PAGE
PRO
O
FS
18
C H A P T E R
Regulatory RNAs
W
E DISCUSSED IN THE PREVIOUS TWO CHAPTERS how transcription is
regulated in prokaryotes and eukaryotes. We learned that this
control is achieved using regulatory proteins—typically, sequence-specific
DNA-binding proteins that either activate or repress
transcription of nearby genes. The mechanistic details of gene regulation
have been studied since François Jacob and Jacques Monod proposed
their model of repression almost 50 years ago (Box 16-2). At
that time, they could not say whether the trans factors (repressors)
were proteins or RNA. It transpired that in the cases they studied (and
indeed most other cases), the regulators were proteins that worked by
binding the operator sites on DNA. But in their original paper they
suggested that the regulators could just as easily be RNA molecules—
indeed, they favored that possibility.
The idea that RNA molecules might be regulators was largely forgotten
as more and more protein regulators were found in both
prokaryotes and eukaryotes. But in recent years, there has been an explosion
in the study of RNA regulators, particularly in eukaryotes,
that operate at the level of transcription and especially translation.
This new field emerged from two sources: the discovery of microRNAs,
first reported in the early 1990s, and then the discovery of the
phenomenon known as RNA interference in the late 1990s. Before we
describe these forms of regulation—how they work and the applications
they afford researchers—we consider cases of RNA-mediated
gene regulation first described in bacteria.
REGULATION BY RNAS IN BACTERIA
Small RNAs have been recognized in prokaryotes for many years.
Some are involved in regulating the replication of plasmids, and others
are involved in regulating gene expression (see the discussion of
Tn10 in Chapter 11). Of the latter group, some of these RNAs control
transcription—the 6S RNA of Escherichia coli, for example. This
RNA binds to the σ70
subunit of RNA polymerase and down-regulates
transcription from many σ70
promoters. The 6S RNA accumulates at
high levels in stationary phase (the growth phase bacteria enter as nutrients
become depleted and the cells stop dividing; see Chapter 22).
In stationary phase, an alternative σ factor, σS
, is made. This σ comO
U T L I N E
Regulation by RNAs in
Bacteria, 633
•
RNA Interference Is a Major Regulatory
Mechanism in Eukaryotes, 641
•
Synthesis and Function of miRNA
Molecules, 643
•
The Evolution and Exploitation of
RNAi, 652
•
Regulatory RNAs and
X-inactivation, 657
•
Visit Web content for Structural
Tutorials and Interactive
Animations
633
PAGE
PRO
O
FS
petes with σ70
for core polymerase and directs the enzyme to promoters
expressing genes for the multiple stress responses needed to survive
stationary phase. By down-regulating transcription from σ70
promoters,
6S RNA helps this shift in expression to σS
promoters.
In recent years, attention has focused on another group of small
RNA molecules in bacteria that regulate translation and mRNA degradation.
Interest in these small RNAs has been heightened by their
similarity to RNAs that regulate gene expression in eukaryotes—the
small interfering and microRNAs we discuss in the second half of this
chapter. The bacterial small RNAs (called sRNAs) are larger (80–110
nucleotides) than those regulatory RNAs from eukaryotes (which
range from 21 to 30 nucleotides). And they are not generally formed
by processing of larger dsRNA precursors (as those eukaryotic RNA
regulators are); instead, they are encoded in their final form by small
genes. Many of these genes have been identified by bioinformatics,
with close to 100 sRNAs being uncovered in E. coli. Of these, about a
dozen have been characterized. Most sRNAs work by base pairing
with complementary sequences within target mRNAs and directing
destruction of the mRNA, inhibiting its translation or even in some
cases stimulating translation.
Binding of an sRNA to its target mRNA is in most cases aided by the
bacterial protein Hfq. This RNA chaperone is needed because the complementarity
between the the sRNAs and their target mRNAs is typically
imperfect and short, and so their interaction is weak. Hfq facilitiates
base pairing; also, by binding the sRNAs even before they are
paired with their targets, Hfq increases the stability of these regulators.
A well-studied sRNA from E. coli is the 81-nucleotide RybB RNA.
This sRNA binds several target mRNAs and triggers their destruction
because the double-strand stretch of heteroduplex formed upon pairing
is recognized as a substrate by the nuclease RNase E. Most of the
mRNAs targeted by RybB encode iron storage proteins. Free iron is required
by the cell under certain circumstances, but high levels are
toxic. RybB regulates the levels of free iron by controlling the levels of
iron storage proteins. RybB is expressed from a promoter recognized
by a special σ factor called σE
(like σS
, a stress response σ factor). Expression
of the gene encoding σE
is itself regulated by RybB, and so
this sRNA is part of an autonegative regulatory loop for σE
.
The stationary-phase σ factor σS
, mentioned above, is encoded by
the rpoS gene of E. coli. Translation of rpoS mRNA is stimulated by
two sRNAs: DsrA and RprA. Activation is achieved by a switch in
alternative RNA base pairing: the small RNAs bind to a region of the
mRNA that otherwise would pair with the ribosome-binding site, inhibiting
translation. The rpoS gene is also acted on negatively by another
small RNA, OxyS. Figure 18-1 shows these two mechanisms.
We shall return to trans-acting regulatory RNAs in the second half
of this chapter where we consider their role in regulating gene expression
in eukaryotes. But before turning to that topic, we consider other
examples in bacteria of gene regulation mediated through alternative
RNA pairing that operate in cis—that is, RNA regulatory elements that
control expression of the genes within whose mRNAs they reside. The
most striking examples are the so-called riboswitches that control
metabolic operons and attenuation in biosynthetic operons. The trp
genes of E. coli are the classic example of the latter mechanism and
are where RNA-mediated regulation was discovered (we shall describe
this case in detail in Box 18-1).
634 Chapter 18
PAGE
PRO
O
FSRiboswitches Reside within the Transcripts of Genes Whose
Expression They Control through Changes in Secondary Structure
Riboswitches control gene expression in response to changes in the
concentrations of small molecules. These regulatory elements are typically
found within the 5′-untranslated regions of the genes they control,
and they can regulate expression at the level of transcription or
translation. They do this through changes in RNA secondary structure,
as we shall see.
Each riboswitch is made up of two components: the aptamer and
the expression platform (Fig. 18-2). The aptamer binds the small-molecule
ligand and, in response, undergoes a conformational change,
which in turn causes a change in the secondary structure of the adjoining
expression platform. These conformational changes alter expression
of the associated gene by either terminating transcription or inhibiting
the initiation of translation. Both mechanisms are illustrated
in the example shown in Figure 18-3, and which we now describe.
Riboswitches are, not surprisingly, typically found upstream of
genes involved in the synthesis of the metabolite ligand recognized
by the riboswitch in question. For example, in Bacillus subtilis, many
genes involved in the use of the amino acid methionine have a 200nucleotide-long
untranslated leader RNA that acts as a SAM (Sadenosylmethionine)-sensing
riboswitch. RNA polymerase initiates
transcription at the promoter and transcribes through this leader region
before entering the coding sequence of the downstream genes.
Once transcribed into RNA, the leader region can adopt alternative
structures through alternative patterns of intramolecular base pairing
Regulatory RNAs 635
F I G U R E 18-1 Activation and repression of translation by sRNAs. RBS indicates the
ribosome-binding site. When this is occluded by base pairing with another RNA molecule—as
in part b—or another region of the same RNA molecule—as in part a—translation
is inhibited. (Adapted, with permission, from Gottesman S. et al. 2006. Cold Spring
Harbor Symp. Quant. Biol. 71: 1–11, Fig. 1. © Cold Spring Harbor Laboratory Press.)
coding sequence
activation
sRNA pairing unmasks the RBS
repression
sRNA pairing inhibits the ribosome binding
RBS
RBS coding sequence
coding sequenceRBS
RBS coding sequence
a
b
sRNA
sRNA
aptamer
expression
platform
coding sequence
RBS
5'
F I G U R E 18-2 Organization of riboswitch
RNAs. As described in the text, the aptamer
binds the controlling metabolite, causing
changes in the structure of the adjoining expression
platform. The aptamers identified to
date vary in size from 70 to 200 nucleotides;
the expression platforms vary more in both
size and character.
PAGE
PRO
O
FS
(Fig. 18-3a). One arrangement includes a stem-loop transcriptional
terminator (see Chapter 12). SAM—the ligand for this riboswitch—
binds to the aptamer and stabilizes the secondary structure that includes
this transcriptional terminator (as shown in the bottom part
of Fig. 18-3a). Under these circumstances, transcription is terminated
before polymerase has a chance to transcribe the downstream protein-coding
segment of the gene. This form of transcriptional regulation
is also called attenuation. In another case—at another gene—a
SAM-sensing riboswitch can work by regulating translation. In that
case, as shown in Figure 18-3b, the alternative secondary structure
stabilized by SAM binding to the aptamer includes a stem loop that,
although not a transcriptional terminator, does include the ribosomebinding
site (RBS; within region 4). This conformational change sequesters
the RBS and blocks ribosomes from initiating translation.
This form of translation inhibition is thus essentially identical to that
described for trans-acting sRNAs above (Fig. 18-1). The details of the
changes in RNA secondary structure induced by SAM binding to a riboswitch
are shown in Figure 18-4.
Many riboswitches have been identified and these respond to a
range of different metabolites. Ligands include lysine and other amino
636 Chapter 18
3'
3'
321
4
ON
b
RBS
SAM
sequestering
helix
5'
21 43
OFF
5'
3'
321
4
ON
a
SAM
transcriptional
terminator
5'
21 43
OFF
5'
F I G U R E 18-3 Riboswitches regulate
transcription termination or translation initiation.
Shown are two examples of a SAMsensing
riboswitch, in one case (a) regulating
transcription termination, in the other (b)
translation initiation. Numbers 1–4 indicate
different sequence elements within the RNA
upstream of the coding region (which is
shown in yellow). In the absence of SAM, regions
2 and 3 form a stem loop; in the presence
of SAM, regions 1 and 2 form a stem
loop, and regions 3 and 4 do likewise. The
consequence of that change in secondary
structure controls transcription or translation
as shown. (a) A stem loop of region 3 and 4
produces a transcriptional terminator, which
triggers RNA polymerase to terminate transcription
immediately after transcribing those
regions and before entering the downstream
coding region. The stem loop in this case is
followed by a stretch of Us in the mRNA, another
feature of transcriptional terminator (Fig.
