Nucleic acid quantification and disease outcome prediction in colorectal cancer
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
TECHNOLOGY R EPORT
Nucleic acid quantification and disease
outcome prediction in colorectal cancer
Stephen A Bustin
Histopathological stage at diagnosis remains the most important prognostic determinant
University of London,
Institute of Cell and for colorectal cancer. However, conventional staging is unable to predict disease outcome
Molecular Science, accurately for each individual patient. This results in considerable prognostic heterogeneity
Barts and the London, within a given tumor stage and is of particular relevance for a subgroup of patients with
Queen Mary’s School of
Medicine and Dentistry, stage II disease that would benefit from adjuvant therapy. The recent advances in
London, UK functional genomics are beginning to have a significant impact on clinical oncology, and
there is widespread interest in using molecular techniques for clinical applications. These
*Centre for Academic Surgery,
3rd Floor Alexandra Wing, have focused on two approaches: the use of polymerase chain reaction (PCR)-based
The Royal London Hospital, methods for the detection of occult disease in lymph nodes, bone marrow and blood and
London E1 1BB, UK the use of microarrays for the expression profiling of primary tumors. The aim is to develop
Tel.: +44 020 7377 7000
Ext. 2431; molecular classifiers that will allow the prediction of disease outcome, thus matching
E-mail: s.a.bustin@ patients with individualized treatment. Despite the obvious attractions of these
qmul.ac.uk approaches, there have been significant technical, biological and analytical problems in
* Address for correspondence
their translation into clinically relevant practice. This is particularly true for colorectal
cancer, the second most common cancer in the western world. Nevertheless, progress is
being made and the improved awareness and appreciation of those difficulties is
beginning to generate results that should prove useful for clinical oncology.
Colorectal cancer, the second most common clinicopathological factors. ANNs are computer-
cancer in the western world [1], is best considered based algorithms that can be used to discover com-
not as a homogeneous disease, but as covering a plex relations within data sets. They permit the rec-
spectrum in terms of its molecular properties ognition of patterns in complex biological data sets
and its clinical and pathological diversity. Unlike that cannot be detected with conventional linear
other cancers, such as those of the lung and statistical analysis by a process of error minimiza-
breast, the size of the primary tumor is often not tion through learning from experience. Although
a primary prognostic indicator. Tumors have a ANNs can increase the accuracy of colon cancer
tendency to grow slowly, enlarging gradually classification and survival prediction when com-
until they eventually penetrate the bowel wall. pared with other statistical or clinicopathological
Surgical resection of the primary cancer contin- method [4], this improvement is still not able to
ues to be the only curative option for colorectal predict survival for each individual patient. Fur-
cancer. The probability of a cure following sur- thermore, ANNs are considered to be ‘black boxes’
gery is directly related to the stage of the cancer, with lengthy development and optimization times,
and lymph node (LN) status is the most influen- as well as huge computational requirements and,
tial prognostic marker with respect to disease- since model development is empirical, ANNs pro-
free and overall survival. However, even after an vide low decision insight. Therefore, while histo-
apparent curative surgical intervention, approxi- pathological staging procedures and extensions
mately 50% of patients experience metastatic such as ANN are useful in defining the extent of
relapse that in most cases eventually leads to and prognosis for colorectal cancer and placing
death. There is considerable prognostic hetero- patients into risk groups, they are unable to accu-
geneity within each tumor stage [2]: up to a third rately identify the risk of treatment failure for every
Keywords: cancer, colon, of LN-negative (N0) patients experience treat- individual patient. Consequently, there continues
metastasis, microarray, occult ment failure and die from disseminated disease, to be widespread interest in using molecular
disease, polymerase chain
reaction, reverse transcription, whereas around a third of stage III (T1–4N1–2M0) techniques in clinical diagnostics.
staging patients are cured [3].
There are many approaches to improving the The challenge
accuracy of clinical staging. One interesting There are various approaches employing molecu-
approach utilizes artificial neural networks lar techniques that can be used to predict disease
(ANNs), trained and validated using various outcome for colorectal cancer patients. One
10.2217/17410541.3.2.207 © 2006 Future Medicine Ltd ISSN 1741-0541 Personalized Medicine (2006) 3(2), 207–216 207TECHNOLOGY REPORT – Bustin
involves attempting to detect circulating tumor expression profile [11]. Expression profiling, in
cells in blood before surgery, coupled with a particular, is expected to lead to the identifica-
molecular analysis of the LN after surgery. This tion of cellular and serum markers for colorectal
is often referred to as molecular staging and aims neoplasia. Ideally, such markers would have sev-
to augment histopathological staging with addi- eral characteristics: first, they should be
tional, patient-specific information. Molecular expressed at high levels in most tumors and at
staging can be followed up with monitoring the greatly reduced levels in normal tissues, thus pro-
appearance of circulating tumor cells in blood ducing high sensitivity and specificity. Second,
over time after surgery, with the aim of detecting any elevated expression should be an early event
the appearance of metastatic recurrence at the and remain at high levels during the neoplastic
earliest possible time. Another approach tries to transformation process. Third, markers should
associate a tumor’s metastatic potential with the be secreted to facilitate their detection. The iden-
genotypic or expression characteristics of the pri- tification of such markers, together with any cor-
mary tumor. All these methods have as their ulti- relation between specific types of genetic changes
mate goal the identification of colorectal cancer and relevant clinical information should be able
patients with actual residual disease, rather than to distinguish clinical prognostic groups.
of patients belonging to a group with a statistical
risk of harboring metastases, and hence in need PCR-based detection of occult disease
of adjuvant treatment. At the same time, these This approach has the longest track-record, and
investigations may also result in the identifica- there are numerous publications identifying a
tion of reliable markers capable of distinguishing huge range of potential prognostic markers [12,13].
