Session 7: Autoimmunity and Allergy

Transcript of Part 2: Transcription Factor Aire Orchestrates T Cell Tolerance

00:00:08.15	So, I'm Diane Mathis.
00:00:10.07	I'm a professor of immunology
00:00:11.29	at Harvard Medical School,
00:00:13.17	and this talk, my second one,
00:00:17.23	is focused on one particular mechanism
00:00:21.29	of enforcing T cell tolerance,
00:00:24.07	which depends on the transcription factor Aire.
00:00:29.11	Now, in my first talk,
00:00:31.09	I explained to you that
00:00:34.13	tolerance mechanisms are needed
00:00:37.25	because of the random manner in which
00:00:40.25	the repertoires of antigen-specific receptors
00:00:44.15	displayed on B cells and T cells
00:00:48.09	are generated during their differentiation
00:00:51.24	in the thymus and bone marrow.
00:00:56.10	And then I went on to
00:00:59.21	define and describe several of the mechanisms,
00:01:03.00	the major mechanisms of imposing tolerance.
00:01:07.13	And then, finally,
00:01:10.10	I explained that these tolerance mechanisms
00:01:15.01	break down, actually, relatively frequently,
00:01:17.20	and the result is the development of autoimmune disease,
00:01:21.15	diseases like type-1 diabetes, or myasthenia gravis,
00:01:25.26	or multiple sclerosis.
00:01:30.11	Now, this talk is focused on
00:01:36.00	one of the more recently discovered
00:01:40.06	means of enforcing tolerance,
00:01:42.17	that is, it was discovered about 15 years ago,
00:01:45.25	and it depends on a transcription factor called Aire.
00:01:51.25	It's an important mechanism
00:01:54.03	and a fascinating mechanism,
00:01:57.08	at least I think so and I hope you'll agree with me
00:01:59.29	by the time we get to the end of the talk.
00:02:03.25	this mechanism actually makes...
00:02:07.26	links together two of the important mechanisms,
00:02:12.14	one in central tolerance
00:02:14.11	and one in peripheral tolerance.
00:02:16.28	So, the Aire story starts with
00:02:20.17	a human disease called APS-1,
00:02:24.17	for autoimmune polyglandular syndrome type-1.
00:02:29.08	And these individuals have
00:02:31.15	severe Candidiasis infections
00:02:34.13	of mucosal surfaces.
00:02:36.02	They also have autoimmune attack
00:02:38.27	on the parathyroid glands
00:02:40.19	and autoimmune attack on the adrenal glands.
00:02:43.25	Now, these are the three most frequent symptoms
00:02:46.01	and they're the symptoms which are used
00:02:49.22	to diagnose the disease,
00:02:51.19	but all of the individuals with APS-1
00:02:55.10	have multiple other autoimmune manifestations.
00:02:58.17	They might have type-1 diabetes
00:03:01.00	or they have autoimmune attack on the ovaries or the liver,
00:03:03.18	and what these organ targets are
00:03:07.08	varies from individual to individual,
00:03:09.02	even for people that are in the same family
00:03:12.02	and have the same mutation.
00:03:16.10	So, in 1997,
00:03:19.21	two groups independently cloned
00:03:23.09	the gene underlying APS-1
00:03:25.08	and they called the protein that it encodes
00:03:29.01	Aire, for autoimmune regulator.
00:03:30.29	Aire was a quite large protein
00:03:33.21	of more than 500 amino acids,
00:03:35.27	and by now more than 50 mutations,
00:03:38.15	scattered through the Aire gene,
00:03:42.29	have been identified in different individuals with APS-1.
00:03:50.11	So, from the beginning,
00:03:52.07	Aire was thought to be some kind of transcriptional regulator,
00:03:57.00	and I'll show you later on that this, indeed,
00:03:59.16	turns out to be the case.
00:04:02.24	So, one important clue
00:04:05.04	to how Aire is working came
00:04:09.05	just from knowing where it's expressed.
00:04:10.28	So, it's expressed primarily in the thymus,
00:04:14.01	not by the differentiating T cells themselves,
00:04:17.11	but rather by the stromal cells,
00:04:19.16	the epithelial cells,
00:04:23.07	which nurture their differentiation and allows the different processes
00:04:27.24	that need to take place for T cell maturation.
00:04:31.16	They're localized...
00:04:34.13	Aire is localized in the medulla
00:04:36.10	and, within the medulla,
00:04:38.21	specifically in a very small subset of epithelial cells
00:04:42.27	which we call medullary epithelial cells,
00:04:45.14	and they make up only 0.5%
00:04:48.27	of the stromal cells in the thymus.
