Components of DNA Histogram
In FCM, the cell cycle is analyzed as a frequency histogram where the fluorescence
intensity of DNA binding dye and the number of cells is plotted across the X-axis
and Y-axis, respectively.[4] In a cell cycle histogram, the fluorescence intensity of DNA binding dyes is expressed
in a linear scale. The linear scale simplifies interpretation of DNA histogram as
each unit in this scale represents an equal interval in increments of fluorescence
(i.e., DNA content per cell). As the data is well spread in a linear scale, individual
phases of cell cycle can be grossly identified by visual inspection itself. Also,
the linear scale facilitates identification of artefacts and sub-G0/G1 peaks, which
might be readily missed in a log-scale.
Due to the stoichiometric binding of DNA-specific dyes, all cells in any given phase
of the cell cycle are expected to have the same amount of DNA and the “expected” cell
cycle must be as the green, blue, and red colored bars depicted in [Fig. 1]. However, due to cell-to-cell variability in dye binding, a gaussian curve with
normal distribution is generated for each phase of the cell cycle, and a multimodal
histogram is generated as depicted in [Fig. 1].
Fig. 1 Flowcytometric DNA assay depicted as a multimodal histogram (shaded curves). The green, blue, and red colored lines represent the theoretical G0/G1, S, and G2/M phases, respectively. Y-axis: number
of cells. X-axis: intensity of staining with FxCycle violet, a DNA-specific fluorescent
dye.
Following are the components of a DNA histogram. The G1/G0 curve represents cells in G0/G1 phases of the cell cycle. Using DNA-specific dyes, a distinction
between G0- and G1-phase cells is not feasible as both G0- and G1-phase cells have
the same DNA content. In contrast to G0 cells, which have only DNA, cells in the G1
phase also have RNA (refer [Table 1]). Combined use of RNA binding dyes like pyronin Y along with a DNA-specific dye
(having a different emission spectrum) can aid distinction between G0- and G1-phase
cells. The G2/M curve represents cells in the G2/M phase of cell cycle. In theory, the G2/M-phase cells
must have exactly double the DNA content of cells in the G0/G1 phase. However, the
extensive DNA–protein interactions in the G2/M-phase cells result in extensive chromatin
compaction, which hinders stoichiometric nucleic acid binding by the dyes. Due to
this, the dye fluorescence intensity of cells in the G2/M phase is 1.97 times that
of the cells in the G0/G1 phase.[7] The S curve is a broad-based histogram between G0/G1 and G2/M curves. This curve reflects the
gradual increase in the DNA content of a cell until it reaches a content equivalent
to a cell in the G2/M phase.
In a DNA histogram, the “origin point” indicates the point on x-axis where the fluoresce
of the DNA binding dye starts. In a typical DNA histogram, the peak staring from this
origin point has the least DNA content, that is, the G0/G1 peak. This origin point
on the x-axis must be clearly defined in the DNA histogram as it facilitates data
normalization for comparing DNA histograms from different experiments or samples or
different cell populations within the same sample. The origin point is also the end
result of variations in sample age, sample processing, and measurement techniques.
The exact start and end points of G1/G0, S, and G2/M histograms are always arbitrary
by manual gating. To identify the correct proportion of cells representing each phase
of a cell cycle, mathematical modeling-based curve fitting algorithms are available
(Dean–Jett model of 1974, Fox-modified Dean–Jett model of 1980, Watson's pragmatic
model of 1987, etc.).
Clinically Relevant Parameters Derived from FCM DNA analysis
In patients diagnosed with B-lineage acute lymphoblastic leukemia (B-ALL) and plasma
cell myeloma (PCM), the DNA index (DI) and S-phase fraction (SPF) are the two parameters
of clinical relevance obtained by FCM DNA analysis.
