Abstract and Introduction

Abstract

The diagnosis of leukemia relies upon a multiparametric approach involving a number of different
pathology disciplines. Molecular methods are increasingly employed to help refine diagnosis,
establish prognosis and determine the most appropriate treatment, including rational therapies
targeting the underlying genetic lesion. This review aims to highlight some of the molecular
techniques commonly used in the diagnosis of leukemia using relevant examples. The focus is on
procedures in current use and technologies showing promise in the research setting that are likely to
enter clinical use in the near future. The list is not exhaustive, and this article concentrates on
diagnosis of leukemia; techniques used to monitor response to therapy and molecular residual
disease are mentioned but have not been covered extensively.

Introduction

Leukemia is characterized by abnormal proliferation of hematopoietic cells as the result of

underlying genetic lesions. Leukemia accounts for approximately 2.5% of all cancers diagnosed in the
UK, with an annual incidence of 13 per 100,000 individuals.[101] Of the four main types, chronic
lymphocytic leukemia (CLL) is the most common, followed by acute myeloid leukemia (AML), acute
lymphoblastic leukemia (ALL) and chronic myeloid leukemia (CML). The WHO classification system
for tumors of the hematopoietic and lymphoid tissues has been widely adopted and endorses a
multiparametric approach to diagnosis.[1] The latest edition published in 2008 recognizes several
entities based solely on their molecular pathogenesis, particularly from within the myeloid lineage.

Leukemia is typically diagnosed via abnormal results on a full blood count followed by examination
of a blood film. The diagnosis is usually confirmed on a bone marrow aspirate and/or trephine
biopsy looking at cell morphology and bone marrow architecture. Additional techniques that may be
used to further classify the type of leukemia include flow cytometric immunophenotyping[2] and
genetic analysis.

Genetic Studies in Leukemia

In many types of leukemia, genetic studies at diagnosis are considered to be crucially important,
especially with regards to prognostication and therapeutic choice. In practice, a bone marrow or
peripheral blood sample is referred to the diagnostic laboratory with a (provisional) diagnosis based
on the initial morphologic and immunophenotypic findings. The techniques employed to define the
genetic aberrations within the leukemia depend on a number of factors, including: the subtype of
leukemia; clinical urgency; type, volume and age of the sample; relevance of the genetic marker; and
also the available technology within the laboratory. An overview of these genetic techniques is
provided in Table 1. The scope of this review is molecular methodologies, illustrated by selected
clinical examples; therefore cytogenetic techniques will only be described in brief.

G-band Metaphase Chromosome Analysis ('Karyotyping')

Conventional cytogenetic analysis in the newly diagnosed leukemia patient relies on the presence of
mitotically active dividing cells, which usually means that a bone marrow specimen must be
cultured, although, if there are circulating leukemia cells then sometimes blood can be used as a
surrogate. Cells are arrested in metaphase of the cell cycle, when the chromosomes are at their
most condensed, and easily visible. The chromosomes in the 'metaphase spread' are processed so
that staining with Giemsa (or Leishman) produces a characteristic banding pattern, allowing
individual chromosomes to be distinguished, and numerical and structural abnormalities to be
identified. A clonal abnormality is usually defined by the presence of at least two cells with the same
aberration (three cells in the case of chromosome loss). Even when the aberration identified is not
recurrent in leukemia, the finding of a clonal chromosomal defect can provide evidence for a
malignant process.

Metaphase cytogenetic analysis of marrow is useful to detect translocations and related changes
such as inversions and aneuplodies, which are common in AML, ALL and CML. Cytogenetic analysis
provides a low-resolution whole-genome scan, and has the advantage of being able to detect
balanced rearrangements, which are relatively common in leukemia. Its main drawbacks are limited
resolution (typically 35 Mb), sensitivity (510%, depending on the number of cells analyzed, and
assuming disease is present in the marrow) and the challenge of leukemia cells with a low mitotic
index (for example in CLL). It is worth noting that gross chromosomal abnormalities can only be
detected in a proportion of leukemia (e.g., in approximately 55% of AML). When metaphase
cytogenetic analysis is applied to the study of hematological malignancies, many chromosomal
defects will remain undetected.[3]

Metaphase cytogenetic analysis is able to provide a definitive diagnosis upon detection of the
pathognomonic rearrangements in CML (t(9;22)(q34;q11), the 'Philadelphia [Ph] chromosome',
creating a BCRABL1 gene fusion), acute promyelocytic leukemia (t(15;17)(q24;q21) PMLRARA) and
core-binding factor AML (either t(8;21)(q22;q22) RUNX1RUNX1T1, or inv(16)(p13q22) or
t(16;16)(p13;q22)CBFBMYH11), even in the absence of other diagnostic features, such as a blast
count >20% in AML.[1] Many other recurrent (and unique) abnormalities are found across all types of
leukemia, and it can be anticipated that at least some of these will be considered pathognomonic of
different subtypes of leukemia in the future.[4]

Chromosomal aberrations may also provide important prognostic information. In AML, cytogenetic
findings are the most important parameter in establishing the prognosis, and three broad prognostic
groups (good, intermediate and poor) are defined on the basis of the diagnostic karyotype, which
can help to define those patients that may benefit from stem cell transplantation.[5] In CML, the
finding of additional karyotypic abnormalities, namely a second Ph chromosome, trisomy 8,
isochromosome 17q or trisomy 19, at presentation may have a negative impact on survival, and may
signify that the leukemia has already progressed to accelerated phase or blast crisis.[6]

Fluorescence in situ Hybridization

FISH provides a useful adjunct to cytogenetic analysis, its main advantage being that it does not rely
on dividing cells and can therefore be performed on cells in interphase. Blood can be used if marrow
is unavailable, and results can be obtained more quickly because cells do not need to be cultured.
FISH is also both more sensitive than conventional cytogenetic analysis (0.51%) and has higher
resolution (100 kb, depending on the probe), although, depending on the type of probe, caution
needs to be applied to avoid false-positive and -negative results due to signal colocalization or drop-
out. Laboratories should establish cut-off values for each probe set to define unambiguous positive
results.

FISH uses fluorescently labeled DNA probes to locate specific sequences of interest and can thus
identify structural and numeric chromosomal changes including balanced rearrangements and
microdeletions. A targeted approach using a limited number of probes is usually employed to look
for the most common or clinically relevant aberrations within the subtype of leukemia, so, unlike
karyotype analysis, the technique does not normally provide a genome-wide assessment. The
exception is multiplex FISH (M-FISH), where paints identify individual metaphase chromosomes,
which can be useful in resolving complex karyotypes or poor morphology metaphase spreads where
chromosomes are difficult to identify by their G-banding pattern alone.

