For over four decades, flow cytometry has supported drug discovery and development, including the development of the first humanized monoclonal antibody therapy alemtuzumab (Campath) for the treatment of B-cell chronic lymphocytic leukemia.1 The technology has become progressively more sensitive and is now routinely used for more in-depth quantitative analyses. In the 1980s, flow cytometry was used as a qualitative tool to demonstrate the removal of lymphocytes, including cancerous B cells, in bone marrow and peripheral blood following treatment with alemtuzumab. Today, flow cytometry is also being used in biomanufacturing to detect small quantitative differences in the lot-to-lot potency of monoclonal antibodies, such as alemtuzumab.
Principles of Flow Cytometry
In brief, cells, bacteria, beads, and particles, including extra- or intra-cellular molecules of interest, are stained with fluorescent dyes and/or fluorophore-labelled antibodies and passed through the flow cytometer. As each cell/particle flows in single file through a laser beam, particle size and density scatter the laser light, which is detected by forward scatter (FSC) and side scatter (SSC) detectors. Fluorescent dyes and fluorophores are excited by the laser and emit light of a longer wavelength, which is then detected and amplified using photo multiplier tubes (PMTs). These data are used to create a profile of light scattering for each cell/particle. Typically, between 5,000 and one million events per sample are collected for analysis, depending on the type of analysis and rarity of the event of interest.
Multi-parameter flow cytometry assays generate a vast amount of data, which is typically displayed graphically using histogram scatter plots. Using boolean gating, a specific region of the histogram (e.g., cells expressing a specific marker) can be analyzed further in separate histogram plots (Figure 1); this is particularly helpful when analyzing biomarker expression on the surface of rare cells. The data presented from flow cytometry experiments are typically reported as the percentage of cells positive or negative for a particular set of markers. In addition, the relative expression levels of important markers are often also reported.
With increased sophistication of the technologies, flow cytometry has become a very versatile tool when employed by experienced users, but results can easily be misinterpreted if the assays are not well designed and appropriately controlled. Data analysis too can be complex if carried out by those without proper training and experience.
Pharmacodynamic Analysis Of Clinical Samples
Flow cytometry has proven to be a useful tool for pharmacodynamic (PD) biomarker analysis in clinical trials, both for measuring protein expression on the surface of cells and for analyzing how a drug candidate is influencing specific cell types within the body. When assessing the PD of a drug candidate, in clinical studies, using cellular analysis such as flow cytometry, there are several important elements that need to be considered at the onset of designing a clinical protocol, including sample collection, processing and storage, reagent qualification, method validation, and data standardization.
Sample collection, processing and storage
Sample collection is the critical first step in the flow cytometry workflow. It is very important that the sample is collected at the right time using the right methods and stored appropriately. Otherwise, study results can be significantly compromised. Therefore, when designing a clinical protocol, several considerations must be made for sample collection:
- What matrix is required? (sputum, ascites, blood, biopsy)
- What sample volume is needed? (1 mL of blood = 1,000,000 PBMCs)
- What is the frequency of the cell population of interest? (rare = less than 5%)
- What precision is needed or is possible? (see Table 1)
- Are absolute counts required?
- Can the population of interest be practically stored/processed or cryopreserved on site?
- Is sample stability affected by transport, incubation conditions, processing or cryostorage?
- Is contingency needed for international shipping, weekend sampling, ad hoc samples?
- Do subjects enrolled in the clinical trial have “normal” cell profiles?
- Do samples need to be kept sterile? (i.e. will they require culture or activation?)
- Analyzed immediately (within two hours) as whole blood using a staining and red blood cell lysis procedure;
- Incubated at ambient temperature for 48 hours and then analyzed using a staining and red blood cell lysis procedure;
- Processed to peripheral blood mononuclear cells (PBMCs), cryopreserved, thawed and then stained for analysis;
- Thawed PBMC samples were also fixed post staining and analyzed the following day.
These data demonstrate that to measure accurate CD16 expression levels, the clinical samples should be analyzed immediately, or at least within 48 hours, and analysis should be carried out at constant post-sampling timepoints to ensure equivalent drift between samples. Cryopreservation should only be used to produce qualitative data (e.g. patient stratification into high and low biomarker expression levels), but not absolute quantitative measurements. If quantitative data are required, clinical sites should ideally be within proximity of the testing facility. If clinical sites are in another country, or widely distributed, cryopreservation of processed blood samples will be required. A plan of how to interpret and compare the resulting qualitative data should be prepared as part of the trial protocol. Inter-study comparisons can be made and are likely less variable than for fresh samples if standardization of sample handling procedures is adopted across studies. In these cases, it is important that all samples are incubated for equivalent periods prior to cryopreservation to ensure that reduction in biomarker expression levels is comparable; this can be monitored using clinical report forms or procedural records. In this example, CD16 expression levels may be restored by resting thawed cells in an incubator overnight. Though possible, this may not be feasible for large sample numbers arriving from multiple clinical sites. In addition, the biological reproduciblity of recovery approaches can prove challenging. Throughput, reproducibility, sample integrity/stability, end use of data, logistics, monitoring, clinical site, technical expertise, time frame, geography and cost must be taken into consideration by CROs or in-house laboratories.
