Ground Zero – Impact of Gut Microbiome on Host Epithelial Functions and Responses – Eugene Chang
Eugene Chang:
I’d like to start by thanking the organizers for allowing me to be part of this discussion. So when Lita asked me to give this talk, she
wanted me to talk about the impact of the gut microbiome on host epithelial functions and
responses. And I’m going to do that, but largely through
inspiration received from Owen White yesterday, I decided to go, rogue, last night, and I shortened
my talk to allow more time for discussion of the gaps, needs, and challenges. I’m still going to talk about what my charge
is, but towards the end, I think you’ll see that I’m going to go outside of that charge
to talk about things that I think broadly apply to research, and hopefully will advance
some of our thinking in approaches to the study of human microbiomes. So this slide just is a compilation of examples
of how critical gut epithelial functions are impacted by microbes, and they include a variety
of functions, for example, barrier function, development, wound healing, immune functions,
cytoprotection, transport, autophagy, and proliferation and apoptosis. And I think that you will appreciate that
many of these processes are quite essential for allowing the gut epithelium to be the
interface between us and the outside world, or in the case of our microbiota, our
inside world.
They provide very important functions in separating
us, but also, they’re a selective barrier so that we absorb critical nutrients, water,
and electrolytes that we need to sustain life. Now what I’m going to do in the next few slides
are show some specific examples, and these are just examples that come from a literature
that is replete with both mechanisms and processes that have been identified that mediate how
microbes affect epithelial function. Shown on the left is just a stain for Ki67,
which is a marker of cellular proliferation. And you can see that in the control, the colonic
crypt is shown here, and the lower half of the colonic crypt are most of the proliferating
cells.
But if you take this mouse, and you treat
this mouse with a cocktail of broad-spectrum antibiotics over a few weeks,
you see a dramatic change. There’s mucosal atrophy with a significant
decrease in cellular proliferation. Now, proliferation is only one side of that
coin, and if you look at apoptosis, which is also important in the maintenance of epithelial
homeostasis, I refer you to this study by Brent Polk and his group where they identified
two peptides that were created by Lactobacillus TG, which they call p40 and p75. And both of these panels show staining for
apoptosis, which is brown, and you can see that after treatment of the mucosa with
TNF, you get a significant increase of apoptosis. But if you pretreat that tissue with either
p75 or p40 you can significantly abrogate the induced apoptosis. The next slide is an example, one of many,
but this one comes from our laboratory, of conditioned media from bifido breves, which
protects both mouse jejunum as well as human caco2 epithelial cells, protects their barrier
function against reactive oxygen species.
And that’s illustrated here; that is, here
are the controls, and, in this case, we’re measuring paracellular flux using tritiated
[spelled phonetically] mannitol. When you add the ROS, you can see that the
barrier function breaks down both here and here. But if you pretreat these tissues with conditioned
media from bifido breves, you can mitigate that effect on barrier function. Now one of my favorite examples, and I think
is a very good example of host-microbe interaction, is the induction of heat shock proteins. So heat shock proteins are highly conserved
proteins, and in the gut, they’re physiologically expressed in two regions of the GI tract:
one is the stomach and the other is the colon.
And I don’t think that that’s by chance. These are the two most hostile areas of the
gastrointestinal tract, and, as I will show you, these heat shock proteins are
essential for the maintenance of intestinal homeostasis. But there is something very unusual that I
wanted to highlight by showing you this Swiss roll of immunostaining for one of the heat
shock proteins, Hsp25. So a Swiss roll is basically where we remove
the colon, and then we cut along the mesenteric border. We roll it up, and then we slice it like a
Swiss roll. Now what that allows us to do is look at protein
expression both along the local vertical axis as well as the longitudinal axis. And I think you can appreciate by the brown
staining that there is a — most of the heat shock protein is expressed by the surface
colonocytes; that is, exactly the cells that are in direct opposition to the luminal fluid
and gut microbes.
The other thing I think you can appreciate
is that there’s a gradient of expression which is largest in the proximal colon, but then
begins to dissipate and is almost undetectable by the time you get to the rectum. Now one of the things that we noticed immediately
was that — or we thought of immediately is that perhaps the reason that these proteins
are being physiologically maintained is that they’re getting cues from microbes. And in support of that, if you look in a germ-free
mouse, or if you antibiotic treat a mouse, what you find is that you get an abrogation
of the expression of these heat shock proteins.
