1.1 Introduction
Flow cytometry (FCM), as indicated by its name, is a semi-automated method which combines two basic approaches: cytometry and flow.
Cytometry is, in essence, the measurement of cell characteristics. More broadly, it can be applied to various types of particles. In fact, the very first application of automated cytometry was invented by Wallace and Joseph Coulter with the first objective of counting paint particles [Reference Simson1]. Before that era, cell counts were performed manually under a microscope with specific calibrated slides called haemocytometers usually bearing the names of their inventors (Malassez, Thoma, Neubauer, Nageotte and others) [Reference Sandhaus2]. Such devices also allow the recognition of some cell types based on their size and granularity, detectable without any staining, by simple phase contrast.
Other cell properties can be examined with an optical bright-field microscope after preparing smears or cytospins where the cells are spread in a thin monolayer fixed on a slide. Such preparations are then stained, most frequently with May Grünwald Giemsa (MGG) or Wright stains [Reference Houwen3]. These panoptic stains contain eosin and methylene blue, plus azure for MGG, and thus make acidic components appear blue and basic components orange-red to violet. Smears, cytospins or cell suspensions can also be labelled with antibodies conjugated to fluorochromes and examined under UV light in specially equipped microscopes.
The flow component of FCM is a liquid sheath that allows conversion of a cell suspension in a narrow linear flux individualising cells or particles. The flow part of the instrument performs what is called hydrodynamic focusing, hydrofocusing, or single cell alignment. Hydrofocusing is achieved by injecting the cell suspension in the core of the sheath fluid at a slightly higher pressure and at a point where the channel becomes smaller. The acceleration of the fluids through this narrow channel and the different speeds of the sheath and the cell suspension result in cell alignment. Of note, there is no mixing of the cells/particle suspension with the sheath liquid. The latter can therefore very well be pure water, its main characteristic being to be devoid of any particle. The flux of cells/particles is guided in a specific device called a flow cell, through which a source of light will illuminate each cell as it passes in front of it.
The major advantages of FCM, compared to the methods briefly mentioned earlier, are that larger numbers of cells will be counted and that many parameters, including immunological characterisation of the cells, will be examined. Moreover, all results will be electronically stored, remaining available for analysis at any time after data acquisition.
1.2 Cell Counting in FCM
Haemocytometers allow the counting of cells in a well-defined chamber of 0.1 mm3 using unmanipulated suspensions (i.e. cerebrospinal fluid [CSF]) or diluted samples where red blood cells have been lysed. Data are then converted by calculation in the usual measurement units of events per mm3 or per litre. For stained cells, typically, between 100 and 500 cells are counted manually when performing cell differentials. Both these methods are prone to errors linked to the small number of counted events, and thus lack precision. This has been well established by Rümke et al., who designed a table displaying the decreasing level of incertitude associated with larger numbers of events counted [Reference Rümke, Bezemer and Kuik4].
Flow cytometry, which examines several thousand events in a few minutes, provides a high level of sensitivity and exactitude. The relative numbers (proportions) of each cell subset acquired will therefore be accurate. However, for exact counting, flow cytometers require the use of standardised bead suspensions with a known number of beads per microlitre mixed volume/volume together with the studied suspension of cells/particles. When this known number of beads has been recorded by the instrument, it can therefore be concluded that 1 μL of the sample has been examined. This can then be extrapolated to the other cells recorded during the same time. Another possibility is to use calibrated volumetric systems built in the flow cell. Because of these large numbers, it also allows the identification of minute populations, likely to be missed with smaller counts.
1.3 FCM and Light
Besides counting particles, FCM allows an appreciation of their physical, chemical or biological properties. The major physical properties of particles/cells exploited by FCM are their ability to diffract, reflect and refract the coherent monochromatic light of a laser beam. Different types of lasers are available. Flow cytometers were initially equipped with gas lasers using argon, krypton, helium-neon or helium-cadmium, some requiring a cooling system. Solid-state lasers are crystals (ruby, yttrium aluminium garnet (YAG)) or ions such as titanium or chromium. More recent instruments use laser diodes based on semiconductors and similar to light-emitting diodes (LEDs) [Reference Telford5].
