Haroid E. Varmus, M.D.

CUMULATIVE PROGRESS REPORT: 1985-PRESENT.

A. Overview.

As anticipated in my original application for an OIG, my laboratory's activities under the terms
of this award have addressed several questions in two main arenas: retroviral replication and
retroviral oncogenesis. In the sections that follow, I summarize our progress towards
understanding the retroviral lifecycle (with emphasis upon viral entry, proviral integration,
synthesis of reverse transcriptase by ribosomal frameshifting, and virus assembly) and some of the
genes (especially src, Wnt-l/int-1, myc, and their relatives) implicated in retrovirus-induced
neoplasia. (References to our published work on these topics are indicated by numbers in the text
and listed at the end of the application.)

At the time of my original application, the remainder of my research program was in the field of
hepadnaviruses and funded by separate grants. During the past few years, I have decreased and
ultimately ceased our work on hepadnaviruses and have consequently not renewed the grants that
supported it. Concomitantly, our group has devoted increased efforts to aspects of retroviral
replication related to AIDS, and I have obtained NIH support (listed elsewhere) for those
extensions of my program.

B. Highlights.
1. Retrovirus replication.

* Discovery of the nucleoprotein complex that mediates retroviral integration
(1) and the development of the first in vitro system that recapitulates the
integration reaction (2). These advances have allowed a precise definition of substrates and
intermediates in the reaction (3) and stimulated the development of in vitro assays (e.g., for att site
binding (29) that use purified components. More recently, we have been able to simulate the in
vivo reaction more closely in vitro by using minichromosomes as targets, and have developed a
PCR-band assay that allows us to survey thousands of integration events in a single gel. (Fig. 1)

* Discovery of the phenomenon of ribosomal frameshifting in the -1 direction
to account for the synthesis of most retroviral pol gene products (4). Our subsequent
studies of this phenomenon in the context of RSV (5), MMTV (6), and HIV (7) produced the first
insights into the viral signals that influence frameshifting (a “slippery” heptanucleotide sequence
and RNA secondary structure downstream) and attracted attention to a translational mechanism for
generating diversity that has now been encountered in £.coli, yeast, and coronaviruses.

* Cloning of the chicken and quail genes encoding the putative receptor for
RSV, subgroup A. Although the sequencing of genes encoding this long-sought receptor is
not yet complete, it will undoubtedly offer a new perspective on viral-host interactions, the
mechanism of early events in the virus life cycle, and the nature of the presumptively polymorphic
avian proteins that recognize the several subgroups of RSV.

* Demonstration that an ectopically expressed transmembrane protein (human
CD4) can be efficiently incorporated into RSV envelopes during virus assembly
(8). This finding casts a new light on the assembly process and provides new opportunities for
targeting retrovirus vectors to host cells and for understanding the determinants of virus entry.

55
Harold E. Varmus, M.D.

2. Retroviral oncogenesis.

* Discovery that the mouse int-1 (now Wnt-1) gene is expressed in a highly
restricted manner (only in the embryonic nervous system and during
spermatogenesis), implying roles in mammalian development (9,10). These findings,
in concert with Nusse's discovery that Wnt-1 is the mouse homolog of the Drosophila segment
polarity gene, wingless, inspired the gene targeting experiments of others, which demonstrated that
Wnt-1 is essential for formation of the cerebellum and midbrain.

* Biochemical and genetic characterization of Wnt-1 protein as a secretory
glycoprotein (11,12) that can function in autocrine or paracrine modes, presumably
by interaction with membrane receptor(s). This work has placed the Wnr-1 proto-
oncogene and other members of the Wnt family among genes that affect differentiation and cell
growth by cell-cell signalling and has focused attention upon the War receptor(s).

* Development of cell culture assays for biological effects of Wni-1 (13) and
production of transgenic mice as a model for multistep mammary tumorigenesis
involving the Wnt-1 gene (14). These advances secured the definition of Wnr-| as a proto-
oncogene, displayed its unusual tissue-restricted potential for cell transformation, showed the
dramatic effects of Wnt-1 on mammary cell proliferation in male and female animals, and permitted
demonstration of collaborative oncogenic effects with int-2 and other proto-oncogenes.

* Characterization of src mutants with host-dependent and dominant-negative
phenotypes that define important functional domains outside the protein-tyrosine
kinase domain of sre proteins (15, 16, 17, 18). In conjunction with work in several other
labs, these mutants brought the Src-homology (SH) domains 2 and 3 to widespread attention,
providing new opportunites for identifying the proteins that are crucial modifiers and targets for
Sre-like kinases.

* Implication of c-sre protein in the cell cycle through the demonstration that it
is a target for phosphorylation by the mitotic kinase, p34, encoded by the
mammalian homolog of cde2 (19). This work, with concurrent studies from Shalloway’s

lab, identified p60o 876 as one of the first specific substrates for the mitotic kinase and suggested
mechanisms to induce those properties of mitotic cells that mimic cell transformation.

* Generation of a large collection of c-myc mutants that allowed definition of
functional domains of Myc proteins (20). These insertion and deletion mutants have been
a valuable resource for the scientific community over the past five years, expediting the rapid
advances in our understanding of Myc proteins as probable transcription factors.

C. Narrative.
1. Retrovirus replication.
a. Retroviral integration.

Retroviral integration remains the best understood example of recombination in higher
eukaryotes, and the components that account for the efficiency and precision of the reaction are of
great interest. About five years ago, we found that newly synthesized viral DNA in the cytoplasm
and nuclei of MLV-infected cells was part of a high molecular weight nucleoprotein complex that
sedimented at 160S in sucrose gradients (1). Predicting that the nucleoprotein complex might
function as an integration machine, we found (in collaboration with Pat Brown in Mike Bishop's
laboratory) that it was able to mediate correct integration of MLV DNA into a target of naked
Harold E. Varmus, M.D.

lambda phage DNA in vitro (2). More detailed studies revealed that the complex contained viral
capsid proteins, in addition to reverse transcriptase and integrase; that it could be considerably
purified without loss of integration activity; that the linear viral DNA lacked two nucleotides from
the 3 ends of each strand as a result of integrase activity in the cytoplasm; and, most importantly,
that linear DNA (rather than circular DNA, as previously thought) was the immediate precursor to
the provirus, with the recessed 3' ends of viral DNA joined to newly created, staggered 5' ends in
the target DNA (1, 3).

Although the in vitro reaction was monitored initially with a genetic assay (in which a supF
gene in the viral LTR conferred replication-competence on a target phage DNA with amber
mutations), the products can also be followed with physical assays, and the target DNA molecules
can be varied (linear or circular, large or small). In the past year, we have shown that chromatin
can also serve as a target; it is used at least as efficiently as naked DNA and allows us to evaluate
the presence and position of nucleosomes (or other chromatin associated proteins) for influences
upon choice of integration sites. Using either cloning and sequencing or a novel PCR-based
method (see Fig. 1) for assay of integration sites, we have studied integration of MLV DNA into a
small yeast minichromosome with two mapped nucleosome-free domains. Unexpectedly, insertion
occurs somewhat more frequently in regions with nucleosomes, and integration sites in
nucleosomes occur with a periodicity of approximately ten bps, suggesting that one face of the
DNA helix is favored for integration. (A similar periodicity is not observed with the same plasmid
in naked form.) Similar studies of MLV integration in SV40 minichromosomes in vitro and in co-
infected monkey cells also indicate that nucleosome-free regions are not favored. The use of
SV40 DNA as a target in infected cells also permits controlled perturbation of the target, using
temperature sensitive $V40 mutants; preliminary results suggest that the DNA replication activity
of the minichromosomes does not affect integration frequency.

We have attempted to exploit the in vitro integration assay to understand the mechanism of
restriction of MLV infection mediated by alleles at the mouse Fv-/ locus. In our hands, a marked
reduction in the level of full length viral DNA in the cytoplasm appears to be the major impairment
during infection of restrictive hosts; nucleoprotein complexes from restrictive cells can mediate
integration in vitro, and extracts from restrictive cells do not affect integration in trans. We have
concluded that the Fv-1 protein affects a step in synthesis, transport, or catalysis of viral DNA,
before the integration event, but we have been unable to identify an alteration in the viral
nucleoprotein complex in restrictive cells.

