Fluorescent protein labeling: a cautionary tale

There was a time when we viewed bacterial cells as mere bags of randomly mixed molecules.  Lacking the obvious compartmentalization of eukaryotic cells, bacteria were viewed as being completely unstructured.  But increasing numbers of studies seem to show clearly defined localization patterns for proteins in bacteria.  One example is that the main proteases responsible for regulated proteolysis in bacteria — the Clp proteases (pronounced “clip”) — have been observed in several studies to form a single bright proteolytic focus, detected by fluorescent protein labeling.

The Paulsson lab spotted these observations and became intrigued.  One of the major interests in the lab is variation between individual cells at the RNA and protein level, and this looks like a potentially significant place where variation may happen.  If all proteolysis in a cell is localized into a single spot, then when a cell divides something interesting must happen: either the spot also divides, or one of the two daughter cells gets all of the Clp proteases in the cell while the other daughter gets nothing.  The second option would lead to a potentially enormous difference in the ability of the two daughters to perform proteolysis.  So a graduate student, Dirk Landgraf, set out to look at whether this difference exists, and if so how long it lasts (Landgraf et al. 2012, Segregation of molecules at cell division reveals native protein localization.  Nature methods doi: 10.1038/nmeth.1955).

The first step was to ask what happens to the proteolytic focus at cell division.  Landgraf et al. made movies of cells carrying fusions of a Clp family member, ClpP, with two different fluorescent proteins, Venus and superfolder GFP.  In each case they saw a single focus of fluorescence, and when the cell divided the whole fluorescent focus went to one daughter.  After a few generations, fluorescent foci (one per cell) reappeared in the line of cells descending from the other daughter.  This strongly suggested that there should be significant variation in the level of proteolysis going on in different cells. If regulated proteolysis is an important function for the cell — which we believe it is — this seems odd, and therefore interesting.  So the authors tested this possibility directly using another fluorescent tag (mCherry)  fused to a Clp substrate, allowing them to measure the variation in the degradation of the substrate in pairs of daughter cells from a single division event.

Clp foci (green) are present in only one of the two daughter cells after cell division.

This is where things get surprising, not to say shocking.  Yes, the lines in which ClpP was labeled with Venus or superfolder GFP showed very significant daughter-to-daughter variation.  But in the wild type strain, in which the ClpP was unmodified, very little daughter-to-daughter variation was seen.  The inescapable conclusion is that the fluorescent protein tags are changing the behavior of the protein being studied.  And this is not a small change: the whole notion that ClpP self-organizes into a single localized focus, which has led for example to the idea that protein degradation needs to be compartmentalized, appears to be an artifact.

Fluorescent proteins have swept the world of cell biology. What better way could there be to study the behavior of your favorite protein than to put a brightly glowing tag on it and watch it going about its normal business?  The images you get are beautiful and compelling, and make great figures in your paper. We’ve become so comfortable with the essential benignity of fluorescent protein fusions that we barely bother to worry about whether adding an extra 238 amino acids to a protein changes its behavior.  Partly this is because we can see so much with fluorescent protein fusions that we could never see before, so there is no easy way to be sure that the behavior of the protein under study hasn’t changed. But partly, too, it’s because the standard in the field has shifted.  Fluorescent proteins are the gold standard now.  If your results from an older and apparently cruder technique, such as immunofluorescence, don’t match the results from live-cell imaging using fluorescent proteins, then the immediate suspicion is that the older technique is wrong.  And probably this is often true.  What Landgraf et al. show, however, is that in the case of the Clp family the older methods are the better methods.  Immunofluorescent staining shows many small Clp foci, probably corresponding to individual protease complexes, located throughout the cell in the wild type, but also detects the large single clump induced when fluorescent tags are added.  

The wisdom of cellular crowds

Once again, an interesting Theory Lunch talk has inspired me to write a blog post.  Last Friday’s talk was from Mike White, who described (among other things) his lab’s efforts to understand the transcriptional behavior of the prolactin gene.  This gene is primarily expressed in the pituitary, and controls the production of milk in breastfeeding mothers.  On a cell-by-cell level, its expression is very variable in pituitary tissue; neighboring cells express the gene to very different extents.  And yet the random expression patterns in individual cells together add up to a coordinated response at the tissue level.  If we hope to build from an understanding of how cells behave to an understanding of how organisms behave, we need to know what underlies this kind of “wisdom of crowds” effect.  And so White and colleagues set out to determine why this gene shows variable expression (Harper et al. 2011.  Dynamic analysis of stochastic transcription cycles.  PLoS Biology doi:10.1371/journal.pbio.1000607) and how this expression might be coordinated on a population level.

