Category Archives: FCS

Transcription caught on camera part 1: Halo transcription factors

Transcription factors (TFs) have a fundamental role is regulating gene expression. The basic model, based on numerous biochemical analyses, has determined where TFs bind (usually at specific sites at or near promoters), when they bind the DNA (at a resolution of minutes/hours) and what do they do there (induce/repress transcription. Duh!).  However, much is yet unknown. One aspect that is fairly unknown is the dynamics of how TFs search for their binding sites, bind them and later dissociate, particularly at the single molecule level. To explore this, the Transcription Imaging Consortium (TIC) at Janelia Research Center (JRC) (it used to be Janelia Farm, but the  “farm” part was removed from the name. oh well) applied sophisticated imaging techniques to measure the dynamics of two TFs, SOX2 and OCT4 in the nuclei of live embryonic stem (ES)cells. Their results were published in Cell almost a year ago.

Continue reading

Silicon nanocrystals – the next generation of fluorescent dyes?

Fluorescent microscopy is the only current method to follow biological structures and molecules in real-time in live specimens. Many advances were made but there are still a few problems with present fluorescent probes. Photobleaching (permanent disappearance of the fluorescent signal due to destruction of the fluorophore) is a major problem for all current fluorescent probes, be it chemical or biological. Photobleaching prevents prolong imaging.

A second problem is blinking of the fluorophore. Although blinking is actually useful in some super resolution techniques, it may still pose a problem for some applications.

Another problem is the size of some fluorescent tags such as fluorescent proteins and quantum dots. FP fusion to the studied protein can affect to protein function, simply by spatial interference.

Now, a new type of fluorophore was developed by the lab of Akihiro Kusumi from Kyoto university. Silicon nanocrystals (SiNCs) were apparently known for two decades as fluorescent, that are stable and can be used for biological applications. However, no serious effort was made to make SiNCs applicable in cell biology studies. This, he says, is due to the difficulty in producing SiNCs, manufacturing them in uniform size (and the size determines the fluorescent properties) and properties and conjugating them to biomolecules.

Kusumi developed a simplified protocol to produce mercaptosilane coated SiNC (mSiNC) that has a hydrodynamic diameter of 4.1 nm.  This was confirmed by three independent methods: gel filtration, FCS and dynamic light scattering (DLS). The hydrodynamic diameter of GFP is ~5.5nm and of Qdot655 – 19.5nm. They further measured by electron microscopy the actual size of the mSiNC core to be 3.2nm.

A & B - Elution traces of gel filtration experiments. The longer the molecule stays in the gel, the smaller it is. C- DLS analysis of mSiNC. D - FCS curves of the indicated molecules. E - image of mSiNCs obtained by electron microscopy (EM). F - distribution of measured diameter based on EM imaging.

A & B – Elution traces of gel filtration experiments. The longer the molecule stays in the gel, the smaller it is. C- DLS analysis of mSiNC. D – FCS curves of the indicated molecules. E – image of mSiNCs obtained by electron microscopy (EM). F – distribution of measured diameter based on EM imaging.

Characterization of the fluorescent properties showed that  the mSiNC fluorescent emission peak is at 655nm with a broad spectrum. Storage of dehydrated mSiNC did not affect the properties for over two months. However, fluorescence intensity decrease if stored in aqueous buffer at 37C (exponential decay time of 42 hours). Exposure of mSiNCs to excitation under fluorescent microscopy conditions showed that they are not blinking of photobleacing for over 300 minutes! Under similar conditions, GFP and the Cy3 dye photobleach within 3 seconds. the Qdot exhibited frequent blinking and fluctuations in fluorescent intensity compared to the mSiNC. mSiNCs fluorescence intensity is lower than Qdots. However, due to the instability of Qdot fluorescence, this difference diminishes over time.