To examine the integration step more precisely, we have used yeast expression vectors to
produce large amounts of MLV and HIV integration proteins, and these have been purified to
virtual homogeneity for structural studies (under the terms of another award) and for biochemical
tests. Both proteins have been shown to bind to their cognate att sites in vitro, using gel
retardation assays (21, 22). The binding sites at each end of MLV DNA are equivalent and

relatively short (13 bp are sufficient); binding is sensitive to mutations that affect integration in |

in

vivo. The two HIV binding sites are not equivalent, a larger region is required for optimal binding,

and mutations that affect binding have yet to be tested in vivo (i.e. the att sites have not been
defined genetically). Recently, we have used our purified HIV integration protein for 3’ end
processing and DNA recombination reactions with putative att site oligonucleotides, as originally
described by Craigie and colleagues. This has allowed us to characterize sequence determinants of
the three defined steps (binding, 3'end processing, and recombination) and to demonstrate the
importance of the conserved CA dinucleotide near the end of the viral av site in both the processing
and recombination reactions.

b. Ribosomal frameshifting.

At the time of my original application for an OIG, we had just begun to take seriously the
possibility that ribosomal frameshifting could explain the synthesis of gag-pol fusion proteins from

57
Haroid E. Varmus, M.D.

overlapping reading frames. By the time the award was implemented, we had shown that -1
frameshifting was indeed responsible for synthesis of the RSV gag-pol precursor protein, using a
now standard assay in which gag-pol transcripts, synthesized in vitro by a bacteriophage RNA
polymerase from cloned retroviral DNA, are translated in a rabbit reticulocyte lysate (4). We then
used this assay to study HIV (7) and MMTV (6), as well as RSV (5), confirming the common use
of -1 frameshifting to make reverse transcriptase; in the case of MMTV, two frameshifts occur, one
to generate a gag-pro fusion protein and a second to generate gag-pro-pol protein. The shifts of
frame occur efficiently (5 to 25% of ribosomes undergo a shift) at sites with the general sequence
.& XXY YYZ (X may equal Y). The importance of these sites was documented by the sequence
of transframe proteins and by site-directed mutagenesis, and prompted the currently accepted
"simultaneous slippage" model for -1 frameshifting (5) in many experimental contexts.

In addition, we found that ribosomal frameshifting may be augmented by secondary
structural features of the RNA downstream from the frameshift site. This was first documented
by site-directed and deletion mutagenesis of a predicted stem-loop in RSV RNA (5) and by finding
that the frameshift sites in MMTV RNA were not sufficient for efficient frameshifting (6).
Subsequently, we have shown that the MMTV gag-pro frameshift site is followed by an RNA
pseudoknot, as earlier found by others for a coronavirus frameshift. On the other hand, a
projected stem-loop in HIV RNA appears to have only weak effects on frameshift efficiency in
vitro (23), although our more recent findings indicate that the HIV stem loop has a dramatic affect
upon frameshifting in ceils.

c. Retroviral receptors

About three years ago we initiated a project to clone chicken genes encoding the several
receptors for RS V-related retroviruses. Since the receptors had been defined genetically and not
biochemically, we have taken a genetic approach: mammalian cells (normally uninfectable by most
strains of RSV) are used as recipients of DNA from receptor-producing chick embryos, and
recipient cells that take up DNA (identified by co-transfected markers) are tested for susceptibility
to RSV vectors of an appropriate subgroup. In this way, we have obtained primary transformants
in COS cells and secondary transformants in mouse 3T3 cells that are now permissive for infection
by subgroup A strains of RSV. We have cloned DNA from the secondary transformants linked to
markers present in the primary transfection, and have obtained a clone that appears by several
criteria to be part of the gene for the subgroup A receptor. Using parts of this clone as probes, we
have subsequently cloned homologous portions of quail and chicken DNA and have shown that a
quail DNA clone can efficiently confer sensitivity to RSV-A upon mouse cells (Fig. 2).
Sequencing of the avian clones reveals probable exons that are being used to isolate cDNA clones.

d. Virus assembly.

We have tested several chimeric envelope proteins and heterologous membrane-associated
proteins for their ability to be incorporated into RSV particles. Initially, RSV env-encoded proteins
with insertions of EGF were made: the only chimeric protein that was incorporated into RSV did
not promote infection of mammalian cells expressing EGF receptors, perhaps because the EGF
moiety was likely to be poorly exposed on the surface of the particle. In efforts to make chimeras
in which the ectodomain was entirely encoded by a cellular gene, we also tested normal human
CD4 and found that it (unlike CD4-RSV-env chimeras) was very efficiently assembled into RSV
particles in quail fibroblasts chronically infected with RSV (8). Preliminary results indicate that
several other foreign proteins (including human CD8, the mouse poly IgA receptor, and influenza
hemagglutinin) can also be incorporated.
Harold E. Varmus, M.D.

2. Retroviral oncogenesis.
a. The sre gene and its relatives.

In a long-standing collaboration with Mike Bishop and his colleagues, we have identified and
cloned several genes belonging to the src gene family, including c-src itself, fgr, and hck (24, 25,
26). Our studies of the viral and cellular src genes and their close relatives are now directed
towards understanding the normal and neoplastic functions of the cytoplasmic protein-tyrosine
kinases they encode, with emphasis upon genetic strategies for idenufying regulatory properties
and protein-protein interactions.

Mutational studies of src have refined the definition of aminoterminal sequences required for
myristylation of pp60 (27) and identified multiple domains in the aminoterminal half of the protein
that localize it variously to the plasma membrane, cytosolic vacuoles, and perinuclear membranes
(28). Nucleotide sequencing of a fortuitously isolated host-range allele of v-src, one that permits
transformation of chicken but not rodent cells (29), demonstrated the importance of a highly
conserved amino acid sequence, FLVRES, upstream of the kinase domain of pp60, in a region
likely to be involved in significant interactions between pp60 and cellular proteins (15). This
sequence appears to be an important component of the portion of pp60 now known as src
homology-2 (SH2), a region conserved among the protein-tyrosine kinases lacking transmembrane
domains and related to regulatory domains of several other signal transduction proteins.

Using site-directed mutagenesis, we have produced a large collection of substitution and
deletion mutants in the SH2 and the adjacent SH3 regions of c-sre protein and characterized their
biological and biochemical properties after expression in chicken and mouse cells (16, 17, 18).
Several mutations in SHZ activate the oncogenic potential of ppoO, some in SH2 or SH3 inhibit the
transforming activity of pp60 activated by mutation of Tyr-527, and several mutations produce
host-dependence for transformation. Two mutants (known as M6 and M9) with lesions in the
FLVRES sequence are hyperactive in transformation of chicken cells, deficient in transformation of
mouse cells, and capable of interfering with transformation of mouse cells by activated c-src (18).
The M6 and M9 proteins also display differences in stability, protein-tyrosine kinase activity, and
substrate preference when expressed in the two cell types, implying differential recognition of
important regulators of ppé0.

As an alternative means to identify interactions between pp60 and cellular proteins, we have
sought host cell mutants resistant to transformation by v-src. A rat cell line expressing three copies
of v-src was constructed, mutagenized, and screened for reversion to a non-transtormed
phenotype. One line was isolated and met the appropriate criteria (loss of transformed properties,
continued expression of wild type v-src, and resistance to retransformation by additional v-sre
genes); somatic cell fusion experiments indicated that reversion is.a dominant phenotype (M.
Schofield, Ph.D. thesis).

To expedite biochemical and structural studies of sre proteins, we have produced the products
of chicken c-sre and nearly a dozen mutant versions of c-src in insect cells, using baculovirus
vectors (19, 30). With immunoaffinity chromatography, we obtain about | mg of protein (over
99% pure and enzymatically active) from a one liter culture. These proteins were first successfully
used to look for an activity in mitotic cells responsible for the phosphorylation of threonine
residues in pp60 reported by Shalloway and his colleagues; we found the mitotic kinase, the
product of the mammalian homolog of cdc2, to be the relevant enzyme, thereby implicating c-sre in
the control of mitotic events (19).

