TREE SHAKING AND JELLY MAKING: GROWING UP WITH RETROVIRUSES Bee 11 T9R5
UCSF LISRARY
LIBRARIAN'S OFFICE

D ECSIVE TM
RECEIVE f}

i

to be given April 4, 1985, as Faculty Research Lecture

» cOlleagues, friends:

One thing I have learned from the experience of preparing this
lecture is the impact of a mystifying title.

We shall come a little later to the origins of the first half of
the title, but I want to speak initially about the idea of growing up
with and living within a scholarly field. One of the novels that
crossed my nightstand in the past year---Disturbances in the Field by
Lynn Sharon Schwartz---uses its characters to address explicitly the
problems of individual initiative and emotional response within a
larger and more powerful network of social behavior. There are
similar tensions in our lives as scientists, between the field in
which we work---a social and intellectual maelstrom with a life of its
own---and our individual aspirations to send out on occasion some
positive signal. This dichotomy will be apparent in my talk. I
will attempt first to outline broad currents in the study of
retroviruses and then give you a partial account of one route I have
taken within this network of people and ideas. As you will see,
Disturbances in the Field, the title of Lynn Schwartz’! novel, could
have worked quite nicely for me.

It will be helpful later if a spend a moment now to tell you how
I got involved in this field. Two factors were decisive. One was

night school at the National Institutes of Health, where medically-
obstacle to the acceptance of the provirus hypothesis. Second, Steve
Martin had isolated a temperature-sensitive mutant of Rous sarcoma
virus, a mutant that clearly distinguished between functions required
for multiplication of retroviruses and a function required for its
neoplastic effects. It was apparent to many that a genetic and
molecular approach to RNA tumor viruses was now possible. Money
flowed, and people too.

(lights down, slide on) During the subsequent decade and a half,
two questions have dominated the thoughts of most of us working with
retroviruses: How do they reproduce? and How do they cause tumors? As
I hope my narrative today will make clear, the answers are
intertwined, and they have raised further questions that were
inconceivable when we began.

(next slide: exog. retros) Most retroviruses are external agents
that invade animal cells. In this sense they resemble viruses of many
other kinds that infect cells in all known biological kingdoms.
Invasion is followed by multiplication of the virus, but with
important differences that distinguish retroviruses from most other
viruses. Retroviruses rarely kill their host cells; instead viral
genes (shown in red throughout most of the slides) are perpetuated in
the infected cells, and the cells may appear either unchanged or given
a growth advantage as a result of neoplastic transformation, which
also affects the appearance of the cells. For over half a century,
tumor virology hobbled along, working mainly with a few exogenous
retroviruses of birds and mice, principally those viruses that will
recur frequently in our discussion today: the leukosis and Rous

Sarcoma viruses of chickens and the mammary tumor and leukemia viruses
of mice. But we now know retroviruses to be widely dispersed in
nature (first colored tree), found certainly in fish and snakes (and
perhaps in non-vertebrates), as well as in several birds, rodents,
ungulates, and primates. Among the most recent and best known
exogenous retroviruses are the apparent causative agents of two human
diseases, adult T cell leukemia and AIDS.

(next slide: info diagram) The biochemical principle that unites
all retroviruses was at one time considered an eccentric violation of
the usual flow of genetic information from DNA to RNA to protein.
Retroviruses encode enzymes, among them reverse transcriptase, that
convert their genes, carried about in an RNA form, to a more stable
DNA form, the provirus ----a form that can be used by host cell
machinery normally entrusted with the job of duplicating and
expressing the cell's own genetic legacy. In the past few years,
retroviruses have come to seem much less odd. Several other kinds of
viruses and genetic elements have been shown or strongly suspected of
transferring their information through an RNA form and back to DNA, as
a means of duplicating the information and moving it within or between
cells. (next fruiting tree: retro elements) Among these are viruses
that carry their genes in a DNA form but convert them transiently to
RNA during virus growth: the several hepatitis B viruses of birds,
rodents, and man and the cauliflower mosaic virus. Other "retro
elements" are components of cell chromosomes that have been generated
by intracellular events involving reverse transcription. These
include a number of mobile genetic elements of yeast, flies, and
vertebrates to which I will return; several kinds of repeated

