64 166 PostTranscriptional Control of Gene Expression RNA Interference

by taratuta

Category: Documents





64 166 PostTranscriptional Control of Gene Expression RNA Interference
wea25324_ch16_471-521.indd Page 488
4:53 PM user-f469
Chapter 16 / Other Post-Transcriptional Events
(a) Low iron
(b) High iron
Figure 16.23 Model for destabilization of TfR mRNA by iron. (a) Under low-iron conditions, the aconitase apoprotein (orange) binds to the IREs
in the 39-UTR of the TfR mRNA. This protects the RNA from degradation by RNases. (b) Under high-iron conditions, iron binds to the aconitase
apoprotein, removing it from the IREs, and opening the IREs up to attack by RNase. The RNase clips the mRNA at least once, exposing its 39-end
to further degradation.
All the data we have considered are consistent with the
following hypothesis (Figure 16.23): When iron concentrations are low, an IRE-binding protein, or iron regulatory
protein (IRP), binds to the rapid turnover determinant in
the 39-UTR of the TfR mRNA. This protein protects the
mRNA from degradation. When iron concentrations are
high, iron binds to the IRE-binding protein, causing it to
dissociate from the rapid turnover determinant, opening it
up to attack by a specific endonuclease that clips off a 1-kb
fragment from the 39-end of the TfR mRNA. This destabilizes the mRNA and leads to its rapid degradation.
One of the proteins (IRP1) that bind to the IREs in both
the transferrin receptor mRNA and the ferritin mRNA
(Chapter 17) has now been identified as a form of aconitase, an enzyme that converts citrate to isocitrate in the
citric acid cycle. The enzymatically active form of aconitase
is an iron-containing protein that does not bind to the
IREs. However, the apoprotein form of aconitase, which
lacks iron, binds to the IREs in mRNAs.
SUMMARY When the iron concentration is high,
the TfR mRNA decays rapidly. When the iron concentration is low, the TfR mRNA decays much
more slowly. This difference in mRNA stability is
about 20-fold and plays a major role in control of
the gene’s expression. The initiating event in TfR
mRNA degradation seems to be an endonucleolytic
cleavage of the mRNA more than 1000 nt from its
39-end, within the IRE region. This cleavage does
not require prior deadenylation of the mRNA. Iron
controls TfR mRNA stability as follows: When iron
concentration is low, aconitase exists at least partly
in an apoprotein form that lacks iron. This protein
binds to the IREs in the TfR mRNA and protects
the RNA against attack by RNases. But when iron
concentration is high, the aconitase apoprotein
binds to iron and therefore cannot bind to the TfR
mRNA IREs. This leaves the RNA vulnerable to
16.6 Post-Transcriptional
Control of Gene Expression:
RNA Interference
For years, molecular biologists have been using antisense
RNA to inhibit expression of selected genes in living cells. At
first, the rationale was that the antisense RNA, which is
complementary to mRNA, would base-pair to the mRNA
and inhibit its translation. The strategy usually worked, but
the rationale was incomplete. As Su Guo and Kenneth Kenphues established in 1995, injecting sense RNA into cells
worked just as well as antisense RNA in blocking expression
of a particular gene. Then, in 1998, Andrew Fire and Craig
Mello and their colleagues showed that double-stranded
RNA (dsRNA) worked much better than either sense or antisense RNA. In fact, the main reason sense and antisense
RNAs worked appears to be that they were contaminated
with (or produced) small amounts of dsRNA, and the
dsRNA actually did the most to block gene expression.
Also, beginning in 1990, molecular biologists began
noticing that placing transgenes into various organisms
sometimes had the opposite of the desired effect. Instead
of turning on the transgene, organisms sometimes turned
off, not only the transgene, but the normal cellular copy of
the gene as well. One of the first examples was an attempt
to intensify the purple color of a petunia by supplying
extra copies of the pigment-producing genes. But in up to
25% of the transformed plants, blossoms were white or
patchy purple and white—the opposite of the intended
wea25324_ch16_471-521.indd Page 489
4:53 PM user-f469
16.6 Post-Transcriptional Control of Gene Expression: RNA Interference
Figure 16.24 Silencing of a purple color gene in petunia by adding
extra copies of the color gene. The central white stripe in each petal
shows where silencing occurred. (Source: Courtesy of Dr. Richard A.
Jorgensen, The Plant Cell.)
effect (Figure 16.24). This phenomenon was called by several names: cosuppression and post-transcriptional gene
silencing (PTGS) in plants, RNA interference (RNAi) in
animals such as nematodes (Caenorhabditis elegans) and
fruit flies, and quelling in fungi. To avoid confusion, we
will refer to this phenomenon as RNAi from now on,
regardless of the species under study.
Figure 16.25 Double-stranded RNA-induced RNA interference
causes destruction of a specific mRNA. Fire and colleagues injected
antisense or dsRNA corresponding to the C. elegans mex-3 mRNA into
C. elegans ovaries. After 24 h, they fixed the embryos in the treated
ovaries and subjected them to in situ hybridization (Chapter 5) with a
probe for mex-3 mRNA. (a) Embryo from a negative control parent with
no hybridization probe. (b) Embryo from a positive control parent that
was not injected with RNA. (c) Embryo from a parent that was injected
with mex-3 antisense RNA. A considerable amount of mex-3 mRNA
remained. (d) Embryo from a parent that was injected with dsRNA
corresponding to part of the mex-3 mRNA. No detectable mex-3
mRNA remained. (Source: Fire, A., S. Xu, M.K. Montgomery, S.A. Kostas, S.E.
Driver, and C.C. Mello, Potent and specific genetic interference by double-stranded
RNA in Caenorhabditis elegans. Nature 391 (1998) f. 3, p. 809. Copyright
© Macmillan Magazines Ltd.)
Mechanism of RNAi
Fire and colleagues showed that injecting C. elegans gonads
with dsRNA (the trigger dsRNA) caused RNAi in the resulting embryos. Furthermore, they detected a loss of the
corresponding mRNA (the target mRNA) in embryos undergoing RNAi (Figure 16.25). However, the dsRNA had
to include exon regions; dsRNA corresponding to introns
and promoter sequences did not cause RNAi. Finally, these
workers demonstrated that the effect of the dsRNA crossed
cell boundaries, at least in C. elegans. That is, the effect
spread throughout the whole organism.
Is this loss of a particular mRNA in response to the corresponding dsRNA caused by repression of transcription
of the gene or destruction of the mRNA? In 1998, Fire and
colleagues, as well as others, demonstrated that RNAi is a
post-transcriptional process that involves mRNA degradation. Several investigators reported the presence of short
pieces of dsRNA called short interfering RNA (siRNA) in
cells undergoing RNAi. In 2000, Scott Hammond and collaborators purified a nuclease from Drosophila embryos
undergoing RNAi that digests the targeted mRNA. The
partially purified preparation that contained this nuclease
activity also contained a 25-nt RNA fraction that could be
detected on Northern blots with probes for either the sense
or antisense strand of the targeted mRNA. Degradation of
the 25-nt RNA with micrococcal nuclease destroyed the
ability of the preparation to digest the mRNA. These data
suggested that a nuclease digests the trigger dsRNA into
fragments about 25 nt long, and these fragments then associate with a nuclease and provide guide sequences that
allow the nuclease to target the corresponding mRNA.
Phillip Zamore and collaborators developed a system
based on Drosophila embryo lysates that carried out RNAi
in vitro. This system allowed these workers to look at individual steps in the RNAi process. The embryos had been
injected with trigger dsRNA corresponding to luciferase
mRNA, so they targeted that mRNA for destruction. First,
Zamore and collaborators showed that RNAi requires
ATP. They depleted their extract of ATP by incubating it
with hexokinase and glucose, which converts ATP to ADP
and transfers the lost phosphate group to glucose. The
ATP-depleted extract no longer carried out the degradation
of the target, luciferase mRNA.
