Fluorescent Protein Indicators

Top

01:00:15.06 Ok, I'm Roger Tsien. And today, I'm gonna talk to you about fluorescent protein indicators.
01:00:21.21 In other words, molecules that have been engineered using fluorescent proteins
01:00:27.00 to then report to us crucial biochemical signals
01:00:31.26 especially their dynamics and spatial localization
01:00:34.09 inside cells.
01:00:36.10 So basically, there are four major classifications of
01:00:41.28 how these indicators have been made.
01:00:43.09 And they're listed, first two, are listed here.
01:00:47.11 The first and simplest class in some ways
01:00:49.24 are the translocating fusions in which in order to monitor the activity
01:00:54.09 of enzymes such as PI-3-kinase
01:00:57.21 that's phosphatidyl inositol-3 kinase
01:01:00.20 or protein kinase C, calmodulin kinase type II
01:01:05.09 et cetera, G-protein coupled receptor desensitization;
01:01:09.22 in all of these cases, the signals have caused the binding protein to move around inside the cell
01:01:18.20 from one location to another.
01:01:20.13 And all we have to do is tag that binding protein or motif
01:01:23.23 with a fluorescent protein and watch where it goes.
01:01:27.04 And when it goes, if we did our biochemistry right,
01:01:30.22 that'll tell us that the original signal was generated.
01:01:33.14 And I'll give you one example of that.
01:01:34.19 The next class is when we engineer a GFP or YFP
01:01:41.26 to monitor things like directly pH or
01:01:45.13 indirectly chloride redox potentials,
01:01:49.04 calcium and so on.
01:01:49.27 And all of these are by introducing mutations into the GFP
01:01:54.23 and sometimes whole chunks of other proteins
01:01:56.29 that enable the equilibrium between
01:02:00.18 the tyrosine anion and the tyrosine neutral form for the chromophore to be
01:02:05.20 altered by the biochemical signals we're trying to measure.
01:02:08.10 And all of these just use a single fluorescent protein.
01:02:12.16 Then a large class of indicators is based on pairs of fluorescent proteins
01:02:19.13 of typically two colors where they are the shorter wavelength one
01:02:25.28 can do fluorescence resonance energy transfer
01:02:29.04 and it serves as the donor; we excite it directly.
01:02:32.12 And if the other protein of the longer wavelength which is called the acceptor is within a few nanometers of distance
01:02:40.05 and is appropriately oriented
01:02:42.09 then the longer wavelength protein steals the energy from the shorter wavelength protein
01:02:47.01 and remits typically.
01:02:50.02 But in any case, the efficiency of that energy stealing or energy transfer depends on distance and orientation
01:02:57.11 And therefore, any biochemical signal that changes the relative distance or orientation
01:03:03.27 between the our pair of fluorescent proteins
01:03:07.20 will cause a change in FRET.
01:03:11.00 And in a way, this is FRET, I should remind you, is somewhat similar to the basic way that the
01:03:15.26 GFP in the jellyfish took the energy from aequorin.
01:03:20.08 Aequorin glows blue, in that case, it is not fluorescent.
01:03:23.24 It's bioluminescence, but once you make the excited state, it doesn't matter.
01:03:27.13 And what the jellyfish modulated or changed blue to green
01:03:32.25 is what we deliberately modulate in order to measure all sorts of signals
01:03:36.05 ranging from calcium, cyclic nucleotides, protein kinase activation,
01:03:41.13 the activation of proteases, even membrane potential and all sorts of protein-protein pair interactions.
01:03:47.23 And finally, I'll give you one brief example
01:03:49.22 of bimolecular fluorescence complementation
01:03:52.06 which uses the special property that some fluorescent proteins can be cut into separate halves
01:03:58.06 not literally that they started as one complete half,
01:04:01.22 and were cut by a protease, but rather we synthesize them
01:04:05.11 as two separate fragments. And those fragments don't work and don't become fluorescent
01:04:09.24 by themselves. But if they're held - welded together in proximity for a little while
01:04:13.29 they can find the partner and generate a fluorescent protein
01:04:18.05 out of their two separate halves.
01:04:20.16 And therefore, anything that brings them into proximity, these fragments, will generate a fluorescent signal.
01:04:26.08 So here's the translocating case.
01:04:29.18 This is class number 1.
01:04:31.21 And I only have this one example, but it's pretty simple
01:04:35.01 in which the activation of phosphatidyl inositol-3 kinase,
01:04:40.10 in this case triggered by the addition of
01:04:43.18 platelet derived growth factor to these fibroblasts,
01:04:46.05 these NIH 3T3 cells,
01:04:47.18 will generate the unusual phospholipid, phosphatidyl inositol(3,4,5) trisphosphate,
01:04:55.21 that's PI(3,4,5)P3.
01:04:58.19 And we have a domain that's long been known.
01:05:02.17 And in this particular case, it was stolen from a protein kinase called
01:05:05.03 AKT, and the domain is called the pleckstrin homology domain
01:05:08.28 That's why it's called PH.
01:05:11.13 And that piece from AKT was fused to GFP
01:05:15.18 and expressed in the cytosol of the cells.
01:05:18.21 Now, before we stimulated, this GFP is just sitting in the cytosol
01:05:22.14 like diffuse GFP, and there's nothing much to see.
01:05:25.14 But that's just before we added the PDGF
01:05:28.24 And when we add the PDGF, or when these workers added these PDGF,
01:05:31.26 this is from Tobias Meyer's lab
01:05:34.18 in the next few minutes, that growth factor acting through the tyrosine kinases initiates a signal transduction cascade
01:05:43.11 that generates this lipid in the plasma membrane
01:05:47.17 and then the AKT-PH domain notices that there's this lipid
01:05:51.29 and wants to go bind to it,
01:05:53.17 so when you look carefully at this micrograph,
01:05:55.19 you will see that the GFP is not completely but to a large extent jumped over to the plasma membrane and lit up the plasma membrane
01:06:02.21 and left the inside of the cell a little bit denuded
01:06:06.04 of proteins, somewhat denuded.
01:06:07.27 And then, later on, just to prove that this was reversible,
01:06:10.22 they put on a drug that inhibits the kinase,
01:06:14.03 and that's called Wortmannin, and that allows the PI(3,4,5)P3 to be broken down
