Bridging Gaps in Infectious Diseases Pathology with Molecular Diagnostic Tools

Bridging Gaps in Infectious Diseases Pathology with Molecular Diagnostic Tools


[MUSIC PLAYING] [APPLAUSE] JOSHUA A. LIEBERMAN:
Good afternoon. Thank you, Steve, for
that warm introduction. It’s my pleasure. I’m lucky to be here. And I’m really
excited to tell you about three projects,
one of which is a longitudinal research
project, that I think is finally in the homestretch. And then two others sort of
assay-development projects, one of which is also, I think,
getting close to wrapping up. And the other is
really just beginning. So I want to start
with a case that has stuck with me
since residency, and really is why this long
research project blossomed. So this was a 67-year-old
guy with relapsed CLL. He was three weeks out from
induction chemotherapy, in anticipation of
CAR T-cell therapy, when he presented with
a fever and headaches, altered mental status
and some falls. And of course in this
patient population, you really don’t
have any neutrophils at this point, which you
can see at admission. Very low neutrophil
count, we’re worried about an invasive
fungal sinusitis. And, of course, by
now, you figured out that’s exactly what this was. So you can see there are
abundant septate non-pigmented hyphae in this tissue. There’s some necrosis
in the background– some fiber in the blood. And then probably most
of his neutrophils are up here, at the
site of infection. And at surgery, I think a
lot of things went right. So he brought in. He had surgery
almost right away. Hyphae, we’re seeing
at frozen section. They were reported. There were no invasive
hyphae identified. On the other hand,
we sort of assume this was an invasive infection. And from the
microbiologic perspective a lot went right as well. Multiple cultures were set up. Multiple PCRs on fresh
tissue were set up. They were all negative. And so, unfortunately,
the patient does not do particularly well. He has an ischemic stroke. He goes to the ICU. The asterisk indicates
where we finally did see invasive hyphae. At this point they’re
in the CNS vasculature, in the cavernous sinus. You can sort of see these
spikes in his neutrophils counts are where he’s getting
granulocyte transfusions. And along the way,
he’s on amphotericin. Eventually, posaconazole. And months later, sort of in a
passing conversation with one of the ID fellows,
who happened to also be taking care of this patient
and was down in the AP resident area to talk about
a different case, it prompted a broad-range
fungal PCR on this. And we identified an
aspergillus fumigatus. He was transitioned
to voriconazole, and was able to go to a
skilled-nursing facility– still with residual
deficits from a stroke. And so I don’t know that, if we
had identified this organism up here and transitioned
him to voriconazole, he actually would have
had a better outcome. But the reason I think
it might be possible, and the reason that this
sort of stuck in my brain and haunts me a little
bit, is because there is a seminal paper in the
early aughts that identified a significant survival
benefit for patients with invasive aspergillosis
who are started on voriconazole and not amphotericin. So this was a randomized
controlled clinical trial across multiple sites, multiple
hospitals, multiple countries– almost 300 patients. And there was a really
significant survival benefit. 53% of people who were
started on voriconazole were alive at 12 weeks, as
opposed to only 31% for those who received amphotericin. Still a pretty dismal
survival with this disease. Now, the idea that you
should treat aspergillus with voriconazole up front
stands in stark contrast to invasive mucormycosis,
where amphotericin is the mainstay of therapy. And in another seminal
paper from the mid-aughts, out of MD Anderson,
the authors identified that if by six days after
your onset of symptoms you start amphotericin,
your mortality is about half compared to if you start later. So at four weeks, it’s about 35%
percent, about 49% at 12 weeks. And that’s in the
appropriately-treated group, where amphotericin
started early. It gets up to the 80%
range at 12 weeks, if you don’t start it in time. And so this is why being able
to identify what fungus you’re dealing with in cases of
invasive fungal disease is particularly important. Before we talk about the
identification steps, I want to just look a little
bit at the patient population. So the initial patient
I was talking about had very low neutrophils. But how low does your neutrophil
count really have to be? So this is a recent study that
came out of our colleagues here in the ENT– Greg Davis, Ian
Humphreys, as well is Rod Schmidt, mostly
retired from pathology. And there are just a
couple of key points to which I want to
draw your attention. The first is that a third– a third of their patients
had absolute neutrophil counts greater than 1,000
cells per microliter. Normally, we think of the
patients being at risk as having less than 500. Interestingly, having a
higher neutrophil count really didn’t correlate with
improved survival. You can sort of get
a sense that maybe at three months in this
high ANC count group there was a little bit
of a survival benefit, but it really didn’t cross
into statistical significance. There were a couple of factors
that did predict outcomes. So having what these authors
termed atypical molds– fusariums, scedosporium,,
alternaria, for example– it was correlated
with a poor outcome. And getting surgery
was correlated with an improved outcome. So these are both surgical
and medical diseases. So in terms of
identifying fungus, the reason we want
to have a sort of broad-range,
organism-agnostic approach, culture, or PCR,
is because trying to do this based off
of histopathology is highly error prone. There’s now a well-developed
body of literature around this, including a couple
of important papers. But this one out of Stanford, by
Sangoi, is one of the better– one of the more well-known. And one of the important
findings is people tend to– the pathology lore is that
one of the reasons you can misidentify
things is patients may have gotten
antifungal therapy, and this may change
the morphology. And actually, that’s
probably not the reason we misidentify them. The reason is they
just look a lot alike. And so there are
evidence-based guidelines that have been developed. Our colleague Jeannette Guarner
at Emory, and Mary Brandt, have published these
in a wonderful review. And so we try and
break down the moulds into septate hyaline moulds,
which includes aspergillus and others. The pigmented moulds
or the dematiaceous moulds, and that includes
organisms like alternaria. But frankly, you often
don’t see the pigments. And so these two groups, there’s
a fair amount of crossover. And then finally, there
the pauci-septate moulds, which really don’t
have septations, except at branch points. And those are the mucorales
or the zygomycosis. So we try and put these into
groups, into broad categories. But it still matters, even
within the septate moulds, if you can get an ID
yeast, you try and break down by small, medium, and