When the narrow and focused coherent light of the laser encounters a cell/particle (a so-called event), its diffraction intensity is proportional to the size of the event. Flow cytometers are equipped with photodiodes collecting the light diffracted by the cell in the path of the laser beam or forward scatter (FSC). The instruments are also equipped with a device (mask) blocking the laser light from the FSC photodiode when no particle crosses it. The beam is widened proportionally to the size of the cell as one enters the laser’s path, and the FSC photodiode can collect and transform it into an electronic signal proportional to the size of the cell/particle (Figure 1.1).
Concomitantly, a second detector (photomultiplier tube (PMT) or, more recently, avalanche diode for better resolution) collects the light reflected by the surface of the cell/particle as well as by any surface inside it (i.e. organelles, vesicles, granules etc.) at a defined angle, lateral to the path of the laser beam. The intensity of this side scatter (SSC) signal will thus be proportional to the granularity of the cell. Typically, in a blood sample, the small erythrocytes with no nuclei will provide very small SSC signals while those generated by the larger and more complex granulocytes will be more intense. The pattern of scatter signals will differ slightly between instruments, depending on the angle of the SSC detector and the number of display channels (see Section 1.5).
The voltage applied to the detector will also modify the intensity of light collected. It must be adapted to the type of cells investigated: that is, it will have to be higher to see small particles such as platelets and lower to see larger cells such as granulocytes.
Chemical parameters can also be measured by flow cytometers, typically based on the properties of fluorochromes. The latter are chemical substances able to absorb light at a defined wavelength and re-emit it at a higher and defined wavelength. This is based on the fact that, in these molecules, absorption of a photon will result in a modification of electrons’ energy state, moving from a ground state to an excited state. When electrons return to their ground state, they partially restore the energy by going through transition stages resulting in the emission of a quantum of light with a lower energy and thus higher wavelength than the excitation light [Reference Noomnarm and Clegg6]. Basically, in FCM, lasers provide excitation light and fluorochromes emission light. To collect emitted light from each fluorochrome, flow cytometers are equipped with dichroic mirrors and bandpass filters before each signal is registered by a dedicated PMT. Dichroic mirrors reflect light at a specific wavelength while letting all other light pass through. They are positioned at an angle from the emission source so that reflected beams make a 90° angle to the mirror and get directed towards the relevant PMT. Just before PMTs, filters of a specific wavelength will narrow the beam of light collected, ideally at the level of peak fluorescence.
The development of multiparameter FCM (MFC) has led to the creation of flow cytometers with multiple lasers, thus broadening the possibilities of staining by using fluorochromes excitable at different wavelengths. Although the basic FSC and SSC parameters are usually measured on the 488 nm blue laser, separate pathways are then used to channel the emitted lights generated by the different lasers [Reference Perfetto, Ambrozak, Nguyen, Chattopadhyay and Roederer7].
1.4 Fluorochromes
Fluorochromes (Table 1.1), also called fluorophores, are aromatic polycyclic carbohydrates which can be found naturally in some algae (i.e. phycoerythrin or rhodamine) or organic synthetised compounds (i.e. fluorescein isothiocyanate) [Reference Chattopadhyay, Hogerkorp and Roederer8]. Some proteins are also fluorescent, such as the green fluorescent protein, which will stain cells transfected with its gene associated with another gene of interest [Reference Cranfill, Sell and Baird9]. Some fluorochromes are used individually, such as propidium iodide (PI), which will fluoresce in red once intercalated in the hydrophobic environment of deoxyribonucleic acid (DNA) and is widely used together with either annexinV to study cell death and apoptosis [Reference Atale, Gupta, Yadav and Rani10]; acridine orange allowing quantification of both RNA and DNA content in two colours [Reference Darzynkiewicz11]; or thiazole orange, which will stain both DNA and RNA and is widely used for the analysis of reticulocytes and platelets [Reference Nygren, Svanvik and Kubista12, Reference Rapi, Ermini and Bartolini13]. Staining nuclear DNA also allows for the use of whole blood or whole bone marrow by selecting the nucleated cells and ignoring red blood cells in the gating strategy [Reference Violidaki, Axler and Jafari14].
