145

SPLIT GENES AND RNA SPLICING

Nobel Lecture, December 8, 1993
by
PHILLIP A. SHARP
Department of Biology and the Center for Cancer Research, Massachusetts
Institute of Technology, Cambridge, MA 02139 - 4307, USA

INTRODUCTION
By the late 70s the physical structure of a gene was firmly established from
work in bacteria. The sequences of the gene, the RNA and the protein were
colinearly organized and expressed. Since the science of genetics suggested
that genes in eukaryotic organisms behaved similarly to those of prokaryotic
organisms, it was naturally assumed that this bacterial gene structure was
universal. It followed that if the gene structure was the same, then the
mechanisms of regulation were probably very similar, and thus what was
true of a bacterium would be true of an elephant.
However, many descriptive biochemical aspects of the genetic material
and its expression in cells with nuclei suggested that the simple molecular
biology of gene expression in bacteria might not be universal. First, obviously, RNAs transcribed from nuclear genes are physically separated from
the translational machinery in the cytoplasm. Thus, the nuclear compartment could be the site of selective RNA processing and transport. Second,
the DNA content of eukaryotic germ cells varied significantly between
organisms without an apparent variation in the number of genes. Some
organisms appeared to have ten times as much DNA as was required to
encode all of the proteins. Third, the previously described phenomenon of
heterogeneous nuclear RNA (hnRNA) suggested that long RNAs were
transcribed from diverse nuclear sequences (1). These hnRNAs had a short
half-life relative to cytoplasmic mRNAs and thus could potentially be precursors to mRNAs. Furthermore, both the long hnRNA and the shorter
mRNA appeared to have modified 5 and 3 termini in common, a
7mGpppX cap (2, 3, 4) and polyadenylation tracts (5, 6, 7), respectively. The
meaning of these observations concerning hnRNAs remained controversial
as it was not possible to establish a precursor-product relationship between
the nuclear RNA population and the cytoplasmic mRNA.
Whether or not the structure of genes in eukaryotic cells was the same as
that in bacteria was not really questioned at that time. The important issue
was to establish the exact biochemical pathway between a gene in the
nucleus and its mRNA in the cytoplasm. This hypothetical pathway was

146

Physiology or Medicine 1993

pictured as beginning with initiation of transcription by RNA polymerase

II(B) which would proceedthrough completion of transcription beyond the
terminus of the gene. The nuclear precursor RNA transcribed from the
gene was potentially processed and the mRNA selectively transported to the
cytoplasm. Regulation of gene expression, the basis of almost all interesting
biology, including cancer, cellular responses to infection, and development,
would primarily result from changes in the rates or efficiencies of the
various steps in the pathway. Thus, understanding the pathway of eukaryotic gene expression promised new insights into the mechanisms of regulation and the biochemical events controlling both diverse biological and
biomedical phenomena.
BACKGROUND
The late stage of adenovirus infection was chosen as the best system for
studying the pathway of nuclear gene + nuclear precursor RNA + cytoplasmic mRNA + protein. We had previously established several restriction
endonuclease cleavage maps of the viral genome (Fig. 1; 8) and had used
fragments of the genome to map the positions of cytoplasmic mRNAs
generated during the early and late stages of infection (9). We had also
determined the abundance (in copies per cell) of both nuclear and cytoplasmic RNAs (l0), and this suggested that RNAs from both compartments
could be obtained in adequate amounts to compare their structures directly
using the electron microscope and the then recently developed RNA.DNA
hybridization methods. Furthermore, these earlier studies established that
sets of viral RNA sequences were restricted to the nucleus, suggesting a
selection in processing and/or transport of only certain RNA sequences to
the cytoplasm (11). Finally, several studies had established that long nuclear
hexon
I
1
B

A
I

I
0

GC
I

I
10

B
I

I
I

I
20

I
30

J D
II
I

I
40

Fig. I: Map of cleavage sites for

I
50

F D E

A
I

I
60

I
70

I Eco RI

H L E
F K
II
I
I I Hind III

I
80

I
90

I MAP UNIT
100

restriction endonucleases EcoRI and HindIII. The approximately

35,000 bp of adcnovirus 2 DNA was assigned as 100 map units. The positions of cleavage sites
are denoted by vertical lines and fragments are lettered on the basis of length. The r and 1 viral
strands are transcribed to the right and left, respectively. The sequences constituting the body of
mRNA specifying the abundant hexon (II) protein are encompassed by the bar about the
boundary of EcoRI fragments A and B.

Phillip A. Sharp

Fig. 2: Electron micrographs of

hybrids of hexon mRNA andfragments of Ad2 DNA (13). Examples of

the latterR-toop hybrids observed after incubation of hexon mRNA and duplex HindIII A
fragment DNA arc shown in A and B and is diagrammed schematically in C. Similarly, two
examples of hybrids of hexon mRNA and the single-stranded HindIII A fragment are shown in
D and E. A schematic. of the hybrid structure shown in E is given in F. The single-stranded RNA
at the end of the hybrid region is represented by a wave-like line. In A, B, D, and E the positions
of the KNA tails at the 5 and 3 ends of the hybrids are denoted by arrows. An example of a
hybrid between single-stranded EcoRI A DNA and hexon RNA is shown in G and diagrammed in
H. The hybrid region is indicated by a heavy line; loops A, B, and C (single-stranded unhybridized DNA) are joined by hybrid regions resulting from annealing of upstream DNA sequences
to the 5 tail of hexon mRNA. Bars on micrographs represent 0.1 M.

148

Physiology or Medicine 1993

RNA was transcribed from the adenovirus genome during the late stages of
infection (12) and the stable cytoplasmic RNAs were shorter than the
predominant nuclear RNAs. Thus, the production of RNAs during the late
stage of adenovirus infection presented a paradigm for the heterogeneous
nuclear RNA phenomenon associated with cellular genes.
DISCOVERY OF RNA SPLICING
Comparison of the structure of a cytoplasmic mRNA to that of a nuclear
precursor RNA (13) required the purification of a specific homogeneous
mRNA. The most abundant mRNA, which encoded the adenovirus hexon
protein, was separated from other viral mRNAs by gel electrophoresis and
used in electron microscope mapping studies (14). Ray White, as a fellow at
Stanford, was the first to recognize that RNA.DNA hybrids were more
stable than DNA.DNA duplexes in high concentrations of formamide (15).
This phenomenon was characterized physically by Davidsons lab (16) and
was the basis of a convenient R-looping technique whereby RNA forms a
hybrid with a DNA strand, displacing the other strand of DNA into an easily
observed loop. The adenovirus mRNA for hexon protein was mapped by
the R-loop method to the HindIII A fragment of adenovirus 2 (13).
Inspection of the R-loops between the hexon-mRNA and the HindIII B
DNA fragment revealed the presence of RNA tails at both the 5 and 3 ends
of the hybrid (Fig. 2A, B and C). The single strand RNA tail at the 3 end was
expected, as the RNA was known to be polyadenylated post-transcriptionally. The single strand RNA tail at the 5 end of the hybrid was not expected;
however, this 5 tail of RNA could have been displaced from the RNA.DNA
hybrids by formation of duplex DNA by a process called branch migration.
In fact, similar 5 tails had been observed previously in R-loop mapping of
adenovirus mRNA and had been ascribed to such branch migration (17,
18). Arguing against branch migration however was the finding that the
lengths of the 5 tails were relatively uniform, 170 nucleotides. To eliminate
this potential of competing DNA sequences displacing RNA sequences, we
used a denatured single-strand of the HindIII fragment (Fig. 2D, E, and F)
to form a RNA.DNA hybrid. Surprisingly, the 5 RNA tail still did not form
a hybrid with the adjacent viral sequences, suggesting that these RNA
sequences were derived from other DNA sequences.
Could the DNA sequences transcribed to form the 5 tail sequence, the
leader RNA sequence, be located upstream of the body of the hexon
mRNA, perhaps as part of the long nuclear RNA? To test this possibility, a
strand of the EcoRI A DNA fragment was hybridized with the hexon mRNA.
This fragment contained all of the linear viral sequences which could have
been transcribed by the polymerase before encountering the body of the
hexon mRNA. Surprisingly, and wonderfully, the leader sequences hybridized to three short tracts of DNA sequences, creating three different size
loops A, B and C of intervening single-strand DNA (Fig. 2G and H). The
length of these loops mapped the positions of the leader sequences, Ll , L2,