01:06:19.28 by other processes. And the whole process reverses,
01:06:20.09 and the GFP goes back to being in the cytosol.
01:06:26.06 And this is the time graph of the relative intensity of the plasma membrane compared to the rest of the cell
01:06:31.19 going first up with a little bit of a delay
01:06:34.13 due to the PDGF and then coming back down with Wortmannin.
01:06:38.16 Now, let me move on to class 2.
01:06:43.03 These are the GFPs that are intrinsically changing their fluorescence properties
01:06:48.20 in response to an external biochemical signal.
01:06:53.07 The original GFP was already somewhat pH sensitive.
01:06:56.18 And these are spectra taken at different pHs
01:06:59.29 and showing that the spectra a little bit goes up and down.
01:07:03.14 But this can be improved or amplified by mutagenesis.
01:07:09.04 If this is what you want, and these proteins called pHluorins, which is a clever pun,
01:07:15.09 pH for the pH, and pHluorins sounds like fluorescence.
01:07:21.22 And these proteins will do all sorts of things, which for example,
01:07:24.27 we can actually change the balance between the left hand or UV peak
01:07:29.18 versus the right hand, which is the visible peak
01:07:32.22 by changing the pH,
01:07:35.13 or we can just make a much bigger change in total intensity
01:07:41.02 - the so-called ecliptic type of pHluorin.
01:07:43.01 That is just a different set of mutations.
01:07:44.12 And the place that is become really widely used in cell and neural biology
01:07:49.05 is to monitor exocytosis
01:07:51.03 because when you put this protein into a secretory vesicle,
01:07:56.22 you should know that the inside of the secreted vesicle is usually quite acidic
01:08:00.29 when it's sitting inside a cell like typically pH 5 or so,
01:08:05.02 but if those contents get exocytosed,
01:08:08.19 then the protein now suddenly sees the extracellular normal pH, which is typically 7.4.
01:08:15.02 So that protein reports that that secretory granule or synaptic vesicle just got exocytosed.
01:08:24.15 And we can even use this now and has been used
01:08:27.16 as in neurobiology
01:08:28.27 where we want to monitor synaptic release,
01:08:31.21 and we can do so. And this is the case in flies
01:08:35.06 where different odorants were put into the antenna of the Drosophila.
01:08:39.28 And now we're imaging with these pH sensitive indicators, which are being expressed in the processing machinery
01:08:47.06 and wherever the right odor causes the cell to exocytose and
01:08:52.03 do its synaptic transmission
01:08:54.05 we get a little burst of higher pH as seen by this synaptopHluorin.
01:08:59.13 In the case of the YFPs, it became possible to find mutants in which the pH dependence was shifted to becoming chloride dependence.
01:09:10.16 Well not entirely shifted, it's still pH sensitive,
01:09:13.01 but now at a fixed pH, the chloride will shift the effective pKa of equilibrium,
01:09:20.07 and so as long as the pH doesn't change, we could also measure chloride.
01:09:24.24 And this, for example, used to screen for drugs that would modulate chloride fluxes,
01:09:30.06 and the place that this was clinically important is that people who suffer from
01:09:34.12 the disease called cystic fibrosis have a problem
01:09:38.26 a mutation in their protein called the CFTR that conducts chloride and bicarbonate.
01:09:45.05 And it was then possible by putting the YFP into cells
01:09:51.05 and exposing them to different chloride concentrations
01:09:54.05 and test cases to monitor whether or not the cells were responding to chloride
01:10:01.11 that was being changed outside
01:10:02.29 or in this case iodide
01:10:05.00 and see whether the fluorescence was changed and determine whether or not drugs that could be applied were fixing the function of CFTR.
01:10:13.09 Now, for the case of calcium,
01:10:16.27 which is one of the most important messengers in lots of signal transduction including neurobiology,
01:10:23.22 it turned out to be necessary to modify the GFP even more drastically.
01:10:29.06 Original GFP has an N-terminus and C-terminus like any old protein,
01:10:34.27 and they sequenced the N-terminus, happens to start with MVSK et cetera
01:10:40.11 and ends at 238 amino acids later with
01:10:43.14 this particular lysine,
01:10:45.27 but there's been a trick known for awhile.
01:10:47.27 It sometimes works on many proteins, not always on every protein,
01:10:51.14 which is that if you can recode this protein so you have basically the same amino acids
01:11:00.02 except that you start them at a new location
01:11:01.29 and end them at a new location,
01:11:03.11 and in the process, you also have to link the old N-terminus and C-terminus together
01:11:08.07 with a few amino acids
01:11:10.09 This, by the way, does require that the N and C-termini of the protein you care about
01:11:15.24 be fairly close to each other in space
01:11:18.03 so that you don't have to make a very long linker.
01:11:19.26 Sometimes, this so-called circularly permutated protein will actually get up and work.
01:11:26.13 And in this case, what it has to do is become fluorescent.
01:11:29.03 So you can express this modified protein, which starts in the middle
01:11:34.18 of what used to be the middle of the old protein,
01:11:37.28 go around, and this will become fluorescence.
01:11:41.11 So in three dimensional terms, this is the 11-stranded beta-barrel
01:11:45.21 of GFP with the alpha-helix running up the center of the chromophore in the middle,
01:11:49.29 and the N and C-termini were up here.
01:11:53.10 That was the old N-terminus, and this is the old C-terminus.
01:11:57.16 And instead, we link these two together genetically
01:12:00.12 and make a nucleic acid that encodes a few amino acids, connect them, but we start the protein now
01:12:07.02 with a new methionine here
01:12:09.06 that becomes the new N-terminus, and it runs through this set of strands
01:12:15.01 around the barrel, more or less as the old one did
01:12:18.09 except it starts from a new place and ends in a new place.
01:12:20.11 And these proteins become fluorescent,
01:12:23.24 and they are a little bit more fragile than the original GFP
01:12:27.28 and that turns out to be useful
01:12:30.11 because we we can then take this new N-terminus
01:12:35.17 and fuse onto it calmodulin,
