Quantifying gene expression with RT-PCR

you are quantifying the cDNA, not the mRNA (RT means reverse
transcription, not real time)
START WITH GOOD RNA; batch for simultaneous RT
Detection in PCR products in gels
run 30-40 cycles and stain gel for presence of amplicon of expected size
not quantitative, ok for there or not there experiments
stop PCR while amplification still is exponential and measure band intensity
trial and error to determine where to stop
optimal # of cycles varies across genes
Add competitive internal standard and measure relative band intensities
competitor uses same primers as target cDNA
competitor amplicon longer or shorter than target amplicon
Real time detection
Has become the gold standard, but you still have to be careful and
understand the pitfalls

Good RNA
ribosomal bands stand out in gel (or Bioanalyzer)
A260/A280 > 1.8 (lower does not necessarily mean bad RNA, but
quantification by A260 is not accurate if ratio low)
partially degraded RNA might be usable [primers must be close to 3 end of
mRNA with oligo(dT) priming, or use random priming]
free of genomic DNA (important for accurate quantification of transcripts
expressed at low levels); DNase it or use RT-negative controls; or design
PCR primers (or probes) so genomic DNA is not amplified
Consistent cDNA synthesis
Prepare both experimental and control cDNA samples at the same time
with the same lot of RT with the same amount of RNA template using the
same kind of primer
Oligo(dT) and random primers can give you different Ct and Ct values, but
Ct should be similar if samples are batched as indicated above
without consistent reverse transcription efficiency, accuracy of method
becomes much more dependent on selection of a good reference gene

Ideally, absorbance ratios should be > 1.8 for both

260 nm/280 nm and 260 nm/230 nm. This allows
more accurate quantification from A260 and makes
it more likely that there are no inhibitors of reverse
transcription.

Two basic approaches to real-time PCR

1. sequence-specific fluorescent probes
adds specificity - nonspecific products should be invisible
multiplexing and allelic discrimination possible
probes are expensive
single reporter molecule per new amplicon molecule produced
2. SYBR Green dye
less expensive than using specific probes
all nonspecific DNA products detected (melting curve or gel analysis
should be done)
cannot multiplex
multiple SYBR Green molecules per amplicon molecule

Rn is ratio of FAM to ROX in

this example

use of a reference dye (ROX

in this case) reduces
variability

no difference in final signal even though 1000-fold difference in input cDNA

Set threshold where these

lines are straight (on log
scale) and parallel across all
samples

SYBR Green assay

3 targets
4 samples/target
all curves look ok

Ct = 27.5-18 = 9.5

Ct is point at which
fluorescence
intensity reaches a
threshold selected to
accommodate all
samples (threshold
can be different for
different genes)

Ct =18.0

Ct =27.5

29.5 = 724-fold
difference if amplicons
same size and PCR
efficiency is 100% for
all targets
As long as threshold
is chosen within
exponential phase of
amplification, Ct
should be the same
regardless of
absolute Ct. If
comparing data
across different
plates, use same
threshold for all
plates to avoid
increasing variability.

Melting curve raw fluorescence

data
% dsDNA drops as T increases
hard to interpret curves in this form
look at derivative instead
derivative is -y/x over very
small intervals of x, in other words
rate of drop in fluorescence as T
increases
rate is fastest at Tm

This is what you want to see

two products, one small (primer

dimer?)

 Two products close in size

abundance of larger varies a lot
in two samples almost looks like a
single peak

From Ct to gene expression

absolute quantification

# molecules mRNA (really cDNA) per ? (cell, mg protein, ng total RNA,

or # molecules reference RNA)

only as accurate as your denominator even if quantification of target

cDNA is very accurate

make RNA standard by IVT if you need to quantify mRNA rather than
cDNA, but even this might not be 100% accurate

relative quantification

Ct: ratio of target cDNA to reference (housekeeping) cDNA

Ct: target/reference ratio in experimental sample relative to

target/reference ratio in calibrator sample

can check amplification efficiencies of target and reference by dilution

curve

reference (housekeeping) genes are not necessarily invariant

Standard curve for absolute quantification

serial 10-fold dilutions of plasmid
amount (log 10) in unknown derived from
slope (a) and intercept (b) of regression line:

Y = aX + b
X = (Y b)/a
ideal slope is -3.32, i.e efficiency* is 1
because:

23.32 = 10
here, slope is -3.58, meaning it takes 3.58
cycles to get 10-fold amplification, so

X3.58 = 10
X = 101/3.58 = 1.9
Efficiency = X 1 = .9
*efficiency is number of amplicon molecules
produced per cycle per number of template
molecules. If one molecule is produced per
template (maximum possible) then number of
amplicons doubles

Issues in absolute quantification

quantification of DNA standard not always accurate
must make RNA standard by in vitro transcription if you want to quantify
RNA rather than cDNA
quantification of standard RNA can be imprecise (fluorescence-based
methods more accurate than UV absorbance)
Even with RNA standard, accuracy of quantification is affected by RNA
instability (freeze small aliquots) and possibly differences in RT efficiency
between RNA standard and target mRNA