FISH is particularly useful in the diagnosis of leukemia with cryptic cytogenetic abnormalities. One
striking example of this is the t(12;21)(p13;q22) translocation resulting in an ETV6RUNX1 gene
fusion, the most frequent translocation in pediatric B-lineage ALL, which as a result of the
breakpoints being relatively close to the telomeres of the short arm of chromosome 12 and the long
arm of chromosome 21 is cytogenetically 'invisible'. It is, however, vital to identify this aberration for
risk-adapted therapeutic protocols, as it is associated with the good risk prognostic group.[7] The
translocation can be detected using FISH or molecular techniques.

Interphase FISH is able to identify genomic aberrations in approximately 80% of CLL cases using a
disease-specific panel of probes,[8] and is particularly useful for identifying TP53 (17p13)
and ATM(11q23) deletions, both being associated with poor prognosis. Interphase FISH analysis is
generally considered to be superior to metaphase cytogenetic analysis in CLL owing to: the
difficulties in persuading B cells to divide in culture (they divide slowly and need mitogen stimulation
to produce enough metaphases for analysis); the fact that most clinically relevant aberrations in CLL
are copy-number changes; the limited sensitivity of metaphase analysis such that small populations
of abnormal cells may be missed; the fact that submicroscopic deletions of prognostic relevance will
not be detected by conventional G-band cytogenetics; and the fact that interphase FISH is quicker
than metaphase analysis and may be cheaper if limited panels of probes are used.

In vitro Molecular Techniques

Molecular techniques that can be employed in the diagnosis of leukemia are many and varied, each
with their own advantages and disadvantages. Detailed below are some of the more common
techniques in routine clinical use, and some techniques that are currently used in the research
setting but that are on the verge of transferring to clinical use. Selected examples of their use in
leukemia diagnostics are mentioned.

Microarray-based Techniques

Whole-genome Scanning by Molecular Karyotyping: Array Comparative Genomic Hybridization

Comparative genomic hybridization (CGH) is used to compare the genetic material from a test
individual, such as a leukemia patient, to that of a reference 'normal' individual (usually DNA pooled
from several subjects), to identify the presence of copy-number changes in the test sample. Test and
reference DNA are digested into small fragments and each labeled with a different fluorophore. In
the past, the DNA was allowed to hybridize to normal metaphase spreads to identify copy-number
changes with a resolution of 23 Mb via differential fluorophore binding, but to improve resolution
metaphase spreads have now been replaced with microarrays. The DNA probes on the array can be
bacterial artificial chromosomes (BACs), or more commonly, oligonucleotides. Resolution and
sensitivity are determined by many factors including the length of the probes, the number of probes
on the array (probe density), probe distribution, size of clonal population, DNA quality and software
analysis algorithms. Deletions or insertions as small as 50 kb can be detected by array CGH (aCGH),
which is a marked improvement over karyotyping. Whole-genome scanning arrays have allowed for
the discovery of chromosomal aberrations in a much higher proportion of patients with leukemia,
and it is hoped that identification of novel aberrations may lead to more precise prognostic
schemes.[9] However, one distinct disadvantage of aCGH is that it is unable to detect balanced
rearrangements, which are relatively common in leukemia.

Array design is critical and is determined by the application. CGH arrays normally have a high
number of probes spaced evenly throughout the genome, with increased density of probes at
regions of particular interest. For example, a 385K array has median probe spacing of 6 kb, but by
targeting specific regions, breakpoints can be mapped within 5-kb intervals.[9] Arrays can also be
designed to be targeted to only specific regions known to be associated with disease. An overview of
the aCGH process is provided in Figure 1.

Figure 1.

Overview of the array comparative genomic hybridization process. Fragmented test DNA and
reference (control) DNA are labeled with different fluorophores and allowed to competitively
hybridize to the microarray, consisting of oligonucleotides or bacterial artificial chromosome clones
spotted onto a slide. If the test DNA contains a region of gain in copy number, it will be in excess and
will preferentially bind to the array, resulting in increased test signal and decreased reference signal
for the probes within the region of gain in copy number. If the test DNA has a decreased copy
number, then reference DNA will be in excess for that region and there will be decreased test signal
intensity (and increased reference signal intensity) for the affected probes.

CLL is ideally suited to analysis with copy-number arrays because the genetic lesions with known
clinical relevance are chromosomal gains and losses. Sargent et al. created and validated a custom
aCGH platform as a clinical assay for CLL genomic profiling, designed to interrogate all known CLL
prognostic loci.[10] Their 60-mer, 44,000-probe oligonucleotide array with a 50-kb average spatial
resolution was augmented with high-density probe tiling at loci that are frequently aberrant in CLL.

The results of these and other studies using array-based karyotyping to evaluate CLL have
consistently reported high concordance with FISH panel results, with instances of nonconcordance
explained by low tumor burden, the presence of small subclones or the relatively low resolution of
the arrays used in the study.[11] aCGH profiling thus represents a feasible routine clinical test for CLL.
Furthermore, Gunn et al. were able to identify clinically significant atypical 11q deletions (atypical
because they did not include deletion of the ATM gene) in CLL with aCGH that may be missed by
FISH panels used for prognostic stratification of the disease.[12]

aCGH has also been used as a research tool to study novel genomic imbalance in CLL; although
recurrent chromosomal alterations occur, relatively few affected tumor suppressors and oncogenes
have been implicated in the disease.[13] Using a BAC array, Gunn et al. found a high proportion of
submicroscopic deletions, both monoallelic and biallelic, of chromosome 22q11.[14] They
subsequently used a higher-resolution oligonucleotide-based array to show that the 22q11 deletions
ranged in size from 0.34 Mb up to approximately 1 Mb, and demonstrated that genes in the
minimally deleted region (including PRAME) had significantly reduced mRNA expression by reverse-
transcription quantitative real-time PCR (RT-qPCR).

Whole-genome Scanning by Molecular Karyotyping: SNP Arrays

The resolution of microarray technology has been improved further still with the introduction of SNP
arrays. In contrast to aCGH, SNP arrays do not rely on competitive binding of reference DNA, but
genotype polymorphisms directly in test DNA, and the hybridization signal strength from individual
probes allows for the estimation of gene copy number. Given the high density of SNPs that can be
evaluated on a genome-wide level (>750,000), very small regions of copy-number alteration can be
identified,[15] and such arrays have again proved popular in the study of CLL, using for example a
250K SNP array.[16] Here, 250K refers to the number of SNPs distributed across the genome (rather
than their spacing), so a 250K array comprises 250,000 SNPs, whereas a 500K array has double the
SNP density. Arrays with even higher numbers of probes are available, which can decrease the
minimal detectable size of a deletion to approximately 25 kb.[17] However, whereas CGH arrays
usually have probes evenly spaced across the genome, the distribution of probes on SNP arrays is
dictated by the location of SNPs, so resolution within 'SNP deserts' can be relatively poor.[18] To
increase resolution, arrays containing both SNPs and genomic probes have been created.[9] One such
example is the Affymetrix CytoScan HD, which contains 750,000 SNPs and 1.9 million
nonpolymorphic probes, and whose coverage includes all 526 Sanger cancer genes, with an average
marker spacing of 553 bp within the cancer genes (>25 markers per 100 kb).[102] This array is capable
of identifying genomic breakpoints at the exonic level throughout the entire genome; however, as
with any leukemia sample, the ability to resolve copy-number changes and loss of heterozygosity
(LOH) depends on the level of mosaicism (proportion of leukemic cells versus normal cells) within
the specimen.