After characterization came the actual test: they conjugated mSiNC to the transferrin protein (mSiNC-Tf) and applied it to cells, to follow the dynamics of cell surface transferrin receptors (TfR). whereas GFP-Tf allowed tracking for only 20 frames (0.67 sec), and Cy3-Tf allowed tracking of 150 frames (5sec), mSiNC-Tf allowed tracking for 3600 frames (120 seconds). This is quite an improvement.

A - a typical image of immobilized single mSiNC molecules. B - Distribution of fluorescence intensities of of single mSiNC spots (top) and single Qdot655 (bottom). C - Typical time-dependent changes of the fluorescence intensities of an mSiNC spot (top) and of a Qdot655 spot (bottom).

A – a typical image of immobilized single mSiNC molecules. B – Distribution of fluorescence intensities of of single mSiNC spots (top) and single Qdot655 (bottom). C – Typical time-dependent changes of the fluorescence intensities of an mSiNC spot (top) and of a Qdot655 spot (bottom).

Usually, due to blinking, single-molecule trajectories have gaps in them. These trajectories are later reconstructed based on the assumption that these gaps are due to blinking. With mSiNC-Tf, no gaps were observed (even of a single frame) for the whole 120 second (3600 frames) tracking. Thus, using mSiNC-Tf, they could follow the dynamics of TfR for long periods of time.

So, the advantages of mSiNC are very clear – they are extremely photostable, which allows for prolong imaging without loss of fluorescence. They are also small, allowing single molecule imaging with minimal perturbation. They can be conjugated to proteins.

Here are links to videos showing the lack of blinking; mSiNC-Tf on cell membrane (exhibiting Brownian motion) and what seems to be receptor internalization.

What are the disadvantages?

Well, first, the protocol to produce them, and verify their uniformity, though simplified, is yet not simple.

Second, the very wide emission spectra limits the use for multicolor imaging.

Last, similar to other dyes, it is excellent for live imaging of outer-membrane molecules, but may be more difficult for intra-cellular tagging for live imaging.

ResearchBlogging.orgNishimura H, Ritchie K, Kasai RS, Goto M, Morone N, Sugimura H, Tanaka K, Sase I, Yoshimura A, Nakano Y, Fujiwara TK, & Kusumi A (2013). Biocompatible fluorescent silicon nanocrystals for single-molecule tracking and fluorescence imaging. The Journal of cell biology, 202 (6), 967-83 PMID: 24043702

Imaging gene expression – methods & protocols

A new book in the “Methods in molecular biology” series, recently published, contains 23 imaging protocols in three major research areas: gene expression & RNA dynamics, genome & chromatin dynamics, and nuclear process & structures.

book cover: Imaging Gene Expression

This is a fairly good overview of the field and can help both beginners and researchers looking for new ideas.

The book can be freely downloaded.

Here’s the contents of the book:

Continue reading

Put that mRNA where you want it

How do you determine the localization of a single mRNA molecule in a living cell? Better yet – can you bring that mRNA where you want it?

I have briefly mentioned the MS2 system a while back. Essentially, The MS2 system to determine mRNA localization is composed of two parts: an RNA stem loop (referred to as MS2-coat protein (MCP) binding site, MBS) and the MCP – the coat protein from bacteriophage MS2.

MCP specifically binds the MBS (as a dimer).

A similar system with bacteriophage PP7 coat protein (PCP) and PBS was developed in our lab.

How does the MS2 system help to visualize mRNAs? Continue reading

Parameters that affect FCS

Last week I did an experiment with a fluorescent dye in solution, just to get a feel of the different parameters that affect the reading. Here I summarize the main points that one should be aware of when performing FCS experiments:

Sampling time

Sampling time is the time interval (∆t) in which you collect photons. The minimum of our photon detector is 25ns. However, short time intervals are not recommended due to a phenomenon called “after pulse”. I do not know the physics behind it. For a biologist like me, it is sufficient to say that there is a Gaussian-distributed probability that after each “real” photon that is detected, a “fake” photon can also be detected. The shorter the sampling time, the higher the probability of after pulse affecting your counts. In my experiment, I changed sampling time from 1µs up to 100µs. At 1-5µs I saw an afterpulse effect.

