Is it possible to detect fluorescence of single fluorophore, like eGFP or Alexa Fluor 488, by confocal microscope? If molecules are separated by distances larger then optical resolution limit, is it enough photons emitted by a single phluorophore to be detected by a photomultiplier tube commonly used for image acquisition on the confocal microscope? Or to see a signal we need more molecules in a volume corresponding to the resolution limit, around 300 nm in xy dimension and around 600 nm in z dimension. For example, I observe fluorescence of eGFP-tagged protein as fluorescent spots in the cell. Could one spot, one foci, represent one molecule of labeled protein or thre are always few molecules?
In my experience conventional photomultiplier used in confocal microscopes do not allow the detection of single fluorophores, because their efficiency is too low. You are likely to bleach the fluorophore before you reach the minimum amount of signal needed to detect it.
Newer detectors like avalanche photodiodes implemented in a confocal microscope can reach single molecule detection.
Depending on your application you could use widefield techniques with an EMCCD or sCMOS cameras which are best for these applications.
Pierre, thanks a lot for the answer. Could you give a rough estimation of the molecules number which we can observe as a one spot in the confocal microscope with conventional photomultiplier? Is it dozens or hundreds?
My bet would be in the dozen range, but that's just a guess and it would depend on conditions (fluorophore, pH/salt conditions -some fluorophores are sensitive to these-, actual photomultiplier sensitivity).
You could try to actually measure it: if you can have solutions of known concentration of fluorophore image them and compare it to imaging in the coverglass (which would give you a control). With a decent estimate of the confocal spot volume, together with concentration of fluorophores, you will have an estimate of the number of fluorophores detected in the confocal spot. I would estimate the concentration sensitivity threshold as the conditions where the average signal in the fluorophore solution is higher than the standard deviation of the signal in your control (using the exact same imaging conditions); this would mean that your signal to noise ratio is higher than 1, which would be to me a decent estimate for the sensitivity threshold.
I'd have to disagree. Fluorochromes can emit thousands of photons before entering a "dark state". As a simple example, fluorescence correlation spectroscopy would be impossible if it were not possible to detect single fluorochromes.
In fact, a method for super-resolution microscopy on a conventional microscope is to simply have your fluorescent proteins or molecules spaced farther apart than the resolution limit (SPDM). Then, each molecule generates a Gaussian intensity profile whose standard deviation corresponds to the number of photons detected before transition to a dark state.
These, and many other methods, rely on being able to detect the fluorescence from a single molecule. However, as fluorescent proteins emit hundreds of photons, on average, before going dark, and molecules can emit thousands of photons, you stand a strong chance of finding the vast majority of individual fluorochromes, however, statistically, you will not find them all.
As for higher sensitivity detectors, they simply increase the proportion of fluorochromes you will detect, not whether you can see all or none. People use EMCCDs for techniques like STORM microscopy, because, as stated above, your standard deviation, and therefore your resolution, is a function of the number of photons detected. Since objective resolution is capped at the refractive index of the lenses within them, the only other way to boost resolution is to boost fluorochrome brightness and/or detector sensitivity.
Benjamin,
I think Anna is using a commercial confocal microscope with standard PMTs (probably not the latest version), which is not the best way to go for FCS. Generally these microscope are not very sensitive. I doubt that people have observed single GFP with those and with the ones with have in my institute clearly not. Did it work for you?
With APDs, hybrid detectors, usually used for FCS that's another story, but it does not look relevant to that question...
In our facility we have an old second hand (i.e. abused) SP2, an equally old, well cared for FV500, and a two year old SP8. The photocathode in the PMTs degrades primarily with light intensity. As a result, the PMTs in the SP2 are fairly insensitive, while the PMTs in the FV500 are nearly as sensitive as the brand new PMT in the SP8.
While APDs and HyDs are more sensitive, PMTs are within an order of magnitude of the sensitivity: http://www.hamamatsu.com/us/en/community/optical_sensors/tutorials/guide_to_detector_selection/index.html
As such, even at 10% detection, that means you would see 100 photons from a fluorochrome that on average would put out 1000 before entering a dark state.