Next, these workers performed experiments in which
they labeled one strand of the dsRNA at a time (or both)
and showed that labeled short siRNAs of 21–23 nt appeared,
no matter which strand was labeled (Figure 16.26). The appearance of the siRNAs did not require the presence of
mRNA (e.g., compare lanes 2 and 3), so these short RNAs
apparently derived from dsRNA, not mRNA. When capped
antisense luciferase RNA was labeled (lanes 11 and 12),
wea25324_ch16_471-521.indd Page 490
4:53 PM user-f469
Chapter 16 / Other Post-Transcriptional Events
nt 233
nt 143
nt 729
nt 644
nt 50
1 2 3 4 5 6 7 8 9 10 11 12
nt 569
496 nt
501 nt
519 nt
Lysate: - + + + + + + - + + + +
mRNA: - - + - + - + - - + - +
a a a
32P-label: a a a
s s s
s s s
dsRNA: Ø A
/ / / s s a a / / / aa
1000 nt
3 0 0.5 1.5 0 0.5 1.5 0 0.5 1.5 h
1.0 kb
Figure 16.26 Generation of 21–23-nt RNA fragments in an RNAicompetent Drosophila embryo extract. Zamore and collaborators
added ds luciferase RNA from Photinus pyralis (Pp-luc RNA) or from
Renilla reniformis (Rr-luc RNA), as indicated at top, to lysates in the
presence or absence of the corresponding mRNA, as indicated at
bottom. The dsRNAs were labeled in the sense strand (s), in the
antisense strand (a), or in both strands (a/s), as indicated at bottom.
RNA markers from 17–27 nt long were included in the lane at left.
Lanes 11 and 12 contained labeled, capped antisense Rr-luc RNA in
the absence and presence of mRNA, respectively. (Source: Zamore, P.D.,
1.5 kb
0.5 kb
230 nt
181 nt
T. Tuschl, P.A. Sharp, and D.P. Bartel, RNAi: Double-stranded RNA directs the
ATP-dependent cleavage of mRNA at 21 to 23 nucleotide intervals. Cell 101 (2000)
f. 3, p. 28. Reprinted by permission of Elsevier Science.)
a small amount of siRNAs appeared, and that amount increased in the presence of mRNA (lane 12). This result suggested that the labeled antisense RNA was hybridizing to
the added mRNA to generate a dsRNA that could be degraded to the short RNA pieces. In summary, all these results suggest that a nuclease degrades the trigger dsRNA
into short pieces. Further work has shown that these siRNAs
are about 21–23 nt long.
Next, Zamore and collaborators showed that the trigger dsRNA dictated where the corresponding mRNA
would be cleaved. They added three different trigger
dsRNAs, whose ends differed by about 100 nt, to their
RNAi extracts, then added 59-labeled mRNA, allowed
RNA cleavage to occur, and electrophoresed the products.
Figure 16.27 shows the results: The dsRNA (C) whose 59-end
was closest to the 59-end of the mRNA yielded the shortest
fragments; the next dsRNA(B), whose 59-end was about
100 nt farther downstream, yielded mRNA fragments
about 100 nt longer; and the third dsRNA, whose 59-end
was about another 100 nt farther downstream, yielded
mRNA fragments about another 100 nt longer. This close
relationship between the position of the trigger dsRNA
relative to the mRNA, and the position at which cleavage
began, strongly suggests that the dsRNA determined the
sites of cleavage of the mRNA.
Next, Zamore and collaborators performed highresolution gel electrophoresis of the mRNA degradation
Rr-Luc mRNA
159 nt
37 nt
Figure 16.27 The trigger dsRNA dictates the boundaries of
cleavage of mRNA in RNAi. Zamore and collaborators added the
three dsRNAs pictured in panel (a) to an embryo extract along with
an Rr-luc mRNA, 59-labeled in one of the phosphates of the cap.
(b) Experimental results. The 59-end-labeled mRNA degradation
products were electrophoresed. The dsRNAs included in the reactions
are indicated and color-coded at top. The first lane, marked 0,
contained no dsRNA. Reactions were incubated for the times (in h)
indicated at top. The arrowhead indicates a faint cleavage site that lies
outside the position of RNA C. Otherwise, the sites cleaved lie within
the positions of the three dsRNAs on the mRNA. (Source: Zamore, P.D.,
T. Tuschl, P.A. Sharp, and D.P. Bartel, 2000. RNAi: Double-stranded RNA directs
the ATP-dependent cleavage of mRNA at 21 to 23 nucleotide intervals. Cell 101
(2000) f. 5, p. 30. Reprinted by permission of Elsevier Science.)
products from Figure 16.27. The results, presented in
Figure 16.28, are striking. The major cleavage sites in the
mRNA are mostly at 21–23-nt intervals, producing a set
of RNA fragments whose lengths differ by multiples of
21–23 nt. The one obvious exception is the site marked
by an arrowhead, which lies only 9 nt from the previous
cleavage site. This exceptional site lies within a run of
seven uracil residues, which is interesting in light of the
fact that 14 of 16 cleavage sites mapped were at uracils.
After this exceptional site, the 21–23-nt interval resumed
wea25324_ch16_471-521.indd Page 491
4:54 PM user-f469
16.6 Post-Transcriptional Control of Gene Expression: RNA Interference
OH 0 20 60 0 20 60 0 20 60 min
~22 nt
(a) Dicer
~21 nt
22 nt
(b) Delivery of ss-siRNA to RISC
21 nt
mRNA: 3′
9 nt
22 nt
23 nt
21 nt
21 nt
Figure 16.29 A simplified model for RNAi. (a) Dicer (yellow)
recognizes and binds to a double-stranded RNA (red and blue), then
cleaves the RNA into siRNAs about 21–23 nt long (depicted here as 10
nt long, for simplicity), with 2-nt 39-overhangs. The ends of the central
siRNA are labeled to illustrate the 39-overhangs. (b) One of the siRNA
strands (red) associates with RISC (orange) and base-pairs to a target
mRNA (blue). (c) The siRNA strand in the RISC complex serves as a
guide RNA to direct the cleavage of the target mRNA in the middle of
the sequence opposite the siRNA.
21 nt
Figure 16.28 Cleavages of target mRNA in RNAi occur at 21–23-nt
intervals. Zamore and collaborators performed high-resolution
denaturing polyacrylamide gel electrophoresis on the products of RNAi
in the presence of all three of the trigger dsRNAs from Figure 16.27.
The cleavages, with one notable exception (arrowhead), occurred at
21–23-nt intervals. The exceptional band indicates a cleavage at
only a 9-nt interval, but cleavages thereafter were at 21–23-nt intervals.
(Source: Zamore, P.D., T. Tuschl, P.A. Sharp, and D.P. Bartel, RNAi: Doublestranded RNA directs the ATP-dependent cleavage of mRNA at 21 to 23 nucleotide
intervals. Cell 101 (2000) f. 6, p. 31. Reprinted by permission of Elsevier Science.)
for the rest of the mapped cleavage sites. These results
support the hypothesis that the 21–23-nt siRNAs determine where the mRNA will be cut and suggest that cleavage takes place preferentially at uracils.