01:12:37.29 which is a famous calcium binding protein
01:12:40.03 and that actually went onto the C-terminus of the so-called circularly permuted enhance GFP.
01:12:50.02 And then into the new amino terminus we fused a peptide called M13 that calmodulin likes to bind to
01:12:59.08 when it sees calcium.
01:13:00.18 So basically, we've sort of created a wound in the side of the fluorescent protein
01:13:05.23 and then that wound sort of hangs open and depresses the fluorescence until the calmodulin sees calcium
01:13:13.09 Then the calmodulin grabs onto the M13,
01:13:15.25 they lock together, and they sort of locally stitch up the wound.
01:13:19.11 And we restore the fluorescence,
01:13:21.09 and this can all happen in just milliseconds or so
01:13:23.29 a few hundred milliseconds, which is the speed of calmodulin.
01:13:28.18 And this has become very popular; it's called GCaMP,
01:13:31.27 which is a sort of pun on calmodulin being stuck inside GFP.
01:13:38.10 And there's another flavor called pericams.
01:13:42.06 And these have been widely used now
01:13:43.03 specially neurobiology to follow calcium transients,
01:13:46.10 which are again good markers of activity,
01:13:48.28 In this particular case, the people doing this work are studying the activity of parallel fibers in the cerebellum, which are being electrically stimulated,
01:14:01.00 and when they are electrically stimulated,
01:14:02.00 they let in some calcium through voltage-operated calcium channels.
01:14:04.23 The GCaMP is expressed inside these cells
01:14:08.26 and momentarily flashes bright, and
01:14:11.04 with that, we can get spatial maps
01:14:13.25 again of where the neural activity is occurring inside the cell.
01:14:17.09 There's a lot more recent work, but I picked this example
01:14:20.18 which is a little bit older
01:14:21.13 because it's a very simple case to explain.
01:14:24.25 So now let me move on to the third class of indicators
01:14:30.08 which is not a single fluorescent protein
01:14:32.15 but to remind you, we have two colors of fluorescent proteins.
01:14:35.21 By far, the most common historically, not necessarily the ones we will use for the rest of scientific progress,
01:14:42.06 but historically, we have tended to use cyan and yellow.
01:14:46.11 They are a good pair that talk to each other nicely
01:14:50.07 and can be excited separately and detected.
01:14:52.06 We excite the cyan with violet light.
01:14:55.08 it tries to glow cyan, but if the yellow protein is nearby, the yellow steals the energy without any photon ever actually coming out.
01:15:02.21 And then the yellow then goes on to re-emit yellow-green as if it had directly been excited in the first place.
01:15:09.03 And the very simplest case is used; tie the two together
01:15:11.26 by a short leash
01:15:14.13 It can be up to 20 or 30 amino acids long.
01:15:17.00 And then as long as that's intact,
01:15:20.24 this energy transfer is very effective.
01:15:23.15 But then, if you come along with a protease that cuts this linker
01:15:27.21 and splits the two apart, now the cyan and the yellow drift apart,
01:15:32.01 they're no longer held together,
01:15:34.12 and all they have to get is a few nanometers apart,
01:15:36.20 and now the cyan is on its own,
01:15:38.09 you excite it with violet light, and it emits cyan
01:15:41.11 as implied by its name.
01:15:43.07 By the way, crucially important in this is the fact that fluorescent proteins themselves are not very easily proteolysed.
01:15:51.15 And fortunately, they are well-folded proteins,
01:15:53.04 and well-folded proteins don't have loose floppy bits,
01:15:56.27 whereas the linker we make is just unstructured amino acids,
01:16:00.21 and typically it is vulnerable to proteolysis.
01:16:04.05 This is an irreversible case
01:16:07.05 once you've broken this thing, you cannot put humpty dumpty back together.
01:16:10.25 The cyan and yellow have no reason to ever get back to each other.
01:16:14.11 So this is a one shot indication that a protease was active and made the transition.
01:16:19.27 Reversible case again uses this calmodulin and M13 motif;
01:16:27.12 remember when calmodulin sees calcium, it then grabs onto M13.
01:16:31.22 So in this case, if we then sandwich that pair,
01:16:36.11 between cyan and yellow fluorescent proteins,
01:16:40.05 the addition of calcium actually brings the cyan and yellow close to each other.
01:16:43.20 They were relatively far apart before;
01:16:46.13 they become closer and better oriented with respect to each other.
01:16:50.15 So in this case, calcium increases the FRET,
01:16:54.03 and when we take the calcium away, it can go right back to its original state.
01:16:57.28 So this process can go back and forth many many times as needed to signal.
01:17:03.16 And this case, we see calcium not as so much as a single intensity increase as it was in the GCaMPs,
01:17:10.14 but rather, a color change - the calcium giving us more yellow and less cyan now.
01:17:15.24 So two more examples, and then I'll show you some of the biology.
01:17:23.05 This is a case where we can look at cyclic AMP.
01:17:26.16 Cyclic AMP's original native biosensor inside most mammalian cells is called cyclic AMP-dependent protein kinase.
01:17:35.11 And cyclic AMP binding to the regulatory subunit, there are two of them in this tetramer,
01:17:42.00 causes a conformational change that makes the regulatory subunit spit out the catalytic subunit
01:17:48.17 stop binding to the catalytic subunit.
01:17:50.12 So it's fairly obvious; hey, put one color, in this case cyan,
01:17:54.28 on the regulatory; put another color on the catalytic
01:17:58.05 in the original form, they are doing FRET to each other.
01:18:02.01 They're held close to each other.
01:18:03.21 When the cyclic AMP comes on, it breaks up the FRET.
01:18:07.12 And this process is also reversible; when the cyclic AMP is destroyed, then these subunits find each other and go back to this state.
01:18:14.21 How can we see not just cyclic AMP but the action of any protein kinase?
01:18:22.07 As you know, protein kinases are the most ubiquitous and universal.
01:18:25.29 And important also the signal mechanisms,
01:18:29.11 and here we exploit the fact that there are certain known motifs that like to bind phosphorylated amino acids
01:18:36.24 but won't bind the original amino acid
01:18:39.07 before it got phosphorylated.