Now implied from that is that if this gradient,
if it is due to cues from microbes, then there must be a gradient or there must be the heterogeneity
of microbes that are found in the proximal versus the distal colon. And I think that through studies that we’ve
done both of 16S as well as metagenomic profiling, this indeed is true. And that’s shown on this slide. So this is a study where we did a colonoscopy
on a healthy human subject, and this was done on a patient or a volunteer that underwent
this procedure without colonic lavage; that is, we didn’t — we tried our best not to perturb
the communities of microbes in the colon.
By doing so, I think that you can appreciate
that the metagenomic profiles of the mucosa-associated microbiota are quite different between the
the right colon and the left colon. And each slice of these pie charts represents
different subsystems, functional subsystems. The fact that these heat shock proteins are
important for the maintenance of intestinal homeostasis is illustrated here, so if you
take a wild-type mouse and treat them with dextran sodium sulfate, an agent that will
induce colitis, wild-type mice typically will develop inflammation, but it will spontaneously
resolve over about a two- to three-week period.
On the other hand, if you have a mouse where
the gene for Hsp70 is deleted, these mice do very poorly. They exhibit much more severe colitis,
and after several treatments of DSS, they go on to chronic-like colitis that is sustained. And the colitis is very typical of what we
see in human ulcerative colitis. There is a presence of crypt passes [spelled
phonetically], branching crypts, and also inflammatory infiltrates. These mice also developed a typical IBD metaplasia
to carcinoma sequence of cancers that we see in our patients. Now a lot is known, and the literature is
quite robust about how microbes affect the host site. What is less well-known are what are
the mediators that microbes use to impact epithelial function? A few have been identified, and I list them
here. They include things like quorum-sensing molecules,
innate ligands, short-chain fatty acids, lactic acid, hydrogen sulfide, chemotactic
peptides, and certain metabolites.
But this is certainly not an all-inclusive list,
and it’s only the tip of the iceberg, and discovery and identification of other agents
awaits better technology to be able to detect them. So this is a segue to my slide on what are
our gaps in our knowledge of gut microbial-epithelial interactions. Well, I think there are three major ones. One is that, as I mentioned, we have rudimentary
knowledge and inventories of bioactive microbe drive factors. But our understanding of the complexity and
heterogeneity of gut epithelial functions, particularly as they relate to microbial selection,
assemblage, and region-specific interactions I think is incomplete.
We also have an incomplete vetting and understanding
of the above in the context of human biology and pathobiology. So at this point, I am going to broaden my
comments to what I think are the challenges and needs, and this not only applies to studies
of the gut epithelium but studies in general in the area of human microbiome research. And I group them into three categories. And the first category is human studies. So HMP1 was largely devoted to the study of
humans. And I think that a lot of useful information
came out of that project. But the limitations are the following. Most of the data that we have to date is observational
or associative. And part of the reason for that is that people
are really difficult to study, and they’re different. You can’t do much with them or with them. The other problem is that we often look at
our microbiota data in the context of the sort of archaic disease classifiers that clinicians
often use to help them manage their patients.
But these disease classifiers are based on
symptoms, course, and disease, and are not necessarily reflective of the kind of underlying
pathophysiology. I think another limitation, and maybe this
was a necessity, is that the HMP1 was a technology-driven effort. And we — it was an important effort and
helped us vet and develop some consensus in the way that we study gut microbes. But I think in doing so, the importance of
biological drivers was minimized, and, in part, they were addressed by the human demonstration
projects.
But I need we need many more human demonstration
projects to move that forward. And associated with that is, you know, we’ve
taken a bottom-up approach. That is, we now have an immense amount of
data that we are sorting through, and from that data, we put together hypotheses that
we then look back on to the clinical situation and try to think if it’s relevant. My view is that maybe that might be backward
and that we ought to really look at the clinical situation, and then, from that, develop
various hypotheses that we will then develop or explore in clinical studies as well as
experimental counterparts. I think that there are major issues that I
brought up in the discussion yesterday which are technical. We still don’t have a consensus on how to sample
patients. And I’ll show you an example of that shortly.
I think quality control is an issue. We don’t have any well-accepted or standardized
operating procedures. And I think that this affects the interpretation
of data. It’s the old garbage in/garbage out story. And I think we have to pay much more attention
to getting together and developing optimal ways to achieve a better upfront result. The other part of the technology is, and I
don’t think that we’ve done a good job in vetting it, so I always tell my colleagues
that — who show me the metatranscriptome or metagenomic data, I ask them, well, what
does it mean, you know? And how do you account for it? I mean, do you know that this data
is true? And I think that we haven’t done a
good job in actually vetting whether our interpretations of these omics data are correct.