Fluorochromes are, however, mostly used conjugated to monoclonal antibodies, allowing visualisation of the structures specifically recognised by the latter. Most of such conjugates (monoclonal antibody/fluorochrome combination) use single fluorochromes.
In some combinations, the properties of two different compounds are used in ‘tandem’ fluorochromes. This allows detection of emitted light from a fluorochrome which cannot be excited by the available laser. The principle of such tandems is to use the light emitted by the first molecule to excite the second fluorochrome tightly bound to it and then measure the second emitted fluorescence. The principle used is also called fluorescence resonance energy transfer (FRET) [Reference Szöllosi, Damjanovich and Mátyus15].
A general very important property of fluorochromes is their sensitivity to light. For this reason, fluorochrome conjugates are provided in light-proof dark vials. Great care must be taken while manipulating these reagents to avoid exposure to direct light at all stages of the experiment. Some fluorochromes are also sensitive to pH yet stable in the neutral conditions associated with most biological applications of MFC [Reference Sugden16]. Finally, tandems can be unstable and get degraded. In this case, the fluorescence collected with them will thus erroneously be that of the first fluorochrome [Reference Hulspas, Dombkowski and Preffer17].
Among the numerous fluorochromes available, the major properties to examine are the wavelengths of their excitation and emission lights (Table 1.1), assuring that they are adapted to the parameters of the flow cytometer to be used. It is also important to know their level of brightness to choose a conjugate best adapted to the level of expression of the molecule to be stained.
1.5 Signal Acquisition
The signals sent to PMTs are very low and need to be amplified, hence the name of these collectors, which increase the signal provided by collected photons. Amplification is performed by increasing the tension applied from 100 to 1,000 V [Reference Bristow, Bundy and Wright18]. Linear amplification can also be obtained by applying a gain value, usually for FSC and SSC. Each signal is transformed electronically by analogic/digital converters (ADC), which generate binary signals of 0 or 1. Typically, no or a low signal is 0 and a higher signal is 1. The sensitivity of the signals collected can be improved by increasing the number of bits, which are the combinations of 0 and 1 that are used to partition the range of voltage of the signal. For instance, a signal between 0 and 10 V can be dichotomised as 0 between 0 and 5 V and 1 between 5 and 10 V in a 1-bit digitalisation. Three-bit digitalisation (i.e. all permutations between 000 and 111) will divide this voltage range in eight levels (23). Typically, flow cytometers use either 14- or 20-bit combinations, which yield 16,384 or 1,048,576 channels, respectively. The 20-bit combination provides a better discrimination for weak signals. Indeed, since the graphic representation of these signals generally uses logarithmic scales, on a four-decade scale the 14-bit configuration will provide ~16 channels in the first decade and ~900 channels in the second (Figure 1.2). The difference will be minimal in the high voltages but will be visible in the low values. More recent instruments, however, can display data on seven decades with up to 24 bits.
These technical considerations may impact the settings of the instruments. When lower numbers of bits are used, the dynamic range of the logarithmic scale can be limited, and the choice of PMT values can impact the resolution of the signals at low intensities.
1.6 Signal Displays
Several software packages have been developed to analyse the signals provided by flow cytometers, some being incorporated in the instrument and used during acquisition and others (or portable versions of the same) used to perform analyses at distance [Reference Chi19].
The simplest way to analyse FCM signals is a monoparametric histogram where the signal intensity is displayed in the abscissa and the number of events in the ordinate. Because most applications of FCM are to biological systems, this usually results in Gaussian peaks, the breadth of which depends on the variability of expression of the marker in the population investigated. For fluorescent beads used to check the proper alignment of the laser, this peak should be very narrow, especially for bright fluorescence, and reflect the sensitivity of the instruments at low fluorescence (see Section 1.5 and Figure 1.3). In a properly aligned instrument, the coefficient of variation of each peak (standard error divided by the mean) should be as low as possible, independent of the apparent width of the peak, which will logically appear larger at low fluorescence values.
Biparametric histograms, also called scattergrams or dot plots, are used to display events based on two of their properties. They will thus appear as dots (each event resulting in a dot at the intersection between its X and Y values) forming one or more clusters depending on the populations present in the sample. The most basic biparametric histogram to check for proper selection of FSC and SSC settings (Figure 1.4a) uses these two parameters, most frequently on linear scales. As mentioned earlier, this is also used by other instruments such as cell counters and does not depend on fluorescence.