between subgroups of patients that respond to Occult disease can be detected by identifying its
different treatment regimens. This review deals DNA (PCR) or its mRNA (reverse transcription
with the first aim, the second one having been [RT]-PCR) in LN, blood or bone marrow. The
dealt with recently [5,6]. targeting of DNA has the advantage of allowing
Two, not necessarily exclusive, molecular the detection of tumor-specific mutations, thus
approaches are used to address these challenges. providing direct evidence for the presence of
The first is based on the assumption that treat- tumor-derived DNA. The targeting of RNA aims
ment failure is due to the presence of occult dis- to detect colon tissue-specific mRNA in a back-
ease, sometimes referred to as ‘micrometastasis’, ground where its expression would not be
which is defined as metastatic cancer of limited expected to occur. However, while simple in prin-
disease burden that cannot be detected by stand- ciple, there are several limitations associated with
ard histopathological techniques. Occult disease the detection of occult disease. First, whilst it
may be revealed by the detection of characteristic would be ideal to be able to detect cancer-specific
mutations and/or the expression of tissue-spe- mutations, there are no known mutations that
cific genes. Since their detection requires occur consistently in colorectal cancers. Second,
extremely sensitive and specific techniques, there even when the same p53 or K-ras mutations are
has been widespread interest in polymerase chain identified in both primary cancers and LN, a sig-
reaction (PCR)-based methods capable of nificant number of patients do not develop distant
detecting such targets [7]. These involve conven- recurrences [14,15]. One reason for this may be that
tional gel-based PCR methods, as well as quanti- the detection of DNA may simply be a reflection
tative methods such as real-time PCR, of tumor burden [16,17], a problem that is not
standardized competitive reverse transcription solved by switching to the detection of tumor-cell-
(StaRT)-PCR and linear-after-the-exponential derived RNA, as even the presence of RNA is not a
(LATE)-PCR. Today, such methods provide the reliable indicator of the presence of viable cells [18].
most sensitive and specific means of detecting Third, there are some concerns regarding the lack
occult disease. of truly tissue-specific markers. For example, the
A second approach arises from the recognition widely used epithelial tissue-specific marker cytok-
that the metastatic potential of a cancer may be eratin 20 (ck20) is expressed in the blood [19–22]
encoded in the bulk of the primary tumor [8] and and LN [23] of healthy controls, with no associa-
that characteristics present in primary tumors tion between ck20 mRNA levels and distant
may predict the course of a patient's disease. recurrence or survival [24,25]. Similar reservations
Consequently, this approach aims to analyze the have been recorded for the other commonly used
primary tumor for microsatellite [9] and tissue-specific markers [12]. Fourth, conventional
ploidy [10] status, and increasingly, its gene RT-PCR assays can be very unreliable [26,27].
208 Personalized Medicine (2006) 3(2)Nucleic acid quantification and disease outcome in colorectal cancer – TECHNOLOGY REPORT
Today there are three PCR-based methods in problems [30] and its appropriate application
common use for the detection of cancer cells from depends on optimizing and validating each indi-
blood, LN and bone marrow. All three are capable vidual step that makes up a qPCR assay [31].
of quantification, thus adding an additional factor, Proper optimization of pre-assay [32] and assay [33]
to the parameters potentially useful for developing conditions, its validation [34] and standardization
a practical, clinical diagnostic assay. [35–37] are crucial to meeting the quality criteria
expected of any laboratory test applied in clinical
Real-time PCR medicine [12]. Arguably the most important over-
The most widely used technology involves the all parameter concerns assay and data standardi-
application of real-time PCR (qPCR) assays [28], zation. A recent survey of qRT-PCR protocols
which represent a significant advance on conven- and analysis procedures revealed huge variability,
tional PCR-based methods [29]. Quantification is even between diagnostic laboratories [38]. Differ-
based on the detection of a fluorescent reporter ences in the pre-analytical steps, such as blood
molecule, and the increase in the amount of flu- sample processing [39,40] and RNA isolation [41]
orescence as PCR products accumulate with lead to heterogeneous results even if evaluated in
each cycle of amplification. There are three main the same laboratory. Similarly, the differences in
chemistries in general use: analytical variables that affect the accuracy of
• Intercalating dyes, such as SYBR®-Green conventional PCR assays [42] affect both qPCR
(Invitrogen, CA, USA), which fluoresce upon assays [43,44]. Standardized data are necessary
light excitation when bound to double when making interexperimental and interlabora-
stranded DNA. These are cheap, easily added tory comparison of gene expression measure-
to legacy assays and amplification products ments, and it is near impossible to develop a
can be verified by the use of melt curves. On meaningful gene expression database with
the other hand, they can lack specificity and nonstandardized data.
fluorescence varies with amplicon length.
StaRT-PCR
• Primers linked to fluorophores, for example,
The problem of data standardization is addressed
Lux™ primers (Invitrogen) or Plexor™
by an alternative method, termed StaRT-PCR [45].
(Promega, CA, USA). Fluorescence detection
StaRT-PCR is a quantitative method for measur-
depends on the primers binding to, and being
ing multigene [46] expression using standardized
extended during the PCR reaction. These are
mixtures of competitive templates that include
cheap, and amplification products can be veri-
defined reference templates for β-actin and glycer-
fied by melt curves. However, specificity
aldehyde-3-phosphate dehydrogenase (GAPDH).
depends on the primers and company-specific
Since these reference competitive templates are
design software needs to be used for optimal
included in each reaction, any expression value
performance.