large, and give a differential. So what if you say, well, Josh,
I’m going to play the odds. Aspergillus is one of the
more common organisms. Why can’t I just say this
is a septate hyaline mould, so you should probably be
thinking about aspergillus? And the reason is this is
data that we collected from the paranasal sinus sites,
that’s a mix of reference-lab work and internal patients
for PCR and internal patients for culture– this is from 2011
through mid-2016. Aspergillus doesn’t
even constitute a plurality of cases,
including multiple species. And so a fume ranges
between about 12% and 23%, depending upon which
assay you’re looking at. But if you include
all aspergilli, you’re probably only getting
up to about 35% of cases. There are two other important
features I want to point out. One is that sometimes we just
can’t get an ID by culture. I think this is less
common, now that there’s fairly close collaboration
between the mycology lab and the molecular lab. So we do a fair amount of
identification by sequencing. The second, not shown here, is
that about 12% of these cases are polyfungal. So the other big
pitfall, of course, is that you may see something
that suggests one organism, and completely miss
the other that’s hiding in the background. And it’s very humbling
when that happened. I’ve made that mistake, and
I know many of my colleagues have as well. And it’s hard, because there’s
a patient on the other end, and you’re really just trying
to get them the best care. So sort of putting
all this together, I developed a
number of questions that just wouldn’t
leave me alone, about how we can use PCR
to really bridge what happens in the clinical lab. And as I showed you
in that first case, sometimes we don’t have an
answer for our colleagues on the clinical side of things. And how do we bridge
that with what’s in the histopathology
lab, where often we have diagnostic material. And so some
important questions I was hoping to answer–
what are the advantages or drawbacks of PCR,
and can we really see if this makes a
difference in patient care? And finally– I think this is
a really important question– whom do we test, and when? And so I’ll try and
answer some of those as we go forward through
this first part of the talk. First, let’s talk a little
bit about what’s known. So this was a work that came out
of Stanford at the end of 2017. They have a very similar assay
to our broad-range PCR assay. And there are a couple of
points to make from this. They sort of had
a two-phase study. First they took 60
known positive cases, where they had an
organism ID by a reference method, 57 cases that were
negative by reference methods. And they use that to test
sensitivity and specificity of their assay. They followed that up with
116 culture-negative patients, all of whom had suspected
invasive fungal disease. So you can see high
sensitivity and specificity. The true specificity
is probably higher, because one of their
negative cases was actually, they believe the PCR
result clinically. And of their positive
cases in this phase two part of the study, almost
60% of those PCR results led to a change in management–
mostly in medication. In a couple of cases it changed
surgical management, as well. So there are two
other key points that I take from this study. One is that the submission
type has an important impact on diagnostic yield. And so what they
found is that if you had an open surgical
resection, where presumably you had bigger tissue, you had a
much higher diagnostic yield– somewhere in the 70% range. Smaller specimens, like
core-needle biopsies, were somewhere around 50%. And they really
didn’t see any yield from fine-needle aspirates The final point that I want
to make about this study is when they compared
the diagnostic yield from FFPE tissue– formalin-fixed paraffin embedded
tissue from the histopathology lab, from their
anatomic-pathology colleagues– performance was about the
same as for fresh tissue. And so we’ll come back to
that and try and explain why. It’s become sort
of popular to refer to some of these broad-range
fungal assays as universal PCR. I think in general, when I
see this in the literature, this means at least two of– usually, University of
Washington’s– broad-range PCRs were ordered on a
specimen– that’s a mix, or all three of the broad-range
micro-bacterial fungal, or bacterial PCRs– often with next-gen
sequencing as a reflex. And so this was a
nice small study, but with really
detailed chart review, in which for all
of their samples, they identified a diagnostic
yield of about 30%. And in about a quarter
of all of those samples, they made a significant
clinical change. Now, that included both positive
and negative results– which I think not looking
at negative results is a shortcoming of
some of these studies, but can be quite informative. The other thing
that I’ll point out is that having corresponding
indicia of inflammation, or ideally organisms
seen by histopathology, increases your diagnostic yield. This is the gospel of
pre-test probability. I don’t think that’s
a surprise, but I do think it’s important
when you’re thinking about what specimens to select. And I think it also
goes to the notion why FFPE specimens probably perform
about as well as breast tissue. In this study, they
used a clinical judgment as their gold standard for
whether patients really had an infection or not. And they estimated
sensitivity of our assays– this group of three PCR
assays– at about 51%– specificity at 94%. And you see the positive and
negative predictive values. So I want to look at
one more study before we get into our data, because there
are a couple of key points. This has just been accepted
in the Journal of Clinical Infectious Diseases. And again, same
sort of situation, where a universal PCR is a
sum combination of our three primary broad-range assays. And what I really
take from this figure is that anatomic site
makes a difference. I think this is
probably more to do with what’s most
likely to be infected, as opposed to anything
intrinsic to the tissue. So this is concordat with our
anecdotal sense in our lab. But heart valves tend to be
have high diagnostic yield– also, pleural fluid. Interestingly, in this data
set, some of the nasal specimens only had about a 10 to
12% diagnostic yield. But most of those results
changed management. So the study only looked
at positive results. And of those positive results,
about half changed management. And like in the
previous studies, the presence of an
organism or inflammation on some sort of concurrent
study increased the probability of diagnostic yield– although
they didn’t quantify that. And they found that
the yield from FFPE was similar to fresh tissue. They didn’t look at volume
of specimen in the same way that the Stanford study did. But they thought that
tissue being submitted, in which category
they lumped FNAs was more likely to have a
diagnostic yield than fluid. Again, that may speak
more to selection of tissue for submission
rather than anything else. So I want to turn our
attention to a study that I started a number of