149

Richard J. Roberts

L3, to approximately 16.9, 19.8, and 26.9 map units on the genome,
respectively. The distances between the bases of the three loops permitted
an estimate of the lengths of the two internal leaders to be 80 and 110
nucleotides, respectively. The 5 proximal leader was estimated to be quite
short, but longer than fifteen nucleotides.
RNA splicing was the mechanism we proposed for generation of the final
hexon mRNA (13). It was known that the nucleus of virus-infected cells
contained long RNAs transcribed from the viral sequences between the 5
most leader Ll , located at 17 map units, and the body of the hexon mRNA,
located between 51.7 and 61.3 map units (19, 20). Thus, the long nuclear
RNA probably contained sequences for all three leaders, Ll , L2 and L3, as
well as for the body of the mRNA. These sequences were conjectured to be
joined by excision of the intervening sequences and ligation of the flanking
RNAs, a process dubbed RNA splicing (Fig. 3). In fact, the nuclear RNAs
were quite abundant and were easily visualized by electron microscopy of
the RNA.DNA hybrids (21). Analysis of the structure of these RNA.DNA
hybrids revealed the presence of potential intermediates, where only Ll had
been spliced to L2 and where only Ll, L2 and L3 had been joined by
splicing:
The RNA splicing hypothesis reconciled many paradoxes. Both the longer nuclear RNAs and the shorter cytoplasmic mRNAs could share 5 cap
termini, 7 mG pppX, and 3 po1yA tracts (22), because internal sequences
were removed from the nuclear precursor. More specifically for adeno-

17 20

27
m
u

52
c

hexon
(AIn
hexon
(A)n
L123

hexon
(A)n

(A)n
mature hexon mRNA
Fig. 3: Proposed RNA splicing mechanism for synthesis of mRNA for hexon protein. A long nuclear
precursor RNA is transcribed from 16.9 map units through the polyA addition site at the end of
the body of hexon mRNA. The region of the Ad2 genome from which the precursor RNA is
transcribed is shown at the top of the figure. The four RNA segments in the cytoplasmic mRNA
are processed from this precursor by excision of intervening sequences (denoted by dashed arrows).

150

Physiology or Medicine 1993

virus, previous evidence had suggested that many different viral mRNAs
appeared to share a common 5 terminal sequence (23). This sequence, a
long Tl oligonucleotide, contained the cap and an apparently unique
sequence eleven residues in length. If many of the late viral mRNAs were
processed by splicing from the same type of precursor RNA, then they
could share a common sequence at their 5 termini. More importantly, the
RNA splicing hypothesis provided an explanation for the hnRNA phenomenology associated with cellular genes. Heterogeneous nuclear RNA transcribed from diverse cellular genes could be processed by RNA splicing into
shorter cytoplasmic mRNAs. Thus, most cellular genes probably contained
sequences which were removed by RNA splicing, i.e. they were split genes.
SPLIT GENES: INTERVENING SEQUENCES OR INTRONS IN
CELLULAR GENES
Shortly after the discovery of RNA splicing and split genes in adenovirus, a
number of cellular genes were also shown to have introns or intervening
sequences. For example, the globin genes contained two intervening sequences (24, 25), the ovalbumin gene was split into eight sets of sequences
(26), and the immunoglobulin genes contained both short and long introns
(27). In fact, the average cellular gene contains approximately eight introns
and the primary transcription unit is typically four times larger than the
final mRNA. Shortly thereafter, it was recognized that there was a limited
set of conserved sequences at each intron boundary (28). Interestingly,
these consensus sequences were common of vertebrate, plant and yeast cells
(29) suggesting the splicing process was evolutionarily general. Introns in
the latter organisms are generally shorter and have more highly conserved
sequences at their boundaries.
Phylogenetic comparison of the sequences of homologous genes from a
variety of organisms revealed that intron sequences had drifted much more
rapidly than exon sequences. This suggested that intron sequences are
generally not functional, at least in the context of requiring long tracts of
specific sequences. Furthermore, the length of introns in homologous
genes varied significantly during evolution, suggesting little constraint.
Finally, it was clear that specific introns could be lost during evolution. The
mechanism responsible for the exact deletion of introns is probably related
to gene conversion using a cDNA copy of the mRNA or a partially spliced
intermediate RNA. This process has been documented for the removal of
introns from yeast genes (30) and raises the question of why introns persisted during evolution.
MUTATIONS IN CELLULAR GENES
Many human diseases are caused by mutations that interfere with RNA
splicing. Approximately one quarter of all mutations in the human globin
genes underlying thalassemia are mutations in sequences specifying correct

151

Phillip A. Sharp

RNA splicing. Thus, conserving information for accurate splicing is a

constraint on genetic systems. The limited nature of the mutations altering
splicing in the globin family is interesting (31). All of the characterized
mutations either alter the conserved 5 or 3 splice site sequences at the
boundaries of the intron, or are changes within introns that generate new
consensus splice sites (Fig. 4). The former is simple - mutation of the highly
conserved GU or AG sequences at the boundaries of the intron commonly
inactivates splicing at that site-this frequently leads to the activation of a
nearby cryptic splice site. The latter is more complicated. In this case, a
mutational change within an intron- CT -, GT activates this site as a 5
splice site and, as a consequence, an upstream AG is activated as a 3 splice
site for further splicing. This results in the generation of a short exon from
the previous intron sequences (Fig. 4). Thus, mutations can alter both the
position of splice sites and the number of exons to generate novel proteins.
Mistakes are probably not uncommon in splicing of RNA from complex
genes. Exons can occasionally be skipped and, in some cases, circular RNAs
can be generated (32, 33). Since specific functions have not been assigned
to many of these RNAs, their generation may reflect noise in the system.
Interestingly, it has recently been proposed that an mRNA surveillance
system exists in cells which destroys RNAs as they enter the cytoplasm if they
contain an open reading frame interrupted by a translational termination
signal (34). This system would probably degrade most mRNAs with splicing
errors as they are transported to the cytoplasm.

cryptic
5 splice sites

-t

new exon
Fig. 4: -Globin mutations. -Globin genes of thalassemia patients possessing mutations which
affect RNA splicing (31). The single base changes in four independent patients are indicated by
arrows from the diagram of the two intron structure of the -globin gene. Three of the
mutations alter conserved sequences of the 5 splice site and a fraction of the mRNA from these
mutant genes are processed at the cryptic 5 splice sites. The fourth mutation is within the
sequences of the second intron and creates a 5 splice site at this position. This results in the
induction of the indicated sequences as a new exon. Exon sequences are represented by
rectangles, intron sequences by lines.

Physiology or Medicine 1993

152

WHY SPLIT GENES AND RNA SPLICING?