Gunnarsson et al. used their 250K SNP array to analyze a cohort of newly diagnosed CLL patients and
to detect clonal evolution in follow-up samples.[16] They identified copy-number aberrations in 90%
of these CLL patients at diagnosis, with 70% carrying known recurrent alterations, including del(13q)
(55%), trisomy 12 (10.5%), del(11q) (10%), and del(17p) (4%).[16] In addition, they identified a small
number of patients with copy number neutral LOH (CN-LOH, also called acquired isodisomy or
acquired uniparental disomy) on 13q, which is the loss of all or part of a chromosome, and a
doubling of the remaining homologous genetic material to restore genomic balance (see next
section for more detail). The ability to detect CN-LOH is a singular feature of SNP arrays, as this
phenomenon is invisible by conventional cytogenetic analysis, FISH and most CGH arrays (SNPs have
recently been introduced within some CGH arrays allowing CN-LOH detection [BlueGnome's
Cytochip Cancer; Oxford Gene Technology's Cytosure ISCA UPD array]). SNP arrays are able to
detect CN-LOH because in addition to measuring copy number, they also provide genotyping
information, and can therefore detect diploid stretches of homozygosity within the genome.
Acquired CN-LOH can also be detected using PCR-based molecular techniques; one of the first
reports of CN-LOH in AML, of the long arm of chromosome 13 encompassing the FLT3 locus, was an
incidental finding after using polymorphic microsatellite markers to study chimerism status post
stem cell transplantation.[19]

Copy-number Neutral LOH Numerous studies have shown that CN-LOH is a frequent event in
leukemia, particularly myeloid malignancy,[20] but one that went mostly unrecognized until SNP
arrays were available for routine use. The utility of SNP arrays to detect CN-LOH in AML was initially
reported in a study of 60 AML patients using 10K arrays.[21] CN-LOH is regularly identified in patients
with a normal karyotype and no other clonal marker, and is particularly common in mixed
myelodysplastic syndrome (MDS)/myeloproliferative neoplasm (MPN),[22] with abnormality rates of
48% in chronic myelomonocytic leukemia and 38% in MDS/MPN-unclassifiable cases,[23] although it
should be noted that this study did not use paired constitutional DNA to validate somatically
acquired changes so these values may be overestimates. Regions of CN-LOH in leukemia often
encompass oncogenes or tumor suppressor genes, facilitating duplication of a mutation with
concomitant loss of the wild-type allele, but without any genomic imbalance. Furthermore, studies
of candidate genes within regions of recurrent CN-LOH identified by SNP arrays have led to the
identification of novel mutated genes such as CBL (associated with CN-LOH 11q in myeloid
malignancy[23]) and TET2 (linked to CN-LOH 4q in MDS and mixed MDS/MPN[24,25]). SNP arrays have
also detected recurrent submicroscopic deletions, which again have enabled the identification of
new cancer-related genes in the minimally deleted region, such as PAX5,IKZF1 and CDKN2A in
pediatric ALL.[2628]

Overall, array-based karyotyping increases diagnostic yield when combined with routine metaphase
cytogenetics.[3] However, with both aCGH and SNP arrays, detection of unbalanced cytogenetic
defects relies on a sufficient number of cells sharing a clonal abnormality, so in cases with a small
clonal population, or multiple subclones, standard metaphase cytogenetics may still be the most
appropriate technique to use in the investigation of leukemia despite its limited resolution and
reliance on dividing cells. A further inherent challenge of array analysis lies in correctly distinguishing
somatically acquired, cancer-specific lesions from patient-specific inherited copy-number variations
or segments of homozygosity. Copy-number variations are surprisingly frequent and not highly
recurrent, making paired studies with matched tumor and germline DNA samples critical for
correctly ascertaining somatically acquired variants.[29]

Gene Expression Profiling

Microarrays can be used to study global gene expression, simultaneously measuring the
transcriptional activity of thousands of genes. The application of microarrays for classification of
leukemia subtypes has been demonstrated in numerous studies, including the multicenter
Microarray Innovations in Leukemia (MILE) project involving 11 laboratories worldwide. The MILE
study assessed the clinical utility of gene expression profiling (GEP) as a single test to subtype
leukemias into conventional categories of myeloid and lymphoid malignancies based upon gene
expression signatures associated with distinct clinical subtypes.[30] Over 3000 patients comprising 16
acute and chronic leukemia subclasses, MDS and a 'none of the target classes' control group were
profiled on a custom 'AmpliChip Leukemia' and results compared to 'gold standard' diagnostic
methods. The AmpliChip Leukemia was able to classify leukemia with a high level of accuracy, was
robust, with a high degree of inter- and intra-laboratory correlation, and was able to generate
results within 48 h.[30] The authors hope that the MILE study will pave the way to the standardized
introduction of microarray technology in the diagnosis and treatment of leukemia, with potential
applicability in developing countries that currently lack the expertise to perform more labor-
intensive and sophisticated diagnostic approaches.

The first commercially available CE-marked chip for the classification of AML was launched in 2011
by Skyline Diagnostics. The AMLprofiler uses distinct gene expression profiles to identify patients
with favorable cytogenetic reciprocal rearrangements and CEBPA bialleic mutations, based on the
earlier work on class prediction by Valk et al..[31] For example, the presence of t(8;21)(q22;q22) is
determined by 31 proprietary gene expression levels, with virtually 100% accuracy. The chip is
unique in that it can also directly detect the common NPM1 gene mutation types A, B and D, as well
as overexpression ofEVI1 and low expression of BAALC. In this way, a single method replaces three
different technologies: cytogenetics, mutation analysis and expression analysis.[103]
One disadvantage of GEP in the classification of AML is that prognostically significant abnormalities
in signaling molecule genes such as FLT3 and RAS appear not to be readily predictable, perhaps on
account of their less direct role in transcriptional modulation.[32] Outside class prediction, GEP has
the potential to facilitate class discovery, referring to the identification of new subtypes of leukemia
by grouping cases according to similar gene expression signatures (often referred to as clustering).
Wouters et al. were able to identify a subgroup of patients with an expression profile resembling
that of AML with bialleic CEBPA mutations despite these patients not carrying the respective
mutations.[33] The subgroup was actually associated with silencing of the CEBPA gene, often due to
promoter hypermethylation.[33] GEP also enables class comparison, identifying genes that are
deregulated in certain subgroups, which may address biologic questions and facilitate the detection
of new molecular targets for therapy.[32,34]

PCR-based Techniques

The cornerstone of molecular diagnosis in leukemia is PCR. Manipulations of the reaction itself and
downstream processing have resulted in a multitude of different techniques; some of the most
relevant to the diagnosis of leukemia are described below.