00:04:52.27	And the reason that this finding elicited interest
00:04:57.29	was that, at that time,
00:05:00.06	there was a body of data that was growing larger and larger
00:05:02.22	that these cells
00:05:06.08	actually express a large repertoire of RNA transcripts
00:05:10.04	encoding what we normally think of as
00:05:14.01	proteins particular for fully differentiated cells.
00:05:18.19	So, for example, one is insulin
00:05:21.09	or there might be myelin basic protein,
00:05:24.10	or a heart protein, or a liver protein.
00:05:28.04	In fact, when it was looked at very carefully,
00:05:32.13	it was found that many tissues in the body,
00:05:36.04	more than 30,
00:05:37.25	are represented by transcripts
00:05:41.13	in this very small population of cells.
00:05:46.17	And so the notion developed that these transcripts
00:05:50.11	would be translated into proteins
00:05:53.02	and, actually, if you have good antibodies
00:05:55.19	you can find these proteins
00:05:57.09	-- you can find insulin in these cells, for example --
00:06:00.29	and these proteins would be degraded
00:06:04.03	by normal mechanisms of antigen processing
00:06:07.08	and loaded onto MHC molecules
00:06:09.17	and be shuttled to the surface of the cell.
00:06:13.07	And then, as the self-reactive, differentiating T cell
00:06:17.23	is percolating through the thymus,
00:06:20.06	if its T cell receptor recognizes
00:06:22.23	this peptide-MHC complex
00:06:25.28	in a particular window of affinity or avidity,
00:06:30.04	T cell tolerance will take place.
00:06:34.03	And it was our hypothesis that,
00:06:36.12	actually, Aire is controlling the transcription
00:06:39.20	of this repertoire of transcripts
00:06:44.11	encoding peripheral tissue antigen...
00:06:48.20	proteins, or PTAs is what we call them.
00:06:51.23	And that was because of the overlap
00:06:55.25	in where Aire is expressed,
00:06:57.17	as well as the fact that individuals that have an Aire mutation
00:07:01.04	have a multi-organ autoimmune disease.
00:07:04.06	So, to evaluate our hypothesis, we
00:07:09.19	-- and, actually, other investigators, independently --
00:07:12.29	made mice which were lacking Aire
00:07:16.02	-- Aire knockout mice.
00:07:18.24	And when we looked at these mice, indeed,
00:07:21.00	they had inflammatory infiltrates in several organs.
00:07:27.03	I'm showing the salivary gland, here, and the thyroid gland,
00:07:29.20	where you can see these infiltrating leukocytes,
00:07:33.18	which you don't see in the wild type...
00:07:36.18	corresponding wild type tissues.
00:07:39.24	These mice also have auto-antibodies,
00:07:42.16	circulating auto-antibodies,
00:07:44.13	against many different organs.
00:07:46.18	So, basically, what you do here is
00:07:49.28	you just take serum from a wild type mouse
00:07:52.22	or a knockout mouse and use it...
00:07:56.04	incubate it with normal tissue
00:07:59.03	and then come in with a secondary step,
00:08:01.24	which allows you to light up
00:08:05.23	those regions where the antibody bound.
00:08:09.06	And so, the serum from the Aire knockout mice,
00:08:11.10	but not from the Aire wild type,
00:08:12.29	lit up a lot of different organs,
00:08:15.01	and actually some specific structures in different organs.
00:08:20.15	So, it was the parietal cells of the stomach
00:08:22.07	and the rods and cones region of the retina, for example.
00:08:26.01	Then a very critical experiment
00:08:30.12	was the one that's depicted here,
00:08:32.10	where we isolated RNA
00:08:35.22	from medullary epithelial cells
00:08:38.14	coming from an Aire wild type mouse,
00:08:40.18	on the x axis,
00:08:42.19	or an Aire knockout mouse, on the y axis,
00:08:44.27	and what I'm showing you is
00:08:47.16	whole-genome expression profiling,
00:08:52.13	where each dots represents expression of a particular gene
00:08:57.10	and how it's expressed in the wild type
00:08:59.24	is its point on the y axis... on the x axis,
00:09:03.09	and how it's expressed in the knockout is its point on the [y axis].
00:09:06.06	And where its induced by Aire
00:09:10.15	falls below the diagonal.
