McTips 2018 PDF (as of 201810002) go to: McTips 2018
The McTips 2018 includes some of my thoughts with respect to Fast Photon Counting (FPC) to make fluorescence confocal microscopy both faster and more quantitative than is now practiced by most biomedical researchers (i.e. twiddle the HV gain and offset values until someone proves their boss' hypothesis ... especially when they are using 'Santa Crap' antibodies and don't bother with controls).
George's Quick Tip on Optimizing Confocal Microscope Image Acquisition
20190603Mon (ok, not that quick ... please be patient)
Applies to: Olympus FV3000RS (analog PMTs), Leica SP8 (photon ounting HyDs), and other point scanning confocal microscopes (that is, whatever instrument you are using, whether at JHU or elsewhere).
Goals - sequential!!!:
1. acquire optimally, then explore your data (initially at the confocal PC, then at your office PC) visualization and quantitation.
2. compare acquired image data to your experiment requirements -- what hypothesis(es) are you testing? how close to resolution limit(s) do you need to be to acquire the data to critically test your hypothesis(es) - then figure out what settings to use in the future.
Quick Summary for 1.4 NA objective lens:
- Use pixel size 60 nm, Z-step size 180 nm.
- Use fast pixel dwell time, ex. 2 microseconds per pixel on FV3000RS ... small pixel dwell time minimizes photobleaching and phototoxicity.
- Use "a few" line averaging, try 3 line avg on FV3000RS, so total pixel dwell time (t-PDT) = 6 microseconds (2 us/pixel * 3 line average). More averaging if needed.
- FV3000RS: low laser power, HV 500 mV, offset = 0 (can subtract later), always keep "Gain" at 1.0 ... Leica SP8 is simple: use HyD photon counting mode.
- Unidrectional scanning -- because (i) less prone to jaggedness artefacts, and (ii) every pixel gets identical "dark time", since flyback is dark. That is, in bidirectional scanning, the end pixels get re-illuminated immediately, whereas middle has delay. For more on jaggedness, and a fix, see Papiez et al 2019.
- I generally operate FV3000RS in galvo mode, 2 microsecond pixel dwell time ("PDT") ... resonant scanner mode is faster (typically average more to get usable "total Pixel Dweel Time") ... in fact, RS mode likely to benefit the most from my proposal below.
Proposal (written here 20190603Mon): Acquire onto MULTIPLE DETECTORS, by setting the detectors to ADJACENT wavelength bands, equal area under the emission spectrum curve (on FV3000RS assumes dichroic beamsplitters enable optimal split).
FV3000RS: 4 internal GaAsP detectors (but also dichroic beamplitters that may not be optimal for your fluorophore ... consider switching to fluorophores that are ... and/or help us buy optimal). Example:
Standard scan: 2 microsecond pixel dwell time. one detector, 500-560 nm, 8 line average.
Proposed scan 2 microsecond pixel dwell time. FOUR detectors, 500-515, 515-530, 530-545, 545-560 nm, 2 line average. Note: FV3000RS (currently., 6/2019) does not have optimized dichroics to take advantage of this hypothetical setting. It would be interesting for Olympus to (i) install 50/50 beamsplitters, and (ii) add to FluoView software the image math to combine (add, 12-bit 0 ... 4095 each --> 14-bit 0 ... 16380) the signal of all four detectors, (iii) optimize C.I. deconvolution to take aqdvantage of extended dynamic range. Bonus: "discard outlier" option when operating PMT's in high HV mode (13-bit if always throw out most extreme value ... Max Krummel has punlished similar idea). I also note ISS.com may have electronics (and software) to enable "fast photon counting" with current FV3000RS PMTs, and also pulsed laser to enable "fast FLIM". going from current analog (0 ... 4095) to photon counts (photon counts!) would be a "game changer" for our FV3000RS in terms of simplfying quantitation (the deconvolution module would also need to be optimized for photon counts).
Leica SP8: 2 HyD photon countng hybrid detectors. Can split the emission spectrum (area under the curve) 50:50, add together ... enables image acquisition in half the time (or get 2x more photons for 'full time). Reminder: scan as fast as possible for whatever zoom you ae using. 600 Hz enables full field of view (0.75x zoom by Leica convention). Example: 600 Hz with 10 Line Accumulation is superior to 60 Hz no accumulation, because less photobleaching and less phototoxicity for 600 Hz. SP8's fastest speed is 1800 Hz (requires 7.5x zoom or greater). Note: SP8 has spacefor two more internal HyD detectors AND Leica introduced (Spring 2018) third generation "SMD HyDs" (cooled, so less thermal noise, resulting in fewer dark counts, and faster counting by about 2x than our 2nd gen HyDs, so effectively bigger dynamic range per pixel dwell time), see below for all four seeing one fluorophore together ... would be great if you could provide us the $$$ to get a full set of four SMD HyD's (trade in our current two) + chiller + a (yes, even more $$ for) pulsed laser to enable the FALCON (Leica Fluorescence lifetime contrast ... generically "Fast FLIM") + PC and Leica LAS X software upgrade.