During a sabbatical at the Whitehead Institute in 1988-89, I ininated attempts to make mutations

of c-sre in cultured cell lines by homologous recombination. Although this effort was made
partially obsolete by Soriano's success in generating a c-src deficient mouse line (see below), it led

59
Harold E. Varmus, M.D.

to the still unexplained observation that embryonic stem (ES) cells and embryonic carinoma (EC)
cells differ with respect to homologous recombination rates at different loci. We have found that
targeting vectors for c-src and HPRT mediate efficient recombination at the endogenous loci in ES
cells, but not in the EC line known as P19; however, an N-myc targeting vector undergoes
homologous recombination with the N-myc locus at similar frequencies in the two lines.

b. The Wnt-1 gene.

The gene originally known as int-1, recently renamed Wnt-1 as the founding member of a large
and highly conserved gene family (31), was discovered here about ten years ago as a common
target for insertional activation in MMTV-induced mammary carcinomas (32). By cloning full
length CDNA from mammary tumors, we predicted Wnt-1 protein to be secreted, cysteine-rich, and
modified by aminoterminal cleavage of a signal peptide and by four N-linked glycosylations (33).
Direct analysis of Wnt-1 proteins synthesized in vitro and in cells programmed to express the gene
confirmed these findings, while also revealing processing and secretion to proceed inefficiently
(11, 12). We have also recently found that the protein is associated with the chaparonin BiP (or
HSP78), presumably in the endoplasmic reticulum.

The protein has proven difficult to express or purify from expression systems in bacteria,
yeast, and insect cells, and we (and others) have failed to demonstrate biological activity in cell-free
preparations of Wnt-1 protein. Instead we have developed several biological assays for the Wat-1
gene. When introduced into a mouse mammary epithelial cell line (CS7MG), the gene induces a
dramatic morphological change and releases cells from growth restraints (13); in the rat
pheochromocytoma line, PC12, the gene causes the cells to flatten and display greater adherance
(Fig. 3). In addition to these direct, autocrine assays, we have developed a paracrine assay, taking
advantage of the observation that many cell types express but fail to respond to the gene: mouse
3T3 cells producing Wni-1 protein induce characteristic changes in surrounding C57MG cells that
are not expressing Wnt-1.

These assays have been used to examine site-directed mutants of Wnt-1. Glycosylation sites
can be eliminated without loss of autocrine activity, although several mutants are deficient in
paracrine acavity. Changes that add or remove cysteine residues often inactivate the gene, findings
consistent with the remarkable conservation of 21 cysteine residues throughout the Wnr gene
family. (One of the glycosylation mutants and one of the cysteine mutants appear unexpectedly to
display a temperature-sensitive transformation phenotype, which will be useful in attempts to study
the Wnt receptor.) Deletion of the signal peptide prevents secretion, processing, and transformation
in paracrine and autocrine assays, but addition of a signal for retention in the endoplasmic
reticulum (SEKDEL) allows autocrine but limits paracrine ransformation. This result suggests
that productive interactions between Wnt-1 protein and its receptor may occur in the endoplasmic
reticulum of C57MG cells. A summary of the mutants and their protein products is presented in
Figure 4.

As another means to validate the role of War-1 as a mammary oncogene, we have developed
lines of transgenic mice carrying a Wat-1 allele with an MMTV LTR upstream from the gene in the
Opposite transcriptional orientation, mimicking the situation that occurs most frequently during
viral oncogenesis (14). The mice express the transgene mainly in the mammary and salivary
glands, manifest mammary hyperplasia in both males and females, and develop mammary (and
occasionally salivary) adenocarcinomas. When crossed with transgenic mice carrying an MMTV-
int-2 transgene (int-2 encodes a fibroblast growth factor and is also frequently activated by MMTV
insertion mutation), bi-transgenic mice develop mammary cancer at a considerably accelerated rate
in males and virgin females (Fig. 5) showing that two oncogenes encoding secretory proteins can
act cooperatively during tumorigenesis (work done in collaboration with Philip Leder, Harvard
Medical School).

60
Harold E. Varmus, M.D.

To explore the role of the virus more fully, we generated molecular clones of MMTV DNA that
proved to direct production of infectious, tumorigenic virus (34), solving a chronic difficulty in
cloning MMTV DNA. In addition to making MMTV vectors that express heterologous genes in a
glucocorticoid-responsive manner, we have used the cloned MMTV to infect Wnt-1 transgenic
mice, again accelerating tumorigenesis (Fig. 6). Tumors from the infected animals have new
MMTV insertion mutations, often in known loci (int-2, int-3, and Ast), but many tumors are clonal
growths with proviruses in unknown sites now under investigation by a former post-doctoral
fellow (Greg Shackleford, USC).

The potential importance of the Wnt-1 gene in development was initially suggested by our
studies of the pattern of expression in the mouse: the gene is expressed principally in parts of the
embryonic central nervous system in mid-gestation and in the round spermatid (a post-meiotic cell)
during spermatogenesis (9, 10). (Wat-1 is now recognized to be the mouse homolog of the
Drosophila segment polarity gene, wingless, and required for development of the cerebellum and
midbrain in the mouse.) To pursue the developmental effects of Wnr genes in another instructive
experimental setting, we have cloned (35) and sequenced most of the only detected Wnt gene and
its CDNA in Caenorhabditis elegans; this gene is also expressed in embryos and has intron-exon
boundaries that differ from those in other Wnz genes. (Fig. 7)

c. The c-myc gene.

In the first couple of years of the OIG, my laboratory was heavily involved in studies of c-myc
and related genes. We helped to refine the structural analysis of chicken c-myc (36), described
some complex rearrangements that had occurred during downstream insertional activation of c-myc
in a bursal lymphoma (37), and characterized amplified c-myc loci in the COLO320 and SEWA cell
lines (38, 39, 40) in collaboration with the Bishop lab. We learned the assay for co-transformation
of rat embryo fibroblasts by myc and ras genes, as originally developed by Robert Weinberg's lab,
and used it to show that elevated expression of a normal human or chicken c-myc gene is sufficient
to collaborate with a mutant ras gene (i.e. mutations of the c-myc coding sequence are not required)
(41) and that N-myc has oncogenic activity in the assay (42). The assay was cenrral to our efforts
to characterize a large collection of insertion and deletion mutants of human c-myc (20); the mutant
proteins were also examined for nuclear localization, stability, ability to transform an established
line of rat fibroblasts, and DNA binding properties. These experiments allowed us to propose the
first rough maps of the functional domains of c-myc protein (20, 43, 44), and the mutants are now
standard reagents for the community that studies the myc gene family.

d. Avian nephroblastoma and c-Ha-ras.

At the inception of this grant period, we were attempting to identify the proto-oncogenes
activated by proviral insertion in MAV-2-induced avian nephroblastoma. We found a single tumor
in which an apparent insertion of MAV-2 DNA initiated high level transcription of the chicken Ha-
ras gene (45). In the absence of further definition of activated genes among our tumors, we have
abandoned work on this problem.

3. Miscellany.

This award has partially supported work only tangentially related to its major themes by a few
especially talented and independent post-doctoral fellows, two of whom have obtained additional
funding as Scholars of the Markey Foundation. Titia DeLange staged a valiant but ultimately
unsuccessful effort to clone cDNA from the Wilms’ Tumor susceptibility locus by subtractive
hybridization (discussed in Ref. 46) and then cloned and examined structural properties of human
telomers (47). David Kaplan, working largely in collaboration with Deborah Morrison in Rusty
Williams’ laboratory, described interactions between the PDGF receptor and several elements in
signal transduction pathways (the Raf kinase, phosphotidylinositol 3‘kinase, and GTPase activator
Harold E. Varmus, M.D.

protein; 48-51, 52) and, in collaborations involving Mike Bishop's and Luis Parada’s laboratories,
examined the tyrosine phosphorylations that follow treatment of PC12 cells with nerve growth
factor (NGF), leading to the suggestion that the high affinity NGF receptor is the product of the mrk
proto-oncogene (53, 54). Peter Sorger has begun a genetic analysis of the role of the CDC28
protein of S. cerevesiae in cell cycle control, with emphasis upon phosphorylation events, with
additional advice from Andrew Murray.

