ARTICLE
In the last few years, most of the attempts to improve the polymerase
chain reaction (PCR) have been dedicated to making it quantitative. Quantification
of DNA sequences or of mRNA transcripts can indeed be an important tool
for a large number of diagnostic tests as well as for medical or environmental
analyses. The measurement of cytokine transcripts is of particular interest,
because the protein products are often below the detection level of the
most sensitive ELISA. Northern or dot blot analysis, or RNase protection
assays, were the first methods used to detect and quantify transcripts.
These methods require a lot of starting material and do not allow simultaneous
handling of large numbers of samples. While studies on animals can use
pooled samples to obtain more material, this is often not possible in
the case of human clinical trials, where each patient has to be assessed
individually. With the recent advent of RT-PCR, the sensitivity of mRNA
detection has increased allowing the detection of specific messages in
a single cell [1]. This sensitivity is an important factor for studying
variations in cytokine production since it allows the determination of
mRNA levels in individuals, and the variability between individuals can
also be assessed. Another important application of cytokine message detection
is that it allows quantification of cytokines from solid tissues, providing
information on local concentrations of cytokines.
Many groups claim to have developed quantitative PCR (Q-PCR) techniques,
many of them being based on the same principles and thus representing
very minor modifications to previously published methods. In this huge
number of publications dealing with Q-PCR, where each group refers to
its own technique, it becomes difficult to find one's own way, and to
clarify what are the essential points from these techniques. Some groups
have compared different techniques for the quantification of the same
samples, but the comparison is usually restricted to a single sequence
of interest [2, 3]. Such prolific literature has very negative effects
on the scientific point of view. Some people even claim that PCR cannot
be relied upon for quantitative measurements owing to the very nature
of the amplification process [4]. This review attempts to assimilate the
available literature (non exhaustively!) on Q-PCR techniques and summarises
the various approaches commonly used to obtain quantitative results. Comments
on the reliability of quantification will be made. The various techniques
used for RNA extraction and reverse transcription have been carefully
compared in a previous review [5] and will not be discussed here.
NORMALISATION OF SAMPLES
Because the efficiency of RNA extraction and reverse transcription can
vary widely between independent experiments, comparison of different cDNA
samples by any Q-PCR technique can only be achieved by normalisation of
the samples. The most commonly used genes for normalisation are the house-keeping
genes GAPDH, HPRT and ß-actin, that are supposed to be expressed
at the same low level in every cell type. This has in fact only been verified
for murine HPRT which has been shown to be expressed at 5 to 10 copies
per cell in four different tissues or cell lines [6]. Because of the existence
of different isoforms of ß-actin, the primers should be designed
very carefully when using ß-actin as a reference. Furthermore, the
expression of ß-actin transcripts has been shown to be eight-fold
lower in mouse liver than in either thymus or spleen [7], and very heterogenous
in Hodgkin's disease lymph nodes [8]. Similarly, GAPDH has been reported
to show different levels of expression in several studies, both in rat
and mouse [8-11]. It thus seems that, at least for studies in the mouse,
HPRT is preferable to either GAPDH or ß-actin for normalisation
of the samples.
The choice of a Q-PCR technique involves consideration of two processes,
which can be treated separately: 1. the principle of the quantification
and 2. the read-out, i.e. the detection of the PCR products. While
most of the read-out methods do not interfere with the quantification
(which will be discussed later), the choice of the principle has a direct
bearing on the reliability of the results obtained. Table
I gives a summary of the technical steps involved in each technique,
as well as the different read-outs that can be used.
PRINCIPLES
Most techniques are defined by their use or non-use of a standard, and
by the character of the standard, if used: external, endogenous internal
or exogenous internal.
Technique without standard.