- if we can get our Leica SP8 upgraded to four SMD HyD's (per above) we would be able to go 8x faster than now (single 2nd gen HyD), by the ~2x faster photon counting rate of SMD HyD * using all four SMD HyDs to observe a single fluorophore.
- SMD HyD's count at 80 MHz, that is 1 photon is detected in each 12.5 nanosecond reptitition.
Vinegoni ... Weissleder 2016 Nature Communications https://www.nature.com/articles/ncomms11077
whose high dynamic range by multiple detectors featured 50/50 beamsplitters and neutral density filters (ND filters kill photons: how dumb!), did have supplemental figure on 'asymmetric' beamsplitters of 90%/10%, so PMT1 = 90%, PMT2 = 10%, PMT3 = 1% (ignoring tiny losses on each optical surface). One unknown to me is how damaging to PMT1 is zapping it with a lot of light in bright area ... ex: at the limit, entire field of view might be bright, so saturate PMT1 the entire time (our Leica SP8 HyD's have a safety cutoff, so prolonged saturatin would simply cutoff the experiment until the user stops scanning).
Pinkard ... Krummel 2016 PloS One https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0150430
Nice approach ... rank filter 50% = median filter, so rank filter 75% could be used to "pick" second highest value of the FV3000RS's internal PMTs ... I suggest "discard outlier" (usually highest intensity) and ACCUMULATE the other three PMTs values (i.e. 3500, 1000, 1000, 1000 --> discard 3500, accumulate 1000+1000+1000 = 3000, data is 16-bit anyway), would work nicely. If doing averaging, PC electronics now fast enough (and RAM cheap enough) that Olympus could keep each value in RAM, wait until that pixel is done acquiring, then rank filter (Max K) or 'discard outlier(s)' and accumulate. Optionally always rescale as if all values were good, i.e. for four data points, discard one, accumulate, multiply by 4/3 ... for 12 data points, if discard two, accumulate ten, multiply by 12/10.
Papiez et al 2019 IEEETBE https://www.ncbi.nlm.nih.gov/pubmed/31034401 and https://ieeexplore.ieee.org/document/8701671
- you have optimally labeled fluorescence specimens, including negative controls (labeling details is outside the scope of this quick tip).
- You are interested in seeing what you have on your specimen, and have time to explore the resolution limits of the specimen and optics.
- You have 'good' understanding of your microscope and/or can work with an expert (ex. me), and have the time and interest (and money) in getting your confocal imaging right.
- Operating at high NA (yes, also assumes you know what NA is!).
- operation at confocal pinhole 1 A.U. (yes, assumes you know what A.U. is). I mention briefly 0.5 A.U. below -- the penalty of 0.5 A.U. is pinhole is 1/4 area, so ~1/4 the number of photons gets through (most confocals can set pinhole to ~0.31, implying 10% area and number of photons compared to 1.0 A.U.).
- your primary interest is in a green fluorophore, emission 520 nm, ex. Alexa Fluor 488, EGFP, mNeonGreen, "mXX" (Nathan Shaner's 5/2019 alias for his new 6x brighter than EGFP green fluorescent protein ... also has new bright YFP). I note that shorter wavelength enables proportionally better resolution, ex. BV421 (em center wavelength 430 nm) enables 1.2 fold better resolution than emission 520 nm (520/430 = 1.2, that is, 520 nm is 1.2 fold worse, so 430 nm is 1.2 fold better).
- Using fluorescence ... I note that reflection from appropriate nanoparticle(s) could have advantages (I would like to see FL-Nanogold tested on our confocals! See www.nanoprobes.com and check for distributors).
- Access to spatial deconvolution algorithm, with GPU accelaration (Olympus: Cellsens "C.I." decovnvolution; Leica: HyVolution = Leica HyD detectors -> Huygens software). With excellent specimens and optimal acquisition settings, this increases resolution by ~10% (if pixel size AND signal to noise ratio are each optimal) and improves contrast and dynamic range. GPU enables (near) instant gratification.
- Ignoring for here AiryScan, STED, etc, optical and engineering 'tricks', just exploring standard point scanning confocal microscope.
- ==>Assumes the microscope is performing well, vibration isolation table is floating (and operating correctly), no air drafts blowing on the specimen or stage.
Theoretical limit of widefield microscope: dxy = 0.61 * Lambda / NA ... dz ~ 3 * dxy. So: dxy = 0.61 * 520 nm / 1.4 NA = 214 nm.