D. Effects of the OIG.

It would be disingenuous for me to say that the OIG has had a dramatic effect upon the
operation or goals of my laboratory or upon the tenor of my own life in science. Stull, the award
has simplified my relationships with granting agencies, since all my support (save my American
Cancer Society Professorship and post-doctoral fellowships) now comes from NIH and mostly
from this award. I have, no doubt, saved some time that would previously have been devoted to
grant renewals and to more lengthy and more numerous progress reports, but I cannot say that
these imposed major burdens (indeed the enforced exercises were often useful).

The stability of funding has, of course, been a highly favorable feature of the OIG, one that has
permitted me to take a somewhat longer range view of our research activities than I might
otherwise have done. This has probably increased my willingness to commit ourselves to some
long-term, high-risk projects; among the successes are the in vitro integration reaction, some
genetic approaches to src, the cloning of avian retroviral receptors, and development of War-lt
transgenic mice, although there are also failures and projects whose outcome remains uncertain. |
have also felt more comfortable about allowing a few highly talented post-doctoral fellows take on
projects that may not have fit closely into work proposed or ongoing; in some of these situations [
have affiliated myself with the work rather loosely and given the fellows the opportunity to publish
independently. or with collaborators from other labs.

The major problem I have encountered has been financial: with most of my research funding
from a single source, budgeted over a seven year period, it is difficult to predict (and obtain
financing for) purchase of major pieces of equipment and to adjust to unanticipated increases in
expenditures. In my case, the most dramatic example of unanticipated increases has been in the
domain of animal-based research, since the use of transgenic animals and our efforts to make gene-
deficient animals by homologous recombination have escalated our vivarium costs and required
employment of an animal care technician at a time when applications for supplementary
appropriations were unlikely to be well received. In addition, the difficulty of buying equipment
within an inflexible long-term budget has made much of our laboratory gear outmoded and subject
to frequent breakdown. Perhaps the OIG program could be modified to allow reevaluation of the
budgets every three years or so.
Harold E. Varmus, M.D.

RESEARCH PROPOSAL
A. Perspective

The vast majority of the work proposed for the next seven years of support by this award
represents a direct extension of studies already underway, so the ensuing sections generally
coincide topically with the progress report. There is a modest change of emphasis, with more
attention given to virus entry and assembly than to ribosomal frameshifting, and with the work on
oncogenes generally restricted to the src and Wnt gene families. (Wherever references are required
in this section, they are given by number to our papers and by letter to papers from other labs also
listed at the end of the proposal.)

B. Retroviral replication.
1. Retroviral integration.

The nucleoprotein complex that mediates the integration reaction remains poorly characterized.
We plan to purify substantial amounts of the MLV complex, as well as complexes formed after
infection with other viruses (RSV, HIV) for determination of their protein constituents and for
visualization in the electron microscope (in frozen hydrated form, in collaboration with Jim Hogle,
Scripps Institute). To determine when the complex dissociates, we will ask whether integration
intermediates can be precipitated with antisera known to recognize gag-derived or pol-derived
components of the complex.

We will look more carefully at the apparent periodicity of retroviral integration into
nucleosomal domains of minichromosomes, using the newly developed PCR assay (Fig. 1) to
analyze mass populations of integration products (rather than cloning and sequencing of individual
products as inthe past) and using an MMTV LTR-based minichromosome on which a nucleosome
has been stringently positioned (a). We will also be evaluating evidence that integration site
preference is biased by sequence as well as nucleosome position, as suggested by recent
experiments (e.g., Fig. 1 and ref. b). To this end, we will consider cloning integration sites
preferred during MLV infection, in the manner described for RSV by Coffin’s laboratory (b); the
cloned DNA could then be studied as targets in vitro, in the form of naked DNA and reconstituted
chromatin. In addition, we have recently found that we can monitor integration events mediated by
either nucleoprotein complexes or purified HIV or MLV IN, using a wide variety of DNA's as
targets. Such experiments should be informative about sequence preferences during integration.

The PCR-based assay of integration events also allows us to assess the effects of protein
binding, DNA bending, and other perturbations of the target upon integration efficiency. (For

example, the yeast «2 protein prevents integration of MLV DNA into the a2 binding site,
simulating a DNAse footprint.) We will use a variety of model systems (e.g., lambda phage azr
sites and integration proteins) to assess the influence of DNA structure and associations upon
integration site use.

Our recent success in obtaining SV40-MLYV cointegrates from coinfected monkey cells allows
us to study the effects of replication, transcription, and packaging of SV40 DNA upon its use as a
target for MLV integration. For this purpose, we have obtained several conditional mutants of
SV40 affecting T antigen and VP1 (c). In addition, we have isolated $V40 minichromosomes tor
use as targets for integration in vitro. Dramatic changes in MLV integration site preference in
$V40 DNA occur during the SV40 life cycle, and we will attempt to identify the determinants of
these changes.

Integration proteins are capable of specific binding to viral att sites, removing 2 nt trom the
3'ends, cutting target DNA, and joining it to the viral substrate, both in vivo and in vitro (55, c’).
Harold E. Varmus, M.D.

To begin to assign domains for these various functions to these relatively small proteins (32-46
Kd), we are enlarging our repertoire of MLV and HIV integrase mutants by site-directed
mutagenesis at conserved residues and producing chimeric (MLV-HIV) proteins; these will be
expressed in yeast (or £.coli) and tested in reactions for each activity in vitro. (The MLV and HIV
components in chimeras can be distinguished by differences in arr site preferences and in the
distances between staggered nicks in target DNA.) Ultimately, we hope to be able to correlate a
genetic definition of the domains with a structural definition of the proteins obtained with Xray
crystallography (a collaboration with Bob Stroud, UCSF, funded by another award).

We also plan to exploit the available assays for integrase function to define the nucleotide
sequence requirements for azt sites and chromosomal targets more rigorously. As a first step, we
will continue to study the several MLV and HIV azz site variants made for binding studies, and we
are examining the utility of various oligonucleotides, varying in length and sequence, as target
sites. (We have recently found that oligonucleotides other than cognate aft sites as used in the
assay described by Craigie [c’] can serve as targets for IN-mediated integration.) Ultimately, we
plan to generate banks of oligonucleotides with randomized sequences at appropriate positions (d)
to ask which sequences favor or interdict the steps in the integration reaction.)

We are continuing to study the mechanism of Fv-1 restriction of MLV infection. The major
new goal is to attempt cloning of the Fv-1 locus, using a gene transfer assay in which resistance
(the dominant trait) will be scored by infecting colonies of cotransfected cells with N or B tropic
strains of MLV. We are also testing whether overexpression of part or all of gag p30 from
restricted strains can abrogate resistance. This might allow improved mapping of the Fv-1 target
and provide a means to isolate the Fv-1 gene product by affinity chromatography or the gene by
protein affinity screening of CDNA expression libraries.

2. Retroviral receptors and virus entry.

As summarized in the progress report, we have recently obtained genomic clones that appear
to encode the subgroup A receptor gene from the chicken and quail m-a locus. We are now
attempting to finish the sequencing of these clones, identify receptor mRNAs, clone and sequence
the cDNAs, and test all relevant clones for biological activity. In addition, we will want to know
whether exonic clones detect non-permissive alleles at the cv-a locus, the other chicken loci (tv-b
and tv-c) believed to encode receptors for the RSV subgroups B, C, D, and E, and related genes in
other birds and mammals. Attempts to understand the normal function of what we expect to be an
extensive, polymorphic gene family will begin with analysis of predicted protein sequences,
production of antibodies, and expression of receptor genes in various contexts. [t would be
premature to guess the nature of the receptors, especially in view of the heterogeneous nature of the
retrovirus receptors whose genes have been cloned (receptors for HIV, ecotropic MLV, and
gibbon ape leukemia virus [e, f, g]).