In 2001, Hammond and colleagues reported that they
had purified from Drosophila the enzyme that cleaves the
trigger double-stranded RNA into short pieces. They
named it Dicer, because it dices double-stranded RNA up
into uniform-sized pieces. Dicer is a member of the RNase III
family discussed earlier in this chapter. In fact, Hammond
and colleagues narrowed their search for Dicer by looking
for enzymes in this family because RNase III was the
only known nuclease specific for dsRNA. Like RNase III,
Dicer leaves 2-nt 39-overhangs (protruding 39-ends) at the
ends of the double-stranded siRNAs, and phosphorylated
Three early lines of evidence implicated Dicer in RNA
cleavage in RNAi. First, dicer, the gene that encodes
Dicer, produces a protein that can cut dsRNA into 22-nt
pieces. Second, antibodies against this protein bind to an
enzyme in Drosophila extracts that cuts dsRNA into
short pieces. Finally, when dicer dsRNA is introduced
into Drosophila cells, it partially blocks RNAi. It is ironic
that Hammond and colleagues could use RNAi to block
RNAi! But, of course, if you think about it, the blockage
could never be complete.
Dicer also has RNA helicase activity, so it can separate
the two strands of the siRNAs it creates, at least in principle. However, Dicer does not carry out the second step in
RNAi, cleavage of the target mRNA. That appears to be
the job of another enzyme, called slicer, which resides in a
complex called the RNA-induced silencing complex (RISC).
Figure 16.29 summarizes what we have learned so far
about the mechanism of RNAi.
Hammond and others have implicated another Drosophila protein, Argonaute, known from genetic experiments to be required for RNAi, in the second (slicer) step.
Argonaute does not have an RNase III motif, so molecular
biologists discounted it at first as a slicer candidate. However, structural, biochemical, and genetic studies of Argonaute carried out by Leemor Joshua-Tor, Gregory Hannon,
and their colleagues in 2004 showed that Argonaute almost
certainly has slicer activity.
These workers had shown in structural studies in 2003
that Argonaute2 of Drosophila contains two characteristic
wea25324_ch16_471-521.indd Page 492
4:54 PM user-f469
Chapter 16 / Other Post-Transcriptional Events
domains, PAZ, and PIWI. (PAZ, from PIWI, Argonaute,
and Zwili, was found only in Argonaute and Dicer; PIWI
was discovered in Drosophila. The acronym stands for Pelement-induced wimpy testis.) They had also determined
the structure of PAZ, and had shown that it contained a
module resembling a so-called OB fold, which can bind
single-stranded RNAs. They also demonstrated by crosslinking studies with labeled siRNAs and cloned GST–PAZ
fusion proteins that the PAZ domain was capable of binding to single-stranded siRNAs, or to the 2-nt single-stranded
overhangs at the 39-ends of double-stranded siRNAs. This
implicated Argonaute in the slicer reaction, at least as a
docking site for the siRNA, but not necessarily as the slicer
enzyme itself.
Next, Joshua-Tor, Hannon, and colleagues performed
x-ray crystallography on the Argonaute-like protein of the
archaeon Pyrococcus furiosus. (No full-length eukaryotic
Argonaute structure could be obtained.) They found that
three domains of the protein (the middle domain, PIWI, and
the N-terminal domain) form a crescent shape at the bottom of the structure, with the PIWI domain in the middle.
The PAZ domain lies above the crescent and is connected to
it by a stalk domain. Figure 16.30 depicts this structure, and
illustrates that the crescent forms a groove, capped by the
PAZ domain. This groove is big enough to accommodate a
double-stranded RNA, and it is lined with basic residues,
which could form electrostatic bridges to an RNA substrate.
However, the most telling part of the structure is that
the PIWI domain resembles a similar domain in RNase H,
which cleaves the RNA strand in an RNA–DNA hybrid.
Thus, RNase H can recognize a double-stranded polynucleotide and cleave one of its strands (the RNA). In addition to their overall architectural similarities, both proteins
have a cluster of three acidic residues (two aspartates and
Figure 16.30 Model for slicer activity of Argonaute. The hybrid
involving an siRNA and a target mRNA is held in the active site, at
least partly due to the interaction between the 39-end of the siRNA
and the PAZ domain of Argonaute. This places the target mRNA in
position to be cut by the slicer active site, represented by the scissors.
Cleavage occurs opposite the middle of the siRNA, which serves as a
guide RNA. The PAZ, middle, PIWI, and N-terminal domains of
Argonaute are labeled. (Source: Adapted from Science, Vol. 305, Ji-Joon
Song, Stephanie K. Smith, Gregory J. Hannon, and Leemor Joshua-Tor,
“Crystal Structure of Argonaute and Its Implications for RISC Slicer Activity,”
Fig. 4, p. 1436, AAAS.)
one glutamate). In RNase H, this carboxylate cluster binds
a Mg21 ion that plays a key role in catalyzing the cleavage
of the RNA strand. These similarities are very interesting
because slicer has an analogous activity: It must also recognize a double-stranded polynucleotide (an siRNA–mRNA
hybrid) and cleave one of its strands (the mRNA). Thus,
Argonaute has all the attributes we expect of slicer: a
domain (PIWI) with a site that appears to be capable of
cleaving one strand of an siRNA–mRNA hybrid, and another domain (PAZ) that can bind to the end of the siRNA.
To investigate further the role of Argonaute in mammals, Hannon, Joshua-Tor, and colleagues performed
genetic and biochemical studies on the Argonaute genes and
proteins in the mouse. Mammals have four Argonaute
proteins, designated Argonaute 1–4. The investigators
transfected cells with genes encoding Argonautes 1–3, along
with an siRNA that targets firefly luciferase mRNA. Then
they immunoprecipitated the RISC complexes and tested
them for ability to cleave luciferase mRNA in vitro. Only
Argonaute2 (Ago2) had this capability.
Next, these workers knocked out the Ago2 gene in mice
and observed that all such animals died in the embryonic
stage of development, with severe developmental defects and
delay. The reason for this profound phenotype is that Ago2
participates, not only in RNAi, but in a normal (and critical)
developmental process involving microRNAs, which we will
discuss later in this chapter. Furthermore, mouse embryo fibroblasts (MEFs) from wild-type cells showed normal RNAi,
but MEFs from Ago2 knockout mice were defective in
RNAi, as expected if Ago2 is important in RNAi.
All of the studies cited so far are consistent with the
hypothesis that Ago2 has slicer activity, but none addressed
this question directly. However, if Argonaute really has
slicer activity, then mutating any of the three acidic amino
acids at the putative active site should block cleavage of
mRNA by RISC. Hannon, Joshua-Tor, and colleagues mutated each of the two key aspartate residues and found that
either mutation abolished the RNAi-mRNA cleavage step
both in vitro and in vivo. Taken together, all this evidence
strongly implicates Ago2 as the slicer enzyme.
In 2005, Joshua-Tor and colleagues demonstrated definitively that human Ago2 really does have slicer activity.