Figure 1.4 Flow cytometry displays: (a) biparametric plot of FSC and SSC, i.e. light diffraction without considering any fluorescence; leukocyte subsets are easily distinguishable and debris can be gated out from gate A; (b) biparametric plot of FL1 against SSC allowing for an easy distinction of positive subsets, here CD64+ monocytes (blue) and a subset of activated neutrophils (lighter blue), by comparison to unstained other blood cell subsets and (c) relationship between two fluorochromes, here CD45 and CD19 in a sample containing B-blasts, T and NK lymphocytes (CD45bright/CD19-), B-lymphocytes and other leucocytes (essentially neutrophils).
Displaying each fluorescence against SSC (Figure 1.4b) allows for a good appreciation of autofluorescence versus specific signals in each population. With such a display for the analysis of examined samples (such as peripheral blood [PB], bone marrow [BM] or other cell suspensions), unstained cells provide an excellent internal control.
Biparametric histograms are also used to examine the relationship between two different fluorochromes (Figure 1.4c). The results can become extremely pertinent to small subsets if a successive selection of relevant subsets is performed by drawing gates. Such gates can be coloured/painted, and the use of a homogeneous colour code on a single platform greatly facilitates data interpretation.
In the example provided in Figure 1.5, the specific staining of leucocytes by a monoclonal antibody directed to CD45 is displayed (see also Chapter 3). Gating strategies have been used in this example to colour granulocytes in red, monocytes in green and lymphocytes in magenta. These subsets have then been backgated on the CD45/SSC scattergram. Selecting only the lymphocyte gate, and a biparametric histogram based on CD3 and CD19 expression, the major lymphocyte subsets are displayed. Gating on CD3+ lymphocytes and displaying CD4 and CD8 allow for analysis and visualisation of the main T-cell subsets.
Figure 1.5 (a) Back-gating of coloured leucocyte subsets on a CD45/SSC ‘cartography’ of BM; (b) gating on lymphocytes can distinguish CD19+ and CD3+ lymphocytes; (c) further gating on the CD3+ population allows the display of T-cell subsets CD4 and CD8; (d) density plot representation of the same plot as in (a); and (e) contour representation of the same plot as in (a).
The gating hierarchy must be kept in mind when performing complex analyses. Most software allows for a visualisation of this hierarchy and labels histograms based on the parameters used. Although software will basically provide letters to identify gates, it is advised to design and save protocols where subsets are more precisely named.
Counts and percentages (called gate statistics), based on the chosen reference population, can be obtained by various means. For monoparametric histograms, integration cursors can be placed encompassing the Gaussian peak of interest. For bi- or multiparameter histograms, gates of different shapes (squares, rectangles, circles, polygons, freehand …) will provide their X and Y coordinates as well as the number and percentage of events they contain. For well-separated subsets, quadrants can be used, dividing each histogram into four regions the size and shape of which can be adapted to best delineate the various subsets visualised (Figure 1.5b and c).
It is important to consider populations and subpopulations as Gaussian clusters and not to divide them either by too short integration cursors or quadrants bisecting clusters. A frequent mistake is to rely too strictly on controls such as irrelevant isotype antibodies stained with the same fluorochrome and consider the brightest signal provided by the Gaussian of such a negative control as the beginning of significant fluorescence. Although this may be the case for brightly stained subsets, often the peak of the population of interest presents with a shift overlapping the ‘negative’ peak [Reference Maecker and Trotter20].
In the examples of biparametric histograms shown so far, individual dots are only seen on the periphery of coloured clusters. A better idea of the density of events in such clusters can be provided by monoparametric histograms but also by variations of biparametric representations such as density or contour plots (Figure 1.5d and e). This can prove very useful to properly delineate cell subsets. The number of cells displayed can also be chosen or the resolution of the plots modified.
Some software packages also provide various approaches of multiparameter representations in spaces with more than one dimension. This can be modelled by mathematical calculation of principal component analysis or drawn by moving the length and angle of the various vectors to best individualise populations of interest (see also Chapter 14).