In the more distant future, we will attempt to define the site and nature of the interaction
between RSV envelope proteins and cognate receptors, making use of cloned RSV env genes and
receptor genes. Chimeric and mutant receptors will be produced, and interactions with RSV or
env proteins will be examined functionally (by tests for permissivity to infection in cell culture) and
biochemically (in collaboration with Judy White, Dept. of Pharmacology, UCSF). Another long-
term goal is an explanation of how reverse transcriptase is activated during virus entry to produce a
substrate for the integration machine. We will also use receptor clones to produce transgenic mice
that can be infected with RSV; this will permit the development of useful experimental reagents---
mice and mouse cell lines that can be efficiently infected with high titre, replication-competent RSV
vectors, which will not produce infectious virus due to intracellular blocks to later stages of the
virus life cycle.

64
Haroid E. Varmus, M.D.

We have recently begun to apply our experience with avian viral receptors to a similar problem:
the cloning of another potentially polymorphic family of transmembrane proteins that serve as
receptors for feline leukemia viruses (FeLVs). Our primary strategy is to use receptorless cells
(mouse 3T3 cells) as recipients for cat DNA, scoring positive cells with virus vectors bearing
various FeLV glycoproteins and selectable genetic markers. In this venture we are collaborating
with Jim Mullins (Stanford) who has shown that pathogenic potential of strains of FeLV can be
ascribed to variations in env genes (h). We hope that the nature and distribution of the receptors
for pathogenic variants will be informative about mechanisms of pathogenesis (e.g. anemia and
immunodeficiency).

3. Viral assembly and cell targeting.

We are pursuing our recent discovery that human CD4 is efficiently incorporated into RSV
particles in several ways. First, we are using a variety of other ectopic membrane-associated
proteins to determine which can be assembled efficiently into RSV..We need to confirm
preliminary evidence for incorporation of CD8, the poly IgA receptor, and influenza HA protein,
and we are also testing the complement (and Epstein-Barr Virus) receptor CR2 and Wnt-1 protein.
When the cDNA for the RSV-A receptor has been cloned, we will also ask whether that receptor
can be incorporated. To explore the criteria for inclusion, we will test several variant versions of
these proteins, including chimeras in which the transmembrane and cytoplasmic domains are
replaced by RSV env-derived sequence (thus far, CD4-RSV chimeras perform much less well than
wild type CD4), a collection of cytoplasmic tail mutants of the poly [gA receptor (provided by
Keith Mostov, UCSF), and proteins (influenza HA and Wnt-1) linked to the membrane by
phosphatidyl-inositol rather than transmembrane domains.

Our currently favored explanation for why ectopic proteins enter virus particles, whereas the
vast majority of endogenous membrane-associated proteins do not, is that membrane proteins
lacking a cytoplasmic partner proceed via a default pathway to regions of the membrane that
ultimately form viral envelopes. We will test this idea by supplying suitable cytoplasmic partners
(e.g. the Lck protein-tyrosine kinase for CD4 and CD8 [i]) or by attempting to anchor proteins in
the cell by adding cytoplasmic tails that should recognize cytoplasmic proteins endogenous to avian
fibroblasts. (For example, the env-sea fusion protein encoded by the $13 avian tumor virus is not
present in virus particles [j], suggesting that addition of sea-encoded sequences to membrane
proteins might impede assembly.)

Finally, we will ask whether any of the viruses carrying ectopic proteins in their envelopes can
be directed to new cell types as a result of interactions with membrane proteins on the target cells.
Although this is being tried initially using HIV gp120-producing human cells as targets for RSV
carrying CD4, we will examine this question with viruses carrying any of the tested proteins.
(Binding of virus particles, synthesis of viral DNA, and expression of a selectable marker in a
virus vector are among the assays that will be used.) These experiments raise questions of obvious
practical significance for virus vectoring, but also address interesting speculative issues about the
range of membrane proteins that can be used as ligands and receptors to mediate virus entry. For
example, can virus infection proceed normally if the usual receptor is in the virus particle and the
viral envelope protein is on the surface of the target cell?

4, Ribosomal frameshifting.

Three components are believed to influence the -1 frameshifts that allow synthesis of pol gene
products: (i) the nucleotide sequence at the frameshift site, (ii) secondary structure of downstream
RNA, and (iii) host tRNA species. (i) We are examining a large number of potential frameshift
sites (naturally occurring sites, synthetic sites that conform to the "rules", and mutant sites) to
observe frameshift frequencies in various RNA contexts; we will then re-search the sequence data
bank to attempt to identify cellular and viral genes that might exploit the frameshifting

BS
Harold E. Varmus, M.D.

Opportunities. (An earlier more restrictive search, based upon a less complete understanding of the
components involved, brought forth few candidates other than retroviruses and coronaviruses.)
(ii) To document the role of RNA secondary structure, we have initiated a collaboration with
Ignacio Tinoco's lab (UC Berkeley), using biochemical and biophysical (NMR) means (k) to study
the pseudoknot that we have recently proposed to exist in MMTV RNA on the basis of site-directed
mutants. We hope to use the structural information to design RNA structures that promote
frameshifting, considering both components folding in cis and those that anneal in trans. We also
plan to examine the physiological significance of the structure more closely: unpublished
experiments that follow nascent polypeptides imply structure-dependent ribosomal pausing at the
frameshift site, but we can now examine this point more precisely with the ribosome mapping
procedure of Wolin and Walter (1). (ii) To identify the tRNAs that mediate frameshifts, we are
collaborating with Dolph Hatfield (NIH), supplying him with a variety of RNAs with wildtype and
mutant frameshift sites for assay in a tRNA-dependent system he has developed in Xenopus
oocytes.

We also plan to judge the influence of frameshift efficiency upon virus production, using site
and structure mutants of RSV and perhaps other viruses; preliminary results indicate that mutants
with modest effects on frameshifting in vitro have dramatic effects on replication in vivo. The role
of trans-acting factors in frameshifting will be evaluated by tests for RNA binding proteins specific
for downstream structural elements (a search newly encouraged by evidence that HIV
frameshifting is more dependent on RNA structure in vivo than in vitro) and by genetic modulation
of frameshifting in budding yeast, using retroviral frameshift signals known to function in
heterologous contexts (m).

C. Retroviral oncogenes.
1. The src gene family.

We are making a major effort to understand the role of the SH2 domain in mediating the
regulation of cytoplasmic protein-tyrosine kinases and the interactions among proteins involved in
signal transduction (n). (i) SH2 mutants of activated c-src that display a host-dependent capacity to
transform cells will be exploited to look for cellular genes whose products have functionally
significant interactions with pp60. By mutagenesis of mouse cells or by transfer of chicken
genomic DNA (or cDNA), we will attempt to induce src-dependent transformation of mouse cells
expressing SH2 mutants previously shown to be hyperactive in chicken cells but biologically
inactive in mouse cells (16, 17). Chicken genes or mutant mouse genes that appear to suppress the
host-range defect will be cloned by gene transfer techniques. (ii) We will pursue the dominant
negative characteristics of SH2 mutants that interfere with transformation of mouse cells by
activated c-sre (18). Dominant-negative alleles, M6 and M9, with additional mutations (e.g.,
deletion of the kinase domain or loss of the myristylation signal) will be tested for inhibitory
activity to identify determinants of the dominant negative phenotype and to isolate mutants with a
more potent phenotype. Appropriate alleles will be tested for their ability to interfere with
transformation by oncogenes in the src gene family (especially Ack and Ick), genes encoding
transmembrane tyrosine kinases, and other classes of oncogenes. We will also ask whether such
alleles can interfere with cell signalling events thought to be mediated by SH2-containing proteins,
including /ck in T cell activation (in collaboration with Dan Littman, UCSF) or Ack in myeloid
differentiation (with N.Quintrell in Mike Bishop's lab). (iii) We will attempt to identify cellular
proteins responsible for the differences in enzymatic activity and substrate preferences of SH2-src
mutants synthesized in chicken and mouse cells (17, 18). Mutant proteins immunoprecipitated
from those cells or purified in large amounts from the baculovirus-based expression system (30)
will be tested for modulation of kinase activity after incubation with extracts from avian and
mammalian cells. We have synthesized wild type and mutant SH2-sre domains in £.coli and will
attempt to use these truncated proteins to inhibit modulating activities and to purify them by affinity
chromatography. The wild type and mutant truncated proteins will also be used as binding
reagents to examine the effects of functionally significant SH2 mutations upon the ability of SH2

AA
Harold E. Varmus, M.D.

domains to recognize other proteins, especially phosphotyrosine-containing proteins. (iv) One of
our host-dependent SH2-sre mutants (M9) induces tyrosine phosphorylation of many proteins in
NIH3T3 cells, although it fails to transform these cells (18). We will look among the proteins
known to be tyrosine phosphorylated in src-transformed cells for those that might be implicated in
the failure of M9 to transform; GTPase activator protein appears to be one such protein and its role
in sre wansformation will be further examined in this context.