For longer sampling time I saw a different phenomenon – undersampling. Undersampling means that the measured particles move in and out of the FCS volume in a shorter time than the sampling time. For instance, the dye I used passes through the volume in 40µs (this is the parameter of diffusion time Td), so any sampling time over 40µs will be undersampling. But why does it affect our result?

The FCS methods measures fluctuations in fluorescence of the particles. That is – different fluorescent intensity when particles entering & leaving the FCS volume. If the particles enter and exit the volume in a time frame that is shorter than the sampling time, then we will just get a sum of the total photon emission of the particles, and no fluctuation. Therefore, if you know the time it takes your particle to cross the FCS volume, you should use a shorter sampling time (preferably at least half of that).  In my experiment, the dye, in water solution, passes the FCS volume within 40µs. Therefore, any sampling time above 40µs will be undersampling.

Undersampling can be calculated and corrected by using a specific algorithm.

Some examples of diffusion times:

  • Alexa488 dye in water solution: 40µs
  • GFP in water solution: 200µs
  • Free GFP in living cells (cytoplasm): 600µs
  • Free GFP in living cells (nucleus): 2ms
  • mRNP particle in living cells: 80ms

Laser power

This is a tricky issue, since laser power depends on the wavelength of the laser and the effect depends also on the fluorophore. However, some important issues:

The laser can melt your mirrors if the power is too high. For instance, in our instrument, going over 50mW can seriously damage the mirrors.

In my experiment I tried a range for power, from 1.5mW to 30mW. In general, the more power you use, the higher the intensity of the emission (which makes sense). This is optimally described by an equation of y=aX2. If your experimental setting is good, your results should give a plot that fits as close to power of 2 as possible. Here’s my result:

At low power you get a problem of signal to noise ratio, i.e. the difference from the background reading. For that reason, I discarded the 1.5mW result from the graph above. To overcome this problem, one should take longer measurements in order to improve the statistical analysis (for instance, I collected samples at 20µs for 60sec, and got rubbish – see graph below. Had I collected samples for 10 min I would improve the statistics and might get a better result).

As we go up in power we get another effect – bleaching. Bleaching means destruction of the fluorophore so that it does not emit photons upon excitation. Bleaching depend on laser power – the higher the power, you get more bleaching. Bleaching also depend on time of exposure to the laser light, and the stability of the fluorophore (for instance GFP is much more stable than RFP).  You can see at the intensity vs. power graph above that the 30mW point does not fit well to the trend-line. This is due to bleaching. If we remove the result of 30mW, the equation “improves” to y = 0.0125x2.0192.

“Up” distance (Z) from the glass-water interface

The distance Z is the distance above the glass-water interface of your sample. For best results, it is recommended that you set Z to the focal point of your laser (see image).

The distance not only affects the intensity (the further you go from that point, the less brightness, and hence less intensity, you get), but also the volume of the measurement. For shorter distances, the volume may be only partially in water, and partially the glass. For longer distances, the laser is not focused and the volume increases – hence there is an increase in the number of particles. However, Td calculation (from the autocorrelation curve) is only slightly affected by changing the Z.

When looking at fluorophore in solution, one should note that the dye may adhere to the glass, so a short distance may cause profound artifacts.  When doing live cells, other parameters come into play, such as internal organs that can obscure fluorescence, membranes that limit your volume, or causes “bouncing back” of the particle into the volume as soon as it exits. Therefore, X,Y & Z should be empirically determined for best result.

The focal point of the laser can be changed by altering the distances of the lenses of the beam expander. This too should be empirically determined for each wavelength used.