They reconstituted a minimal RISC with human recombinant Ago2 and an siRNA, which could accurately cleave a
substrate RNA complementary to the siRNA. Figure 16.31
shows the results. The first siRNA (siRNA1) caused cleavage of the substrate RNA (S500) about 180 nt from its 39-end,
yielding a 39-product about 180 nt long and a 59-product
about 320 nt long. The second siRNA (siRNA2) caused
cleavage of the S500 about 140 nt from its 59-end, yielding
a 59-product about 140 nt long and a 39-product about
360 nt long. As expected, no products were produced in the
absence of siRNA. Nor did products appear in the absence
of Mg21, showing that a divalent metal ion is required for
slicer activity.
wea25324_ch16_471-521.indd Page 493
4:54 PM user-f469
16.6 Post-Transcriptional Control of Gene Expression: RNA Interference
+ Mg2+
No Mg2+
500 b
5′ product
3′ product
300 b
200 b
5′ product
150 b
3′ product
125 b
Figure 16.31 Ago2 plus an siRNA form a minimal RISC with slicer
activity in vitro. Joshua-Tor and colleagues mixed recombinant
human Ago2 (produced in bacteria) with either of two siRNAs that
were specific for two different sites on a target 500-nt RNA, as shown
at bottom. Then they added the labeled target RNA in the presence or
absence of Mg21 ions, as indicated at top. The siRNA used (either #1,
or #2, or neither) is also indicated at top. Finally, they displayed the
labeled RNA products by gel electrophoresis. Cleavage depended on
Mg21 and on an siRNA. The two siRNAs yielded different products,
whose sizes were predicted from the known sites on the target RNA to
which they hybridized. (Source: Reprinted from Nature Structural & Molecular
Biology, vol 12, Fabiola V Rivas, Niraj H Tolia, Ji-Joon Song, Juan P Aragon, Jidong
Liu, Gregory J. Hannon, Leemor Joshua-Tor, “Purified Argonaute2 and an siRNA
form recombinant human RISC,” fig. 1d, p. 341, Copyright 2005, reprinted by
permission from Macmillan Publishers Ltd)
For mRNA cleavage to occur, a catalytically active
RISC must form (Figure 16.32). We have seen that an
Argonaute protein contains the slicer active site in a RISC,
and we also know that a single-stranded siRNA must be
present to serve as a guide to select mRNAs to degrade. So
Ago2 plus siRNA constitutes a minimal RISC, at least in
mammalian cells. But this complex does not form directly.
Instead, siRNA must be delivered to Ago2 by a RISC loading
complex (RLC). The composition of the RLC is presumed
to include at least Dicer and a Dicer-associated protein,
cutely-named R2D2, in addition to siRNA, and it could
also include Armitage, which is essential for converting an
RLC to a RISC in Drosophila.
What is the role of R2D2? It is not required for doublestranded siRNA formation, as Dicer can carry out this process efficiently without R2D2 in vitro. However, gel
mobility shift and protein–RNA cross-linking experiments
have shown that Dicer alone cannot retain contact with
siRNAs once it has made them, but Dicer plus R2D2 can.
Furthermore, R2D2 contains two double-stranded RNAbinding domains, and mutations in these domains render
the Dicer–R2D2 complex incapable of binding doublestranded siRNAs. Thus, it appears that R2D2 is an essential part of the RLC because it can shepherd the siRNA
between the time it is formed by Dicer and the time it is
delivered to the RISC.
Slicer clipping
+ fragment removal
Figure 16.32 Delivery of single-stranded siRNA to RISC. The names
of the proteins are from Drosophila, in which this process has been well
studied. (a) Ago2 is attracted to a Dicer (DCR-2)-R2D2-dsRNA, forming
a pre-RISC complex. The ds siRNA has already been created by DCR-2,
leaving phosphorylated 59-ends and 2-nt 39-overhangs. (b) The slicer
activity of Ago2 cuts the passenger strand (top) in half, weakening its
base-pairing to the guide strand. The passenger strand fragments are
lost, leaving the guide strand bound to Ago2, which is the catalytic
center of the mature RISC. Other proteins besides Ago2 are part of
mature RISC, though they are not shown here.
How are the two strands of the ds-siRNA separated to
yield the ss-siRNA that ultimately associates with the
RISC? An early hypothesis was that Armitage, which has
RNA helicase activity, separated the two strands. However,
that would require ATP, and the two RNA strands can be
separated without ATP, at least in Drosophila. Figure 16.32
presents a model that incorporates that fact and other data.
A complex composed of double-stranded siRNA plus Dicer
(DCR-2 in Drosophila) and R2D2 attracts an Argonaute
protein (Ago2 in Drosophila). Then Ago2 cleaves the
passenger strand (the discarded strand) of the siRNA in the
middle, weakening its grip on the guide strand (the strand
that will associate with the RISC), so the passenger strand
fragments are lost. This leaves a RISC active center composed of Ago2 and the siRNA guide strand.
What determines which strand is the guide strand, and
which is the discarded passenger strand of the siRNA? This
distinction is made in a complex that forms before the
RLC, and contains Dicer and R2D2, each of which binds
to an end of the double-stranded siRNA. The two proteins
appear to bind asymmetrically, with Dicer associated with
wea25324_ch16_471-521.indd Page 494
4:54 PM user-f469
Chapter 16 / Other Post-Transcriptional Events
the less stable end (the one in which the base pairs are
easiest to dissociate). And the strand with its 59-end bound
to Dicer is the one that becomes the guide strand.
X-ray crystallography studies on complexes between
siRNAs and Argonaute-like proteins have shown that the
siRNA guide strand binds with 39-end in the PAZ domain.
This places the active site of Argonaute between residues
10 and 11 of the siRNA, so the mRNA would be cleaved
right in the middle of the siRNA–mRNA hybrid.
What is the physiological significance of RNAi? True
double-stranded RNA does not normally occur in eukaryotic
cells, but it does occur during infection by certain RNA
viruses that replicate through dsRNA intermediates. So one
important function of RNAi may be to inhibit the replication
of viruses by degrading their mRNAs. But Fire and other
investigators have also found that some of the genes required
for RNAi are also required to prevent certain transposons
from transposing within the genome. Indeed, Titia Sijen and
Ronald Plasterk showed in 2003 that transposition of the Tc1
transposon in C. elegans germ cells is silenced by RNAi. What
double-stranded RNA triggers this RNAi? It appears that
transcription of the terminal inverted repeats of the transposon yields an RNA that can form a stem-loop structure, which
is double-stranded in the stem portion. Thus, RNAi can protect cells not only against viruses, but also against transposition that can threaten the genomic integrity of germ cells.
RNAi can also silence transgenes and their genomic
homologs. How is double-stranded RNA made from transgenes? It seems that some transcription of both strands of
transgenes occurs, in contrast to the behavior of normal
genes. This symmetric transcription yields enough doublestranded RNA to trigger RNAi.
Aside from its natural functions, RNAi has been a terrific boon to molecular biologists because it enables them
to inactivate genes at will, simply by introducing doublestranded RNAs corresponding to the target genes. This
process, known as knockdown, is usually much more convenient than the laborious process of producing knockout
organisms, as described in Chapter 5. Also, it has not escaped the notice of the biotechnology industry that RNAi
represents a potential bonanza. We know of many genes
which, when overactive, can have devastating effects. For
example, many oncogenes become hyperactive in various
cancer cells, and that hyperactivity is what drives the cancer cells to lose control over their growth. RNAi directed
against these oncogenes could control their activities, and
thereby restore growth control to the cancer cells.
In spite of all this optimism, some caution is warranted because data began accumulating in 2004 that
RNAi is not as exquisitely specific as had been thought.
Genes that do not match the trigger double-stranded
RNA perfectly are still targeted for repression to some
extent. We do not know yet whether this nonspecificity
will seriously compromise the effectiveness of RNAi in
research and medicine.
Furthermore, if scientists want to use RNAi to investigate human gene function, or even to combat human disease, they will have to take account of another fact: Unlike
in roundworms and fruit flies, the RNAi induced by adding
dsRNA to mammalian cells is transient. But there is a way
around this problem: Lasting RNAi can be induced by
transforming mammalian cells with genes encoding RNAs
with inverted repeats that form hairpins. These genes provide a continuous supply of double-stranded RNA in the
form of hairpins, and that is enough to keep the RNAi process going. By 2004, researchers had already built libraries
of genes encoding short hairpin RNAs (shRNAs) that targeted almost 10,000 human genes. These represent a valuable resource for research, and perhaps even intervention
in human disease.