00:09:12.14	And when we looked at this,
00:09:14.28	what we found was that
00:09:17.11	Aire actually induced hundreds of transcripts
00:09:19.25	in medullary epithelial cells,
00:09:21.24	in particular, transcripts that were encoding these PTA proteins,
00:09:27.11	or peripheral tissue antigens.
00:09:31.05	Now, these data which I've shown you
00:09:33.13	were the original data that we got back in 2002,
00:09:37.19	and we could see that Aire was controlling hundreds of transcripts.
00:09:40.12	And more recently, there are more performant methods
00:09:43.24	to look at gene expression,
00:09:47.24	and, for example, what we call RNAseq,
00:09:51.08	which allows us to look at a lot more transcripts
00:09:53.12	and in a more quantitative way,
00:09:55.10	and what we found there surprised us in the fact that
00:09:59.10	Aire actually induces thousands of transcripts,
00:10:02.28	more than a quarter of the genome.
00:10:08.18	And then, the last experiment is
00:10:12.03	what I call closing the circle,
00:10:15.01	and what we did there was to take Aire knockout mice,
00:10:18.02	and in those mice there are circulating auto-antibodies,
00:10:22.15	and some of these are against stomach proteins,
00:10:26.02	which I showed you on the previous slide.
00:10:29.18	We then isolated...
00:10:31.04	we then identified the antigen
00:10:33.09	that these stomach antibodies saw
00:10:37.00	and found out that it's the mucin protein.
00:10:41.02	And then we went back
00:10:43.02	and looked into the medullary epithelial
00:10:47.26	gene expression profile and found that, indeed,
00:10:50.25	Aire was regulating mucin transcripts,
00:10:53.16	because when Aire wasn't there
00:10:56.13	mucin transcript levels in the thymus were reduced.
00:10:59.04	Other investigators did similar experiments,
00:11:02.24	one with an eye antigen
00:11:05.26	and another with salivary gland antigen,
00:11:08.29	and came out with the same conclusion.
00:11:11.21	So, I think that the model which we proposed
00:11:16.17	actually turned out to be correct,
00:11:18.23	according to a number of criteria,
00:11:20.19	and the model was close enough to the human disease
00:11:24.23	that it could be used to dissect specific mechanisms.
00:11:31.15	So, let's look a little bit at the cellular mechanisms involved.
00:11:36.07	First of all, you might have noticed that
00:11:40.04	I was very vague about what happened
00:11:42.18	once the differentiating thymocytes
00:11:44.23	saw the MHC-self-peptide complex.
00:11:46.19	I just said T cell tolerance takes place.
00:11:49.23	Now, there are several mechanisms
00:11:52.28	by which this might happen.
00:11:55.26	It could be that Aire promotes negative selection,
00:11:59.11	or clonal deletion,
00:12:01.14	of the self-reactive effector cells
00:12:03.25	that are going to get out into the periphery
00:12:07.08	and wreak havoc.
00:12:09.05	It could instead, or in addition,
00:12:11.15	be that Aire promotes positive selection
00:12:14.09	of regulatory T cells
00:12:16.25	that control the activities of these effector T cells.
00:12:21.13	Or Aire could do something completely different,
00:12:23.12	maybe something totally unexpected,
00:12:25.16	for example, it could control
00:12:28.15	antigen-presenting cells
00:12:30.20	or some other kind of innate cell
00:12:34.14	that starts off an autoimmune response.
00:12:40.00	Indeed, we found that Aire does promote negative selection,
00:12:43.23	or clonal deletion of self-reactive cells,
00:12:45.22	and I'll show you the original experiment
00:12:51.03	which demonstrated that.
00:12:53.11	So, this takes advantage of a trick
00:12:58.14	that immunologists use,
00:13:00.10	the creation of T cell receptor transgenic mice,
00:13:02.22	and this just tries to get around the fact that
00:13:08.06	any particular antigen specificity
00:13:10.09	is usually at a very low frequency,
00:13:12.17	so only 1 in 10^4 to 1 in 10^6 T cells
00:13:17.09	will be specific for a particular antigen.
00:13:19.19	And so, what one can do is
00:13:22.16	isolate T cell receptor transgenes
00:13:25.08	from a particular T cell clone
00:13:27.07	that are already rearranged,
00:13:29.10	introduce those into mice,
00:13:31.13	and that will shut off endogenous T cell receptor gene rearrangements,
00:13:38.16	and so the mouse will have a repertoire
00:13:40.14	that's highly skewed for that antigen specificity.