years ago as a resident– and really has been
completed by Alex Mays, who’s built a really wonderful
set of informatics tools, along with Patrick
and Nathan Breit, that I’ll tell you
about as we go along. But we pulled data
from the LIS, from 2011 through 2019 inclusive– which represent for all fungal
testing that we’ve done. It includes about 38,000
broad-range fungal PCRs, 28,000 fungal cultures,
representing nearly 50,000 distinct patients. We excluded the vast
majority of these, because they are not from
the paranasal sinus sites. And we were trying
to limit our focus. And then when we just look
at sinonasal specimens, we see there are about 640
fungal PCRs, representing nearly 600 patients,
427 cultures. Really all of those come from
the inpatient setting here at UW, or one of our
associated institutions. So for our inpatients, there
were only about 55 patients, which is a tractable number for
chart review, which is great. And then they had about 78
fungal PCRs between them. So let’s try and
march through this, and see how our
assay performs in a high-probability-of-infection
setting. So first what we noticed is that
the positivity rate is higher for broad-range fungal PCR. I was actually surprised
that it was almost 50%. It came in just at about 49%. Culture also came in higher
than some preliminary data I collected, right around 20%. And so we asked the question,
is this reflective of all tests, or is this really something
about our patient population? And so the way we
went about that was to say, all right,
here on the left– so this graph is all
patients with samplings from the paranasal sinuses. And so that includes our
reference-lab clients. For all broad range
fungal PCRs, right about 10%, which is
consistent with some of the previous
literature I cited. But in the sinuses,
even when you look at our reference-lab
clients, we’re up around 36%. And so the way I
make sense of this, particularly now that I’ve had
a chance to sign-out a bunch of these cases and review the
pathology reports that come in with them, there
are a lot of specimens, particularly in this– this is all sites– but
in the non-sinus category that come in. And there’s granulomatous
inflammation. It’s a nice dense
epitheliod granuloma. And the suspicion is sarcoid. And the clinicians
would like to do broad-range fungal
mycobacterial and PCR testing, to rule out an
infection– so they can put the patient on steroids
and treat their sarcoid. And so that emphasizes that
while I can’t quantify it yet, these negative results are
actually still oftentimes quite informative,
and also explains why the sinus
population, I think, has a really high
positivity rate. I don’t think of sarcoid as
being particularly prominent in the sinuses. But I do think if
you have a reason to order a broad-range fungal
PCR on a sinus specimen, you’re probably thinking about
someone who’s quite sick, and may have an invasive
infection there. So we divvied this up a
little bit differently, because we did have
some outpatients listed. The outpatients– so
you see the counts here on the left and the
proportion positive more clearly on the right. So obviously, the
bulk of our patients are from outside clients. We have a small
population from here. And then we have
a few outpatients. Their PCRs are not
positive as frequently as in the inpatients. And again, I think this just
goes back to who’s sick. And presumably if you’re getting
sampled as an outpatient, you may be suspicious for an
invasive fungal infection, but you may also be sampled
for an allergic process. And so we’re probably
including people who don’t have disease
in here, but where we want to rule things out. As opposed to the inpatients,
we’re probably actually quite worried. So I want to turn our attention
to this question about formalin fixation, because I think it’s
a really important question. We’ve certainly done
experiments in the lab where we’ve seen that prolonged
formalin fixation time destroys DNA. This is, I think,
a known phenomenon. It’s a complex
biochemical process, that leads to DNA
cross-linking and shearing, and reduces PCR yield. And so we tried to break this
out, first, by count, and then by proportion. So same data displayed
in two different ways. And what you see is that
actually it’s about the same. This difference in
percent positive is not statistically different. It’s about 42% positive
in the FFPE samples. It’s only about 35% in
the fresh specimens. And so I think the
question is, how do you overcome the damage caused
by formalin fixation, or other pre-analytical
variables? And so I think if
you’re a pathologist, or if you’re working with a
pathologist, you get to say, oh look, block A1 has
not many organisms. Block A3 has a
bunch of organisms, but it also has bone–
and went through a strong acid de-calcification,
which is much worse for DNA, I would argue, than formalin. And so I’m going to block A2,
which has plenty of organisms. So I really think that’s
one of the key links between the molecular
assays, the clinical lab, and the anatomic-pathology side. So one of the hard questions
to answer with this process is, who wins the race? Is it PCR, or is it culture? And so the big caveat here is
this is all final-result dates. And I think we’re all
pretty accustomed to having preliminary results come out
well before the final dates. So we see the negatives and
positives, in red and teal respectively, for the
broad-range fungal PCR and our fungal cultures. And we’re several days faster
by final result on the PCR. There are a couple of
other complexities here. One is, I haven’t tried
to take into account how– I mentioned how the
prelims come early. In the case of the
fungal PCR, it’s probably anywhere from
six-to-36 hours earlier. I think that range is
much broader in culture, where sometimes rapid
growing moulds– we can give a presumptive
diagnosis quite early. I also think there are
probably some outliers that are positive out here,
which may be because we’re waiting to final them until
we have some sensitivity data back. But I have to dive into some of
those cases a little bit more. And then the other sort of big
elephant in the room that I haven’t touched is, how do
other assays that we run in the molecular-micro lab, that
we often set-up up front with broad-range PCR– like our
aspergillus-fumigatus-specific PCR, or
zygomycete-group-specific PCR, how do they factor into this? Because we result those– or they’re able to be
finaled about two-days faster than the broad-range fungal PCR. They don’t have the same
breadth of organisms that they can detect. But for the organisms
that they do detect, they’re more sensitive. And so in terms of how to
piece all these together into a better model
of how we actually should be algorithmically
testing patients, I don’t have a great answer yet. One of the things I do want to
point out about turnaround time is that, while turnaround time
is pretty constant for culture over time, we’ve gotten faster
in the molecular-micro lab. And so it’ll be interesting
to see if this drop in about 2 and 1/2 days turnaround time
continues, as we’ve now– [? Tatiana ?] [? Vlasic ?]