Finally, for samples assayed with the same antibody combinations, the merge function of some software allows for a direct comparison of different conditions (i.e. diagnosis and relapse) or for the concomitant analysis of larger numbers of events. For example, merging six different samples of normal BM will provide a good idea of what normal is by smoothing out individual variations (Figure 1.6). An important parameter for such displays is the time of acquisition, which must be systematically recorded. This allows, during merge analysis, to always individualise each sample.
Figure 1.6 Example of the merge of six PB samples. The bottom right histogram shows the merged samples according to a ‘time’ abscissa that allows to discriminate each sample as a single column. All cells in this histogram (i.e. merged samples) have been used then to examine the florescence of each of the markers in this common tube using SSC as ordinate. The top left histogram allows for eliminating debris on an FSC/SSC display. The second top histogram is the ‘cartography’ of CD45/SSC display with backgating of neutrophils in red, monocytes in green and lymphocytes in magenta. The third histogram shows the strong positivity of neutrophils for CD65 and intermediate labelling of monocytes. The fourth histogram shows the strong positivity of monocytes for CD14. On the second row, histograms show CD13-positive monocytes and neutrophils, strongly CD33-positive monocytes with lower staining of neutrophils and, in the last two histograms, absence of immature myeloid cells expressing CD34 or CD117, respectively. The three bottom left histograms show CD7-positive lymphocytes, CD11b-positive monocytes and neutrophils, CD16bright neutrophils and finally CD16intermediate NK-lymphocytes.
1.7 Compensation
Emission spectra of fluorochromes appear as semi-Gaussian curves with a maximal peak that dictates the choice of the bandpass filters placed before the PMTs (Figure 1.7a). Yet, there is frequent overlap of these emission spectra, and it may be necessary to eliminate the contaminating signal [Reference Tung, Parks, Moore, Herzenberg and Herzenberg21]. This is achieved mathematically by subtracting a percentage of the overlapping signal(s) in each fluorescence channel (Figure 1.7b).
Figure 1.7 (a) spectral overlap of four commonly used fluorochromes, ordered by emission wavelength (abscissa and spectrum below); (b) strategy for compensation calculations, indicating the percentage of overspill of FITC in higher wavelengths; (c) distortion or spreading or ‘trumpet effect’: compensations are perfect with 50% of the spread on both sides of the negative signal for the fluorescence on the ordinate.
Basically, compensations will be calculated by performing single staining of cells or beads with each of the antibodies intended to be used in a panel. Each preparation will then be examined in each possible emission fluorescence channel, and the compensation percentage adjusted to remove overlapping signals. Currently, beads coated with antibodies to mouse immunoglobulins allow compensation of settings for any antibody/fluorochrome combination [Reference Maecker and Trotter20]. If different conjugates are obtained from the same manufacturer, using the same fluorochromes, a compensation matrix adequate for different panels can be used. The beads stained with the antibodies are mixed with uncoated beads, which will provide the baseline signal. Each biparametric histogram combining the fluorescence tested versus each other channel will be examined to provide for the absent fluorochromes the same signal as that of the negative beads. It is also possible to perform or check for proper compensation by using the ‘fluorescence minus one’ or FMO method. This technique consists of testing a relevant sample in a series of tubes where, for each, one of the markers intended to be used in the panel is missing. This allows us to visualise/check the overspill or spreading of other emitted fluorescences in the fluorescence of interest where it is known that no antibody has been added [Reference Tung, Heydari and Tirouvanziam22].
When checking for proper compensation, or building a compensation matrix with cells, it is wise to use a biexponential or logical representation of low signals, since some staining will lead to an unavoidable spreading or trumpet effect (Figure 1.7c). This means that light from labelled cells will spill over in a different channel with a broader yet symmetrical display, which is also called signal distortion or spreading [Reference Roederer23–Reference Nguyen, Perfetto, Mahnke, Chattopadhyay and Roederer25].
With the development of MFC, interferences are complex. Most software provides a ‘wizard tool’ that automatically performs compensations using the acquisition file of each single labelling.
1.8 FCM Settings
For any given experiment, the parameters of the instruments must be defined beforehand. The PMTs must be adjusted in such a way that the signals recorded are well defined yet do not saturate in the brightest decades.