sequences; and apparently worthless copies of genes.
The connections that I will emphasize today between retroviruses
and certain mobile components of chromosomes were presaged almost
twenty years ago by evidence that retroviral genes could be
transmitted through the germ line, (next slide: genetic transmission)
in a form we now recognize---according to rules I'll discuss later---
to be endogenous proviruses. Such proviruses are components of normal
chromosomes, passed on to future progeny, and presumed to be the
consequences of infection of germinal cells in the recent or distant
past. In somatic cells, endogenous proviruses may be completely
silent or they may be expressed. Some can generate fully infectious
viruses, and the endogenous viruses may even be tumorigenic. Most
of these proviruses, however, appear to be functionally incompetent.
Though first studied in mice and chickens, inherited, provirus-like
elements are now known to be abundant in virtually all eukaryotes
(next fruited tree), present in yeast, flies, and all vertebrates
including man. There are, at a minimum, a few thousand such elements
in our chromosomes, approaching one percent of our total DNA. What
are they doing there? It is anyone's guess. The pundits offer many
possibilities: the elements may hasten evolutionary change and hence
be good for us; they may cause considerable damage but be tolerated
if they do not; or their existence may be largely unrelated to effects
upon their hosts, and instead reflect only the elements' selfish
interest in survival.

(next slide: expression) Whether exogenous or endogenous,
retroviruses exploit mechanisms provided by their host cells and
contribute their own idiosyncratic mechanisms to produce viral RNA,

protein, and ultimately infectious particles from proviral DNA.
Consequently, (next fruited tree) retroviruses have been important
reagents in efforts to learn how eukaryotic cells regulate their
genes---through synthesis of RNA, with promoters, enhancers, hormonal
regulators, and splicing; through synthesis and modification of
proteins; and through coordinated assembly and release of virus
Particles competent to infect a new host cell efficiently.

A closer look at retroviruses reveals an extraordinary potential
for complex genetic interaction with each other and with their hosts.
(next slide) Retroviruses carry two RNA copies of their genes~-~hence,
like us, but unlike other viruses, they are diploid. When two
distinguishable retroviruses infect the same cells a number of
consequences are possible: heterozygotes, with two different sets of
genes; recombinants, with new combinations of genes; or pseudotypes,
with the proteins of one virus wrapping up the genes of another. As
we shall explore in greater detail later, the interaction between
viral and host genes is even more remarkable, with proviral DNA
covalently joined to the host chromosome. This may lead to mutation
of host genes and occasionally to incorporation of host genes into
progeny viruses, a process called transduction. (next fruiting tree)
These phenomena (and others) provide the principles for a new phase
of retrovirology, in which retroviruses are modified by design for use
as genetic vectors. Within the next year or two, we will likely see
the first attempts to cure certain inherited human diseases with such
vectors.

(next slide: transduction) The most potent of the oncogenic
retroviruses are naturally-occuring genetic vectors, having captured

host genes during passage of wild type retroviruses through cells.
The transduced genes are responsible for the neoplastic activities of