And in that regard, I think we need to do a
better job at building a toolbox that can be more direct measures of microbial function. The third category I lumped into was experimental
and data analysis. And here I think that I’m just going to echo
some of the things that Sarkis and others have said. I think we have to think about experimental
and animal models as part of studies of the human microbiome. Just studying the human is not going to get
you far enough, and I think the approach has to be where you have to do these two approaches
in tandem.
I think that the integration of large data
sets is a problem, but also the fact that location is very important. For example, if you do a stool sample, and
you’re looking at an inflammatory process that may be localized to a different part
of the GI tract, how do you truly know that the alterations that you see in the stool
are going to be reflective of what is happening at the local level. I think we need to think about or strive
to complete our inventories of microbial transcriptomes, proteomes, and metabolomes because, at the
current point, I think that this is a major bottleneck. Now let me give you some specific examples
of the points that I tried to make. So I mentioned yesterday that inflammatory
bowel diseases are described as two clinical phenotypes: one, is Crohn’s disease, and the other
called ulcerative colitis.
They don’t necessarily reflect the pathophysiology. And the pathophysiology can be quite varied. So when we’re trying to relate that to microbial
data or host-microbe interactions, we may, in some instances, be comparing apples and
oranges. But to add to that, you can look at what happens
in patients with Crohn’s disease. So this is data that came out of a study that
was published many years ago but reflects the fact that IBD, in this case, Crohn’s disease,
is progressive. And the natural histories among patients vary. So what you see in the early part of the disease,
which is mostly an inflammatory type, may not be what you see later in the disease. And I think that that affects the interpretation
of your data.
So if you did your sampling at T-1, it’s going
to be very different from what happens on T-2. And if you sample at T-0, that is, before
the onset of disease, this is an ideal time point to be able to capture events before
the disease happens so you know a little bit of what cause and effect are. But the problem is that we have no way of
predicting or identifying people who are at high risk for developing inflammatory bowel diseases,
so most studies — in fact, I would say all studies, pretty much, have looked at post-inflammatory
events. Now another challenge is this. So this is a lesion, or ulceration, in a patient
with Crohn’s disease, and so what typifies these diseases is their topography. Ulcerative colitis only involves the colon;
always starts in the rectum. It’s a mucosal disease. Crohn’s disease, on the other hand, is patchy, and it can involve any part of the GI tract. So if you look at this lesion, this is inflamed
and ulcerated, but over here, it’s normal. So when we do a stool sample, what does it
reflect? We need to biopsy or sample around
this lesion.
Now if we sample, do we sample here, here,
or in here? And how do we sample? Do we do a pinch biopsy, or do we do a brushing? So none of these things have been truly resolved,
and there’s no consensus. But they will affect the outcome of your data
interpretation. So I may seem, having presented this, as
a pessimist, and — but I can tell you those people in my lab know that I am an eternal
optimist, and I believe there’s a solution to every problem.
So one of them is that we were fortunately
funded by the Human Microbiome Project to do one of the demonstration projects, and
we selected pouchitis. This is a surgical procedure that’s performed
in patients with severe ulcerative colitis. They get their colon removed, which is curative. But for continence what happens is the surgeons
fashion a pouch that represents a pseudo-rectum, which is anastomosed to the anus so that they
can have voluntary bowel movements. Now half or more of these patients would develop
a condition that looks very much like the original ulcerative colitis. And the other interesting thing is that if
you do the same procedure in a patient that doesn’t have IBD, somebody who has a familial
form of colon cancer, very few of those patients develop pouchitis.
So this is something that we think is unique
to ulcerative colitis and may recapitulate some of the pathophysiological events related
to that disease. We know it’s microbe dependent because the
treatment of choice is antibiotics. This is — most of the patients will develop
this condition within a year, which makes an ideal project because you can do it prospectively,
you can sequentially sample your patients unprepped and over time, and be able to maybe
capture events before they happen. And then patients serve as their controls,
which I think is very important to design in most human trials of — that look at the
microbiome. Now a lot of useful information has been gained
from this study, and Vince Young is going to present that later in the morning so I’m
not going to get into it.
But even after looking at that data and trying
to project out what it means, I think that we’re encumbered by the fact that we still
are — they have correlations and associations. So my argument is that we have to develop
better animal models. And in this regard, I think that we propose
a model that’s been around a long time that fulfills a number of the criteria that recapitulates
some of the events in the human pouchitis condition.