12-12). (b) The stem loop formed by regions 3
and 4 inhibits translation initiation by sequestering
the ribosome binding site, as shown.
PAGE
PRO
O
FS
Regulatory RNAs 637
acids, vitamin B12, coenzyme thiamine pyrophosphate (TPP), flavin
mononucleotide (FMN), and guanine (Fig. 18-5).
Another kind of riboswitch responds to uncharged tRNAs, rather
than to small-molecule ligands. Thus, certain genes, notably genes for
aminoacyl-tRNA synthetases (see Chapter 14), are controlled by attenuation
mediated by a 200–300-nucleotide-long, untranslated, leader RNA
that directly and specifically interacts with the cognate, uncharged
tRNA for the synthetase; the charged form of the tRNA does not fit in
the binding pocket provided by the RNA secondary structure. Binding
of uncharged tRNA stabilizes the leader RNA in its antitermination
structure so that transcription into the adjacent synthetase gene can
proceed. Specificity is achieved in part by a “codon–anticodon” interaction
between the tRNA and the leader RNA. Because uncharged (but
not charged) tRNA can bind to the leader, transcriptional readthrough is
only stimulated when the cognate amino acid is in short supply and the
level of uncharged tRNA in the cell rises.
Although most prevalent in bacteria, riboswitches are found in
other organisms as well, including archaea, fungi, and plants. In some
cases in these higher organisms, riboswitches are even involved in
controlling alternative splicing (Chapter 13). Thus, for example, in a
recent case described in the fungus Neurospora crassa, three TPP aptamers
were identified, two of which inhibited, and the third stimulated,
expression of genes through regulation of RNA splicing.
+
1
1 2 3 4
2
3
4
3'
5'
SAM
SAM
SAM
C G
U A
A U
U A
U A
C G
U A
U A
U GA A GA A GGGGU U U
A
U
U
C
U
U
G C
A
A
G
U
U
C
CC
CG
CU
AU
A A
3'
5'
G
A
U
A
A
G
A
A
U
C
C
C
C
A
A
A
A
G
G
U
C
U
A
U A
U U
U A
C G
G
A
G
G
U
U
UG A A G C G U U
C
U
A
U
U
C
U
U
F I G U R E 18-4 Changes in secondary structure of a SAM-sensing riboswitch. The sequences
of regions 1–4 (described in Fig. 18-3) are here shown in detail and color-coded.
The base pairing found in the two alternative secondary structures—that is, with and without
SAM bound—are shown. (Adapted from Winkler W.C. et al. 2003. Nat. Struct. Biol.
10: 701–707, Fig. 5b. © Macmillan.)
PAGE
PRO
O
FS
638 Chapter 18
Attenuation mediated by alternative RNA secondary structures was
first discovered in studies of the tryptophan operon of Escherichia
coli, as described in Box 18-1, Amino Acid Biosynthetic Operons Are
Controlled by Attenuation. The trp operon contains genes responsible
for the biosynthesis of tryptophan. Their expression is controlled in
response to the level of tryptophan in the cell as measured by the
availability of charged tRNAtrp
. This example turned out to be one of
several similar cases—attenuation linked to amino acid availability is
a common control mechanism for amino acid biosynthetic operons.
Understanding how the trp operon was controlled was the first time
that alternative RNA secondary structures were shown to regulate
gene expression in any organism.
N
N
N
N
O
NH2
a
c
b
HO OH
N
N
N
N
O
HO
O
P–
O O
OH
NH2 O
P O
–
O
N
N
+
N
S
O
P O–
O
OH
O
–
P O
O
O
OH
OHOH
N
NH
N
N
OH
O
O–
O
H3N
+
N
+
H3
O
S
+
–
O
OHHO
CH3
O
+
NH3
N
N
N
N
NH2
N
H
N
N
N
H
NH2
O
N
N
N
N
H
NH2
O
O
CO–
O
O C
b
HN
O
O
O N N
NN
NH2
NH2
O
NH2
NH2
H2N
H2N
a
B12 riboswitch coenzyme B12
TPP riboswitch thiamine pyrophosphate FMN riboswitch flavin mononucleotide
SAM riboswitch S-adenosyl-methionine lysine riboswitch lysine
guanine riboswitch guanine adenine riboswitch adenine
P5
P6 P7 P8
P12
P11
P9
P10
P3
P5
P4
P2
P3a P3
P4P3
P2
P1
P3
P2 P4
P1
P2b P2a
P2 P2
P1 P1
P3P3
P2
P3 P4
P5
P1
P6
P5
P1
P2 P1
P4
5'
5'
5'
5' 5'
5' 5'
F I G U R E 18-5 Riboswitches respond to a range of metabolites. The secondary structure of
seven riboswitches and the metabolites they sense are shown here. (Adapted, with permission,
from Mandal M. et al. 2003. Cell 113: 577–586, Fig. 7A. © Elsevier.)
PAGE
PRO
O
FS
B O X 18-1 Amino Acid Biosynthetic Operons Are Controlled by Attenuation
In E. coli, the five contiguous trp genes encode enzymes that
synthesize the amino acid tryptophan. These genes are expressed
efficiently only when tryptophan is limiting (Box 18-1
Fig. 1). The genes are controlled by a repressor, just as the lac
genes are, although in this case, it is the absence of its ligand
(tryptophan) that relieves repression.
Even after RNA polymerase has initiated a trp mRNA molecule,
however, it does not always complete the full transcript.
As with riboswitches, the decision to make a complete transcript
is controlled by attenuation; in this case, most transcripts
are terminated prematurely, before they include even the first
trp gene (trpE). But attenuation is overcome if tryptophan levels
are low in the cell; when tryptophan is limiting, polymerase
does not terminate and instead transcribes all of the
trp genes. Whether or not attenuation occurs depends on the
ability of RNAs to form alternative secondary structures, just as
it did with the riboswitches. In this case, however, the choice
between alternative structures formed by the leader RNA is
not controlled by binding of ligand directly to that RNA; instead,
the choice of alternatives relies on the coupling of transcription
and translation in bacteria.
The sequence of the 5′ end of trp operon mRNA includes
a 161-nucleotide leader sequence upstream of the first codon
of trpE (Box 18-1 Fig. 2). Near the end of this leader sequence,
and before trpE, is a transcription terminator, composed
of a characteristic hairpin loop in the RNA (made from
sequences in regions 3 and 4 of Box 18-1 Fig. 2), followed by
eight uridine residues (see Fig. 12-12). Transcription usually
stops after this terminator (and, we might have thought,
should always stop), yielding a leader RNA 139 nucleotides
long. This is the RNA product seen in the presence of high
levels of tryptophan.
Three features of the leader sequence allow the terminator
to be passed by RNA polymerase when the cellular concentration
of tryptophan is low. First, there is a second hairpin
(besides the terminator hairpin) that can form between regions
1 and 2 of the leader (see Box 18-1 Fig. 2). Second, region
2 also is complementary to region 3; thus, yet another
hairpin consisting of regions 2 and 3 can form, and when it
does, it prevents the terminator hairpin (3, 4) from forming.
Third, the leader RNA contains an open reading frame encoding
a short “leader” peptide of 14 amino acids, and this
open reading frame is preceded by a strong ribosome-binding
site (see Box 18-1 Fig. 2).
The sequence encoding the leader peptide has a striking
feature: two tryptophan codons in a row. When tryptophan
is scarce, there is very little charged tryptophan tRNA available,
and the ribosome stalls when it reaches the two tryptophan
codons. Under these circumstances, RNA around the
tryptophan codons is within the ribosome and cannot be part
of a hairpin loop. The consequence of this is shown in Box
18-1 Figure 3.
A ribosome caught at the tryptophan codons (part b) masks
region 1, leaving region 2 free to pair with region 3; thus, the
terminator hairpin (formed by regions 3 and 4) cannot be
made, and transcription is not attenuated. If, on the other
hand, there is enough tryptophan (and therefore enough
charged Trp tRNA) for the ribosome to proceed through the
tryptophan codons, the ribosome blocks sequence 2 by the
time RNA containing regions 3 and 4 has been made. Thus,
the terminator forms, attenuating transcription, and the trp
genes are not transcribed.
The trp operon is controlled by repression and attenuation,
providing a two-stage response to progressively more stringent
tryptophan starvation. But attenuation alone can provide robust
regulation: other amino acid operons such as leu and his
rely entirely on attenuation for their control. In the case of the
leucine operon, its leader peptide has four adjacent leucine
codons, and the histidine operon leader peptide has seven
histidine codons in a row.
B O X 18-1 F I G U R E 1 The trp operon. The tryptophan operon of E. coli, showing the relationship of the leader (see text) to the structural
genes that code for the Trp enzymes. The gene products are anthranilate synthetase (product of trpE), phosphoribosyl anthranilate
transferase (trpD), phosphoribosyl anthranilate isomerase-indole glycerol phosphate synthetase (trpC), tryptophan synthetase β (trpB), and
tryptophan synthetase α (trpA).
promoter
leader trpBtrpCtrpDtrpE trpA
operator
trp mRNA
RNA
A D V A N C E D C O N C E P T S
Regulatory RNAs 639
PAGE
PRO
O
FS
B O X 18-1 (Continued)
UAA
A
AAGG
GUAUC
G A C A A A A A
A
A A A A A AA U
U U
U U U U UU U U U UUG G G GG GG G G G AAAGC
C C
C C C C C C C
C
GGGCA
G
GUAUUCCCCCCCCCCC
C
CC
U
A
A
A
U G G C G
AAAAAAAAAAA
UUUU
GGGGG
G U
U G
G G
GG G G G GCU U U U U U U U U U U UA A A A A A A A A A A A A A A A . . .
. . .
A AC C C C
leader peptide
Met Lys Ala Ile Phe Val Leu Lys Gly TrpTrp TrpTrpTrp Trp Arg
1
3
4
139
2
Met Gln Thr
Thr Ser (stop)
161
end of leader
(site of attenuation)
trpE polypeptide
B O X 18-1 F I G U R E 2 trp operator leader RNA. Features of the nucleotide sequence of the trp operon leader RNA.