these viruses and are known as viral oncogenes or v-onc's; their

cellular progenitors are called c-onc's or proto-oncogenes. Following
the elucidation of these principles with the sre gene of Rous sarcoma
virus a decade ago, (next fruited tree) sre has been joined by about
twenty other genes, some closely related and some not, some first
encountered in other highly oncogenic viruses of birds, some in
viruses of mice, rats, and cats, one in a virus of monkeys. Each of
these viral oncogenes is derived from a cellular proto-oncogene,
genes that are carefully conserved in evolution---and hence found in
all vertebrates and sometimes in insects, sponges, yeast, and other
organisms. These cellular progenitors of viral oncogenes are in one
sense misnamed: (next slide) under normal conditions, proto-oncogenes
appear to supply functions essential for the growth, development, and
function of cells. They provide extracellular factors that signal
cells to divide; cell membrane proteins that serve as receptors for
such growth factors and possess protein kinase activity specific for
tyrosine residues in target proteins; other kinases, some also
specific for tyrosine, but without known physiological function:
GTPases that are presumptive modulators of external signals for
growth; and nuclear proteins that may regulate the expression of
other genes. (next slide) On the other hand, in many cancers of man
and animals, proto-oncogenes fulfill the promise of their name, having
been converted to apparently active oncogenes by a variety of
mutational mechanisms, including gene amplification, chromosomal
translocations, nearby insertion of mobile elements, rearrangements

that truncate genes, or simple point mutations. Naturally, these
observations have been the seeds for additional trees (next Slide),
one of which is illustrated by the kinase family of oncogenes and
proto-oncogenes.

(composite tree slide) I have tried to convince you that the
orchard of retrovirology has flowered and fruited marvellously over
the past fifteen years, nourished in some collective sense by those of
us who work in it. As we turn to consider individual contributions to
the discipline, I find myself shifting metaphors slightly, thinking
more about harvesting than nurturing, but wondering what determines
how each of us fares in this garden. To what extent does a good crop
signify character, luck, insight, or a transient stage of life or
career? (slide off; lights)

It was a coincidence of vague thoughts of this sort with a small
item in last November's newpaper that provided the mysterious title of
today's talk. Following the defeat of the Democrats in last year's
election, some of Jesse Jackson's supporters were asked what he
planned to do with the public platform he had built during the
campaign. "We want him to be a tree-shaker, not a jelly-maker," one
of them replied. The reporter went on to explain that Jackson was
expected to speak and act in ways that knocked issues out of the trees
So they could be seen and acted upon; the pitting, boiling, sugaring,
and bottling (read litigation, legislation, enforcement) would be left
to others less imaginative.

These are concerns that I suspect many of us have about our
work. We all seek those startling results that shift paradigms and
found new disciplines. But is it possble to make a meaningful

distinction between tree-shaking and jelly-making in daily laboratory
life? Is it possible to start each day with a revolutionary plan? Or
are we just as likely to meet with thunderous success by hewing
closely to Pasteur's dictum, "Chance and the prepared mind"? As we
grow with a discipline, we become attached to our own findings and our
own reagents---cell lines, mutants, techniques ---often to the point
at which it is legitible to wonder, "Would I be trying to solve the
problem at the heart of my research efforts in some better way if I
now entered it anew and attacked it without the baggage of the past? "

I don't see simple answers to these questions. I have had the
good luck to be associated some years ago with one episode of what now
seems to be a candidate for tree-shaking: the discovery that
retroviral oncogenes are derived from normal cellular genes. Yet even
that discovery was not instanteously perceived as tree-shaking by
us---and certainly not by our audience---until a lot of jelly-making,
and even some additional tree-shaking, had occurred in a large number
of labs.

I would like to explore these considerations further by telling a
story that begins with a simple and potentially specialized (i.e.
jelly-making) question about how retroviruses reproduce. The tale
leads first to a description of proviral DNA that links retroviruses
with other kinds of mobile genetic elements; it then provides the
first evidence that cellular proto-oncogenes can become true oncogenes
without conversion to viral oncogenes; and it ends with the discovery
of an novel oncogene implicated in mammary carcinomas of mice.

By 1975, the provirus hypothesis was firmly established. (lights
down: slide) Work here and elsewhere had shown that viral DNA is made

in the cytoplasm of retrovirus-infected cells by reverse transcriptase
and later integrataed covalently into host chromosomes, where it then
serves as a template for synthesis of viral RNA. Ram Guntaka had
recently discovered here that retroviral DNA is present in a circular
as well as a double stranded linear form early after infection. The
model shown was proposed at that point for early events in the virus
life cycle, showing linear DNA, approximately the size of its viral
RNA template, as precursor to circular DNA, circular DNA integrating
into host DNA, and proviral DNA being transcribed into viral RNA.
Attention was focused upon the organization of viral DNA, the manner
in which it was synthesized and integrated, and the positions in host
chromosomes at which integration occurred.