00:13:43.24	And in this case we add an additional twist onto that,
00:13:47.04	in that we create one mouse
00:13:50.18	that has a neo-self-antigen
00:13:53.19	expressed somewhere,
00:13:55.06	and then we create a T cell receptor transgenic mouse
00:13:57.20	that's capable of seeing this neo-self antigen.
00:14:02.08	So, an example is, for the neo-self antigen,
00:14:06.07	we make a mouse which is expressed ovalbumin,
00:14:10.12	membrane-bound ovalbumin,
00:14:13.04	under the dictates of the rat insulin promoter,
00:14:16.05	so it should be expressed in the pancreas
00:14:19.15	and, as I now should have convinced you,
00:14:21.24	is also expressed in the thymus.
00:14:25.26	And the second mouse, the reporter mouse,
00:14:28.05	is a T cell receptor transgenic mouse
00:14:33.17	which is capable of seeing a peptide
00:14:37.11	for membrane-ovalbumin.
00:14:39.25	So, you can see that when we cross these two mice together,
00:14:42.14	we create a potentially explosive situation
00:14:47.12	where there are many, many T cells
00:14:50.05	expressing this self-reactive specificity.
00:14:53.11	So, this slide shows you what happens
00:14:56.29	in the thymus of these mice.
00:14:59.17	We're looking at thymocytes
00:15:03.09	by staining for the CD4 and CD8 co-receptors,
00:15:07.21	and what you see is that,
00:15:09.15	in the absence of the neo-self antigen,
00:15:13.23	just the plain T cell receptor transgenic,
00:15:15.26	we find a lot of CD4...
00:15:18.15	fully mature CD4 single-positive T cells maturing.
00:15:24.15	And that's because the original clone
00:15:27.04	that we started with was a CD4+ T cell clone,
00:15:30.04	so that's expected.
00:15:31.28	And this is what happens
00:15:34.20	when you now cross the mice to the transgenic line
00:15:38.01	which is expressing the neo-self antigen.
00:15:41.03	In the presence of Aire, these CD45...
00:15:44.08	these CD4+ T cells are absent
00:15:48.24	-- they disappear, they're clonally deleted.
00:15:50.24	However, in the absence of Aire,
00:15:54.14	those T cells come back.
00:15:58.20	So, that clonal deletion does not take place
00:16:01.22	when Aire's not there.
00:16:03.16	And, actually, the animals develop
00:16:06.28	autoreactivity of the pancreas.
00:16:13.12	So, other labs did similar experiments
00:16:17.01	using other TCR transgenic
00:16:20.27	neo-self antigen models
00:16:23.01	and, in addition,
00:16:25.01	non-transgenic systems
00:16:28.04	could demonstrate that Aire was required
00:16:31.12	for clonal deletion.
00:16:33.04	So, I think we can be more specific here
00:16:35.06	and say that Aire is
00:16:39.05	controlling negative selection of self-reactive effector T cells.
00:16:43.16	Now, it turns out that Aire also controls
00:16:46.09	positive selection of regulatory T cells.
00:16:49.29	When we first started studying this,
00:16:51.29	we didn't think that that was the case
00:16:54.06	because we looked at a lot of mice
00:16:56.01	and we didn't see much difference
00:16:58.18	in the frequency or the type of Treg cell
00:17:03.17	that these mice had
00:17:06.13	-- so, CD4+/Foxp3+ regulatory T cells --
00:17:10.28	and that's because we were always looking at adult mice.
00:17:14.05	Later, we looked earlier
00:17:17.14	and did find that there was a deficit
00:17:19.26	in this regulatory T cell population
00:17:22.29	before 10 days of age.
00:17:25.14	Now, we wondered if this was very meaningful,
00:17:30.01	because it's not a huge difference
00:17:32.06	-- the cells aren't totally gone --
00:17:34.09	and it's only during this quite narrow time window
00:17:38.01	where one sees the difference.
00:17:40.00	So, we designed an experiment
00:17:42.05	which allowed us to show
00:17:45.03	whether regulatory T cells, during that time period,
00:17:47.20	are important.
00:17:49.07	So, we could do both a loss-of-function experiment
00:17:52.01	and a gain-of-function experiment.
00:17:54.22	So, in the loss-of-function experiment,
00:17:56.24	we took a mouse
00:18:00.19	where we could turn Foxp3 off
00:18:03.20	whenever we wanted to.
00:18:05.28	And so if we turn off Foxp3
00:18:08.14	then we won't get Tregs made during that time window.