Cell-to-cell variability in the levels of proteins and mRNAs has been much studied in bacteria, where at least two factors are likely to be important: first, key regulatory molecules may be present in the cell at very low numbers, leading to randomness in gene activation; and second, unequal partitioning of components at cell division may create additional variation that is pretty much indistinguishable from the fluctuations caused by sporadic gene activation.  There have been fewer studies in eukaryotes, so far, but people are already speculating about additional sources of differences: perhaps genes are moved in and out of “transcription factories” at different times in different cells, or perhaps the differences are caused by chromatin remodeling.

How tissues self-construct

The process of development is an astounding journey from simplicity to complexity.  You start with a single cell, the fertilized egg, and you end up with a complete multicellular organism, made up of tissues that self-organize from many individual cells of different types.  The question of how cells know who to be and where to go has many layers to it, starting with the question of how you lay down the basic body plan (head here, tail there, which side is left and where does the heart go?) and continuing on down to microscopic structures, with questions such as how and where to form the small tubes that will allow blood to permeate through apparently solid tissues.  This kind of self-organizing behavior is deeply interesting to robotics researchers (who would love to copy it) and tissue engineers (who would like to manipulate it).

A recent paper (Parsa et al. 2011. Uncovering the behaviors of individual cells within a multicellular microvascular community.  PNAS doi:10.1073/pnas.1007508108) takes a close look at self-organization on the micro level.  It turns out that if you take human endothelial cells and put them in a soft gel, they will spontaneously move around and form small tubes.  Parsa et al. tracked the behavior and morphology of individual cells from the moment they were seeded into the gel to the point when they have formed a connected network that will eventually turn into capillary-like structures.  This wasn’t an easy task: the cells move in three dimensions in the gel, the gel itself can shrink over time, and each cell is making many contacts with other cells; in many cases the cells are literally crawling over each other. To help track individual cells, they made a mixed population of cells that were labeled with 6 different combinations of fluorescent dyes, so that they had a good chance of being able to distinguish two neighboring cells using color.  And to shape the gel in a way that gives reasonable optical imaging they designed a PDMS mold with a removable cap that was used to flatten the gel’s top surface.

Miniaturizing medicine

You’ve probably seen NMR machines at some point during your career.  They usually have their own room, often with an extra-high ceiling to allow the operator to insert the sample without bumping his or her head.  So it may surprise you to know that a miniaturized NMR machine that you can literally hold in the palm of your hand has now been developed by Ralph Weissleder’s group.  And yes, there’s an app for that: the instrument is operated via a smartphone, making it possible to use NMR analysis of clinical samples literally at the bedside (Haun et al. 2011. Micro-NMR for rapid molecular analysis of human tumor samples. Sci Transl Med 3, 71ra16, doi:10.1126/scitranslmed.3002048).

The world's smallest NMR machine, so far

The Weissleder lab and their collaborators have been working on this miniaturized NMR machine and the imaging reagents required to use it for several years now; Ralph brought a prototype to our faculty lunch meeting about a year ago.  The goal is to use NMR as a sensitive way of detecting specific markers on very small samples of cells from patients who may or may not have a malignant tumor.  The device uses a miniaturized magnet to create the field inducing the magnetic resonance, solenoidal microcoils to detect the signal with high sensitivity, and tiny fluid channels, embedded in polydimethylsiloxane beside the microcoils, into which the sample is injected.  The whole device has a footprint of 10cm square.  The other part of the magic is in the imaging reagents.  These are magnetic nanoparticles that can be linked to a variety of monoclonal antibodies via some clever chemistry that allows the linking to be done in the presence of whole blood (for more details, come to Neal Devaraj’s Pizza Talk at 12.30 today).  Using this system, it’s possible to get a reliable measurement of the level of a specific marker from just ~200 cells.  And the measurement is amazingly quick: it takes under an hour.

Repost: Controls are Cool

Many of you know that as a post-doc in Uri Alon’s lab, Galit Lahav caused a small revolution in our understanding of how the p53 network responds to DNA damage.  By looking at single cells instead of populations, she showed that individual cells responding to the damage caused by gamma-irradiation show a series of stereotyped pulses (shown in this movie); different cells show different numbers of pulses, and as you increase the amount of damage, the number of pulses per cell increases.  Now the Lahav lab has identified another previously unsuspected feature of the p53 response (Loewer A, Batchelor E, Gaglia G, Lahav G.  2010.  Basal Dynamics of p53 Reveal Transcriptionally Attenuated Pulses in Cycling Cells Cell 142 89-100. PMID: 20598361). It turns out that p53 is being activated in normal growing cells all the time.  Because the cell cycle of cells in culture is unsynchronized, this activation can only be seen by looking at single cells. Since p53 may be the most studied protein on the planet, discovering something completely new and unexpected about its activities isn’t an everyday event.