This scheme was correct in general. But thinking about some
crucial details was troubled by the following dilemma. (next slide)
Several laboratories had examined the initial steps in reverse
transcription by working with the viral enzyme in vitro. The primer
for DNA synthesis was found to be a host tRNA bound to viral RNA near
its 5' end. Synthesis of DNA could proceed for only a very short way,
about 1% of the RNA template, before something unusual had to occur.
Because viral RNA was shown to be terminally redundant---that is, to
have exactly the same short sequence, called R, at each end---and
because reverse transcriptase is associated with another enzymatic
activity that can remove RNA from an RNA:DNA hybrid, it was proposed
(and later proven) that the DNA copy of R (called R') is exposed to
form a hybrid with the other (3') end of viral RNA. This allows the
nascent DNA to prime synthesis along the rest of viral RNA. While it
solves the problem of extending the DNA strand, this manuever creates

another problem: namely, a copy of R has apparently been sacrificed.
(slide) At first glance this would be disasterous for the virus
life cycle: copies of R must be present at both ends of proviral DNA
to provide an appropriate template for synthesis of viral RNA. A copy
of R must be regained, but when and how? It is sobering to reveal
to you the answers that I proposed at the time to my colleagues in
public places. We then called the R sequence X, and the schemes
included (slide) integration at a host site that contained a second
copy of X, (slide) staggered cutting of circular DNA to duplicate X,
and (slide) tandemly integrated proviruses, each providing a single
copy of X. Each of these seems reasonably clever and would solve the
problem during the integration step; but all are wrong and
essentially less interesting than the real answer.

Finding the answer was technically demanding---remember these not-
so-long ago days preceded molecular cloning of retroviral DNA---and it
represented one of the first applications of Southern's DNA transfer
procedure to eukaryotic sequences present in very few copies per
cell.(slide) The successful strategies made use of molecular probes
specific for the region of viral RNA positioned between the tRNA
primer and R---a sequence known hereafter as U5---and for a region
situated near the other copy of R (and known as U3). We were also
helped by recognition signals for certain restriction enzymes in these
regions (e.g. an Eco RI site in U3).

The picture of unintegrated linear DNA that emerged from this work
(performed by Peter Shank, Steve Hughes, and HJ Kung) was
extraordinary and unexpected (slide): both U3 and US (and, by
implication, a copy of R) were present at both ends of linear DNA.

This meant that the copy of R was restored during DNA synthesis rather
than during integration; that sequences present only once in viral RNA
(i.e. U3 and U5) were present twice in DNA; that the arrangement of
these sequences created terminal repeats in the DNA that were
substantially longer than the terminal repeat in the RNA.

Furthermore, creation of these long repeats required a second
acrobatic event during DNA synthesis, one resembling the first jump
between templates described earlier. (This event, the second jump, is
now understood but a distraction from the story I want to tell.)

(next slide) Linear DNA was then shown to be converted to two
forms of circular DNA, a form with one copy of the long terminal
repeat (or LTR) and one with two copies that was recently proven to be
the integrating species. Integrated proviral DNA itself was found to
resemble the linear DNA, save for the fact that it was joined to host
DNA at both ends, with each provirus linked to a different part of the
host genome. According to this strategy, proviral DNA was
beautifully equipped to be a template for production of viral RNA---
the R sequence was present near both ends, and there was additional]
sequence that might encode regulatory signals, such as transcriptional
promoters, and would reside upstream of the appropriate initiation
site. The results also implied that integration of viral DNA was
quite precise, always joining the same region of viral DNA to
cellular DNA, but random or quasi-random with respect to the cellular
site at which integration occurs.