B O X 18-1 F I G U R E 3 Transcription termination at the trp attenuator. The figure shows how transcription termination at the trp
operon attenuator is controlled by the availability of tryptophan. The blue box shows the leader peptide coding region. (a) Conditions of high
tryptophan: sequence 3 can pair with sequence 4 to form the transcription termination hairpin. (b) Conditions of low tryptophan: the ribosome
stalls at adjacent tryptophan codons, leaving sequence 2 free to pair with sequence 3, thereby preventing formation of the 3-4 termination
hairpin. (c) No protein synthesis: if no ribosome begins translation of the leader peptide AUG, the hairpin forms by pairing of sequences 1
and 2, preventing formation of the 2-3 hairpin, and allowing formation of the hairpin at sequences 3-4. The Trp enzymes are not expressed.
a high tryptophan
b low tryptophan
c no protein synthesis
1 2
1
2 3
4
1 2
3 4
3 4
leader peptide
coding region
leader peptide
tryptophan codons
trp operon mRNA
PAGE
PRO
O
FS
RNA INTERFERENCE IS A MAJOR REGULATORY
MECHANISM IN EUKARYOTES
We have seen how RNA molecules can regulate expression of genes
in prokaryotes. We have (in Chapter 17) already seen examples of regulatory
RNA elements in the transcripts of some eukaryotic genes,
but these function through binding regulatory proteins (the HIV TAT
protein was one such case). It is, however, now apparent that RNAs
have a more widespread and unanticipated role in gene regulation in
eukaryotes. These new modes of regulation are described in this sec-
tion.
Several types of very short RNAs repress—or silence—expression
of genes with homology to those short RNAs. This silencing, called
RNA interference (RNAi), manifests in different ways: sometimes by
inhibiting translation of the mRNA, in other cases through destruction
of the mRNA, and in yet others by transcriptional silencing of the promoter
that directs expression of that mRNA. As we describe below,
these short RNAs are generated by special enzymes from longer double-stranded
RNAs (dsRNAs) of various origins.
It is still unclear how widespread the action of regulatory RNAs
will turn out to be, and the detailed mechanism used to silence the
target genes in any given case is still emerging. But it is already clear
that the roles of these RNAs range from developmental regulation
(well studied in the worm Caenorhabditis elegans and the plant Arabidopsis;
see Chapter 22) to mechanisms that protect organisms
against viral infection. Furthermore, RNAi has been adapted for use
in several organisms as a powerful experimental tool, providing an
easy means of “turning off” expression of any specific gene.
Short RNAs That Silence Genes Are Produced from a Variety of
Sources and Direct the Silencing of Genes in Three Different Ways
Before describing aspects of the production and function of these
short silencing RNAs in more detail, we first provide an overview of
how this type of silencing works (illustrated in Fig. 18-6).
The small RNAs have different names depending on their origin.
Those made artificially or produced in vivo from dsRNA precursors are
typically called small interfering RNAs (siRNAs). Another group of
regulatory RNAs are the microRNAs (miRNAs). The latter are derived
from precursor RNAs which are encoded by genes expressed in cells
where those miRNAs have specific regulatory functions.
Both siRNAs and miRNAs are generated from longer RNA molecules
by the enzyme Dicer, an RNase-III-like enzyme (Fig. 5-24) that
recognizes and digests longer dsRNA or the stem-loop structures
formed by miRNA precursors (see below). The siRNA and miRNA
products are typically 21–23 nucleotides long; their production is
shown as the first step in Figure 18-6.
These small RNAs inhibit expression of homologous target genes in
three ways: they trigger destruction of the mRNA encoded by the target
gene, they inhibit translation of the mRNA, or they induce chromatin
modifications within the target gene and thereby silence its
transcription. Remarkably, whichever route is used in any given case,
much of the same machinery is required. This machinery includes a
complex called the RNA-induced silencing complex (RISC). A RISC
Regulatory RNAs 641
PAGE
PRO
O
FS
contains, in addition to the siRNA or miRNA, various proteins including
a member of the Argonaute family.
The siRNA or miRNA must be denatured to give a guide RNA—the
strand that gives the RISC specificity, as we shall see—and a passenger
RNA, which usually gets discarded. The resulting complex, the mature
RISC, is then directed to target RNAs containing sequences complementary
to the guide RNA. These target RNAs are degraded or their translation
is inhibited. Typically, the choice depends in part on how closely
the guide RNA matches the target mRNA: if the sequences are highly
complementary, the target is degraded; if the match is not as good (i.e.,
if there are several base-pairing mismatches), the response is more often
inhibition of translation. In those cases where the target RNA is degraded,
Argonaute is the catalytic subunit that carries out the initial
mRNA cleavage; for this reason Argonaute is often called “Slicer” and
mRNA cleavage is called “slicing.”
A RISC can also be directed into the nucleus where it recruits other
proteins that modify the chromatin around the promoter of the gene
complementary to the guide RNA (shown on the left of Fig. 18-6). This
642 Chapter 18
cell membrane
chromatin
remodeling
RISC*
amplification
degradation
translational
inhibition
RISC
nucleus
pre-
miRNA
cytoplasmDicer
dsRNA
siRNAs and
miRNAs
F I G U R E 18-6 RNAi silencing. RNAi
switches off the expression of a gene when
dsRNA molecules that have homology to that
gene are introduced, or made, in the cell. This
effect involves processing of the dsRNA to
make siRNAs and miRNAs by the enzyme
Dicer. Another enzyme involved only in the
case of miRNAs—Drosha—is not shown here,
but is described later. The siRNAs and miRNAs
direct a complex called RISC (RNA-induced silencing
complex) to repress genes in three
ways. It attacks and digests mRNA that has homology
with the siRNA; it interferes with translation
of those mRNAs; or it directs chromatinmodifying
enzymes to the promoters that
direct expression of those mRNAs. (Adapted,
with permission, from Hannon G.J. 2002. Nature
418: 244–251, Fig. 5. © Macmillan.)
PAGE
PRO
O
FS
Regulatory RNAs 643
modification leads to silencing of transcription (Chapter 17). Establishing
silencing in the centromeric regions of the yeast Schizosaccharomyces
pombe, for example, requires the RNAi machinery. In this
case, it is believed that regions of the centromere (see Chapter 7) are
transcribed to produce RNAs that hybridize with other RNAs from the
same region. The resulting dsRNAs are recognized by Dicer and
cleaved to produce the siRNAs responsible for directing the RNAi machinery
to the centromeres.
Another feature of RNAi silencing is worth noting—its extreme efficiency.
Thus, very small amounts of dsRNA are often enough to induce
a near complete shutdown of target gene expression. A factor adding to
efficiency could be the action of an RNA-dependent RNA polymerase
(RdRP), an additional enzyme required in many cases of RNAi including
centromeric silencing in fission yeast. This polymerase can amplify
the inhibitory signal: the RdRP generates dsRNA after recruitment to
the mRNA by the original siRNA (shown on the right of Fig. 18-6). This
feedback process generates large amounts of siRNA. RdRP has not yet
been identified in mammalian cells, but high efficiency likely still results
from the fact that slicing is catalytic—that is, each RISC can cleave
several mRNAs.
Thus, although in the first section of this chapter we saw examples
of small RNAs regulating gene expression in trans in bacteria, the
mechanism of both production and action of such RNAs in eukaryotes
is very different.
SYNTHESIS AND FUNCTION OF MIRNA MOLECULES
miRNAs Have a Characteristic Structure That Assists in
Identifying Them and Their Target Genes
As we have seen, miRNAs are one class of regulatory RNA that silences
gene expression through the RNAi pathway. Indeed, in terms of
their production and function, miRNAs are perhaps the best understood
class. As mentioned above, miRNAs are encoded in the genome
as segments of longer transcripts. Their characteristic structure helps
identify them and predict the target genes they might regulate.
The functional form of an miRNA is typically about 21 or 22 nucleotides
(it can vary from 19 to 25 nucleotides). These short RNAs are
generated by two RNA cleavage reactions from a longer RNA transcript
(called a pri-miRNA, for primary) that carries a hairpin-shaped secondary
structure. The first cleavage liberates the stem loop, called the
pre-miRNA; the second generates the mature miRNA from the premiRNA.
One of the first identified, and best-characterized, miRNAs is
let-7, which regulates development at the larval-to-adult transition in
the worm C. elegans (see Chapter 22). The structure of the pre-miRNAs
for let-7 and some other naturally occurring miRNAs are shown in Figure
18-7.
It was thought initially that one “arm” of pre-miRNA stem-loop structure
would be the regulatory miRNA. Recently, however, numerous examples
have been identified where both “arms” of the structure give rise
to functional miRNAs, each with its own set of target genes (in these
cases the two miRNAs are labeled red and blue in Fig. 18-7). It now appears
that having miRNAs produced from both arms is common. The
pre-miRNAs can be encoded by any part of a transcript: that is, they
PAGE
PRO
O
FS
might fall within coding regions, within leader regions, or within introns
(Fig. 18-8).
The distinctive secondary structure of a primary transcript carrying
an miRNA (pri-miRNA) has made it possible to predict their presence
based on the calculated secondary-structure fold of the RNA sequence.
Furthermore, in many cases, candidates for the regulated target genes
can also be predicted, because silencing depends on sequence complementarity
between the target and the mature miRNA. The base pairing
between miRNA and target RNA is initiated by interactions of so-called
“seed residues”—typically the sequence between bases 2 and 9 of the
22-nucleotide miRNA. This is the region of highest complementarity,
and so it is the region most useful in identifying candidate target genes.