SUMMARY RNA interference (RNAi) occurs when
a cell encounters dsRNA from a virus, a transposon,
or a transgene (or experimentally added dsRNA),
and results in destruction of the mRNA corresponding to the trigger dsRNA. The mechanism of RNAi
in Drosophila is as follows: The trigger dsRNA is
degraded into 21–23-nt fragments (siRNAs) by an
RNase III-like enzyme called Dicer. The doublestranded siRNA, with Dicer and the Dicer-associated
protein R2D2 recruit Ago2 to form a pre-RISC
complex that can separate the siRNA into its two
component strands: the guide strand, which will
base-pair with the target mRNA in the RNAinduced silencing complex (RISC) and guide cleavage
of the mRNA, and the passenger strand, which will
be discarded. Ago2 cleaves the passenger strand,
which then falls off the pre-RISC complex. The
guide strand of the siRNA then base-pairs with the
target mRNA in the active site in the PIWI domain
of Ago2, which is an RNase H-like enzyme, also
known as slicer. Slicer cleaves the target mRNA in
the middle of the region of its base-pairing with the
siRNA. In an ATP-dependent step, the cleaved
mRNA is ejected from the RISC, which can then
accept a new molecule of mRNA to be degraded.
Amplification of siRNA
One aspect of RNAi in some organisms, including plants
and nematodes, has been difficult to explain: its great sensitivity. Just a few molecules of dsRNA can set in motion a
process that totally silences a gene, not only in one cell, but
in a whole organism—and even the descendants of that
organism. This phenomenon led to the proposal that the
process is catalytic. Indeed, Dicer does create many molecules of siRNA out of the trigger dsRNA and the target
mRNA, but that seems insufficient to explain the power of
wea25324_ch16_471-521.indd Page 495
4:54 PM user-f469
16.6 Post-Transcriptional Control of Gene Expression: RNA Interference
dsRNA trigger
(a) Dicer
(b) Unwinding
(c) Priming
(e) Dicer
Target mRNA
SUMMARY In certain organisms, including C. elegans,
siRNA is amplified during RNAi. This happens
when antisense siRNAs hybridize to target mRNA
and prime synthesis of full-length antisense RNA by
an RNA-dependent RNA polymerase. This new
dsRNA is then digested by Dicer into new pieces of
New dsRNA
Figure 16.33 Amplification of siRNA. (a) Dicer chops up trigger
dsRNA to make siRNA. (b) The antisense strands of siRNA hybridize
to target mRNA. (c) RdRP uses the siRNA antisense strands as
primers and target mRNA as template to make long antisense strands.
(d) The product of step (c) is new trigger dsRNA. (e) Dicer chops up
the new trigger dsRNA to make more siRNA, which can start a new
round of priming and siRNA amplification. (Source: Adapted from
Nishikura. Cell 107 (2001) f. 1, p. 416.)
RNAi in organisms like C. elegans. Fire and colleagues
solved this riddle by showing that C. elegans cells employ
an enzyme: RNA-directed RNA polymerase (RdRP) that
uses antisense siRNAs as primers to make many copies of
siRNA, as shown in Figure 16.33.
To test this hypothesis, Fire and colleagues used an
RNase protection assay with a labeled sense strand probe
to detect antisense siRNA in C. elegans fed on bacteria
expressing trigger dsRNA at high levels. They used two
different triggers and found large amounts of new siRNA
produced in both cases. In addition, they discovered some
secondary siRNAs outside the bounds of the trigger RNA.
It is significant that these secondary siRNAs always corresponded only to the mRNA region upstream of the trigger sequence. This finding makes sense in the context of
RdRP activity, because the trigger siRNA should prime
synthesis toward the 59 (upstream)-end of the mRNA. Thus,
the discovery of secondary siRNAs also supports the
hypothesis that an RdRP amplifies the siRNA, using the
target mRNA as the template.
Thus, a mechanism does exist for amplifying the input dsRNA, and this could explain the great power of
RNAi. The first round of this mechanism depends on
priming by antisense siRNA on an mRNA template. This
model can explain the earlier finding of Fire and colleagues that modification of the antisense, but not the
sense, strand of the trigger dsRNA blocks RNAi. The
model is also compatible with the earlier discovery of an
RdRP in tomato cells, and the presence of homologous
genes in fungi, and other plants, that are required for
efficiency of RNAi.
Role of the RNAi Machinery
in Heterochromatin Formation
and Gene Silencing
In 2002, evidence began accumulating that implicated the
RNAi machinery in heterochromatin formation and gene
silencing, known as transcriptional gene silencing (TGS), as
well as in RNAi itself. Then investigators found that
siRNA-induced gene silencing can target a gene’s control
region through DNA and histone methylation.
RNAi and Heterochromatization Shiv Grewal, Robert
Martienssen, and their colleagues deleted the RNAi genes
encoding Dicer, Argonaute, and RdRP (dcr1, ago1, and
rdp1, respectively) in the fission yeast Schizosaccharomyces
pombe and found that all of these mutants were defective
in the silencing that normally affects transgenes inserted
near the centromere. That is, these transgenes became active in the RNAi mutants. Note that no trigger dsRNAs for
the transgenes had been added, so RNAi was not directly
involved in silencing the transgenes.
The investigators also looked to see whether the repeated DNA sequences (cen3 sequences) at the centromere
were transcribed in wild-type cells and in the mutants. Using Northern blots, they found no trace of such transcripts
in wild-type cells, but they found three abundant transcripts
in the RNAi mutants. A more detailed investigation using
RNA dot blots showed that the reverse transcript of the
cen3 sequences appeared in wild-type and mutant cells, but
the forward transcript appeared only in the mutants. Furthermore, nuclear run-on analysis demonstrated the same
pattern: forward transcripts only in the mutants. Thus, the
concentration of cen3 transcripts is controlled at the transcriptional, rather than the post-transcriptional, level.
Next, the investigators examined specific core histone
methylation in centromeric repeats using ChIP with antibodies against methylated histone H3 lysine 4 and lysine 9.
As we learned in Chapter 13, methylated lysine 4 of histone
H3 is associated with active genes, whereas methylated
lysine 9 correlates with heterochromatin and gene inactivity.
As expected from the activities we have already discussed,
wild-type cells had lysines 4 and 9 that were both methylated
in the centromeric region, but all three RNAi mutants showed
an aberrant pattern of centromeric histone H3 methylation:
wea25324_ch16_471-521.indd Page 496
4:54 PM user-f469
Chapter 16 / Other Post-Transcriptional Events
a high level of lysine 4 methylation, but a very low level of
lysine 9 methylation. The same pattern was found in a ura41
transgene placed in the outermost centromere region (otr): a
high level of lysine 9 methylation in wild-type cells, but a
greatly depressed level in all three RNAi mutants.
Is RNAi responsible for histone methylation, and the
resulting heterochromatization at the centromere? If so, we
would expect at least some RNAi proteins to interact with
centromeric chromatin, and we would also expect to find
siRNAs corresponding to centromeric RNA. Martienssen and
colleagues did indeed find that the Rdp1 part of the RNAi
machinery binds to centromeric chromatin. And B.J. Reinhard
and David Bartel had already found evidence to support the
second prediction of the hypothesis when they cloned apparent Dicer products from wild-type cells and showed that
all 12 clones came from transcripts of the centromeric region.