It is diffkult to confidently account for why introns have been conserved
during evolution within the genes of most, if not all, eukaryotes. Clearly the
intron-exon structure of genes has been very important in the generation of
new genes during evolution. For example, the fibronectin gene of man is
composed of three types of exons, each of which encodes a specific type of
protein folding domain (Fig. 5; 35). These same three protein-folding
domains and corresponding exon structures are found in other genes, some
of which encode cell surface receptors and blood coagulation proteins.
Thus, the fibronectin gene was created by tandem and dispersed duplication of exon units using breakage and joining within the intron sequences.
Since these exon units are assembled precisely by RNA splicing and encode
functional protein domains, the final protein is stable and has multiple
functions. This theme of duplication and utilization of an exon unit has
been a common mechanism for the generation of new genes encoding many
cell surface receptors and other types of proteins in vertebrates. Thus, the
presence of exon units which correspond to functional protein domains has
been critical for the evolution of complex organisms.

Fibronectin Protein:

Fibronectin Gene:

Protein Subdomain
Type I

0 - Also Found In Tissue Plasminogen Activator

Type ll ()

- Also Found In Blood Coagulation Proteins

Type Ill 0

- Also Found In Cell Surface Receptors And Other

Extracellular Matrix Proteins

Fig. 5: Fibronectin gene evolved by exon duplication. The large extracellular matrix protein, fibronectin, is primarily composed of three types of protein domains, I, II, and III denoted as tilted
ellipsoids, vertical ellipsoids and circles, respectively. As indicated at the bottom, homologous
domains are found in other cellular proteins. Each of these domains are encoded by a defined
exon pattern, one or two exons, denoted as verticle rectangles in the middle. These exon units
are duplicated in the ftbronectin gene and also in the other genes containing these domains. In
two cases, larger rectangles, the intron separating the typical two exon structures of the type III
subdomain, have been deleted. These patterns suggest that descendants of a common progenitor exon configuration was used to generate parts of these genes (36).

Phillip A. Sharp

153

The ability to alternatively select different combinations of exons at the

stage of RNA splicing to generate proteins with different functions is clearly
critical for the viability of many vertebrate organisms. Approximately, one
of every twenty genes is expressed by alternative pathways of RNA splicing
in different cell types or growth states. For example, nuclear precursor
RNAs from the fibronectin gene are alternatively spliced into mRNAs that
encode 20 different proteins. These proteins have slightly different functions reflecting their variant structures. Furthermore, a set of exons is not
included in mRNAs processed in liver cells but is included in mRNAs in
other cell types (Fig. 6). The fibronectin secreted by the liver more readily
circulates in the blood stream because it has fewer protein domains which
are recognized by receptors on the cell surface than the other forms of
fibronectin (36). The fibronectin secreted by other cells is more typically
found in the intracellular matrix of solid tissue. A similar variation in
cellular adhesion is observed with alternative splicing patterns in another
part of the fibronectin mRNA. Thus, through alternative splicing, the same
set of gene sequences can be utilized for different functions.
Alternative splicing of precursor RNAs in different cell types must reflect
differences in the factors regulating the splicing process. Only in few cases
have these factors been identified, primarily by the analysis of mutants
which are defective in the regulatory step. The development of sexual
dimorphism in the fruit fly Drosophila is regulated at the level of alternative
RNA splicing by expression of a cascade of genes. Early in development, the

FIBROBLASTS
LIVER

Fig. 6: Alternative splicing of RNA encoding fibronectin. The exons denoted as EIIIB and EIIIA are
spliced into the mature RNA synthesized in many cell types including fibroblasts. However, in
the liver, these two exons are not included, are skipped, in synthesis of the mRNA. Furthermore, five variations for the splicing patterns of the V region are shown. All of these variations
on splicing are found in most cell types. The fibronectin proteins encoded by the first and third
mRNA bind differentially to receptors on lymphocytes. In total, at least twenty different
proteins are synthesized from the single fibronectin gene.

154

Physiology or Medicine 1993

female or male specific pattern of splicing is set by a reading of the ratio of

sex chromosomes to autosomal chromosomes. After setting this switch, the
developmental process is controlled by an autoregulatory process which
maintains the male or female splicing pattern in many if not all cells (37).
Interestingly, most of the proteins encoded by these sex regulating genes
are members of a family whose common features are one or more RNA
recognition subdomains and a pronounced tract of the repeating amino
acid sequence Arg-Ser (38). It is likely that these factors directly interact
through the Arg-Ser tract with other proteins critical for formation of the
splicing machinery.
SPLIT GENES IN THE PROGENOTE?
The split gene structure must be very old, clearly predating the divergence
of organisms that gave rise to plants, yeast and man (29). First, as will be
described later, a complex splicing machinery, which involves greater than
50 - 100 proteins and five small RNAs, exists in the nucleus of all eukaryotes. This machinery is highly conserved and could not have arisen spontaneously in the various lineages. Second, introns are conserved in positions
within homologous genes found in the lineages of plants and animals. This
evidence establishes that split genes and RNA splicing were present in
common progenitor organisms a billion years ago.
Several scientists have speculated that genes originally evolved as exons
and that the progenote organism from which current prokaryotic and
eukaryotic organisms evolved may have had a split gene structure (39 - 42).
These primordial exons are pictured as encoding sequences for stable
protein folding domains. Assembly of a number of exon sequences by RNA
splicing would be expected to produce a protein composed of stable folding
domains which have a high probability of being functional either structurally or catalytically. If genes originally evolved in this fashion, the positions of
introns in relationship to protein secondary structure might not be random.
Evidence to support this hypothesis has been sought in the exon-intron
structure of evolutionarily old proteins critical for energy metabolism. For
example, the enzyme pyruvate kinase (PK) is conserved in structure and
functions in bacteria, yeast and chicken (Fig. 7; 43). The gene is not
interrupted with introns in bacteria and yeast, but contains ten introns in
chicken, nine within the coding sequences. The positions of these nine
introns are not random when compared to the protein structure. First, in
the N-terminal part of the protein, the first three introns are positioned
between repeating structural motifs of helix- sheet. Second, in the
mononucleotide binding fold. PK shares a common intron position with
one other old metabolic enzyme, alcohol dehydrogenase. In PK, the eighth
intron occupies a position similar to that of an intron in the dehydrogenase.
Since the generation of these two enzymes from a common protein subdomain clearly predated the evolutionary divergence of prokaryotic and eukaryotic organisms, these results suggest that the progenitor organisms for

Phillip A. Sharp

Fig. 7: The structure of pyruvate kinase and intron positions. Pyruvate kinase is shown schematically
as a protein with secondary and tertiary structure on the left. On the right, this structure is
expanded at the positions of introns in the tertiary structure of the protein. Note that the first
three introns all fall between a-helix (barrels) and -sheet (arrows) repeats. Intron 8 is positioned in approximately equivalent positions in the mononucleotide binding fold of PK and
maize alcohol dehydrogenase. Reprinte d with permission from ref 44.

both prokaryotic and eukaryotic organisms may have had a split gene
structure more typical of a current eukaryotic cell (for discussion see 44).
IS THE GENE AN EXON?
The gene was first genetically defined as a unit of inheritance which is
associated with a locus on a chromosome. The chemical definition of a gene
has become much more difficult however, as the complexity of genetic
information and its modifications are discovered. The existence of alternative splicing of exons, where information at an exon unit can be optionally
expressed in some cells and not in others, suggests that an exon might
correspond to a gene. That is, an exon corresponds to the minimal amount
of information which is expressed as a discrete unit. This concept becomes
particularly relevant when the trans-splicing process is considered (45). In
this case, exons transcribed from different loci, and in many cases different
chromosomes, are joined by RNA splicing. Trans-splicing of exons and
introns has been established in the parasite trypanosomes (46), the flat
worm C. elegans (47), and suggested for some human genes (48). Clearly, in
the case of trans-splicing, both the unit of inheritance and the locus on a
chromosome can correspond to a single exon.