DNA Index
The DI is calculated as the ratio of the G0/G1 phase of malignant cells to that of
the G0/G1 phase of normal diploid control cells.[4] There is near 90% correlation between the ploidy status assessed by DI and cytogenetics
(CG).[8]
[9]
[10] There are published cutoffs (refer to [Supplementary Table S2], available in the online version) that have correlated the range of DI with the
corresponding number of chromosomes in a population of cells.[8]
[11]
[12]
[13] The advantages of ploidy assessment by DI are that the technique is more rapid and
sensitive than CG, and can aid in ploidy identification even in samples with less
than 1% cells of interest.[8] This technique can identify cell population–specific ploidy even in samples with
markedly heterogeneous cellular composition (e.g., bone marrow). However, ploidy evaluation
by FCM is not sensitive in assessing aneuploidies involving smaller chromosomes.[9]
Role of DNA Index in B-Lineage Acute Lymphoblastic Leukemias
Hyperdiploidy (51–65 chromosomes) is observed in nearly 30% of pediatric patients
diagnosed with B-ALL and is associated with a favorable prognosis.[14]
[15]
[16] These hyperdiploid blasts are highly susceptible to apoptosis and hence require
stringent conditions for their ex vivo survival. This is reflected in frequent metaphase
failures and/or poor morphology of chromosomes during conventional CG evaluation of
these patients. Under ambient culture conditions used for conventional karyotyping,
the normal diploid cells in the sample can overwhelm the hyperdiploid blasts and result
in pseudodiploidy. Ploidy assessment by DI is effective in overcoming these limitations
of conventional CG-based ploidy assessment.
The relevance of DI in pediatric patients with B-ALL was observed as early as 1985.[17] A DI of ≥1.16 in these patients is consistently associated with hyperdiploidy and
hence better outcomes.[11]
[13]
[17]
[18]
[19]
[20]
[21] Although there is more than 95% concordance in hyperdiploidy identified by DI and
karyotyping, not all B lymphoblasts with 51 to 65 chromosomes have DI ≥1.16.[8]
[9]
[22] In this context, it has been proposed to consider B lymphoblasts having 51 to 55
chromosomes and DI of 1.10 to 1.15 as “low DI-high hyperdiploid” (LDI-HHD). These
patients with LDI-HHD have a poor prognosis as compared to patients with DI ≥1.16
and 51 to 65 chromosomes, that is, the “high DI-high hyperdiploid” category.[23] In fact, patients with DI ≥1.24 (corresponding ≥58 modal chromosomes) have the best
prognosis.[23]
[Fig. 2] demonstrates the FCM DNA analysis histograms of individual patients with diploid
([Fig. 2a–c]) and hyperdiploid ([Fig. 2d–f]) B lymphoblasts.
Fig. 2 DNA index assessment in B-lymphoblastic acute lymphoblastic leukemia using FxCycle
violet, a DNA-specific dye. (a–c) From a sample with diploid blasts. (d–f) From a sample with high hyperdiploid blasts. Overlay histograms (c,f) show the G0-/G1-phase cells of both blasts and lymphocytes. The fluorescence intensity
of FxCycle violet is depicted as Geometric mean (GMean) of expression.