Assumptions about the roles of c-src in normal cell growth and development have recently been
teevaluated in the light of Soriano's revelation that c-src deficient mice survive through
embryogenesis and early post-natal life, usually succumbing to malnutrition and osteopetrosis as 4a
result of impaired osteoclast function (0). These findings suggest that in the vast majority of cells
in vertebrate organisms c-src is not an essential gene, despite dramatic conservation of coding
sequence and widespread expression of the gene. Our group and others are endeavoring to make
mice deficient in other members of the src gene family to test the hypothesis that the functions of
sre can be complemented by other members of the family. Because of our familiarity with hck and
fgr, two sre-like genes expressed predominantly in the myeloid lineage (25,26, p), we have
initiated efforts to interrupt these genes in ES cells (Bradley's AB1 line) and to produce chimeric,
heterozygous, and ultimately homozygous mutant animals by blastocyst injection and breeding.
As discussed in the progress report, we have learned to mutate c-sre by homologous recombination
in ES cells (obtaining recombinants at a frequency of about | per 100 cells incorporating DNA after
electroporation), we have made similar targeting vectors with genomic clones of Ack and fgr, and
we have recently obtained many clones of ES cells with targeted mutations of Ack. Stem cells
identified (by PCR and Southern blotting) to have Ack or fgr mutations resulting from homologous
recombination will be injected into blastocysts (in collaboration with P. Soriano, Baylor); chimeric
progeny will be tested for their ability to transmit the mutant gene; and mice homozygous for the
deficiency will be bred and examined for phenotypic changes in several tissues and cell lineages at
different developmental stages. Naturally, special emphasis will be given to the observation and
manipulation of hematopoietic lineages in vitro and in vivo. The heterozygous animals will be
crossed as appropriate with mice carrying null mutations in other src gene family members (and
mice with mutations in other genes important in hematopoeisis, e.g. ab) as those animals become
available. One obvious question to be addressed by such studies is whether certain differentiated
properties or cell growth are affected in mice lacking multiple family members, suggesting that src-
type genes can complement each other.

We are using $V40-immortalized, src-deficient mouse fibroblasts obtained through a
collaboration with P.Soriano (Baylor) to study the location and enzymatic activity of c-sre protein
during the cell cycle. As noted earlier, we and Shalloway's group have shown that during mitosis

pp60" "¢ is more active as a kinase and phosphorylated on serine and threonine residues near the
amino terminus by the cdce2 kinase (q,19). We have made mutations at the mitotic phosphorylation
sites in human c-sre protein to gauge the influence of these mutations on the kinase activity and the
location of pp60 during interphase and mitosis in the "background free” cell context. Preliminary
results with c-src endogenous to, or over-expressed in, normal rat cells using immunofluoresence

methods indicate that ppeoe> ”° is located in several sites in addition to the plasma membrane.
These include an association with microtubules (also see ref.r), the microtubule organizing center,
the mitotic spindle, and perinuclear membranes that appear to be late endosomes, as they are
enriched for the mannose-6-phosphate//IGF2 receptor. We are attempting to confirm and refine
these results with biochemical fractionation of cell organelles and immunoelectron microscopy in
various cell types. Potential substrates for the src kinase will also be sought in relevant fractions.

2. The Wnt gene family in development and oncogenesis.

The search for receptor(s) that mediate the developmental and oncogenic actions of War family
proteins is central to our current and future studies of Wnz genes. After considerable unsuccessful

67
Harold E. Varmus, M.D.

efforts to purify biologically active Wnz-1 protein to serve as a radioactive ligand in this search, we
have initiated several other approaches to the receptor and its gene. In the long run, we aspire to
characterize what is likely to be a family of receptor genes for the structure and function of the
proteins, the pattern of expression, and the mode and pathway of signal transmission. A sampler
of possible routes to these genes follows.

(i) In hopes that NIH3T3 cells, which do not respond to Wnt-1, will respond if provided with
receptor, we will transfect cells programmed to express Wat-1 with cDNA expression libraries
prepared from responsive cells (PC12 and C57MG cells) and with human genomic DNA.
Although we will look initially for traditional properties of ransformed cells (focus formation or
growth in agar), we are also planning to identify genes that respond to a Wnt-1 signal by
augmented expression. This wouid allow us to use reporter genes linked to promoters from such
genes to moniter for a Wnt-1 response in cells that might not manifest an otherwise recognizabie
phenotype. We will use differential cDNA cloning and some educated guesses (e.g., genes known
to respond to growth factors [s]) to find such genes. (We have recent evidence that this approach
may work since we have found that one early response gene is induced after temperature shift-
down of C57MG cells expressing a ts allele of Wnt-1.) We will be able to distinguish spontaneous
transformants from Wni-dependent transformants, using our newly isolated ts alleles of Wnt-1 or
Wnt-1 retrovirus vectors that also carry a gene (HSV tk) we can select against. (ii) In hopes of
producing a Wnt protein active in soluble form, we are making a series of deletion mutants and
fusion proteins (e.g. Wnt-1 - IL2 chimeras), and we will screen the mutants-for biological activity
in a new and simplified paracrine assay using transiently transfected quail cells and CS7MG cells
as responders. (This assay reduces the time required from 2 weeks or more to 4 days.) (iii) We are
testing Wnt genes from other organisms (fly, worm) and other murine members of the Wnt gene
family for their ability to ransform CS7MG cells. If any fail to ransform and thus do not properly
activate the Wnt receptor that must already be present in those cells, we will attempt to isolate the
cognate receptor from the appropriate organism, using transformation of CS7MG cells as an assay.
(iv) We are attempting to make retroviruses that carry Wnt-1 proteins on their surface, inspired by
our success in producing RSV particles that contain CD4 (see above). Such viruses will be tested
for biological activity in ransformation assays and for use as radioactive ligands for Wnt receptors.
(v) Armadillo, a gene that acts downstream of wingless in the fly, encodes the insect homolog of
plakoglobin, a constituent of desmosomes and some cell-cell junctions (t). We have obtained
antibodies (from Drs. P. Cowin, NYU; J. Stanley, NIH; and B. Gumbiner, UCSF) against
several membrane glycoproteins associated with plakoglobin-containing structures (desmogleins,
desmocollins, desmoplakin and cadherins) to attempt to inhibit transformation in paracrine assays,
on the assumption that one of those glycoproteins might be a Wnt receptor. (vi) We have
preliminary evidence (based upon in vitro kinase assays of CS7MG membrane proteins precipitated
from control and Wnt-1-expressing cells with and-phosphotyrosine antibodies) to suggest that Wnt
receptors might be protein-tyrosine kinases. We will explore this further in other cell types and
through better characterization of the proteins phosphorylated in the Wnr-dependent assay. (vii)
We are using a PCR-based approach to find mutants in mutagenized worm populations with
deletions or base substitutions that eliminate restriction sites in the C.elegans Wnt gene (in
collaboration with Cynthia Kenyon, UCSF). After characterizing any mutants obtained, we will
look for additional mutants with similar phenotypes or for suppressor mutants, in hopes of finding
abnormal receptor alleles.