The one thing that is harder to do with an insensitive detector is to determine how many fluorochromes are within a single diffraction limited volume. However, methods such as number and brightness: users.df.uba.ar/mad/Workshop/Gratton%20N&B.pdf
allow a for statistical sampling across time to infer the subresolution assembly and disassembly of fluorescently labelled molecules/proteins.
Therefore, the sensitivity of a PMT is sufficient to detect the fluorescence from the majority of single fluorescent molecules, but you won't have much information about these molecules other than their presence vs. absence.
One other point, SPADs can actually be a bit crummy at detecting single molecules, depending on their rate of movement. This is because while SPADs are sensitive, they also have a low saturation threshold (i.e. the need to drop the bias voltage after each detection limits the rate of detection). Therefore, if you have a sufficiently bright light source, a sensitive PMT may prove superior for detecting fast moving stable fluorochromes, as they have a much higher saturation threshold. But, there is also the trade-off that PMTs have an inferior temporal resolution.
In summary, most detector types should be able to detect a fraction of the total number of single molecules, where the proportion of molecules detected is a function of the fluorescence half-life of the molecule and the sensitivity of the detector. However, depending on the type of information you require about the molecule, different detectors and different methodologies would be most appropriate for a given question.
http://www.hamamatsu.com/us/en/community/optical_sensors/tutorials/guide_to_detector_selection/index.html
users.df.uba.ar/mad/Workshop/Gratton%20N&B.pdf
Dear Benjamin, I see your points. Indeed for the LSM I can use only PMT tube as a detector but not CCD camera. I understand and agree with arguments about increase of resolution. But my question is exactly about quantity of molecules which can be seen as one foci in conventional microscopy. I am working with mitochondrial proteins. Some of them, labeled with eGFP, or by antibodies from Alexa family, appears as just a few foci located in mitochondria. Can this one foci be a one molecule of the protein?
Virtually, I could check it by PALM – one foci in conventional technique, will become few foci under higher resolution in PALM microscope if few molecules are involved and divided by dozens nm?
For this, I would use the number and brightness (N&B) method. The method is ideal when used with photon counting detectors such as hybrid detectors, SPADs, or FLIM-PMTs. But, there is a correction worked out for the analog nature of PMTs (see attached PDF, page 16).
Basically, you monitor a single optical section and take an XYT series. You can then compare the single pixel intensities across time, and deconvolve your way back to roughly integer multiples of fluorescent molecules per pixel, per frame. This allows you to not only try and elucidate the number of molecules per unit brightness, but also the rate of assembly and disassembly of the molecules.
Effectively, N&B is a mathematically simplified form of RICS, making it a much more tenable approach. Also, the math underlying RICS is published, so you can setup your own analysis algorithm in R or Matlab, or you can download the SimFCS software package from the Gratton laboratory, which has the algorithm already programmed in.
We have performed N&B ourselves using a hybrid detector and the SimFCS package, and it proved fairly data rich for our own experiments looking at rate of membrane bound receptor assembly in response to various stimuli.
- Dear Benjamin, thanks for the answer and the N&B presentation! I only wonder, if it is possible to get a signal from single fluorophore, why people are using signal amplification methods. For example, for fluorescence in situ hybridisation for detection of small RNAs localisation on intracellular level. These molecules are not numerous, so no overlap in resoluble distances, but direct labeling of oligonucleotides with irganic dyes doesn't help a lot.
This is because of photon noise (aka shot noise): https://en.wikipedia.org/?title=Shot_noise
Basically, photons are discrete objects, and therefore, the dimmer the signal the greater the variance in intensity.
This ties back to single molecule imaging, where we know that since a fluorochrome transitioning to a dark state is a random event with a fixed half-life, some fluorochromes will go dark before you could capture enough light from them to sense them above the noise of your detector (i.e. to know that the molecule was there with a given statistical confidence). Therefore the brighter the fluorochrome and/or the the more sensitive the detector, the greater percentage of total single molecules you can detect with confidence. Even with a cheap webcam and a crappy fluorochrome, you will still detect some single molecules (i.e. those that had a disproportionately long lifetime) but you would only see a very small fraction of the total number of molecules.
A way to think about this would be to think of an experiment where you want to measure the number of cars (photons) that go under an overpass (your detector) on an interstate per minute using pressure plates.