Hypodiploidy (<46 chromosomes) is observed in less than 5% of pediatric B-ALL patients.[15] In this category, patients with 31 to 39 chromosomes (low hypodiploid) and 24 to
29 chromosomes (near haploid) have dismal prognosis.[15]
[24]
[25]
Endoreduplication is a phenomenon where a cell undergoes DNA replication without undergoing
cell division.[24] This results in the cell having multiple copies of its chromosomes within a single
nucleus. Although endoreduplication is a physiological process in certain human cells
(e.g., megakaryocytes), its occurrence in blasts is pathological and complicates ploidy
assessment.[24] Nearly 65% of patients with near haploidy and 45% of patients with low hypodiploidy
undergo endoreduplication.[24] Due to this phenomenon, CG can misinterpret endoreduplicated near-haploid and low-hypodiploid
lymphoblasts as hyperdiploid (50–60 chromosomes) and near-triploid (up to 78 chromosomes)
lymphoblasts, respectively.[24] Recently, both low-hypodiploid and near-triploid blasts have been recognized as
a unified genetic entity involving TP53 mutations.[24]
[25]
[26] Hence, misinterpretation of low hypodiploidy as near triploidy might not affect
the patient's prognosis. However, misclassification of endoreduplicated near haploidy
as high hyperdiploidy is detrimental.[24] In a sample where the near-haploid blasts have endoreduplicated, the ensuing hyperdiploid
blasts are larger in size than their founder cell.[24]
[25]
[26]
[27] Using blast size–specific DI assessment, both hypodiploid and endoreduplicated hyperdiploid
blasts in a sample can be identified in isolation.[9]
Role of DNA Index in Plasma Cell Myeloma
Malignant plasma cells in nearly 50% of patients with PCM are hyperdiploid due to
trisomies involving odd-numbered chromosomes.[28]
[29]
[30]
[31]
[32] These hyperdiploid PCM patients have a better prognosis as compared to hypodiploid
and IgH translocated PCM patients.[33]
[34] As plasma cells are terminally differentiated, conventional CG results in normal
diploid metaphases (possibly of myeloid origin) in nearly 70% of PCM samples.[34] In this context, a multiparametric FCM DI analysis can identify ploidy in both normal
and malignant plasma cells.[10]
[35] Using normal plasma cells as the reference population, malignant plasma cells with
DI of 0.95 to 1.05, 1.06 to 1.50, 1.51 to 1.70, and greater than 1.71 are considered
diploid, hyperdiploid, near tetraploid, and tetraploid, respectively.[35] Using lymphocytes as diploid control, an alternative cutoff of ≤0.95 (hypodiploid),
0.95 to ≤ 1.05 (diploid), 1.06 to 1.15 (near hyperdiploid), and ≥1.16 (hyperdiploid)
to define DI-based ploidy has also been described.[10]
S-Phase Fraction of Malignant Cells
In a cell cycle, cells in the S phase are those that undergo active DNA replication
(refer to [Table 1]). As cells in the S phase eventually result in mitosis (i.e., physical doubling
of the cells), the SPF reflects the proliferative capability of the tumor.
Role of S-Phase Fraction Assessment in Acute Lymphoblastic Leukemias
The relevance of SPF assessment in acute leukemias dates back to 1980 when the presence
of greater than 6% lymphoblasts in the S phase was associated with poor prognosis
among pediatric patients diagnosed with ALL.[36] This 6% cutoff was established based on the median SPF (range of 1–60% and mean
of 9%) among 65 ALL samples with greater than 80% lymphoblasts, irrespective of their
B- or T-lineage origin.[36] The prognostic relevance of this 6% SPF was further confirmed by contemporary literature
from both the Dutch and British groups.[11]
[36] However, further literature regarding the relevance of SPF in pediatric patients
with ALL is sparse and has not claimed any prognostic relevance.[18]
[37]
[38] Due to the emergence of more robust risk assessment modalities, SPF is no longer
used as a prognostic tool in any present-day pediatric ALL treatment protocols.
Role of S-Phase Fraction Assessment in Plasma Cell Myeloma
The proliferative capability of plasma cells, as indicated by the plasma cell labeling
index (PCLI), is a time-tested prognostic factor in patients diagnosed with plasma
cell dyscrasias.[39] Initial PCLI identification assay using radiolabeled thymidine incorporation followed
by autoradiography was both technically challenging and hazardous. In 1985, PCLI by
a smear-based immunofluorescence assay utilizing an anti-5-Bromo-2'-deoxyuridine (BrdU)
antibody was introduced. A PCLI of less than 1, 1 to 3, and ≥3% identified by this
technique was associated with low, intermediate, and high-risk diseases, respectively.[40]
[41]
[42]
[43]
[44] Although this BrdU-based PCLI technique is safer and relatively easier than the
radiolabeled thymidine technique, it is not widely utilized in routine practice.[39]
In 1994, Orfäo et al described an FCM-based CD38/propidium iodide dual staining technique
to analyze the DNA content of plasma cells in the bone marrow of patients diagnosed
with PCM.[45] According to that study conducted among 120 treatment naïve PCM patients, those
patients with ≥3% malignant plasma cells in the S phase had inferior relapse-free
and overall survival.[30] To date, higher SPF of malignant plasma cells identified by FCM is considered one
of the parameters associated with high-risk disease across the spectrum of plasma
cell dyscrasias and at various phases of treatment.[39]
[46]
[47]
[48]
Laboratory Aspects of Flow Cytometric Ploidy Assessment
A successful FCM DNA analysis warrants fulfilling of the prerequisites discussed in
the following sections.