and Dan [INAUDIBLE] have built a really phenomenal
informatics architecture, that allows us to track,
manage, and sign-out our cases. And my suspicion is
that over time this is going to shave off a little
bit more time from our results. So for those 50-odd patients
who are UW inpatients, I was able to dive in
and do a chart review, and tease out a couple of, I
think, really useful findings. The first is that
positive PCR results led to a medication change
about 18% or 19% of the time. PCR was the only positive
diagnostic test in about 9% of cases. And so if I understand
how to do this correctly, that means the number
needed to test is about 11. I think that’s a fairly
conservative estimate, but a reasonable
first-order approximation. I’ll come back to that
on the next slide. There are a couple of
points I want to make, which is that negative
testing is still valuable. So there were a
number of cases– about 12% of all
the tests ordered– where the clinicians
were more comfortable not adding antifungal
medications because the testing was negative. Some of that includes
culture results. But one of the
big advantages you have from the
broad-range fungal PCR is we result it
potentially weeks faster than a negative fungal culture. So there are two cases
where that were deemed to be colonizers or bystanders. And this is really an
issue of clinical judgment. In one case, they were
able to de-escalate, and the other they did not
add additional therapy. And I think it was
really striking. Because you read
the notes and it’s pretty clear that this patient
is not immunocompromised. They’ve been using afrin a lot. And so that probably explains
the local necrosis in the nose. And so then you have super
infection or colonization by some of these
potential pathogens, and you just have to watch
the patient clinically. The other thing I’ll point
out is that this is still not good for patients. The 30-day mortality, just
looking at these cases, is about 14%. So I want to talk a little
bit about that number needed to test. Because one of the questions
is, how much does it cost? And obviously, I hate to boil
down a patient’s life to, how expensive is
it to treat you, but I think that in a day when
resources are constrained, it’s important to at
least think about. And so the cost of 11
fungal PCRs is about $5,000, in 2019 or 2020 dollars. In 2017 dollars, the cost of
one day of a regular hospital patient admission in
Washington state– on average, hospital expenses– is about $3,500. That trend– the rate of
increase has not abated. And it strikes me that shaving
off one regular inpatient day is about the minimum
impact you can have. And so I can’t yet attribute
changes in length of stay to whether patients got
broad-range fungal PCR, but if that’s the minimum,
I think we have real opportunities to at the
very least break even– to say nothing about
getting patients on the appropriate therapy
early and potentially improving mortality– and to continue to
be crassly financial, protect the $200,000
to $800,000 investment in their leukemia care. So I do think there are
potential for cost savings. And it does make financial
sense to add these up-front, particularly when we look at
the ability to catch infections we might otherwise miss. So I’m hoping to answer
those questions with help from our fabulous
informatics team. So Nathan Breit and I
just spoke yesterday about how to make
use of this data, that he and Patrick were
able to able to pull– in an architecture that
seems to dovetail well with the electronic
data warehouse. And maybe like the analysis
package that Alex Mays built, is reusable for other tests
and other anatomic sites. So Nathan pulled a mix of
emission discharge and transfer data, from which will
derive length of stay data– and sort of a robust look
at medication changes. And the goal is to compare
our 255 patients who only received culture to
those who only received PCR. And one can imagine also doing
this for just crude aggregate hospital costs, to try and see
if this single variable makes it makes a difference. So I want to wrap up
this part of the talk by really emphasizing
that I think broad-range fungal PCR
does not replace culture, but is a really important
high-yield component, particularly for
high-risk patients. And so this may be the
only diagnostic result. And working with the
pathologist to make sure that you pull tissue from
your anatomic-pathology lab is a really important way
to not miss infections. Just to reemphasize–
medication changes do happen with positive
results, but negative results are informative too. So I want to move on,
to talk about an assay that we have in development,
where I think we’ve made some really nice progress. And this is really
sort of thinking about what happens when small
intracellular pathogens look alike, and you’re trying
to tease them apart. Or, as in this first case,
when you actually don’t see anything. And so this was a case
that came through when I was a resident, then on CP. And this was a
30-year-old guy, who was returning from the Peace
Corps in South America. He had a history of
latent TB, for which he had claimed he’d been treated. He also had an extensive
travel history, throughout South and Central
America and Australasia. No history of cutaneous lesions. No other symptoms. And this mucosal lesion
in his nasal cavity had been there
for several months by the time he presented
to our ENT docs– who were thinking about
chronic corrosive processes, like rhinoscleroma which is a
chronic Klebsiella infection, or mycobacterial infections. And this pathology is actually
not all that informative, although it is striking. There’s a rich influx of
lymphocytes and histiocytes, which we see here. There’s a little
bit of necrosis. That’s sort of patchy. But nothing really diagnostic,
even after multiple rounds of culture, multiple biopsies– the whole panoply
of special stains for microorganisms in the
anatomic-pathology lab. And so ultimately
an astute ID fellow was examining this
patient and palpated an enlarged cervical