A good way of defining this is to use unstained lysed blood and adjust all PMTs so that more than 85% of the cells can be seen in the first decade, i.e. above the first channel. Such adjustments can then be used to identify the fluorescence channels where a specific batch of beads, fluorescing in all channels, will appear. This strategy, adopted in the Harmonemia initiative, allows for excellent reproducibility between instruments, new channels being recorded at each change of beads batch [Reference Lacombe, Bernal and Bloxham26].
Lasers’ alignment with the flow cell must be checked regularly, although current instruments are usually extremely stable. This is performed by recording the fluorescence emitted by specific beads, with target limits of coefficients of variation of the peaks obtained [Reference Tangri, Vall and Kaplan27]. Such controls belong to the daily assessment of laser alignment and result in the generation of traceable data, most frequently in the form of Levy-Jennings graphs. The mandatory pre-analytical precautions to ensure that proper data are generated have been extensively described by an expert working group [Reference Tangri, Vall and Kaplan27].
Depending on the number of cells that have to be analysed and the precision of the detection, the pressure of the sample sheath fluid can be adjusted between low for the more precise measurements and intermediate or high for less sensitive signals.
The tubing of the instrument must be kept extremely clean, and regular cleansing cycles with chlorine and several tubes of distilled water are recommended on top of the automatic cleanses performed by most machines between samples.
1.9 Panel Design
MFC presents the great advantage of being able to test numerous characteristics of a suspension at the same time. Besides FSC and SSC, 8 and 10 colours flow cytometers provide up to 10 or 12 parameters, or 14 and 15 for more recent 12 or 13 colours instruments and even more in certain combinations of lasers allowing for more than 20 concomitant fluorescences [Reference Heubeck, Savage and Henderson28]. This has proven to be extremely valuable in immunophenotyping for the diagnosis of haematological malignancies, especially when only small samples such as fine-needle aspirates (FNA) or CSF are available.
Two types of considerations can be taken into account when designing a panel [Reference Nguyen, Perfetto, Mahnke, Chattopadhyay and Roederer25, Reference Heubeck, Savage and Henderson28]. One is to use a variety of markers allowing characterisation of a maximum of different subsets, such as markers for granulocytes, monocytes, lymphocytes and progenitors. The other option is to investigate the co-expression of several markers on the same cells. This is applied if the goal is to define the maturation stage or activation status of a given subset. In that case, it is important to choose wisely the fluorochromes associated with the various potentially co-expressed monoclonal antibodies used.
In all cases, bright fluorochromes must be preferred for antigens expressed at low levels, and dimmer fluorochromes for densely expressed antigens. When co-expression is expected, the overlap of fluorochromes must be accounted for to avoid the necessity of too much (or impossible) compensation. Several rules apply for this case. The first channel of the blue laser (488 nm excitation), generally used for FITC conjugates, is never impacted by other fluorochromes. The same is true for the first channel of the violet laser (405 nm excitation) used for Pacific Blue® or Brilliant® Violet 421. These two channels can be used for weakly expressed antigens, since no compensation needs to be applied. The second channel of the violet laser conversely never overlaps other channels. It can thus be used for bright markers and/or ‘parent’ markers that will contain subsets. This is, for instance, the case for the pan-leukocyte CD45. In the most recent instruments, addition of an ultraviolet laser allows for even more flexibility, since signals emitted from excitation at these low wavelengths do not interfere with any others. The degree of spillover will, however, vary between fluorochrome combinations which should also guide their choice. ‘Parent’ and ‘children’ relationships between markers characterise the fact that all ‘children’ are also stained by the ‘parent’ marker. Spillover can then be accepted from a ‘child’ to the ‘parent’, since the primary staining will prevail. Conversely, ‘parents’ should have no or low spray on ‘children’ to avoid erroneous appreciation or fuzzy images [Reference Lugli, Roederer and Cossarizza29]. It is also important to know that co-expressed markers can lead to dot plots with an angled shape, suggesting poor compensation, although it is just due to the dual staining.