00:18:12.14	So, we used that mouse to
00:18:15.03	deplete regulatory T cells
00:18:20.14	that are being generated during the first ten days of life,
00:18:23.16	and when we did that the mice developed multi-organ autoimmunity,
00:18:25.18	as indicated by the shaded blocks
00:18:29.28	that I show you, here.
00:18:32.03	If we did the same experiment,
00:18:34.10	however we took an adult mouse
00:18:37.16	where we depleted Tregs for 10 days,
00:18:39.06	we saw only rare and quite sporadic autoimmunity.
00:18:45.09	Now, for the gain-of-function experiment,
00:18:48.01	what we did was take Aire knockout mice,
00:18:50.15	which, as you saw before,
00:18:52.11	will develop multi-organ autoimmunity,
00:18:55.09	and then we took Tregs that were generated specifically
00:18:58.27	during that 10-day age window just after birth,
00:19:02.14	and we added those in,
00:19:05.20	and those mice were highly protected from autoimmunity.
00:19:08.27	However, if we took regulatory T cells
00:19:12.12	that were made during a 10-day window in an adult,
00:19:16.01	they were not protected.
00:19:17.18	So, by both the loss-of-function and gain-of-function experiments,
00:19:22.14	it was clear that Aire
00:19:27.16	was controlling an important population of regulatory T cells,
00:19:31.16	and this was specifically
00:19:34.29	during a very early time window.
00:19:38.05	Now, that made sense to us,
00:19:40.03	because we had an earlier finding
00:19:42.20	which we were not able to understand
00:19:45.11	until we got these more recent results,
00:19:47.21	and that is that we made a mouse
00:19:49.27	where we could turn Aire on and off at will.
00:19:53.28	So, if we had Aire on during the whole life of the mouse,
00:19:59.11	there was tolerance and the mouse didn't develop any autoimmunity.
00:20:03.10	And if we had Aire off during the whole life of the mouse,
00:20:06.24	it did develop autoimmunity.
00:20:09.23	That wasn't very surprising;
00:20:11.10	that's what we expected at that point.
00:20:13.14	However, what was surprising was that
00:20:17.00	if we had Aire on only for the first 7 or 10 days of life,
00:20:22.15	and then turned it off,
00:20:25.21	the mice were tolerant -- they didn't develop any signs of autoimmunity.
00:20:28.10	And then, conversely,
00:20:31.18	if we had Aire off during the first 7-10 days of life
00:20:35.10	and then turned it on for the rest of the life of the animal,
00:20:38.15	they did develop severe autoimmunity.
00:20:41.15	So, the expression of Aire during this very early,
00:20:47.06	this perinatal time window,
00:20:49.18	seemed to be necessary and sufficient to predict...
00:20:53.13	to protect from the autoimmune disease
00:20:56.02	that develops in the absence of Aire.
00:20:59.23	And so, we can also add that
00:21:03.07	Aire is also functioning
00:21:07.02	to promote positive selection of regulatory T cells.
00:21:10.05	I would be remiss if I didn't add that
00:21:13.13	other experiments in other laboratories
00:21:15.21	have suggested that Aire might have additional functions.
00:21:19.12	It might be involved in
00:21:22.01	differentiation of medullary epithelial cells,
00:21:24.11	differentiation of 𝛾:δ cells,
00:21:27.00	the turnover of medullary epithelial cells,
00:21:31.00	or peripheral tolerance.
00:21:35.15	Okay, ummm...
00:21:37.23	so, I think those are the major points
00:21:40.07	I'd like to make about the cellular mechanisms of Aire.
00:21:42.17	Now, I'd like to turn to the molecular mechanisms.
00:21:45.20	And I'll start out by saying this is an area
00:21:47.28	where it's been very dark for some time,
00:21:52.24	and light is just beginning to be shed.
00:21:57.29	People have been fascinated
00:22:00.04	by the molecular mechanism of Aire
00:22:02.02	since the beginning,
00:22:03.14	and the reason for that is that
00:22:05.22	here is a transcription factor
00:22:08.06	that's controlling thousands of genes
00:22:10.26	in a very small population of cells in the thymus.
00:22:16.11	And these genes, in the periphery,
00:22:20.00	are expressed very differently.
00:22:23.08	They're expressed in different cells,
00:22:25.11	they're expressed at different levels,
00:22:27.07	and they're expressed at different times
00:22:30.10	during the ontogeny of the individual,
00:22:33.22	so how can this one transcription factor do that?
00:22:38.22	So, from the beginning,
00:22:40.08	it was thought that Aire
00:22:43.19	was a transcriptional regulator,
00:22:45.09	and that's because it has
00:22:49.08	structural and functional features
00:22:52.05	of a transcriptional regulator.
