Cell fate decisions in hematopoiesis

Transcript

• 00:00Why don't we go ahead and get started?
• 00:03So I'm Jeannie Hendrickson.
• 00:04I'm representing the enrichment
• 00:06program of the Yale Cooperative
• 00:08Center of Excellence in Hematology,
• 00:10or super excited on this Snowy Friday
• 00:12afternoon to have Marjorie Brands
• 00:14visiting us for a talk that's called
• 00:16cell fate decisions in hematopoiesis.
• 00:18She's visiting us,
• 00:19and virtually she's a professor in the
• 00:22Department of Cellular and Molecular
• 00:24Medicine and a senior scientist at the
• 00:27Ottawa Hospital Research Institute,
• 00:28and really enjoy the picture of her lab here.
• 00:31So we're thrilled.
• 00:33She's with us this afternoon.
• 00:35And at the end of her talk we
• 00:37should be able to have some
• 00:40questions and answers as well.
• 00:42So thanks again Marjorie for visiting.
• 00:44We really appreciate it.
• 00:46Thank you very much.
• 00:48It's a great pleasure to be here.
• 00:50Thanks Pat for the invitation and I really
• 00:53enjoyed my discussions that I had today,
• 00:56so I'm going to start by 10.
• 00:59So the as we all know,
• 01:01hematopoiesis is a multi step process
• 01:04through which the hematopoietic stem cell.
• 01:06Shades to give rise to all the cells
• 01:09that are present in the blood,
• 01:12and this process is highly regulated
• 01:14by a number of transcription
• 01:16factors and cofactors,
• 01:17and these factors work together as
• 01:20networks to promote differentiation to
• 01:22our specific lineages such As for example,
• 01:25the average ridlen age that I'm going
• 01:28to focus on today and at the same
• 01:32time these transcription factors.
• 01:34Inhibiting the other alternate
• 01:36hematopoietic lineages.
• 01:36Now what's interesting is that the
• 01:39transcription factors that drive
• 01:41every trade or other lineages
• 01:43drive differentiation.
• 01:44There are expressed at very high levels
• 01:47at the specific stage when they are
• 01:50important to drive differentiation,
• 01:52usually towards the later stages.
• 01:54But most of these factors are
• 01:57also expressed in hematopoietic
• 01:59stem and progenitor cells,
• 02:01although at very low levels.
• 02:03Which suggests that the amount also
• 02:07dosage of transcription factors play
• 02:09critical role for cell fate determination.
• 02:13Now the importance of the dosage
• 02:15of transcription factors for the
• 02:17process of cell fate decision
• 02:19has been established early on by
• 02:22reprogramming experiments in cell lines.
• 02:24So, for example,
• 02:25the lab of tumors graph in 1995
• 02:28was able to reprogram myeloblastic
• 02:30cell lines 2 hours different fate
• 02:33depending on the amount of the data.
• 02:36One footings that was Ectopically
• 02:38expressed so specifically,
• 02:39the lab showed that if that high
• 02:42levels of get a one.
• 02:44We drive differentiation to us every
• 02:46trade lineages or megakaryocytes,
• 02:48while intermediate levels of get
• 02:50a one will drive differentiation
• 02:53too as the elzina fear.
• 02:54Now these early experiment really
• 02:57demonstrated that different
• 02:58amounts of the data one put in
• 03:01can promote alternate cell fate.
• 03:03Now, more recently,
• 03:04linear programming experiment in
• 03:06in primary cells have also shown
• 03:09the importance of the dosage of
• 03:12transcription factors for the
• 03:13process of cell fate decisions.
• 03:15So for example, in this publication here,
• 03:19the authors have reprogrammed
• 03:21human fibroblasts into every trade
• 03:23progenitors by using this cocktail
• 03:26of four transcription factors get
• 03:28away until one element 2 and C myc,
• 03:31which they called the GTLM.
• 03:33Cocktail, and for these reprogramming
• 03:35experiments they have used a retroviruses,
• 03:37and each retrovirus expresses one
• 03:39of these transcription factors.
• 03:41And what's very interesting is they
• 03:43found that depending on the relative
• 03:46levels between their retroviruses
• 03:47that they use in their experiments,
• 03:50they can obtain completely different results.
• 03:52So for example,
• 03:53if they use a GT LM ratio of 2111,
• 03:57then these these reprogram cells will
• 03:59produce exclusively red cell colony,
• 04:02which suggests that those are
• 04:04functionally which would progenitors.