Compensations are usually easily done for lights emitted after excitation with blue and violet lasers. They are more complicated for excitation with the red laser, and mutually exclusive markers can be preferred for these channels, where spillover will not interfere.
1.10 Sample Handling
Flow cytometry requires cells in suspension, implying that PB or BM are collected on anticoagulated tubes [Reference Davis, Dasgupta, Kussick, Han and Estrellado30, 31]. Depending on local habits and on other tests liable to be performed on the same sample, ethylene diamine tetraacetic acid (EDTA) or heparin can be used. It is also recommended to perform MFC analyses as rapidly as possible after collection. For fragile samples such as CSF, it may be interesting to add a specific preservative solution to collect more accurate information and avoid cell loss in the inhospitable low-protein content of this liquid [Reference de Jongste, Kraan and van den Broek32]. Because staining and acquisition are relatively rapid, this allows quick answers to the clinicians, nearly at the same time as morphologic analyses. When transport is needed, the time to processing for samples other than CSF should not exceed 72 h [Reference Tangri, Vall and Kaplan27]. Of note, the characterisation of leukaemic cells in a heavily infiltrated sample usually will not be modified by some delay, possibly leading to a degree of apoptosis. Conversely, the search for minute subsets in the context of minimal residual disease can be impacted by extended delays. Also, some tumour cells such as high-grade B-cell lymphoma or plasma cells in myeloma may be more prone to apoptosis.
Another pre-analytical aspect of possible impact is the haemodilution of BM samples [Reference Lugli, Roederer and Cossarizza29, Reference Loken, Chu, Fritschle, Kalnoski and Wells33]. It is important to remember that this will not impact the characterisation of malignant cells which can be isolated by gating strategies. However, the great quality of MFC to be able to provide exact counts is lost in haemodiluted samples. It therefore cannot be used to make a diagnosis of leukaemia or myelodysplasia where blast percentages impact the result. However, this is well used to enumerate mobilised CD34+ progenitors in PB before stem-cell transplantation [Reference Schmidt-Lucke, Fichtlscherer and Aicher34].
The choice of the panel of antibodies to test will depend on clinical information provided when sending the sample. Screening or disease-specific panels (described in detail in other chapters) are used. Those applied daily can be prepared in advance (i.e., weekly on Monday) but have to be carefully stored in the dark and validated for use on several consecutive days. It is also important that each antibody is titrated to use the most adequate concentration for saturation yet limiting non-specific labelling (see next paragraph). This is also true when antibody cocktails are prepared extemporaneously. Great care must be taken in these preparations in terms of exposure to light, meticulous pipetting (changing pipetting cones between each antibody to avoid contamination) and strict control of the sequential addition of antibodies, not to forget or duplicate any given marker. Interesting alternatives have appeared which provide customised or generic ready-to-use antibody combinations [Reference Hedley, Keeney, Popma and Chin-Yee35, Reference Chan, Kotner, Chuang and Gaur36]. The latter can be premixed cocktails allowing for a single pipetting step, or individual MFC tubes containing lyophilised or dried antibodies. Thorough vortexing of the sample and the antibody mixture must be performed before incubation in the dark for 10–15 min at room temperature (20°C). Incubation at 4°C avoids the undesirable patching and capping of the antibodies that may lead to erroneous labelling negativity but requires a 15–20 min incubation.
Guidelines concerning sample handling for several haematological disorders have been published [Reference Sędek, Flores-Montero and van der Sluijs37–Reference Soh and Wallace39].
For most applications, whole blood or BM is the sample of choice. They can be used directly by mixing an aliquot of 50 to 100 μL of sample with the antibody mixture. Various lysis reagents can be applied after the incubation time. Home-made ammonium chloride (NH4Cl) has long been used, but there are many different versions of this lysis that may not lead to highly reproducible data. Very stable lysis solutions are available commercially, some of them being highly selective in interfering only with the physiology of red cells (Table 1.2). After lysis, the sample can be directly processed for acquisition. Such lysis–no wash methods allow to examine the entire cell population of the sample. To minimise non-specific staining, it is recommended to use smaller amounts of antibodies, determined after titrating the reagents as mentioned above. Otherwise, a washing step may follow the lysis, at the risk of losing some cells [Reference Tangri, Vall and Kaplan27]. Only when investigating for surface immunoglobulins, a pre-incubation wash is mandatory to eliminate plasma antibodies [Reference Tangri, Vall and Kaplan27]. Washing steps are also necessary when permeabilisation is required for the investigation of intracytoplasmic or nuclear antigens. Finally, if a search for a small population is required, bulk lysis of a larger sample can be performed prior to incubation, followed by centrifugation to increase the cell concentration. It must, however, be remembered that each washing step induces random cell losses and can change the proportions of cell subsets.