Quantitative Techniques

Reverse-transcription PCR & Reverse Transcription Quantitative Real-time PCR

RNA is a prerequisite for the GEP techniques described above. RNA is also used as the template of
choice when studying the molecular counterparts of chromosomal rearrangements (i.e., gene
fusions) and for a few other selected applications within the genetics laboratory. This is not
necessarily because the test relies on expression per se, but because RNA devoid of introns makes a
more convenient template. Translocation breakpoints are usually patient specific within the DNA
sequence, although they do have a tendency to cluster within particular regions (usually within
certain introns, but occasionally within exons). By using RNA (or cDNA) as the template and locating
primers just outside the breakpoint cluster regions, a common set of primers can be employed to
detect the fusion gene counterparts of chromosomal translocations in the majority of patients,
except those that have rare fusion subtypes with breakpoints outside the common regions (Figure
2). The disadvantage of using RNA is that it is labile; samples need to reach the laboratory quickly
and ideally be extracted within 4872 h to prevent the action of RNase enzymes that degrade
RNA.[35]

Figure 2.

The t(8;21)(q22;q22) translocation resulting in aRUNX1 RUNX1T1 gene fusion. Breakpoints occur
within intron 5 of RUNX1 and within intron 1 of RUNX1T1. To circumvent the need to design patient-
specific primers near to the exact breakpoints in the genomic DNA sequence, cDNA (reverse
transcribed from mRNA) can be amplified using a common pair of external primers (E) located within
exon 4 of RUNX1, and exon 3 of RUNX1T1. Increased sensitivity for minimal residual disease
monitoring can be achieved by performing a second round of reverse-transcription PCR with primers
positioned internally (I) to the first-round primers (nested reverse-transcription PCR). Cen:
Centromere; Tel: Telomere. Courtesy of Jane Bryon, West Midlands Regional Genetics Laboratory
(Birmingham, UK).

End-point reverse-transcription PCR (RT-PCR), where products are assessed after the completion of
the amplification reaction, provides a nonquantitative (or if PCR is kept in the exponential phase, a
semiquantitative) assessment of the presence or absence of the specific product of interest. This is a
useful technique at leukemia presentation to confirm or exclude the presence of a gene fusion: it is
quick; sensitive relative to FISH and cytogenetics; does not rely on dividing cells so can potentially be
performed on blood samples; can detect cytogenetically and FISH cryptic rearrangements; and may
be able to detect rare variants in addition to standard gene fusions depending on the location of the
primers. We have been able to identify a total of 19 leukemia patients with rare variant BCR
ABL1,RUNX1RUNX1T1, CBFBMYH11 or PMLRARA transcripts using our RT-PCR primers that
cannot be amplified using the standard reverse transcription quantitative real-time PCR (RT-qPCR)
primers in routine use. These would have been missed if screening at presentation relied on RT-
qPCR, with one variant CBFBMYH11 fusion also being cytogenetically cryptic.[35]

Qualitative end-point RT-PCR can be used for minimal residual disease (MRD) monitoring,[36] but has
been largely overtaken by quantitative techniques that allow estimation of disease levels, and in the
case of negative results, give an indication of quality and sensitivity to avoid 'false negatives' due to
suboptimal samples. RT-qPCR can be performed using various fluorescent chemistries (e.g., DNA
binding dyes, TaqMan hydrolysis probes [Roche], Lightcycler dual hybridization probes [Roche],
Molecular Beacons, Locked Nucleic Acid probes) and on a range of platforms (such as ABI's
7500/7900/ViiA Real-Time PCR systems, Roche's LightCycler, Bio-Rad's CFX96 Real-Time PCR
Detection System, Cepheid's SmartCycler and Qiagen's Rotor-Gene Q). The one thing they all have
in common is that the reaction is monitored in real time, with fluorescently labeled products
detected and measured in the exponential (also known as log-linear) phase. During this phase, the
PCR product doubles after every cycle, assuming 100% reaction efficiency. A fractional PCR cycle
early in this phase, when amplification can first be detected above background noise, is used for
quantification, known variously as the threshold cycle, crossing point, take-off point or quantification
cycle. This value is representative of the starting copy number in the original template; the greater
the starting amount, the lower the PCR cycle at which fluorescence (amplification) can first be
detected.

There are two main methods of quantification: absolute and relative. Absolute quantitation uses
serially diluted standards of known concentrations to generate a standard curve from which the
concentration of unknowns based on their quantification cycle values can be determined. With
relative quantification, changes in sample gene expression are measured based on either an external
standard or a reference sample, also known as a calibrator.[37] Pros and cons of the two
methodologies and further detailed information on the design and theory behind RT-qPCR
experiments can be found in the excellent publications by Wong and Medrano[37] and Bustin et al. [38]

Fluorescence-based real-time PCR is one of the most widely used methods of quantification because
it has a high dynamic range, is very sensitive and specific, and requires no postamplification
processing.[37,38] It is commonly used to measure response to therapy and monitor MRD in
leukemia.[39] European standardized protocols for the measurement of fusion gene transcripts in
acute leukemia and CML by RT-qPCR were published in 2003 and are still in widespread use
today.[40] It is beyond the scope of this review to cover molecular monitoring of leukemia, but a
wealth of information is available in the published literature on both methodologies and the clinical
application of such monitoring.
Real-time PCR can also make use of DNA as a template. Several studies have shown that levels of
MRD, measured at critical timepoints in both childhood and adult ALL, significantly correlate with
clinical outcome.[41] There are a number of methods available, including identification and
subsequent monitoring of clonal immunoglobulin and T-cell receptor gene rearrangements by PCR
amplification of DNA, where patient-specific primers are designed complementary to the junctional
sequences of the target. Again, European standardized assays[42,43] and consensus guidelines on
interpretation of results[44] have been published. Other methods of monitoring MRD in ALL include
RT-qPCR for fusion genes, as discussed above, and flow cytometry based on a leukemia-associated
immunophenotype.[45]

Quantitative techniques based on real-time PCR have been available for leukemia diagnostics for a
relatively long time; other innovative methods of molecular quantification are available, such as
multiplex ligation probe amplification (MLPA; MRC-Holland) and digital PCR, which are discussed
below, and pyrosequencing, which is considered in the section on sequencing-based techniques.