Thus, at least one component of the RNAi machinery is
found at the centromere, and siRNAs are made from centromeric transcripts. All these data, and more, led
Martienssen and colleagues to propose that RNAi is involved in heterochromatic silencing at the centromere
(Figure 16.34). In particular, they proposed that the abundant reverse transcripts of the otr region base-pair with forward transcripts produced occasionally by RNA polymerase
II, or perhaps by RdRP, to form trigger dsRNA. Dicer then
digests this dsRNA to produce siRNA, and the siRNA associates with an Argonaute1 protein (Ago1) in a complex
called RITS (for RNA-induced transcriptional silencing
complex). This complex can then attract RdRP in a complex known as RDRC (for RNA-directed RNA polymerase
complex) which amplifies the double-stranded siRNA. By
base-pairing either to the DNA directly or to transcripts of
the DNA, the siRNA then escorts RITS to corresponding
sites on the genome. RITS then causes recruitment of a histone H3 lysine 9 methyltransferase. Once a lysine 9 is methylated, it can recruit Swi6, which is required for forming
heterochromatin. Other proteins may be required, but the
end result is spreading of heterochromatin to the otr region
of the centromere. Whatever the mechanism, it is likely to
be highly conserved, because mammalian pericentromeric
heterochromatin structure also involves histone H3 lysine 9
modification and some RNase-sensitive substance, which
could be one or more of the RNAi intermediates.
Does the RITS complex associate directly with DNA, or
is it attracted by transcripts of chromatin regions that are
targeted for silencing? In 2006, Danesh Moazed and colleagues provided evidence for the importance of transcripts
in this process by showing that artificially tethering RITS
to a nascent transcript of the ura41 gene resulted in silencing of this normally active gene.
and reverse
Swi6 Swi6
(c) Dicer
(and RdRP)
Swi6 Swi6 H3
(e) RDRC
Swi6 Swi6 H3
Swi6 Swi6 H3
Figure 16.34 A model for the involvement of the RNAi machinery
in the heterochromatization at the S. pombe centromere. (a) The
outermost region (otr) of the centromere is constantly being
transcribed to produce reverse transcripts, and production of forward
transcripts probably also occurs at a low (undetectable) level. (b) After
transcription and reverse transcription (or after reverse transcription
and RdRP action), we have double-stranded RNA (dsRNA). (c) Dicer
cuts the dsRNA into siRNAs. (d) Ago1 (yellow, perhaps along with
other proteins) associates with single-stranded siRNAs to produce the
RITS. (e) The RdRP in the RDRC amplifies the siRNA, producing
double-stranded siRNAs. (f) The RITS, through its siRNA, associates
with the otr, either through direct interaction with the DNA, or through
interaction with transcripts in this region. (g) The RITS attracts a
histone methyltransferase (HMT, green) to the otr. (h) The HMT
methylates the lysine 9 of a histone H3 (blue). Of course, this histone
is part of a nucleosome, which is not shown here, for simplicity. (i) This
methylation in turn attracts more Swi6 (red), which helps to spread
wea25324_ch16_471-521.indd Page 497
4:54 PM user-f469
16.6 Post-Transcriptional Control of Gene Expression: RNA Interference
It seems paradoxical that, in order for a region like a
centromere to be silenced, it has to be expressed. How, then,
does expression occur after mitosis to preserve heterochromatization in the genomes of both progeny cells? A solution
to this paradox was proposed by Rob Martienssen and colleagues and Grewal and colleagues in 2008. Together, the
work of these two groups showed that serine 10 of histone
H3 in centromeric heterochromatin in S. pombe becomes
phosphorylated during mitosis, and that this results in the
loss of methylation of lysine 9 of histone H3, and therefore
in the loss of the Swi6 protein that is necessary for heterochromatization. As a result, the chromatin opens up enough
that it is transcribed during the S phase. This produces centromere transcripts, presumably in both directions, that
attract the RNAi machinery, so the centromere can be heterochromatized again during the ensuing long G2 phase.
This hypothesis views heterochromatin as more dynamic than the traditional view of a static, condensed, inactive structure. Does it also open up the possibility of real
expression of centromeric DNA? Apparently not. For one
thing, centromeric transcription is confined to the S phase,
in which gene expression is very restricted. For another, the
centromeric transcripts are rapidly degraded, either by the
RNAi machinery, or by other RNA-degrading systems that
recognize aberrant transcripts.
Grewal and colleagues noted that centromere-like sequences are also found at sites such as the silent matingtype region, which lies far from the centromere but is also
silenced by heterochromatization. In separate experiments,
these workers showed that the RNAi machinery is required
for initiating heterochromatization at the silent matingtype region, but is expendable for maintaining and inheriting the silencing. Swi6 is apparently sufficient for such
heterochromatin maintenance.
The role of the RNAi machinery in centromeric events
is not confined to lower organisms. In 2004, Tatsuo
Fukagawa and colleagues reported tests on a chicken–
human hybrid cell line whose only human chromosome
was chromosome 21. These workers then made the Dicer
gene tetracycline-repressible in these hybrid cells and observed
what happened, particularly to human chromosome 21,
when Dicer expression was blocked by tetracycline. The
most obvious effect of the loss of Dicer was that the cells
died after about five days.
Moreover, the specific pathologies of these cells point to
problems with the centromere: The cells showed abnormal
mitoses with evidence of premature sister chromatid separation. As in yeast cells with defective RNAi, these vertebrate cells exhibited abnormal buildup of transcripts of the
centromeric repeat region of human chromosome 21. They
also showed abnormal localization of some, but not all,
centromeric proteins. The problems at the centromere were
presumably caused by the loss of Dicer, and this in turn led
to the failure of cell division and to cell death.
We assume that the events that occur in the centromeric
region in fission yeast, illustrated in Figure 16.34, help to
explain these results in cells from higher organisms. However,
one caveat to bear in mind is that mammals appear to lack an
RdRP. So any dsRNA that appears at the centromere in
mammals must be made by bidirectional transcription of this
region, or of a homologous region elsewhere in the genome.
Another major difference between heterochromatization in fission yeast and in plants and mammals is that the
latter organisms experience DNA methylation in addition
to histone methylation. The methyl groups are added to the
C’s of CpG sequences in both strands, and these help to
attract the proteins that induce heterochromatization.
Again, the presence of double-stranded RNA appears to
play a key role by recruiting the RNAi machinery, which
stimulates DNA methylation.
One significant advantage of this mechanism is that it is
permanent. Once the DNA is methylated on the C’s of both
strands of a CpG sequence, this methylation is inherited
from one cell generation to the next, as the methylated C on
one strand ensures that the new C on the opposite strand
will also be methylated after DNA replication. Although
this methylation is permanent, it is not a true genetic change,
which would be a change of one base to another (e.g., a C
changed to a T). Instead, we call it an epigenetic modification of the DNA. It is every bit as important as a genetic
change because it can cause the silencing of a gene or even
heterochromatization of a whole region of a chromosome.
RNAi may also play a role in X chromosome inactivation in mammals. In each cell of a female mammal, one of
the X chromosomes is inactivated by heterochromatization.
This prevents the very deleterious consequences of elevated
levels of X chromosome products. One of the first steps in
X chromosome inactivation is histone H3 lysine 9 methylation. And this methylation occurs immediately after the appearance of a noncoding transcript of the Xist locus. We
also know that Xist is controlled by the antisense RNA,
Tsix, and by Xist promoter methylation. The presence of
Tsix and Xist transcripts in the same cell would of course
invoke the RNAi system, and that could recruit the histone
methylase that kicks off the formation of heterochromatin.
SUMMARY The RNAi machinery is involved in het-
erochromatization at yeast centromeres and silent
mating-type regions and is also involved in heterochromatization in other organisms. At the outermost regions of centromeres of fission yeast, active
transcription of the reverse strand occurs. Occasional forward transcripts, or forward transcripts
made by RdRP, base-pair with the reverse transcripts to kick off RNAi, which in turn recruits a
histone methyltransferase, which methylates lysine
9 of histone H3, which recruits Swi6, which causes
heterochromatization. In plants and mammals, this
process is abetted by DNA methylation, which can
also attract the heterochromatization machinery.
wea25324_ch16_471-521.indd Page 498
4:54 PM user-f469
Chapter 16 / Other Post-Transcriptional Events
Transcriptional Gene Silencing Induced by siRNA Directed
at a Gene’s Control Region Kevin Morris and colleagues
found in 2004 that mammalian genes can also be silenced
by the RNAi machinery and, as we have seen with heterochromatization in plants and mammals, this silencing
involves DNA methylation. Furthermore, in contrast to
normal RNAi, this silencing involves an siRNA directed at
the control region, rather than the coding region, of a gene.