lymph node– at which point she
said, I know he doesn’t have cutaneous lesions
and claims to have never had them, but he has an
erosive mucosal lesion. He’s been in an
endemic area, and he’s got an associated
lymphadenopathy. I think this is
mucosal Leishmaniasis. And about three months after
he first presented to us, the CDC came back having
detected Leishmania braziliensis in this patient. So you may be more
familiar looking at slides like this, where you can see
amastigotes running around inside of histiocytes
although it’s not unreasonable to also consider
something like histoplasma in the differential. And in the first case
I just showed you, even though we did
broad-range fungal PCR, we did not detect Leishmania. In this case we did. We reported as an
incidental finding. The 28S locus that
we use is a region of fairly conserved sequence. And so that’s how we usually
detect these organisms. The flip side, though, is
that locus is conserved enough that we can’t
identify the species. And so that’s why I’m focusing
on mucosal Leishmaniasis. So I want to just take a minute
to introduce that disease. Mucocutaneous
Leishmaniasis is usually thought of part of a spectrum. There about a billion people
who live in endemic areas. There have been about a million
cases over the last five years. And mucosal Leishmaniasis
can not only be disfiguring, but places patients at
risk for super infection. And so on the right, we see
sort of a classic cutaneous Leishmaniasis lesion, which
can either progress or present, without a cutaneous lesion,
into some of these erosive, highly-destructive lesions– which are, if nothing
else, disfiguring. This is actually not
an extreme presentation of mucosal Leishmania. I had the opportunity to see
a number of mucosal cases when I worked in Brazil,
in a strange sojourn we can talk about some other time. But this is not an
unreasonable presentation. So the two other
things that I want to draw your attention to as
I introduce this disease– first is, so there’s
this sort of mucosal belt across northern
South America, where mucosal Leishmania or
Leishmaniasis is more common. The other thing, as
you can see, it’s distributed around the world. And these different colors
indicate different species. And I don’t want
you to pay attention to which species is where
necessarily– simply to note that there is
significant overlap between their
geographic regions. And so that’s not enough to tell
you what species of Leishmania the patient has. And actually the species
that the patient has matters quite a bit,
in terms of treatment. There are other factors
that are involved in deciding whether
or not you’re going to treat the patient. How big is the lesion? How many lesions are there? Are they immunocompromised? Do they have visceral disease? I’m not going to talk
about kala-azar today, but it’s highly fatal,
so it’s always treated with pretty nasty drugs. And then the cutaneous disease– so we can really divide
our Leishmania species into either old or new world. And the old-world leishmania,
I think the treatment decision is really, is it
Leishmania tropica, which is more likely
to resolve on its own– although expert opinion
diverges somewhat. And then within the
new-world species, subgenus viannia is
the real question. This includes Leish
braziliensis, guyanensis, and panamensis, which are
much more likely to produce this mucosal disease. They have this mucosal tropism. So we wanted to
develop, ideally, a single-locus PCR assay,
that could distinguish these different species. And so when Kendal Jensen,
our Molecular Pathology fellow was on rotation with us, he
did a really helpful literature review, and identified
several candidates. The other features of
this potential assay would be that you have flanking
conserved primer-binding sites, with an intervening
region of high sequence and potentially
length polymorphism. And it has to work across
a pretty wide range of, from a PCR perspective,
fairly hostile conditions. So for comparison E.
coli is about 50% GC. Human on average is about 41%. True leishmania genomes
are up in the 70% to 80% range, and the viannia subgenus,
down around a scant 60%. So really kind of a wide
range of molecular conditions to deal with. And so the target
that we thought had really good phylogenetic
discriminatory power was something called the
mini-exon repeat gene, or the spliced-leader gene. And this is both cool biology
and a couple of useful factors as we designed the assay. So it’s, first of all, arrayed
in about 100-to-200 tandem repeats, which means there’s
a sort of antecedent template amplification step. So a single genome
has multiple targets. They’re highly conserved,
potential primer-binding sites. And what’s really cool is this
39-nucleotide, highly-conserved region gets transcribed
and then transspliced onto the five-prime end
of other mRNA species throughout the organism–
which is just really kind of cool biology. It doesn’t really
influence what we’re doing, but it is really kind of cool. So we were lucky enough to
have one of our talented MLS students, Karissa Crawford,
join us in the lab. And in a short month,
she was able to take this from concept to really what
seems to be a working assay. And I should point out that
these many of these strains we either purchased from ATCC,
and then cultured and extracted DNA from those promastigotes. Or we were kind enough to
get cryostocks for culture, or extracted-genomic DNA,
from two of our collaborators in infectious diseases, Wes
van Voorhis and Fred Buckner. They and their staff
have been really generous with their time and their
reagents and their expertise. And so you can
see where it falls in the gel that’s over
here, that I’ll walk through in just a second, what we
estimated the amplicoon length to be and what
we expected the amplicon length to be. And so there’s a
range of intensity of bands, which
probably inhibited our sequencing– which I’ll
talk about in just a second. But one of the things that I
really thought was interesting was for this L.