Of course, establishing that an miRNA really exists requires that its
644 Chapter 18
lin-4RNA
let-7RNA
miR-1*
miR-1
miR-34
miR-34*
miR-124
3'5' 3'5'
3'5' 3'5'
3'5'
G
A
G
U
G
U
G
A
C
U
C
A
G
A
G
C
C
C
G
U
C
C
A
U
G
U
G
C
U
U
C
A
C
A
C
U
G
G
G
U
C
U
C
G
G
G
C
A
G
G
U
A
C
U
C
U
C
G
A
A
U
U
C
C
U
C
A
U
A
U
C
A
U
A
C
C
U
A
C
A
U
U
C
U
U
C
A
U
A
C
G
U
C
G
C
C
G
C
C
A
A
U
G
G
U
A
U
G
G
A
U
G
U
A
A
G
A
A
G
U
A
U
G
U
A
G
C
G
G
U
G
G
U
C
G
C
A
A
U A
A A
C
A
U
A
A
C
G
U
U
G
U
C
G
A
U
G
G
G
U
G
A
C
G
G
G
G
C
U
G
U
A
C
A
G
G
U
G
C
A
A
C
G
G
C
U
A
C
C
C
A
C
U
G
C
C
C
C
G
A
C
A
U
G
U
C
C
G
U
U
A
A
A
G
C
U
AC
U
A
U
U
C
C
A
G
G
U
U
U
G
A
U
A
U
G
U
U
G
G
A
G
A
U
G
G
A
G
U
A
A
C
C
U
C
C
G
A
A
C
U
A
U
G
C
A
A
U
U
U
C
U
A
C
C
U
U
A
C
C
G
G
A
U
U
A
A
C
C
A
A A
U
G
G
U
U U
A
U
C
U
U
U U
C
C
A
G
G
U
G
A
U
U
U
C
G
U
G
C
C
A
C
U
A
C
G
G
U
U
A
C
U
G
C
A
C
C
C
G
U
C
U
A
C
U
A
A
G
G
C
A
C
G
G
U
G
A
U
G
C
C
A
G
U
G
G
C
G
U
G
G
G
A
A
C
G
UC
CA
G
C
A
C
U
U
C
C
U
A
A
C U
U
A
A
A
G A
G C
U
U
A
A
G
A
G
G
A
U
F I G U R E 18-7 Structure of some pre-miRNAs prior to processing to generate the mature
miRNAs. The sequences in red are miRNAs. In some cases, both “arms” of a stem
loop can generate a functional miRNA. In such cases, the second miRNA is shown in
blue—for example, miR-1 (red) and mi-R1* (blue), as well as with miR-34 (red) and
miR34* (blue). The miRNAs shown are all from the worm. lin-4 and let-7 were identified
genetically; those called miR are found by bioinformatics. (Modified, with permission, from
Lim L.P. et al. 2003. Genes Dev. 17: 991, Fig. 6. © Cold Spring Harbor Laboratory Press.)
PAGE
PRO
O
FS
presence be detected in cells (e.g., by northern blotting) and that gene
expression from target mRNA be affected by its presence.
The two cleavage reactions required to generate the miRNA from
these primary transcripts are mediated by two distinct RNases. One is
Dicer, which we have already introduced, and is required for almost all
cases of RNAi. The other, specifically required for miRNA processing, is
Drosha. A characteristic of both these enzymes is that they recognize
and cleave RNAs on the basis of the structure of their substrates rather
than their specific sequence. We now turn our attention to how these
enzymes work.
An Active miRNA Is Generated through a Two-Step
Nucleolytic Processing
Two specialized RNA-cleaving enzymes are required to process the
initial pri-miRNA transcript, containing the stem-loop structure, into
the mature miRNA. The first enzyme is Drosha, a member of the
RNase III family of enzymes. Drosha makes two cleavages that cut
the stem-loop region of the RNA (pre-miRNA) out of the primary
transcript RNA (pri-miRNA). This enzyme works together with an
essential specificity subunit protein (called Pasha in some organisms
and DGCR8 in others), and together these two proteins form an active
microprocessor complex. The pre-miRNA generated by Drosha
Regulatory RNAs 645
start
codon
stop
codon
5'
pre-miRNA in coding region
poly-A
(AAUAAA)
cap AAAAn
AAAAn
AAAAn
AAAAn
5' cap
5'
pre-miRNA in a noncoding region
cap
5'
pre-miRNA in intron of noncoding RNA
cap
AAAAn5'
pre-miRNA in intron of a protein-coding pre-mRNA
cap
F I G U R E 18-8 miRNAs are coded in both introns and exons in RNA. Intronic sequences
are shown in light green.
PAGE
PRO
O
FSis usually approximately 65–70 nucleotides long. Drosha resides in
the cell’s nucleus, and the Drosha-catalyzed cleavage event occurs in
this cellular compartment.
The base-paired stem in the pri-miRNA is typically about 33 bp in
length (three helical turns of dsRNA) and contains only a few mismatches
(Fig. 18-9). At the “top” of the stem is a loop of variable size
(usually relativity large, about 10 nucleotides); the sequence of this
loop region is not critical for the processing reactions. Importantly, for
processing by Drosha, single-stranded RNA (ssRNA), lacking significant
secondary structure, is needed flanking each side (5′ and 3′) of
the stem-loop. It is the ssRNA–dsRNA junctions that are in large part
responsible for determining the cleavage specificity of Drosha.
The stem region can be divided into two functional segments: an
approximately 11-bp lower stem and an approximately 22-bp upper
stem (Fig. 18-9). Drosha cleaves 11 bp away from the dsRNA–ssRNA
junctions—that is, between the lower and upper stems in the primiRNA
(Fig. 18-10). The two cleavages thus generate the approximately
65-nucleotide pre-miRNA composed of the 22 bp (two helical
turns) of dsRNA and the top loop. The RNase III family enzymes are
specific for dsRNA and cleave it in a manner that leaves a 2 nucleotide
overhang on the 3′ ends of the dsRNA product. This 3′ overhang
is important for recognition of that RNA molecule by the next
enzyme in the pathway, Dicer.
Dicer Is the Second RNA-Cleaving Enzyme Involved in
miRNA Production
The pre-miRNA liberated by Drosha is exported to the cytoplasm
where the second RNA cleavage reaction, carried out by Dicer, takes
place. As with Drosha, Dicer selects its cleavage sites using a measuring,
rather than sequence-specific, mechanism. A high-resolution
structure of Dicer provides insight into how this likely occurs.
Dicer is constructed of three modules (Fig. 5-23): two RNase III domains
and a dsRNA-binding domain called the PAZ domain (named for
three proteins that contain this domain: Piwi, Argonaute, and Zwille).
Figure 18-11a shows in cartoon form the organization of the Dicer
protein and how it is believed to interact with a dsRNA molecule. In
the right panel, Figure 18-11b, is the structure of Dicer, modeled with
a substrate RNA. The overall structure of the protein is shaped like a
hatchet. The PAZ domain is at the bottom of the handle, where it
forms a binding pocket for the 3′ end of the dsRNA substrate. The
646 Chapter 18
pri-miRNA
substrate
recognition step
DroshaDGCR8
~11 bp
~11 bp
~11 bp
~22 bp
~22 bp
catalytic reaction
step
products
F1 F2 (pre-miRNA)
F3
processing center
F I G U R E 18-10 Recognition and cleavage
of pri-miRNA by the microprocessor complex.
Three fragments are generated by cleavage,
labeled F1, F2 (the pre-miRNA), and F3.
basal
segments
lower stem
(~11 bp)
upper stem
(~22 bp)
terminal
loop
cleavage
by Drosha
mature miRNA
+1
–2
cleavage
by Dicer
5' N N N N N
N N N N N N N N N N N N N N N N N N N N N N N N N N N N
N
N
N N N N
N
N N
N N N N N N N N
3' N N N N N
N N N N N N N N N N N N N N N N N N N N N N N N N N N N
N N N
N N
N N N N N N N N
F I G U R E 18-9 Overview of the structure of pri-RNA showing Dicer and Drosha
cleavage sites. The region in red becomes the mature miRNA. Note that the basal segments
must be single-stranded for proper recognition by the Drosha complex.
PAGE
PRO
O
FS
handle of the hatchet is formed by a linker domain and contains a
positively charged binding surface for the RNA molecule. The top
“blade” region comprises the two RNase domains, arranged in a symmetrical
dimer. Each RNase domain carries an active site and is responsible
for cleaving one of the two strands of the substrate RNA.
Thus, Dicer will act on any dsRNA, regardless of sequence, and will
cleave this molecule 22 nucleotides from its end. The PAZ domain anchors
the 3′ terminus of the substrate RNA to position the active sites
of the enzyme approximately 22 nucleotides away in a ruler-like fashion
(see Fig. 18-11). Indeed, the occurrence of differently sized PAZ
domains correlates with the different sizes of Dicer products found in
different organisms.
Incorporation of a Guide Strand RNA into RISC Makes the Mature
Complex That Is Ready to Silence Gene Expression
The actions of Drosha and then Dicer generate the 21–25-nucleotide
RNA molecule that will guide regulation of gene expression. The active
form of the regulatory miRNA is the single-strand form—at this stage
called the “guide” RNA—incorporated into a RISC protein complex.
Within this complex, the RNA guide strand recruits RISC to a target
RNA. It has been argued that the length of approximately 22 nucleotides
is just long enough to uniquely specify a single target gene in
the large genomes of complex eukaryotes using RNA–RNA base pairing.
Regulatory RNAs 647
3'
blade
Dicer model
handle
F I G U R E 18-11 Dicer structure and organization. (a) The scheme shows Dicer organization.
(b) Dicer structure modeled with dsRNA reveals how length is measured. The protein
is shown in gray, with nuclease active sites indicated by the red spheres (and as black
dots in part a). The RNA is in green. The structure shown contains only the RNase III and
PAZ domains. The Dicer protein also contains ATPase and other domains. (b, MacRae I.J.
et al. 2006. Science 311: 195–198. PDB Code: 2FFL; note that the RNA was modeled into
the structure and was not part of the crystal structure.) Image prepared with MolScript,
BobScript, and Raster 3-D.
a b
PAZ domain
3′ end of RNA
PAGE
PRO
O
FS
RISC is a multicomponent complex that includes the guide
miRNA. The guide RNA base-pairs with the target mRNA, and, as a
consequence, promotes silencing of gene expression. The central component
of this regulatory complex is a protein called Argonaute,
which is in many cases an RNA-cleaving enzyme. The best understood
mechanism of gene silencing is RISC-mediated cleavage—or
slicing—of the target mRNA. However, many organisms have multiple
members of the Argonaute protein family. For example, there are eight
distinct Argonautes in humans, but not all of these Argonautes, when
incorporated into a RISC complex, have slicer activity. RISCs containing
other Argonautes must silence gene expression using non-slicerdependent
mechanisms, such as repression of translation.