can be related to the expression of one or both of
• Hybridization-probe-based methods, for these two genes. The inclusion of the reference
example, TaqMan® (Applied Biosystems) or genes in each mixture also allows for direct com-
molecular beacons. These are the most spe- parison of values obtained with different mixtures,
cific methods, as products are only detected if or different experiments. The competitive tem-
the probes hybridize to the appropriate plates for each gene are commercially designed,
amplification products. There are many vari- manufactured and validated. Each gene expres-
ations on this theme, with melt curve analysis sion result is reported as the number of molecules
possible for some chemistries. Their main dis- of mRNA for gene ‘X’ per 106 molecules of refer-
advantages are cost, complexity and occa- ence gene. Serial dilutions of the standardized
sional fragility of probe synthesis and mixes allow quantitative measurements over a 6.5
potential problems associated with the fact log range. All data are directly comparable as data
that probe-based assays do not report primer are measured using standardized lots of the inter-
dimers that can interfere with the efficiency nal standards. With such standardized data,
of the amplification reaction. expression levels of genes in diseased vs normal tis-
If real-time PCR is preceded by a RT step to sue can be compared across experiments and other
quantitate mRNA or viral RNA targets, this laboratories. In addition, an internal standard
method is often termed qRT-PCR or, alterna- inherently corrects for any inhibitors that may be
tively RT-qPCR. Not surprisingly, whilst simple present in tissue extracts, a problem that requires
in concept, this technology has its own additional assays with real-time PCR analysis. The
www.futuremedicine.com 209TECHNOLOGY REPORT – Bustin
establishment of a gene expression database that is patients remains unclear. This is due to concep-
annotated with respect to phenotype (e.g., respon- tual reservations, as well as technical limitations
siveness to particular chemotherapeutic agents) that interfere with the diagnostic specificity of
makes it possible to compare and evaluate the util- qRT-PCR assays. The lack of clinical validation
ity of gene expression data from new clinical spec- of RT-PCR-based assays in prospective studies
imens [47]. StaRT-PCR has been used to develop a suggests that although RT-PCR assays can iden-
method that is 100% accurate for diagnosing lung tify LN harboring colon-derived cells that are
cancer in cytological specimen, improving on the not detected by conventional histopathology,
80% sensitivity rate currently obtained for mor- these are not likely to progress to distant metas-
phological criteria [48]. Results obtained using tases. This implies a conceptual flaw, based on a
StaRT-PCR correlate well with those obtained simplistic view of the biology and kinetics of
using TaqMan assays [49], and this technology may tumor cell traffic through the lymphatic and sys-
be better for high-throughput quantitative temic circulation and subsequent metastasis
transcript measurements. development [12]. Critically, RT-PCR detection
strategies do not provide any information about
LATE-PCR the metastatic potential of the cells they detect.
A third alternative is LATE-PCR, which is particu- Even unequivocal detection of tumor cells by
larly useful when targeting low copy number tar- mutant allele-specific PCR fails to predict relapse
gets [50]. This method is similar to asymmetric within 5 years of surgery in between 27 [14] and
PCR, in that the two PCR primers are used at dif- 70% [15] of patients. Indeed, a patient in the LN-
ferent concentrations. However, LATE-PCR takes negative group also died. In general, when
into account the difference in melting temperature RT-PCR assays are used to detect ‘tissue-specific’
(Tm) of two primers present at different concentra- mRNA expression, tumor recurrence is observed
tions, thus improving both efficiency and specifi- in only 14–50% of patients whose LN are
city. LATE-PCR also makes it possible to use RT-PCR positive. Together with the results
lower temperature detection, since the probe does described above, this suggests, as has been previ-
not need to compete with hybridization and ously asserted [52], that while RT-PCR may be
extension of the limiting primer during the early, detecting cancer cells, the majority of these may
exponential phase of the reaction. Hybridization not have any metastatic potential. Hence their
of probe and target is unimpeded once the limit- detection may be no more prognostic for the
ing primer is depleted and can be done either by individual patient than conventional staging.
lowering the annealing temperature at that point,
or by introducing a low-temperature detection Microarrays
step between the extension and melting steps. DNA microarrays rely on nucleic acid hybridiza-
Probes with lower melting temperatures are easier tion and the use of nucleic acid polymers, immo-
to design, more allele discriminating, and have bilized on a solid surface, as probes for
lower background fluorescence. Moreover, as the complementary nucleic acid target sequences.
probe dissociates from its target strand well below They can be used to monitor simultaneously the
the extension temperature of the reaction, suffi- expression of thousands of genes from human
cient probe can be added to the reaction to meas- tumor samples and can be applied to large num-
ure all product strands without inhibiting the bers of samples in parallel. There are two plat-
amplification reaction. Consequently, LATE-PCR forms, based on cDNA or oligonucleotide
increases the signal strength and allele discrimina- microarrays. cDNA microarrays consist of cDNA
tion capability of fluorescent probes and reduces clone inserts representing genes of interest spotted
variability among replicate samples and provides a onto a solid support. They are commonly queried
means for improving the accuracy of single-cell with cDNA derived from experimental and refer-
genetic diagnosis [51]. Therefore, it is suited for ence RNA samples that have been differentially
detecting and/or quantitating nucleic acids from labeled with two fluorophores to allow for the
the very limited amounts of tumor material quantification of differential gene expression, and
available for analysis following a resection. expression values are reported as ratios between
two fluorescent values. Oligonucleotide arrays use
Usefulness of PCR-based methods a single fluorescent label and experimental RNA is
The prognostic value of detecting disease-associ- amplified, biotinylated for detection, hybridized
ated mRNA from circulating colorectal cancer and detected through the binding of a fluorescent
cells in the blood, LN or bone marrow of streptavidin-bound compound [53].
210 Personalized Medicine (2006) 3(2)Nucleic acid quantification and disease outcome in colorectal cancer – TECHNOLOGY REPORT
There have been many attempts to find novel non-metastatic tumors [61]. The authors identi-
markers to identify patients with the potential for fied a 200-odd gene set that divided patients
colon cancer progression [54]. The concept of dis- with significantly different 5-year survival rates.
ease classifiers based on differential gene expression Importantly, most non-metastatic tumors that
supposes that metastatic potential is associated clustered with metastatic cases eventually devel-
with the expression levels of a given set of genes. oped metastasis, confirming the notion that met-
Hence the characterization of global mRNA pat- astatic potential can be predicted from the
terns is expected to reveal clues about regulatory transcriptome of the primary tumor. Discrimi-
mechanisms, biochemical pathways and broader nator genes were associated with various cellular
cellular function and so advance the understand- processes (e.g., cell-cycle regulation, cell adhe-
ing of disease pathology and increase the accuracy sion and angiogenesis). However, in common
of staging. Furthermore, transcriptome analysis with other findings they do not point to an obvi-
could lead to the identification of key markers ous discriminatory mechanism for metastasis.
among the complex network of gene products Although many of the genes identifying LN
involved in metastasis and their association with metastasis and predicting distant metastasis are
previously undetectable features of the molecular the same, there are differences. This is a reflec-
basis of individual tumor characteristics. tion of the clinical observation that the two are
not perfectly correlated and emphasizes expres-
Profiling of colorectal cancers sion of the underlying biological differences.