donovani species, that we bought from ATCC. It was 300 bases
when we amplified it out of the
genomic DNA– when we amplified the mini-exon gene. And that’s significantly
lower than we would expect. And so at this point
it was about the end of Karissa’s rotation. I was getting ready
for this talk. And so we’re able
to get the sequence in a clinical molecular lab. Ichih McGuire was able
to find time in between– I’m very impressed–
reviewing cases for NGS. And I think we
have some answers. So the first is that we just
picked a one-size-fits-all dilution for these. And so I think we wound up
both over and under deluding them, which explains a lot
of the poor sequence quality that we see, and some of
the mis-identifications. But mostly we’re able to
do a good job identifying. So let’s just walk
through this briefly. The positive control, we took
the mini-exon gene from lizard Leishmania– L. tarentolae. And I changed every fifth base. Synthesize that in
a plasmid from IDT. And sure enough,
our positive control was 81% identical to
lizard leishmania, 100% identical to
our positive control. So great, that’s working well. There was high sequence quality. We had one sort of
challenging identification, where this L. major strain was
either L. Major or Leishmania mexicana. I’m a little bit curious
to dig into the strain and just make sure it’s
accurately annotated. The main theory that we had
was this L. tropica strain was identified as either
L. major or L. Mexicana– but again, low sequence quality. So it’ll be interesting
to see if we can troubleshoot the
sequencing and resolve this. The L. Donovani strain, that
didn’t make any sense to us, turned out to be
Leishmania amazonensis So we were working with
the Leish amazonensis. And we sort of went through the
standard lab medicine resident call– when could the specimens
have been swapped? s and I were very
careful doing this. I really don’t think
that we swapped them. We do have other specimens from
other points in the culture, and so we’ll do some
testing of this. But ATCC performs a really
incredibly valuable task. But bless their hearts,
they handle a huge number of strains. And getting a strain that
is mis-annotated from them is not unheard
of, nor is getting a micoplasma contaminated cell
culture strain, or a HeLa-cell contaminated strain. It’s a big problem,
starting in the 1960s. So I think what
we’ve done, actually, is just found an error
in ATCC’s catalog, that we can now correct–
and sort of proves the point that I think
this assay is working, or will be working soon. The limit of detection
is extremely good. So we have some more
replicates to do, but we’re getting down
to two-or-three genomes per reaction. And I think our validation
plan is interesting. We’ve been able to
bank some DNA, where we’ve been able to identify
incidentally leishmania in the clinical lab. We’ll also send some of
our promastigote pellets through our extraction process. And then we have a
number of FFPE cases, where we either weren’t able to
identify a leishmania organism, but it was identified at
the CDC in that specimen, or in another specimen
for that patient. We still have some
residual genomic DNA samples that we have extracted. And then there
are two FFPE cases that I’ve seen recently,
where leishmania cropped up in the differential. But I will be shocked
if that’s what it is. And so I want to know. So I want to move on to
the last sort of vignette that I have, in the
last few minutes, which is what happens when
I don’t get a PCR result, there’s fungus and tissue,
and I want to try and make a diagnosis. And so I’ll start again
with this one case, which was shared by me with my
colleague and teacher, Larry True. And for those of you
who train with me, you might be wondering
what on earth this is doing in a prostate. This was a cystoprostatectomy
from a 62-year-old guy with a high-grade
urothelial neoplasia. And there was one block of
the prostate that was full of granulomas, including
a number of these really well-preserved yeast forms– which in several
cases appeared to have broad-based budding, a nice
double-contoured cell wall. And they ranged in size
from about 12-to-23 microns. And so this sounds
like blastomyces , but it’s really hard to rule out
cryptococcus– particularly C. gattii, given that he
doesn’t have a travel history to a blastomyces-endemic region. It’s also hard to rule out
definitively coccidiodomycosis, although it is a
bit small for that. And blastomyces is more
likely to disseminate. At any rate, the
patient did not have any lesions in their
lungs, and was not going to receive
systemic chemotherapy. So despite the fact that
we didn’t get a PCR result, we wound up not pursuing
species-specific PCRs for either crypto or cocci,
which we thought about. And ID was sort of comfortable
calling this probable blasto, but they recognized the pitfalls
of trying to be specific– even with some of these
sort of classic features. So thinking about this case
and other related cases where I don’t know why we
don’t have a PCR amplification, I’d really like to develop this
multiplexed fungal FISH assay, where we can encode on
species-specific nucleic acid probes information using
combinatorial fluorescent labeling. And so what I mean
by that is basically if you mix up any combination
of these six fluorophores, you can get 15
possible combinations. And so you can pretty
rapidly expand the number of organisms that
you can identify, or the number of targets you can
throw at a sample all at once– and image across
multiple spectra. And so you could think
about a couple of ways to deploy this– either
as a single pool, or as sort of multiple
cascading pools of fish probes. This has been done in
bacterial colonies. So on the left–
this is not my work– semi-dispersed dental plaque. This is the fused image of
imaging in six channels. And then after linear and
mixing and sort of decoding these combinatorial labels,
you can then assign taxons and get a false-color image. And you can actually see