00:22:53.27	So, structurally, it has a SAND domain
00:22:58.09	and, in other transcription factors,
00:23:00.28	the SAND domain is a DNA-binding domain.
00:23:03.12	However, I should mention that the important
00:23:09.24	amino acids in Aire...
00:23:11.26	the important amino acids for DNA binding in other proteins
00:23:15.25	are mutated in Aire
00:23:18.03	-- they're not the same as they are in these other proteins.
00:23:20.19	Aire has a nuclear localization signal.
00:23:23.10	It has a CARD domain,
00:23:25.11	which is important for homo-oligomerization.
00:23:29.10	It also has two PHD domains,
00:23:31.23	which are used in various types of protein-protein interactions.
00:23:41.02	As far as functional features,
00:23:42.27	Aire is localized in the nucleus,
00:23:45.01	as one would expect for a transcription factor.
00:23:47.06	It can induce transcription
00:23:50.02	-- so, if you make an expression vector where...
00:23:56.25	which will allow Aire expression if you transduce or transfect
00:24:01.17	the plasmid into a cultured cell,
00:24:08.19	and you also transfect a reporter gene
00:24:12.27	driven by the interferon promoter,
00:24:16.20	Aire will induce expression of that reporter.
00:24:20.06	So, it also binds to known transcription factors.
00:24:24.00	The first one identified was CBP,
00:24:27.10	or CREB-binding protein.
00:24:29.09	Now, Aire only binds very weakly
00:24:32.07	and non-specifically to DNA,
00:24:34.22	even though it has the SAND domain,
00:24:38.18	which I mentioned.
00:24:40.04	It seems to be more involved in binding
00:24:43.07	to different chromatin proteins
00:24:45.05	than directly to DNA itself.
00:24:50.17	So, even though it looked and smelled like a transcription factor,
00:24:52.29	there were always some odd things about Aire
00:24:55.18	which made one question
00:24:58.26	whether it was a classical transcription factor
00:25:01.29	that would bind to a promoter and induce or repress transcription
00:25:06.14	of a particular locus.
00:25:09.17	So, first of all, it's regulating thousands of genes,
00:25:14.01	so it seemed unlikely that they would
00:25:16.26	all have a specific binding site for Aire.
00:25:20.06	Secondly, it induces genes
00:25:25.26	which encode proteins that are found in many different cell types.
00:25:29.23	And, thirdly, it's possible to
00:25:34.04	introduce Aire artificially into different cell types,
00:25:38.04	either in culture or making transgenic mice,
00:25:41.21	for example, putting Aire behind the rat insulin promoter,
00:25:44.06	and in all these different cell types
00:25:46.27	Aire will induce batteries of transcripts.
00:25:49.14	However, the particular transcripts
00:25:52.17	that it does induce
00:25:54.27	differ from cell type to cell type,
00:25:56.23	and usually there's only about a 10-20% overlap
00:26:00.01	when you're comparing different cell types.
00:26:04.14	So, people have been working quite hard on this puzzle,
00:26:09.13	and I have to say that
00:26:13.16	we don't have a complete answer yet,
00:26:15.19	but we have learned some very important things
00:26:18.25	over the past couple of years.
00:26:20.21	So, one of them is that Aire
00:26:24.08	partners with many different proteins.
00:26:29.20	Now, the way that this experiment was done
00:26:33.23	was to take chromatin
00:26:37.09	from an Aire-expressing cell,
00:26:39.23	immunoprecipitate Aire
00:26:42.22	and then use mass spectromety...
00:26:47.05	spectrometry...
00:26:49.17	to identify the proteins that are binding to Aire.
00:26:52.05	And of course you have to do
00:26:54.23	many different types of controls
00:26:57.09	to prove the validity of this assay,
00:27:01.06	which were done.
00:27:03.00	And, even after these controls,
00:27:05.19	it became clear that Aire binds to...
00:27:08.20	interacts with scores of proteins,
00:27:10.22	either directly or within the same complex.
00:27:15.10	Up here, I'm showing about 40,
00:27:17.13	but even since this figure was made
00:27:20.07	there are another 5-10 which have been identified.
00:27:23.21	Now, some of these... and these proteins fall into four different classes:
00:27:27.24	nuclear transport,
00:27:30.12	pre-messenger RNA processing,
00:27:32.11	chromatin function,
00:27:34.10	and the DMA-damage response
00:27:38.24	and some transcriptional activation function.