01:18:40.25 So now, this is a variant of the calcium sensor.
01:18:44.27 Now, the pink bit is the floppy peptide that's in the driver seat.
01:18:49.05 It's not M13 anymore; it's a substrate that the kinase wants to phosphorylate
01:18:54.08 if the kinase is active.
01:18:55.24 Once it gets phosphorylated, now there's a phosphate stuck on a serine, threonine, or tyrosine.
01:19:01.19 And if we have taken this binding domain correctly,
01:19:04.06 it will then want to eat the phosphoate or grab onto the phosphate
01:19:08.23 like a fist closing around it,
01:19:10.19 and that's shown here where the orange dot symbolizes the phosphate,
01:19:15.13 which is being grabbed by the catcher's mitt,
01:19:17.06 which is the binding domain.
01:19:19.28 And in this process, we change the relative orientation or distance between the cyan and yellow.
01:19:24.24 As I drew it here, it increases FRET, but depending on the details
01:19:28.10 of what this is really here, sometimes there there is a decrease in FRET.
01:19:31.09 But we don't really care as long as we get a big signal.
01:19:34.03 So here's some biological examples.
01:19:38.02 This is the protease sensor for apoptosis,
01:19:42.12 and you know that in apoptosis, there's the massive activation of a whole sequence of proteases
01:19:47.21 including the famous caspases.
01:19:49.13 And they like to cut at the sequence DEVD.
01:19:54.01 That's their favorite substrate known from other biochemistry.
01:20:00.00 This isn't all there is, but basically, this is the crucial bit
01:20:02.20 of the substrate.
01:20:03.17 And we put that between CFP and YFP
01:20:06.23 and we get strong FRET beforehand
01:20:09.14 and in this particular case, cells were induced to apoptose.
01:20:13.13 And what's interesting is that in any one cell, the transition is very sharp.
01:20:18.24 It only takes a few minutes to go from complete FRET to complete destruction of FRET
01:20:25.05 and complete proteolysis
01:20:27.22 of this molecule.
01:20:29.04 And the reason that this is biologically important
01:20:31.08 is that up to now,
01:20:32.15 before these experiments, people mostly had to watch apoptosis as a population of millions of cells
01:20:40.13 by the traditional biochemical techniques.
01:20:43.11 And when you look at millions of cells, they each do their transition
01:20:47.10 at a different time,
01:20:48.26 so in the population, they transition between not active and active caspase takes hours.
01:20:56.11 But when you can watch the single cells, you see that every cell does it in a few minutes
01:21:00.03 but that different cells do it at different times,
01:21:02.11 and that's what generates the spread of times.
01:21:05.14 A different form of proteolysis that doesn't actually use FRET,
01:21:10.14 but a very beautiful and interesting example is shown here
01:21:14.17 where the Miyawaki lab decided to make an indicator
01:21:19.01 for the cell cycle.
01:21:20.10 And in the cell cycle, we also have sets of proteases that turn on at different phases.
01:21:26.23 And particularly, we have some proteases that destroy proteins during mitosis
01:21:30.25 and other proteases that destroy proteins every time but when you're in mitosis.
01:21:35.08 And they manage to put YFP under the control of one of them
01:21:41.08 so that YFP is highly expressed during mitosis
01:21:50.00 and destroyed at other times;
01:21:52.24 whereas the red protein mCherry is exactly opposite.
01:21:57.27 And it's made persists when the cell is not in mitosis,
01:22:03.00 and then during mitosis, it gets destroyed.
01:22:05.27 So they're exactly out of phase with each other
01:22:08.18 and the fortunate easy way to remember this,
01:22:12.12 these workers picked the color scheme right so that green means go,
01:22:16.13 means you're in mitosis,
01:22:17.16 and red means stop
01:22:18.24 like traffic lights.
01:22:20.08 They could have perfectly well have done it the other way around,
01:22:22.08 but it would be harder for us to remember.
01:22:24.20 And so here's the movie
01:22:26.13 of a time-lapse of one week's worth
01:22:28.28 in the life of these two cells
01:22:30.15 which are dividing,
01:22:32.13 and as they divide, they flash through green to red to green to red
01:22:34.24 to green to red, and we can watch the cell cycle biochemically
01:22:39.03 and all the highly controlled activation destruction of these proteases.
01:22:46.11 Now you might say, why'd you need this?
01:22:49.16 Hey, we could have watched it, you know, just by seeing when one cell splits into two
01:22:53.26 we knew when mitosis occurred,
01:22:55.26 so why do you need a biochemical indicator?
01:22:58.09 Well, the place you need it is say when you are going to work in an intact mouse
01:23:01.09 and now you have a cancer that's been implanted into the mouse.
01:23:08.12 And you don't have the spatial resolution nor the continuous imaging to watch every single cell
01:23:13.13 to track when it does mitosis
01:23:17.01 and follow the history of every single cell.
01:23:18.27 But now, just by looking, we can see
01:23:21.10 that this cell population one day after implantation
01:23:25.08 is a sort of yellowish color
01:23:26.07 which is a mixture, therefore, of mitotic and non-mitotic cells,
01:23:31.19 but by sixteen days later,
01:23:32.11 it's all red,
01:23:33.12 which says that even without high resolution imaging,
01:23:36.02 all of these cells have stopped dividing.
01:23:38.27 That's what makes this a benign cancer cell.
01:23:41.11 This is a malignant cancer, and even through the flank of the animal,
01:23:46.03 it is visible.
01:23:47.28 You can detect that it's continually dividing,
01:23:50.21 and when we then do microscopy,
01:23:54.01 we can see the distribution of the mitotic cells
01:23:56.15 which in this case are close to the blood vessel
01:23:59.09 and the non-mitotic cells that have no longer rapidly dividing,
01:24:03.27 which are further from the blood vessel,
01:24:06.03 and this is just from a single snapshot
01:24:08.03 without having to watch the entire movie
01:24:10.22 for twenty seven days.
01:24:11.20 So here again are the cameleons
01:24:15.14 which is a blow-up showing again the mechanism by which calcium
01:24:21.02 can control the FRET between a cyan and yellow fluorescent protein
01:24:25.17 in a reversible way.