Since the first report comparing expression pro- Another recent study examined the expression
files of tumor and normal colon tissue [55], there profile from 74 patients with stage II colorectal
have been several studies analyzing expression cancer and identified a 23-gene signature profile
profiles from different disease stages. Two have that predicted recurrence. The signature was val-
analyzed molecular differences between adeno- idated in 36 independent patients with an over-
mas and carcinomas and identified 1800 [56] and all accuracy of 78%: 13 out of 18 patients with
50 [57] discriminating genes capable of distin- relapse and 15 of 18 disease-free patients were
guishing the two stages. Another study reported correctly predicted, representing an odds ratio of
the construction of a ‘colonochip’, a microarray 13 [11]. A third report used a 43-gene signature to
specifying some 4600 genes that are expressed in predict 3-year survival with a 90% accuracy
colorectal cancer, normal colonic mucosa and (93% sensitivity and 84% specificity) in predict-
liver metastatic cancer tissues and its use to ing 36-month overall survival in 78 patients and,
identify expression profiles associated with most importantly, was able to discriminate a sur-
colorectal tumor progress. Expression profiling vival difference in an independent test set [62]. In
of highly metastatic cell lines relative to their another study, a 20-gene expression profile that
poorly metastatic parental cell line has identified predicted distant recurrence was obtained from
metastasis-associated expression pattern changes the primary tumors of stage III patients that had
in 170 genes [58]. Another study compared the not received any adjuvant therapy [63]. Although
expression profiles of LN-positive and LN-nega- the ability to predict a poor clinical prognosis
tive cancers using a 4608 cDNA array and iden- does not necessarily indicate a benefit from an
tified 60 genes that could correctly place 10/12 intervention, it certainly seems capable of defin-
cancers into their appropriate groups [59]. Yet ing a group of patients that will not require any
another study collected primary and metastatic adjuvant treatment.
tissue from patients undergoing simultaneous
primary resection from the colon and metastatic Usefulness of microarray-based methods
resection from the liver [60]. Samples were iso- The identification of genes associated with
lated using laser capture microdissection (LCM) metastasis demonstrates the feasibility of com-
and analyzed using a 9121-gene cDNA microar- bining large-scale molecular analysis of expres-
ray. A total of 40 genes were upregulated signifi- sion profiles with classic morphological and
cantly, while seven genes were downregulated clinical methods of staging and grading cancer
significantly in metastatic as compared with for better outcome prediction. The studies
primary tumor. above suggest that global expression profiling is
A recent report addresses the issue of progno- beginning to identify patterns of gene expres-
sis and suggests that expression profiling of sion that appear to correlate with metastatic
colorectal cancers is able to distinguish clinically potential and may identify molecular markers
relevant subgroups and metastatic versus for colon cancer prognosis.
www.futuremedicine.com 211TECHNOLOGY REPORT – Bustin
However, there are a number of limitations of the host [69]. In addition, there is an important
that result in uncertainty concerning the reliabil- contribution from the tumor microenvironment
ity of DNA microarray measurements. First, and metastasis can be modified by stromal
since the hybridization conditions are variable events. Consequently, it is plausible that multi-
for different genes, accurate measurements of ple signatures could be a consequence of the
absolute expression levels and the reliable detec- multiple pathways through which colorectal
tion of low-abundance genes are currently metastases arise [70], as demonstrated by the fact
beyond the reach of microarray technology. that LN and liver metastases from the same
Consequently, although the direction of change patient do not always show the same genetic
indicated by microarray experiments is usually aberrations [71]. This implies that the accuracy of
reliable, the magnitude of gene expression multigenic classifiers may be improved if their
changes are much less so [54]. However, it is likely selection is based on better understanding of the
that technological advances in extraction, underlying tumor biology. The finding of a
hybridization and detection methods will molecular signature in primary cancers predic-
improve the measurement potential of micro- tive of metastasis [8] suggests that the metastatic
arrays. Second, differential gene expression and potential of many cancers is encoded in the bulk
prognostic value of expression patterns are not of the primary tumor and raises the expectation
necessarily synonymous and there is little con- that this will result in a prognostic application.
sistency between studies used to predict clinical Nonetheless, it is notable that colorectal cancers
outcome [64]. A recent tissue microarray study were not included in the demonstration of clini-
was unable to show any association between the cal utility of this metastasis-associated signature,
expression profiles of several cell cycle regulatory although another recent study does describe such
or proliferation-related markers previously corre- a signature for stage III colorectal cancer
lated with prognostic relevance and disease-free patients [63]. Interestingly, this study also sug-
survival in node-negative rectal cancers treated gests that detection of the expression of a single
by surgery alone [65]. It is apparent that single mRNA, the ras homolog gene family member A
microarray data are prone to false findings [66] (RHOA), can identify a subset of patients that
and that different studies result in different gene may benefit from adjuvant chemotherapy. How-
lists; indeed different gene selection methods can ever, this depended on selecting appropriate cut-
lead to strikingly different gene lists from the off levels for mRNA levels, with the associated
same experiment [67]. This may be caused by problems this entails [12].