Sample-Related Prerequisites
FCM DNA analysis pertinent to hematological malignancies is commonly performed on
fresh, unfixed, fluid-state samples like peripheral blood, bone marrow, body fluids,
and fine needle aspirates.[18] In indications requiring sample fixation (like transported samples), precipitating
fixatives (acetone and alcohol) are preferred over cross-linking fixatives (formaldehyde
and glutaraldehyde).[4] This is because the nucleic acid cross-linking resulting from these fixatives will
interfere with the desired stoichiometric biding of the dyes.[4] In the era of univariate and bivariate FCM DNA analysis, it was recommended to perform
the assay in samples with a minimum of 15 to 20% malignant cells of interest.[49] With the advent of multiparametric FCM DNA analysis, the assay could be successfully
performed even in samples with less than 1% malignant cells.[8]
If the assay has to be performed on archived paraffin blocks, the following factors
have to be taken care of. A cell suspension must be prepared from a section of the
paraffin block that contains both the tumor cells and normal resident cells (selected
by light microscopic evaluation). The thickness of the section is determined by the
size of the tumor cell's nucleus (a 50-nm-thick section would be optimal).[5] The quality of DNA binding dye staining is relatively inferior as the tissue is
already exposed to formaldehyde during fixation; hence, stoichiometric staining might
not be optimal.[3]
[4] In this scenario, 4',6-diamidino-2-phenylindole (DAPI) is the preferred nucleic
acid binding dye as it is least affected by formaldehyde-induced cross-linking.[4] Since the nucleus of cells is cut through during sectioning, it is common to get
a sub-G0/G1 population during analysis. This population must not be interpreted as
a hypodiploid clone of cells. In such situations, curve-fitting algorithms/software
(Dean–Jett model of 1974, Fox-modified Dean–Jett model of 1980, Watson's pragmatic
model of 1987, MultiCycle of 1988, etc.) can be used to analyze the cell cycle.
Assay-Related Prerequisites
For any FCM DNA assay, it is mandatory to validate the linearity, degree of resolution,
and robustness of sample processing in a machine and panel-specific setting. Commercially
available chicken or trout (they inherently have a mixture of single, double, triple,
and quadruple nucleated red blood cells [RBCs]) erythrocytes are used for this purpose.[4]
[10]
According to the 1993 DNA cytometry consensus conference's recommendations, indigenous
normal cells (stromal cells in the case of paraffin-embedded tissues) present in a
tumor-rich sample should be used as internal diploid controls. In scenarios where
such internal controls are not available/sparsely present within the sample, normal
diploid cells sourced from healthy controls can be spiked to the tumor sample before
processing.
In the context of FCM DNA analysis pertinent to hematological malignancies, lymphocytes
are used as diploid controls. Neutrophils and monocytes are not ideal controls as
the former are highly fragile and the latter have increased stainability (especially
for propidium iodide) due to their unique chromatin conformation.[4]
Sample Processing–Related Prerequisites
The protocols pertinent to sample processing toward FCM DNA analysis depend on the
nature of the sample, the cell-permeant ability of the dye, and its extent of RNA
binding capability (refer to [Supplementary Table S1], available in the online version). Detailed discussion regarding the specific modifications
required for each dye is beyond the scope of this review but is readily available
in the literature.[2]
[3]
[50]
[51] The vital features are that the samples must be processed within 24 to 48 hours
of collection. A stable dye binding equilibrium is achieved only if the concentration
of dye used for staining is nearly 100 times above the concentration of the cells
to be stained. Hence, the optimal concentrations of cells and the dye used must be
standardized to achieve stoichiometric staining.[52] The processed samples must be acquired immediately and the acquisition must be at
a low event rate, around 100 to 200 events per second.[52] A minimum of 200 diploid control cells must be acquired.[52]
Analysis-Related Prerequisites
For optimal FCM DNA evaluation, a minimum of 90% of cells must be stained with the
DNA binding dye.[52] The quality of an FCM DNA assay is evaluated based on the coefficient of variation
(CV) of the G0/G1 phase of diploid control cells present in the processed sample.