The story started with an experiment that was originally intended as a control, looking at unstressed cells.  Unexpectedly, in these unstressed, undamaged cells they found p53 pulses that are indistinguishable in shape from the pulses seen in gamma-irradiated cells. The first clue to where these pulses come from was the observation that they’re correlated with specific stages of the cell cycle, primarily happening right after mitosis.  Loewer et al. used a Cdk inhibitor to show that when the cell cycle is stopped, the pulses go away.  And the pulses were also selectively stopped when the ATM/DNA-PK pathway, which monitors double-stranded DNA breaks, was inhibited. It appears that these pulses are triggered by transient DNA damage that is a routine part of the cell cycle.

The field will never be the same

A report of the use of the OMX microscope in high-speed live-cell imaging just came out in PNAS (Carlton et al. 2010 Fast live simultaneous multiwavelength four-dimensional optical microscopy.  Proc. Natl. Acad Sci. USA 107 16016-16022 PMID: 20705899).  In an accompanying commentary, Jason Swedlow offers the opinion that the field of live cell imaging will never be the same.

No need to FRET

I just found this very pretty story about a new kind of kinase sensor from Barbara Imperiali’s lab (Luković et al. 2009 Monitoring Protein Kinases in Cellular Media with Highly Selective Chimeric Reporters. Angew Chem Int Ed Engl. 48 6828-31).  I don’t have to tell you why kinases are important; phosphorylation is one of the most significant things you can do to a protein (short of cleavage) and it’s used to create new binding sites and change conformations, causing profound changes in protein activity, in an enormous variety of settings in the cell.  There are over 500 kinases that phosphorylate proteins in the human genome, and the fraction of eukaryotic proteins that are phosphorylated in eukaryotes is huge — at least 30%, and the number goes up as methods get better. Misregulation of protein kinases is common in cancer, and anti-cancer drugs like Gleevec and Iressa have their effect by inhibiting the activity of specific kinases.  Because kinases are so common, it can be hard to see the activity of an individual kinase; you often don’t know whether the activity you’re watching is the kinase you care about, or a different one with overlapping specificity.

FRET can be helpful as a kinase monitor, but it’s not a simple technology — a bit fretful (ha!) in application [“marked by worry and distress; troublesome”] — and while it’s useful in a live-cell setting, it’s not a general solution.  In particular FRET is not ideal for high-throughput screening.  One promising alternative is a chromophore called Sox, for sulphonamido-oxine, that only fluoresces if it’s interacting with magnesium ions.  But it can’t effectively chelate the magnesium on its own; instead, it’s dependent on the presence of a nearby phosphate group that provides two of the four ligands needed for tetrahedral coordination of the magnesium.  So when the peptide is not phosphorylated, there’s no magnesium chelation and no fluorescence.  When the peptide is phosphorylated, fluorescence goes up significantly; the phosphate essentially becomes part of the chromophore.  This is the kind of design where, when you look at it, you go “oh, that’s clever”.  It’s reminiscent of the FlAsH system, which is also very clever.  In that system, a tag added to your protein of interest, containing four cysteines, coordinates a metal (in this case arsenic), causing a non-fluorescent chromophore to become fluorescent.  But in the case of Sox you can see the activity of an enzyme, not just its presence.

The double-edged sword of Damocles

Apoptosis is everywhere: it formed the spaces between your fingers and toes as you developed in utero, it prevents the development of T cells that would attack the cells of your body, and it shapes the structure of your brain.  Too little apoptosis can cause cancer; over-zealous apoptosis causes much of the damage in a stroke or heart attack, and contributes to the failure of organ transplants.  Not surprisingly, apoptosis is extremely tightly controlled at many levels.  The key event in the decision to go ahead and die is the permeabilization of the mitochondrial outer membrane (a.k.a. MOMP).  Cells given an apoptotic signal may wait for quite a long time before MOMP happens, and can be rescued if the conditions around them change; but once MOMP begins, under normal circumstances, there’s no going back.  Proteins released from inside the mitochondrion unleash the activity of the so-called “executioner” caspases, setting off a vicious cycle that ends with most of the contents of the cell being chewed up.  In fact, MOMP itself is usually enough to kill a cell, even when the downstream caspases are inhibited.