156

Physiology or Medicine 1993

It is unlikely, however, that the current working definition of a gene as a

linear collection of exons which are joined by RNA splicing will be radically
altered in the near future. This is probably wise, as the existence of multiple
processes collectively called RNA editing further complicates the biochemical definition of a gene (49). However, given the possibility that the earliest
unit of genetic information may have evolved as an exon, the general
concept of exons as gene units may be more valid than any other proposal.
SPLICING OF NUCLEAR PRECURSORS TO mRNAs
The development of a reaction composed of soluble cellular components
which accurately processed precursors to mRNAs (pre-mRNAs) was critical
for advancing our understanding of splicing (50). When combined with the
use of highly radioactive pre-mRNA substrates, a biochemical analysis of the
splicing process became feasible (51, 52). Not surprisingly, kinetic RNA
intermediates in the splicing reaction were soon identified (53, 54). Surprisingly, these intermediates had a lariat structure where the 5 most nucleotide of the intron was joined in a 2-5 phosphodiester bond to an adenosine
within the intron (55). Since the adenosine is covalently bonded through
both 3-5and 2-5 phosphodiester linkages, this forms an RNA branch. The
existence of such branches had just been described from nuclease digestion
studies of total nuclear RNA from human cells (56). Formation of the
branch appears simultaneously with cleavage at the 5 splice site and generates the lariat RNA which typically migrates more slowly than the premRNA during electrophoresis through a tight porosity polyacrylamide gel.
The splicing of pre-mRNA proceeds in two steps (Fig. 8, left). As mentioned above, the first step consists of cleavage at the 5 splice site with the
concomitant formation of the branch. At this stage, the 5 exon RNA has a
3 hydroxyl group and the lariat intermediate RNA contains the intron and
3 exon. The second step consists of cleavage of the RNA at the 3 splice site
with concomitant joining of the two exons. The intron is released as a lariat
RNA and is reasonably stable in reactions in vitro. This contrasts with the
situation in vivo where intron RNAs are almost always rapidly degraded.
The fact that the intermediate state, consisting of two RNAs, was efficiently converted to the final products strongly suggested that these RNAs
remain bound in a complex. The complex was identified by its rate of
sedimentation in a glycerol gradient, 60S, and was designated to be a
spliceosome or splicing body (57, 58). As anticipated from earlier work
suggesting the importance of small nuclear ribonucleoprotein particles in
splicing, the spliceosome contained the small nuclear RNAs (snRNAs) U2,
U4, U5 and U6 and, under certain conditions, Ul (59). Thus, the spliceosome, much like a ribosome, contains a substrate RNA and a number of
stable cellular RNA-protein components.

Phillip A. Sharp

Spliceosome
Nuclear pre-mRNA

-p-

157

Self-splicing
Group

P-

-Ip-

Fig. 8: Comparison of self-splicing and nuclear pre-mRNA splicing mechanism. The first column
outlines the mRNA precursor splicing mechanism. The shaded circle represents a multicomponent complex, the spliceosome, which promotes the splicing reaction. The second column
outlines the splicing mechanism of self-splicing introns of the group I type. This process is
catalyzed by RNA structures within the intron (dark semicircle), which contains a guanosine
binding site, and utilizes a guanosine (C) factor in the first step. The third column outlines the
splicing mechanism of self-splicing introns of the group II type. This process is also catalyzed by
RNA structures within the intron but utilizes, instead of a cofactor, an adenosine residue (A)
within the intron to form a lariat RNA. All three mechanisms proceed by two steps, reaction at
the 5 splice site and then reaction at the 3 splice site. The fate of the phosphates at the 5 and 3
splice sites is indicated.

GROUP I, GROUP II AND THE SPLICEOSOME

Comparison of the two steps in splicing by the spliceosome to the RNAcatalyzed self-splicing reactions of group I and II introns shows some
striking similarities (Fig. 8). In all three cases, the first step is cleavage at the
5 splice site. In group I introns, this cleavage requires a guanosine which
specifically occupies a binding site in the catalytic intron sequences (60).
The 3 hydroxyl group on this guanosine is activated and through a transesterification reaction displaces the 3 hydroxyl of the 5 exon. Group II selfsplicing introns cleave at the 5 splice site by activating the 2 OH at the
branch site, producing a lariat RNA (61, 62) much like that produced by the
spliceosome. The second step for all three processes involves a reaction at
the 3 splice site to join the exons and displace the intron. The simplest
mechanism for the second step of splicing of the group I and II introns is a
single transesterification reaction. This has been proven to be the case for
reactions by group I introns.

158

Physiology or Medicine 1993

The similarities between the spliceosomal process and the self-splicing

introns suggested that these reactions might be evolutionarily related. This
is particularly the case for the group II and spliceosome reactions. Thus, the
snRNAs in the spliceosome might be pictured as a group II intron in pieces,
where these RNAs form the catalytic sites for both reactions. The proteins
in the spliceosomes could be necessary for recognition of the precursor
RNA, arranging the snRNAs in catalytic structures and rearrangement of
the components to facilitate completion of the process. Early in evolution
all introns might have been self-splicing. Then, during the passage of time,
tram-acting components may have developed which executed the splicing of
introns by recognition of sequences at the splice sites, thus permitting
atrophy of the cis-sequence within each intron (63). Trans-acting segments
of group II introns which complement the splicing of introns with partial
catalytic structure have been documented (see reference 63).
Years after the discovery of the lariat intermediate in the cis-splicing
process, studies of the trans-splicing reaction revealed a branched intermediate state. Thus, the chemistry of the truns-splicing reaction is very similar
to that of the cis-splicing process (64). It is therefore highly likely that the
cis- and truns-splicing processes are variations of a single fundamental
mechanism. Trypanosome parasites which exclusively synthesize mRNA by
truns-splicing of a short leader RNA do not apparently express Ul snRNA
or U5 snRNA, but do express U2, U4 and U6 snRNAs (65).
The sequence complementarity between the 5 end of Ul snRNA and the
consensus sequences at the 5 splice site led to the hypothesis that this
interaction was critical for splicing (66, 67). This hypothesis has been
confirmed by a number of studies including the inhibition of splicing in
vitro by addition of anti-sera specific for Ul snRNP (68). After the discovery
of lariat RNA, a consensus sequence was recognized in the region flanking
the branch site. This consensus sequence is complementary to a conserved
internal sequence in U2 snRNA, and mutational analysis has shown that this
interaction is also critical in splicing (69). Ul and U2 snRNAs recognize
consensus sequences at the 5 splice site and branch site as early steps in the
splicing reaction (Fig. 9A). At the same stage of splicing, U4 and U6
snRNAs are bound to one another through an extended region of complementarity (70). SnRNA U6 is unique relative to the other snRNAs in that: it
is not bound by core peptides recognized by the Sm lupus antisera, it is
transcribed by a different polymerase, and it has a different type of modification at its 5 terminus. U6 snRNA is also the most conserved member of
the snRNA family and may have a unique catalytic role. The U4/6 snRNP
forms a specific complex with U5 snRNP and this tri-snRNP complex is
thought to bind to the other components in formation of the spliceosome
(71).
After formation of the spliceosome, there are major rearrangements of
the snRNAs (Fig. 9B). Both genetic and biochemical experiments show that
U6 and U2 snRNAs are paired through an extensive tract of complementarity in the spliceosome (72, 73). Formation of the U2 and U6 structure

Phillip A. Sharp

Fig. 9: RNA interactions between spliceosomal snRNAs and pre-mRNA substrates. (A) (Top) Basepairing interactions between Ul and U2 snRNAs and pre-mRNA are indicated on left and right
of intron, respectively. (Bottom) Extensive base-pairing between U4 and U6 snRNAs. (B) Interactions between Ul, U2, U5, and U6 snRNAs and pre-mRNA. In both A and B, pre-mRNA
consensus sequences and snRNA sequences are those of S. cerevisiae;
uppercase nucleotides are
highly conserved between S. cerevisiae
and the known sequences of other organisms (excluding
trypanosomes, which do not have a GUAGUA sequence in U2 snRNA). Different shaded areas
highlight sequences in U2 and U6 snRNAs that change base-pairing partners during the
spliceosome cycle. Internal snRNA secondary structures that do not change between A and B
are shown as stylized stems and loops. Asterisks indicate snRNA positions at which mutations
specifically block the second step of splicing.