• 04:06However, if they use a
• 04:07slightly different GT LM ratio,
• 04:09for example 1112,
• 04:11then the reprogram cells will
• 04:12not produce red cells colonies,
• 04:14suggesting that these are not functional.
• 04:17So these these experiments have
• 04:18shown that it's not only the dosage
• 04:21of the amount of transcription
• 04:23factors that's important,
• 04:24but really the whole ative levels,
• 04:27or just to comma tree of
• 04:29transcription factors.
• 04:30That is key for reprogramming efficiency,
• 04:32and it is the same for native hematopoiesis.
• 04:36Where it has been proposed that the
• 04:38transcription factors that drive
• 04:40differentiation towards alternate
• 04:42hematopoietic lineages Co expressing
• 04:45bipotential progenitors and that those
• 04:47changes in us to comma tree is what
• 04:50is important to drive differentiation
• 04:53to US1 fate or another phase.
• 04:57Now a number of gene regulatory
• 04:59networks have been established in
• 05:01an attempt to model this self,
• 05:03a choice that occur in this
• 05:05bipotential progenitors,
• 05:06and for example,
• 05:07here is a network model of the
• 05:09selfish choice that occur in the MVP,
• 05:12just the megakaryocytes every
• 05:14trade progenitors and those cells
• 05:16must choose between an arbitrate
• 05:18fate or a megakaryocytic faith.
• 05:20Now,
• 05:20this hematopoietic tree of
• 05:22differentiation that I have shown
• 05:24you so far was a commonly accepted
• 05:27model for hematopoiesis until
• 05:28about 6:00 or seven years ago,
• 05:30when the use of single cell RNA
• 05:33seek transformed it into something
• 05:35that looks more like this.
• 05:37Now,
• 05:37in this knew continuous model
• 05:39of hematopoiesis,
• 05:40there are no differentiation steps process,
• 05:42but instead the cells are gradually
• 05:45transitioning along hematopoietic lineages.
• 05:46And really in these types of models,
• 05:49the.
• 05:49Hematopoiesis is actually based on a
• 05:52continuum of changing failed probability.
• 05:54So in in that type of model they are
• 05:57known by potential opportunities,
• 05:59but they're only hematopoietic.
• 06:01For generators with sulfate
• 06:02probabilities that are becoming
• 06:04more and more restricted as the
• 06:05cells gradually transition along
• 06:07the hematopoietic trajectories.
• 06:08Now in his new types of models
• 06:11for him at a crisis,
• 06:13and the exact exact role of
• 06:15transcription factors and the
• 06:17importance of their quantitative
• 06:18changes for the process of self.
• 06:21Decision remains unclear,
• 06:22so our goal is to establish a dynamic
• 06:26model of erythropoiesis that can
• 06:29integrate those quantitative changes in
• 06:32the level of transcription factors overtime.
• 06:35Now I will.
• 06:36I will model system to study human.
• 06:38Every troop Oasis is an ex vivo
• 06:41differentiation protocol where we
• 06:42isolate the city certified positive
• 06:44stomach progenitor cells from cord
• 06:46blood or peripheral blood or bone
• 06:48marrow and those cells are then
• 06:50differentiated to observe Richard
• 06:52lineages by using a cocktail of
• 06:54growth factors and cytokines.
• 06:56Now here are the sum of the
• 06:58morphological changes that occur
• 06:59during this ex vivo electrophoresis.
• 07:01You can see at the at the beginning at
• 07:03Day zero we can see the hematopoietic
• 07:05stem and progenitor cells and then there
• 07:08are dramatic changes in morphology and
• 07:10at the end of differentiation we obtain
• 07:12complete indication on those cells.
• 07:14That we wanted to further study every
• 07:17true policies at the single cell level.
• 07:19And for this we decided to use
• 07:21my cytometry or site off,
• 07:23which allows one to measure
• 07:25putains in single cell.
• 07:27So for the site of experiment,
• 07:29we harvested cells at regular
• 07:31intervals during the crossover.
• 07:33Every trait differentiation and
• 07:34then we barcoded those cells are
• 07:37separately at different time points
• 07:39with Palladium isotopes prior to
• 07:41combining them and standing them
• 07:43together with a cocktail of antibodies
• 07:46contain including cell surface markers
• 07:48as well as transcription factors.
• 07:51And then we did a clustering analysis
• 07:53on the 27 markers at a different time
• 07:56points and this allowed us to identify
• 07:5818 clusters or cell populations.