Fluorophore concentration

Fluorophore concentration also affects the quality of the FCS. To little particles and you will have a lousy signal-noise ratio. Too high a concentration and you get a problem of “dead time”.  Dead time is the time needed after a photoncounting detector fires for the detector to recover so that it is ready to register a new photon. This recovery time may be due to the detector, the processing electronics, or some combination of the two. In our lab it is approx. 10ns.

Consequently, if you have a high concentration of fluorophore, you get a high intensity – i.e. lots of photons. Due to the dead time, some of these photons may not register, leading to under-estimation of the intensity, the number of particles and further calculations.

There are other parameters that can affect the FCS measurement, such as temperature, quality of the optics, the properties of the fluorophore itself (e.g. size, brightness, stability),  the solution in which it resides (water, cell cytoplasm, cell nucleus) and more.

FCS data analysis

I started with the data analysis of my experiment. The data is far from being “good”, we can still extract some information out of it.

First, let’s look at the Intensity data.  The intensity, I, is measured in 50µs time intervals for a total time of 5 min (300sec). We can therefore plot I(t) vs as a function of time. However, we have 6 million data points (do the math – how many 50µs intervals so you have in 5 minutes). Therefore, plotting it would be a very dense graph. We therefore average the points for each second (20,000 data points) to get 300 points. This can be plotted with ease:

Plot 1 (cell number 6): Intensity (y axis) vs time (sec) (x axis)

This is a fairly good data set, with no large “jumps”. Compare to this one:

Plot 2 (cell number 1): Intensity (y axis) vs time (sec) (x axis)

In this second plot, we have time points with large fluctuations (the increase at time 30sec, the decrease in time 200-270sec). This is probably due to a movement of the cell or internal organs, relative to the FCS volume that we measure. These data points can be cropped out (unless you have too many of them).

Next, we look at the Brightness. The Brightness is a quality of the fluorofore, and should not change much over the course of the measurement. Here is the Brightness plot associated with plot 1:

Plot 3 (cell number 6): Brighness (y axis) vs data segment (x axis)

This is considered a “good”, consistent data set. Note that the data is not show per second, but per segment – the data was segmented to 183 pieces of 215 data points (215 x 50µs x 183 = 300 seconds).

This segmentation is, I assume, helpful for the autocorrelation calculations later. I still need to figure it out.

In plot 4, you can see a peak. This indicates some drastic change. Dependeing on your experiment – it can indicate a significant event, or an artifact (as in my case).

Plot 4 (cell number 1): Brighness (y axis) vs data segment (x axis)

The “bad” segments can be taken out of further calculations, which is what we did.

Next, and most popular use of the FCS, is the autocorrelation curve.

The following curve was fitted to a theoretical two population curve.

Plot 5 (cell number 1): Autocorrelation curve. Circles – data segments. Green curve – Theoretical fitted curve.

Note that the point at the earlier time points, and the last time points, usually do not fit well.

g01 and g02 are the autocorrelation calculations for each population. td1 and td2 represent the mobility of the particles measured. A lower td value indicate faster movement.

Below the autocorrelation curve you can see the “residual” curve – this is the plot of the differnce (“residual”)between the actual data and the theoretical data. If the actual data fits well, the residual plot should be randomly-distributed small flactuations around zero.

I still need to go deeper into the physics and math to get a better understanding of the system and the analysis.

This week, or next week, I’m going to do a tutorial, just with fluorofor in solution. I will learn how to operate the FCS microscope and up to the complete data analysis.

Fluorescence Correlation Spectroscopy – some technical notes

Today I had the pleasure of doing my first FCS experiment. Actually, I sat and watched as the FCS expert in the lab analyzed my cells.  The purpose of my experiment was to try to determine if a certain protein (and its mutant forms)  is associated with RNA in the cytoplasm or in the nucleus of yeast cells.

First point- yeast cells are small. You will need a lot of patience.