160

Physiology or Medicine 1993

requires dissociation of U4 snRNA from U6 snRNA. In fact, U4 snRNA can

be released from the active spliceosome after this transition. The complementarity between Ul snRNA and the 5 splice site sequence is probably
also dissociated before the first step of splicing. The 5 splice site almost
certainly pairs with another region of U6 snRNA (74, 75). U5 snRNA is
thought to be important in recognition of the exon sequences immediately
flanking the splice sites. Mutations in these flanking sequences frequently
have significant effects on splicing efficiency and these effects can be
reduced by complementary changes in U5 snRNAs (76). It is possible that
there are further rearrangements in the secondary structures formed by the
snRNAs between the first and second steps in splicing. There are mutational
changes in U2 and U6 (positions denoted by *Fig. 9) which only inactivate
the second step (72, 73). Inspection of the network of complementarity
formed between snRNAs and the pre-mRNA in the spliceosome reveals a
concentration of snRNA structure, perhaps tertiary, near the branch site
and 5 splice site sequences. This arrangement of snRNAs might form the
catalytic site for the first step. Perhaps further rearrangements would form
another catalytic site for the second step.
FORMATION OF THE SPLICEOSOME
A number of stable complexes containing snRNAs and pre-mRNA have
been partially characterized and placed in a spliceosome cycle (Fig. 10; 77).
The commitment complex, CC, forms on the pre-mRNA by recognition of
the 5 splice site sequence by Ul snRNP and sequences encompassing the
branch site and 3 splice sites (78, 79, 80). For mammalian systems, a
pyrimidine tract near the 3 splice site and branch site is particularly
important. The subsequent binding of U2 snRNP results in formation of
the very stable A complex. The stable binding of U2 snRNP to the premRNA depends critically on a protein U2AF (8l), which recognizes the
polypyrimidine tract through prototypical RNA-binding domains and signals interactions with the other splicing components through a tract of ArgSer amino acid repeats. Interestingly, Ul snRNP which is also required for
stable U2 snRNP binding, has a bound protein with a tract of Arg-Ser
repeats, the Ul - 70 kd polypeptide. Whether U2AF and Ul - 70 kd directly communicate across the intron is not clear.
The Bl spliceosome forms when the U4/U6/U5 tri-snRNA complex
binds the A complex. It is probably in Bl complex that U4 snRNA dissociates from U6 snRNA and perhaps Ul snRNA from the 5 splice site. These
events define the B2 complex which is the precursor of the Cl complex.
The first step of splicing has occurred in the Cl complex. In generation of
the C2 complex, it is likely that there are further rearrangements that are
necessary to catalyze the second step in splicing. The C2 complex dissociates
and the newly joined exons are released from the snRNP.intron complex I.
In the cell, the spliced product migrates to the cytoplasm. The lariat intron
is retained with the snRNPs and this I complex turns over. The snRNPs are

Phillip A. Sharp

5 ss

bp

3 ss

Py
a BAG

5--GU

B - 3

Pre-mRNA

5_ - . ..--q-3
mRNA
Fig. 10: Schematic representation of the spliceosome cycle in terms of the role of small nuclear ribonucleoprotein (snRNP) particles in pre-mRNA splicing. Pre-mRNA (top line), containing two exons
separated by an intron, enters splicing complexes with snRNPs and exits as mRNA (bottom line)
and excised lariat intron (left border ). Other, non-snRNP factors are required for spliceosome
formation, but have been omitted for simplicity. CC, A, Bl, B2, Cl, C2, and I represent
complexes within the splicing pathway that have been distinguished biochemically and/or
genetically. 5 SS, 3 SS, bs, and Py indicate 5 and 3 splice sites, branch site, and polypyrimidine
tract, respectively. The individual snRNPs indicated are Ul, U2, U4, U5, and U6.

162

Physiology or Medicine 1993

ss

5--GU

bp

3 ss
PY

A BAG--~
Pre-mRNA

mRNA
Fig. 11: Transitions in the spliceosome cycle which require a PRPprotein. The particular PRP mutant
is listed beside the arrow indicating the transition in the cycle in vitro which requires the mutant
protein.

Phillip A. Sharp

163

recycled for further splicing in vivo, while the intron RNA is degraded.
Since all snRNAs are very stable, a particular snRNA likely participates in
many splicing cycles.
Assembly and rearrangements of complexes as elaborate as those in the
spliceosome cycle would be expected to require a number of proteins.
Sophisticated genetic analysis has identified a number of yeast gene products that are important for splicing either in vivo or in vitro (82, 83).
Temperature sensitive mutations in PRP (precursor RNA processing) genes
cause the accumulation of unspliced pre-mRNA in the nucleus and/or
generate extracts defective for splicing in vitro (Fig. 11). Surprisingly, it is
estimated that at least a hundred different genes encode products important for RNA splicing, about 2% of the total yeast genome. Thus, the
splicing apparatus is a significant component of the nucleus. The amino
acid sequences predicted for some of the PRP genes suggests that they have
functions such as RNA binding, RNA helicase and protein-protein interaction properties. Matching these hypothetical functions to processes in spliceosomes will be challenging.
Reassuringly, the steps in the spliceosome cycle where particular PRP
proteins are required (Fig. 11) are consistent with the cycle as defined by
kinetic and biochemical methods. Most transitions between specific forms
of the spliceosome require one or more specific proteins. Furthermore, a
number of PRP mutants are defective in splicing because of their inability to
reassemble snRNPs for further splicing. Thus, both genetic and biochemical results prove that the spliceosome cycle is the process responsible for
excision of introns from split genes.
THE CHEMISTRY OF THE SPLICEOSOME
Both steps in the spliceosome process involve reactions between hydroxyl
groups and a phosphodiester bond. The products and substrates in both
steps contain the same number of covalent bonds and thus each step could
be accomplished with a single transesterification reaction. This has been
shown to be the case for group I self-splicing introns by analysis of the
stereochemistry of the two reactions (Fig. 12; 84, 85, 86). Phosphorothioates in diester bonds of RNA are chiral, having either a Rp or Sp
stereochemistry depending upon the position of the sulfur atom. The
specific stereochemistry can be determined by its sensitivity to particular
nucleases. If a chiral phosphorothioate group participates in a single transesterification reaction, its stereochemistry is inverted, for example from
Rp to Sp. This fact permits counting of the number of transesterification
reactions in a process and, thus, the detection of any potential transient
intermediate.
As implied above, Rp and Sp phosphorothioates are not equally active in
many catalytic sites. This is commonly interpreted as reflecting the unique
role of one of the oxygen groups on the phosphate in interaction with a
metal ion, a role which could not be equally well-fulfilled by a sulfur group.

164

Physiology or Medicine 1993

-Binding Site

First Step: RP active

G-lvs

-Binding Site

Second Step: Sp active

Fig. 12: The stereochemistry of the first and second step reactions of the group I self-splicing site. A sulfur

atom is indicated at the positions where it does not inhibit the reactions, Rp and Sp positions for
the first and second steps, respectively (84, 85, 86). The interactions between the oxygen group
on the phosphate and the Mg+ + groups are hypothetical, proposed to explain why a phosphorothioate with the sulfur atom in this position is not reactive.