• 08:01And because we did a temporal
• 08:02because we did re bar coded the
• 08:05samples at different time points,
• 08:07we could do a temporal analysis of
• 08:10those cell populations which allowed us
• 08:12to reconstruct the dynamic trajectory
• 08:14from the early March put on progenitor
• 08:16cells to the differentiated auto chromatic,
• 08:18which replies.
• 08:19And then we looked at.
• 08:21The transcription factors expression
• 08:23in single cells within these different
• 08:26populations along the retreat trajectory.
• 08:28And our first question was whether
• 08:30there is Co expression of antagonist
• 08:33transcription factors in early progenitors.
• 08:35So we looked at fly one which
• 08:37promotes differentiation towards
• 08:39megakaryocytes and PLF,
• 08:40one which promotes differentiation
• 08:41to repair itself,
• 08:43and we looked at them in the MVP
• 08:45populations that we gated with this
• 08:48combination of seven cell surface
• 08:50markers and what we found is that the
• 08:53vast majority of the MVP Dooku Express
• 08:56is antagonist transcription factors,
• 08:57KLF one and fly one.
• 09:00In single serve.
• 09:01So next we wanted to follow the
• 09:03changes in put in levels of KLF one
• 09:05and fly one as the sales proceeds
• 09:08towards irritated differentiation
• 09:10and you can see these results at
• 09:12the single cell level here and when
• 09:15we aggregate the results we found
• 09:17that there is a gradual increase
• 09:19in put in levels of KLF one and
• 09:22at the same time gradual decrease
• 09:24in foot in levels of fly one.
• 09:26But there is no abrupt switch
• 09:28after a specific population.
• 09:30What happens is that those
• 09:32gradual changes actually.
• 09:33Taking several populations to occur
• 09:36belongs in between the trajectory.
• 09:40And then we we decided to Ectopically
• 09:43Express a flag tag version of fly,
• 09:46one in early progenitors.
• 09:48And then again,
• 09:50we followed differentiation by
• 09:51site of analysis and this time we
• 09:54identified not one but two trajectories
• 09:56in addition to the average rate
• 09:58trajectory that you can see here.
• 10:00By high level of expression of the
• 10:02Alpha globin we also found a second
• 10:05trajectory here which correspond to
• 10:07the megakaryocytic trajectory as
• 10:08shown by high level of expression of
• 10:11the city 41 marker.
• 10:13And then when we followed the cells
• 10:16that express flag fly one by including
• 10:18a flag antibody in our cocktail,
• 10:21we could determine that only the
• 10:23cells that express flag fly one shown
• 10:26in blue are differentiating towards
• 10:28the megakaryocytic trajectory,
• 10:30while the cells that do not
• 10:32express factor one shown here in
• 10:34green continue to differentiate
• 10:36along the imagery trajectory.
• 10:38So this experiment showed that the
• 10:41ectopic expression of Flight 1 is
• 10:43able to deviate the cells from their
• 10:45preferred every trade tragic tree
• 10:47to take on a megakaryocytic faith.
• 10:50So overall,
• 10:50these results are supports the
• 10:53concepts that quantitative changes
• 10:54in transcription factor put in
• 10:57levels in individual hematopoietic
• 10:59progenitors are key determinants
• 11:00of the cell fate decisions.
• 11:02Now in our site of experiment we
• 11:05could account for all older cell
• 11:07populations that were present
• 11:09in our differentiation media and
• 11:12that includes the hematopoetic
• 11:14progenitors erythroid cells.
• 11:15Also some megakaryocytes and Milo itself.
• 11:18However,
• 11:19there was one major cell population here
• 11:22for which the identity was less clear.
• 11:26And as you can see,
• 11:28the cell population starts to accumulate
• 11:30from day four of differentiation and
• 11:32then it increases until the 11th
• 11:34and then starts to progressively
• 11:36decrease and then completely
• 11:38disappears from our differentiation.
• 11:42Such a medium.
• 11:43Now looking at the cell surface
• 11:45markers for this population,
• 11:47we notice that it expresses
• 11:48very high levels of CD 44,
• 11:50which you can see here on the heat map,
• 11:54but also on this Disney plot.
• 11:56And we also notice that this
• 11:58sector population expresses
• 11:59intermediate levels of CD123.
• 12:00But when we look closer on a Disney product,
• 12:03you can see the expression
• 12:05of CD123 is heterogeneous,
• 12:06with some cells expressing
• 12:08extremely high levels an other
• 12:10cells are moderate levels of CD123.
