Tip:
Highlight text to annotate it
X
Michael Parfenov: Thank you for the opportunity to present our
results. The Harvard GCC team generates and performs analysis of low pass whole genome
sequence in data for different TCGA tumors. And in our case, low pass means 7x coverage
on average. One aspect of our analysis is the search for viruses and bacteria in tumors,
and looking for possible integration events and mechanisms for such integrations. And
today I'm going to present our results for head and neck carcinomas and bladder cancers.
So why the search for viruses? Actually, it's a well-known fact that viral infection is
one of the important risk factors for different cancer types. A significant proportion of
head and neck carcinomas are causes by human papillomavirus. As far as bladder cancer,
the results are somehow controversial, but several previous studies reported moderate
association between the bladder cancer and HPV and some papilloma [spelled phonetically]
viruses.
So what did we find in our TCGA datasets? We performed analysis of whole genome sequencing
data for 113 head and neck tumors, tumor, and control pairs, and 105 bladder cancer
control pairs. Generally, we could divide the detected viruses into three groups. Different
types of HPV viruses, human herpesviruses, and one case of polyomavirus posted bladder
tumor.
HPV infected 8 percent of head and neck tumors and 4 percent of bladder cancers. And most
cases are represented by so-called higher-risk types of HPV, such as type 16, 33, and 56.
And we do not see these viruses in the control pairs except for one case, but this positive
control case base has also positive tumor pair. And probably this tumor continuation
event like was discussed on yesterday's talks.
So let's look more closely to HPV-positive samples. As you can see, the entire, in most
cases, the entire viral genome presents in infected cells. And the estimated number of
HPV copies per cell varies from one to 30. However, sometimes, we see only 80 or even
less percent of viral genome in the cell. Such stations observed in only two positive
control samples and a couple of positive tumors. And we think that it happens not only because
of a low number of copies per cell in these cases, but, in some cases, it happens due
to large deletions in the viral genome.
So here's a visualization of -- for sequencing data for some HPV-positive tumors, and for
those who are not familiar with IGV browser, each gray rectangle represents a sequence
and read map to reference viral genome, and I want to mention that it's a linear representation
of HPV genome which actually is a circle genome. So you can see here that genome coverage varies
not only across the different samples but also within the same sample. Some regions
are amplified whereas some other regions are completely lost like I mentioned before.
So, okay, we detected virus positive tumors. But the critical issue, at least in the case
of HPV, is the physical [unintelligible] DNA in the infected cells. Normally, presents
there as an episome. However, I wanted to integrate into a human genome, but things
begin to happen. Typically, integration leads to a disruption of the viral E2 gene, which
normally represses viral oncogenes E6 and E7. And as a result, E6 and E7 start to overexpress,
and their products don't relate human tumor suppressors P53 and PRB. So that's why -- and
this leads to malignant progression, so that's why it's interesting to detect integration
sites.
So in order to detect integration events, we look for clusters of discordant read pairs.
We take a bunch of pair and sequencing, and once the discordant read pairs clusters, where
one end of a pair maps to a viral genome and the other end map to a human genome.
So here are some detailed examples of detected integration events. And then besides HPV we
found two integration events for polyomavirus posted bladder tumor. And in one case, polyomavirus
integrates in the FIGN, a gene which involved in mitosis regulation. In the case of ***
virus six, integration happens in the telomeric regions. As for HPV integrations, I want to
bring to your attention the fact that many targeted human genes are actually members
of a different cancer pathways. You can easily recognize some very familiar genes, like NOTCH1,
TP63, RAD51B, involved in DNA reparation; or BCL2L1, which is either a pro- or anti-apoptotic
regulator.
So such results support the idea that integration events probably might contribute to cancer
genetics, not only from a viral oncogene expression, but also from identification of host tumor
suppressors and oncogenes.
So to sum up, we detected integration events in 70 HPV- or polyomavirus-positive tumors,
and two-thirds of them have at least one integration event involved cancer-related genes. However,
we also noticed another interesting fact, that almost half of all integration events
are accompanied by somatic copy number changes. For example, in one bladder tumor, HPV insertion
replaces large part of NOTCH1 gene leading to a heterozygous loss of this region. However,
I think that the most fascinating cases are those where we see amplification of both viral
and human regions. And in at least four tumors, we think that we see such amplification happens
after the formation of circle chimeric viral human episomes. And in the next couple of
slides, I will show you one example of such episome.