These principles of phosphorothioate chemistry have been analyzed for the

group I self-splicing intron and now have been studied in the spliceosome.
The two steps of splicing by the group I intron are probably executed as a
forward and reverse reaction at a single catalytic site (Fig. 12). In the first
step, the guanosine cofactor is bound to the G binding site and its 3 OH
group is activated for the transesterilication reaction. In this reaction, the
Rp chiral phosphorothioate is active while the Sp is not. This suggests that
the oxygen group in the equivalent Sp position is uniquely recognized in
this reaction, perhaps by a metal ion. Consistent with a single transesterification mechanism during the first step, the Rp chiral center is converted to
a Sp phosphorothioate product (84). In the second step, the guanosine
group at the 3 splice site occupies the G binding site and the 3 OH of the
5 exon is activated for the transesterification reaction. As would necessarily
be the case by the principle of microscopic-reversibility, the Sp chiral
phosphorothioate is active in this reaction and the Rp is not (86). The Sp
substrate phosphodiester bond generates a Rp product. Thus, stereochemical analysis of the group I process strongly supports the hypothesis of a
single catalytic center for both steps.
The stereochemistry of the two steps in the spliceosome process was
investigated by synthesis of substrate RNAs containing either a Rp or Sp
phosphorothioate at either the 5 or 3 splice site (87). The particular
stereoisomer was incorporated by a combination of chemical synthesis,
oligo-primed transcription and ligation of RNAs by use of a bridging
oligodeoxynucleotide and T4 DNA ligase (88). Splicing of the phosphorothioate-containing RNA substrates was tested in nuclear extracts after

165

Phillip A. Sharp

Group I

Spliceosome

RP

tl
sp

1st step

t1

1st step

t1

2nd step

@I

G
P-

y?H

SP

t1

RP

2nd step

Gp&
-p-

Fig. 13: Chemical mechanisms and stereochemical configurations of the two steps of pre-mRNA splicing
(right) and group I intron self-splicing (left) First-step differences between spliceosomal and group

I splicing include opposite phosphorothioate diastereomer preferences, different reaction

nucleophiles (Y-OH versus 3-OH), differential placement (5 versus 3) of the conserved
guanosine (G) with respect to the phosphate at the 5 splice site. Second-step similarities include
preference for the Sp phosphorothioate diastereomer, use of a 3-OH as the nucleophile and a
conserved guanosine (C) attached through a 3 oxygen to the phosphate at the 3 splice site. The
phosphorothioate diastereomer (Rp or Sp), which is not significantly inhibitory, is indicated for
the substrate for each step. The diastereomer that results is indicated for the product of each
step. Small dashed arrows indicate the potential (but not yet observed) reversibility of each step
of nuclear pre-mRNA splicing. The dark hemicircle, indicated in the spliceosome for the second
step, designates a hypothetical site similar to that of group I introns.

166

Physiology or Medicine 1993

different periods of incubation. Analysis of the substrate and product RNAs

showed that (a) both steps in splicing involve a single transesterification
reaction with inversion of the chiral stereochemistry of the active phosphorothioate, and (b) the Sp diastereomer was the active phosphorothioate in
both steps, while the Rp diastereomer was not (87).
A comparison of the stereochemistry and critical components in the two
steps in splicing for the group I self-splicing process and the spliceosome
process is informative (Fig. 13). As discussed before, the two steps in the
group I case are forward and reverse reactions within a single catalytic
center. This is almost certainly not the situation for the spliceosome process. In both steps of the latter process, the Sp phosphorothioate was active;
thus, the second step is not a reverse of the first step. These results are most
consistent with the proposal that the spliceosome generates two different
catalytic centers for the two steps. The constituents in these centers may
partially overlap but the centers must be distinct. This is also consistent with
the different chemical nature of reaction components in the two steps. A 2
hydroxyl on an adenosine is activated in the first step, while a 3 hydroxyl
group on the 5 exon is activated during the second step. Furthermore, the
conserved guanosine residue is linked to the activated phosphate by a 5
bond in the first step and a 3 bond in the second. Thus, the spliceosome
process, and probably the group II self-splicing process, must involve two
distinct catalytic sites.
The catalytic site in the spliceosome responsible for the second step is
probably similar to that of the group I self-splicing introns. These two sites
share several common characteristics. Both catalyze the transesterification
reaction using (a) a Sp, and not Rp, phosphorothioate at the active site, (b)
the activated phosphate is linked to the 3 position to a conserved guanosine
residue, and (c) both activate a 3 hydroxyl group on the 5 exon. These
similarities are probably determined by some common chemical structure
among the two catalytic sites. Perhaps this common micro-structure reflects
a shared RNA tertiary structure for the catalytic centers and a shared
evolutionary origin.
NUCLEAR STRUCTURE AND RNA SPLICING
The processing of introns occurs from both nascent RNA which is being
extended by polymerase and from post-transcriptional precursor RNA (89).
Thus, the splicing apparatus must be located proximal to the gene, as well as
in the regions between the gene and the nuclear pores. After completion of
splicing, the mRNA is transported to the cytoplasm through one of the
approximately 4,000 nuclear pores of a typical mammalian cell (90). The
mechanism by which nuclear RNAs are transported from sites of transcription and splicing to pores remains a mystery.
The subnuclear locations of a number of proteins and RNAs important
for RNA splicing have been studied by light and electron microscopy. The
most striking feature is the concentration of snRNPs, U2, U4/6 and U5 in

,speckles
poly A+ RNA
Ul. U2. LE. U4/U 6snRNPs
SC35. SF2IASF

Ul snRNP
U2AF

coiled bodies
Ul. U2. U5. U4/U 6snRNPs
U2AF

Fig. 14: The nucleus. Subnuclear localization of splicing factors. Specific antisera and in situ
hybridization methods have been used to localize components of the spliceosome and splicing
factors in the nucleus of mammalian cells. The outline of the cell is shown without structure in
the cytoplasm. Three nuclear pores are diagrammed in the nuclear membrane. The large dark
structures are nucleoli,.site of ribosomal RNA synthesis. Some of the components of the
speckles, regions containing pre-mRNA, are listed. The coiled bodies are different and do not
appear to contain pre-mRNAs (93). The general shadowing represents the nuclear distribution
of Ul snRNP and U2AF.

20 - 50 speckled structures and also in 1 - 5 foci called coiled bodies (Fig.

14; 91, 92). This contrasts with the subnuclear location of Ul snRNP,
which, although also concentrated in speckles and foci, is more uniformly
dispersed throughout the nucleus. These speckled structures correspond to
regions described by electron microscopists as interchromatin granule clusters and the perichromatin fibril network. High resolution in situ hybridization methods and pulse labeling studies locate newly synthesized RNA in the
speckled regions and also in curvilinear tracks which extend from the gene
toward the nuclear periphery (93, 94). The further the nuclear precursor
RNA is located along the tracks which lead from its gene of origin, the lower
the relative concentration of intron sequences as compared to exon sequences. This suggests that the precursor RNA moves through the speckled
structures as it is being processed and transported (94). Active spliceosomes
are probably concentrated in these regions.
Several proteins important for RNA splicing are also concentrated in
speckled structures with the snRNPs. These include the Arg/Ser proteins
SF2/ASF (95, 96) and SC35 (97). Surprisingly, these proteins, as well as the
snRNPs in active spliceosomes, are probably attached to an operationally
defined nuclear matrix. This matrix is the structure that remains in the