With later application of molecular cloning and nucleotide
sequencing techniques, each of these deductions was confirmed and
additional symmetries were perceived. (next slide) First, the LTRs

were found to have internal symmetries, with their terminal
nucleotides arranged as inversions of the same sequence. Second, the
site in viral DNA at which linkage to host DNA occurs is precise to
the nucleotide, with exactly two base pairs lost from each LTR. (This
loss is tolerated because it involves a region that is not transcribed
into RNA.) Third, though the host integration sites are always
different for each provirus, a few nucleotides at the insertion site
are precisely duplicated to flank the provirus.

Finally, (next slide: LTR anatomy) both structural and functional
studies show that the LTR is truly the hub of the proviral universe---
defined by initiation sites for synthesis of both DNA strands,
containing the sites for integration, and equipped with signals to
start RNA at the upstream end of a provirus and to end it at the
downstream end.

In mid-1978, just after LTRs had come into view, we moved to
London to spend a sabbatical at the Imperial Cancer Research Fund and
in a Georgian row house in Islington (slide). Having plenty of time
to ruminate, I began to think seriously about the possibility that
integration of retroviral DNA would, in effect, mutate the host
chromosome. (next slide: structural homologies) This idea had
actually been self-evident from the time proviruses were first
proposed to integrate into host DNA, but it took added force from
description of the LTRs. It was now apparent that the proviruses of
retroviruses---here represented by the two we had studied, Rous
sarcoma and mouse mammary tumor virus---were affiliated, at least on
structural grounds, with mobile genetic elements of bacteria, yeast,
and Drosophila. (The latter, as you have already heard, are also

functionally related to retroviral proviruses, since they pass through
RNA intermediates.) In Beneral, when these elements move (or
transpose) they enter host chromosomes promiscuously; (next slide:
insertion muts) hence they are prone to make functionally significant
mutations, either by disrupting and inactivating a normal gene or by
augmenting its expression from a nearby insertion site.

I decided to show directly that retroviruses can act as mutagens by
interrupting a pre-existing gene in a cellular chromosome. I chose to
use aS a genetic target for an invading provirus not a cellular gene,
but instead (next slide) the single Rous sarcoma provirus whose sre
gene and protein product had been responsible for neoplastic
transformation of a normal rat cell, producing the line called B31.
Superinfection of B31 cells with another retrovirus---a murine
leukemia virus that lacks an oncogene and does not transform cells in
culture---should revert the behavior of B3l cells to normal in those
rare cells, roughly one ina million, whose Rous provirus is
interrupted by a newly acquired murine leukemia provirus. Moreover,
if the latter provirus were then excised, the cells might return to
the transformed state. (next slide: ins1,2) The expected mutants
were indeed rare, but Suzanne Ortiz and I were able to isolate two of
them, and they exhibited the predicted properties. In both cases, an
MLV provirus was stuck in the middle of the RSV provirus---
interestingly, not in the coding region for sre but in the middle of a
region normally removed by splicing to produce sre messenger RNA. As
a result, little or no sre mRNA or protein was found in the mutant
cells and the cells lost their transformed phenotype. In one of the
two lines, most of the MLV provirus was eliminated at low frequency,

by recombination between the LTRs. This rare event was readily
detectable, however, because it restored expression of the src gene
and the transformed phenotype. In short, retroviruses were proven
competent to make insertion mutations that inactivate genes, at about
the frequency expected for a randomly integrating element. ( This
conclusion has now been corroborated by other laboratories, in both
germinal and somatic cells, using true host genes as targets.)

But what about insertion mutations that activate cellular genes?
About the time of my sabbatical, Nancy Quintrell performed an
experiment that I now like to describe as heuristic. (slide of NQ's
expt) In surveys of viral DNA and RNA in Rous sarcoma virus
transformed mammalian cells, she happened upon one cell line in which
the provirus appeared to be promoting transcription of adjacent
cellular DNA, since a novel messenger RNA was shown to contain host
sequences linked to sequences---R and U5---from the viral LTR. No
matter that the conclusion shown here is inaccurate in certain
details, or that the phenomenon had no apparent consequences for this
cell. Its effect was instead upon our thinking about another
situation: (slide) how does a retrovirus that lacks an oncogene manage
to induce a tumor?