So this is a blind loop that’s surgically
created; so that this is the mainstream of the ileum, and then off to the side is this
pouch. And if you orient the motility this way, it
empties; if you orient the motility this way, it fills. But over days to weeks, this self-filling
the loop gets colonized and pretty much begins to look like a colon. As you can see, this looks somewhat like this;
this looks like the ileum. And if you look at the microbiota, indeed
they develop a colonic-like microbiota, as we see in the human ileal pouches, that clusters
very closely to the sham-operated colon, whereas the self-emptying clusters very close to the
sham-operated ileum. And then if you look at how the microbiota
from these pouches compare to the human pouch you can see that up here in the green is the
self-emptying loop, which is away from the human samples here, whereas the self-filling
is somewhere in between but closer to the human sample, suggesting that we’re recapitulating
some of the changes in the human microbiota.
Now the other thing is, you have to ask why
pouchitis if it reflects ulcerative colitis, occurs in a small intestinal tissue
because that’s what the small — the pouch is made from. And we have to hypothesize that they’re transformations
of the small intestine into a large intestine. And this is seen in human pouches,
but we also show that that is true in the mouse pouch. So this is just to show you the normal colon
in the mouse, and then with the filling loop, you see that the histology with these elongated
test tube-like crypts resembles that we see in the normal colon, whereas the emptying
loop stays pretty much like the small intestine. And then the gene signatures are also very
similar; that is, the filling loop is very similar to the colon, whereas the emptying
is close to the ileum.
And — but we notice that even though we get
colonization, that is insufficient by itself to cause inflammation. And — so we added to this a genetic variable
with these IL-10 mice were subjected to the same procedure. You can see that, with the genetic susceptibility,
they develop information, but if — up here in the self-filling loop, but not the self-emptying
loop. So the working model that we have from this
mouse model is that this is a three-legged stool, representing the importance of colonic metaplasia,
the development of a colonic microbiota, maybe dysbiosis, and some form of genetic susceptibility. None of these are sufficient by themselves,
and you need to have the perfect storm for this event or pouchitis, and maybe
ulcerative colitis to develop.
So, in moving forward, we’re beginning to
go back to our patients and beginning to look at the host side, because looking at microbiota
alone out of context with the host, I think we lost a lot of opportunities. But to look at the host side, we have to
understand that the gut mucosa is a complex multicellular tissue, and that’s shown by
this picture here. This is the microbial world here, and then
there is a mucus layer in between, and then this is us. But our mucosa is made up of many different
types of cell types, and if you just take a biopsy and you do, for example, gene expression,
it’s going to represent an admixture of cells.
And I think through a variety of technologies,
for example, laser capture, we can actually look at individual compartments, and then
through a systems biology approach, determine how they fit together like a puzzle. And then there are other new technologies,
like the development of steroids, where we can develop fairly mature structures that
we can then look at host-microbe relationships. So with that, I’m going to end, and I’d be
happy to entertain any further discussion or questions.
Thank you. [applause] Male Speaker:
Gene, thank you for a great overview, and for, I think, clearly articulating some of
the gaps, needs, and challenges. I think we can have time for maybe one question. While Alfred [spelled phonetically] is coming
up to the mic, maybe I could just ask quickly, in these entered models, are indigenous microbiota
being retained as you develop these enteroids? Eugene Chang:
No.
So these steroids are derived from stem cells
that you can harvest. From the mouse, it’s fairly easy to do, and
the technology is now feasible even for humans. But they have to be established in sterile
conditions. Male Speaker:
[unintelligible] North Carolina. Gene, great discussion points. We need a full afternoon, but I support
your idea of host-microbial interactions. We need to genotype all of these patients
that we’re getting the microbiota on. And fully also support the mechanistic necessity
for mechanism exploration and appropriately-chosen mouse models, making sure that they are appropriate. And I love your self-filling [unintelligible]. Quick question and that is on the chronicity
of your Hsp70 knockout mouse with — I couldn’t tell whether it’s single are several doses
of DSS — did that extend into the small intestine, and could you comment quickly on the differences
in the protective mechanisms and resiliency of the small intestine recovery versus the
colon.
Eugene Chang:
Yeah, so — thanks for asking that. So actually what happens is that they — as
you know, the IL10 knockout mice — or the DSS is typical distal colitis. What we see is that the colitis will extend,
like ulcerative colitis, out to the proximal colon. But it doesn’t go into the ileum. Male Speaker:
Yeah, but it can. So in our RGM knockout mice, we get
ileitis in the susceptible host. And DSS certainly is pervasive throughout
the intestine because you give it orally. So it’s just that there is a particular propensity.
But I was just wondering if it extended in
that particular setting. It can be in other genetically-susceptible hosts. Eugene Chang:
Thanks. Female Speaker:
Thank you. [applause] Our next speaker is Dr. Susan Erdman from
Massachusetts Institute of Technology. Her talk will be Wound Healing Longevity:
Harnessing the Microbe-Induced Hormonal and Immune Proficiency for Human Health.