As whole blood cannot be directly cryopreserved, PBMC samples are widely used. This is especially true for analysis of clinical study samples from immune system targeting drug programs. Validated practical procedures exist for the enrichment and cryopreservation of PBMCs and often cryopreservation stabilizes binding of drug to cells, thus enabling shipment to testing facilities. This approach allows clinical samples to be batched together (pre- and post-dose) for analysis, reducing inter-batch variation and greatly improving scientific validity and statistical power; logistical costs are also reduced by batchwise shipping, while shipping conditions, such as temperature, are more easily controlled. Cryopreservation can also provide a contingency plan. For example, when European air traffic was suspended due to volcanic eruptions in Iceland, shipping times across Europe were increased from hours to days, ruining whole blood analysis relying on 24-hour shipments; for cryopreserved PMBC samples this was no issue. There are now standard approaches that can be used to process PBMCs, which simply require a tabletop centrifuge, cell freezing container, and a -70°C freezer. Cryopreserved samples for non-functional analysis can usually be stored for 12+ weeks at -70°C. If the samples are needed for functional assays they should be quickly transferred from -70°C to liquid nitrogen storage. Although cryopreserved PBMCs can accurately represent the patient profile at the time of blood collection, this approach is not applicable for all cell types or when requiring absolute cell counts.
Once a sample collection technique has been derived for a study (e.g. PBMC collection, stabilization buffer, etc.), a standardized collection kit can then be developed and delivered to clinical sites for sample processing. The use of sample collection kits can significantly improve data integrity, but operator training and assessment on the use of the kits is critical to ensure that all samples are treated in the same manner according to the defined protocol. To control the quality of sample collection, processing, storage, and shipment, training programs for clinical site staff should be implemented. The most successful programs should consist of theory presentations, onsite practical sessions, certification of trained staff, procedural videos to facilitate refresher training and ongoing technical feedback and support. These strategies are even applicable to experienced sites, as small procedural differences can bias results. Trial runs are particularly useful and often the samples generated can be used in robustness validation tests at analytical test sites.
There are many suppliers of flow cytometry reagents and antibodies; some are of high quality, supplied with qualification data, and others are not. It is critical to conduct analysis on any new reagent or new production batch/lot of reagent to ensure sufficient quality and equivalent results between assay validation and sample analysis.
The design of a reagent qualification test is important, and often it is most suitable to use an alternative staining procedure or data analysis setup than you would in the final assay. To demonstrate the importance of reagent qualification, we purchased two lots of anti-CD4:RPE-Alexa Fluor 750 monoclonal antibody. These two lots, A and B, were prequalified for flow cytometry by the supplier, and we then performed reagent qualification using an aliquot of bulk purchased, quality-controlled PBMC sample. Using the analytical procedure designed for sample analysis, both lots were shown to be equivalent for the percentage proportion of CD4 positive cells within the lymphocyte population. However, using a simple reagent qualification data analysis template and single staining (e.g. not the full antibody cocktail used in sample analysis), fluorescence shift from the target FL5 channel into the RPE FL2 channel was detected in Lot B. This was due to breakdown of the tandem dye (Figure 3). Using an inappropriate reagent qualification test would have caused this unsuitable reagent to be used for sample analysis and would have provided a false positive signal in the RPE detection channel. This experiment highlights the importance of reagent qualification tests before new reagents are used on clinical samples, to prevent erroneous results from being incorporated into clinical studies.