This CV is the result of sample quality, sample processing errors, heterogeneity in
individual cells' DNA content, and the degree of chromatin compaction. For optimal
FCM DNA analysis, the CV of the G0/G1 phase of diploid control must be less than 6%,
with CV less than 3% yielding the best results.[4]
[18]
[52]
The common pitfalls encountered during FCM DNA analysis are depicted in [Table 3].[3]
[4]
[52]
Table 3
Common issues encountered during flow cytometric DNA analysis and the possible reasoning
and solutions to consider
|
Issue
|
Reasons
|
Solution
|
|
Doublets and cell aggregates
|
Higher relative centrifugal force (RCF)
|
The ideal RCF has to be calculated to achieve an optimal centrifugation force of 500–540
g
[52]
|
|
Aged sample. Autolysis of cells cause release of sticky nuclear content, causing cell-to-cell
adhesions
|
Sample should be processed as fresh as possible, ideally within 24–48 h from collection[52]
|
|
Sample acquired at high acquisition rate
|
The sample must be acquired at a low event rate, ideally 100–200 events/s[52]
|
|
Harsh handling of samples during processing
|
Mixing steps involved during processing must be gentle, preferably using a Pasteur
pipette[52]
|
|
Nonstoichiometric staining (ideally >90% cells in a sample must be stained with the
dye)[52]
|
Low concentration of dye used for staining
|
Re-standardize the assay with gradients of dye concentration to identify optimal dye
concentration to be used. Note: the cell concentration and incubation time have to
be kept constant
|
|
Excess of cells in the cell suspension being stained
|
Re-standardize the assay with gradients of cell concentrations to identify optimal
cell concentration to be used. Note: the dye concentration and incubation time have
to be kept constant
|
|
Insufficient incubation time
|
Re-standardize the assay with gradients of incubation time to identify optimal incubation
time. Note: the dye and cell concentration have to be kept constant
|
|
Samples from patients on DNA binding anticancer drugs
|
Repeat the assay in a treatment naïve sample (if available)
|
|
Undesired RNA binding[4]
|
The degree of undesired RNA binding depends on the dye's RNA binding propensity
|
• While using dyes with higher RNA binding propensity (e.g., propidium iodide), the
sample has to be treated with RNAse before incubation with the dye[4]
• A dye without RNA binding capability (e.g., FxCycle violet) can be used[9]
[52]
|
|
Sub-G0/G1 population[4]
(the cells that have DNA content less than that of a normal diploid control)
|
Apoptosis: apoptosis results in DNA fragmentation of specific lengths. During sample
processing, apoptotic cells leak out their intracellular contents (including smaller
DNA fragments). Hence, apoptotic cells have a lower total DNA content than normal
diploid control and manifest as a distinct sub-G0/G1 peak
|
• If a sub-G0/G1 peak is also present in diploid controls, then the sub-G0/G1 population
present in the malignant cells can be considered apoptotic and ignored; however, this
approach is subjective and risks exclusion of true hypodiploid malignant cells[9]
• Concurrent staining with annexin V will aid objective identification of apoptotic
cells[9]
|
|
Necrosis: necrosis involves random-sized DNA fragmentation, resulting in a sub-G0/G1
“smearing pattern” against a perfect peak generated by apoptosis/hypodiploid cells
|
Samples should be fresh and processed within 24–48 h of collection[52]
|