I indicated earlier that not all oncogenic retroviruses carry
viral oncogenes---genes derived from cellular proto-oncogenes. Those
that don't, tend to produce tumors Sluggishly and are incompetent to
transform cultured cells. Since these viruses are vigorously
infectious, it seemed plausible to ask whether their LTRs might drive
inappropriate expression of some proto-oncogene that happened to
reside near (presumably downstream from) an inserted provirus. Even
if this event occurred extremely rarely, it should be detectable when

the affected cel] grew clonally to form a tumor.
Greg Payne, then a graduate student in our group, considered this
possibility in his work with the avian leukosis virus--~ a chicken
retrovirus devoid of a viral oncogene but still able to produce
tumors, most commonly lymphomas in the B cell organ of the chicken,
the bursa of Fabricius. (slide: models and LL1) Some tumors did in
fact contain species of RNA similar to that Nancy had described, with
LTR sequences linked to unidentified host sequences. Moreover, such
tumors--~-for example, tumor LL1--- usua)ly had only a single provirus
and the provirus was often aberrant, so that it was not possible to
express viral genes. This suggested the viral genes used for virus
replication were not important for tumor growth, but whatever
Sequences resided downstream from the provirus might be.

It was obviously critical to know what those sequences were.
Physical mapping with restriction enzymes, both by Greg and in
parallel work by Hayward, Astrin, and their colleagues, suggested that
the activated host sequences might be the same in different tumors.
At this point Hayward caused what I would now call "a disturbance in
the field." ( I still lose sleep over this one, though I assume
Hayward does not.) He used available probes for several retroviral
oncogenes, and found (slide) that the vast majority of bursal
lymphomas contained ALV proviruses adjacent to a now efficiently
transcribed cellular mye gene, the proto-oncogene previously
discovered by Mike and Diana Sheiness when they traced the cellular
origins of the v-mye oncogene of a highly oncogenic avian
retroviruses.

Here was a tree-shaker: for the first time, cellular proto-

oncogenes appeared to deserve their perjorative name after all: they
could be activated in their original settings and thereby contribute
to neoplasia. A few of you have heard me speak previously about the
musicality of discovery. For example, the finding that viral oncogenes
are derived from cellular genes seems Wagnerian in texture: the piece
draws much of its strength from repeated restatement of the theme.
Hayward's experiment was Mozartian in character: a single,
transporting phrase, one that seemed effortless but suggested infinite
possibilities. The physical execution of his experiment was trivial;
we were able to confirm the results within days of his phone call.
Why didn't we do it earlier? It will cost you a beer to find out.

When we regained our composure, we realized that our collection of
tumors differed in an interesting way from Hayward's: (next slide:
diagram of myc) all of his tumors contained proviruses positioned so
that an LTR could provide a promoter for c-myc. However, several of
our tumors had proviruses arranged differently within the c-myc locus.
In such cases, with the provirus upstream from c-myc but in the
opposite transcriptional orientation or with the provirus downstream
from c-myc, another property of the provirus must be responsible for
augmented expression. This was one of the first pieces of evidence
that proviruses contained transcriptional enhancers, sequences that
improve the performance of adjacent promoters over relatively large
distances and in a fashion independent of orientation.

Further study of the mutated c-mye loci by Greg and Dave Westaway
showed that the insertion of ALV DNA was not the sole lesion: usually
the proviruses were damaged by deletions and sometimes the coding
sequence of c-myc was also altered. These secondary mutations are

consistent with clinical opinion that the development of a cancer
requires multiple rare events, reinforcing our confidence that tumor
induction by viruses without oncogenes is a credible model for human
cancers.