• 12:12We also notice that these cells
• 12:14express very high levels of data.
• 12:16Two,
• 12:16it's actually the cell population that
• 12:19expresses the highest level of of getting 2.
• 12:22In in our experiment,
• 12:24and it does express graded levels
• 12:27of the transferring receptor CD
• 12:3071 but is completely negative
• 12:33for the early marker CD 34.
• 12:36So based on these combination
• 12:38of cell surface markers,
• 12:40it suggested that these populations
• 12:42could represent battlefields which
• 12:44at first we found a surprising
• 12:46because basil fees are supposed
• 12:48to emerge from the granulocytic.
• 12:50So the model with branch of differentiation,
• 12:52not the average rate branch.
• 12:54However,
• 12:55these results are consistent with
• 12:57a single scientific data that have
• 12:59been published both from mouse bone
• 13:01marrow and from human bone marrow
• 13:04that have identified a cell population.
• 13:06In close proximity to the error rate,
• 13:09cells that with the gene expression
• 13:10profile that that is consistent with the
• 13:13battlefield identity transcription factors.
• 13:15Now why do we want to
• 13:17know Putin's documetary?
• 13:18First of all,
• 13:19because we want to better understand
• 13:21how this changes in stock.
• 13:23Yama tree, Dr,
• 13:24Cell fate decisions.
• 13:25But also because despite everything
• 13:27that we have learned from genomic
• 13:29studies and cheap seek about
• 13:31the binding of transcription
• 13:32factors to their target side,
• 13:34we still don't know how
• 13:36many transcription factors.
• 13:37Are there compared to the number of
• 13:39binding sites and we still don't
• 13:42know if cofactors are limiting
• 13:43compared to transcription factors,
• 13:46in which case the transcription
• 13:48factors must compete to recruit them,
• 13:50or whether cofactors are
• 13:52present in large excess,
• 13:53in which case there equipment would
• 13:56be highly facilitated.
• 13:58Now, if one wants to measure
• 14:00documetary between Putin's,
• 14:02it is necessary to use approaches
• 14:04that provide an absolute
• 14:05quantification of these proteins.
• 14:07So we decided to use a targeted
• 14:10mass spectrometry approach,
• 14:12which is called selected reaction
• 14:14monitoring our SRM, which,
• 14:15when coupled with the spiking of known
• 14:18amounts of isotopically labeled peptides,
• 14:20can provide an absolute
• 14:22quantification of footings.
• 14:23Now recently,
• 14:24a number of publications came
• 14:26out using quantitative mass
• 14:28spectrometry approaches.
• 14:29Which we are not SRM,
• 14:31so I just wanted to take a few
• 14:34minutes to explain why we have
• 14:37decided to use this SRM approach
• 14:39for our for our own purposes.
• 14:42So first of all I will start by
• 14:44saying that my spectrometry is not
• 14:46inherently quantitative an what it
• 14:48means is that the intensity of the
• 14:51signal that is measured by the mass
• 14:53spectrometer does not only depend
• 14:55on the abundance of the peptide,
• 14:57but also on a number of other criteria,
• 15:00such as the amino acid composition
• 15:02or the ionization efficiency,
• 15:03or other parameters that are
• 15:05not fully understood and what it
• 15:08means is that it's not unusual to
• 15:10be in situation like this one.
• 15:12When you have a very high abundant
• 15:14peptide like the blue peptide
• 15:16producing a low intensity signal by
• 15:19mass spectrometry or lower opponents
• 15:21peptide like the red peptide,
• 15:23producing a very high intensity signal.
• 15:27Now to overcome this limitation and
• 15:29still being able to extract quantitative
• 15:31information from my spec data,
• 15:33a number of strategies have been proposed.
• 15:36Anna very popular strategy is
• 15:38called the Proteomic Ruler Method,
• 15:40which has been established by the
• 15:42lab of Matthias Mann to estimate
• 15:44copy number of proteins preserve,
• 15:46so this method is using the mass
• 15:49spectrometry signals from histones
• 15:50as an internal standard to quantify
• 15:52all the other footings,
• 15:54and this is this is a very effective
• 15:57method to estimate putting copy numbers.
• 16:00It's also high throughput.
• 16:01You can measure thousands of
• 16:03proteins in a single experiment,
• 16:05and it is quite easy to do and it
• 16:08has been used in and in a lot of
• 16:11different publications and I have
• 16:13highlighted some of these publications
• 16:16here for high throughput studies.
