growth and production study. Samples were taken daily. These samples were assayed using a protein A HPLC method. This method was chosen and developed for it’s rapidity, accuracy and cost effectiveness. Bio-Monolith Protein A separates IgG in 1 to 2 minutes, high recovery (>95%). This method has a large linear range as is demonstrated by the typical standard “curve” given in Figure 21. Additionally, the assay was able to give a linear standard curve over the concentration range of 0.01 to 0.2 g/L). This is demonstrated in Figure 22. Furthermore, peak profile and retention time remained consistent which demonstrates column performance is maintained over time. This is illustrated in Figure 23. The max MAb titres from the fed batch are illustrated in Figure 24.
Having established the fed batch antibody titres and clone specific chemical responses for the PM cells, the ability to predict MAb titre using information from growth on the chemical plates was explored. A number of modelling techniques were explored including PLS, PCA and multiple linear modelling. In essence, no model could be achieved which was substantially predictive. The best models achieved was a two parameter linear model. Using the “fitbest” package in R, the best 2 parameter achieved an R2 of of 0.43. The model fit is illustrated in the Figure 25. It was concluded that it is not possible to give industrially useful predictions of MAb titre using information from the cell specific chemical growth response we re-evaluated with a number of potential customers and it was outlined that in fact it would be more interesting and much more valuable if an association could be demonstrated between the product levels in the plate and the product levels at Fed-batch. The consortium considered this for some time and came up with a design of experiments to support evaluation of same which was not performed by November 2013 but was performed subsequently. This data, along with an additional application of the technology is most succinctly described in poster presentations as figures 26, 27 & 28.
Potential Impact:
The final result of this project will be generation of the PM-CELL product line to support the rapidly growing biopharmaceutical market. Our business model is focused on the installation of our fluorescence reading platforms and to supply a suite of niche process improvement solutions [96-well microtitre plates] for various parts of the Biopharmaceutical production process – all focused on the following primary drivers:
• Speeding up the Biopharmaceutical development process;
• Enhancing traceability in Biopharmaceutical development;
• Enhancing safety in Biopharmaceutical development;
PM-CELL is already, and will continue to be, a true International partnership:
• Technopath [Ireland] – Product concept, R&D, Validation, Regulatory approval, Logistics, IP, Branding, Commercialisation, Sales, Marketing, Logistics & Customer support;
• Dorteegelund [Denmark] – R&D, Sales, Marketing & Customer support for Norway, Sweden & Denmark;
• IUL Instruments [Germany] - R&D, Sales, Marketing & Customer support for Germany, Austria & Switzerland;
• University of Sheffield [UK] – PM-CELL R&D Centre of Excellence.
• AHVLA [UK] – Modeling support centre during PM-CELL 286065.
• Euformatics [Finland] – PM-CELL Software Development and Support Centre;
• BMG Labtech [Germany] – Harware OEM Manufacturer;
• Microcoat [Germany] – Microtitre plate OEM Manufacturer.
The project has facilitated close interaction between SMEs & RTD performers from different member sates, providing significant insight on the product development and commercialisation process. Projects/Initiatives being explored by consortium members include:
• Technopath and AHVLA are discussing collaboration on an existing modeling project for the dairy industry.
• Technopath and EuFor are discussing collaboration on commercial software development projects.
Furthermore, the SMEs have been introduced to RTD collaboration which should lead to future project development activity and funding acquisition. The highest impact and value of this project will be realised upon delivery of the PM-CELL product, this will have significant impact for the SME beneficiaries, the European and Global Biopharmaceutical sector.
Each niche process improvement solution will involve a different assay kit [see Figure 29], which in effect will be a unique suite of synthetic microenvironments which collectively are capable of establishing a unique Stress response fingerprint for each individual cell line which is associated with/predictive of particular bioreactor performance variables.
We currently have beta versions of the following process improvement solutions:
• CHO Clone Bioprocess performance;
• CHO Clone tracking and profiling.
We continue to work on the proof of concept of the technology in relation to Stability testing.
In its original manifestation, the PM-CELL product was focused on decreasing cell selection time by 33%, equating to approximately 8 weeks. Considering that each week a biopharmaceutical product is not on the market results in estimated losses of €2.02 million the PM-CELL product will significantly reduce costs and time to market, thus increasing sales and profitability. This is still very valid.
In addition, novel regulatory, cell harvesting, cell line development and media development & optimization are also core. This project through its advancement in mammalian cell selection for the biopharmaceutical industry is an excellent example for the Innovation Union, one of the flagship programmes of Europe’s 2020 strategy to deliver “a smart, sustainable and inclusive economy” to sustain Europe’s competitiveness.
Socioeconomic Impact
For the four PM-CELL SMEs significant benefits will be gained with development of the PM-CELL product(s). The SMEs will benefit from economic growth and increased employment due to IP ownership and selling rights. The PM-CELL product(s) will improve SMEs competitiveness, as key suppliers to the highly regulated biopharmaceutical industry. The PM-CELL product(s) will open up new target markets providing economies of scale and improved return. Furthermore, there is significant potential to transfer the PM-CELL product(s) to other biopharmaceutical sectors including the bio-similar and cell vaccine wherein the EU are the current market leader – this may be further extrapolated to the Stem Cell market which is a key target area for the EU. The SMEs will have a competitive advantage in driving sales into new export markets of North America, Asia and China. The SMEs in this project represent the lifeline of regional and national economies as 99% of EU companies are SMEs accounting for 67% of jobs. With trans-national cooperation in the PM-CELL consortium the SMEs collaborate with SMEs and RTD performers from different states. Such collaborative efforts have the potential to result in new business opportunities, increased sales, and identification of new commercial partners to improve products or processes and access additional know-how. Trans-national cooperation will encourage rapid market penetration and generate the sales revenue that will be required to expand sales of the product into export and other markets. To the end-user, the biopharmaceutical industry, the SMEs‟ will supply a novel high-throughput cell screening product, which will result in a marked reduction in R&D time and costs and greatly reduce the time to market for new products resulting in increased sales revenue and greater profitability. Currently, the length of time to produce a high yielding cell line is 25 weeks of which 11 weeks are devoted to cell selection. The SMEs PM-CELL product can reduce the cell selection time by 8 weeks. To the EU citizen, as a net receiver of medical research and targeted therapies, the PM-CELL product has the potential of accelerating product development and also, as a direct result of the cell identification work which resulted from this consortium but which was not included in the original project scope, the industry and patient will greatly benefit from a robust, inexpensive, tracking mechanism for the bioproduction machinery.
Dissemination activities to date have included:
Bioproduction 2013, 22nd & 23rd October 2013, Dublin, Ireland – Poster Presentations [Dr. Ben Thompson & Dr. Robert Whitfield].
‘Microplate based stress resistance fingerprinting technology for rapid, high-throughput cell-line development’.
‘Microplate based stress response “barcoding” technology for rapid and facile CHO cell type identification for quality control.’ Posters won Best Communication at the meeting.