This episode with bursal lymphomas has had a dramatic effect upon
many aspects of retrovirology and oncology. (slide: rearrangements)
The most far-ranging influence of insertion mutations in the c-myc
locus has been to serve as harbingers of other kinds of rearrangements
of DNA, alterations of chromosomal architecture that affect other
proto-oncogenes as well as c-myc, in both human and experimental
tumors. This slide illustrates some well known examples ---
chromosomal translocations in which c-myc genes in Burkitt lymphomas
are fused to immunoglobulin genes on other chromosomes, and localized
amplifications of cellular DNA that augment the number of copies of c-
myc genes and the abundance of the gene products.

Proviral insertions in c-myc also stimulated our thinking about
the mechanism by which retroviruses capture host proto-oncogenes and
convert them to viral oncogenes. (slide) As shown in this widely-
accepted scheme, whose principal architect was Ron Swanstrom,
transduction is proposed to begin with a proviral insertion mutation
upstream from a proto-oncogene and in the same transcriptional
orientation. I like to think that this event initiates uncontrolled
cell growth, expanding the number of cells in which the subsequent
rare events might occur. A deletion mutation then removes the
downstream part of the provirus, making a hybrid transcriptional unit.
After splicing, messenger RNA from such units would have one end that
resembles normal retroviral RNA and includes signals for packaging

into virus particles. The other end contains some of the exons from
the cellular proto-oncogene. If another normal provirus is in the
same cell, heterozygote genomes can form, and recombination is likely
to follow. The final product is an RNA molecule with all the features
required for replication as a retroviral genome, provided that viral
proteins are supplied by an accompanying helper virus. This is the
case for the vast majority of retroviruses bearing oncogenes. Several
factors may contribute to the physiological differences between an
oncogenic v-one and a benign c-onc: sustained and efficient expression
from viral LTRs, truncations of the coding regions, and more subtle
nucleotide sequence changes that occur during virus multiplication.

Let's pause briefly here to review where we have been. You have
seen how proviruses were found to resemble transposable elements, and
I have shown you examples of three sorts of proviral insertion
mutations: (slide: IM1) mutations that inactivate genes, (slide:
IM2)mutations that activate genes by promoter insertion and could
represent the first step in the transduction of oncogenes, and (slide:
IM3) mutations that activate genes by an enhancer mechanism. One of
the rewards of mutations should be the identification and isolation of
new genes, something I have yet to describe. This brings me to the
final stage of today's story.

(slide of C3H mouse) Our group has been interested since the
early 1970's in the mammary carcinomas caused by the mouse mammary
tumor virus; (slide: summary) MMTV is a retrovirus commonly
transmitted through milk to nursing mice. Like avian leukosis virus,
it lacks a viral oncogene, and induces clonal neoplastic growths that
contain new proviruses, suggesting that an integration event may be

important in carcinogenesis. Finally, the MMTV LTR harbors an
unusual promoter with glucocorticoid responsiveness, suggesting that
an atypical regulatory phenomenon may also be important.

In 1981, Roel Nusse and I set out (slide: bicyclists) to ask
whether MMTV proviruses were causing insertion mutations in these
tumors (slide: ?) and, if so, to isolate the affected proto-
oncogene~-~ hoping that we would be led to a novel gene, not back to
c-myc or some other known proto-oncogene. In the wake of Hayward's
results, the latter would only be interesting jelly-making.

Our strategy was similar to one that had been recently used to
isolate an eye color gene from Drosophila melanogaster. The mobile
genetic element, copia, had caused a mutation in the white locus; by
cloning that copy of the copia element, it was possible to gain a
foothold in the gene, because a piece of it was adjacent to copia in
the molecular clone. (slide) To practice this art, now known as
transposon tagging, Roel sought and found a mouse mammary tumor
bearing only a single new MMTV provirus. According to the hypothesis,
this provirus should be adjacent to an activated proto-oncogene. If
tumorigenesis depends upon the rare insertion of viral DNA near this
proto-oncogene, other tumors should also have proviruses somewhere in
the vicinity. To test these predictions, part of the provirus was
cloned from the initial tumor with flanking host DNA still attached;
the host sequences were then separated from viral DNA and used asa
probe to ask two questions: whether the integration site was
interrupted in other tumors as a result of MMTV insertions, and
whether the insertions altered expression of any genes in the region.