where the different organisms are interacting. So this is useful on
the basic-science side, trying to understand how
complex communities interact. But I think for us
it really becomes a potentially powerful
diagnostic tool when PCR doesn’t work. So have folks done this? And the answer is, yes. So first in the clinical
microbiology lab, we have FDA-cleared
kits that we use, either to detect mycobacteria,
as we do here. Or other labs will
sometimes use these for detecting certain
bacteria, or endemic yeasts. Going into FFPE tissue is
a bit more challenging. But folks have worked on at
least genus-specific probes for the last 20 years. And high specificity is
certainly achievable. I think we can do better. Work in Steve Salipante’s lab,
I think, is getting to that. And then the sensitivity
is also relatively high. Probably not as high as
PCR, so not a replacement. But I think filling in
a gap, where sometimes all of your tools don’t work. There are other ways to
sort of biochemically enhance the binding, as well
as to enhance the signal amplification, so the
pathology presents. There will be a fabulous
speaker, Brian Beliveau next week at 4:30, who’s
talking about saber fish– which is one of these possibilities
for chaining together different molecules, to
encode more information and to amplify the
fluorescent signal. So I will be there. And then I think the
reason this hasn’t taken hold in anatomic-pathology
labs is, one, you have to see enough
fungal disease for this to really be useful. And two, it’s
just, I think, been eclipsed by the excitement
around PCR, followed by sequencing, and now
next-generation sequencing. Just a couple of
very brief examples. So Kathy Montone’s
group at the Penn developed both
aspergillus, fusarium, and broad dematiaceious probes. This is not fish, but ISH,
much like our EBV ISH probe that we use in hematopathology,
that many of you are probably familiar with. And so in H&E, you see that the
aspergillus and the fusarium both look a lot alike. So having some sort of
organism-agnostic test is really important. The aspergillosis
ISH probe readily detects the aspergillus hyphae,
but not the fusarium hyphae. There is a little
background staining– I think probably just
from a counter stain. And so she’s published
this in a number of ways over the last, well, decade. And then about a decade
ago, our own David Fredricks at the Fred Hutch worked
with a European collaborator to compare fish in FFPE tissue
to a broad-range fungal assay that they had in their lab,
that’s similar to ours. And they found that the
fish worked, but was not as sensitive as a PCR. And there are about
three of these 40 cases in which fish was
positive and PCR was negative. And in all of these cases, the
histopathology was positive. So I think there really
is a niche for this assay. The way we’ve gone about
trying to start this is with collaborating with my colleague
Lori Bourassa and mycology staff is to collect interesting,
clinically-relevant moulds– cut them up and embed them
as formalin-fixed paraffin embedded tissue. And so you can see sort
of the range of organisms. A bunch of different
aspergilli, in case we really want to
get organism specific and not just leave it
at the genus level. T. marneffei, and penicilliium,
fusarium, alternaria. And then a slew of– these are mostly
dematiaceious moulds. And I’ve used the pathologist’s
trick of inking them, so I can tell them apart
and trace them back. And so I think this is a
really nice library of tools that we can then use
to develop this assay. I will also say it is really
challenging to cut FFPE tissue. So for residents, be kind,
cut your sections thin, and remember that this is hard. So appreciate your
histopathologist. You may be able to see
that these have been cut. And so we have slides,
we have probes in-house, and we’re getting
going on that project. In terms of the use case,
I think there are really two opportunities. One is when PCR fails. As I mentioned,
the other, you can imagine that this
could be useful as sort of a perioperative type
of rapid diagnostic. So the Montone group has been
able to get the staining time down to three hours or less. So you’d still have to
be able to cut them. But one could imagine having an
intra-operative consult, where frozen section is positive,
and then moving very quickly to the sort of fish assay. Again, patient selection
would be important, but just trying to
think ahead and decide when we might use that. So I want to close
by just making sure I acknowledge people. And I apologize, I
probably let people off. But I’ve highlighted
Alex and Karissa and Kendal’s involvement. Andrew Bryan was really
instrumental in getting the fungal-utilization
PCR off the ground. Patrick and Nathan
have been really essential in getting a lot
of the informatics support for that. And then very helpful
conversations along the way with Karen, Brad, Dhruba. And then Rod in AP,
Ian and Greg in ENT. Dhruba and Kyoiko
have really just been essential sources
of support, both for me clinically as I get involved
in the molecular-micro lab. And Kyoiko has been
really essential in both of these projects,
making sure we have DNA for validation studies,
reagents to build our assays. And then I mentioned
Ichih the sequencing. Steve and Brad have been
involved in this project, and Steve on a little
bit on the fungal fish project– more so in the future. And I want to recognize our
collaborators, Fred, Wes, and folks in both their
lab and in Steve’s lab. And then finally, on
the fungal fish project, I’ve been working with
Frank in Behzad’s lab. And Behzad Najafian
and Ram Akilesh, my collaborators in AP,
have been kind enough to share their time
and expertise with me. So I’m very grateful. I want to close by
just acknowledging the molecular-micro lab. This is a really fabulous team. I’m really lucky to be
a part of this group. I think they turn out
great results every day. And with that I’ll close