Morris and colleagues targeted a green fluorescent
protein reporter gene driven by the human elongation
factor 1a gene (EF1A) promoter-enhancer region. They
transduced human cells with feline immunodeficiency
virus (FIV) containing this reporter construct, which
caused integration of the reporter gene and its control
region into the human genome. The FIV vector also made
the nuclear membrane permeable to the siRNA, which
otherwise would not have been taken up by the mammalian nuclei.
Because the siRNA in this case was directed against the
gene’s control region, and not its coding region, we would
predict that it could not cause mRNA destruction or block
translation. Indeed, we would predict that it would block
transcription, and indeed that is what Morris and colleagues showed. Using real-time RT-PCR (Chapter 4), they
demonstrated almost total disappearance of the GFP transcript upon transducing cells with the EF52 siRNA, which
targets the control region of the fusion gene. By contrast,
an siRNA that targets the coding region of the GFP mRNA
caused a relatively modest 78% reduction in the concentration of the GFP transcript (Figure 16.35a).
Because a common feature of transcriptional silencing
in mammals is histone and DNA (cytosine) methylation,
Morris and colleagues tested the effect of trichostatin (TSA)
and 5-azacytidine (5-azaC), which inhibit histone and
DNA methylation, respectively. These drugs completely reversed the silencing caused by the EF52 siRNA, but had no
effect on silencing caused by the GFP coding region siRNA.
These results supported the hypothesis that DNA and/or
histone methylation are involved in silencing caused by the
EF52 siRNA.
To check whether the silencing by the EF52 siRNA was
at the transcription level, Morris and colleagues performed
nuclear run-on assays (Chapter 5). Figure 16.35b shows
that EF52 did indeed dramatically reduce the number of
initiated GFP transcripts, while it had no effect on irrelevant glyceraldehyde-phosphate dehydrogenase (GAPDH)
To see whether DNA in the gene’s control region was
really methylated during transcriptional silencing, Morris
and colleagues used HinP1I, a restriction enzyme that cuts
at a site that includes a CpG. If the C in this sequence is
unmethylated, HinP1I will cut, but if it is methylated it will
not. There is a HinP1I site in the control region of the
EF1A gene. Thus, if this site is methylated, it will be protected from HinP1I cleavage, and PCR using primers on
opposite sides of the site will produce a product. On the
other hand, if the site is unmethylated, HinP1I will cut it,
and no PCR product will appear.
Figure 16.36 shows the results of this experiment. The
control in lane 1 shows that a plasmid with a HinP1I site
methylated in vitro really does yield a PCR product, even
after attempted cleavage with HinP1I. Lanes 2 and 3 are
controls with DNA from cells that had been transduced
with an irrelevant siRNA or a GFP coding region siRNA,
GFP mRNA expression
No drug
TSA + 5-azaC
Control GFP
siRNA treatment
EF52 Control GFP
Figure 16.35 Silencing by an siRNA targeting the EF1A gene
control region. (a) Real-time PCR assay for GFP mRNA in human
cells bearing a GFP gene driven by the EF1A gene promoterenhancer region. Cells were transduced with FIV bearing the GFP
gene construct, and then siRNAs were added in the absence (no
drug), or presence of TSA and 5-azaC. Then real-time PCR was
performed to measure the concentration of GFP mRNA. The bars
(and corresponding quantifications) show the results with no siRNA
(control), an siRNA that targets the coding region of the mRNA
(GFP), and an siRNA that targets the EF1A gene control region
(EF52). (b) Nuclear run-on assay for transcription. Nuclei were
isolated from cells transduced with the EF1A-GFP construct, plus
either the EF52 siRNA or no siRNA (control). Labeled nuclear
run-on mRNA was synthesized and hybridized to blots of GFP
DNA, or GAPDH DNA, as indicated at left. The EF52 siRNA silenced
the GFP gene, but not the GAPDH gene, at the transcriptional
level. (Source: Reprinted with permission from Science, Vol. 305, Kevin V.
Morris, Simon W.-L. Chan, Steven E. Jacobsen, and David J. Looney, “Small
Interfering RNA-Induced Transcriptional Gene Silencing in Human Cells,”
Fig. 1, p. 1290, Copyright 2004, AAAS.)
wea25324_ch16_471-521.indd Page 499
4:54 PM user-f469
16.6 Post-Transcriptional Control of Gene Expression: RNA Interference
HinP1I digest + EF1A promoter PCR
FIV transduced,
siRNA transfected cells
No drug
Figure 16.36 Demonstration of methylation of the EF1A gene
control region in response to siRNA. Morris and colleagues tested
for methylation of a CpG sequence in the EF1A control region by
cleavage with HinP1I, which cleaves unmethylated, but not methylated
sites including CpG sequences. They performed the cleavage on DNA
from cells either untreated (top row, “No drug”) or treated (bottom row)
with TSA plus 5-azaC to block methylation of CpG sequences. After
treatment with HinP1I, they performed PCR with primers flanking the
CpG site. Only uncut (methylated) DNA should yield a signal. Lane 1,
positive control with synthetically methylated site. Lane 2, negative
control with irrelevant siRNA. Lane 3, negative control with an siRNA
directed against the GFP coding region, rather than the control region.
Lane 4, experimental result with an siRNA that targets the control
region. With this siRNA, the CpG is methylated (uncut, and therefore
yields a PCR signal) in the absence of drug, but is not methylated
when the methylation blocker was included. (Source: Reprinted with
permission from Science, Vol. 305, Kevin V. Morris, Simon W.-L. Chan, Steven E.
Jacobsen, and David J. Looney, “Small Interfering RNA-Induced Transcriptional
Gene Silencing in Human Cells,” Fig. 1, p. 1290, Copyright 2004, AAAS.)
respectively. Lane 4 shows the results with cells transduced
with the EF52 siRNA. The top row shows that the DNA
must have been methylated, because it was protected from
HinP1I cleavage, and a PCR product appeared. However,
the bottom row shows that the methylation-blocking drugs
TSA and 5-azaC, blocked methylation, rendering the
HinP1I site cleavable, so no PCR product appeared.
All of the experiments described so far used cells that
were transduced with FIV, which inserted the EF1A gene into
the human genome, but not in its natural location. To check
for siRNA silencing of the endogenous human gene, Morris
and colleagues performed the same kinds of experiments as
in Figures 16.35 and 16.36, but with cells rendered permeable to siRNAs with MPG, a fusion peptide that contains an
HIV-1 transmembrane peptide linked to the nuclear localization signal from SV40 virus. In these experiments, no EF1A
gene was introduced into the cells, so only the endogenous
gene was present, and it was silenced (though not as dramatically as in the previous experiments) by the EF52 siRNA.
As before, this silencing was accompanied by DNA methylation, and could be blocked by methylation inhibitors.
Where does the siRNA in these experiments come
from? After all, it is directed at the control region, not the
coding region, of the gene, so it cannot come from a normal gene transcript. Morris and colleagues showed that the
sense strand part of the siRNA probably came from a
59-extended transcript of the EF1a gene—that is, a transcript that started in the promoter, upstream of the normal
transcription start site. They detected this extended transcript with an RNA pull-down procedure that used a
59-biotin-labeled promoter antisense RNA and avidin
bound to magnetic beads. The biotin-labeled promoter antisense RNA hybridized in vivo to the RNA transcribed
through the promoter region, and the avidin-tagged beads
bound to the biotin, allowing the whole RNA-RNA-bead
complex to be isolated (“pulled down”) magnetically.