Generation of the active RISC and slicing occur as follows. The short
dsRNA generated by Dicer is incorporated into RISC. The dsRNA is denatured
to provide the guide strand and the passenger strand. According
to one model, the passenger strand is cleaved by Argonaute and
then ejected from the complex; according to another, the passenger
strand is removed without first being cleaved, likely by an RNA helicase.
The resulting RISC—called mature RISC—with its single-stranded
guide RNA is now ready to recognize and slice the target mRNA.
As we saw with Dicer, the structure of an Argonaute protein provides a
framework for understanding the mechanism of target RNA recognition
and cleavage by RISC (Fig. 18-12; see also Structural Tutorial 18-1). Like
Dicer, Argonaute has both a PAZ domain and an RNase domain. The PAZ
domain specifically recognizes the 3′ end of the guide RNA. The bound
guide RNA is base-paired to the target RNA, and the architecture of the
complex is such that this binding positions the active site of the RNase
648 Chapter 18
F I G U R E 18-12 Argonaute structure, showing RNA-binding regions and an RNase
H–like nuclease domain. (a) Crystal structure of Argonaute. The domains are colored as in
part b, with the blue domain being the amino-terminal part of the protein, and the green
domain in the middle. (b) Cartoon of the Argonaute domains. The arrow shows the RNase
active site positioned to cleave in the middle of the paired region between small RNA and
miRNA. (a, Song J.J. et al. 2004. Science 305: 1434–1437, Fig. 4C. PDB Code: Iu04. Image
prepared with MolScript, BobScript, and Raster 3-D. b, Adapted, with permission,
from Song J.J. et al. 2004. Science 305: 1434–1437, Fig. 4C. © AAAS.)
PAZ domain
recognizes 3′
end of RNA
RNase active site positioned to cleave in the
middle of the paired region between small
RNA and mRNA
a b
PAGE
PRO
O
FS
Regulatory RNAs 649
domain appropriately to cleave the target RNA strand. Cleavage occurs
nearly in the middle of the guide RNA–target RNA duplex, between the
10th and 11th nucleotide from the 5′ end of the guide RNA.
As we have already mentioned, in some cases, mature RISC can inhibit
translation rather than slicing mRNA. The mechanism of this
translational repression is still under scrutiny, but it appears that
miRNAs lead, in some cases, to the sequestration of mRNA in socalled
processing bodies (P-bodies) within the cytoplasm where translation
is repressed. Binding of miRNA can also destabilize the poly-A
tail (see Chapter 12) of the target mRNA, disrupting translation initiation
(Chapter 14).
siRNAs Are Regulatory RNAs Generated from Long
Double-Stranded RNAs
As we have seen, the critical steps in synthesis of an active miRNA are
Drosha cleavage, followed by Dicer cleavage, followed by incorporation
of the guide RNA into the RISC complex. But only miRNAs are
made from large hairpin precursors. In contrast, the precursor RNA for
the siRNA pathway is a longer dsRNA. As a consequence of this different
initial substrate, Drosha is not needed for the generation of siRNAs.
Cleavage by Dicer is still required, however, and again generates a suitable
21–22-nucleotide RNA for incorporation into RISC. In plants,
even miRNAs are generated by Dicer alone; it is not clear how they
manage to forgo prior action of Drosha.
A fascinating series of observations and experiments led to our current
understanding of small regulatory RNAs in eukaryotes. These began
in the late 1980s with seemingly mystifying results of attempts to
overexpress pigment genes in petunias (to make them a deeper purple,
but ending up with the flowers turning white). Next was the surprising
discovery of regulatory genes from worms whose products turned
out to be miRNAs, and then experiments showing that introducing
dsRNAs into worms silenced complementary gene expression. This
story is described in Box 18-2, History of miRNAs and RNAi.
Small RNAs Can Transcriptionally Silence Genes by Directing
Chromatin Modification
We have now seen how miRNAs and siRNAs can silence genes by inhibiting
the translation of target mRNAs or directing their destruction.
Regulatory RNAs can also act at the level of transcription, switching
off expression of target genes by directing histone modification of the
promoter. This mechanism has been most extensively studied in centromeric
silencing in the fission yeast S. pombe.
We noted in Chapter 17 that in yeast, genes placed in certain regions
of the genome are typically silenced. In the case we described in
detail in that chapter, genes placed near the telomeres in Saccharomyces
cerevisiae were silenced. Genes in the mating-type locus of
that yeast and of S. pombe are also silenced. In S. pombe, the centromeres
are another silenced region of the genome. In both organisms,
silencing involves histone modifications. But unlike cases of silencing
in S. cerevisiae, which lacks the RNAi machinery, centromeric
silencing in S. pombe requires that pathway.
The centromeres of S. pombe have a sequence organization more
like that of higher eukaryotes (e.g., flies and humans) than that of S.
cerevisiae (see Fig. 7-8). Each centromere has a central region, of
PAGE
PRO
O
FS
B O X 18-2 History of miRNAs and RNAi
K E Y E X P E R I M E N T S
In 1989, Richard Jorgensen, working at the biotech company
Advanced Genetic Sciences in Oakland, California, was trying
to make petunia plants with more deeply purple flowers than
existing strains. The strategy seemed straightfoward: he would
introduce into the plants an additional copy of the pigment
gene (encoding chalcone synthase) under the control of a
strong promoter. These plants would make more chalcone
synthase and the flowers would be more purple. What he actually
got were plants with varying degress of paler flowers,
many that were sectored—with purple and white regions—
and even some that were completely white (Box 18-2 Fig. 1).
Although disappointing, these results were intriguing. In attempting
to understand what was going on, Jorgensen uncovered
various features of the phenomenon, called cosupression
(because expression of both the transgene and the endogenous
gene are repressed). The greater the expression of the transgene,
the lower the level of chalcone synthase; this was true
whether increased expression resulted from multiple copies of
the transgene or from use of stronger promoters driving the
transgene. It was also noted that some plants had variegated
patterns of pigmentation, and that different variegation patterns
could be found in different flowers on the same plant. These
patterns were sometimes inherited, but on other occasions apparently
altered at random. These observations suggested to
Jorgensen and others (particularly Marjori Matzke, who also
was investigating this phenomenon) that they were dealing with
an epigenetic phenomenon.
Other investigators were trying to make plants resistant to viral
infection. One approach was to overexpress in plants a dominant-negative
derivative of a common viral replication factor:
this protein was expected to block replication of any infecting
virus that used this common replicative mechanism. Although
the dominant-negative viral product blocked replication of the
potato virus from which it was derived, its specificity of action
was surprisingly tightly restricted to that virus. It was also shown
that the protein itself was not even needed—just the RNA.
Meanwhile other researchers were using antisense RNA to
knock down expression of the par-1 gene in worms. Their intention
was to prove that this gene was responsible for a particular
developmental phenotype. Antisense RNA produced in
vitro and injected into the developing worm induced the phenotype
predicted for the loss of par-1 expression. But it was
found that sense RNA had the same effect. This was only included
in the experiment as a negative control of course; it was
not expected to have any effect on expression. RNAs unrelated
to the par-1 gene had no effect.
The explanation for this RNA-dependent gene repression was
provided by Andrew Fire and Craig Mello in experiments that
earned them the 2006 Nobel Prize in Physiology or Medicine.
They demonstrated that it was in fact neither sense nor antisense
RNA that silenced the gene—it was the dsRNA produced by a
mix of the two. It turned out that the RNA preparations of sense
or antisense were both contaminated with small amounts of the
opposite strand, and it was the resulting double-strand populaB
O X 18-2 F I G U R E 1 Petunia flower. An example of the effects
of overexpressing the pigment gene chalcone synthase in
what would otherwise be a completely purple petunia flower.
(Courtesy of Richard A. Jorgensen, University of Arizona.)
tion that caused silencing. When dsRNAs were deliberately prepared,
they were shown to be very potent in eliminating expression
of the target gene. Hence, the phenomenon of RNAi had
been discovered, a finding published in 1998.
Mechanistic insights came thick and fast from several labs.
First, dsRNAs were shown to trigger degradation of homologous
mRNAs in extracts from Drosophila cells, an assay that
led to the identification of RISC. The identifcation of siRNAs—
the species that directs RISC to the target genes—was reported
in plants in 1999. Dicer, the nuclease that creates them, was
described in 2001. And the final major component of the
pathway, Slicer, was identified in 2005 when the crystal structure
of Argonaute revealed the protein to be an RNase.
In addition to showing that Dicer was needed to generate
the siRNAs, it was shown to also be needed for miRNA genes to
function during worm development. The first miRNA and its target
had been described in 1993, by Victor Ambros and Gary
Ruvkun, respectively. At the time, this observation was seen as a
neat but eccentric oddity; the lin-4 gene encoded a small RNA
that acted on a target gene, lin-14, by virtue of sequence complementarity
between the miRNA and regions of the 5′ UTR of
the target genes (Box 18-2 Fig. 2). Later, other miRNAs were
found in worms, some with homology to similar genes in animals
and plants, suggesting that this mechanism of regulation
was more widespread. Everything was tied together when it became
clear that siRNAs and miRNAs work through the same
pathways. Thus, the picture emerged of a world of tiny RNAs
involved in gene regulation—some exogenously supplied, others
inbuilt as part of gene regulatory programs employed during
development. The field developed very rapidly, as the dates in
this account reveal, moving from obscure phenomenology to a
Nobel Prize and demanding of its own chapter in textbooks, in
just 15 years. The accelerated progress was perhaps largely a
consequence of the range of species (yeast, plant, and worm)
studied and approaches (genetics, biochemistry, structural studies,
and bioinformatics) employed.