technical variability, where the methods by Cancer analysis by expression profiling cannot
which the classifiers are extracted from the afford to ignore the molecular complexity of
microarray data generate variable results. In this colorectal cancer. There is extensive intratumor
case the use of appropriate ‘models of consist- genetic heterogeneity in colorectal cancers, with
ency’ and standard operating procedures may between two and six different clonal genotypes
resolve the underlying biological phenomena of per tumor [72,73], with obvious implications for
the experiments. Furthermore, the establishment expression heterogeneity. Furthermore, gene
of gene coexpression networks for functionally expression profiles of colorectal cancers are
related genes is also likely to improve the biolog- affected by environmental conditions, such as dif-
ical validity of microarray data. A recent study ferences in diet or the preparation of the bowel
combined independent datasets on different prior to surgery. It must also be remembered that
types of cancer to explore transcriptional changes colorectal cancer biopsies are composed of epithe-
in terms of gene interactions rather than at the lial cells, as well as numerous other cell types. Var-
level of individual genes [68]. Two distinct net- iable proportions of these nonepithelial and non-
works were able to detect biological changes and malignant cells could lead to inconsistent and bio-
identify gene groups whose co-regulation might logically inaccurate gene expression patterns.
contribute to malignant transformation. Analysis of microdissected cell populations partly
On the other hand, recent evidence implies a addresses these problems, and has been applied to
significant role in the metastatic process for the analysis of profiles associated with colorectal
germline polymorphisms. In practice this means cancer development [74]. However, analysis of
that malignancy is influenced not just by envi- microdissected cell populations has its own disad-
ronmental stimuli, and multiple genetic and epi- vantages; a significant proportion of metastasis-
genetic events that arise within the malignant associated signatures appear to be derived from
epithelium, but also by the genetic background the nonepithelial component of the tumor and
212 Personalized Medicine (2006) 3(2)Nucleic acid quantification and disease outcome in colorectal cancer – TECHNOLOGY REPORT
microdissection would result in the loss of that marker will be validated if it is able to predict the
component [8]. The case for considering expres- risk of distant recurrence accurately for every indi-
sion profiles of both epithelial and stromal cells is vidual patient. Whether this will turn out to be a
further strengthened by the extraordinary plastic- blood or LN-based marker, a multigene mRNA
ity of cells: signals arising in a tumor can repro- expression profile obtained from primary tumors,
gram both normal and cancer cells. Cancer cells or protein/peptide or metabolome profiles
may be reprogrammed to take on the structure obtained from primary tumor, LN or blood,
and function of vascular cells in a process called remains an issue open for debate. However, there
vascular mimicry [75] and tumors can organize can be no doubt that in the not-too-distant future
regional fibroblasts and endothelial cells into a there will be a more accurate classification system
collaborating metabolic domain that is able to sus- that will allow accurate prognostic stratification of
tain cancer cell survival [76]. All these factors may individual patients and the administration of
generate noise that could mask expression patterns molecular therapeutics that show optimal activity
relevant for outcome prediction. in particular subgroups of patients.
Conclusion & expert commentary Outlook
Colorectal cancer is a complex disease arising from Host genetics plays an increasingly recognized
the interplay of genetic and environmental fac- critical role in determining tumor behavior as
tors. The achievements of genome sequencing well as patient response to adjuvant therapies.
projects and technological advances in functional Consequently, new molecular therapeutics are
genomics have facilitated the transition to large- likely to find optimal activity in particular, well-
scale systemic approaches to studying this disease defined subgroups of patients. Furthermore,
and, undoubtedly, the quantification of mRNA whereas there are numerous tumor-type specific
by RT-PCR methods and expression profiling by changes on the path to malignancy and metasta-
microarrays has provided important insights into sis, it appears that there may be fewer, and com-
its biology. Less clear is whether these technologies mon changes in the cancer microenvironment,
are more accurate at predicting disease outcome raising the hope that therapeutic targeting of
for individual patients at risk of treatment failure these events could be generally applicable [77].
following supposedly curative surgery. First, they Interestingly, if polymorphisms, rather than
generate yet more information without any clear oncogenic events, induce the metastasis-predic-
evidence that this is any more relevant for patient tive gene signatures, then it may not be necessary
management. The results obtained from using to obtain tumor biopsies for prognosis. Instead,
RT-PCR to quantitate ‘tissue-specific’ markers, in normal colonic biopsies or even more easily
particular, are not compelling. Second, there are accessible tissues, for example, blood, could be
numerous technical issues that need to be useful for identifying patients at risk before they
addressed and it is essential that technical variabil- develop tumors. This issues are well worth exam-
ity is minimized by using standard operating pro- ining and it is essential to obtain epidemiological
cedures that implement standards at each step of evidence linking human polymorphisms with
the procedures, especially with regard to careful metastasis. In any event, the molecular networks
quality control of experiments with associated involved in regulating the metastatic cascade are
quality metrics. Again, the biological relevance of slowly being untangled and molecular medicine,
data obtained using real-time PCR, the most by helping with more accurate prognostic strati-
widely used RT-PCR technology, is obscured by fication of colorectal cancer patients, will play a
its lack of any meaningful standardization. An leading role in the move toward individualized,
alternative PCR-based method, such as STaRT- genome-based, combinatorial cancer therapy. In
PCR may well be a more useful tool due to its combination with an increasing array of addi-
very standardized protocol. Third, larger studies tional tools, such as tissue-specific genetic
with appropriate clinical design, adjustment for knockouts, siRNA-mediated knockdown, the
known predictors and correct validation inde- development of protein, tissue and interaction
pendent data sets are essential before clinical deci- arrays, these serve to increase our understanding
sions can be based on results obtained from either of the metastatic process and the role played by
technology. Ultimately, any set of molecular the tumor microenvironment.
www.futuremedicine.com 213TECHNOLOGY REPORT – Bustin
Highlights
• Colorectal cancer describes a disease spectrum characterized by genetic and epigenetic heterogeneity.
• There is a clear association between different subtypes of colorectal cancer and patient prognosis.
• Molecular techniques such as polymerase chain reaction (PCR) and microarray analysis are being used in attempts to improve the
accuracy of conventional histopathological staging for individual patients.
• While numerous individual markers and several expression profiles have been reported as being independent predictors of disease
outcome, none have been universally validated.