In 1990, Brady and collaborators described a novel method by which the
entire mRNA population was amplified while preserving relative sequence
representation in the population [12]. An adaptation of this technique
has recently been shown to be semi-quantitative when used on low-abundance
messages, thus showing that it is suitable for quantification in single
cells [13]. This method uses oligo-dT-primed reverse transcription followed
by poly-A-tailing via a terminal transferase-catalysed reaction
to generate a 5'-oligo-dT-transcript-poly-A-3' first strand cDNA. This
is followed by oligo-dT-primed PCR to amplify a representative population
of cDNAs. Southern blot analysis is then performed with specific probes,
and hybridised blots are analysed by PhosphorImager quantification. While
this technique may be for the moment the only one suitable for semi-quantification
of messages in single cells, it in fact only represents an increase in
the detection levels of Northern blots, with all the drawbacks of this
technique (use of radioactivity, time-consuming, not applicable to the
handling of large numbers of samples). Furthermore, because of the different
kinetics of amplification of templates with very different sequences between
the primers (further discussed below), the assumption that the non-specific
amplification conserves the relative representation of the different populations
should be treated with caution. Another modification of this technique
of non-specific RNA amplification has been reported, in which the expression
level of different genes was ultimately assessed by dot blotting [14].
Techniques using standards.
All the other published Q-PCR techniques employ gene-specific amplification
of a cDNA. At this point, two distinct options can be chosen: co-amplification
with an internal standard, or reference to a standard curve that is obtained
from distinct amplifications (external standard).
Standards can be either RNA, added to the samples prior to reverse transcription,
or DNA, used only at the amplification step. RNA standards have been claimed
to be more reliable because they are included earlier in the quantification
process. However, they do not provide any control for the yield of RNA
extraction, and thus do not prevent the need for normalisation. In our
experiments, we have always noticed that reverse transcription was far
more reproducible than RNA extraction, i.e. while the extraction
of the same numbe r of cells provided very different amounts of RNA, different
reverse transcriptions of the same RNA usually provided a cDNA with about
the same number of HPRT copies per unit. Furthermore, DNA standards are
more stable and are thus easier to handle.
Endogenous internal standard. The use of endogenous internal standards
for Q-PCR can be achieved in two ways: the co-amplification (in the same
reaction tube) of the sequence of interest and of a house-keeping gene
using different primers [15] or the co-amplification of the sequence of
interest together with an endogenous variant using the same primers [3].
The second option is of course restricted to specific purposes, such as
the analysis of deletions involved in some diseases [3] and cannot be
applied to cytokine quantitations. The co-amplification of a specific
gene together with a house-keeping gene is very attractive, mainly because
it does not require the construction of an artificial standard, however
it is in fact quite difficult to achieve. The main problem is that it
requires amplification of two different products with two pairs of primers
in the same tube, often starting from very different amounts of specific
template prior to amplification; both amplifications should be analysed
in the exponential phase. These conditions exist in a very limited number
of samples.
External standard. Good quantification can be achieved by the use of
an external standard in two ways: comparison of the amounts of PCR products
to a standard curve, either from 1. the same amount of cDNA after different
numbers of amplification cycles, or 2. from several dilutions of a given
cDNA after a fixed number of cycles. It seems that, providing that the
standard curve shows that the amplification was exponential, these two
methods give similar results. Several groups have described techniques
based on sequential sampling of PCR products during amplification [15-17].
The main problem of such a technique, if it is not completely automated,
is that it is technically very difficult to perform: it is time-consuming
and it is quite easy to create contamination problems because of the repeated
samplings. The principle has been used by Applied BioSystem for their
newest PCR machine, the ABI PRISM 7700 Sequence Detector. During PCR,
a fluorogenic probe, consisting of an oligonucleotide with both a reporter
and a quencher dye attached, anneals specifically between the two amplification
primers. During elongation, the probe is cleaved by the 5' nuclease activity
of the DNA polymerase, and the reporter dye is separated from the quencher
dye, providing a sequence specific signal. At each cycle, additional reporter
dye molecules are cleaved from their probes, and the fluorescence intensity
is monitored during the PCR. Results are available immediately after the
amplification is complete. While this machine may provide the answer to
achieving cytokine quantification of multiple samples (no published data),
most laboratories will probably be unable to afford it.
The alternative to differential sampling is to amplify different dilutions
of a cDNA for a given number of cycles [18, 19]. This requires definition
of the right number of cycles to use prior to quantification. Nevertheless
the fact that the amount of PCR product should be directly proportional
to the dilution used ensures that the PCR is performed in conditions allowing
quantification.