01:24:26.17 And just to show you a little more about, in this case,
01:24:31.07 the technical detail; we monitor the emission
01:24:33.17 in the yellow band
01:24:34.18 from the YFP as well as the cyan band from that one.
01:24:38.12 And what calcium does is it takes away cyan and gives it to the yellow band
01:24:44.01 because that's the increase in FRET.
01:24:47.06 And this ratio between yellow and cyan is particularly helpful to cancel out all sorts of artifacts
01:24:54.23 like the absolute amount of protein
01:24:55.19 or the absolute strength in illumination
01:24:58.14 or the absolute sensitivity of your camera
01:25:00.13 or all these sorts of fluctuations are cancelled out when we take the ratio
01:25:04.17 of two wavelengths.
01:25:06.09 And so here's an example from the earlist form of development.
01:25:11.00 This is a zebrafish embryo that has just been fertilized.
01:25:13.15 It is expressing cameleon as we call it,
01:25:18.16 which is again a pun on the word calmodulin
01:25:20.13 and the fact that it changes color.
01:25:22.17 And here, this is 30 minutes post fertilization.
01:25:27.20 And as the zebrafish begins to divide,
01:25:30.21 you will see cleavage furrows form,
01:25:32.23 and each time a cleavage furrow is beginning to form,
01:25:36.15 there's this flood of high calcium
01:25:38.18 preceding the actual constriction - the belt tightening -
01:25:42.14 that then makes the next cleavage furrow.
01:25:44.26 So that's two cell divisions into a group of four;
01:25:48.17 now we're going into eight.
01:25:50.08 And that's the next cleavage furrow
01:25:51.23 - time going by in minutes in timelapse
01:25:54.25 So this is demonstrating the tremendous importance of calcium in causing the motility
01:26:02.26 and guiding development in order to decide where the cleavage furrow is going to occur.
01:26:07.18 As another example of the biological usefulness of the cameleons,
01:26:12.16 and particularly here, it was useful to have a ratiometric indicator
01:26:17.18 that changed color, which made it possible to try to calibrate
01:26:21.08 the actual yellow-to-cyan ratio
01:26:23.21 in terms of an absolute calcium level.
01:26:26.05 And this was work on the effects of Alzheimer's protein, the famous Abeta
01:26:32.21 being misregulated and expressed.
01:26:35.27 Now, first we have here, looking at normal brain, this is actual
01:26:40.06 in vivo microscopy in the intact brain.
01:26:43.03 it's lightly labeled; you don't see all of the different dendrites and axons
01:26:48.14 because the genetics only cause maybe one in a hundred or one in a thousand cells
01:26:52.25 to express, but we can see these
01:26:54.27 sort of like tree trunks
01:26:56.20 and roots of the different neurons.
01:27:00.04 And they're all pretty much uniform blue-green pseudocolor.
01:27:04.16 This is not a real color.
01:27:05.15 This is the computer coding of what the ratio is.
01:27:08.22 And this is the color scale.
01:27:10.17 And a blue-green is somewhere around here.
01:27:12.27 And that tells you that the average calcium was on the order of about 80 nanomolars
01:27:18.03 - quite tightly distributed in this histogram.
01:27:21.03 This is the normal brain,
01:27:22.29 but after you have made the brain express
01:27:25.20 the Alzheimer's protein with the appropriate mutations,
01:27:29.08 then it turns out that some, not all, of the processes start to turn this
01:27:34.22 have a higher ratio of more FRET, indicating more calcium,
01:27:42.05 and there's still a normal population,
01:27:44.10 which is more or less the same.
01:27:45.08 But there is this additional small population,
01:27:48.12 which actually turns out to be about 20% of the processes
01:27:51.25 that has this elevated calcium
01:27:54.25 And this lasts and is sort of continuous;
01:27:59.03 it's not a sudden, you know, minute-by-minute fluctuation.
01:28:01.26 This is high, and it turns out these are highest when they are close to the plaques
01:28:06.18 that are the damaging deposits made in these mutants.
01:28:09.23 So here was essential to be able to calibrate the calcium scale - here -
01:28:15.03 in absolute terms,
01:28:16.04 and that's much more difficult with all of the other forms of indicators
01:28:20.01 than this two-color version.
01:28:22.09 A quick example on this cyclic AMP;
01:28:25.12 again, this is the basic mechanism that I have already described
01:28:28.15 where the regulatory subunit binds cyclic AMP,
01:28:31.01 splits itself off from the catalytic subunit,
01:28:33.22 thereby disrupting the close localization
01:28:37.01 of cyan with yellow.
01:28:39.08 And here, that pair of constructs - remember,
01:28:43.26 these are two separate proteins,
01:28:45.13 the regulatory that has the cyan,
01:28:47.12 and the catalytic that has the yellow.
01:28:49.09 They both have to be transfected at roughly equal amounts,
01:28:52.29 in this case, into a cardiomyocyte
01:28:56.07 and this was the first evidence that really demonstrated directly by imaging
01:29:01.22 that the cyclic AMP elevation
01:29:05.15 caused by different agents such as norepinephrine,
01:29:09.02 which is the hormone inside you that when you are getting excited
01:29:12.16 and your heart is beating faster,
01:29:13.17 it speeds up the heartbeat,
01:29:16.16 but it doesn't actually raise the cyclic AMP uniformly.
01:29:19.17 It uses micron scale variations within the heart muscle,
01:29:25.29 and that was shown here
01:29:27.06 by this imaging of the activation of this indicator.
01:29:30.24 And that turns out to be very important for the biology of agents
01:29:35.03 like these beta-adrenergic agents that they localized cyclic AMP signal, not just a general widespread one.
01:29:43.21 And this is the kinase phosphatase reporter
01:29:49.00 again using the substrate that gets phosphorylated,
01:29:53.07 and then that substrate, if phosphorylated, then
01:29:56.04 binds intramolecularly
01:29:57.21 to the phosphoserine, phosphothreonine, or phosphotyrosine
01:30:02.06 binding site.
01:30:03.12 And the CFP and YFP that are attached on either end of this internal construct then change the FRET
01:30:11.18 if we have arranged the engineering correctly.
01:30:14.06 And this movie is sort of cute. This is a fibroblast
01:30:18.17 that is expressing the reporter for the activity of the
01:30:23.10 epidermal growth factor receptor;
01:30:25.23 that's a famous tyrosine kinase