Cell Line Development & Engineering 2014, February 13th 2014, Vienna, Austria – Keynote [Prof. David James].‘The Influence of media with cell line engineering’.
Bioprocess Industry Day 2014, 6th February 2014, Sheffield, UK – Invited talk [Dr. Ben Thompson].‘Rapid prediction of CHO Cell Clone Functional Performance by Microplate-based Stress Response Profiling’.
Invitation to speak at Cell Culture Engineering XIV, May 4th-9th 2014, Quebec City, Canada.
Invitation to speak at Bioproduction 2014, 8th – 9th October 2014, Barcelona, Spain.
It is important to outline that the consortium has agreed to continue beyond the duration of the EU-funded project. TPath have entered a 10-month Research Agreement with UFSD in the first instance which will be paid for directly by TPath.
It is equally important to outline that PMCELL have captured, documented and filed 3 patents and we intend to submit a further patent by the end of March 2014:
Cell ranking/Cell performance application [Cell growth]
On June 26th 2013 we filed a new European Patent Application [1317387.8] entitled ‘A rapid, high throughput, method of predicting relative fed batch performance of a panel of clonally derived cells from a host cell population’. We received an extended search report and initial opinion on 13th November 2013 and have been discussing same electronically. The consortium has a scheduled workshop on January 20th to consider this patent in full. Additional applications claiming priority from the above application must be filed before 26th June 2014.
Cell Identification application
On September 25th 2013 we filed a new European Patent Application [13185979] entitled ‘A method of identifying a cell’. We are currently awaiting an extended search report and initial opinion.
Cell ranking/Cell performance application [Product titre]
On 30th January 2014 we filed a new European Patent Application [14153333.1] entitled ‘A method of predicting relative fed-batch production titer of a panel of clonally-derived producer cells’. We are currently awaiting an extended search report and initial opinion.
As the production of data in relation to this project is an ongoing development, the consortium have decided to adopt the existing mechanism of dealing with the publishing of results. In summary the data is reviewed at the individual RTD performer level, again at a consortium level and structurally, on a monthly basis by the Exploitation Manager & Commercial Director at Technopath, along with Mr. Barry Purdy of Purdy Lucey IP. This provides a very robust identification and analysis of IP Generation and if following analysis IP has been deemed to be identified then it is immediately captured and protected. Periodically we also add to our core Freedom to Operate database.
Any plans to publish information or results by any member of the consortium must be approved by the Project Management Committee and all approvals are associated with a formal sign-off by the exploitation manager, Commercial Director of TPath and Barry Purdy prior to submission for consideration for publication.
As mentioned earlier this system has worked very well and we have already approved a number of presentations, both Oral and Poster. We have also approved the preparation of a full manuscript which will be reviewed prior to submission.
As the IP ownership is clearly defined in all the PMCELL agreements, and there are no separate ‘fields of use’ limitations associated with individual consortium partners, the consortium deemed it not necessary to prepare a structure Table of ownership by project result.
The parties to the consortium have made no changes to the treatment of background intellectual property from that outlined and clearly defined in the consortium agreement, including both the background included and the background excluded.
The parties continuing to work on the project i.e. UFSD, DortE, IUL, TPath & Eufor are adopting identical treatment of the background IP conditions, specifics of which are included in the various contractual arrangements.
Our current target for formal launch of the product(s) is the Bioprocess International Conference & Exhibition at the Hynes Convention Centre in Boston, USA on October 20-23, 2014.
List of Websites:
The project website [www.pm-cell.eu] was developed as planned. As we will shortly enter the formal naming/branding of the product, it is likely that the product will not be named PMCELL and thus we will transfer a certain amount of the information from the current website to the product website. We intend to perpetually acknowledge the source of the seed funding for this product.
Identifying cell lines that stably produce high protein titres is a critical part of biopharmaceutical development – this is traditionally a time-consuming, labour intensive process because their productivity and stability can vary enormously and large numbers of cell lines must be screened. Cell-line development is unsurprisingly seen as a major bottleneck in biopharmaceutical development. As cultivated mammalian cells have become the dominant system for the production of biopharmaceuticals, the characterisation and selection of highly productive cell lines is a critical step for efficient industrial manufacturing processes.
The PMCELL project was conceived to focus on the development of rapid, cost saving, regulatory focused, screening assays for the Biopharmaceutical industry with an initial focus on the development of an innovative consumable product for the rapid determination of CHO Clone Bioprocess performance and we believe we have made significant progress in relation to this. We have also encountered a novel application of our findings towards CHO Clone stability, tracking and profiling.
Prediction of fed-batch growth – using multiple linear modeling, a predictive mode was built using the chemical specific growth response as an input and to output the predicted fed batch IVCD50. This model was validated via 3-fold internal cross validation and was found to predict whether clones were good or bad groers with 85% accuracy.
Prediction of fed-batch titre – using multiple linear modeling, a predictive mode was built using the chemical specific growth response as an input and to output the predicted fed batch maximum titre. This model was validated with 12-fold cross validation and was found to predict maximum fed batch titre with an R2 value of 0.8.
Growth fingerprint for cell type ID/stability – it was observed that cell growth response fingerprints were highly characteristic for particular cell lines. It was demonstrated that initially clones could be identified by their fingerprints with 100% accuracy, 2 passages later with 100% accuracy and 4 passages later with 89% accuracy.
The PMCELL project has been successful in the generation of a technology capable of enabling the rapid and facile generation of a chemical specific growth and productivity fingerprint. Using this fingerprint we have been able to produce accurate predictions of fed batch cell growth and maximum titre. This illustrates the potential use of our product as an alternative to traditional, expensive, small-scale bioreactors. These plates have a minimal footprint thus facilitating the screening of much larger numbers of clones than traditional SS bioreactors. Additionally, as added value, we have demonstrated the use of our product to identify and verify the stability of cell lines – this is a comparatively rapid and in expensive alternative to the existing method employed i.e. genetic identification.
Project Context and Objectives:
Identifying cell lines that stably produce high protein titres is a critical part of biopharmaceutical development – this is traditionally a time-consuming, labour intensive process because their productivity and stability can vary enormously and large numbers of cell lines must be screened. Cell-line development is unsurprisingly seen as a major bottleneck in biopharmaceutical development. As cultivated mammalian cells have become the dominant system for the production of biopharmaceuticals, the characterisation and selection of highly productive cell lines is a critical step for efficient industrial manufacturing processes.
The primary scientific rationale underpinning the project was that in order to achieve high growth and productivity a clone must be able to proliferate / produce product while under intrinsic and extrinsic stress for extended culture periods in a bioreactor - can stress response fingerprinting predict clone bioprocess performance? The experimental concept is to create a high-throughout platform [Microtitre plate] which individually and collectively simulates the stressor environment [through a plurality of inducers of intrinsic and extrinsic stress in individual wells of the Microtitre plate] of a bioreactor and can thus be used to rapidly generate a ‘Stress Response fingerprint’ for any cell line or clone with the hypothesis that elements of that ‘Stress Response fingerprint’ will be capable of predicting, individually & collectively, elements of the Bioprocess performance/Identity of any clone.