Fit for Purpose Validation
The extent to which a flow cytometric assay needs to be validated depends on how the data generated is going to be used.2 The validation of an assay is a test that the analytical method precisely, reproducibly, robustly (if applicable), and accurately produces data that will be reported during sample analysis. The industry adheres to guidances from the regulatory authorities and industry whitepapers describing current best practices.2,3
Accuracy & Precision
Accuracy of a reported parameter is the relative error of an experimental measurement compared to the actual value of an analyte in a sample. It is typically not possible for PD biomarker flow cytometry assays to measure accuracy, as the true value of an analyte in a sample is not usually known. However, if sufficient precision can be shown, accuracy is usually not required. Factors that cause variation in precision include multiple operators, inter-assay variation, variations in sample handling procedures and the use of multiple flow cytometer instruments. It is typically not possible to determine the true upper limit of quantitation (ULOQ) and lower limit of quantitation (LLOQ), as QC material within a characterized range is not available (e.g. immunophenotyping assays). If reference material is available within an appropriate analyte range, the acceptance criteria for inter-assay precision of the assay typically should be ≤ 35% CV at the LLOQ and ULOQ QC levels, in comparison to 25% CV for samples within a reliable range. For rare event detection, low levels of biomarker expression, overlapping populations and, when imprecision is high, use of the Limit of Detection (LOD) to establish signal above background is important, especially if the LLOQ is not known.
Sample stability varies greatly depending on storage conditions, the nature of samples and whether the biomarker assay requires cellular functional activity. Cellular PD biomarker assay sample stability is assessed by investigation over the period for which samples are likely to await processing and longer term storage after collection at the clinical site. The period tested should incorporate realistic time intervals that actual samples may incur, and the data compared to promptly processed samples (baseline). The time interval can be communicated to clinical sites and monitored via worksheet to control collection procedures.
Specificity & Selectivity
Specificity and selectivity for measuring the intended cell population or the target compound should be investigated during the development phase. Robustness/ruggedness of the assay should also be investigated in order to measure the data during small but deliberate variations in assay procedures, i.e., incubation times, time interval for analysis after fixation, or collection of reducing numbers of events.
Most flow cytometry assays are quasi-quantitative, but they can be quantitative with proper calibration and QC material. For example, receptor saturation and modulation is a quantitative PD assay that relates to actual antibody molecules detected and therefore receptor levels, not relative fluorescent units (MFI). In this type of analysis, graded fluorescent beads are characterized for anti-Ig molecules to allow calculations of Antibody Binding Capacity (ABC), and the MFI of the beads with a known ABC are used to plot a calibration curve (Figure 4). By reading the fluorescence of the marker of interest, the calibration curve can quantify not only the number of receptor sites that are present on the cell surface, but the inferred receptor saturation and modulation with the appropriate saturated QC samples.
Data standardization is a critical component of flow cytometry analyses. It is important that data are standardized and do not rely on mean fluorescence units (MFI) if comparing data from different analytical runs. Standards can and should be developed for all flow cytometry assays including immunophenotyping (e.g., fluorescence minus one (FMO) controls and/or isotypes, rules for operator gating for samples with fluorescent shifts, batches of QC samples for use in analytical runs to measure in study variation), receptor occupancy (e.g. fluorescent calibration standards, non-competing antibodies for receptor and bound drug detection, internal cellular calibration QC samples) and protein expression (e.g., standardize fluorescence MESF and ABC or MFI ratio to a known negative population within the sample). Additionally, reagents, instrument and instrument settings are also factors to take into consideration when developing standards that can normalize for these factors.
There are a number of common issues that are encountered during the development, validation and sample analysis of flow cytometry assays. Such issues can be exacerbated when the project is outsourced to a service provider. Simple solutions using planning and comprehensive project design can help mitigate these risks. Sponsors of flow cytometry assays should ensure the availability of experienced trained flow cytometry users to quality check all data, adjusting gates when justified and needed. All new lots of reagents should be qualified before use.
Templates of data analysis protocols, gating hierarchy and histogram plots, including controls, should be fully documented before sample analysis. Other considerations include: supply of raw unanalyzed data files to the end user, data supplied is standardized to suitable controls, and real-time feedback is provided for sample quality at all phases of a project to provide in-study troubleshooting to clinical sites as required.
- Hale. G. et al., Removal of T cells from bone marrow for transplantation: a monoclonal anti-lymphocyte antibody that fixes human complement. (1983) Blood Vol. 62, No. 4 pp 873-882.
- Lee, L.W., et al., Fit-for-Purpose Method Development and Validation for Successful Biomarker Measurement, Pharmaceutical Research (2006), Volume 23, No. 2. DOI: 10.1007/s11095-005-9045-3.
- O’Hara, D.M., et al., Recommendations for the validation of flow cytometric testing during drug development: II assays, J. Immunol. Methods (2010), doi:10.1016/j.jim.2010.09.036.
Kevin Maskell is principal scientific director, Discovery and Development Solutions, at Merck Millipore in Oxford, UK. He can be reached at firstname.lastname@example.org.