01:30:27.15 it's when epidermal growth factor binds to the outside,
01:30:32.12 then it triggers the dimerization of this receptor.
01:30:37.07 And then tyrosine phosphorylation occurs,
01:30:42.14 and we arranged this reporter to be phosphorylated along with many other things
01:30:47.08 And then that phosphorylation causes internal binding of SH2 domain,
01:30:53.18 which is a famous domain that likes phosphotyrosine
01:30:56.12 and we get an increase in FRET.
01:30:58.14 So here is the cell before it's been stimulated,
01:31:01.19 and then in a moment, and it started,
01:31:04.15 and when we apply EGF
01:31:05.22 - now -
01:31:06.17 and there's a flood of high FRET
01:31:10.06 that starts from the end of this thinnest dendrite
01:31:14.10 goes up into the center, and by now,
01:31:15.18 the signal is largely gone away,
01:31:17.13 but you may have seen that there was a flood of red
01:31:20.13 that went up into the center
01:31:21.18 and then we, what I didn't have time to tell you,
01:31:23.27 is that we then washed out the EGF.
01:31:26.05 And the signal went back, and the red tie sort of receded
01:31:31.12 back the way it had come.
01:31:33.14 Though up to now, I've been talking about sort of this traditional biochemical signals
01:31:42.01 that are mediated, that are soluble, things like calcium or cyclic AMP
01:31:47.17 or so on.
01:31:49.24 If you just want to watch when a pair of proteins gets together,
01:31:53.28 as a pair of proteins,
01:31:56.01 this form of induced protein-protein interaction, it's simple.
01:32:00.02 You put CFP on one end, on one of the proteins
01:32:02.10 YFP on the other.
01:32:03.14 There's not a guarantee that every such construct would put the cyan and yellow close enough
01:32:09.21 for good FRET.
01:32:10.17 But there's a reasonable chance, and you try things
01:32:13.03 until you find ones that work.
01:32:15.02 And in this particular case, our colleagues asked us to help them
01:32:18.14 look at the binding of nuclear hormone receptors -
01:32:23.27 basically steroid receptor types -
01:32:25.28 with other key proteins that are called coactivators
01:32:29.28 that are famous for helping drive transcription
01:32:33.23 once the hormone binds.
01:32:35.24 So here's a cartoon in which this is the hormone receptor.
01:32:39.06 The ligand is the hormone;
01:32:40.19 it comes in and causes a change in conformation that
01:32:43.18 makes the coactivator bind, and so we tag them with CFP
01:32:47.09 and YFP.
01:32:48.00 And looked at the cells, and that's the CFP on the retinoic acid receptor
01:32:53.18 and YFP attached to the so-called SRC-1
01:32:57.12 it's one of the famous coactivators.
01:32:59.18 We put on the hormone at this vertical line
01:33:01.24 or the drug that simulates the retinoic acid
01:33:06.07 and within twenty or thirty seconds,
01:33:08.24 already, we have a big increase in FRET
01:33:12.04 that levels off rapidly
01:33:13.07 And you can do it also with a real, natural ligand
01:33:17.00 trans-retinoic acid
01:33:18.13 Now, who cares about this?
01:33:19.21 Once again, we get the kinetics; this is done in a single cell.
01:33:23.02 And this showed the whole process that this biochemistry
01:33:26.24 happened in just a few tens of seconds.
01:33:29.26 Until this point, all the biochemical assays require
01:33:33.24 that you check for the affinity of
01:33:36.00 one protein for another
01:33:37.03 one of them was hung up on an affinity column,
01:33:39.15 we poured the other protein past it,
01:33:41.27 and the assay took twenty or thirty minutes.
01:33:44.22 And everyone thought that this sort of event took tens of minutes,
01:33:49.02 but it actually takes tens of seconds.
01:33:51.01 And that's at room temperature;
01:33:52.07 it's probably five seconds or so at body temperature.
01:33:55.09 My last example is now class 4,
01:33:59.18 which I haven't talked about up to now.
01:34:01.14 And that's the so-called bimolecular fluorescence complementation
01:34:05.10 where we take these two fragments
01:34:08.18 of fluorescent proteins
01:34:10.10 that each by themselves is impotent
01:34:13.16 unable to make fluorescence,
01:34:15.27 and we clamp them together for a little while in proximity
01:34:18.17 and in a sort of strange variation on circular permutation,
01:34:22.06 these two fragments can now cooperate
01:34:25.00 with each other.
01:34:26.07 In the circular permutation, they were held together,
01:34:28.20 but just out of the normal order
01:34:30.29 so back to front.
01:34:32.02 Here, we take some other scaffolding
01:34:34.25 and hold them together,
01:34:36.28 and if we do it right,
01:34:38.08 we can generate fluorescence where there was none before.
01:34:41.07 And there wouldn't be any if we didn't bring those two fragments
01:34:44.19 in proximity.
01:34:45.23 And this very clever application
01:34:48.01 called GRASP,
01:34:49.04 this is for looking at when a post- and pre-synaptic neuron get together
01:34:55.07 close enough to form a synapse
01:34:57.11 And the way you do it is that you express one half of the protein
01:35:01.08 the first fragment on the presynaptic cell
01:35:04.04 and the complementary fragment on the post-synaptic cell
01:35:07.25 and if they're not close enough to quite reach across the gap,
01:35:12.04 you get no fluorescence.
01:35:13.01 But when they finally do get to reach across the gap, then you generate fluorescence.
01:35:17.05 And down here is a micrograph of some examples
01:35:20.24 where you have pre- and post-synaptic cells.
01:35:23.12 Now these, just for visibility purposes separately marked with other colors
01:35:26.24 like blue, for the pre-, I think, and red for the post.
01:35:30.11 But where they come into contact,
01:35:32.10 that's where the GFP, the green got reconstituted from the two fragments
01:35:37.19 and that marks a synapse that got close enough to actually be functional
01:35:44.24 as a neuron, as a synapse as well as generate the fluorescence.
01:35:49.28 So what are genetically targetable sensors good for?
01:35:55.11 And in general, we can now look at some of the most important and popular signals inside cells
01:36:00.06 ranging from just pH to calcium, even zinc if you want to
01:36:05.26 I didn't talk about, cyclic nucleotides,
01:36:08.00 I didn't show redox potential
01:36:09.17 but there are all the ways in which redox potential