• 16:18So,
• 16:19however,
• 16:19in our case we want it to be
• 16:21able to distinguish even subtle
• 16:24differences in the circulatory
• 16:26between transcription factors.
• 16:28So so instead we decided to
• 16:30use a different approach,
• 16:32which is based on the spiking of
• 16:35known amount of isotopically labeled
• 16:37peptides that are used as internal controls.
• 16:40So in this approach,
• 16:42each individual peptide is
• 16:44quantified using an isotopically
• 16:45labeled version of itself,
• 16:47as illustrated here.
• 16:49Which we called the the silk
• 16:51peptide and this approach has been
• 16:54shown to be extremely sensitive
• 16:56to dynamic range,
• 16:57is very large and it's currently
• 16:59considered as a gold standard for putting
• 17:03quantification by mass spectrometry.
• 17:05So to use to establish SRM assay
• 17:08for putting quantification,
• 17:09we collaborated with Jeff and Ishan
• 17:12Margulis pee from the Institute
• 17:14for Systems Biology in Seattle,
• 17:16and we developed SRM assay and
• 17:19optimize Sri massive over 100 Putin's
• 17:21and we decided to use a different
• 17:24types of proteins, including.
• 17:26DNA binding transcription factors,
• 17:28but also compressors, coactivators,
• 17:30some chromatin remodeling enzyme,
• 17:32as well as some subunits of the
• 17:35general transcription machinery.
• 17:36And overall we quantified 103
• 17:38proteins at 13 time points during the
• 17:41course of erythroid differentiation.
• 17:43Now I should mention that even though
• 17:46these SRM assay are quite expensive
• 17:48to develop and they take time,
• 17:51but once they have been developed
• 17:55they can be reused.
• 17:57Much cheaper way and fast way to be
• 18:00able to quantify the same protein in
• 18:04in a different cellular environment.
• 18:06So because we wanted to build
• 18:09gene regulatory network,
• 18:10we also had to measure changes
• 18:12in our near levels.
• 18:13So again,
• 18:14we have just excels at regular intervals
• 18:17during the course of their true policies,
• 18:19and we measured our in a level by RNA
• 18:24sequencing and put in levels by SRAM.
• 18:27And first,
• 18:27focusing on the RNA data.
• 18:29Here I'm showing this irony Sickbay
• 18:32Suretrade trajectory overtime along
• 18:33these two principal components,
• 18:35and as you can see,
• 18:37there is a nice purple nice progression
• 18:39from the hematopoietic stem and
• 18:42progenitor cells at day zero to the
• 18:44differentiated erythroid cells shown here.
• 18:47Now looking at the proteins that we
• 18:49have measured by hospital clustering,
• 18:51you can see that there are different
• 18:54clusters and these clusters appear
• 18:57to be a temporary Co regulated.
• 19:00Now the first question we wanted
• 19:02to ask was whether there is a good
• 19:05correlation between putting an M
• 19:06RNA during erythropoiesis and 1st.
• 19:09We looked at global correlation
• 19:11as illustrated here.
• 19:12Looking at M RNA level versus
• 19:14protein levels and what we found is
• 19:17that there is really quite a good
• 19:19correlation in differentiating cells.
• 19:21However,
• 19:22the correlation is extremely low in
• 19:24hematopoietic stem and progenitor cells,
• 19:26which is consistent with published
• 19:28data that have shown that the
• 19:30protein translation rate.
• 19:32Is extremely low in the hematopoietic
• 19:34stem and progenitor cells.
• 19:36So next we wanted to look at the
• 19:39correlation between our input in
• 19:41changes overtime during differentiation
• 19:43as illustrated here and what we
• 19:45found is that for most genes there
• 19:48is a positive correlation.
• 19:49However,
• 19:49for some genes there is low correlation
• 19:52and even sometimes a negative correlation.
• 19:54So for example here if we look
• 19:57at the fly one or get the two.
• 20:00She and you can see that there is
• 20:02a very good correlation between
• 20:04changes in putting an M RNA levels.
• 20:07However,
• 20:08for other genes such as get our Nortel one,
• 20:11you can see the Putin level is
• 20:13increasing much faster than the our name,
• 20:16suggesting an important contribution
• 20:17of the post transcriptional
• 20:19regulatory mechanisms.
• 20:20So if you are interested
• 20:22in looking at other genes
• 20:24that we have measured,
• 20:25we have we have done a website here
• 20:28which I'm going to try to access and I
• 20:31hope it is going to work so I'm just
• 20:35gonna stop sharing for a few seconds.