In this example, cars striking the plate will be true detections, while vibrations and other false positives will be your noise floor.
A low sensitivity detector would only be able to detect cars if hit just right. If there is heavy traffic flow (i.e. a bright, stable fluorochrome), then you will be able to still get enough cars hitting the plate to still get a good estimate of the rate of traffic flow.
However, in light traffic, in a given minute, there is a reasonable chance that, in one minute, no cars go by or the few that do fail to hit the plate hard enough, meaning you think there is no traffic at all. Conversely, several cars, by random chance, will go by in a minute and hid the plate dead center, and you will not only detect the presence of traffic, but will over estimate the rate, since by chance there were more cars than normal in that minute.
In this scenario, having both a pressure plate sensitive enough to detect even motorcycles (i.e. a sensitive detector) and heavy traffic flow (i.e. a bright and stable fluorochrome), would give you a near guarantee of not only detecting nearly all traffic, but also the rate of traffic flow.
There is one more variable to also consider, which is numerical aperture. Flourescence, for an intensive purposes in biological specimens, emits light as a perfect sphere. Therefore, the higher the numerical aperture of your objective, the larger the fraction of this emitted light you will detect. Going back to our interstate example, a low NA objective would be like monitoring a single lane on a 16 lane interstate, while a high NA objective would be like monitoring 8 lanes simultaneously on the same interstate.
So, and ideal setup would be a highly sensitive detector (like a SPAD or EMCCD camera), a super-stable fluorochrome, like quantum dots, and a high numerical aperture objective. However, less than ideal setups can still work, just with a lower efficiency.
For example, in STORM, detecting all the light emitted by every molecule is critical, since the more light detected, the higher the resolution (again due to shot noise). However, if all you need is a is my molecule there or not, and you don't care about exactly where and/or how many, than a lower sensitivity system will be more than adequate.
O yes Benjamin, I understood. Statistic! I especially appreciated the end of your messeage. I analyse intracellular localisation of some RNAs directly labeled by Alexa or ATTO derivatives. They expected to be in mitochondria. But I never sow the signal in these organells. My coallegues suggest that I will see it by STORM. But I am sure, and after your detailed answer, especially, that if you see in LSM nothing, you will not get more by super-resolution techniques.
For sure, you can detect all fluorescence signals individually as well as in combination of two or three or may be four, depending on the range of detection of your confocal microscope.
Well, on the other hand, I cannot recall anyone doing STORM with LSM... Usually it is used in widefield technique (with several possible types of illumination) and always with EMCCD or sCMOS camera. Alexa dyes can be seen individually but just not with traditional confocal microscope. So your colleagues suggestion is right, but not with the same setup.
The only chance you would have to see a single RNA with LSM is to use techniques where you can have multiple labels on a single RNA, in case this is compatible with your experiment.
One of my favorite lines from EM or super-resolution microscopy is that if it looks bad under LM, it will look really bad at higher resolution. STORM requires extremely bright samples to get enough detection events to recreate the entire structure before the fluorochromes fade out. If you are having a hard time detecting the signal with LSCM, you definitely will not see it with STORM.
STORM is also normally done on a TIRF platform, rather than widefield. This is because fluorescence from away from the focal plane adds an additional level of complexity to the analysis. Even on a TIRF platform, the STORM resolution is high enough to detect variations in Z, so when reconstructing an image, one of the steps is to filter out detection events exceeding a certain diameter, and retaining only the light from near the focal plane.
There are ways to get Z info back, such as astigmatism or the newer double helix phase mask, but the max Z resolution still tends to be in the tens of nanometers, meaning a TIRF platform is still ideal for STORM.
The other huge advantage of using TIRF for STORM is the very high NA used for TIRF means max detection efficiency for STORM.
This papers says about internationalist of synthesized porphyrins and Hoechst staining in candida cells
https://www.researchgate.net/publication/279459630_Axial_Ligand_Modified_High_Valent_Tin%28IV%29_Porphyrins_Synthesis_Structure_Photophysical_Studies_and_Photodynamic_Antimicrobial_Activities_on_Candida_Albicans
Article Axial Ligand Modified High Valent Tin(IV) Porphyrins: Synthe...
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Physical Phenomena
- Luminescence
- Fluorescence