01:36:12.06 will modulate these types of signals.
01:36:15.15 and membrane potential, proteases, kinase, phosphatases, and so on.
01:36:19.28 Basically, just about any biochemical change that alters protein expression, localization, as in class 1,
01:36:27.28 conformation, proximity, specially class 4,
01:36:31.18 any of these can be sensed if you're willing to go
01:36:34.09 and engineer the molecule,
01:36:36.13 and once you've got it engineered,
01:36:38.00 it may take you some trial and error
01:36:39.21 sometimes a little computer modeling
01:36:43.24 has been known to help.
01:36:44.25 Once you make it, then you can transfer it over and over again
01:36:48.13 to new systems again and look at your favorite
01:36:50.11 signal that way.
01:36:51.28 And the fact that it's genetically encoded
01:36:53.10 greatly simplifies delivery into almost any organism
01:36:56.24 that you can transfect
01:36:58.23 or into a tissue;
01:37:00.15 you can force it to go into subcellular locations.
01:37:03.22 Suppose you want to see calcium in the mitochondria
01:37:06.18 not everywhere else.
01:37:08.02 Well, you just put a mitochondrial targeting signal on it,
01:37:10.07 or if you want to see protein kinase activity right at the plasma membrane
01:37:14.09 and not in the golgi
01:37:15.14 or if you want to see it in the golgi but not on the plasma membrane
01:37:17.29 or in the nucleus; all of those are now controllable
01:37:21.16 by this standard trafficking signals that we know about.
01:37:24.20 And it makes it mutagenic improvement; there's all sorts of tricks for doing mutagenesis
01:37:30.11 and screening huge libraries rapidly for your favorite new response.
01:37:37.20 And it even makes it easier to distribute around the world because
01:37:41.06 we don't have to dispense fragile organic chemicals;
01:37:44.22 we just have to dispense DNA,
01:37:46.26 which you can just do by spotting on a piece of filter paper and dropping it in the mail.
01:37:50.20 And it will survive mailing for some considerable distance.
01:37:54.24 What are the exceptions?
01:37:55.29 As I mentioned, these current fluorescent proteins are based on the jellyfish and corals
01:38:01.08 cannot be done on obligate anaerobes because they all need oxygen at some stage
01:38:06.05 in their life cycle
01:38:07.00 to become fluorescent
01:38:08.05 And not to mention the last one
01:38:11.07 is that intact human beings do have the problem that ethically,
01:38:14.24 are we allowed to transfect a sick person?
01:38:18.22 We, assuming that we have obeyed appropriate
01:38:22.03 animal welfare regulations, we are allowed to do that in mice.
01:38:26.07 We can make transgenic mice
01:38:27.06 transgenic flies, or transfect them acutely.
01:38:30.15 There's nothing says that human beings are biologically different;
01:38:33.21 you can certainly make these work in human cells in tissue culture.
01:38:36.24 But the question is whether you can do it in an intact human being
01:38:40.02 when they come to you as a sick person,
01:38:43.03 that's a whole other issue
01:38:44.10 which much more has to do with ethics and regulations
01:38:47.09 than with molecular biology.
01:38:51.20 I've thrown these four different sensing mechanisms at you -
01:38:55.09 translocation, single fluorescence intensity, dual FP, fluorescent protein FRET, and bimolecular fluorescence complementation.
01:39:04.03 And without going through all of this, each one of them has its own set of advantages and disadvantages;
01:39:09.07 otherwise, we could throw one of them out if it was decisively
01:39:13.10 bad in all sorts of way.
01:39:15.05 For example, the translocation one is relatively easy to engineer
01:39:19.03 if you've got a good binding domain already.
01:39:21.27 The fluorescent protein is along for the ride;
01:39:24.07 it doesn't have to change its properties or do anything subtle.
01:39:27.05 It just, you just have to be able to image it.
01:39:28.04 So it's relatively easy
01:39:30.21 to engineer.
01:39:31.22 On the other hand, it's not particularly easy to calibrate
01:39:35.10 which is another factor.
01:39:38.01 The single fluorescent protein intensity is nicely reversible, very fast,
01:39:43.04 takes quite a bit of engineering.
01:39:46.05 It's a little tricky to build all of this into a single fluorescent protein.
01:39:49.17 The FRET indicators are the most widely used because
01:39:56.02 so many things will change the relative proximity
01:40:01.09 or orientation of a pair of proteins.
01:40:03.13 They take a fair amount of engineering as well.
01:40:07.01 And both the fluorescent - single and double - are highly reversible.
01:40:11.20 The bimolecular fluorescence complementation, it's great advantage is that it has a tremendous signal over background
01:40:18.03 because if you don't get combination,
01:40:21.01 there is no signal.
01:40:21.25 There's just zero.
01:40:22.18 And then when you do manage to make even some fluorescent protein reconstitute,
01:40:29.00 you can get a big enhancement over zero.
01:40:32.28 And so it gives this huge signal-to-background contrast;
01:40:36.05 whereas the translocation, you saw that there was some stuff already hanging around
01:40:40.08 in the cytosol; it wasn't an all-or-nothing transition.
01:40:43.05 And discriminating the two perfectly by microscopy is not always easy and so on
01:40:47.20 In the case of the FRET, often, there is barely any change in the FRET,
01:40:52.20 and we only have a few percent change in the ratio of say yellow to cyan.
01:40:57.18 And it takes good instrumentation, so that has the least signal-to-background
01:41:00.25 But it's the most reversible.
01:41:02.09 The bimolecular fluorescence complementation, once that protein has folded out of the two fragments,
01:41:09.02 it doesn't need anymore scaffolding.
01:41:11.19 It will stay together until it's destroyed.
01:41:13.09 So it's not reversible basically.
01:41:15.15 And it also takes time of protein folding to respond
01:41:18.07 so that's why it's relatively low in reversibility
01:41:22.06 and speed of response.
01:41:23.24 So that's why we need all these different mechanisms because for different applications
01:41:29.11 each one may be advantageous or not.
01:41:32.04 So with that, I thank you very much
01:41:36.12 and hope that you've learned something from this lecture.