A number of additional experiments will address the role of Wnr genes in development. (i) We
are using a sensitive and quantitative PCR-based assay to measure the amount of Wnt-1 RNA ina
wide variety of mouse tissues, revealing previously undetected RNA in some organs (¢.g. one
transcript per 50 cells in the thymus) and stringently documenting the absence of RNA in others
(e.g. less than one transcript per 100,000 cells in the mammary gland). This assay will soon be
extended to other members of the War family (31, u). (ii) We are trying to idenufy the region of
the Wnt-1 gene responsible for regulated expression during mouse development. Since over 5 kb
of sequence upstream of the transcriptional start has proven insufficient to drive appropriate

68
Harold E. Varmus, M.D.

expression of a beta-galactosidase reporter gene in transgenic animals, we are making constructs
that contain introns, exons, and downstream sequence as well. If successful, the resulting
transgenic mice will permit a more precise definition of the cells that express Wnz-1 in the embryo
and adult. In addition, we will be able to make transgenic mice that express other members of the
Wnt family andWni-1 genes with site-directed mutants active in autocrine but not paracrine assays,
to ask, in collaboration with Mario Capecchi (Univ. of Utah), whether such transgenes can
complement a Wnt-1 deficiency (v). (iii) Preliminary evidence, obtained with the modified
paracrine assay for Wnr-1, indicates that some of our site-directed mutants interfere with the
paracrine effects of wild type Wnz-1 (i.e., are dominant negative mutants). If so, these alleles will
be tested for developmental effects in Xenopus (w), in collaboration with M. Kirschner (UCSF),
or in mice. (iv) We will more fully characterize the pattern of expression of the C.elegans Wnt
gene, using in situ hybridization and Wnt antibodies. Overexpression of the gene and expression
in ectopic sites will be attempted using new methods for gene transfer (w'). These studies could
be influenced by placement ofWnr on the “contig map" of the C.elegans genome (the gene is now
mapped to an unattached DNA “island”) or by success in our other attempts to make Wnz mutants
in the worm.

We will also be extending our efforts to use MMTV and Wnr-1 in models for understanding
multi-step carcinogenesis. (i) We plan to cross Wat-1 transgenics and Wnt-\/int-2 bi-transgenics
with animals carrying other oncogenic transgenes: an MMTV-TGF-alpha transgene [x; from Bob
Coffey, Vanderbilt] and an MMTV-inz-3 transgene [from Bob Callahan, NIH], or inactivated
tumor suppressor genes (Rb and p53, from Tyler Jacks in Bob Weinberg's lab, MIT). Each of
these has its attractions. TGF-alpha is another secretory protein, one that (like int-2 protein)
activates a transmembrane receptor with protein-tyrosine kinase activity and acts as a weak
mammary oncogene; it is of interest to know whether three secretory proteins (Wnt-1, Int-2, and

TGFa), at least two of which act through similar signalling pathways, can synergize during
tumorigenesis. /nt-3 encodes a (truncated) transmembrane protein without a kinase domain; as a
transgene it involutes mammary epithelium (in contrast to the proliferative effect of Wnr-1) but it
also leads to mammary tumors; how wiil the Wn-1 and int-3 wransgenes interact in preneoplastic
tissue? Mutant tumor suppressor genes are often found in human mammary carcinoma, but their
role in tumorigenesis has been difficult to study, and our surveys with p53 antisera have failed to
show p53 mutations in our transgenic tumors. A transgenic model for cooperation between
dominant and recessive mutations is potentially powerful for simulating the interactions that appear
to occur in the human disease. (ii) Viral determinants of MMTV-induced tumors are not well
understood. We have made mutations in the open reading frame of the LTRs in our infectious
clone of MMTV; provisional evidence indicates that the ORF mutants replicate well in culture and
in the animal, but are deficient in tumor induction. We are repeating this experiment on a larger
scale in BALB/c animals and in Czech 3 mice, devoid of endogenous MMTY proviruses (in
collaboration with Bob Callahan, NIH). These experiments take on new significance in view of
the recent discovery that MMTV ORFs encode superantigens that activate or delete T cells of
defined V-beta subclasses (y). We will devise experiments to ask whether the nature of
endogenous or exogenous MMTV ORFs influence mammary tumorigenesis after virus infection or
in the transgenic model by deletion or activation of T cells that might be implicated in tumor
immunity. (iii) MMTV-induced and transgenic mammary tumors are generally invasive and
sometimes metastatic; we will ask whether such tumors are associated with activation of genes that
have been implicated in tumor progression, including expression of matrix metalloproteases in
stromal cells (such as stromolysin -3 [z]) and angiogenic factors.

“a
Harold E. Varmus, M.D.

Use of Vertebrate Animals

1. We are studying the genetic basis of cancer, using a variety of retroviruses and viral and
cellular oncogenes to induce tumors in widely-studied laboratory animals (chickens, mice and
rats). Our ultimate objective is to identify and characterize several genes that contribute to the
development of cancers in animals, in hopes that this information will be useful in designing
attacks upon human cancer and germane to an understanding of normal regulation of growth and
development. Studies of the sort we describe here have been important in transforming a
completely mysterious process (the origins of a cancer cell) to one we can now attribute with
confidence to mutations in a few isolated genes in many circumstances. Continued study of such
oncogenes biochemically in test tubes and functionally both in cultured cells and in experimental
animals seems very likely to have a significant impact upon cancer in the next decade. Although
we and others do as much as possible with cultured cells, there is no completely satisfactory
correlation between the tumorigenic effects observed in the animal and changes noted in cells ina
petri dish. Hence we must continue to rely upon assays in animals, particularly when certain cell
types (mammary cells, blood cells, etc.) are involved. Finally, biochemical study of oncogenes
and viruses depends heavily upon appropriate antibodies that are most reliably produced by
immunization of rabbits - and sometimes by mice or chickens. This too involves time-honored and.
virtually painless methods.

Allocation of Animals and Rationales

For studies involving tests of the tumor-inducing capacity of viruses or cell lines in chickens,
mice and rats, a minimum of 8-12 animals is required in each experiment to obtain a meaningful
result, since at least two dilutions of the inducing material must generally be used. Naturally it is
impossible to estimate how many experiments will be done in any given year; the number of
animals requested is based upon past performance. For studies involving the induction of
antibodies in rabbits, chickens, or mice, it is customary to immunize 3-4 animals for each antigenic
preparation, since animals may differ appreciably in their immunological response. For studies
involving the generation of transgenic mice, it is generally necessary to use 10-30 animals to obtain
a single "founder" animal in which the foreign DNA has been inserted in the germ line. Since 5-10
“founders” are required to obtain meaningful results with each transgene, single experiments may
use as many as 300 mice.

Anticipated Results

Our general goals are to understand the molecular basis of cancer and the manner in which
retroviruses grow. The few types of experiments in which we must rely upon animals rather than
cultured cells are intended to give us the following kinds of information:

i). To assess the cancer-causing effects of an oncogene, it is necessary to examine the
oncogene directly in an animal. Through such tests and correlations with biochemical
properties of oncogenic proteins measured in the test tube, we expect to learn the attributes
of such proteins that are responsible for their tumor-causing potential.

ii). In studies of oncogenic proteins and of viral and cellular proteins that participate in the
virus multiplication cycle, it is necessary to have antisera that can specifically recognize
proteins of interest, both for purposes of detection and isolation of those proteins. We
expect the immunization studies to provide us with the necessary sera.

70
Harold E. Varmus, M.D.

The use of transgenic mice for the study of oncogenes has revolutionized cancer research because it
is now possible to produce animals in which all of the target cells are afflicted by an initial step in
carcinogenesis. This provides an ideal setting in which to seek the still largely mysterious
secondary events in neoplasia. In the experiments we are performing, we expect to learn about
such secondary events in breast cancer and, in addition, to learn about the role of proto-oncogenes
(the normal forms of oncogenes) in development of an organism.

We will use both male and female mice from strains CS7B6, BALB/cB45, SJL, and FVB.
Ages will vary from 3 weeks to 12 months. We will maintain approximately 1,500 mice per year.