Relative quantification
XC = X0 (1+E)C
where XC is amount of amplicon after C cycles, X0 is number of template
molecules before amplification, and E is amplification efficiency.
We want to know X0; X0 = XC (1+E)C
Because relation between XC and detected fluorescence is not necessarily
constant for all values of C, we get relative quantity by picking a fluorescence
level, well above the background fluorescence, that all amplification curves cross
while still in exponential phase of amplification (linear part of amplification plot).
XC at Ct is a constant (all samples have the same fluorescence level), which for
simplicity is set at an arbitrary value of 1. Thus,
Relative X0 (arbitrary units) = 1 (1+E)Ct = (1+E)-Ct
If we are sure that every sample in the experiment has an identical amount of
total cDNA of the same quality, this information might be enough to compare
across samples. Usually, we do not want to make this assumption but prefer to
express X0 in relation to a reference gene to control for variations in amount of
RNA in RT reaction, RNA quality, RT efficiency, etc. In other words, we want to
know X0/X0ref.

Since X0ref is calculated the same way as X0

X0/X0ref = (1+E)-Ct (1+Eref)-Ctref
If E = Eref, then you can just subtract exponents
X0/X0ref = (1+E)-Ct (-Ctref) = (1+E)Ctref-Ct
Ctref Ct is known as Ct, and often it is assumed that E =1, so
X0/X0ref = 2Ct
[although note that Ct often refers to Ct-Ctref, so 2-Ct is used]

Use of Ct
This method is used to normalize all X0/X0ref ratios relative to a single sample
(often a control sample, but not necessarily). The X0/X0ref ratio in the
calibrator sample is set arbitrarily at a value of 1.0.
Example: calibrator Ct (ref gene of interest) is -5, experimental Ct is -10,
so gene is downregulated by experimental condition relative to calibrator by:
2Ct = 2[-5 (-10)] = 25 (32-fold)
This calculation should be used only if E = 1for both reference gene and gene
of interest (use X0/X0ref = (1+E)-Ct (1+Eref)-Ctref if E and Eref are not same).
Alternatively (my preference), you can just compare average X0/X0ref in
experimental samples with X0/X0ref in a group of control samples (convert Ct
to X0/X0ref before averaging). This gives you the same relative quantification
except absolute units are different (i.e., no sample has to have a value of 1.0).

Magnitude of errors when E is different for gene of interest and reference gene,
but you use 2 Ct anyway:
Suppose actual Eref = 1 and actual Egoi = 0.9, sample 1 is calibrator
ref Ct 1

ref Ct 2

goi Ct 1

goi Ct 2

2^ddCt

correct ratio

%error

15

15

20

22

0.25

0.2770

10.8

15

18

21

26

0.25

0.3231

29.2

15

20

22

29

0.25

0.3580

43.2

15

22

23

32

0.25

0.3967

58.7

15

25

20

32

0.25

0.4627

85.1

12

22

17

29

0.25

0.4627

85.1

12

12

17

19

0.25

0.2770

10.8

12

12

17

21

0.0625

0.0767

22.8

12

12

17

23

0.01563

0.0213

36.0

32

Validating Ct method

Ct Mstn

dilute cDNA to determine

efficiency for both gene of
interest and control gene (use
the strongest sample for the
dilution series)

30

slope = -.99
28

26

24

22
-12

-10

-8

-6

-4

-2

26

Ct GAPDH

For log2 data, slope should be

close to -1, for log10 data, close
to -3.32

24

slope = -1.01

for slope from log2 data

22

E = 2(-1/slope) 1

20

= 1.01 for Mstn, 0.99 for GAPDH

18

for slope from log10 data

16
-12

-10

-8

-6

Log2 total cDNA

-4

-2

E = 10(-1/slope) 1

Ct for Mstn Ct GAPDH Ct

7
6.8
6.6

slope = 0.023 Ct per 2-fold increase cDNA

6.4
6.2

Ct

6
5.8
5.6
5.4
5.2
5
-12

-10

-8

-6

-4

-2

Log2 amount total cDNA

If amplification efficiencies are similar enough to use the simple (1+E)Ct
calculation method, then slope must be close to zero

Selecting a reference gene

consistent expression in all samples both within and across experimental conditions
easier to choose if validating microarray data, otherwise usually you are not certain that reference
gene is totally unaffected by experimental conditions
can try several reference genes, or see if Ctref changes with experimental conditions when you
carefully standardize cDNA input
more critical if hoping to find relatively small effects
An alternative is to measure total cDNA concentration, use same amount of total cDNA in every well,
and calculate relative X0 (=[1+E]-Ct), for every well without worrying about reference genes

Replicate, Replicate, Replicate

Always run replicate wells (we prefer 3) of each assay for each sample.
Precision (SD) with careful pipetting can be less than 0.1 Ct units (on our
7900HT instrument). Thus, it is possible to detect small differences
between two samples with very careful plate set up.

Do as many biological replicates as you can if you want to detect
relatively small effects of a treatment. Selecting a good reference
transcript (stable expression across conditions) or accurate determination
of total cDNA concentrations will enhance the ability to detect small
effects. With 8-10 subjects per group we have been able to detect effects
on gene expression as small as 25% at P < 0.05.