Multiplex Ligation Probe Amplification

MLPA is a technique that was initially developed to identify copy-number changes, from whole-
chromosome aneuploidy to single exon deletions and duplications,[46] that has since been further
enhanced for methylation profiling and mutation detection. MLPA is based on the hybridization and
subsequent ligation of two separate oligonucleotide probes specific to immediately adjacent target
sequences. Ligation will only occur when both oligonucleotides anneal to the template. At either end
of the ligated probe are universal primer sequences, which enable the probe (not the template) to
be amplified in a multiplex PCR reaction, and a stuffer sequence of variable length, which facilitates
discrimination of different probes based on size, allowing up to 50 targets to be multiplexed.
Quantification of probe ligation products (via peak height or area following capillary electrophoresis;
see later section titled 'Fragment analysis') provides a measure of the number of target sequences in
the sample, and by comparing the peak pattern obtained to that of reference samples, indicates
which sequences show aberrant copy numbers. Probe oligonucleotides that are not ligated only
contain one primer sequence and as a consequence they cannot be amplified and will not generate a
signal. An example of MLPA showing a deletion in NSD1 is shown in Figure 3. Commercial MLPA kits
are available for CLL, MDS, intrachromosomal amplification of chromosome 21, ALL and
'hematological malignancies', the latter kit containing probes for several genes and chromosomal
regions known to have a significant diagnostic or prognostic role in ALL, AML, MDS, CML and CLL,
including 2p23 (ALK), 7p12 (IKZF1), 8q24 (MYC), 10q23 (PTEN), 11q23 (ATM), 12p13 (ETV6), 13q14
(RB1, MIR15A), 17p13 (TP53) and 21q22 (RUNX1) deletions and duplications.[104]

Figure 3.

Multiplex ligation probe amplification (MLPA) analysis of the NSD1 gene. The lighter bars (left peak
within each doublet) are the normal reference sample with no copy-number change inNSD1. The
darker bars (right peaks), representing the test sample, are approximately the same height as the
reference sample for all exons except 19, 20 and 21 (highlighted by arrows), revealing a
partial NSD1 deletion.

One undesirable feature of MLPA is that a SNP within the probe binding site can prevent
hybridization of the probe to the target sequence, and can thus generate a false-positive (apparent
deletion) result. This feature has been exploited by the designers of MLPA assays in order that MLPA
can also be used for mutation detection by deliberately locating probes at the site of a base change.
A kit is due to be launched for 'myeloproliferative neoplasms' that has been designed to detect
the JAK2 mutations V617F, E543_D544del and N542_E543del, MPL W515K and W515L, IDH1 R132S
and R132C, IDH2R140Q, IKZF1 and EZH2 deletions and FIP1L1PDGFRA fusions. MLPA is thus one of
very few techniques that allows detection of mutations and copy-number changes in a single
reaction, and is superior to FISH in being a multiplex technique capable of detecting even single exon
deletions or duplications.

Ligation-dependent PCR (LD-PCR), based on a subtle modification of the MLPA technique, has been
developed for the quantification of point mutations within the ABL1 kinase domain (AKD) of BCR
ABL1.[47] Such mutations can interfere with binding of tyrosine kinase inhibitors to the BCRABL1
protein in patients with CML and Ph chromosome-positive ALL, a major cause of resistance to
therapy. Over 100 different amino acids have been shown to be mutated to date, but only a small
number of residues account for the majority of resistance-causing mutations. LD-PCR assays were
developed for 18 of the most common mutations, including the T315I, which confers resistance to
imatinib, dasatinib and nilotinib. LD-PCR has the advantage over some other detection techniques of
being able to accurately quantify the proportion of mutant clones. Qualitative detection of an AKD
mutation does not necessarily imply impending onset of clinically resistant disease, especially if the
clone size is small, but demonstration of expansion of a mutant clone over time can provide
evidence of clinical relevance.

Digital PCR

Digital PCR, by array or droplet technology, transforms the exponential, analog nature of PCR into a
digital signal suitable for detecting predefined mutations present in a minor fraction of a cell
population.[48] Single molecules of the target sequence are isolated by dilution and individually
amplified by PCR in multiple individual parallel reactions, so that the resultant PCR products are
homogeneous (i.e., completely mutant or completely wild-type). The homogeneity of these PCR
products makes them easy to distinguish.

The limit of detection of digital PCR is defined by the number of individual reactions performed.
Microfluidic sample handling systems, such as Fluidigm's BioMark digital array, have been
developed, which split one sample into thousands of individual reaction chambers that reside on an
'integrated fluidic circuit'. This miniaturization of the PCR, whereby reactions are performed in
nanoliter volumes on the chip compared with milliliter volumes typically used in conventional PCR,
provides a significant reagent cost saving. The miniaturized process allows performance of 39,960
individual PCR reactions per chip (48.770 Fluidigm Digital Array IFC[105]). Alternative and more
sensitive approaches are offered by Bio-rad's Quantalife and RainDance RainDrop droplet digital
PCR technologies. In the latter, millions of emulsified picoliter droplets are generated, which act as
individual reaction chambers within the RainDrop Digital PCR System. By combining this technology
with differentially labeled probes to mutant and wild-type, quantitative detection of low copy
targets with sensitivities in the order of one in more than 1 million can be achieved, and multiplexing
is also possible.[106,107]
Digital PCR shows some promise for MRD monitoring in leukemia, and is adequately sensitive,
successfully detecting BCRABL1 transcripts in samples with low-level MRD that were not detectable
by conventional RT-qPCR.[49] It has also been used to identify and quantify the T315I AKD mutation in
CML patients on the Fluidigm BioMark digital array, with a theoretical 1000-fold improvement in
detection sensitivity versus conventional PCR, and was able to detect as few as three T315I-mutated
molecules in a total background of 100,000 unmutated ABL1 molecules.[50] At diagnosis it may be
applicable: for detection of tumor cells present at very low levels relative to a high background of
normal DNA, such as KIT D816V-positive mast cells in systemic mastocytosis;[51,52] for detection of
low-level subclones harboring key mutations; in the early stages of cancer development; or for
detection of circulating tumor cells or cell-free DNA, particularly those derived from solid tumors,
present in plasma. One potential example of the latter scenario may be confirmation of a CNS
relapse of ALL, where the marrow tests negative.