• The acceptance of any markers into clinical practice must be supported by reproducible validation using standardized assay
conditions or diagnostic criteria in multiple independent large sets of samples, as well as by different laboratories.
• The identification of the role germline polymorphisms have in cancer progression suggests it will be possible to match therapy to
individuals, based on their susceptibilities and response profiles.
Bibliography carcinoma: a biologically distinct and individuals. Clin. Chem. 48, 1212–1217
Papers of special note have been highlighted as aggressive subset of sporadic colorectal (2002).
either of interest (•) or of considerable interest (••) cancer. Br. J. Cancer 84, 232–236 (2001). 19. Bustin SA, Gyselman VG, Williams NS,
to readers. 11. Wang Y, Jatkoe T, Zhang Y et al.: Gene Dorudi S: Detection of cytokeratins 19/20
1. Boyle P, Ferlay J: Cancer incidence and expression profiles and molecular markers to and guanylyl cyclase C in peripheral blood
mortality in Europe, 2004. Ann Oncol. 16, predict recurrence of Dukes' B colon cancer. of colorectal cancer patients. Br. J. Cancer
481–488 (2005). J. Clin. Oncol. 22, 1564–1571 (2004). 79, 1813–1820 (1999).
2. Jemal A, Murray T, Ward E et al.: Cancer • Identification of a set of prognostic 20. Wyld DK, Selby P, Perren TJ et al.:
statistics, 2005. CA Cancer J. Clin. 55, markers for disease recurrence in lymph Detection of colorectal cancer cells in
10–30 (2005). node negative cancers. peripheral blood by reverse-transcriptase
3. Hermanek P: pTNM and residual tumor 12. Bustin SA, Mueller R: Real-time reverse polymerase chain reaction for cytokeratin
classifications: problems of assessment and transcription PCR (qRT-PCR) and its 20. Int. J. Cancer 79, 288–293 (1998).
prognostic significance. World J. Surg. 19, potential use in clinical diagnosis. Clin. Sci. 21. Champelovier P, Mongelard F,
184–190 (1995). (Lond.) 109, 365–379 (2005). Seigneurin D: CK20 gene expression:
4. Ahmed FE: Artificial neural networks for 13. Bustin SA, Mueller R: Real-time reverse technical limits for the detection of
diagnosis and survival prediction in colon transcription PCR and the detection of circulating tumor cells. Anticancer Res. 19,
cancer. Mol. Cancer 4, 29 (2005). occult disease in colorectal cancer. Mol. 2073–2078 (1999).
5. Ahmed FE: Molecular markers that predict Aspects Med. 27, 192–223 (2006). 22. Wharton RQ, Jonas SK, Glover C et al.:
response to colon cancer therapy. Expert Rev. 14. Hayashi N, Ito I, Yanagisawa A et al.: Increased detection of circulating tumor
Mol. Diagn. 5, 353–375 (2005). Genetic diagnosis of lymph-node metastasis cells in the blood of colorectal carcinoma
6. Ahmed FE: Development of novel in colorectal cancer. Lancet 345, 1257–1259 patients using two reverse transcription-
diagnostic and prognostic molecular (1995). PCR assays and multiple blood samples.
markers for sporadic colon cancer. Expert •• The first description of the use of a Clin. Cancer Res. 5, 4158–4163 (1999).
Rev. Mol. Diagn. 5, 337–352 (2005). polymerase chain reaction (PCR)-based 23. Bustin SA, Gyselman VG, Siddiqi S,
7. Bustin SA, Dorudi S: Molecular assessment assay to try and improve staging accuracy Dorudi S: Cytokeratin 20 is not a
of tumor stage and disease recurrence using in colorectal cancer. tissue-specific marker for the detection of
PCR-based assays. Mol. Med. Today 4, 15. Clarke GA, Ryan E, Crowe JP, O'Keane JC, malignant epithelial cells in the blood of
389–396 (1998). MacMathuna P: Tumor-derived mutated colorectal cancer patients. Int. J. Surg. Invest.
8. Ramaswamy S, Ross KN, Lander ES, K-ras codon 12 expression in regional 2, 49–57 (2000).
Golub TR: A molecular signature of lymph nodes of stage II colorectal cancer 24. Bustin SA, Siddiqi S, Ahmed S, Hands R,
metastasis in primary solid tumors. Nature patients is not associated with increased risk Dorudi S: Quantification of cytokeratin 20,
Genet. 33, 49–54 (2003). of cancer-related death. Int. J. Colorectal Dis. carcinoembryonic antigen and guanylyl
•• Describes the identification of an 16, 108–111 (2001). cyclase C mRNA levels in lymph nodes may
expression signature in the early stages of 16. Anker P, Lefort F, Vasioukhin V et al.: K-ras not predict treatment failure in colorectal
several solid tumor types that support the mutations are found in DNA extracted from cancer patients. Int. J. Cancer 108, 412–417
idea that some primary tumors are the plasma of patients with colorectal cancer. (2004).
preconfigured to metastasize, and that this Gastroenterology 112, 1114–1120 (1997). 25. Oberg AN, Lindmark GE, Israelsson AC,
propensity is detectable at the time of 17. Lecomte T, Berger A, Zinzindohoue F et al.: Hammarstrom SG, Hammarstrom ML:
initial diagnosis. Detection of free-circulating tumor- Detection of occult tumor cells in lymph
9. Ribic CM, Sargent DJ, Moore MJ et al.: associated DNA in plasma of colorectal nodes of colorectal cancer patients using
Tumor microsatellite-instability status as a cancer patients and its association with real-time quantitative RT-PCR for CEA and
predictor of benefit from fluorouracil-based prognosis. Int. J. Cancer 100, 542–548 CK20 mRNAs. Int. J. Cancer 111, 101–110
adjuvant chemotherapy for colon cancer. (2002). (2004).