In these different techniques, the standard curve is provided by performing
the same amplifications in separate tubes of the standard, whose amount
is known prior to amplification. Comparison of the curve obtained from
the sample under analysis, with the standard curve, gives the number of
the messages of interest in the sample. The same methods can be applied
with no reference to a standard curve, but simply to compare the amounts
of a given mRNA in different samples. This is also accurate, and is usually
called semi-quantification because it does not provide an absolute number
of copies per sample. Quantification using this technique also has to
be performed during the exponential phase of amplification.
Internal standard. The most frequently used Q-PCR techniques are based
on the co-amplification of the sample of interest and of a known amount
of an artificial standard using only one pair of primers. Three alternatives
are proposed for the standard: the sequence of the standard between the
primers can be as close as possible to the wild-type sequence [6, 20-22],
completely unrelated [23-26], or intermediate, such as the use of genomic
DNA containing a small intron [21]. The use of a plasmid containing the
sequences of several pairs of primers is very attractive because the same
standard can be used to quantify a number of different cytokines. There
has been (and still is) much controversy about the relative merits of
these different standards. When the sequences of the target and the control
are very close, the situation approaches the "equivalency of replication
efficiencies" as defined by Nedelman et al. [27]. While it is clear
that a standard bearing minor modifications, such as a single point mutation
or a deletion or addition of a few base pairs, will follow the same amplification
kinetics as the wild type sequence, it has been shown by several groups
that it is not the case for a standard exhibiting an large sequence difference
between the two primers. Zhang and collaborators have shown that sequences
with large difference in size could exhibit different amplification kinetics,
though they were amplified in the same tube, using the same primers [3].
Similarly, Pannetier and collaborators have shown that standards containing
irrelevant sequences between the amplification primers could not be used
with confidence when PCR was run to saturation [6].
READ-OUT
The detection method must also be taken into consideration as a part
of the overall quantitative process. The general principle is that the
more sensitive the read-out, the fewer the number of cycles that are required
and the more precise the method. Many attempts have been made to standardise
routine procedures in order to simplify this step of the quantification
procedure.
The choice of the Q-PCR technique can restrict the possible use of read-out,
but most are compatible with different ways of detecting the PCR products
(Table I). The early quantitative
PCR techniques that were developed used gel electrophoresis after amplification.
Agarose gels can be used when the difference in the sizes of the products
is sufficient to allow the discrimination of the wild-type- and control-product,
or when using a standard that carries a specific restriction site [20,
21]. When the standard is very close in size to the wild-type sequence,
polyacrylamide gels have to be used to discriminate between the two products.
Gels are then analysed using ethidium bromide staining, blotting and hybridisation
with specific radioactive probes, or run on automated systems that allow
a higher sensitivity while avoiding the use of radioactivity. However
gel electrophoresis is not easily compatible with the routine handling
of many samples, so many groups have developed read-out systems based
on ELISA [28, 29], EIA [26], ELOSA [2], or electroluminescence [30, 31].
Other colorimetric Q-PCR techniques have been described, mainly for the
quantification of viral load in samples [32-35], but their accuracy and
sensitivity have not been well enough documented to assess whether they
could be applied to accurate measurement of cytokines. While the electroluminescent
detection of PCR products requires specialised equipment, the other techniques
can be performed by most laboratories, and do not differ much in their
principles. Used to detect products from co-amplification with an internal
standard, these techniques have been shown to be easy to handle, accurate,
and sensitive enough for the quantification of cytokine messengers in
small samples.
Analysis of PCR products based on agarose or
polyacrylamide gel electrophoresis is the most widely employed method,
but two major problems have been recently described. Non-denaturating
electrophoresis has been reported to allow formation of heteroduplexes,
thus greatly reducing the sensitivity and accuracy of quantification [36].
This formation of heteroduplexes is of course enhanced when the internal
standard has similarities with the target sequence. Furthermore, when
gels are stained with ethidium bromide, the technique has low sensitivity
and requires a greater amount of amplified product to allow detection.