Mutation Detection

Co-amplification at Lower Denaturation Temperature PCR

A major limitation of PCR-based methods is the inability to selectively amplify low levels of
mutations in a wild-type background, as is frequently the case in leukemia. Co-amplification at lower
denaturation temperature PCR (COLD-PCR) is an elegant technique for enriching minor populations,
and mutation detection sensitivity can be improved by up to 100-fold by replacing regular PCR with
COLD-PCR before sequencing or genotyping assays.[53] The cycling conditions facilitate cross-
hybridization of mutant and wild-type alleles. COLD-PCR relies on identifying the critical temperature
for denaturation of heteroduplexes, and because mutant alleles are in the minority, most mutant
alleles will end up in a heteroduplex that has a lower melting temperature than fully matched wild-
type or mutant homoduplexes. Heteroduplexes are then selectively denatured and amplified at
critical temperature, whereas homoduplexes remain double-stranded and do not amplify efficiently.
Li et al., who first described the technique, demonstrated vastly improved sensitivity of existing
mutation detection methodologies and identified new mutations in TP53, EGFR and KRAS in tumor
specimens simply by replacing regular PCR with COLD-PCR.[54]

Fragment Analysis

Fragment analysis discriminates PCR products based on size and is useful for the detection of
mutations, such as insertions, deletions and duplications, and for identifying gene fusion subtypes.
The process can be thought of as a more accurate and sensitive form of gel electrophoresis; PCR
products are fluorescently labeled and, along with appropriate size standards, loaded onto a
capillary filled with polymer and detected within an automated analyzer.

Two of the most common mutations found in AML can be detected in this way. FLT3 internal tandem
duplications (FLT3-ITDs) are in-frame insertions of three to approximately 400 bp within the
juxtamembrane domain of the FLT3 tyrosine kinase gene,[55] causing constitutive activation of the
receptor. NPM1 gene mutations, which usually involve insertion of four bases in exon 12, cause a
frame shift in the C terminus, disrupting the nucleolar-localization signal or generating a nuclear
export motif, resulting in aberrant cytoplasmic accumulation of NPM.[56] Both types of mutation can
therefore be detected by fragment analysis following a simple PCR using primers either side of the
commonly duplicated/inserted region (Figure 4). Given that both mutations have prognostic
relevance in AML, and it is recommended that both should be ascertained at presentation, many
laboratories have combined the two tests into one essay.[57]
Figure 4.

Fragment analysis for NPM1 insertion and FLT3- ITD mutations. NPM1 alleles on the left; wild-type
allele 167 bp, mutant allele 171 bp representing the common type A 4-bp insertion. FLT3alleles on
the right; wild-type FLT3 236 bp (cDNA) or 326 bp (genomic DNA), ITD allele 275 bp (cDNA) or 365 bp
(genomic DNA), representing a 39-bp duplication. ITD: Internal tandem duplication. Courtesy of
Kerry Wall, West Midlands Regional Genetics Laboratory (Birmingham, UK).

Allele-specific Oligonucleotide Analysis

Fragment analysis can be used in conjunction with allele-specific oligonucleotide analysis (ASO-PCR)
in the detection of the JAK2 V617F mutation. This point mutation in the JAK2 gene, resulting in
constitutive activation of the tyrosine kinase that it encodes, is found in a high proportion of MPN.
The rationale behind the ASO assay is that a single nucleotide mismatch at the 3' end of a primer will
prevent primer binding and extension. A mutant-specific primer is therefore included in the reaction,
which will only bind to an allele that contains a thymine at nucleotide position 1849 in JAK2 (i.e., the
V617F mutation), and will not bind to the wild-type sequence with a guanine at that
position[58] (Figure 5). Wild-type and mutant products are discriminated by size.

Figure 5.

Detection of the JAK2 V617F mutation. (A) JAK2 V617F allele-specific oligonucleotide assay. When
only wild-type JAK2 is present, the F and R primers bind to the DNA, amplifying a 364-bp product,
which can be detected via gel electrophoresis, or if one of the primers is fluorescently tagged, by
fragment analysis. The M primer will only hybridize if the V617F single nucleotide substitution
mutation (c.1849G>T, indicated by white diamond) is present, in which case a 203-bp amplicon is
generated in addition to the 364-bp control fragment. (B) Fragment analysis trace showing a patient
with the JAK2 V617F mutation using the ASO assay. F: Forward; M: Mutant; R: Reverse; V617F:
Mutant allele; WT: Wild-type allele.

In 2011, a point mutation in the BRAF gene was reported to be present in all cases of hairy cell
leukemia (HCL), and absent from other B-cell neoplasms.[59] HCL can be difficult to differentiate from
HCL-like disorders such as splenic marginal zone lymphoma and HCL variant, and in addition, and in
contrast to other chronic B-cell leukemias, HCL cells circulate at low percentages in the blood.[1] A
genetic test is therefore very useful for accurate diagnosis. The BRAF V600E mutation, the disease-
defining genetic event in HCL, is easily detected using ASO-PCR at diagnosis, and in HCL samples
containing as few as 0.1% leukemic cells from patients after therapy in complete flow-cytometric
remission.[60]

Amplification Refractory Mutation System

Amplification refractory mutation system (ARMS) is similar to ASO-PCR, exploiting the fact that
oligonucleotide primers must be perfectly annealed at their 3' ends for PCR to take place. In ARMS,
two primers are designed to anneal to the sequence/nucleotide of interest, one specific to the
mutant sequence at the 3' end, and one specific to the wild-type sequence at the 3' end. In the
detection ofJAK2 V617F, this enables homozygotes to be distinguished from heterozygotes, which
may be clinically relevant.[61]

Restriction Enzyme Digestion/Restriction Fragment Length Polymorphism Analysis

Restriction enzyme digestion can also be combined with fragment analysis to detect specific point
mutations, using fluorescently labeled primers for the PCR step. Following PCR, an enzyme is chosen
that will cut mutant sequence but not wild-type, or vice versa, and the products are electophoresed
to separate digested from intact amplicons. Care must be taken to avoid false-positive or -negative
results caused by incomplete digestion. Again this method is commonly employed for the detection
of JAK2V617F, where the mutation abolishes the BsaXI recognition site, which will therefore only
cleave wild-type JAK2 sequence.[62]

Other Mutation Detection Techniques

Often a technique is required that will scan for multiple/unknown mutations in a gene, rather than a
single specific mutation. Traditional techniques such as single-strand conformation polymorphism
analysis, denaturing gradient gel electrophoresis and denaturing high-performance liquid
chromatography analysis have been largely surpassed by high-resolution melt analysis (HRM) in
many diagnostic laboratories. HRM is sensitive, inexpensive, rapid, simple to perform and
downstream analysis is performed in a single tube immediately following PCR.

High-resolution Melt Analysis

HRM characterizes PCR products based on their denaturation (melting) behavior. Samples can be
discriminated according to their sequence, length or methylation status (following bisulfite
treatment), and the resolution is so good that even SNPs can be readily identified. The PCR is
performed in the presence of an intercalating dye, such as LCGreen or EvaGreen, which binds to
double-stranded DNA and fluoresces. Once cycling is complete, the PCR products are heated to
approximately 95oC whilst being monitored on an HRM-enabled real-time PCR instrument. As the
double-stranded product denatures, the fluorescent dye is released and the fluorescence profile
changes. This is detected as a sequence-specific melt curve, and subtle differences in melting
behavior from the normal wild-type product can be easily distinguished. Sometimes mutations
produce characteristic profiles and are thus recognized without further analysis, but often a second
technique, such as sequence analysis, will need to be performed to define the exact nature of the
change.