The size (x,y) of the laser beam is ~0.3-0.4 micron. that is only 1/10 of the size of yeast cell, about half or third of the size of its nucleus. Furthermore, yeast cells have vacuole that take a considerable volume of the cell (1/4-1/2). It is therefore very difficult to set the laser at the required point in the cell (be it cytoplasm or, even more difficult – nucleus). It is even more difficult to keep it stable on that spot for the duration of data collection (2-5 minutes). Also, since these are live cells, the organelles inside the cell can move around, disrupting your measurement.

Even small movements (0.1 micron) can drastically affect the reading, especially for small cells like yeast. Therefore, the cells are immobilized on the surface of the coverslide with a lectin (Concanavalin A). The proteins on the cell surface bind to the conA coated glass, thus the cells (hopefully) do not move. However, we cannot control the movement of organelles. If we measure the protein in the nucleus, and it moves, it can affect our measurements. We can usually detect sure movements by a drastic change in the intensity readings. If it happens momentarily, one can crop out the “bad” data points. However, If it happens too much during the measurement, you’re screwed – you have to start over, or pick a different cell.
All in all, we measured a total of 24 cells (3 strains, 8 cells each), and it took us only six hours.

Second point – Bleaching

I was surprised that there was no apparent bleaching of the GFP. Bleaching is dimming of the emission light, usually due to destruction of the fluorofor. However, in this experiment we used 2-photon excitation (in short, if your fluorofor is excited at 450nm, you can shine light at a 900nm wavelength. There is a very narrow, specific, area where the photon concentration is high enough that two photons hit the fluorofor almost simultaneously (about 1 femtosecond time frame). Two photons at 900nm have the same energy as one photon at 450nm (more or less). The longer wavelength reduced photodestruction of molecules. We also used low-power laser, which contributed to the stability of the protein.

Third point – Temperature.

The FCS microscope is capable of heating the objective lens, as well as the specimen stage to a fixed and stable temperature.  Since yeast normally grow at 30C, we pre-warmed the objective to 30C. the yeast sample was also heated. Actually, we did not use a simple glass slide. I used deltaT dishes, which are just small dishes with thin glass for bottom. After you immobilize the yeast, you add a small volume (1.5ml) of yeast media to keep the cells viable. However, this should also be heated. The temperature can affect not only the optics of the system, but also the behavior of the molecules at the microsecond range. Therefore, whenever we changed dishes (for the next strain), we had to wait ~20min for the temperature to stabilize.

At first, I made a mistake and used a different type of dish which cannot fit in the heating stage adaptor, and the readings were really unstable.

Fourth point – measurements.

Even if your fluorescence is low, FCS can detect it and gain very useful data. The data that is gathered is, for instance,  on the Intensity of the field and Brightness of the protein. The Intensity tells you how many photons are emitted in your field for a set time differential (e.g. 50 microsecond). The Brightness is a measure of the photon emission of a single particle that moves through your field. for instance, if you have Intensity of 100, if could be due to 100 particles with Brightness of 1, or 50 particles with brightness of 2.

If you know that your protein is fused to only a single GFP, but you get twice the Brightness of a single GFP, it could mean that your protein travels as a dimer. That is very important conclusion that can be derived from these experiments.

A different kind of data is the autocorrelation which we briefly touched in the previous post. This data can tell us how fast the particles are traveling through the field of measurement. A free protein would move much faster than a protein bound to a large complex (say, mRNP). It is also possible to determine the percentage of bound/unbound proteins.

A third type of information that you can measure by FCS is concentration of your particle. Based on the Intensity, brightness and mobility, one can calculate the concentration of the particle in the field. This could be useful, for instance in comparing cytoplasmic vs. nuclear concentrations, particularly if you wish to compare mutants of your protein.

Last point –

It was exhausting but interesting. I’m am still far from being an expert, but its a start. We analyzed the data for only a few measurements, but this was just to get an general idea of the expected results. We still need to systematically analyze all the data. When I have that, I may write about it here. Show some real data.