This requirement further favours the formation of heteroduplexes, thus
greatly distorting the quantification. Another problem of quantification
using image analytic assessment of agarose gel-separated PCR products
is its lack of reproducibility. Routine use of such instrumentation has
been shown to exhibit about 20% variability in the measurement of individual
bands [37]. Furthermore, when the PCR has to be analysed during the exponential
phase, the range of detection of agarose gels is yet another limitation.
It is thus clear that data obtained from Q-PCR analysed on agarose gels
should be interpreted very carefully, and should be considered as semi-quantitative,
failing to measure with accuracy a difference of less than 50-100 fold
between different samples.
It is very clear that none of the techniques described above can be
used to quantify mRNA with a single reaction. Statements like "specific
messages were measured by limiting both the cycle number and the amount
of starting material such that the efficiency of the reaction remained
constant" cannot be taken as scientifically valid. Although many people
do not seem to be aware of it, it is in fact, quite difficult to achieve
PCR in the exponential phase. Indeed, the increase in the amount of amplicons
remains exponential only for a limited number of cycles, after which the
amplification rate reaches a plateau. In cytokine quantification experiments,
the end of the exponential phase usually corresponds to the number of
cycles providing an amplified product which is barely detectable using
agarose gels stained with ethidium bromide, while the plateau is reached
only 5 to 8 cycles later (J.P. Levraud, personal communication). Many
parameters are involved in the kinetics of a given PCR: primer sequences,
Taq polymerase used, amount of starting material, etc. It has been clearly
shown that the use of a single point for quantification not only prevents
the rejection of failed quantification, but also fails to detect any contamination
as well as producing less acurate results [28]. Furthermore, the use of
several points for a given quantification ensures that the experiment
is performed in conditions that allow quantification and thus provides
an integral control.
Minor changes between two different PCR experiments, should imply that
the limits of the assay were reevaluated. The extreme sensitivity of quantitative
PCR to any changes in the conditions of the assay demands additional effort
by the investigator to ensure that the analysis is performed within the
established limits of the method. Common sense controls are no longer
sufficient to generate enough confidence in the quantitative power of
the assay. One of the most important points to bear in mind when performing
Q-PCR is that it requires one to define parameters to accept or reject
a result, and that these parameters should be defined once and for all.
When several dilutions or numbers of cycles are used, plotting the quantity
of amplified product (whatever the unit is, depending on the read-out)
as a function of dilution, copy number of standard, or number of cycles
should provide a straight line on a log-log scale. This plotting also
provides important controls, such as the slope of the curve, and the dispersion
of the experimental points, that should fall within a pre-determined range.
For example, it has been shown that the requirement for a slope of 1 is
imperative [38], while reference to this slope is often omitted in the
published techniques. Although it is very tempting to reduce the number
of amplifications per sample, when there are many, it can have heavy consequences
on the accuracy of the results. I would like to conclude by reminding
users of these different techniques that none of them can be relied upon
to be quantitative, unless each experiment is carefully controlled.
CONCLUSION
Acknowledgement.
I would like to thank J.P. Levraud and G. May for helpful discussions
and careful reading of the manuscript.
REFERENCES
1. Rappolee D A, Wang A, Mark D, Werb Z. 1989. Novel method for studying
mRNA phenotypes in single or small numbers of cells. J. Cell. Biochem.
39: 1.
2. Mallet F, Hebrard C, Livrozet J M, Lees O, Tron F, Touraine J L,
Mandrand B. 1995. Quantitation of human immunodeficiency virus type 1
DNA by two PCR procedures coupled with enzyme-linked oligosorbent assay.
J. Clin. Microbiol. 33: 3201.
3. Zhang C, Peters L E, Linnane A W, Nagley P. 1996. Comparison of different
quantitative PCR procedures in the analysis of the 4977-bp deletion in
human mitochondrial DNA. Biochem. Biophys. Res. Comm. 223: 450.
4. Hengen P N. 1995. Methods and reagents. Quantitative PCR: an accurate
measure of mRNA? Trends Biochem. Sci. 20: 476.