00:27:41.19	So, it's not surprising that Aire would bind to proteins
00:27:47.08	involved in nuclear transport,
00:27:48.23	because it is in the nucleus,
00:27:50.27	or chromatin function,
00:27:53.01	because we know it's a transcriptional regulator somehow,
00:27:55.12	but some of the other classes of proteins
00:27:57.26	were somewhat surprising,
00:27:59.21	like pre-mRNA processing.
00:28:01.16	And, after this was found, it...
00:28:05.10	experiments were done which showed that
00:28:08.12	Aire does actually control pre-mRNA processing.
00:28:10.26	And then, lastly,
00:28:14.17	these proteins that are involved in transcription
00:28:19.00	but are also part of the DNA-damage response.
00:28:24.01	So, that's one important thing.
00:28:25.29	The second important thing is that
00:28:28.08	Aire seems to be operating very early
00:28:30.28	in the process of gene transcription
00:28:33.25	by RNA polymerase-II.
00:28:37.05	So, when I was learning about transcription,
00:28:41.25	some years ago,
00:28:44.14	we used to think that there were two types of genes:
00:28:47.24	genes that had RNA polymerase on them,
00:28:50.02	and they were transcribed;
00:28:52.07	and genes that didn't have RNA polymerase on them,
00:28:55.15	and they weren't transcribed.
00:28:56.20	But, over the last, say, 5-8 years,
00:28:59.14	it's become clear that there's a third class of genes
00:29:03.13	which is very important,
00:29:06.09	and these genes are ones which RNA polymerase binds,
00:29:11.01	it moves a little bit into the gene,
00:29:13.09	and then just stops.
00:29:14.29	And these are called paused or stalled polymerase genes,
00:29:20.13	and it's clear that every cell type
00:29:23.11	that's been looked at so far
00:29:25.17	has a group of genes
00:29:27.27	which have paused polymerase on them.
00:29:29.29	So, to explain this a little better,
00:29:34.06	let me give you some detail.
00:29:36.27	So, what happens is that
00:29:40.09	RNA polymerase binds to the transcriptional start site
00:29:43.12	and it has at its C-terminal end
00:29:47.09	a set of repeats of a particular sequence
00:29:50.16	that has serines in it that become phosphorylated.
00:29:54.23	And so it binds at the transcriptional start site
00:29:57.28	and lets its C-terminal domain hang out,
00:30:02.18	and then it clears the promoter,
00:30:06.03	once there has been phosphorylation
00:30:11.06	of the serine residues at position 5 in this repeat
00:30:14.21	by the transcriptional factor TFIIH.
00:30:20.09	It proceeds a little further
00:30:23.22	and then just stops.
00:30:26.07	This is a paused polymerase
00:30:28.21	and it has been thought that one of the reasons that it pauses
00:30:32.11	is to allow time to allow capping of the RNA to take place.
00:30:36.22	Now, pausing is enforced by these two transcription factors,
00:30:40.28	NELF and DSIF,
00:30:44.04	and it's lifted, or released,
00:30:47.16	by the combined action of Brd4 and the heterodimer P-TEFb.
00:30:55.07	So, this set of proteins comes in,
00:30:57.28	causes additional phosphorylation of the C-terminal domain,
00:31:02.09	causes phosphorylation of NELF,
00:31:05.04	which releases it,
00:31:07.13	and turns DSIF into a positive transcriptional regulator.
00:31:13.13	And then elongation can proceed.
00:31:15.29	So, what Aire does is that
00:31:21.26	it releases RNA polymerase pausing
00:31:26.03	on paused genes.
00:31:27.20	And it's now become clear that
00:31:30.24	this site of RNA polymerase pausing
00:31:33.24	is a site of impact for many important types of
00:31:38.15	transcriptional regulation.
00:31:40.02	It's where c-Myc works,
00:31:41.27	it's where much of the regulation of LPS
00:31:44.28	-- lipopolysaccharide --
00:31:48.22	induced genes in macrophages is controlled,
00:31:53.07	and this is where Aire is operating.
00:31:56.15	So, we have a number of piece of evidence that point to that.
00:32:01.12	First of all, if we go back to
00:32:05.05	our genome-wide gene expression profiles
00:32:07.08	and, instead of looking at the whole gene...
00:32:10.14	the whole...
00:32:13.02	a group of Aire-induced genes and looking at the whole gene,
00:32:15.29	we look at different regions along the gene,
00:32:19.10	what we find is that Aire's impact
00:32:22.11	at the beginning of the gene is relatively low,
00:32:24.29	but it's impact later on is much higher,
00:32:27.26	which is indicative of an effect on elongation.