Quantification of the promoter-associated RNA and
the normal EF1a transcripts by real-time RT-PCR yielded a
ratio of about 1:570. Thus, about one in 570 transcripts of
the EF1a gene begins within the promoter. A 59-RACE procedure (Chapter 5) showed that these promoter-associated
transcripts begin about 230 bp upstream of the normal
transcription start site, and a 39-RACE procedure showed
that these transcripts extend as far in the 39-direction as the
normal transcripts and are spliced and polyadenylated.
Does the promoter-associated RNA play a role in transcriptional gene silencing (TGS)? To answer this question,
Morris and colleagues targeted the promoter-associated RNA
for destruction by RNase H (Chapter 14), by transfecting cells
with a promoter-associated RNA-specific phosphorothioate
oligonucleotide, which acts like a deoxyribo-oligonucleotide
in this procedure. The destruction of the EF1a promoterassociated RNA abolished transcriptional silencing by
added promoter-associated siRNA. By contrast, RNase
H-mediated destruction of a promoter-associated RNA
from another gene (CCR5) had no effect on TGS of the
EF1a gene. Thus, a promoter-associated RNA appears to
be essential for TGS.
One of the epigenetic changes that occurs in the EF1a
control region during gene silencing is a trimethylation of
lysine 27 of histone H3 (H3K27me3) in a nucleosome at
that site. Does the promoter-associated RNA play a role in
this epigenetic change? A pull-down assay showed that it
does. When the EF1a promoter-associated RNA was
destroyed by oligonucleotide and RNase treatment, the
chromatin could no longer be precipitated with an antiH3K27me3 antibody. On the other hand, treatment with
the irrelevant oligonucleotide directed at the CCR5 control
region did not block precipitation of the EF1a promoterassociated nucleosome with an anti-H3K27me3 antibody.
Thus, the presence of the promoter-associated RNA is
required for the silencing methylation of H3K27. The exact
nature of that requirement is still unclear, but one can imagine that the promoter-associated RNA would hybridize to an
antisense RNA (perhaps the antisense strand of an siRNA).
This hybrid would in turn recruit a chromatin remodeling
complex, including the H3K27 methyltransferase, which
would trimethylate H3K27, helping to silence the gene.
All of the silencing we have discussed so far is due to
epigenetic modification (usually methylation) of chromatin.
wea25324_ch16_471-521.indd Page 500
4:54 PM user-f469
Chapter 16 / Other Post-Transcriptional Events
Another silencing mechanism targets nuclear RNA:
Endogenous double-stranded siRNAs can enter the nucleus
and cause degradation of nuclear RNAs by the familiar
RNAi mechanism. Scott Kennedy and colleagues showed in
2008 that siRNAs bind to an Argonaute protein (NRDE-3
in C. elegans) in the cytoplasm. NRDE-3 has a nuclear localization signal that targets it to the nucleus, so the siRNANRDE-3 complex can enter the nucleus and collaborate in
the destruction of cognate nuclear pre-mRNAs. Note that
the nuclear location distinguishes this mechanism from
ordinary RNAi, which occurs in the cytoplasm.
SUMMARY Individual genes in mammals can also
be silenced by an RNAi mechanism that targets the
control region, rather than the coding region, of the
gene. This silencing process involves DNA and histone methylation, rather than mRNA destruction.
One requirement for such histone methylation in
siRNA-induced gene silencing, at least in some
genes, is production of a 59-extended transcript that
begins within the gene’s control region (a
promoter-associated transcript). This transcript presumably associates with an antisense RNA, and
then recruits a chromatin remodeling complex, including a histone methyltransferase, which methylates H3K27 on a nearby nucleosome, helping to
silence the gene. Genes can also be silenced by a
nuclear RNAi process that involves Argonaute proteins that are targeted to the nucleus by a nuclear
localization signal.
Transcriptional Gene Silencing in Plants The short
RNAs required for TGS in fission yeast and animals are
made by RNA polymerase II. But in TGS in flowering
plants, two other polymerases, RNA polymerase IV and
RNA polymerase V, which are evolutionarily derived
from polymerase II, play the key roles. Polymerase IV
produces the 24-nt heterochromatic siRNAs whose yeast
and animal counterparts are made by polymerase II. The
role of polymerase V is more subtle, and was therefore
more difficult to unravel.
Polymerase V produces transcripts of non-coding regions that are more than 200 nt long, have either caps or
triphosphates at their 59-ends, and are not polyadenylated.
Transcripts in a given region have multiple 59-ends, which
suggests they are made in a promoter-independent manner.
In 2008, Craig Pikaard and colleagues demonstrated the
involvement of polymerase V in transcriptional gene silencing by mutating the largest subunit of the enzyme. They
observed, in addition to loss of polymerase V activity, loss
of transcripts of certain non-coding regions, and defective
silencing in overlapping and adjacent chromatin regions.
Furthermore, they found that some of the hallmarks of
heterochromatin, including histone and DNA methylation,
were lost in cells lacking polymerase V activity.
How do the polymerase V transcripts attract the silencing machinery? Pikaard and colleagues proposed a model
very similar to that in Figure 16.34, except that polymerases IV and V play roles performed by polymerase II in
fungi and animals. The polymerase V transcripts attract a
complex composed of Argonaute 4 (Ago4) and siRNA
(made by polymerase IV). This complex in turn attracts the
silencing machinery. In 2009, Pikaard and colleagues provided more support for this hypothesis, as follows. First,
they performed ChIP analysis with chromatin from Arabidopsis plants that produce mutant Ago4 and polymerase V.
They found that both wild-type Ago4 and polymerase V
bound to transposon genes that are normally silenced, but
mutations in either the Ago4 gene or the nrpe1 gene, which
encodes the largest polymerase V subunit, abolished this
association. Thus, Ago4 and polymerase V are necessary
for Ago4 to associate with chromatin that is to be silenced.
To test whether polymerase V transcripts are required
to recruit Ago4 to chromatin, Pikaard and colleagues performed ChiP analysis in wild-type plants, and in plants
bearing a mutation at the active site of the largest subunit
of polymerase V. The mutant polypeptide is stable and can
still bind normally to the second-largest subunit, but it is
utterly incapable of making transcripts. ChIP analysis
showed no binding of Ago4 to target chromatin sites in the
mutant plants. This binding could be restored by transforming plants with the wild-type nrpe1 gene, but not with
the mutant gene. Thus, transcription by polymerase V is
required to recruit Ago4, in accord with the hypothesis.
It is important to note that polymerase V transcripts are
found throughout the genome of Arabidopsis thaliana, a
member of the mustard family, in heterochromatic and
euchromatic regions alike. How then do the euchromatic
regions avoid silencing? Pikaard and colleagues proposed
that polymerase V transcripts are necessary, but not sufficient, for silencing. The silencing process also requires
siRNAs. Therefore, because euchromatic regions do not
give rise to siRNAs, they are not silenced.
Earlier in this chapter, we discussed the paradox that
silenced chromatin must be transcribed in order to be
silenced. The existence of polymerases IV and V gives
flowering plants a way to deal with this problem: These
polymerases appear not to initiate at promoters, and they
are not subject to the same rules as polymerase II. Thus,
they can presumably initiate transcription even in chromatin regions that are silenced with respect to polymerase II.
SUMMARY Flowering plants have two nuclear RNA
polymerases, polymerase IV and polymerase V, that
are not found in animals and fungi. Polymerase IV
makes siRNAs corresponding to chromatin regions
Fly UP