PAGE
PRO
O
FS
B O X 18-2 (Continued)
C. elegans lin-4
pre-miRNA
lin-4 miRNA
binding sites
a
lin-14 3'-UTR An
1 2 3 4 5 6 7
b
lin-14
lin-4
c
1
processed miRNA
G U U
C C C
U
A U
G G G
C C U C A
2
U G U G A G U
3
G U G AG A G U G A G U C C U U G
C
G A G A
C
C U C U
C
C U C A
A
G G G U
C
C U
G U U
A
A
U G C U C U C A G G A A C
A
UC
CU
A
U
G A G U C C U U G
G A C
U C
G C
A UA C
C U C G C A U U U
4
C U C A G G G A A C
G A G U G U G A G A G U C C C U U G
C A
A
CU C
U
U U A
U
G A G U G G A G U C C U U G
U A
G C
A C
AC U
U C
G A G G U G G A G U C C C U U G
U
C
U A A
A C
C
U
C
U U C U A C C U C A G G G A A C
A C A U U C A A A A C U C A G G A A U
U G A G U G U G A A G U C C U U G
A G
C A
U
C
C
G C U U A U G U U A A A A U C A G G A A C
U
U G A G U G U G A G U C C C U GA
A C
G
A C
CU
A C U C A C A A C C A A C U C A G G G A C
U C U C A U U G A A C U C A G G A A UU C U C A U
5
6
7
A A
C C
B O X 18-2 F I G U R E 2 microRNA lin-4 binds within the 3′-untranslated region (UTR) of its target gene lin-14. (a) The lin-4 premiRNA
before processing by Dicer. The sequence of the miRNA is shown in blue. (b) The seven sequences within the lin-4 3′-UTR that
can base-pair with the lin-4 miRNA to various extents, as shown in part c. (Modified from Ha I. 1996. Genes Dev. 10: 3041–3050, Fig.
1. © Cold Spring Harbor Laboratory Press.)
PAGE
PRO
O
FS
largely unique sequence, flanked by a series of repeats common to all
centromeres. The repeats are important to function and contribute to
the formation of heterochromatin and the transcriptional silencing associated
with the region, as we shall see. Histones within the heterochromatin
carry repressing markers: low levels of acetylation, and
methylation on lysine 9 of histone H3 tail (H3K9) (see Table 7-7).
S. pombe has only a single gene for each of the major components of
the RNAi pathway—Dicer and Argonaute. Higher organisms have multiple
Dicer and Argonaute genes with partially redundant functions, making
genetic manipulation of the pathway more difficult. In addition, unlike
the situation in flies and worms, loss of the RNAi pathway is not
lethal to S. pombe, although it does make the cells grow poorly by, for
example, disrupting chromosome segregation. It was a surprise, however,
to discover that loss of any component of the RNAi pathway led to
loss of histone H3K9 methylation and therefor also loss of gene silencing
at the centromeres, particularly as this silencing was known to be
transcriptional; until this discovery, RNAi had been thought to act only
posttranscriptionally.
The key to understanding this transcriptional silencing seems to be
the centromeric repeats themselves; these sequence elements are transcribed
from both strands by RNA polymerase II, producing complementary
transcripts that can hybridize to form dsRNAs. The RNAs are
in turn acted on by the RNAi machinery to generate siRNAs that somehow—and
quite how remains unclear—direct an Argonaute-containing
RISC-like complex (called RNA-induced transcriptional silencing
[RITS] complex) to the centromeres. The siRNAs could in theory do this
by recognizing DNA at the centromeres, through sequence-specific base
pairing directly with the DNA template. But more likely are models in
which the siRNAs recruit RITS to transcripts tethered to the centromere
by RNA polymerase II. Recruitment results in slicing of centromic transcripts,
which is, in turn, required for spreading of the histone modification
apparatus along the centromere (Fig. 18-13). Thus transcription itself
may itself spread silencing, when transcripts are targeted by RNAi.
As we mentioned above, the mating-type loci of S. pombe are also
transcriptionally silenced, and here the silencing is not lost in mutant
strains defective for RNAi. It is believed that RNAi acts in this case as
well, but only in initially establishing the silenced state—it is not required
for maintaining silencing once it is established; other, proteinbased
mechanisms sustain the repressed state—just as they do in S.
cerevisiae (Chapter 17). RNAi is also believed to play a part in heterochromatin
silencing in other organisms, ranging from flies to plants.
Silencing of unwanted transcription from transposons also appears to
be RNA-mediated, as we describe below.
THE EVOLUTION AND EXPLOITATION OF RNAI
Did RNAi Evolve As an Immune System?
The RNAi machinery is widespread in eukaryotes, although not ubiquitous.
It does not occur in S. cerevisiae, for example, as we just
noted. It is believed, however, that at least the basic system existed in
the most recent common ancestor to all eukaryotes, but was subsequently
lost in some lineages.
But what does RNAi do, biologically? There are miRNAs of course—
and the RNAi machinery is required to produce and use those regula-
652 Chapter 18
PAGE
PRO
O
FS
Regulatory RNAs 653
dsRNA
Dicer
Clr4
Swi6
spreading
Chpl
RITS
Argonaute
centromere
repeat Me
Ac
siRNAs
F I G U R E 18-13 A model for RITS recruitment
and the silencing of centromeres.
Shown at the top are nucleosomes around the
repeat sequences at a centromere in S.
pombe. The repeat sequences are transcribed
from both strands by RNA polymerase II, generating
dsRNA that is a substrate for Dicer.
The siRNAs thereby produced are loaded into
the Argonaute-containing complex RITS. As
shown in the middle, the siRNA-containing
RITS is then recruited to the PolII-tethered
transcripts being generated by continued transcription
of the centromeric repeats, through
complementarity between the siRNA and the
transcript. This complex then recruits factors
(Clr4 and Swi6) that locally modify nucleosomes
by adding the H3K9 silencing markers.
Another subunit of RITS, Chp1, contains a
chromodomain (Fig. 7-40), which, by binding
to the methylated nucleosomes, likely stabilizes
the binding of RITS. Although not shown
in the figure, “slicing” of the transcripts by
Argonaute (within RITS) generates substrate
RNAs for the RdRP, which synthesizes a complementary
strand and thus generates further
substrate for Dicer. This process is required for
the nucleosome modification—and thus the
region silenced—to spread.
tors—but some organisms have the RNAi machinery and no miRNAs
(including S. pombe). It is, in fact, believed that miRNAs evolved to
take advantage of the existence of the RNAi machinery rather than being
the reason that machinery exists. One ancient function the RNAi
machinery might have served (and still serves) is protecting organisms
from transposons and viruses.
Transposons are found in all eukaryotes and, in some cases, make up a
substantial amount of a genome. In humans, for example, about 45% of
our genome is made of sequences that were once transposons. Transposons
are often transcriptionally silent and packaged into heterochromatin.
In some RNAi mutants, however, the histone modifications associated
with transposon silencing are lost. In addition, in plants and worms,
several siRNAs have been identified that correspond to transposons. And
in some cases in both these organisms, the loss of RNAi reactivates transposons,
causes them to jump, and leads to high levels of spontaneous
PAGE
PRO
O
FS
mutagenesis. Not as many transposons are reactivated as are known to
generate siRNAs, however. This might reflect a situation similar to that
described above for the mating-type silencing in S. pombe: RNAi might
be essential for initiating silencing of some transposons, but the silencing
then becomes self-sustaining without further need for the siRNAs.
The protective effect of RNAi on viral infection of plants has been
widely observed. Indeed, the effects were recognized long before
RNAi was known to be an underlying mechanism. When one leaf on a
plant is infected by a virus, a factor able to silence replication of the
virus is spread systematically throughout the whole plant. This factor
does not protect that originally infected leaf, but it does stop the infection
from spreading. In plants mutated in the Argonaute or Dicer
genes, infection spreads unhindered and viral replication is much
higher. The protecting signal comprises siRNAs generated from the viral
genome itself. Viruses have retaliated: they often carry genes
whose function is to protect the infecting virus from host RNAi. One
example is HcPro from potato virus Y, which acts to reduce production
or stability of siRNAs. Other viral products effect other steps in
the defense mechanism, including the systemic spreading of siRNAs.
Links between RNAi and human disease are described in Box 18-3,
RNAi and Human Disease.
RNAi Has Become a Powerful Tool for Manipulating Gene
Expression
The discovery of RNAi arose from observations made by investigators
attempting to manipulate gene expression (see Box 18-2). In the cases of
both cosuppression in plants and antisense RNA in worms, it was attempts
to understand unexpected blips in those manipulations that led
to the discovery of RNAi. It was therefore perhaps not surprising that
once understood, RNAi was quickly exploited as a tool for manipulating
gene expression. In worms, this is now routinely done. Libraries exist
that encode dsRNAs that can target any gene in the worm genome.
These libraries can be used to screen worms for the consequences of inhibiting
expression of any given gene. The general way in which this is
done is shown in Figure 18-14. Worms feed on bacteria. In the lab, they
are fed E. coli, and it turns out that the quickest route to a worm’s
genome is through its stomach: any desired dsRNA can be expressed in
the E. coli on which the worms feed, and this delivers enough substrate
for the RNAi response to be triggered in cells of the worm, switching off
genes homologous to the original dsRNA.
It would, of course, be of great benefit to screen for genes in this
way in mammalian cells, where traditional genetic screens are not feasible.
It was established that artificially synthesized siRNAs, made in
vitro and introduced into mammalian cells in culture, trigger an RNAi
response and down-regulate appropriate target genes, but the efficiency
of transfection (getting the RNAs into the cells) was low.
Longer dsRNA molecules are also problematic because they trigger a
response that shuts down all translation in the cell, a response
evolved to block viral replication; many viruses have RNA genomes.
Despite these drawbacks this approach can still be fruitfully em-
ployed.
Being able to express the RNAs in cells (rather than having to introduce
them in high concentrations) would get around the problems
mentioned above and have a further benefit: the RNAs would be pro-
654 Chapter 18
PAGE
PRO
O
FS
PCR amplify portion of
gene of interest
ligate into
plasmid
introduce plasmid into E. coli
strain and express shRNAs
as explained in the legend
screen for phenotype of interest
add worms to
each well in plate
T7 RNA
polymerase lac
promoter
AT
T
A
T7T7T7
AmpAmpAmp
T7T7T7
A
A
F I G U R E 18-14 RNA interference can be
induced in worms by feeding bacteria expressing
dsRNAs. See Chapter 21 for details
of the molecular manipulations required in
the first steps of this scheme. The shRNA expression
from the plasmid is under the control
of a promoter recognized not by E. coli
RNA polymerase, but rather by a single-subunit
RNA polymerase from a phage called
T7. The gene for that polymerase is expressed
artificially in the cells used in this
scheme, under control of the lac promoter
(Chapter 16). Thus, production of the
shRNAs can be controlled using an inducer
of the lac promoter.
duced in a sustained manner rather than being present transiently as
they are with transfection. But this approach presents its own formidable
hurdles: how can one ensure that both complementary strands
are expressed in the same cell and find each other efficiently to form
the active dsRNA substrate for RNAi?