168

Physiology or Medicine 1993

nucleus after almost all of the chromatin proteins and DNA are extracted by
sequential treatments with DNase 1 and high salt (98). The integrity of the
fibrillar matrix which remains, is sensitive to RNase suggesting a major role
for RNA in its structure. Spliceosomes have been shown to be associated
with the nuclear matrix as it has been possible to process pulse-labeled
endogenous precursor RNA extracted as part of the matrix (98).
Active spliceosomes may be associated with the nuclear matrix through
interactions with proteins in the Arg/Ser family (99, 100). A panel of
monoclonal antibodies has been generated to components in the nuclear
matrix of human cells. Three of these monoclonal antibodies specifically
stain the matrix and extensively co-localize with the snRNPs in the speckled
structures. Surprisingly, these three antibodies specifically immunoprecipitate spliceosome complexes which contain exon sequences (101). Thus, the
antibodies immunoprecipitate a fraction of the precursor RNA, the lariat
intermediate RNA and the associated 5 exon, and the spliced exon product
RNAs. The antibodies do not immunoprecipitate the intron containing
spliceosome complex I even though it contains most of the snRNPs and
associated protein components. These remarkable results suggest that the
matrix components bind to exon RNA sequences that are specifically assembled into splicing complexes.
The proteins recognized by the monoclonal antibodies to the nuclear
matrix are most likely members of the Arg/Ser family of proteins or are
associated with the Arg/Ser proteins. Precursor RNAs are not immunoprecipitated from reactions containing nuclear extracts depleted of Arg/Ser
proteins. Further addition of preparations of purified Arg/Ser proteins will
block specific immunofluoresence staining of subnuclear structures. These
results are consistent with a model where the Arg/Ser family of proteins are
associated with or form the nuclear matrix (101). Newly synthesized RNA
sequences are recognized by the Arg/Ser proteins such as SC35, ASF, the
U1 70 kd and UPAF and become associated with the matrix, perhaps,
through interactions of the Arg/Ser tracts. On this structure, complete
spliceosomes would form encompassing the matrix proximal splice sites and
execute splicing. The exon product RNAs would remain matrix bound and
move by some unknown mechanism to the nuclear pore.
CONCLUSION
The discovery of split genes and RNA splicing has been critical for studies of
the biology of eukaryotic organisms. Gene regulation is central to all biological phenomena and RNA splicing is important in the regulation of
genes, particularly when precursor RNAs are processed by alternative pathways to generate mRNAs encoding different proteins. The mechanism of
splicing by the spliceosome is probably related to the self-splicing process of
group I and II introns. The spliceosome process is old in an evolutionary
sense, perhaps as old as the ribosomal process responsible for translation.
Thus, the eukaryotic cell can be conjectured as consisting of two compart-

Phillip A. Sharp

169

ments, the nucleus where the spliceosome processes RNA precursors by

RNA catalysis and the cytoplasm where the ribosome translates mRNAs by
RNA catalysis. The distinct subnuclear locations of spliceosome-related
components suggest a compartmentalized organization for the nucleus.
Further studies of RNA splicing and transport hold promise of revealing
the nature of the organizational specificity of the nucleus.

REFERENCES
1.
2.

Darnell, J. E., Jr. (1975) Harvey Lectures, 69 , l-47.

Furuichi, Y., Morgan, M., Muthukrishnan, S. and Shatkin, A. J. (1975) Proc.
Natl. Acad. Sci. USA, 72, 362 - 366.
3. Wei, C. M. and Moss, B. (1974) Proc. Natl. Acad. Sci. USA, 71, 3014-3018.
4. Rottman, F., Shatkin, A. and Perry, R. P. (1974) Cell, 3, 197-199.
5. Edmonds, M., Vaughan, M. H., Jr., and Nakazoto, H. (1971) Proc. Natl. Acad.
Sci. USA , 68, 1336- 1340.
6. Lee, S. Y., Mendecki, J. and Brawerman, G. (1971) Proc. Natl. Acad. Sci. USA,
68, 1331- 1335.
7. Darnell, J. E., Jr., Wall, R. and Tushinski, R. J. (1971) Proc. Natl. Acad. Sci.
USA , 68, 1321- 1325.
8. Pettersson, U., Mulder, C., Delius, H. and Sharp, P. A. (1973) Proc. Natl. Acad.
Sci. US A, 70, 200 - 204.
9. Sharp, P. A., Gallimore, P. H. and Flint, S. J. (1974) Cold Spring Harbor Symp.
Quant. Biol., 34, 457 - 474.
10. Flint, S. J., Gallimore, P. H., and Sharp, P. A. (1975)J. Mol. Biol., 96, 47 - 68.
11. Flint, S. J. and Sharp, P. A. (1976)J. Mol. Biol., 106, 749-771.
12. Bachenheimer, S. and Darnell, J. E. (1975) Proc. Natl. Acad. Sci. USA, 72,
4445 - 4449.
13. Berget, S. M., Moore, C. and Sharp, P. A. (1977) Proc. Natl. Acad. Sci. USA, 74,
3171-3175.
14. Lewis, J., Atkins, J. F., Anderson, C., Baum, P. R. and Gesteland, R. F. (1975)
Proc. Natl. Acad. Sci. USA, 72, 4445-4449.
15. Thomas, M., White, R. L. and Davis, R. W. (1976) Proc. Natl. Acad. Sci. USA,
73,2294 - 2298.
16. Casey, J. and Davidson, N. (1977) Nucleic Acid Res., 4, 1539-1552.
17. Westphal, H., Meyer, J. and Maizel, J. (1976) Proc. Natl. Acad. Sci. USA, 73,

2069-2071.
18. Chow, L. T., Roberts, J. M., Lewis, J. B. and Broker, T. R. (1977) Cell, 11,
819-836.
19. Goldberg, S., Weber, J. and Darnell, J. E., Jr. (1977) Cell, 10, 617-622.
20. Weber, J., Jelinek, W. and Darnell, J. E., Jr. (1977) Cell, 10, 611-616.
21. Berget, S. M. and Sharp, P. A. (1979)J. Mol. Biol., 129, 547-565.
22. Perry, R. P. and Kelley, D. E. (1976) Cell, 8, 433 -442.
23. Gelinas, R. E. and Roberts, R. J. (1977) Cell, 11, 533-544.
24. Jeffreys, A. J. and Flavell, R. A. (1977) Cell, 12, 1097- 1108.
25. Tilghman, S. M., Tiemeier, D. C., Seidman, J. G., Peterlin, B. M., Sullivan, M.,
Maizel, J. V. and Leder, P. (1978) Proc. Natl. Acad. Sci. USA, 75, 725 - 729.
26. Breathnach, R., Manel, J-L. and Chambon, P. (1977): Nature, 270, 314-319.
27. Tonegawa, S., Maxam, A. M., Tizard, R., Bernard, 0. and Gilbert, W. (1978)
Proc. Natl. Acad. Sci. USA, 75, 1485 - 1489.
28. Breathnach, R. and Chambon, P. (1981) Annu. Rev. Biochem., 50, 349-384.
29. Padgett, R. A., Grabowski, P. J., Konarska, M. M., Seiler, S. and Sharp, P. A.
(1986) Annu . Rev. Biochem., 5 5, 1119-1150.