Usually, but not quite always.  Recently, it’s been noticed that MOMP alone is not always a death sentence.  If the caspases are inhibited, a few cells survive.  Galit Lahav pointed me to an interesting paper that shows that the cells that survive do so because they have a few intact mitochondria left, which can divide and gradually repopulate the cell (Tait et al. 2010 Resistance to caspase-independent cell death requires persistence of intact mitochondria. Dev Cell. 18 802-13. PMID: 20493813)

Resources roundup

Having listened to Naama Geva-Zetorsky’s seminar yesterday, I felt bad that I hadn’t been advertising the wonderful resource she helped build in her time in the Alon lab.  So I’ve added it under the list of “databases and tools” links (Dynamic Proteomics). What you will get if you go there is a database of localization and dynamics on 1164 different genes (at the time of writing; this is, after all a dynamic database), tagged with YFP and studied in the H1299 non-small lung cell carcinoma line.  The YFP is inserted by exon tagging, and each labeled gene is therefore under its endogenous promoter.  You can look at images showing protein localization, with quantitation of nucleus/cytoplasm levels, and movies showing protein dynamics after exposure to the DNA-damaging drug camptothecin.  It’s a remarkable resource.

And perhaps it’s not a bad idea to say a few words about what else is under there.

BioNumbers is a project Ron Milo, Paul Jorgensen and Mike Springer started while sharing a bay in the Kirschner lab.  It’s a database that collects “useful” biological numbers — how much, where, how big, how fast — with references to the literature where the number was found.  Ron Milo recently published a sampling of the data, which I wrote about here.

DataRail is an open source MATLAB toolbox for managing, transforming, visualizing, and modeling data, in particular high-throughput data.  It was  developed in the Sorger and Lauffenburger labs, primarily by Julio Saez-Rodriguez and Arthur Goldsipe, with help from Jeremy Muhlich and Bjorn Millard. I wrote a little about what it has been used for here.

GoFigure is the Megason lab’s software platform for quantitating  4D in vivo microscopy based data in high-throughput at the level of the cell, which is being developed by Arnaud Gelas, Kishore Mosaliganti, and Lydie Souhait.  There’s a snippet more about it here.

little b is an open source language for building models that allows the re-use and modification of shared parts.  It also provides custom notations that make models easier to read and write. It was developed in the Gunawardena lab by Aneil Mallavarapu.

MitoCarta is an inventory of 1098 mouse genes encoding proteins with strong support of mitochondrial localization. The Mootha lab performed mass spectrometry of mitochondria isolated from fourteen tissues, assessed protein localization through large-scale GFP tagging/microscopy, and integrated these results with six other genome-scale datasets of mitochondrial localization.  You can search human and mouse datasets, and view images of 131 GFP-tagged proteins with mitochondrial localization.

Rule-based modeling is a rule-based language for modeling protein interaction networks.  It allows you to write general rules about how proteins interact, creating executable models of protein networks.  It’s based on the kappa language, orginally written by Jérôme Feret and Jean Krivine, working with Walter Fontana.

Do you know about tools that were developed to help understand biological systems at the cell/organelle/pathway level?  Send me an e-mail at becky[at]hms.harvard.edu and I’ll link it.  Thanks!

GFP gets flashier yet

A recent report in Chemistry & Biology (Subach et al 2010 Red fluorescent protein with reversibly photoswitchable absorbance for photochromic FRET.  Chem Biol. 17 745-55. PMID: 20659687) describes the discovery of the first red fluorescent protein that has switchable absorbance spectra. The switch is thought to happen because the chromophore undergoes a cis–trans isomerization in response to certain wavelengths of light, in this case blue or yellow light.  The switchable RFP described here changes its absorbance from an “on” state with an absorbance maximum of 567 nm to an “off” state with absorbance peaking around 440 nm.  In the “off” state, the emission intensity (at 585 nm) is also dramatically reduced, possibly because the chromophore is more flexible in this state.

One reason to be interested in such fluors is that they may add new power to in vivo FRET (Förster resonance energy transfer, a.k.a. fluorescent resonance energy transfer).  In a form of FRET called photochromic FRET (pcFRET), which had only previously been shown to be possible with photoswitchable dyes, you can arrange matters such that the donor fluor’s emissions overlap well with the acceptor fluor’s absorbance when the acceptor is in the “on” state, and overlap poorly with absorbance in the “off” state.  You can then measure the reduction in the emission from the donor when the acceptor is turned on, and see it go up again when the acceptor is turned off.  This gives you a new — possibly more accurate and sensitive — way to measure  proximity between the two fluors, instead of by using the output of the acceptor.  If your acceptor can switch reversibly and repeatedly, which this one can, then you can switch the acceptor on and off multiple times and get a more accurate measurement of fluorescence transfer by averaging the many readings.  Even if you don’t use them for FRET, you can follow biological dynamics, or label several subcellular regions one after another, or use them for super-resolution microscopy approaches such as those based on PALM (photoactivated localization microscopy).