and take questions. Thank you. [APPLAUSE] AUDIENCE: So in
the first section you said negative results is
really useful, or some people it’s really useful. But you also said the
size of the sample makes a huge contribution
to the meaning of the data. So do you have
anything more to say about negatives in tiny samples
and negatives in big samples. JOSHUA A. LIEBERMAN: Yeah. So Sean is asking
about, when do you determine whether a
negative is a true negative, and is therefore informative
because you’re ruling out infection, or it’s just
too small a sample, and so you didn’t
get an amplification? I don’t have a good answer now. I think if you have
pathology to look at, that’s really informative. And so if it’s a small sample
but there’s no inflammation, or it’s something that
looks like, I don’t know, lung rejection, for example, I
think it’s much easier to say, that’s a true negative. There have been a couple of
cases where the first AFB stain that we got on a case had
acid-fast organisms. . But the pattern of
inflammation didn’t really fit. There’s nothing in the PCR. And so then we went
back and looked at it, ordered a follow-up AFB stain,
which was dead negative. And so we really think that that
was contamination of the stain. So I think trying to put all
those together really matters a lot. One of the questions
that I hope to look at is that precise question. We record the size
of our specimens. And so you could also imagine
that it goes the other way. So that prostate case that
I showed, one of my hunches is it may have been
negative because it was such a big sample. But maybe we just
gummed up the columns, and it needed a
manual extraction. So I think there are pitfalls
in both ends of that spectrum. And hopefully the next
time we’re in this venue, I can answer that for you. It’s one of my
questions going forward. Thank you. Other questions? Susan? AUDIENCE: That was
a great talk, Josh. In your first
part, you mentioned that discarded a lot of the
data because it wasn’t sinus. And so I wondered
if have you have plans to go back to those data. JOSHUA A. LIEBERMAN: Yeah. So Susan has asked whether
the data that we didn’t look at– non-sinus sites– is informative, and what are
we going to do with that? And I think there are a number
of really immediate questions we can ask. I’m really interested in
other pulmonary sites. And thinking about if you have
an invasive fungal infection, or just fungus detected
in your perinasal sinuses, is it in your lungs too? And there is some
literature around that. I think we can ask
questions about, for the aspergillus PCR,
what’s the utility in that? Because it is a very
fast turnaround time. It’s used a lot in BAL
and in pleural fluid and on lung specimens. So I think that’s another really
important question to ask. I think there’s a wealth of
mycobacterial data, which was not included
in that data pull, but could be subjected to
the same sort of analysis. So yeah, there’s a
lot to mine in there. I’m sorry I don’t
have more information to tell you about it now. Yes? AUDIENCE: It sounds like
there’s a perception that it’s difficult to get
robust amplification from FFPE-preserved tissue. But you found pretty concordant
results, it seems like, with FFPE. Did you have to do any kind
of special optimizations in primer design, or do
you think [INAUDIBLE]?? JOSHUA A. LIEBERMAN:
That’s a great question. So do we do anything
different molecularly for FFPE tissue
versus fresh tissue? The short answer is, no. I think that the longer answer
is, we intentionally target– try and minimize the
application length. Because if you think
about DNA shearing, that’s wrapped around
histone shearing, you sort of get a size limit. And so shorter
amplicons are better. But same primer set. We try really hard for us to
be the ones to cut the tissue, just because it gets handled
a lot in a non-sterile way. And so we try and get rid of
contaminating microbial DNA. But other than that,
no, there’s not a lot that’s special that we do. Good management, good staff,
and longitudinal investment in people– but maybe that’s self-evident. AUDIENCE: I have
one more question. So you had this 80%
mortality when the treatment was delayed till 12 days. So it seems like there’s a race
for both sensitivity and speed. You’re down to five days– are these people just
presenting so sick that most of the pathology
is a foregone conclusion and you will can’t get
below 5%, 10% mortality? Or if we had a one-day
result, can you estimate the magnitude
of the clinical impact? JOSHUA A. LIEBERMAN: I can try. So the question is,
basically, at what point is fast fast
enough, for particularly diagnosing mucormycosis? And are the patients so
sick that they’re already committed to a terminal fate? So I guess borrow from
cell-developmental biology. I think early on there probably
is real benefit to starting earlier. And it’s not just this
binary cut point at six days. This is only 70 patients. So it’d be interesting to see
this with a much larger sample set. I think one of the
problems is actually patients don’t necessarily
present that sick. And so if you have a little
bit of eye pain and kind of a stuffy nose,
most of us, I think, might ignore that
as seasonal allergy. But if you are three weeks
out from chemotherapy, that’s potentially a devastating
symptom, even without a fever. So I think speed
probably does matter. And getting down to
one day or two day and starting therapy
early matters. I think at this
point, usually, when people are highly suspicious
for an infection like this, you get upfront and
empiric antifungals– often combination therapy. And then the question becomes,
how quickly can we pull off the amphotericin to try and
protect someone’s kidneys, or how quickly can we
pull off of voriconazole because they’re having
some sort of CNS toxicity? So thank you, that’s
a great question. All right. Thank you very much. [APPLAUSE] [MUSIC PLAYING]

Leave a Reply

Your email address will not be published. Required fields are marked *