Since 1982 the biopharmaceutical industry has been concentrated among relatively large Western companies. Today, 88% of global biomanufacturing capacity (as measured by total liters of installed cell culture capacity) is located in North America, Western Europe, Ireland, and the United Kingdom. High-cost, high-margin biologics are made in those locations primarily in massive, hard-piped stainless steel and glass production systems that have defined the state of the industry for decades. In recent years, the first generation of biomanufacturing has approached maturity, with new products flowing through established pipelines and revenue growth leading to overall industry profitability following decades of large capital investments.
New technologies, especially single-use production systems, are dramatically lowering building costs (and in many cases also operating costs) of biomanufacturing plants. Modular systems make it easier to produce multiple products in a single site, adding flexibility by significantly reducing lead times needed to modify a process for each new product.
Process improvements have increased productivity significantly more than industrial experts had predicted just 10 years ago, making the 2,000-L scale the workhorse for most biologics, including antibodies. The desire of many governments to have local vaccine and biologics manufacturing capacity to serve their populations is driving demand. Such evolutionary pressures will expand biomanufacturing from its current concentration in the West to countries across Asia, Eastern Europe, South America, and Africa — specifically, China, Russia, Korea, Brazil, and Malaysia in the near term. Bioprocessing will migrate from large-scale single-product installations to smaller and more flexible multiproduct systems. Looking ahead five years and considering projects that are now in various stages of planning or development, Asia and Eastern Europe will likely see the fastest growth in new biomanufacturing capacity. There is also significant interest in South America for manufacturing a range of biotherapeutics and in Africa for local vaccine production. Scanning the landscape of the evolving global biomanufacturing industry, seven general categories of enterprises that are either considering or already moving to build biomanufacturing capacity in new geographies exist:
1. Large, incumbent biopharmaceutical and vaccine companies;
2. Established midsized biotechnology companies;
3. New contract manufacturers;
4. New vaccine manufacturers;
5. Government-backed initiatives focused on biodefense and local populations;
6. Emerging biotechnology companies working on new drugs and biosimilar development;
7. Small-molecule companies seeking better margins by transitioning to biologics.
Established mid-sized biotechnology companies that have experience with both traditional and flexible single-use systems are well-positioned to move into many emerging markets, and many are planning to do so. With biosimilar-licensure regulations in place in Europe and on a path to being established in the United States, the number of new manufacturers seeking to capitalize on those pathways will grow.
In countries with nationalized healthcare, governments are the principal therapeutics buyers. Many such governments are getting involved with international partners to establish or incentivize construction of biomanufacturing capacity within national borders. MAbs now dominate the list of top-selling drugs worldwide. Industry reports indicate that nearly half of all drugs now in development are biologics. There is resurgence in the vaccine sector for biodefense and for life-saving medicines to treat local populations. The advent of biosimilar licensure for drugs coming off patent protection adds a new and powerful segment. Success in this new, globally distributed biomanufacturing industry will be a function of how well companies can leverage new technologies and manage the challenges of “in market, for market” manufacturing.
The principle objective of the PM-CELL project was to develop an innovative consumable product for rapid determination of CHO Clone Bioprocess performance [Cell Characterisation Technology] which will result in time and cost savings for the biopharmaceutical industry [including CMO] customers of the participating SMEs. As the project evolved and data began to be generated it became clear to the consortium that there were a number of additional potential applications of the ‘Stress response fingerprint’ being generated by the aforementioned consumable product which can be grouped into two distinct groups:
Application No. 1 - An innovative and rapid cell line selection technology to identify the most industrially relevant cell line applicable to a particular product through a relative ranking of Bioprocess performance prediction.
The annual global biopharmaceutical market is currently valued at $50 billion (€37.4 billion), projected to reach $70 billion (€52.4 billion) by the end of the decade. The demand for cell-derived biopharmaceutical products is increasing. This demand is caused not only by the fact that there are an estimated 800 biopharmaceuticals currently in the therapeutic pipeline but also because, patent expiry on many big-selling popular branded products is giving way to a biosimilar market. It is estimated that for every 2-months a product is not on the market, the average revenue loss is $2 million.
Application No. 2 - An innovative and rapid cell line identification and quality control technology which can be used to specifically identify and track the stability of any cell line at any stage of the development/production process.
Cell line authentication is now required by certain journals prior to publication and in some cases is mandatory before receiving funding from small granting agencies. The FDA has also instituted a requirement for the authentication of cell lines used to produce pharmaceuticals in their General Requirements for Laboratory Controls and the General Standards for Biological Products (21 CFR 211.160 (b) and 21 CFR 610.18 (b)). There are methods in place for authenticating human cell lines using multiplex PCR assays that target short tandem repeat (STR) markers in the human genome Although there are successful methods in place for human cell line authentication, methods for nonhuman cell lines are not well established.
15-30% of all cell lines are cross-contaminated with other cell lines or are misidentified. This means that billions of dollars have been wasted over the past 45 years on producing misleading or false data. There have been retractions of journal articles due to this problem – one example is the 2005 discovery that stem cells could seed cancer (Garcia-Castro, J. et al, Can Res, 2005. In 2011, Bayer conducted a study that led to the cessation of two-thirds of its drug target validation projects because their in-house experimental data failed to reproduce the published academic literature claims in 65% of projects examined.
The PMCELL project was conceived to focus on the development of rapid, cost saving, regulatory focused, screening assays for the Biopharmaceutical industry.
Workpackages 1 & 2 are primarily concerned with project management and Cell model establishment. The primary product concept and development workpackages are outlined as follows:
WP 3 – Characterise Microenvironments
WP3 is focused on characterizing the microenvironments which will provide a platform for future R&D activities of PM-CELL. WP3 has 3 No. objectives and 5 No. Deliverables.
Objectives
1. Identify and simulate the environmental conditions that influence cell culture in industrial bioprocess.
2. Comparatively quantify detection methods for the evaluation of cellular metabolic phenotypes in microplate format.
3. Create database to manage and support software application development.
4. Profile variant clones in culture environments.
WP 4 – Develop Microplate
WP4 is focused on market analysis, microplate design and software interface development and has 4 No. Objectives and 6 No. Deliverables.
Objectives
1. Perform a market survey to position the product in the biopharmaceutical industry.
2. Design the microplate for maximum efficacy.
3. Identify key performance indicators (microenvironments) that most effectively assess cell line performance i.e. cell performance index.
4. Create and develop the software interface.
WP 5 – Validate in Working Models
WP5 is focused on validating the core scientific principles in working models and has 3 No. Objectives and 5 No. Deliverables.
Objectives
1. Manufacture the prototype plate.
2. Validate the microplate in industrial producing cell models.
3. To use the designed microarray plate to differentiate industrial cell lines with known differences in growth and production.