(slide: int insertions) This diagram summarizes the conclusions

drawn from many experiments. About three-fourths of mammary
carcinomas in C3H mice have MMTV proviruses inserted within a 20--30 kb
domain, roughly equally distributed on both sides of a gene that we
now call int-1. Though silent in normal mammary tissue, the gene is
expressed when an MMTV provirus is nearby. You will notice that the
proviruses are not positioned to allow the LTR to act as a promoter
for int-l; virtually all the proviruses are either upstream from int-l
in the opposite transcriptional orientation or downstream in the same
orientation. We do not yet know why this is so, but the result
implies that an MMTV enhancer function is fundamental to its oncogenic
potential.

What do we know about the int-1 gene? First, it appears, as hoped,
to be an entirely novel proto-oncogene, unrelated to any of the known
progenitors of retroviral oncogenes. [ The only gene to which it bears
reasonable comparison is another gene, called int-2, that a group in
England found to be activated in about half of another collection of
MMTV-induced mammary tumors.] (slide: int gene) Structural examination
of the gene--- by Roel back in Amsterdam--— and of cDNA clones copied
here from int-1 mRNA shows that the gene is composed of at least four
exons and encodes a protein of moderate size, with a very hydrophobic
amino terminus and a cysteine-rich carboxyterminus. The biochemical
function of the protein is presently obscure, but we are inclined to
believe it is important to normal animals, since, like other proto-
oncogenes, it is highly conserved in evolution: Roel has found that
the product of the human int-1 gene differs in only four of four
hundred amino acids from the mouse int-] protein. Thus far, there is

no evidence that int-] has a role in human breast cancer. (lights on)
This is as far as I can take you on the circuitous path that began
with a search for a short nucleotide sequence lost during reverse
transcription and has led to a search for the function of a mammary
oncogene. Is there a message in all this about how to foretell
jelly-making from tree-shaking? I doubt it. Good luck to Reverend
Jackson.

I would like to conclude with a few words about the pleasures of
scientific life. The first faculty research lecture I remember
attending was given by Gordon Tomkins. He began by claiming that
science is fundamentally measurement, and that the first scientist was
the caveman who wondered how long it took for a rock to fall from his
hand to the ground. I was and am enamoured of this idea, but I am
also struck by our need to interpret and represent the things we have
measured. I imagine Gordon's first scientist finding complete
satisfaction only when he tried to tell a colleague what the speed of
the stone's fall meant for the organization of his world. The
diagrams that have illustrated this lecture have served a similar
purpose for me, reducing an awful lot of numbers and radiographic
Signals to convenient metaphors for the world of viruses and cancers
we seek to understand.

When Tom Sawyer learned how to persuade his friends to paint his
fence, he "comprehended that Work consists of whatever a body is
obliged to do, and that Play consists of whatever a body is not
obliged to do." One of the great pleasures in our lives as academic
scientists is the priviledge to engage in extended play. JI hope it is
not too obviously self-serving to say that it is the mark of an
enlightened society---and university---to recognize and subsidize such

play, be it jelly-making or tree~shaking.
A word about my playmates. (final fruited tree) Science has long
ceased to be a solitary endeavor. Among the best things retroviruses
have done for me has been to foster collegial relations with an
amazingly large and admirable group of people, including faculty,
research staff, and trainees of various sorts. The artist failed to
provide enough fruit for everyone; what is shown is a partial
accounting of those past and those persistant. The one name that must
be spoken is that of Suzanne Ortiz; without her help, I would not have
a story to tell.

Finally, (slide:final tree) a word of thanks to the committee that
invited me to give this talk and to those of you who came to listen.
Some months ago, my wife told me that Bessie Smith used to sing "If
you don't want my peaches, don't shake my tree." I am honored that

you chose to shake mine.