• 20:38And try to share.
• 20:46My other sites. OK,
• 20:47so I hope you can see it so so perfect.
• 20:52Thank you so on these websites we
• 20:54have to drop down menu here with the
• 20:57proteins that you have measured.
• 20:59So I just wanted to show you an
• 21:01example of proteins that going up
• 21:03during differentiation where there
• 21:05is extremely good correlation so
• 21:07you can see so on the left side.
• 21:09Here we have put in copy number per
• 21:12cell and the blue graph correspond to
• 21:15put in an on the right we have the RNA.
• 21:18And from left to right we have
• 21:21a differentiation from MPP to
• 21:22Basel figurative glass,
• 21:23so you can see this two hour and then
• 21:26put in correlate extremely well for six,
• 21:296, but actually for for a lot of
• 21:31proteins there is a very low correlation
• 21:33or or even a negative correlation.
• 21:36I just want to show you one more one.
• 21:39Another example kept away which
• 21:41is a histo nasty transfer is also
• 21:44called GCN 5 and you can see for
• 21:46this put in the RNA and put in.
• 21:49Growth curves have almost
• 21:51nothing to do with each other,
• 21:54suggesting a highly complex
• 21:57posttranscriptional regulation mechanism.
• 21:59So.
• 22:01I will just go back to my presentation now.
• 22:06OK, perfect,
• 22:06so as you could see,
• 22:08we found that there are really
• 22:11major discrepancies between
• 22:13mRNA and put in and put in.
• 22:15Abundance is an.
• 22:16This is for master regulators
• 22:18of erythropoiesis,
• 22:19suggesting that gene regulatory networks
• 22:21should not be limited to M RNA but
• 22:25should also integrate put things.
• 22:27And this is what we tried to
• 22:29do in collaboration with the
• 22:31computational biology step.
• 22:32Perkins and Daniel Sanchez start
• 22:34level from our Institute and
• 22:36basically they they wanted to build
• 22:38a dynamic network model of every
• 22:40trade commitments that incorporates
• 22:41the quantitative changes in
• 22:43transcription factor protein levels.
• 22:45So in the model,
• 22:46M RNA are considered as the targets and
• 22:49Putin's are considered as the regulators.
• 22:51Putins can be activated so
• 22:53they can be repressors.
• 22:55And basically what they did is to
• 22:59use ordinary differential equations
• 23:01to try to predict the kinetic
• 23:03of the change in M RNA levels by
• 23:07the changes in the quantitative
• 23:09changes in protein levels.
• 23:10And here is the the models that
• 23:13they they established around 14
• 23:16transcription factors at different
• 23:18days during differentiation.
• 23:20Now what distinguishes this model
• 23:22from previously published model is
• 23:24that in quantifies the strength of
• 23:27the identified regulatory relationships
• 23:29as a measure of the contribution of
• 23:32all proteins and this is shown by
• 23:35different levels of transparency.
• 23:37So for example here,
• 23:38if we look at ranks one and get it to,
• 23:41you can see that the activation
• 23:43of getting 2 by ranks one is much
• 23:45stronger than the activation of ranks.
• 23:47One by getting 2.
• 23:49The model is also dynamic in
• 23:51that it reveals changes in these
• 23:54regulatory relationships overtime,
• 23:55so you can see, for example,
• 23:58at day zero,
• 23:59the strongest regulatory relationships.
• 24:01Involve proteins that are important
• 24:03in progenitor cells,
• 24:04but those links are progressively
• 24:07decreasing in strength and they
• 24:09are fading away and that
• 24:11they tend in committed cells.
• 24:13They have been replaced by other links.
• 24:17That that binds proteins that are known
• 24:19to be important in differentiated cells.
• 24:22We also notice that our model is able to
• 24:25correctly recapitulates the timing of the
• 24:27transcription factor across antagonisms
• 24:29that underlies the cell fate decisions.
• 24:32So here you can see at day zero,
• 24:35the model was able to capture the P1,
• 24:38get a one across antagonism,
• 24:39which regulates a very early self,
• 24:42a choice between my
• 24:44Louisiana Richard lineage.
• 24:45And that they too were able to capture
• 24:48a second subsequent cross antagonism,
• 24:50the KLF one,
• 24:52Fly 1 cross antagonism that regulates
• 24:54cell fate decision that took on later
• 24:57between the erythroid cells Anna
• 24:59megakaryocytes and both of these
• 25:01cross antagonisms have completely
• 25:03disappeared at day 10 in committed cells.