HRM has been used to detect mutations in exon 12 of the JAK2 gene, which are associated with the
MPN polycythemia vera. HRM was chosen by one group to replace the individual allele-specific PCR
reactions that had been developed for some of the recognized exon 12 mutations, in the knowledge
that additional mutations of exon 12 had since been identified and novel mutations may
exist.[63] HRM proved to be a more generally applicable diagnostic assay, capable of detecting at least
seven different types of exon 12 mutation, including duplications, with a sensitivity as low as 5% for
some mutations.

An HRM assay for detection of BRAF V600E mutations in HCL has been described recently. This assay
was able to detect the mutation with 100% specificity, when hairy cells were present at only 510%
in a sample.[64]

MALDI-TOF

MALDI-TOF is a form of mass spectrometry specially designed for the analysis of biomolecules, such
as nucleic acids, that can fragment when ionized by more conventional ionization methods. A laser is
fired at the matrix, which absorbs the laser energy and transfers protons to the analyte molecules,
thus charging the analyte. The mass of the molecule is calculated with great accuracy, which in turn
can enable the sequence composition of the nucleic acid to be determined based on the molecular
weights of the individual bases, and can also reveal the methylation status of the molecule. MALDI-
TOF has a sensitivity limit of approximately 510% for low-level mutation detection. This technique
has been used to identify novel AML subgroups based on distinct DNA methylation patterns, and in
turn to generate a methylation-based outcome predictor, supporting the use of genomic
methylation markers for improved molecular classification and prognostication in adult AML.[65]

Sequencing Techniques

The gold-standard method of molecular analysis is to sequence the nucleic acid to determine the
exact base composition and characterize changes at single nucleotide resolution. Sanger sequencing
on capillary sequencers has been the dominant technology for two decades and was the workhorse
of the human genome project. A major drawback, especially with regard to leukemia diagnostics and
minor cell populations, is that it can only reliably identify mutations when the fraction of mutated
alleles is greater than 1020%, although COLD-PCR (described earlier) can be used to enhance this
sensitivity. Sanger sequencing, being such a prevalent and familiar technique, will not be discussed
further, but pyrosequencing and next-generation sequencing techniques are described in some
detail below.

Pyrosequencing

In contrast to Sanger sequencing, which is based on dye terminator chemistry, pyrosequencing

involves sequencing by synthesis, and relies on the ultimate detection of pyrophosphate release on
nucleotide incorporation, which as a result of a chemical cascade is converted to light. Nucleotides
are sequentially added to the reaction, and light is produced only when the nucleotide complements
the first unpaired base of the template (Figure 6).

Figure 6.

Pyrosequencing reaction: chemiluminescent cascade to detect incorporation of nucleotides

according to the sequence of the template strand. dGTP: Deoxyguanosine triphosphate; PPi:
Pyrophosphate.
Advantages of pyrosequencing include the quantitative measurement of alleles, which is not reliable
with Sanger sequencing, simplicity and speed, as sequence data are available in real time without an
additional post-PCR analysis procedure. Pyrosequencing is also mutation tolerant; unlike
hybridization-based assays, pyrosequencing analysis generates a correct sequence regardless of the
appearance of new, unexpected mutations.[108] One of the drawbacks is the maximum read length of
only 5060 bases, making it best suited to targeted analysis of regions of interest within genes,
rather than sequencing of whole genes.

Pyrosequencing has been used for detecting mutations in some of the genes already mentioned,
such as JAK2 V617F[62] and BRAF V600E,[66] and KIT D816V, where the quantitative nature of the
pyrosequencing assay proved to be advantageous in aiding better disease classification.[67] It has also
been developed for the detection of AKD mutations conferring tyrosine kinase inhibitor resistance in
CML and ALL. Pyrosequencing is useful in this context because whilst being more sensitive than
Sanger sequencing, it has the added advantage of quantifying mutant alleles, which may provide
important clinical information on drug response.[68]

Next-generation Sequencing

Pyrosequencing is just one of a number of methodologies that have been further developed for
extremely high-throughput sequencing, techniques known collectively as next-generation
sequencing (NGS) or massively parallel sequencing. These technologies promise to transform
molecular medicine, with the potential to facilitate the cost-effective sequencing of entire genomes
in a matter of hours.

NGS holds great promise for the study of leukemia, and is already moving into clinical diagnostics; it
has been suggested that this technology has the potential to eventually replace all other genetic
analyses at diagnosis.[29] NGS has the ability to fully sequence thousands of genes in a single test and
simultaneously detect deletions, insertions, base substitutions, copy-number alterations and
translocations, including balanced rearrangements.

There are a number of different platforms for NGS on the market using different chemistries, each
with their own advantages and disadvantages, and it is beyond the scope of this review to describe
them in detail. The interested reader is directed to the excellent reviews by Shendure and Ji,[69] ten
Bosch and Grody,[70] Metzker[71] and Su et al..[72] Natrajan et al. [73] and Ross et al. [74] have recently
reviewed the application of NGS to cancer diagnosis and prognostication.

One of the first cancer genomes to be sequenced using NGS was that of an AML patient with an
apparently normal karyotype.[75] The AML genome was sequenced to a depth of >30-fold coverage,
and matched normal DNA from the patient's skin was used to exclude almost 98% of variants found
from further study on the basis that they were considered to have been inherited. The AML genome
contained ten nonsynonymous somatic mutations thought to be pathogenetically relevant; two
were known recurrent mutations in AML (FLT3-ITD and NPM1 4-bp insertion), and eight were new
mutations, all single base changes, as yet undescribed in AML yet present in virtually all tumor cells
at presentation and relapse. Half of the affected genes were already associated with cancer
pathogenesis, but not AML, and the other four somatic mutations occurred in genes not previously
implicated in cancer. All eight genes are now the focus of further study. A further 5001000
additional noncoding and nongenic somatic variants were identified in the AML genome, some of
which may in time also prove to be of clinical significance. Whole-genome sequencing may thus be
the means for discovering all of the mutations that are relevant for cancer pathogenesis.
The authors of the above study sequenced a second AML genome and investigated whether any of
the variants identified were recurrent in additional AML tumors, which in the absence of functional
validation is considered to be a good test of the relevance of individual mutations.[76] This time, 12
nonsynonymous mutations were identified as most likely to be relevant for pathogenesis, since they
could potentially alter the function of expressed genes. Three of these mutations were found in
some of the other AML samples, including mutations in NPM1 and NRAS (already associated with
AML) and the glioblastoma-associated cancer gene IDH1, which encodes a metabolic enzyme.
The IDH1 mutation was identified in 16% of samples from patients with cytogenetically normal AML,
and has since been confirmed to be a new recurrent mutation with potential prognostic significance
in AML,[77]demonstrating the potential of an unbiased sequencing approach to discover previously
unsuspected recurring mutations in cancer. Recurrent acquired mutations in DNMT3A, a de
novo DNA methyltransferase, have also been identified in AML using NGS.[7880]