5. Zamorano P L, Mahesh V B, Brann D W. 1996. Quantitative RT-PCR for
neuroendocrine studies. A minireview (Review) Neuroendocrinology
63: 397.
6. Pannetier C, Delassus S, Darche S, Saucier C, Kourilsky P. 1993.
Quantitative titration of nucleic acids by enzymatic amplification reactions
run to saturation. Nucleic Acids Res. 21: 577.
7. Biragyn A, Arkins S, Kelley K W. 1994. Riboprobe expression cassettes
for measuring IGF-I, beta-actin and glyceraldehyde 3-phosphate dehydrogenase
transcripts. J. Immunol. Methods 168: 235.
8. Perfetti V, Manenti G, Dragani T A. 1991. Expression of housekeeping
genes in Hodgkin's disease lymph nodes. Leukemia 5: 1110.
9. Funabashi T, Brooks P J, Weesner G D, Pfaff D W. 1994. Luteinizing
hormone-releasing hormone receptor messenger ribonucleic acid expression
in the rat pituitaty during lactation and the estrous cycle. J. Neuroendocrinol.
6: 261.
10. Ferber S, Meyerovitch J, Kriauciunas K M, Kahn C R. 1994. Vanadate
normalizes hyperglycemia and phosphoenolpyruvate carboxykinase mRNA levels
in ob/ob mice. Metabolism 43: 1346.
11. Bhatia P, Taylor W R, Greenberg A H, Wright J A. 1994. Comparison
of glyceraldehyde-3-phosphate dehydrogenase and 28S-ribosomal RNA gene
expression as RNA loading controls for northern blot analysis of cell
lines of varying malignant potential. Anal. Biochem. 216: 223.
12. Brady G, Barbara M, Iscove N N. 1990. Representative in vitro
cDNA amplification from individual hemopoitic cells and colonies. Meth.
Mol. Cell. Biol. 2: 17.
13. Cheng T, Shen H, Giokas D, Gere J, Tenen D G, Scadden D T. 1996.
Temporal mapping of gene expression levels during the differentiation
of individual primary hematopoietic cells. Proc. Natl. Acad. Sci. USA
93: 13158.
14. Hamoui S, Benedetto J P, Garret M, Bonnet J. 1994. Quantitation
of mRNA species by RT-PCR on total mRNA population. PCR Methods Appl.
4: 160.
15. Chelly J, Montarras D, Pinset C, Berwald-Netter Y, Kaplan J C, Kahn
A. 1990. Quantitative estimation of minor mRNAs by cDNA-polymerase chain
reaction. Application to dystrophin mRNA in cultured myogenic and brain
cells. Eur. J. Biochem. 187: 691.
16. Alard P, Lantz O, Sebagh M, Calvo C F, Weill D, Chavanel G, Senik
A, Charpentier B. 1993. A versatile ELISA-PCR assay for mRNA quantitation
from a few cells. Biotechniques 15: 730.
17. Hoof T, Riordan J R, Tümmler B. 1991. Quantitation of mRNA
by the kinetic polymerase chain reaction assay: a tool for monitoring
P-glycoprotein gene expression. Anal. Biochem. 196: 161.
18. Azuara V, Levraud J P, Lembezat M P, Pereira P. 1997. A novel subset
of adult gammadelta thymocytes that secretes a distinct pattern of cytokines
and expresses a very restricted T cell receptor repertoire. Eur. J.
Immunol. 27: 544.
19. Ballagi-Pordany A, Ballagi-Pordani A, Funa K. 1991. Quantitative
determination of mRNA phenotypes by the polymerase chain reaction. Anal.
Biochem. 196: 89.
20. Becker-Andre M, Hahlbrock K. 1989. Absolute mRNA quantification
using the polymerase chain reaction (PCR). A novel approach by PCR aided
transcript titration assay (PATTY). Nucleic Acids Res. 17: 9437.
21. Gilliland G, Perrin S, Blanchard K, Bunn H F. 1990. Analysis of
cytokine mRNA and DNA: detection and quantitation by competitive polymerase
chain reaction. Proc. Natl. Acad. Sci. USA 87: 2725.