• 25:07Now I will another interesting
• 25:09aspect of the model is that it allows
• 25:11us to compare the strength of the
• 25:13different transcription factors to
• 25:15the regulation of their target genes.
• 25:17So for example,
• 25:18if we look at the data 1 gene,
• 25:21you can see that the contribution of
• 25:23tell one to the activation of get
• 25:26a one is about twice as important
• 25:28compared to the contribution of GATA 2.
• 25:31And surprisingly,
• 25:32this is not because tell one is a
• 25:34stronger activator of this gene,
• 25:36because when we look.
• 25:38Get the activation strength and one
• 25:40is in fact a weaker activator of the
• 25:42ghetto engine compared to get that too.
• 25:44But what happens is that 10 one is
• 25:463 times more abundant than get a 2.
• 25:49So basically what this shows is
• 25:51that Taiwan is a week,
• 25:53but opponents of activators of the
• 25:55data 1 gene and get a 2 even though
• 25:58it's a stronger activator of the gene.
• 26:01Overall it contributes less to its
• 26:03activation because it is less abundant.
• 26:05So with this type of models we
• 26:07are really able to dissect the
• 26:09quantitative relationships between
• 26:11these transcription factors.
• 26:12And that's very important if we
• 26:15want to be able to control every
• 26:18trapezes either in vitro or in vivo.
• 26:21So because they built different gene
• 26:24regulatory networks at different time points,
• 26:26we pieced them all together and
• 26:29made little movies where you can see
• 26:32the quantitative changes in this
• 26:35regulatory relationships as the cells
• 26:37proceed towards every trait differentiation.
• 26:40So you can see that this early
• 26:43regulatory here relationships
• 26:44decreases with time and are replaced
• 26:47by other regulatory relationships
• 26:49in more differentiated cells.
• 26:54OK, So what about cofactors?
• 26:56So everything I showed you so far was about
• 26:59the DNA binding transcription factors,
• 27:02but we know that these proteins works
• 27:04with the recruitment of cofactors,
• 27:06so we also wanted to measure cofactors by SM.
• 27:11And we started with the
• 27:13histone acety transfers P300.
• 27:15Now, according to the current literature,
• 27:17P300 has been shown to interact with
• 27:20dozens of transcription factors,
• 27:22so we expected it to be highly abundant
• 27:24in the nucleus, but much more,
• 27:27well, surprise.
• 27:27We found that not only P 300 is not in
• 27:30excess of the transcription factors,
• 27:33it is in fact present at a lower
• 27:35opponents than almost every single
• 27:38transcription factor taken individually.
• 27:40So then we looked at the other histone
• 27:44acetyltransferase is CBP and catuara
• 27:46GCN 5 and as you can see those are
• 27:50just as well in the nucleus as P300.
• 27:53So the next logical thing was
• 27:55to look at the antagonist,
• 27:57enzymes, histone deacetylase,
• 27:58and as you can see,
• 28:00those are much more abundant than
• 28:02the histo nasty transparency.
• 28:03So there is a quantitative
• 28:05imbalance between the edge ducks
• 28:07and their hearts in the nucleus.
• 28:09And then we looked at a broader
• 28:12range of cofactors.
• 28:13And what we found is all the
• 28:15coactivators that we have measured
• 28:17are actually really low abundance
• 28:19compared to the compressors.
• 28:21So for example,
• 28:22if we focus on CHD for sociology,
• 28:24far is one of the enzymatic subunits
• 28:26of the Nerd core oppressor complex.
• 28:29And as you can see,
• 28:31there is an almost 50 fold excess
• 28:33of CHD 4 compared to the to
• 28:35the Co activators in nucleus,
• 28:37and these results likely explain the
• 28:39fact that they're not complex has been
• 28:42identified as an interacting Putin.
• 28:44In almost every single put on the screen,
• 28:47now that have been published so far.
• 28:50So we found that the compressors
• 28:52are highly opponent.
• 28:53The coactivators are rare,
• 28:55so where do the transcription
• 28:58factor of fit on that scale?
• 29:00And the answer is that they
• 29:02fit somewhere in the middle.
• 29:04So here on this graph we have the
• 29:06copy number of Putin's in log scale,
• 29:09and as you can see,
• 29:11there is about an order of magnitude
• 29:13difference in your balance between Co
• 29:15activators and transcription factors.