So in this tumor, HPV integrates in the gene TRPC4AP, involved in cell cycle control. Here
is a visualization for HPV sequencing, and I want to remind you that these ends are joined
to each other, and here are integration break points. And these regions of viral genome
are involved in integration. And we see that they are amplified, 60x compared with 30x.
So let's look at the human side. This 9KB region of TRPC involved in the integration,
and again, we see amplification of this region, 40x comparing with 10x. Chimeric reads detected
-- chimeric reads suggests that HPV sequence encompasses this human region from both sides.
One possible explanation that we see some very, very complex standing [spelled phonetically]
duplications, something like seven duplications in a row. But the simplest and the most probable
explanation is that we see the actual circle chimeric fragment, sounds like episome.
So we suggest the fuller model of integration event. This region of viral DNA integrates
into the TRCP4 gene, and it -- this chimeric fragment undergoes excision and circularization.
But in such case one might expect to see traces of deletion on the place of excised chimeric
fragment. And actually we do not see such deletion. Is it possible? And the answer is
yes. We think that in the case of chimeric viral episome, we see the same mechanism which
a responsible for formation of double minute extra chromosome observed in different cancer
types.
So there are some probable models of this mechanism. One model is segregation-based
model, and here excision and circularization happen after the replication and after mitosis.
One daughter cell gets deleted copy for gene and another daughter cell gets two intact
corpus plus episome. And maybe such combination confers select for much of cells. So at the
end of the day we see only descendants of this cell, and we've amplified episomes because
episome has viral origin of replication.
Another possible model is the rereplication-based model. And here happen at reparation of deleted
region and the rereplication. And after mitosis, both daughter cells get two intact corpus
of gene but only one gets episome. And again, due to possible selected for advantage, at
the end, we see only the descendants of this daughter cell; we've amplified episomes. That's
why we see many corpus of episomes, chimeric episomes, and we do not see any trace of the
chromosomes.
So, as conclusions, even low pass whole genome sequencing data together with our pipeline
allow -- effectively detect not only the viral presence in the cells but -- in the samples,
but also integration events and mechanisms of such integrations. We detected integrations
in 70 percent of virus positive tumors, and we think that integration events influence
of malignant progression through both human genes and viral genes. And finally, almost
a quarter of all HPV integration events, we think that we see formation of chimeric events,
human viral episome. And that might have some connections with some clinical outcomes, for
example, it could be a cause of different treatment responses. But that requires further
investigation.
So with that, I'd like to thank everybody in the Harvard GCC team, and especially Raju,
Angela, and Jon Seidman. And thank you for your attention.
[applause]
Lou Staudt: So, I'll ask the first question, which is,
do these events -- first of all, is the virus doing this at 100 percent of the malignant
cells? Do we know in some clonal way that this is in all the cells in the population?
And then another question would be, if this can pop in and out of the genome, does it
do that just on once in a given cancer, or do you see multiple events of sort of going
in, picking up genes, and coming out as an episome?
Michael Parfenov: For second question, I think that we see two
such events in one tumor sample. Yes. As far as first question, in some cases, all clones
have this event. But in some cases, I think not.
Lou Staudt: So in those cases would this -- would that
be against the hypothesis that this virus was picking up a driver oncogene?
Michael Parfenov: Probably yes. Probably yes.
Female Speaker: Hi, I was just wondering why you think you
have episomes as opposed to tandem duplications of these integration?
Michael Parfenov: Well, because it's relatively hard to imagine
that we -- that cell undergoes just, for example, exact same -- seven -- the same tandem duplications,
why? And then, what's the purpose of these four cell? The formation of episome, like
formation of double minute extra chromosomes, is more simple explanation. And it's more
likely explanation. It's more likely event in terms of probability.
Female Speaker: So part of the reason I ask is that I have
some data that's not published but shows that we can -- like in the -- in one of the projects
I work on that we have tandem integrations of bacterial genomes, so six tandem integrations
in a fly, and we're pretty sure that they're not episomes. And so I think there's this
notion that these tandem duplications, large tandem duplications, don't occur and aren't
stable, and so I just would put forward that as maybe another explanation. Because the
episome argument looks -- is very compelling until you start having to invoke all these
different mechanisms to get rid of the chromosomal scar [spelled phonetically].
Michael Parfenov: Well, yeah. Actually, that's just a probable
model. The ultimate answer on this question would be FISH analysis. We will see it or
not, but --
Lou Staudt: Thanks.