WP 6 – Prototype testing
WP6 is focused on consortium training and obtaining as much user feedback on the prototype as possible and has 3 No. Objectives and 2 No. Deliverables.
Objectives
1. Provide scientific training and transfer knowledge on the product to technical SME staff.
2. To demonstrate the prototype product to industry users and enable on-site testing in industry environments.
3. Obtain feedback on product performance in end-user sites.
WP 7 – Dissemination
WP7 is focused on Dissemination and has 6 No. Objectives and 8 No. Deliverables.
Objectives
1. To ensure expedient exchange of results between the RTD Performers and SMEs in the consortium.
2. To prepare identify, capture and protect Intellectual Property for the SMEs in the consortium.
3. To prepare exploitation plans for the new technology.
4. To prepare market penetration strategies based on the Market Survey.
5. To disseminate the potential for the microplate based cell analysis platform to the end-users.
6. To plan for scale up from prototype to manufacture and commercialisation of the product.
Project Results:
The author has considered the PMCELL project into 2 distinct phases i.e. Project establishment, Cell model establishment & project management and Product development, Proof of concept testing, Prototype development, Validation and Dissemination. The first phase conveniently encompasses Period 1 and Phase 2 encompasses Period 2.
As Period 2 is concerned with the main S & T results, the author has chosen to go into a lot more detail on Phase 2/Period 2 than Phase 1/Period 1.
It is however important to note that PMCELL made progress during Phase1/Period 1 and ultimately produced an industrially relevant cell model which could be used for the remainder of the project – this was the crucial practical outcome of this phase as without it the project could not have progressed. The consortium was not able to develop the Cell model as submitted in the original application and thus we obtained an industrially relevant cell line from Cobra Biologics [Keele, UK]. Progress was also made on the assay format; a panel of analytes were selected and underwent development and optimisation according to a specific set of criteria devised to ensure the bioprocess, manufacturing and commercial focus of the final PMCELL product. A number of potential detection methods were identified and undergone preliminary evaluation. Whilst there were some delays and challenges, the PMCELL ‘Cell model’ was established for Phase 2/Period 2.
Microenvironment selection
In order to determine if the PM-CELL micro-environments selected for evaluation accurately predict CHO cell performance in bioreactor conditions it is necessary to generate two distinct clone sets - a training set and the validation set. The training set will assess if each selected micro-environment is able to differentiate between clones with varying cellular performances in simulated bioprocess conditions (fed-batch culture). Thus if a specific microenvironment is observed to exert the same effect upon each clone tested regardless of bioprocess performance it would therefore not be deemed suitable for further use. Micro-environments that appear to exert differential outputs will be incorporated for further analysis. Data from the training set clones subjected to all potential microenvironments will be sent to EuFor and AHVLA for statistical analysis to select the most effective micro-environments with the highest statistical power. Once selected, the validation set will be used to determine if these micro-environments can actually predict clonal cellular bioreactor performance.
Analysis was performed on the data collected from the fed-batch - obvious anomalous data points were removed from each data series. These artefacts arose from clumping issues that occurred during fed-batch studies resulting in erroneous readings generated by the Vi-cell. This is a common issue observed with CHO-S cells especially at high viable cell concentrations which renders the vi-cell unable to determine the existence of single cells. Any anomalous results were replaced by points extrapolated from the readings before and after the anomalous result. Various performance characteristics were considered to differentiate ‘good from bad’ clones in terms of growth and performance in fed batch culture. These include maximal viable cell density, culture duration to 90% and 50% viability and integral viable cell density (IVCD) to both 90% and 50% viability. These particular viability percentages were chosen as 90% highlights the start of the decline phase of culture. USFD was advised by industry experts that 50% is the standard point of harvest from bioreactor runs for the majority of easy to fold recombinant proteins. Bioreactor productivity is a function of both IVCD and cell specific productivity. Therefore IVCD was identified as a key culture parameter.
One-way ANOVAs were performed on all data described. Prior to this, post-hoc tests were performed to assess significance of individual differences. For this, Tukeys multiple comparisons test was performed (which takes into account increasing type I error potential with multiple comparisons). From this, “honestly significant difference” confidence limits (alpha at 0.05) were determined for ease of visualisation of significant differences. Review of all data revealed that the most promising parameter to differentiate the clones was IVCD at 50% culture viability. Statistical analysis of this parameter revealed 14 (5 low, 5 high, 2 very high IVCD 50%) significantly different clones from the 35 Cobra clones generated. This number is insufficient for a complete training set to be able to accurately test the predictive power of the microenvironments. USFD has access to 29 alternative CHO clones and 11 CHO-S non-producing clones, all of which possess a high probability of clonality. Fed-batch studies are currently in progress to enable statistical analysis.
We previously demonstrated that clones within the training set would be cultured in mini-parallel bioreactors to monitor performance in simulated bioprocess culture processes. Although fed-batch shake flask culture is a good representation of how cells behave within a bioreactor process there are certain differences between the processes, which can alter cellular behaviour. For example cells in fed-batch shake culture are maintained in suspension by rotation of the incubator plate that the culture flasks attach to. However, in bioreactors suspension is achieved by stirring the cells with an impeller or air-lift reactors. These processes can cause an additional stress to the cells. Also gassing mechanisms are different between both processes. In bioreactors the cells are sparged whilst fed-batch cultures are maintained in a CO2 environment. Sparging has been well characterised to exert detrimental effects upon cell growth and recombinant protein production.
USFD negotiated access to trial the Hexascreen automated multiple mini-bioreactor system (Hexascreen, Spain; www.hexascreen.com). This system allows for 6 mini-bioreactors to be performed in parallel (10-15 mL total volume). Culture information such as temperature, dissolved oxygen, optical density and media pH can be monitored and recorded. Discussions with Hexacsreen confirmed that although automatic feeds / sampling could not be set up it was possible to manually add feeds to the individual bioreactors and remove samples removed for subsequent analysis. The Hexascreen system was installed into USFD on 3rd July for testing to determine its suitability for this task. During the four weeks trial period USFD experienced a number of problems including persistent erroneous readings from one of the optical density sensors and issues with pausing the system to collect samples and add feeds (Fig 1).
Alternative small-scale bioreactor models exist however USFD was unable to negotiate rental of these systems. Following the failure of the Hexascreen system for the purposes intended for this project USFD entered discussions with Cobra and Life technologies regarding the expected similarity between fed-batch clone data and scale down model systems, such as the Hexascreen system. Life technologies indicated that ranking of clones based on their performance in shake flask fed-batch culture should be mirrored in scale down models when the same processes are imposed (e.g feeding regimes). Furthermore Cobra Biologics, who generated the parental cell model employed in this project, stated that they have not observed any significant differences in clonal performance between shake flask fed-batch and bioreactor systems.