Some lysing preparations may contain fixatives that will stabilise the staining by modifying the plasticity of plasmatic membranes. This is interesting when performing large batches of staining in that it allows a delayed acquisition. It is, however, mandatory, even after fixation, to prevent exposure of the samples to light, since the fluorochromes retain their sensitivity.
Of note, all these pipetting, staining and eventual washing steps are more and more allotted to sophisticated instruments including artificial intelligence software checking all the sensitive steps [Reference Al-Attar, Kumar, Untersee, O’Driscoll, Ventura and Lin40]. In a very near future, flow cytometrists are likely going to dedicate their time in critical data analyses from well-established panels, or in designing even more comprehensive panels liable to provide extensive information.
Another point requires some attention, regarding accreditation procedures and, in Europe especially, the IVDR (in vitro diagnosis regulation). Flow cytometry assays, owing to the inherent incredible diversity they allow, are rightly considered laboratory-developed tests or LDTs. Yet, although the flexibility of flow cytometrists facing complicated and unusual cases must be preserved, it is important to apply consensual documented strategies verifying quality criteria [Reference Lagoo41].
1.11 Post-Analytical Procedures
As mentioned earlier, MFC experiments generate acquisition files which can be analysed or re-analysed at any time. It is good practice to establish analysis matrices (protocols), which will easily provide all the information needed in any specific condition in a reproducible fashion. As also mentioned, the use of systematic colour codes makes interpretation easier.
For immunophenotypic patterns, especially in leukaemia, the pathological clone must be identified at best on scatter properties and CD45 expression (see also Chapter 5). This will usually be enough for acute leukaemia or myelodysplasia where the population involved is that of progenitors. The latter can be defined using a Boolean equation excluding from analysis of the whole sample the mature granulocytes, monocytes and lymphocytes. This can be helped by a combination of CD11a, CD16 and CD14 that will stain the first two subsets. Lymphocytes are characterised by their low SSC and bright CD45 expression and are usually easy to gate, especially when using a density plot display [Reference Arnoulet, Béné and Durrieu42]. For lymphoproliferative disorders of B-lineage, a CD19 gate can be useful, better established on the whole sample not to miss proliferations with a high SSC such as hairy cells or diffuse large B-cell lymphomas. A lymphocyte gate will be preferable for T-lineage proliferations, since not all express surface CD3.
Once the malignant population is isolated and enumerated, percentages should not be used afterwards to describe the immunophenotype, except for clear partial expressions.
The report should mention the type of sample analysed, the specificities tested and those that yielded a relevant positive or negative signal contributing to the interpretation, together with a conclusion on the likely diagnosis. Some recommendations can be found in the literature regarding suggested report formats [Reference Del Vecchio, Brando and Lanza43, Reference Hrušák, Basso and Ratei44]. In any case, an integrated report, at least combining morphology and MFC and further incorporating cytogenetics and molecular findings when available, is recommended. It is also important to mention that unsupervised methods which allow for automatic detection and quantitation of cells clusters of interest, together with comparison to normal counterparts, are of growing interest. This approach will be discussed in depth in Chapter 14.
1.12 Conclusion
Many of the principles and technologies described in this chapter are embedded in the instruments and monitored without necessarily knowing what happens technically. However, a good knowledge of them can prove useful for troubleshooting when unexpected data are obtained. The most sensitive part is probably the stability of fluorochromes, and great care must be taken when manipulating labelled reagents to preserve their quality. At the end of the day, what will in fact matter for haematological malignancies is the result obtained and provided to the clinicians in charge of the patient’s management. Flow cytometry results will also be important to guide further analyses as stated above. Production of an integrated conclusive report following the WHO classification is certainly an aim to target.