• 29:17And then another order of magnitude
• 29:19difference between the transcription
• 29:21factors and the Co oppressors.
• 29:22And Interestingly,
• 29:23we could detect these differences
• 29:25in abundance only because we
• 29:27looked at Putin's because when we
• 29:29do the same analysis with M RNA.
• 29:31As you can see,
• 29:33there is no statistically significant
• 29:35differences in the abundance of
• 29:37the transcripts that code for
• 29:39these different classes of factors,
• 29:41which suggests that these very
• 29:43large differences in abundance are
• 29:45regulated post transcriptionally.
• 29:47And indeed, consistent with that,
• 29:49we found that when we inhibits
• 29:51protein translation by using by
• 29:53treating the cells with cycloheximide,
• 29:55you can see that the coactivators
• 29:57appear to be much less stable compare.
• 30:00This occurred pressers.
• 30:02So at least one of the reason why
• 30:05the coactivators are so where in
• 30:08a cell is
• 30:09because those appears to be
• 30:12extremely unstable footing.
• 30:14And this is true throughout every trip Oasis.
• 30:17Here I have plotted the Putin
• 30:20transcript average and as you can see
• 30:23the coactivators here are present
• 30:25between 100 and 1000 copies per
• 30:28Nicholas there transcription factors.
• 30:31On average, between 1000 and 10,000 and
• 30:35the oppressors between 10,000 and 100,000.
• 30:39So overall, these results suggest that
• 30:41the nucleus is a highly repressive
• 30:43environment where coactivators are limiting,
• 30:46which implies that the DNA bound
• 30:49transcription factors must compete with
• 30:51each other to mediate the recruitment
• 30:53of this limited number of coactivators.
• 30:56Now, one way through which the
• 30:59transcription factors can increase their
• 31:01likelihood of being able to recruit them,
• 31:04the limiting coactivators is through
• 31:06assembling themselves onto enhancers.
• 31:08So we wanted to ask the question,
• 31:11how do the number of active enhancers
• 31:13and coactivator molecules compare?
• 31:15So we looked at Co activators that are
• 31:18known to associate to active enhancers,
• 31:20which are the histone acetyl transferases,
• 31:22CPP 300 and the HTK for monometer
• 31:26transfer is is a mileage 3 MLN 4?
• 31:29And we have estimated the number of
• 31:32active enhancers by doing a taxic
• 31:34experiments which we have analyzed by
• 31:36hint Attack to identify transcription
• 31:39factor footprint and then intersected
• 31:42these with predicted enhancers from database.
• 31:44And this is the number of estimated
• 31:48active enhancers that we have obtained
• 31:50and as you can see these numbers are
• 31:54in the same order of magnitude as the
• 31:57number of molecules of coactivators.
• 31:59Suggesting that maybe the formation of
• 32:02these active in home sales depend on
• 32:05the availability of these coactivators.
• 32:07So this is of all our model where we have
• 32:10the transcription factors recruiting
• 32:12the RECO activators and the rest of
• 32:16the nucleus containing corepressor's.
• 32:18And we think that this.
• 32:23Hello hello hi how are you doing?
• 32:28Well, not there are you in
• 32:30the wilderness or in the camp.
• 32:32Sorry, my treat doctor Wiseman.
• 32:34Can you hear me OK?
• 32:37Actually going out on Sunday.
• 32:39So OK, so you missed some buddies
• 32:42working before Doctor Wiseman.
• 32:44Please mute yourself.
• 32:46I muted him. Thank you.
• 32:49I I'm almost done anyway.
• 32:52Thank you so so we think basically
• 32:55that this restriction of abundance
• 32:57of coactivators could be an
• 32:59important mechanism for concerted
• 33:01gene regulation and we think it
• 33:04may be particularly important in
• 33:06multipotent progenitors to prevent
• 33:08high levels of Co expression of
• 33:10genes from different lineages.
• 33:12So basically decreasing the
• 33:14transcriptional noise as well.
• 33:18And I will stop here.
• 33:20And I've knowledge people who did the work,
• 33:24so I already acknowledged all
• 33:26our collaborators an from my lab.
• 33:28This project was done mainly by two
• 33:31very talented postdoctoral fellow,
• 33:32common Poly who did most of the wet lab work.
• 33:36Xena Motory, who did the informatic.
• 33:39Analysis, With some help
• 33:41from a Kathy Silverman,
• 33:42who also did the informatic analysis,
• 33:44and I thank you all for listening and
• 33:47I'll be happy to take any question.