2. We have chosen mice because for our work, they are the cheapest mammalian species available
and possess a fast reproductive cycle. The estimated size of our colony reflects our need for an
ample statistical base for these studies.

3.-5. Veterinary care and euthanasia procedures comply with stringent guidelines promulgated by
UCSF's Committee on Animal Research. These guidelines are based upon PHS regulations.

71
Harold E. Varmus, M.D.

PRIMERS

-BASED ANALYSIS OF MLV INTEGRATION INTO
re TA MINICHROMOSOMES AND NAKED DNA

Mc BNA Mc DNA Mt DNA MC DNA

 

 

 

 

Cri" Cr" fri fT"
(Lo: |

NUCLEOSOME
FREE
REGION

 

ah pial ene TOT

‘

 

 

 

Figure 1. Polymerase chain reactions with 32p-1abeled
oligonucleotide primers were used to characterize the
products of integration of murine leukemia virus DNA by
nucleoprotein complexes into naked Trp-Ars plasmid DNA or
minichromosomes prepared from S.cerevectiae (2). Each
primer set included one primét for a sequence near the end
of an MLV LTR and one for each of four positions on the 1.5
Kb Trp-Ars DNA. Each reaction was performed in duplicate on
both naked DNA (DNA Lanes) and minichromosomes (MC), and
markers (M) of known size were electrophoresed in parallel.
The sausage-like drawings indicate the positions of mapped
nucleosomes; two nucleosome-free regions, representing the
origin of replication and the transcriptional control
region, are also shown.

At least four conclusions can be drawn from this
analysis: (1) integration does not occur randomly in naked
DNA, since some sites are clearly preferred; (ii) there is
little or no difference between use of nucleosome- free
regions of MC DNA and the homologous regions of naked DNA;
(1i1) different sites are favored for integration events in
nucleosome-containing regions of MC DNA and many of these
sites are used more efficiently than the favored sites in
naked DNA; and (iv) many of the favored sites in nucleosomal
domains are positioned 10-11 bp apart (as denoted by the
black dots). This pertodicity confirms earlier analyses by
cloning and sequencing of integration products and predicts
that integration events occur preferentially on one exposed
face of the DNA helix wound around phased nucleosomes.

72
Harold E. Varms, M.D.

# HygromycinB-resistant colonies

Virus 3T3 cells 3T3_transfectants
AH (iml) 0 confluent
AH (0.1ml) 0 357

AH (0.01ml) 0 80

BH (1ml) 0 30

None 0 0

Fig.2, A 5.5kb quail genomic DNA clone confers susceptibility to ALV infection upon mouse 3T3
cells. A 5.5kb quail genomic DNA clone derived from the ALV-receptor locus and a plasmid
conferring resistance to G418 were cotransfected into mouse 3T3 fibroblasts and transtectants
were selected using G418. 15x10 cells derived from this pool of transfectants were infected with
either subgroupA-specific (AH) or subgroup B-specific (BH) ALV vectors containing a
hygromycinB-resistance gene. An equivalent number of untransfected 3T3 ceils were challenged
with these viruses for control purposes. Infected cells were selected using hygromycinB.
Preliminary analysis suggests that the AH and BH virus stocks used in these experiments were of
equivalent titre. These findings indicate that introduction of the quail DNA clone into mouse cells
confers a high degree of susceptibility to subgroup A RSV vectors: subgroup B vectors are at least
100-fold less able to infect the transfected cells, confirming that the cloned receptor locus is rva.

73
Harold E. Varmus, M.D.

 

PC 12 Cells

Figure 3. The rat pheochromocytoma line, PCI2, was infected with MLV vectors
carrying the v-sre or Wnt-1 penes, plus a bacterial neomycin
phosphotransferase pene as a selectable marker. Colonies of cells pictured

here show the induction of neurite outprewlhs previously observed after
expression of v-stre and the Flattened, adherent behavior of cells expressing
Wnt-1.

T71.
Mutant Autocrine Paracrine Harold E. Varmus, M.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

+ +
+ +
>
> R100Q + N.D.
C121s + =
C143S - -
—— C151S ts N.O.
+ R157Q t! +
N316Q + N.D.
S319C - -
> N346Q + =
> N359Q + (ts) +/~
bet © CIESW - =
™ SEKDEL + -)
-3N + -
-4N + “

Fig.4A Mutants of mouse Wnt-l were generated by site-directed mutagenesis and
tested biologically for autocrine and paracrine transformation of the mammary
epithelial cell Line, C5S7MG, after introduction into CS7MG or NIH3T3 cells in
MLV vectors. The diagram at the left shows the predicted primary product of
translation of Wnt-1 mRNA, with the shaded signal peptide at the aminoterminus
(top). N-linked glycosylation sites are denoted by forked lines; the arrows
indicate dibasic peptides that are possible sites for serine proteases
(probably not used); and horizontal lines show the positions of cysteine

residues. ts, inactive at 40°C but active at 34-37 C; N.D., not done.
7c
Harold #. Varmus, M

 

 

 

 

<+ sip

 

 

 

Fig.4B. Proteins synthesized by the Wnt-1 mutants Listed in Figure 4A were
detected by transient transfection of quail tumor cells (QT6) with plasmid
vectors containing the indicated alleles driven by a cytomegalovirus immediate
early promoter. Proteins were labeled for three hours with 29S methionine and
cysteine before immunoprecipitation with a Wnt-l-specific monoclonal antibody
and subjected to polyacrylamide gel electrophoresis. The unglycosylated and
three most abundant glycosylated forms of Wnt-1 protein are observed in
extracts from cells expressing the wild type Wnt-1 gene; the analysis of
glycosylation site mutants shows that N346 is the site least efficiently used.
(The faint band representing Wnt-1 protein glycosylated at all four sites is
not visible in these analyses.) Also indicated is the 78Kd co-precipitated
protein that we have identified as the BiP chaparonin.

76
Harold E. Varmus, M.D.

    

| Wht-t/int-2
Ai=di

Tumor Free Virgin
Female Mice (%)
-

Q
j

 

 

   
      
 

Qo
Q
i
a
a
a
a
a
a
Rn
f+]
nu
a
a
a
a
a
n
3
é
pM

én

Q
Qo
i

4

ray)
oO
t

p
ao
1

Wot-tAat-2
237

Tumor Frea
Male Mice (%)

nN
a
Ieneerberommend
oO

 

 

Go * ee ™ —t . t * T
Q 2 4 § 8 19 12
Age (me.)

Figure 5. Oncogenic cooperation between Wnt-1 and int-2 transgenes. Sibling
transgenic mice bearing an MMTV-Wne-1 transgene, an MMTV-int-2 transgene, or
both transgenes were monitered for one year for the appearance of palpable
mammary tumors, which were then confirmed histologically as adenocarcinomas.
As indicated in the graphs, the presence of the int-2 transgene shortens the
latency for tumorigenesis in Wnt-l transgenic animals by about two months in
virgin females; an even more dramatic effect is observed in males.
Harold E. Varmus, M.D.

% Tumor Free

   

 

 

MMTV Infection Accelerates Mammary Tumorigenesis
in Female Wnt-1 Transgenic Mice

Virgin Mice Breeding Mice

  

 

 

 

 

00 _ MMTV F. (N=40): 100 * gr ere, 0)
NI. F. (N=10)
80 4 3 80-
. | , MMTV F.(N=42)
60 + . 604
J 7 |
E
o _ 407 MMTV.
| * 7 Wnt-1 F.
20- 20 4 (N=39)
° 0 — a
Age (mo.) Age(mos)

Figure 6. Wnt-1 transgenic female mice were infected with MMTV as neonates,
by intraperitoneal injection with a rat cell line expressing an infectious,
tumorigenic clone of MMTV DNA. The infected animals and uninfected control
animals were observed for the appearance of mammary carcinomas for one year.
The gtaphs show that virus infection accelerates the tumorigenesis by one to
two months; in addition, the number of tumors per animal was generally greater
(up to ten per animal) in the infected cohort. Tumors from the infected

animals contained new MMTV proviruses, often in known proto-oncogenic loci as
discussed in the text.

70