N. Engl. J. Med. 349, 247–257 (2003). 18. Ng EK, Tsui NB, Lam NY et al.: Presence of 26. Keilholz U, Willhauck M, Rimoldi D et al.:
10. Hawkins NJ, Tomlinson I, Meagher A, filterable and nonfilterable mRNA in the Reliability of reverse transcription-
Ward RL: Microsatellite-stable diploid plasma of cancer patients and healthy polymerase chain reaction (RT-PCR)- based
214 Personalized Medicine (2006) 3(2)Nucleic acid quantification and disease outcome in colorectal cancer – TECHNOLOGY REPORT
assays for the detection of circulating tumor enrichment methods. Clin. Chem. 47, standardized-competitive RT-PCR for
cells: a quality-assurance initiative of the 1860–1863 (2001). transcript quantification and its comparison
EORTC Melanoma Cooperative Group. 40. Vlems F, Soong R, Diepstra H et al.: Effect with real time RT-PCR. Mol. Cancer 3, 5
Eur. J. Cancer 34, 750–753 (1998). of blood sample handling and reverse (2004).
• Demonstration of the lack of reliability of transcriptase-polymerase chain reaction 50. Sanchez JA, Pierce KE, Rice JE, Wangh LJ:
conventional reverse transcriptase assay sensitivity on detection of CK20 Linear-after-the-exponential (LATE)-PCR:
(RT-PCR)-based assay. expression in healthy donor blood. Diagn. an advanced method of asymmetric PCR
27. Lacroix J, Doeberitz MK: Technical aspects Mol. Pathol. 11, 90–97 (2002). and its uses in quantitative real-time
of minimal residual disease detection in 41. Max N, Willhauck M, Wolf K et al.: analysis. Proc. Natl Acad. Sci. USA 101,
carcinoma patients. Semin. Surg. Oncol. 20, Reliability of PCR-based detection of occult 1933–1938 (2004).
252–264 (2001). tumor cells: lessons from real-time RT-PCR. 51. Hartshorn C, Anshelevich A, Wangh LJ:
28. Heid CA, Stevens J, Livak KJ, Williams PM: Melanoma Res. 11, 371–378 (2001). Rapid, single-tube method for quantitative
Real time quantitative PCR. Genome Res. 6, 42. Zippelius A, Lutterbuse R, Riethmuller G, preparation and analysis of RNA and DNA
986–994 (1996). Pantel K: Analytical variables of reverse in samples as small as one cell. BMC
29. Bustin SA: Absolute quantification of transcription-polymerase chain reaction- Biotechnol. 5, 2 (2005).
mRNA using real-time reverse transcription based detection of disseminated prostate 52. Zetter BR: The cellular basis of site-specific
polymerase chain reaction assays. J. Mol. cancer cells. Clin. Cancer Res. 6, 2741–2750 tumor metastasis. N. Engl. J. Med. 322,
Endocrinol. 25, 169–193 (2000). (2000). 605–612 (1990).
30. Bustin SA, Nolan T: Pitfalls of quantitative 43. Vlems FA, Diepstra JH, Cornelissen IM 53. Ramaswamy S, Golub TR: DNA
real-time reverse-transcription polymerase et al.: Limitations of cytokeratin 20 microarrays in clinical oncology. J. Clin.
chain reaction. J. Biomol. Tech. 15, 155–166 RT-PCR to detect disseminated tumor cells Oncol. 20, 1932–1941 (2002).
(2004). in blood and bone marrow of patients with 54. Shih W, Chetty R, Tsao MS: Expression
31. Bustin SA: Quantification of mRNA using colorectal cancer: expression in controls and profiling by microarrays in colorectal cancer.
real-time reverse transcription PCR (RT- downregulation in tumor tissue. Mol. Pathol. Oncol. Rep. 13, 517–524 (2005).
PCR): trends and problems. J. Mol. 55, 156–163 (2002). 55. Alon U, Barkai N, Notterman DA et al.:
Endocrinol. 29, 23–39 (2002). 44. Raggi CC, Verderio P, Pazzagli M et al.: Broad patterns of gene expression revealed
32. Muller MC, Hordt T, Paschka P et al.: An Italian program of external quality by clustering analysis of tumor and normal
Standardization of preanalytical factors for control for quantitative assays based on colon tissues probed by oligonucleotide
minimal residual disease analysis in chronic real-time PCR with Taq-Man probes. Clin. arrays. Proc. Natl Acad. Sci. USA 96,
myelogenous leukemia. Acta Haematol. 112, Chem. Lab. Med. 43, 542–548 (2005). 6745–6750 (1999).
30–33 (2004). • Demonstration of the need for strict • The first report comparing expression
33. Bustin SA: A-Z of quantitative PCR. IUL quality control of real-time PCR assays. profiles in normal colon and colon cancer.
Press, La Jolla, CA, USA (2004). 45. Willey JC, Crawford EL, Jackson CM et al.: 56. Notterman DA, Alon U, Sierk AJ,
34. Afzal MA, Osterhaus AD, Cosby SL et al.: Expression measurement of many genes Levine AJ: Transcriptional gene expression
Comparative evaluation of measles virus- simultaneously by quantitative RT-PCR profiles of colorectal adenoma,
specific RT-PCR methods through an using standardized mixtures of competitive adenocarcinoma, and normal tissue
international collaborative study. J. Med. templates. Am. J. Respir. Cell Mol. Biol. 19, examined by oligonucleotide arrays. Cancer
Virol. 70, 171–176 (2003). 6–17 (1998). Res. 61, 3124–3130 (2001).
35. Niesters HG: Quantitation of viral load • Application of a fully standardized PCR 57. Lin YM, Furukawa Y, Tsunoda T et al.:
using real-time amplification techniques. technology to expression quantification. Molecular diagnosis of colorectal tumors by
Methods 25, 419–429 (2001). 46. Crawford EL, Warner KA, Khuder SA, expression profiles of 50 genes expressed
36. Gabert J, Beillard E, van der Velden VH Zahorchak RJ, Willey JC: Multiplex differentially in adenomas and carcinomas.
et al.: Standardization and quality control standardized RT-PCR for expression analysis Oncogene 21, 4120–4128 (2002).
studies of 'real-time' quantitative reverse of many genes in small samples. Biochem. 58. Hegde P, Qi R, Gaspard R et al.:
transcriptase polymerase chain reaction of Biophys. Res. Commun. 293, 509–516 Identification of tumor markers in models
fusion gene transcripts for residual disease (2002). of human colorectal cancer using a 19,200-
detection in leukemia – a Europe Against 47. Willey JC, Crawford EL, Knight CR et al.: element complementary DNA microarray.