This Talk

Speaker: Roger Tsien

Audience:

Researcher

Recorded: May 2012

Talk Overview

In this talk titled “Fluorescent Protein Indicators,” Roger Tsien discusses how fluorescent proteins have been turned into indicators for a wide variety of biological properties, including pH, redox potential, as well as for signaling molecules such as phosphoinositides. The talk also describes how reporters are used to measure the activity of enzymes like kinases, phosphatases, and proteases. It covers both single proteins whose intensity or wavelength changes, as well as reporters using Forster resonance energy transfer (FRET).

Questions

What changes in a cell can fluorescent indicators measure? (select all that apply):
1. Signaling pathway activation
2. Change in pH
3. Protein concentration increase
4. Ca+2 signaling
5. Cell cycle procession
Briefly describe the four categories of fluorescent indicators.
FRET sensors can be used to measure which of the following (choose all that apply):
1. Changes in pH
2. Protein-protein interactions
3. Ca 2+ levels
4. Protease activity
What is the biggest limitation of fluorescent protein pH indicators?
What is the biggest advantage to using bimolecular fluorescence complementation?

Answers

View Answers

Speaker Bio

Roger Tsien

Dr. Tsien was a Professor at the University of California, San Diego, a Howard Hughes Medical Institute Investigator and a member of the National Academy of Sciences. In 2008, Tsien shared the Nobel Prize in Chemistry for the discovery and development of green fluorescent protein, GFP. His lab continues to develop new fluorescent proteins as… Continue Reading