USFD and ITT generated a detailed database listing potential culture microenvironments (i.e. chemicals/biochemical), replicating a wide variety of stresses encountered during a typical bioreactor run, for inclusion on the final PM-CELL plate (Table 1). However, this range of chemicals was substantially reduced due to limited solubility in water/media. Following unsuccessful preliminary efforts to solubilise the compounds, it was deemed that a major criterion for compound selection was solubility. Additional criteria were also considered whilst selecting which compounds detailed in the database to test. Overall, the aforementioned criteria can be considered as primary and secondary considerations depending on importance.
Primary considerations are as follows:
(1) Solubility: Originally thought that all compounds must either dissolve in water or media. This attribute would greatly facilitate the manufacture, distribution and storage of the plate as no ‘solubilising agents’ such as organics would be required;
2) Manufacturing compatibility: Manufacturing compatibility is an issue of high significance, as compounds may potentially have toxic effects in the PM-CELL plate manufacturing process and/or for the end user. Compounds with a higher category toxicity rating of grade III were eliminated;
Secondary conditions are as follows:
(1) Stability: Compounds ideally should be stable in light to increase the shelf life of the product. For example, Actinomycin D was immediately ruled out as in powder format, it is sensitive to light and must be sealed and protected from light and moisture at 2-8OC. Dilute solutions are also very sensitive to light and must be discarded and not stored long term;
(2) Cost: This is a major consideration in relation to the economics of final product manufacture and retail price. Estimated price per well was calculated from reported effective concentrations in the literature and price of compounds from the Sigma Aldrich catalogue. Compounds with an estimated cost greater than 10p per well were not usually considered unless alternative compounds simulating the same effect could not be identified.
A screen of compounds to test were ordered and tested on the Cobra parental cell line. Initial investigations were performed in small-scale shake flask culture, cultiflasks (Satrorius Stadium) total volume 10ml, whilst the microplate detection method was being optimised. Cultiflasks were seeded at standard concentrations (0.2x106 cells mL-1) with a range of concentrations per compound. Appropriate concentration ranges were selected following a detailed search of the literature. Cell counts were collected daily for 96 hrs (Figure 2). From the data collected an IC50 concentration (concentration required to inhibit 50% of response, growth in this case) was estimated for each micro-environment (Table 1). From this 26 compatible microenvironments have been identified simulating a wide range of bioreactor stresses which permitted the elucidation of an IC50 value.
Following compound selection and screening in ‘cultiflask’ format, for plate design, the compounds were to be tested in 96-well plate format with the finalised detection method. This would enable determination of compound specific microwell compatibility and microwell specific IC50 in addition to evaluation of detection method compatibility.
Detection method selection
Clone Select Imager (CSI)
To evaluate the CSI as a potential detection method it was initially optimised to detect CHO cell growth over 96 hrs. During optimisation of the seeding density it became apparent that there was a large variation in estimated confluency between wells, in some cases a 2-fold difference. This variation is due to a phenonemon deemed the ‘edge effect’ (figure 4), whereby the cells appear to settle around the edge of a well. We speculate that the variation in confluency between an ‘edge effect’ well and one where the cells are evenly distributed across a well is due to the fact that once cells are settled at the edge of a well there is limited room for growth. Therefore the CSI is unable to differentiate between cells in mutliple layers and underestimates the well confluency. To investigate the cause of this effect and potentially reduce well-well varation within a plate a number of experiments were performed, testing various hypotheses.
(1) Cell line specific effect
We hypothesized from this work (figure 5) that the ‘edge effect’ arises from the cells proprensity to clump together, an artefact exhibited with CHO-S cells.
(2) Anti-clumping agents
We tested 2 anti-clumping agents at various dilutions to investigate their effects upon the edge effect. Suspension flasks (total volume 30ml) were seeded to monitor the agents’ effects upon Cobra cell line growth (figure 6) and plates seeded for analysis on the CSI. All dilutions tested for the anti-clumping agent from Invitrogen and PF-68 (0.1%) increased final viable cell density compared to control flasks. PF-68 at 1% appeared to reduce cell proliferation rate and was therefore not tested in the plate format. Despite the positive effects upon viable cell density no differences in the edge effect were observed with no change in well-well variation.
(3) Trypsin
We investigated the possibility of using trypsin to detach the cells, back into suspension mode, to reduce the well-well variation. A wide variation was still apparent the post trypsin incubation which was reduced following the 30 min settling period.
Nephlometer
NephloSTAR (BMG Labtech, UK) is a laser-based microplate nephlometer that measures suspsended particles in liquid by light scattering. This system fulfils the requirements to be considered as a potential detection method as it is very easy to implement following successful generation of a standard curve to extrapolate unknown concentrations. It is a relatively common system in many cell line development companies especially those working in the high-throughput microplate format. A standard curve was generated using the Cobra cell line with limited variability between replicates (figure 7a). However these measurements were taken immediately post seeding prior to settling and occurrence of the edge effect. However, once the cells start to adhere huge variations between replicates develop (figure 7b).
Presto blue
Presto blue (Life Technologies) is a cell permeable resazurin-based solution that functions as a cell viability indicator by utilizing the reducing power of living cells to quantitatively measure the cellular proliferation. Resazurin is reduced by cellular NADH to give fluorescent resorufin which can be detected using a standard fluorescent plate reader (Figure 6). Compared to alternative viability assays such as MTT, Presto blue is very easy to use with a single addition of the solution and no cell lysis required. USFD evaluated Presto blue as a potential detection method as it fulfills all necessary requirements. Firstly, the assay is easy to implement. Secondly, the fluorescence change is detected by a standard fluorescent plate reader, a common piece of equipment in all biopharmaceutical companies. Finally the output, metabolic activity, is directly linked to bioreactor performance. Following optimization of the protocol it was found to give accurate determination of cellular density with very low variability (figure 8).
Presto blue has been implemented to initiate microenvironment IC50s determination in microplates using optimized seeding density (0.2x106 cells mL-1). Plates were seeded and Presto blue measurements performed every 24 h. Various concentrations were tested, selected using the cultiflask suspension data (Deliverable 3.1). Comparable replicate variation was observed (figure 9).
A total of 5 detection methods have been evaluated by USFD and ITT. Two potential detection methods have been identified which fulfil all necessary requirements (USFD). Flow Cytometer has been proven unsuitable for use with the PM-CELL end product. Fluorometric measurement with Prestoblue® is the detection method of choice as it is easy to implement, detectable by a standard plate reader and metabolic activity as the output is linked to bioreactor performance.
Microenvironment characterisation
Following this decision, all the above mentioned compounds were screened in microwell format using the ‘PrestoBlue’ (PB) assay. Ranges of compound concentrations were screened based on those used in the 10ml format. Due to the kinetics of the cell growth in the 96-well format, i.e. there is no substantial cell growth after day 3, it was decided that 72 hours would be the optimal time of incubation of cells in the microenvironments since this maximises control cell growth (this signal to noise ratio in model) while minimising the time period.