00:32:33.11	Secondly, Aire interacts directly with
00:32:38.29	Brd4 protein and also P-TEBb proteins.
00:32:43.03	And, thirdly, if we take inhibitors of
00:32:47.01	Brd4 or P-TEFb
00:32:49.28	or things which lift polymerase pausing,
00:32:54.05	those inhibitors also inhibit
00:32:57.15	Aire-induced gene transcription.
00:33:03.10	So, we're quite convinced that this is
00:33:06.27	a key element of Aire's molecular mechanism of action.
00:33:11.12	So, in fact, this mechanism goes quite far
00:33:14.15	in explaining many of these Aire oddities,
00:33:18.26	which I mentioned before.
00:33:21.08	It's clear, then, that it can
00:33:24.01	simultaneously control thousands of genes,
00:33:25.28	because thousands of genes
00:33:27.27	have polymerase paused on them.
00:33:29.15	It can induce
00:33:32.04	peripheral tissue antigen genes
00:33:34.23	associated with many cell types,
00:33:36.15	because, in fact, it's been found that,
00:33:38.22	preferentially, genes that are paused
00:33:42.01	are genes which are going to later be induced
00:33:44.20	or are later important for cell type differentiation.
00:33:49.26	And then, finally, it explains why,
00:33:52.19	if you put Aire into different cell types,
00:33:55.17	different sets of transcripts are induced,
00:33:57.16	and that's because different cell types
00:33:59.28	have different repertoires of paused genes.
00:34:04.18	And it also explains the large number of protein partners
00:34:09.07	that Aire has,
00:34:11.08	and that's because this scaffold,
00:34:14.21	this tail, this C-terminal domain tail,
00:34:18.10	which becomes phosphorylated
00:34:21.04	when polymerase pausing is lifted,
00:34:24.23	is an important scaffold for a number of cellular processes.
00:34:29.06	So, this... these phosphorylation events control RNA capping.
00:34:36.06	They'll also be very important in regulating splicing,
00:34:39.21	elongation,
00:34:42.10	and even nuclear export.
00:34:45.15	And then, lastly, both DSIF
00:34:48.19	and the C-terminal domain
00:34:50.16	have been implicated in controlling different histone modifications.
00:34:56.26	So, that's where we stand today
00:35:01.24	with both the cellular and molecular mechanism of Aire,
00:35:05.12	and of course there are some very interesting questions
00:35:08.00	which remain to be answered.
00:35:10.15	If we look back at the cellular mechanisms,
00:35:13.00	I think the most pressing issue is to determine...
00:35:18.15	is to determine,
00:35:22.16	what's so special about this repertoire of regulatory T cells
00:35:25.04	that develops in the first 10 days of a mouse?
00:35:27.22	Do they have a particular repertoire?
00:35:29.29	What are the antigens that they are seeing?
00:35:32.25	And why are they so important
00:35:35.13	for protecting against autoimmunity,
00:35:37.28	when regulatory T cells are of course generated
00:35:40.17	throughout the life of the animal?
00:35:42.17	If we look at the molecular mechanism,
00:35:45.01	it's clear that we have some important bits and pieces
00:35:51.03	about how Aire operates at the molecular level,
00:35:55.19	but we need to learn, still, some important things,
00:35:58.29	like, how does Aire actually target the genes
00:36:01.22	that it is going to induce?
00:36:04.05	Is it sufficient that the gene just be paused
00:36:06.08	and there's something about that configuration
00:36:08.21	which Aire can recognize,
00:36:10.16	or is there some other element
00:36:14.03	which is used for recognition?
00:36:15.24	And then, finally, what I haven't told you yet
00:36:19.10	is that Aire does control
00:36:22.08	a lot of these peripheral tissue antigen transcripts
00:36:25.11	in medullary epithelial cells,
00:36:27.15	but there are also some that Aire doesn't control.
00:36:30.19	So, is there an Aire-2?
00:36:33.24	And, if that's the cause, what's its identity?
00:36:38.10	there's no homologous protein,
00:36:40.29	or no highly homologous protein, I should say.
00:36:43.16	So, what's its identity,
00:36:46.05	and does it operate in the same way that Aire does.
00:36:49.22	And, lastly, I'd like to thank or acknowledge
00:36:53.26	the people in the lab who have worked on Aire
00:36:56.26	over the past 15 years,
00:37:01.08	and I think found some very interesting things.