Instead of trying to deliver dsRNA into cells, investigators found it
more fruitful to mimic miRNAs. To this end, libraries have been generated
in which short genes are synthesized as oligonucleotides and
cloned in plasmids. Each short gene is designed to give a transcript
that will fold into a stem loop. These are processed by dicer in the
cell to generate an siRNA that will direct silencing of its target genes.
These short synthetic genes are called short hairpin RNA genes
(shRNAs). By using an appropriately designed shRNA, any individual
gene in the genome can be targeted. Or, with a suitable library, a geRegulatory
RNAs 655
PAGE
PRO
O
FS
B O X 18-3 RNAi and Human Disease
M E D I C A L C O N N E C T I O N S
Cancer
A general decrease in levels of many miRNAs is often seen in
cancers. This decrease has been taken to indicate that those
miRNAs usually have a tumor-suppressing effect. Despite this
general trend, other specific miRNAs are up-regulated in some
cancers. Analogous to protein-coding genes implicated in cancer,
the miRNAs in question are described as being tumor suppressors
(if their absence increases cancer) or oncogenic (if
their increased expression leads to cancer). Their targets tend
to be genes involved in cell cycle progression (proliferation) or
apoptosis.
Of the 300 miRNAs identified in humans, more than onehalf
are located in regions of the genome regularly disrupted in
cancers. Thus, in many cancers, the genes for these miRNAs are
deleted or amplified, depending on the nature of the chromosomal
rearrangement. So, for example, the miRNAs miR-15 and
miR-16 induce apoptosis of cells by down-regulating the BCL2
gene (see Box 18-3 Fig. 1). The most common form of adult
leukemia in the western hemisphere is chronic lymphocytic
leukemia (CLL), a disease associated with deletions in a region of
chromosome 13 (13q14). This region of the genome contains
the miRNA miR-15a and miR-16a genes; indeed, these are the
only two genes included in the smallest deletions associated with
CLL. Thus, when these genes are deleted, apoptosis is down-regulated
and tumors can more readily arise and develop.
In another region of chromosome 13 (13q31) is found miR-
17-92, an oncogenic miRNA. Compared to normal tissue, extumor
suppressor
miRNAs
MiR-15a and Mir-16
Let-7a
MiR-17–92 MiR-155
MiR-21
low expression
high expression
oncogenic miRNAs
oncogene mRNA
BCL-2
RAS
tumor suppressor mRNA
PTEN
RB2
a
b
pression of this gene is significantly increased in many cancers,
including lung cancer, and especially in its most aggressive
forms (e.g., small-cell lung cancer). In addition, overexpression
of this miRNA in transgenic mice drives tumorigenesis. Among
the many predicted targets of miR-17-92 are two tumor suppressor
genes, PTEN and RB2. One definite target is the cell
cycle progression regulator E2F1. Both these and other examples
of miRNAs in cancer are shown in Box 18-3 Figure 1.
Fragile X Mental Retardation
Through biochemical analysis of the RISC complex, a number of
associated proteins have been identified. One of these is the
Fragile X mental retardation protein (FMRP). The gene encoding
this protein (FMR1) is X-linked and its mutation is the cause of
the most common inherited form of mental retardation. FMRP
is an RNA-binding protein involved in gene regulation. Patients
lacking FMRP have a range of developmental defects, as well as
the mental retardation, due to disrupted gene expression.
Drosophila has an FMRP homolog. In flies deficient for this
gene, unusual synaptic connections between neurons and
muscles were observed. One of the Drosophila Argonaute proteins
was found to be associated with FMRP, whereas separate
studies of Argonaute found that FMRP was bound to that component
of the RNAi machinery. Similar findings followed in
human cells as well, indicating an intriguing connection between
the Fragile X condition and the RNAi gene regulatory
pathway.
B O X 18-3 F I G U R E 1 miRNAs as tumor suppressors of oncogenes. (a) In this model, an miRNA that normally down-regulates an
oncogene can function as a tumor suppressor gene. The loss of function of the miRNA by mutation or deletion, for example, might result
in an abnormal expression of the target oncogene, which would then contribute to tumor formation. (b) Here, the amplification or overexpression
of an miRNA that down-regulates a tumor suppressor or other important genes involved in differentiation may contribute to tumor
formation by stimulating proliferation, angiogenesis, and invasion. (Redrawn, with permission, from Garzon R. et al. 2006. Trends
Mol. Med. 12: 580–587, Fig. 2. © Elsevier.)
PAGE
PRO
O
FS
Regulatory RNAs 657
netic screen can be carried out. In such a library, for example, each
plasmid would encode a shRNA directed against a different gene. The
whole library is transfected into cells such that each cell receives a
different shRNA. Cells with a particular phenotype are chosen, and
the gene whose repression led to that phenotype can be identified.
REGULATORY RNAS AND X-INACTIVATION
X-inactivation Creates Mosaic Individuals
Female mammals have two X chromosomes, whereas males have only
one X and a Y chromosome. Although this is the basis of sex determination—what
enables males to differ from females—it also poses a
problem: any gene encoded by the X chromosome would, if left
unchecked, be expressed at twice the level in females as in males.
This imbalance would potentially cause disruption to metabolic and
other cellular processes. Avoiding such problems requires what is
called dosage compensation. In mammals this is achieved by females
inactivating one of their two X chromosomes. This action results in
none of the genes on that copy of the chromosome being expressed. In
placental mammals, inactivation occurs at the 32- to 64-cell stage, and
the choice of which X chromosome to inactivate—the maternal or paternal
copy—is apparently made at random in each cell at that time.
Once selected in each cell, the same copy remains inactivated in all
the descendants of that cell.
A consequence of inactivation being random in each cell is that females
are mosaics—some of their cells express the paternal and others
the maternal X chromosome. This is usually of little consequence, although
it can influence the severity of symptoms of X-linked diseases,
depending on the proportion of cells in which the mutated gene is expressed
or silenced. A more familiar example is the calico (or tortoiseshell)
cat (Fig 18-15). In cats, one gene on the X chromosome influences
whether fur is orange or black—one allele of that gene gives rise
to orange fur, another allele gives black. In cats heterozygous for this
gene, the different patches of black and orange fur reveal regions
made up of cells in which one or the other X chromosome was inactivated.
This observation also explains why all calico cats are female.
The white comes from effects of an autosomal gene.
Xist Is an RNA Regulator That Inactivates a Single X Chromosome in
Female Mammals
How is an X chromosome inactivated, and how is inactivation inherited
through the remainder of development? The initiating regulator is
an RNA molecule called Xist. This RNA is encoded within the locus
known to be vital for X-inactivation, the Xic (X-inactivation center) on
the X chromosome. Xist RNA coats the X chromosome from which it
is expressed. This is shown in the in situ in Figure 18-16a. It is not
clear what causes this coating nor how it is restricted to one X chromosome
(i.e., why it acts only in cis). It is, however, known that the
action of Xist is central to inactivation and does not require other Xchromosomal
sequences beyond Xic: when expressed ectopically
from an autosomal location (i.e., from a non-sex chromosome), Xist
can, to varying extents, silence genes along that chromosome. That is,
it “inactivates” the autosome from which it is expressed.
F I G U R E 18-15 Visualizing X-inactivation:
the calico cat. The patches of orange
and black fur provide an indirect visualization
of X-chromosome inactivation, as described
in the text. (Courtesy VG.)
PAGE
PRO
O
FS
Xist RNA itself does not cause silencing, but it recruits other factors
that modify and condense chromatin and perhaps methylate
DNA as well (just as we saw in other examples of silencing in Chapter
17). It is these modifications that cause silencing and ensure it is
inherited; once firmly established, Xist itself is no longer required.
One difference in histone modification of the inactivated X chromosome
is shown in Figure 18-16b. There, the single inactivated X chromosome
is much less acetylated than is the rest of the genome. As we
saw in earlier chapters (7 and 17), deacetylated histones are associated
with regions of the genome that are not transcribed.
How does a cell choose which X chromosome to inactivate? The
answer is not yet known. But another RNA regulator might be key.
This other RNA is also encoded by the Xic locus but on the opposite
strand and overlapping the Xist gene. It is called Tsix (Xist spelled
backward) and acts as a negative regulator of Xist. Indeed, if Tsix is
mutated on a given X chromosome, it is that chromosome that will be
chosen for inactivation. Thus, a balance between the production and
stability of the Xist and Tsix RNAs may tilt the outcome one way or
the other in each cell.
Dosage compensation is necessary in all animals—for example,
worms and flies—just as it is in mammals. But in each case the mechanisms
for achieving compensation are different. For example, in
Drosophila, it is achieved by increasing expression of X-linked genes
in the male (rather than decreasing them in the female). But there, too,
the mechanism involves noncoding regulatory RNAs. In this case, the
RNAs (called roX1 and roX2) are involved in recruiting chromatinmodifying
complexes to genes on the X chromosome in males, where
they help activate transcription.
Noncoding RNAs are also found within some clusters of imprinted
genes in the mouse. We discussed imprinting in gene expression in
Chapter 17. In addition to the mechanisms described there, it is now
believed that in some cases regulatory RNAs play a part, working in a
manner analogous to Xist and Tsix in X-inactivation. The noncoding
RNAs in question are regulated by the imprinting control region (ICR;
see Fig. 17-26) found at each cluster, and they silence nearby genes on
the same chromosome. Both DNA and histone methylation have been
implicated in the silencing mechanism.
658 Chapter 18
Xist at metaphase hypoacetylated X
a b
F I G U R E 18-16 Visualizing X-inactivation: molecular markers. (a) Localization of Xist
RNA along the inactive X-chromosome is shown by in situ hybridization in metaphase cells.
(b) Chromosomes are stained for acetylation on histone H4. The arrow points to the inactivated
X chromosome, which has much lower levels of acetylation than the other chromosomes.
(Reprinted, with permission, from Brockdorff N. and Turner B.M. 2007. Epigenetics
[ed. Allis et al.], p. 327. © Cold Spring Harbor Laboratory Press.)