An envisaged issue with using a chemical based detection method is that the microenvironments may directly interfere with the assay chemical which may compromise the accuracy of cell growth determination in certain microenvironments. To investigate this, a plate containing cells and microenvironments at relevant concentrations were assayed at 0 hours post seeding. This would allow a measure of direct influence of compound on the absorbance value from the PB compound. From this it is evident that none of the microenvironments tested directly affect the PB assay to any substantial degree.
IC50s for 72 hours were determined for each microenvironment in 96-well format utilising the PB assay. From this data it is evident that a clear dose response curve can be obtained for each microenvironment in 96-well format.
Following this work, it is now necessary to construct a prototype clone screening plate to test for repeatable IC50s measured in high throughput format, plate-plate variation and within plate variation.
A schematic overview [figure 10] of creation and utilisation of microplates (1) 45 ul of 2x IC50 chemical (in appropriate media) was added to the plates en masse, (2) 45ul of clonal cells were added to the plate prior to incubation at 37C for 3 days, (3) 20ul of ‘Presto-Blue’ (diluted 1:1 in CD-CHO) was added to the plates. Plates were mixed by shaking and incubated at 37°C for 30 minutes prior to reading absorbance.
From each clone, a spread-sheet of growth in chemical inhibitors was generated. These data were the average of 9 technical replicates for each well (9 readings were taken to cover the whole well area). Subsequently the mean of the triplicates was taken to give a specific growth for each chemical microenvironment.
To give a ‘clone specific response’ to a chemical microenvironment, growth in a ‘chemical well’ was normalised to the control well growth (i.e. control well growth was taken as 100% and growth in chemicals was relative to this). Subsequently, for each clonal group (either PM cobra or P2) to the chemicals were normalised, i.e. the mean of the cell specific response for a chemical for the cell line in question was subtracted and divided by the standard deviation of the response to the chemical.
In addition, cell lines were given a rank according to their performance in the fed batch study. This ‘ranking’ enabled comparisons between the cell lines (PM and P2). Subsequently the data files from the 2 cell lines were merged. This merging allowed statistical analysis to be performed on the entire data set.
Statistical analysis was performed in R. The initial approach of the analysis was to model the ‘standardised cell specific microenvironment response’ (CSCR) (explanatory variable) to the standardised rank IVCD50 (SRI) (response variable).
Following analysis of individual correlations between ‘CSCR’ and ‘SRI’ it was evident that (a) there were few ‘significant’ individual correlations and (b) from a view of scatter plots of CSCR and SRI, it was clear that there were no ‘non-linear’ relationships present.
From this, it was decided to build a multiple linear model between the various CSCRs and the SRI. The overall approach to building this model is illustrated in Figure 11.
From this modelling approach, a model was elucidated which is in the Table below. From initial analysis the model built is highly significant. i.e. there is a 0.00000008249% chance that this model could occur by chance. The model R2 was around 0.9 indicating that approximately 90% of the variance in the data can be explained by the model. The adjusted R2 was also considered very good (this is a version of R2 which is penalised for the number of regression terms used).
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 1.24442 0.01765 70.509 < 2e-16 ***
BCH -0.14454 0.03969 -3.642 0.00298 **
Sodium Butyrate -0.14459 0.03858 -3.748 0.00244 **
Cadmium Acetate 0.09827 0.03190 3.080 0.00877 **
Citric Acid 0.14028 0.03348 4.190 0.00106 **
Sodium Lactate -0.11869 0.03096 -3.834 0.00207 **
3 Bromopyruvate 0.13289 0.03244 4.097 0.00126 **
---Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.07893 on 13 degrees of freedom
Multiple R-squared: 0.9003,Adjusted R-squared: 0.8543 F-statistic: 19.57 on 6 and 13 DF, p-value: 8.249e-06
Microplate development
Optimal concentrations of chemical inhibitors were determined. Proof of concept was produced and consequently the status of the project is “GO”. Following discussions with the end plate manufacturer, Microcoat, the final plate format was agreed [figure 12] and designs for pre-beta testing plates were finalised.
End-user protocol
To define the end user protocol, two main process variables were investigated. These were the time post-subculture [see Figure 13, 14 & 15] for plate assay and also, passage of the cells. Information from these will allow assessment of the impact of these process parameters to accordingly design the end user protocol. Other aspects of the user protocol such as the set up of the plates and the assay of cell growth in the plates has been established. The following end user protocol is proposed:
(1) Use clones at the same passage where possible;
(2) Set up plates using cells at day 3 post subculture (or a fixed time post subculture where day 3 is not appropriate e.g. when this is not ‘mid exponential’;
(3) Dilute cells to a density of 0.2x106 cells/ml in growth media;
(4) Add 90ul of cells to the chemical plate;
(5) Incubate plate for 3 days in a static, humidified incubator with conditions optimised for CHO cell growth (such as 37C, 5% CO2, humidified);
(6) After 3 days, dilute presto blue 1:1 in growth media, allow to equilibrate to room temperature. Ensure this is thoroughly mixed. Add 20ul of the presto mix to each well;
(7) Shake for 10 seconds (orbital, 1200rmp, 1mm diameter);
(8) Incubate to 30 minutes (although this may be optimised by the end user and may depend on the rate of growth of the cells in question);
(9) Shake as above, and read fluorescence (Abs 535, Emi 620) using the supplied BMG plate reader.
Cellular-performance Index [CPI]
The CPI focussed on utilising the biological data to accurately model the optimal predictive capacity of the various environments thereby establishing a model which can be further evolved through beta testing. Two panels of industrially relevant clones were created and their fed batch performance was assessed. Here, our metric of “fed batch performance” was integral of viable cell density using a cut-off were cells dropped below 50% viability (figure 16). A wide range of chemical inhibitors of cell growth and metabolism were selected. These stress the cells in a wide “multi-dimensional” way. These were chosen to simulate relevant stresses and to generally test the resilience of the cells in a in a number of different ‘metabolic dimensions’ e.g. resistance to oxidative stress/cell cycle inhibition/inhibition of protein production/ NAD production etc. For a selected panel of clones, each was subjected to the chemicals in a 96 well plate format. This enabled a “cell specific” chemical growth profile to be generated. This can be viewed as a “metabolic fingerprint”. Using aspects of clone metabolic fingerprint, a multi-linear model was build with the power to accurately predict the relative cell performance of a clone (IVCD50 generated from fed batch studies) using information from the metabolic fingerprint. The Model was validated using an “n fold cross validation” technique. Data was randomly split into “training and “validation” sets. Models were built using training sets, and their predictive power was tested using validation sets. Essentially, this model was able to predict new data Points and classify new clones into “good” and “bad” performers with 85% accuracy (figure 17 & 18). Using our plate based “stress” plates to generate a “stress resistance profile” we were able to use aspects of this to accurately predict bioreactor performance (relative fed batch IVCD50) with 85% accuracy. Using our technology it is speculated that this will allow very high through put screening of hundreds of ‘candidate’ clones with a minimal fingerprint (all that is required is static incubator space – average incubator capacity could easily exceed 200 plates), using a process which requires little specialist knowledge or training. Moreover due to the ‘user-friendly nature of this technology is it easily incorporated into an automated system making it compatible with large scale workflows.