NGS has since been applied to the sequencing of eight relapsed AML genomes to ascertain the
mutational spectrum associated with relapse.[81] This study revealed more novel, recurrently
mutated genes in AML and also found two major clonal evolution patterns associated with AML
relapse: either the founding clone in the primary tumor gains mutations and evolves into the relapse
clone; or a subclone of the founding clone survives initial therapy, gains additional mutations and
expands. The same group, using a similar study design, have also looked at the clonal evolution of
secondary AML from antecedent MDS to identify the genetic changes underlying
progression.[82] Other examples of NGS used to investigate the mutational spectrum of leukemia
include those of Puente et al. [83] and Wang et al. [84] in CLL.

The development of NGS strategies as diagnostic tools will lead to some novel interpretational
problems. Sequence variations that differ from reference sequences will need to be defined as
either germline or somatically acquired; and then it must be determined whether the acquired
change is likely to be a contributing oncogenic event (driver mutation) or a passenger event. For the
former, parallel analysis of germline DNA may be required to determine the somatic changes a
reasonable option technically as NGS allows DNA barcoding of each sample. The latter is in part
determined by known recurrent oncogenic mutations, but will increasingly require the development
of functional assays, such as that designed for FLT3 mutations,[85] and bioinformatics resources to
provide annotated information on the significance of each event. In addition, the expansion into
analysis of multiple genes will increase the likelihood of detecting significant germline mutations in
the gene involved with implications for predisposition to cancer which can extend to other family
members who may also carry the mutation.

NGS heralds a new era of molecular medicine and has great potential in the diagnostic setting for a
heterogeneous disease such as AML, where it could facilitate the detection of all prognostically
relevant mutations with a single test, which will in turn enable better classification of AML and a
personalized approach to treatment. NGS is already being used to detect insertions, deletions and
point mutations ofCEBPA with 397-fold to 1194-fold depth of coverage.[86] An international
consortium has assessed the robustness, precision and reproducibility of NGS for investigation
of TET2, CBL and KRAS mutations in chronic myelomonocytic leukemia in the clinical laboratory
setting.[87] The study was designed to test the utility of amplicon deep sequencing, where multiple
genes or hotspot regions are sequenced in multiple patients in a massively parallel fashion. A median
of 500-fold coverage per amplicon was achieved, and the authors concluded that the technique
shows clinical applicability, with high concordance across multiple centers, and improved sensitivity
compared with Sanger sequencing (12 vs 20%). While NGS makes sequencing of a limited
repertoire of relevant genes at very high sensitivity possible, it can also be used to sequence the
following: whole genomes, such as the examples of the AML genomes mentioned above; whole
exomes, thus concentrating on the coding regions most likely to be of clinical relevance; or even
whole transcriptomes. Exome sequencing is unable to detect most structural variants, such as
chromosomal translocations with intronic breakpoints, so may be of limited utility in leukemia,
whereas transcriptome sequencing can detect fusion transcripts produced by chromosomal
rearrangements and also provides quantitative information about gene expression levels.[88]

Expert Commentary

Since the success of the Human Genome Mapping Project, knowledge of the underlying genetic
alterations driving leukemia has expanded rapidly. A diverse range of genetic mutations has been
identified, building on previous knowledge derived initially from cytogenetic analysis acting as a
surrogate marker for gene mutations and rearrangements. In addition to gene fusions, amplification
and deletion, we now have lists of genes with aberrant expression or mutations. Mutations can be
activating or inactivating, by direct mutation or by epigenetic promotion or silencing. Mutations can
be single base changes found at multiple sites across the majority of a gene (usually inactivating) or
very specific mutations at a single codon (usually activating), and involve base changes, deletions,
insertions and duplications. The molecular analysis of these events is being successfully delivered by
a range of technical approaches which continue to evolve but the challenge remains of how to
deliver ever-increasing numbers of relevant mutation tests on each patient at an affordable cost in a
clinical timescale. Such growth of testing needs is unsustainable with the technologies commonly in
use in most laboratories hence we will need to look to more recent or new technologies to help us
deliver testing on the scale likely to be required.

Each newly identified gene and/or mutation will need to be assessed for associations, prognostic
significance and potential suitability to direct targeted therapies. The role of sample archives from
clinical trials will be of great importance in allowing rapid retrospective studies on patient cohorts
with defined treatments and outcome data.

There is a trend led by the pharmaceutical industry towards targeted therapies with approved
specific 'companion diagnostic' tests, but this is not the future. As illustrated by the discussions
above, there are many technological approaches to achieving the same result, which is often a
simple sequence change in the DNA, and a need to develop ever more informative and cheaper
technologies to deliver these results. The technology used to deliver the companion diagnostic
information should not be fixed because this will inhibit development of better or cheaper
strategies. However, robust mechanisms for external quality assessment to validate a laboratory's
ability to deliver the results will be required. The importance of quality control and standardization
should not be underestimated; comparable assay results from all laboratories performing a test for
the same analyte are critical for interpreting results and making clinical decisions. Where they exist,
this can be accomplished through: adherence to 'best practice guidelines' and formulation of
standard operating procedures; the use of quality-marked commercially available assays (in
preference to 'in-house' methods); use of reference materials; and participation in external quality
assurance schemes. Arguably the most important features for reassuring users that a clinical
laboratory can produce reliable results are its accreditation and that its staff be registered with
statutory bodies, wherever this is possible. Further information about standardization can be found
in a previous edition of this journal in an article by Holden et al. [89]

Five-year View
The potential of NGS is so great that in 5 years we can expect to see most molecular diagnosis in
leukemia to be delivered through NGS technology, replacing today's multiple single-test approach
with gene mutation panels providing multiplex, multigene, multipatient analysis (Patel et al. have
recently published an example of how this 'integrated genetic profiling' approach could be applied
for risk stratification in AML,[90]which is further discussed in the accompanying editorial[91]). Whole-
exome or whole-genome sequencing will be viewed on a rapidly approaching horizon. Layered on
top of sequence-based gene mutation analysis will be panels for gene fusions, expression of key
genes and epigenetic markers. The costs of the technologies involved will have reduced so that such
an approach is technically affordable. Such screening will produce genetic markers that will allow
effective residual disease monitoring in all patients. The changing technologies will support the
increased use of targeted therapies, the monitoring of responses to therapy and improved outcomes
for patients.