PAGE
PRO
O
FS
Regulatory RNAs 659
SUMMARY
Despite it being proposed as long ago as 1961 that RNA
molecules were likely agents of gene regulation, it is only
in the last few years that their widespread occurrence and
significance in that role have come to light. Before that, attenuation
of the trp operon in E. coli was a rare case where
RNA sequences in the 5′ region of an mRNA were known
to control expression of the downstream genes. In that
case, alternative patterns of intramolecular base pairing
within that region of the RNA give rise to alternative secondary
structures that communicate different outcomes to
the genes. In one conformation, transcription is terminated
before it enters the coding region of the downstream genes,
whereas in another conformation, it allows that transcription
to continue, and the genes are expressed. Those genes
encode enzymes involved in the synthesis of tryptophan,
and the decision of whether or not to express them is based
on the availability of uncharged tRNAtrp
: if the uncharged
species is rare in the cell, more Trp is needed and the genes
are expressed; if there is ample charged tRNAtrp
in the cell,
transcription of the trp genes is terminated.
Riboswitches control genes in a similar way: alternative
secondary structures in the 5′-untranslated region of
genes determine whether transcription of those genes
continues (or, in other cases, whether translation is initiated).
With riboswitches, the choice of alternative secondary
structure depends on the direct binding to the
RNA of the ligands that control the given gene. These are
small metabolites (vitamin B12, SAM, amino acids) and
the genes regulated by each one encode enzymes that use
the metabolite in question.
E. coli also encodes small RNAs (called sRNAs) that
act in trans to regulate genes. Thus, small genes encode
short RNAs that base-pair with mRNAs bearing complementary
sequences. This situation either inhibits translation
of those target mRNAs, triggers their destruction, or
even, in some cases, stimulates their translation. The actions
of bacterial small RNAs are similar in many regards
to those of small RNAs that regulate genes in eukaryotic
cells, although the machinery used to produce these eukaryotic
RNA regulators and the machinery used in carrying
out their effects on target genes are quite different.
Eukaryotes use dsRNA as an agent of gene silencing, in
a process called RNAi. Special enzymes (Drosha and, most
generally, Dicer) recognize dsRNA and generate from that
short (21–22-nucleotide) RNAs that are the active species
for gene silencing. The dsRNA substrates from which
these are generated can arise from two complementary
strands base-pairing or a single molecule folding into a
secondary structure with a characteristic double-strand region
in an appropriate context. The former are acted on by
Dicer, and the short products are called siRNAs; the latter
require the action of Drosha followed by Dicer, and the
products are called miRNAs. Both Dicer and Drosha have
RNase domains and cut the substrate RNA on the basis of
size and structure, rather than specific sequence.
Once produced, siRNAs and miRNAs act in essentially
the same way. They are incorporated into a machine called
RISC where one of the RNA strands is selected as the socalled
guide RNA and directs the mature RISC complex to
target RNAs with complementarity to that guide RNA. Once
there, RISC either “slices” the RNA (through its catalytic
subunit Argonaut, which includes an RNaseH-related domain)
or inhibits the translation of the mRNA. Which route
to silencing is chosen depends largely on how good the
base-pairing match is between the guide RNA and the target—the
higher the match, the more likely it is to trigger
slicing. The guide RNA can also direct RISC with associated
histone-modifying complexes to promoter regions where it
silences genes transcriptionally by modifying their promoters.
It is likely that even in these cases, recruitment to the
promoter is through base pairing between the guide RNA
and an mRNA, but in this instance, one still being made and
thus still associated with RNA polymerase II at the gene.
miRNAs are encoded by genes within organisms where
they typically act as regulators of genes involved in development—those
from worms and plants are well-studied examples.
miRNAs have also been associated with cancer,
with some miRNAs being classified as tumor suppressors
and others as oncogenes. The dsRNAs that give rise to
siRNAs can arise from various sources ranging from infecting
viruses, to transcribed repeat regions (centromeres or
transposons), to dsRNA introduced into a cell deliberately
by an experimenter who wants to down-regulate expression
of a specific gene. This latter use of RNAi has become a regular
tool and is particularly useful in systems where traditional
genetics is not feasible.
Finally, we saw another case where noncoding RNAs
regulate an important biological process. Female animals
have two X chromosomes, whereas males have just one
(and a Y chromosome). To ensure both sexes express comparable
amounts of X-chromosome gene products, a mechanism
of dosage compensation must correct for this unequal
chromosome number. Mammals do this by
inactivating one of the X chromosomes in females. An
RNA molecule (Xist) encoded on the X chromosome regulates
this process.
PAGE
PRO
O
FS
660 Chapter 18
BIBLIOGRAPHY
Books
Hannon G.J. 2003. RNAi: A guide to gene silencing. Cold Spring
Harbor Laboratory Press, Cold Spring Harbor, New York.
Regulatory RNAs. 2006. Cold Spring Harbor Symp. Quant.
Biol., vol. 71. Cold Spring Harbor Laboratory Press, Cold
Spring Harbor, New York.
Bacterial Small RNAs
Coppins R.L., Hall K.B., and Groisman E.A. 2007. The intricate
world of riboswitches. Curr. Opin. Microbiol. 10: 176–181.
Gottesman S. 2005. Micros for microbes: Non-coding regulatory
RNAs in bacteria. Trends Genet. 21: 399–404. Review.
Storz G. and Haas D. 2007. A guide to small RNAs in microorganisms.
Curr. Opin. Microbiol. 10: 93–95.
Riboswitches and Attenuation
Gilbert S.D. and Batey R.T. 2006. Riboswitches: Fold and
function. Chem. Biol. 13: 805–807.
Gollnick P., Babitzke P., Antson A., and Yanofsky C. 2005.
Complexity in regulation of tryptophan biosynthesis in
Bacillus subtilis. Annu. Rev. Genet. 39: 47–68.
Winkler W.C. 2005. Riboswitches and the role of noncoding
RNAs in bacterial metabolic control. Curr. Opin. Chem.
Biol. 9: 594–602.
Winkler W.C. and Breaker R.R. 2005. Regulation of bacterial
gene expression by riboswitches. Annu. Rev. Microbiol. 59:
487–517.
Yanofsy C. 2000. Transcription attenuation: Once viewed as a
novel regulatory strategy. J. Bacteriol. 182: 1–8.
Mechanisms of RNAi
Baulcombe D. 2005. RNA silencing. Trends Biochem. Sci. 30:
290–293.
Cook A. and Conti E. 2006. Dicer measures up. Nat. Struct.
Mol. Biol. 13: 190–192.
Hannon G.J. and Rossi J.J. 2004. Unlocking the potential of the
human genome with RNA interference. Nature 431: 371–378.
Pei Y. and Tuschl T. 2006. On the art of identifying effective
and specific siRNAs. Nat. Methods 3: 670–676.
Peters L. and Meister G. 2007. Argonaute proteins: Mediators
of RNA silencing. Mol. Cell 26: 611–623.
Rana T.M. 2007. Illuminating the silence: Understanding the
structure and function of small RNAs. Nat. Rev. Mol. Cell
Biol. 8: 23–36.
Tang G. 2005. siRNA and miRNA: An insight into RISCs.
Trends Biochem. Sci. 30: 106–114.
Tolia N.H. and Joshua-Tor L. 2007. Slicer and the Argonautes.
Nat. Chem. Biol. 3: 36–43.
Valencia-Sanchez M.A., Liu J., Hannon G.J., and Parker R.
2006. Control of translation and mRNA degradation by
miRNAs and siRNAs. Genes Dev. 20: 515–524.
Zaratiegui M., Irvine D.V., and Martienssen R.A. 2007. Noncoding
RNAs and gene silencing. Cell 128: 763–776.
MicroRNAs
Ambros V. 2004. The functions of animal microRNAs. Nature
431: 350–355.
Bushati N. and Cohen S.M. 2007. microRNA functions. Annu.
Rev. Cell Dev. Biol. 23: 175–205.
Hannon G.J., Rivas F.V., Murchison E.P., and Steitz J.A. 2006.
The expanding universe of noncoding RNAs. Cold Spring
Harbor Symp. Quant. Biol. 71: 551–564.
History of miRNAs and RNAi
Baulcombe D. 1994. Replicase-mediated resistance: A novel
type of virus resistance in transgenic plants? Trends Microbiol.
2: 60–63.
Lee R., Feinbaum R., and Ambros V. 2004. A short history of a
short RNA. Cell (suppl. 2) 116: S89–S92, 1 p. following S96.
Ruvkun G., Wightman B., and Ha I. 2004. The 20 years it took
to recognize the importance of tiny RNAs. Cell 116(suppl.
2): S93-S96, 2 p. following S96.
MicroRNAs and Disease
Calin G.A. and Croce C.M. 2006. MicroRNA signatures in human
cancers. Nat. Rev. Cancer 6: 857–866.
Dalmay T. and Edwards D.R. 2006. MicroRNAs and the hallmarks
of cancer. Oncogene 25: 6170–6175.
Garzon R., Fabbri M., Cimmino A., Calin G.A., and Croce
C.M. 2006. MicroRNA expression and function in cancer.
Trends Mol. Med. 12: 580-587.
Zhang B., Pan X., Cobb G.P., and Anderson T.A. 2007. microRNAs
as oncogenes and tumor suppressors. Dev. Biol. 302:
1–12.
Xist RNA and X-inactivation
Alexander M.K. and Panning B. 2005. Counting chromosomes:
Not as easy as 1, 2, 3. Curr. Biol. 15: R834–R836.
Deng X and Meller V.K. 2006. Non-coding RNA in fly dosage
compensation. Trends Biochem. Sci. 31: 526–532.
Heard E. 2005. Delving into the diversity of facultative heterochromatin:
The epigenetics of the inactive X chromosome.
Curr. Opin. Genet. Dev. 15: 482–489.
Ng K., Pullirsch D., Leeb M., and Wutz A. 2006. Xist and the
order of silencing. EMBO Rep. 8: 34–39.
Pauler F.M., Koerner M.V., and Barlow D.P. 2007. Silencing by
imprinted noncoding RNAs: Is transcription the answer?
Trends Genet. 23: 284–292.