Software Interface
To enable correlation between the microplates phenotype fingerprinting of the antibody producing cell lines, the cell lines will be classified into high and low producers based on calculations made on the measures obtained from growth capacity in multiple chemical micro-environments. The role of the SW is to provide a user platform to
1. Import learning data from know cell lines
2. Create a classification model from the learning data
3. Import test data from unknown cell lines to be classified
4. Classify the unknown cell lines based on the learned model into high and low producers
5. Report the classification results to the user for decision making and for further integration into other laboratory management systems.
A beta-version (figure 19) of the SW has been achieved and can now be tested with users in a laboratory environment.
Plate Manufacture
Following proof of concept, the next step towards commercialisation is to make a technology transfer from wet “stress profile plates” manufactured ‘in-house’ to dry stress profile plates. We carried out an extensive evaluation of the requirements and technologies required to successfully manufacture the plates internally at Technopath and came to the conclusion that in the time frames permitted there was no possible way this could be achieved. We chose instead to partner with an expert in the field and we carried out an extensive review of the various key players operating in this space in Ireland, UK and across the EU. We eventually settled on a company called Microcoat and the TPath Technology transfer and validation team visited the Microcoat facilities in Germany. MicroCoat Biotechnologie GmbH are based at Am Neuland 3, 82347 Bernried, Germany, www.microcoat.de.
A plethora of manufacturing conditions were evaluated. Using the reconstitution conditions of “Just add cells”, these manufacturing conditions were compared. Criteria considered were sterility, inter-plate CV, and level of chemical inhibition achieved.
With respect to sterility, there was only one contamination and that was present on an 'Unradiated' plate. This is not strong evidence for a systematic problem with unradiated plates since there was only 1 well contaminated, despite many unradiated plates being examined. Nonetheless, it would be best to proceed with radiated plates as this certainly did not have an apparent detrimental effect on the performance.
With respect to all other manufacturing conditions, the plate performances were all very similar. Given this, the best manufacturing conditions for us to proceed with are those which gave a high level of chemical inhibition and a low CV. These manufacturing conditions were:
Nunc plates, 12h drying time, 30C drying temperature, 4G irradiated.
These conditions gave good levels of inhibition and a inter-plate CV value of 4.5% from a 4 plate replicate. The levels of chemical inhibition achieved with these conditions is illustrated in figure 20. The dotted lines represent 100%, 75% and 50% growth. Bars highlighted in red are chemical which exceed the 75% inhibition limit. The error bars are standard deviations from 4 separate plates.
The limit of acceptability was 75% growth inhibition (I have highlighted any chemical which goes above this limit in red). There are 5 chemicals, which exceed this limit. Overall the majority of the chemicals give adequate levels of inhibition. Also, it is clear, there is very little variation between the plates, which is very promising.
Product level assessment
An in-house panel of clones which produce IgG was sourced. These were subjected to a fed batch growth and production study. Samples were taken daily. These samples were assayed using a protein A HPLC method. This method was chosen and developed for it’s rapidity, accuracy and cost effectiveness. Bio-Monolith Protein A separates IgG in 1 to 2 minutes, high recovery (>95%). This method has a large linear range as is demonstrated by the typical standard “curve” given in Figure 21. Additionally, the assay was able to give a linear standard curve over the concentration range of 0.01 to 0.2 g/L). This is demonstrated in Figure 22. Furthermore, peak profile and retention time remained consistent which demonstrates column performance is maintained over time. This is illustrated in Figure 23. The max MAb titres from the fed batch are illustrated in Figure 24.
Having established the fed batch antibody titres and clone specific chemical responses for the PM cells, the ability to predict MAb titre using information from growth on the chemical plates was explored. A number of modelling techniques were explored including PLS, PCA and multiple linear modelling. In essence, no model could be achieved which was substantially predictive. The best models achieved was a two parameter linear model. Using the “fitbest” package in R, the best 2 parameter achieved an R2 of of 0.43. The model fit is illustrated in the Figure 25. It was concluded that it is not possible to give industrially useful predictions of MAb titre using information from the cell specific chemical growth response we re-evaluated with a number of potential customers and it was outlined that in fact it would be more interesting and much more valuable if an association could be demonstrated between the product levels in the plate and the product levels at Fed-batch. The consortium considered this for some time and came up with a design of experiments to support evaluation of same which was not performed by November 2013 but was performed subsequently. This data, along with an additional application of the technology is most succinctly described in poster presentations as figures 26, 27 & 28.
Potential Impact:
The final result of this project will be generation of the PM-CELL product line to support the rapidly growing biopharmaceutical market. Our business model is focused on the installation of our fluorescence reading platforms and to supply a suite of niche process improvement solutions [96-well microtitre plates] for various parts of the Biopharmaceutical production process – all focused on the following primary drivers:
• Speeding up the Biopharmaceutical development process;
• Enhancing traceability in Biopharmaceutical development;
• Enhancing safety in Biopharmaceutical development;
PM-CELL is already, and will continue to be, a true International partnership:
• Technopath [Ireland] – Product concept, R&D, Validation, Regulatory approval, Logistics, IP, Branding, Commercialisation, Sales, Marketing, Logistics & Customer support;
• Dorteegelund [Denmark] – R&D, Sales, Marketing & Customer support for Norway, Sweden & Denmark;
• IUL Instruments [Germany] - R&D, Sales, Marketing & Customer support for Germany, Austria & Switzerland;
• University of Sheffield [UK] – PM-CELL R&D Centre of Excellence.
• AHVLA [UK] – Modeling support centre during PM-CELL 286065.
• Euformatics [Finland] – PM-CELL Software Development and Support Centre;
• BMG Labtech [Germany] – Harware OEM Manufacturer;
• Microcoat [Germany] – Microtitre plate OEM Manufacturer.
The project has facilitated close interaction between SMEs & RTD performers from different member sates, providing significant insight on the product development and commercialisation process. Projects/Initiatives being explored by consortium members include:
• Technopath and AHVLA are discussing collaboration on an existing modeling project for the dairy industry.
• Technopath and EuFor are discussing collaboration on commercial software development projects.
Furthermore, the SMEs have been introduced to RTD collaboration which should lead to future project development activity and funding acquisition. The highest impact and value of this project will be realised upon delivery of the PM-CELL product, this will have significant impact for the SME beneficiaries, the European and Global Biopharmaceutical sector.
Each niche process improvement solution will involve a different assay kit [see Figure 29], which in effect will be a unique suite of synthetic microenvironments which collectively are capable of establishing a unique Stress response fingerprint for each individual cell line which is associated with/predictive of particular bioreactor performance variables.