22. Porcher C, Malinge M C, Picat C, Grandchamp B. 1992. A simplified
method for determination of specific DNA or RNA copy number using quantitative
PCR and an automatic DNA sequencer. Biotechniques 13: 106.
23. Wang A M, Doyle M V, Mark D F. 1989. Quantitation of mRNA by the
polymerase chain reaction. Proc. Natl. Acad. Sci. USA 86: 9717.
24. Platzer C, Richter G, Überla K, Müller W, Blöcker
H, Diamantstein T, Blankenstein T. 1992. Analysis of cytokine mRNA levels
in interleukin-4-transgenic mice by quantitative polymerase chain reaction.
Eur. J. Immunol. 22: 1179.
25. Legoux P, Minty C, Delpech B, Minty A J, Shire D. 1992. Simultaneous
quantification of cytokine mRNAs in interleukin-1ß stimulated U373
human astrocytoma cells by a polymerase chain reaction method involving
co-amplification with an internal multispecific control. Eur. Cytokine
Netw. 3: 553.
26. Hockett R D Jr, Janowski K M, Bucy R P. 1995. Simultaneous quantitation
of multiple cytokine mRNAs by RT-PCR utilizing plate based EIA methodology.
J. Immunol. Meth. 187: 273.
27. Nedelman J, Heagerty P, Lawrence C. 1992. Quantitative PCR: procedures
and precisions. Bull. Math. Biol. 54: 477.
28. Demay F, Darche S, Kourilsky P, Delassus S. 1996. Quantitative PCR
with colorimetric detection. Res. Immunol. 147: 403.
29. Zou W, Durand-Gasselin I, Dulioust A, Maillot M, Galanaud P, Emilie
D. 1995. Quantification of cytokine gene expression by competitive PCR
using a colorimetric assay. Eur. Cytokine Netw. 6: 257.
30. DiCesare J, Grossman B, Katz E, Picozza E, Ragusa R, Woudenberg
T. 1993. A high-sensitivity electrochemiluminescence-based detection system
for automated PCR product quantitation. Biotechniques 15: 152.
31. Wages J M, Dolenga L, Fowler A K. 1993. Electrochemiluminescent
detection and quantitation of PCR-amplified DNA. Amplification
10: 1.
32. Yun Z, Lundeberg J, Johansson B, Hedrum A, Weiland O, Uhlen M, Sönnerborg
A. 1994. Colorimetric detection of competitive PCR products for quantification
of hepatitis C viremia. J. Virol. Methods 47: 1.
33. Shindo M, Di Bisceglie A M, Silver J, Limjoco T, Hoofnagle J H,
Feinstone S M. 1994. Detection and quantitation of hepatitis C virus RNA
in serum using PCR and a colorimetric detection system. J. Virol. Methods
48: 65.
34. Ranki M, Schätzl HM, Zachoval R, Uusi-Oukari M, Lehtovaara
P. 1995. Quantification of hepatitis virus DNA over a wide range from
serum for studying viral replicative activity in response to treatment
and in recurrent infection. Hepatology 21: 1492.
35. Mulder J, McKinney N, Christopherson C, Sninsky J, Greenfield L,
Kwok S. 1994. Rapid and simple PCR assay for quantitation of human immunodeficiency
virus type 1 RNA in plasma: application to acute retroviral infection.
J. Clin. Microbiol. 32: 292.
36. Henley W N, Schuebel K E, Nielsen D A. 1996. Limitations imposed
by heteroduplex formation on quantitative RT-PCR. Biochem. Biophys.
Res. Comm. 226: 113.
37. Sundfors C, Collan Y. 1995. Basics of quantitative PCR 1: image
analysis and quantitation of PCR products. Cell. Mol. Biol. (Noisy-le-Grand)
41: 671.
38. Raeymaekers L. 1993. Quantitative PCR: theoretical considerations
with practical implications. Anal. Biochem. 214: 582.
39. Delassus S, Coutinho G C, Saucier C, Darche S, Kourilsky P. 1994.
Differential cytokine expression in maternal blood and placenta during
murine gestation. J. Immunol. 152: 2411.
|
Printable version