170

Physiology or Medicine 1993

30. Fink, G. R. (1987) Cell, 49, 5 - 6.

31. Treisman, R, Orkin, S. and Maniatis, T. (1983) Nature, 302, 591-596.
32. Nigro, J. M., Cho, K. R., Fearson, E. R., Kern, S. E., Ruppert, J. M., Oliver, J.
D., Kinzler, D. W. and Vogelstein, B. (1991) Cell, 64, 607-613.
33. Cocquerelle, C., Daubersies, P., Majerus, M.-A., Kerckaert, J.-P. and Bailleul,
B. (1992) EMBO J., 11, 1 0 9 5 - 1 0 9 8 .
34. Pulok, R. and Anderson, P. (1993) Genes Dev. , 7, 1885- 1897.
35. Patel, R. S., Odermatt, E., Schwarzbauer, J. E. and Hynes, R. 0. (1986) EMBO
J., 6 , 2 5 6 5 - 2 5 7 2 .
36. Hynes, R. 0. (1989) Fibronectin , Springer-Verlag, New York.
37. Baker, B. S. (1989) Nature , 340, 521-524.
38. Tian, M. and Maniatis, T. (1992) Science , 256, 237-240.
39. Gilbert, W. (1978) Nature , 271, 501.
40. Darnell, J. E., Jr. (1978) Science , 202, 1257.
41. Doolittle, W. F. (1978) Nature, 272, 581.
42. Blake, C. C. F. (1978) Nature, 273, 267.
43. Stone, E. M. and Schwartz, R. J., editors (1990) Intervening Sequences in
Evolution and Development , Oxford University Press, New York.
44. Lonberg, N. and Gilbert, W. (1985) Cell , 40, 81-90.
45. Agabian, N. (1990) Cell, 61, 1157-1160.
46. Borst, P. (1986) Ann. Rev. Riochem., 55, 701-732.
47. Krause, M. and Hirsh, D. (1987) Cell, 49, 753-761.
48. Vellard, M., Sureau, A., Soret, J., Martinerie, C. and Perbal, B. (1992) Proc.
Natl. Acad. Sci. USA, 89, 25 1 l - 2 5 1 5 .
49. Benne, R., van den Burg, J., Branhenhoff,J. P. J., Sloof, P., van Boom, J. H.
and Tramp, M. C. (1986) Cell , 46, 819-826.
50. Padgett, R. A., Hardy, S. F. and Sharp, P. A. (1983) Proc. Natl. Acad. Sci. USA,
80,523O - 5234.
51. Green, M. R., Maniatis, T. and Melton, D. A. (1983) Cell, 32, 681-694.
52. Hernandez, N. and Keller, W. (1983) Cell, 35, 89-99.
53. Grabowski, P. J., Padgett, R. A. and Sharp, P. A. (1984) Cell , 37, 415-427.
54. Krainer, A. R., Maniatis, T., Ruskin, B. and Green, M. R. (1984) Cell, 36,
993 - 1005.
55. Konarska, M. M., Grabowski, P. J., Padgett, R. A. and Sharp, P. A. (1985)
Nature, 313, 552-557.
56. Wallace, J. C. and Edmonds, M. (1983) Proc. Natl. Acad. Sci. USA, 80, 950954.
57. Grabowski, P. J., Seiler, S. R. and Sharp, P. A. (1985) Cell, 42, 345 -353.
58. Brody, E. and Abelson, J. (1985) Science, 228, 963-967.
59. Steitz, J. A., Black, D. L., Gerke, V., Parker, K. A., Kramer, A., Frendeway,
D.,and Keller, W. (1988) In Structure and function of major and minor small
nuclear ribonucleoprotein particle s (ed. M. L. Birnstiel), pp. 115 - 154, SpringerVerlag, New York.
60. Cech, T. R. (1985) Cell, 43, 713-716.
61. Peebles, C. L., Perlman, P. S., Mecklenburg, K. L., Petrillo, M. L., Tabor, J.
H., Jarrell, K. A. and Cheng, H.-L. (1986) Cell, 44, 213-223.
62. van der Veen, R., Arnberg, A. C., van der Horst, G., Bonen, L., Tabak, H. F.
and Grivell, L. A. (1986) Cell, 44, 225-234.
63. Sharp, P. A. (1991) Science, 254, 663.
64. Hannon, G. J., Maroney, P. A., Denker, J. A. and Nilsen, T. W. (1990) Cell, 61,
1247-1255.
65. Watkins, K. R., Dungan, J. M. and Agabian, N. (1993) Cell , in press.
66. Lerner, M. R., Boyle, J. A., Mount, S. M., Wolin, S. L. and Steitz, J. A. (1980)
Nature, 283, 220- 224.
67. Rogers, J. and Wall, R. (1980) Proc. Natl. Acad. Sci. USA, 77. 1877- 1879.

Phillip A. Sharp

171

68. Padgett, K. A., Mount, S. M., Steitz, J. A. and Sharp, P. A. (1983) Cell, 35,
101 - 107.
69. Parker, R., Siliciano, P. G. and Guthrie, C. (1987) Cell, 49, 229-239.
70. Guthrie, C. and Patterson, B. (1988) Annu. Rev. Genet., 22, 387-419.
71. Konarska, M. M. and Sharp, P. A. (1987) Cell, 49, 763-774.
72. Madhani, H. D. and Guthrie, C. (1992) Cell, 71, 803-817.
73. McPheeters, D. S. and Abelson, J. (1992) Cell, 71, 819-831.
74. Sawa, H. and Abelson, J. (1992) Proc. Natl. Acad. Sci. USA, 89, 11269- 11273.
75. Wassarman, D. A. and Steitz, J. A. (1992) Science, 257, 1918- 1925.
76. Newman, A. and Norman, C. (1992) Cell, 68, 743-754.
77. Moore, M. J., Query, C. C. and Sharp, P. A. (1993) In The RNA World, Cold
Spring Harbor Laboratory Press, pp. 303 - 357.
78. Seraphin, B. and Rosbash, M. (1991) EMBO J., 10, 1209-1216.
79. Michaud, S. and Reed, R. (1991). Genes Dev., 5, 2534-2546.
80. Jamison, S. F. and Garcia-Blanco, M. A. (1992) Proc. Natl. Acad. Sci. USA, 89,
5482 - 5486.
81. Ruskin, B., Zamore, P. D. and Green, M. R. (1988) Cell, 52, 207-219.
82. Ruby, S. W. and Abelson, J. (1988) Trends Genet., 7, 79-85.
83. Guthrie, C. (1991) Science, 253, 157-163.
84. McSwiggen, J. A. and Cech, T. R. (1989) Science, 244, 679-683.
85. Rajagopal, J., Doudna, J. A. and Szostak, J. W. (1989) Science, 244, 692-694.
86. Suh, E.-R. and Waring, R. B. (1992) Nucleic Acids Res., 20, 6303 -6309.
87. Moore, M. J. and Sharp, P. A. (1993) Nature, 365, 364- 368.
88. Moore, M. J. and Sharp, P. A. (1992) Science, 256, 992-997.
89. Beyer, A. L. and Osheim, Y. N. (1988) Genes Dev., 2, 754- 765.
90. Mehlin, H., Daneholt, B. and Skoglund, U. (1992) Cell, 69, 605-613.
91. Spector, D. L. (1993) Annu. Rev. Cell Biol., 9, 265-315.
92. Nyman, U., Hallman, I-I., Hadlaczky, G., Pettersson, I., Sharp, G. and Ringertz
N. R. (1986) J. Cell. Biol., 102, 137- 144.
93. Carter, K. C., Bowman, D., Carrington, W., Fogarty, K., McNeil, J. A., Fay, F.
S. and Lawrence,J. B. (1993) Science, 259, 1330 - 1334.
94. Xing, Y., Johnson, C. V., Dobner, P. R. and Lawrence, J. B. (1993) Science 259,
1326-1330.
95. Ge, H. and Manley, J. L. (1990) Cell, 62, 25-34.
96. Krainer, A. R., Conway, G. C. and Kozak, D. (1990b) Cell, 62, 35 -42.
97. Fu, X.-D. and Maniatis, T. (1990) Nature, 343, 437-441.
98. Zeitlin, S., Wilson, R. C. and Efstratiadis, A. (1989) J. Cell. Biol., 108, 765777.
99. Roth, M. B., Murphy, C. and Gall, J. G. (199O)J. Cell. Biol., 111, 2217-2223.
100. Roth, M. B., Zahler, A. M. and Stolk, J. A. (1991)J. Cell. Biol., 115, 587-596.
101. Blencowe, B. J., Nickerson, J. A., Issner, R., Penman, S. and Sharp, P. A.
(1993) in preparation.