Cancer program. Leukemia 17, 2318–2357 Standardized RT-PCR and the standardized Cancer Res. 61, 7792–7797 (2001).
(2003). expression measurement center. Methods 59. Kwon HC, Kim SH, Roh MS et al.: Gene
37. Niesters HG: Molecular and diagnostic Mol. Biol. 258, 13–41 (2004). expression profiling in lymph node-positive
clinical virology in real time. Clin. 48. Willey JC, Knight CR, Crawford EL et al.: and lymph node-negative colorectal cancer.
Microbiol. Infect. 10, 5–11 (2004). Use of standardized reverse transcription- Dis Colon Rectum 47, 141–152 (2004).
38. Bustin SA: Real-time, fluorescence-based polymerase chain reaction and the 60. Yanagawa R, Furukawa Y, Tsunoda T et al.:
quantitative PCR: a snapshot of current standardized expression measurement center Genome-wide screening of genes showing
procedures and preferences. Expert Rev. Mol. in multi-institutional trials to develop altered expression in liver metastases of
Diagn. 5, 493–498 (2005). meaningful lung cancer classification based human colorectal cancers by cDNA
39. Ladanyi A, Soong R, Tabiti K, Molnar B, on molecular genetic criteria. Chest 125, microarray. Neoplasia 3, 395–401 (2001).
Tulassay Z: Quantitative reverse 155S–156S (2004). 61. Bertucci F, Salas S, Eysteries S et al.: Gene
transcription-PCR comparison of tumor cell 49. Pagliarulo V, George B, Beil SJ et al.: expression profiling of colon cancer by
Sensitivity and reproducibility of DNA microarrays and correlation with
www.futuremedicine.com 215TECHNOLOGY REPORT – Bustin
histoclinical parameters. Oncogene 23, biological themes within lists of genes with
1377–1391 (2004). EASE. Genome Biol. 4, R70 (2003). 73. Losi L, Baisse B, Bouzourene H,
62. Eschrich S, Yang I, Bloom G et al.: 68. Choi JK, Yu U, Yoo OJ, Kim S: Differential Benhattar J: Evolution of intratumoral
Molecular staging for survival prediction of coexpression analysis using microarray data genetic heterogeneity during colorectal
colorectal cancer patients. J. Clin. Oncol. 23, and its application to human cancer. cancer progression. Carcinogenesis 26,
3526–3535 (2005). Bioinformatics 21, 4348–4355 (2005). 916–922 (2005).
63. Arango D, Laiho P, Kokko A et al.: Gene- 69. Hunter K: Host genetics influence tumor 74. Kitahara O, Furukawa Y, Tanaka T et al.:
expression profiling predicts recurrence in metastasis. Nature Rev. Cancer 6, 141–146 Alterations of gene expression during
Dukes' C colorectal cancer. Gastroenterology (2006). colorectal carcinogenesis revealed by cDNA
129, 874–884 (2005). •• Lucid and thought-provoking review microarrays after laser-capture
• Identification of a set of prognostic describing the critical role played by the microdissection of tumor tissues and normal
markers for disease recurrence in lymph host genetic background in the epithelia. Cancer Res. 61, 3544–3549
node (LN)-positive tumors. development of metastases. (2001).
64. Bustin SA, Dorudi S: The value of 70. Boedefeld WM 2nd, Bland KI, Heslin MJ: • Demonstration of the use of laser capture
microarray techniques for quantitative gene Recent insights into angiogenesis, apoptosis, microdissection for identifying epithelium-
profiling in molecular diagnostics. Trends invasion, and metastasis in colorectal specific expression patterns.
Mol. Med. 8, 269–272 (2002). carcinoma. Ann. Surg. Oncol. 10, 839–851 75. Hendrix MJ, Seftor EA, Meltzer PS et al.:
65. Hoos A, Nissan A, Stojadinovic A et al.: (2003). Expression and functional significance of
Tissue microarray molecular profiling of 71. Al-Mulla F, Keith WN, Pickford IR, VE-cadherin in aggressive human melanoma
early, node-negative adenocarcinoma of the Going JJ, Birnie GD: Comparative genomic cells: role in vasculogenic mimicry. Proc.
rectum: a comprehensive analysis. Clin hybridization analysis of primary colorectal Natl Acad. Sci. USA 98, 8018–8023 (2001).
Cancer Res. 8, 3841–3849 (2002). carcinomas and their synchronous 76. Koukourakis MI, Giatromanolaki A,
• Describes the problems involved with metastases. Genes Chromosomes Cancer 24, Harris AL, Sivridis E: Comparison of
trying to associate expression profiles 306–314 (1999). metabolic pathways between cancer cells
with prognosis. 72. Baisse B, Bouzourene H, Saraga EP, Bosman and stromal cells in colorectal carcinomas: a
66. Choi JK, Choi JY, Kim DG et al.: FT, Benhattar J: Intratumor genetic metabolic survival role for tumor-associated
Integrative analysis of multiple gene heterogeneity in advanced human colorectal stroma. Cancer Res. 66, 632–637 (2006).
expression profiles applied to liver cancer adenocarcinoma. Int. J. Cancer 93, 346–352 77. Joyce JA: Therapeutic targeting of the tumor
study. FEBS Lett. 565, 93–100 (2004). (2001). microenvironment. Cancer Cell 7, 513–520
67. Hosack DA, Dennis G Jr, Sherman BT, • Molecular description of colorectal (2005).
Lane HC, Lempicki RA: Identifying cancer heterogeneity.
216 Personalized Medicine (2006) 3(2)You can also read