We currently have beta versions of the following process improvement solutions:
• CHO Clone Bioprocess performance;
• CHO Clone tracking and profiling.
We continue to work on the proof of concept of the technology in relation to Stability testing.
In its original manifestation, the PM-CELL product was focused on decreasing cell selection time by 33%, equating to approximately 8 weeks. Considering that each week a biopharmaceutical product is not on the market results in estimated losses of €2.02 million the PM-CELL product will significantly reduce costs and time to market, thus increasing sales and profitability. This is still very valid.
In addition, novel regulatory, cell harvesting, cell line development and media development & optimization are also core. This project through its advancement in mammalian cell selection for the biopharmaceutical industry is an excellent example for the Innovation Union, one of the flagship programmes of Europe’s 2020 strategy to deliver “a smart, sustainable and inclusive economy” to sustain Europe’s competitiveness.
Socioeconomic Impact
For the four PM-CELL SMEs significant benefits will be gained with development of the PM-CELL product(s). The SMEs will benefit from economic growth and increased employment due to IP ownership and selling rights. The PM-CELL product(s) will improve SMEs competitiveness, as key suppliers to the highly regulated biopharmaceutical industry. The PM-CELL product(s) will open up new target markets providing economies of scale and improved return. Furthermore, there is significant potential to transfer the PM-CELL product(s) to other biopharmaceutical sectors including the bio-similar and cell vaccine wherein the EU are the current market leader – this may be further extrapolated to the Stem Cell market which is a key target area for the EU. The SMEs will have a competitive advantage in driving sales into new export markets of North America, Asia and China. The SMEs in this project represent the lifeline of regional and national economies as 99% of EU companies are SMEs accounting for 67% of jobs. With trans-national cooperation in the PM-CELL consortium the SMEs collaborate with SMEs and RTD performers from different states. Such collaborative efforts have the potential to result in new business opportunities, increased sales, and identification of new commercial partners to improve products or processes and access additional know-how. Trans-national cooperation will encourage rapid market penetration and generate the sales revenue that will be required to expand sales of the product into export and other markets. To the end-user, the biopharmaceutical industry, the SMEs‟ will supply a novel high-throughput cell screening product, which will result in a marked reduction in R&D time and costs and greatly reduce the time to market for new products resulting in increased sales revenue and greater profitability. Currently, the length of time to produce a high yielding cell line is 25 weeks of which 11 weeks are devoted to cell selection. The SMEs PM-CELL product can reduce the cell selection time by 8 weeks. To the EU citizen, as a net receiver of medical research and targeted therapies, the PM-CELL product has the potential of accelerating product development and also, as a direct result of the cell identification work which resulted from this consortium but which was not included in the original project scope, the industry and patient will greatly benefit from a robust, inexpensive, tracking mechanism for the bioproduction machinery.
Dissemination activities to date have included:
Bioproduction 2013, 22nd & 23rd October 2013, Dublin, Ireland – Poster Presentations [Dr. Ben Thompson & Dr. Robert Whitfield].
‘Microplate based stress resistance fingerprinting technology for rapid, high-throughput cell-line development’.
‘Microplate based stress response “barcoding” technology for rapid and facile CHO cell type identification for quality control.’ Posters won Best Communication at the meeting.
Cell Line Development & Engineering 2014, February 13th 2014, Vienna, Austria – Keynote [Prof. David James].‘The Influence of media with cell line engineering’.
Bioprocess Industry Day 2014, 6th February 2014, Sheffield, UK – Invited talk [Dr. Ben Thompson].‘Rapid prediction of CHO Cell Clone Functional Performance by Microplate-based Stress Response Profiling’.
Invitation to speak at Cell Culture Engineering XIV, May 4th-9th 2014, Quebec City, Canada.
Invitation to speak at Bioproduction 2014, 8th – 9th October 2014, Barcelona, Spain.
It is important to outline that the consortium has agreed to continue beyond the duration of the EU-funded project. TPath have entered a 10-month Research Agreement with UFSD in the first instance which will be paid for directly by TPath.
It is equally important to outline that PMCELL have captured, documented and filed 3 patents and we intend to submit a further patent by the end of March 2014:
Cell ranking/Cell performance application [Cell growth]
On June 26th 2013 we filed a new European Patent Application [1317387.8] entitled ‘A rapid, high throughput, method of predicting relative fed batch performance of a panel of clonally derived cells from a host cell population’. We received an extended search report and initial opinion on 13th November 2013 and have been discussing same electronically. The consortium has a scheduled workshop on January 20th to consider this patent in full. Additional applications claiming priority from the above application must be filed before 26th June 2014.
Cell Identification application
On September 25th 2013 we filed a new European Patent Application [13185979] entitled ‘A method of identifying a cell’. We are currently awaiting an extended search report and initial opinion.
Cell ranking/Cell performance application [Product titre]
On 30th January 2014 we filed a new European Patent Application [14153333.1] entitled ‘A method of predicting relative fed-batch production titer of a panel of clonally-derived producer cells’. We are currently awaiting an extended search report and initial opinion.
As the production of data in relation to this project is an ongoing development, the consortium have decided to adopt the existing mechanism of dealing with the publishing of results. In summary the data is reviewed at the individual RTD performer level, again at a consortium level and structurally, on a monthly basis by the Exploitation Manager & Commercial Director at Technopath, along with Mr. Barry Purdy of Purdy Lucey IP. This provides a very robust identification and analysis of IP Generation and if following analysis IP has been deemed to be identified then it is immediately captured and protected. Periodically we also add to our core Freedom to Operate database.
Any plans to publish information or results by any member of the consortium must be approved by the Project Management Committee and all approvals are associated with a formal sign-off by the exploitation manager, Commercial Director of TPath and Barry Purdy prior to submission for consideration for publication.
As mentioned earlier this system has worked very well and we have already approved a number of presentations, both Oral and Poster. We have also approved the preparation of a full manuscript which will be reviewed prior to submission.
As the IP ownership is clearly defined in all the PMCELL agreements, and there are no separate ‘fields of use’ limitations associated with individual consortium partners, the consortium deemed it not necessary to prepare a structure Table of ownership by project result.
The parties to the consortium have made no changes to the treatment of background intellectual property from that outlined and clearly defined in the consortium agreement, including both the background included and the background excluded.
The parties continuing to work on the project i.e. UFSD, DortE, IUL, TPath & Eufor are adopting identical treatment of the background IP conditions, specifics of which are included in the various contractual arrangements.
Our current target for formal launch of the product(s) is the Bioprocess International Conference & Exhibition at the Hynes Convention Centre in Boston, USA on October 20-23, 2014.
List of Websites:
The project website [www.pm-cell.eu] was developed as planned. As we will shortly enter the formal naming/branding of the product, it is likely that the product will not be named PMCELL and thus we will transfer a certain amount of the information from the current website to the product website. We intend to perpetually acknowledge the source of the seed funding for this product.