Theoretical limit Confocal 1.0 A.U. microscope: dxy = 0.51 * Lambda / NA ... dz ~ 3 * dxy. So: dxy = 0.51 * 520 nm / 1.4 NA = 189 nm.
Theoretical limit Confocal 0.5 A.U. microscope: dxy = (0.51 * Lambda / NA)/1.2 ... dz ~ 3 * dxy. So: dxy = (0.51 * 520 nm / 1.4 NA)/1.2 = 158 nm.
George's Quick Rules of Confocal Settings:
- Refractive index match your specimen mounting medium to the microscope objective lens immersion oil
- Olympuws FV3000RS: 1.405 R.I. silicone oil (see Boothe 2018 ELife for use of OptiPrep if using live cells).
- Leica SP8: 1.518 R.I. standard immersion oil (we use leica oil).
- Tip: I encourage all specimens to be imaged in 35 mm (or larger) imaging dishes, non-fluorescent glass ... ex. Mattek.com (~$2/dish), WPI FluoroDishes, CellVis (~$/dish, though the latter may have 'autofluorescent stuff' on it, obliterating the $1/dish cost savings).
- use 1.4 NA objective lens (or close as possible on the specific microscope ... if you would like to donate a 1.45 NA lens for our confocals, that would improve resolution by 3% (1.45/1.40 = 1.03) -- which may be quantifyable if you perfectly refractive index match. Note: High NA objective lenses are designed for imaging at the coverglass, and have limited working distance (usually less than 100 um). Can extend working distance by ~50 um by using #0 thickness coverglass (nominally ~120 um) instead of standard #1.5 coverglass (nominally 170 um thick), assuming you refractive index match (i.e. 1.518 R.I.) and/or deconvolution software can correct for thin coverglass (don't count on it).
- George's take on Nyquist Sampling Theorem (which was developed for sine waves, and states 2.2 data points needed for one wave) ... confocal images are acquired with 2D pixels, or 3D voxels, and specimens are not usually simple sine waves oriented along one axis, SO: Pixel size = oversample by 3 fold from the dxy above. Optionally, oversample even more (maybe 3.5 fold, unlikely to benefit from 4.0 fold, but if you find >3 fold useful, please share with me).
- For 1.0 Airy unit, fluorescence emission 520 nm (and photostable fluorophore, optimally matched mounting medium):
- for dxy = 189 nm, George recommends 60 nm XY pixel size ... optional: SVI.nl Huygens suggests 43nm (see below).
- for dz = 3 * dxy, use 180 nm Z step size (ok to use 200 nm)... optional: SVI.nl Huygens suggests 131 nm.
- I note that 60/43 = 1.395 and 1.395^2 = 1.947, so my recommended setting results in ~2-fold more photons and 2-fold less time, tthan SVI's recommendation ... assuming identical pixel dwell time (ex. 2 microsecond per pixel * 3 line averaging on FV3000RS = 6 microseconds 'total pixel dwell time').
- SVI.nl Huygens - online calculator https://svi.nl/NyquistCalculator:
https://www.fpbase.org - Fluorescent Proteins Database - including Spectral Viewer and FRET Ro Calculator
Spectral Viewer https://www.fpbase.org/spectra/
FRET Calculator https://www.fpbase.org/fret/
FRET equation from above:
QY = Quantum Yield, EC = Extinction Coefficient, J(λ) = Overlap Integral, R0 = Förster Radius, �� = refractive index, κ2 = orientation factor
Wu & Brand (1994). Resonance Energy Transfer: Methods and Applications. Analytical Biochem. 218
Note: GM also has a FRET calculator Excel file inside PubSpectra ZIP file.
PubSpectra web page https://works.bepress.com/gmcnamara/9
PubSpectra download link https://works.bepress.com/gmcnamara/9/download
20190528U: stay tuned late 2019 for Nathan Shaner's improved GFP and YFP. The GFP is 2x brighter than his mNeonGreen, so 6x brighter than EGFP. Time has come to retire EGFP!!!
ThermoFisher Prolong Glass (without DAPI) is now the best choice, if imaging fixed specimens with oil immersion objective lens.
==> Or- from Marker Gene Technologies (see MGT web site for more options for each product): *
* Opti-Bryt (fixed cells) https://www.markergene.com/opti-bryt-trade-perm-antifade-mount.html
* Opti-Klear (live cells) https://www.markergene.com/opti-klear-live-cell-imaging-buffer-5x.html
Prolong Glass info states needs to cure for 30+ hours.
* grow cells in imaging dishes (mattek or WPI-Inc ... or ibidi imaging quality coverglass chambers)
** at no time should cells be allowed to "air dry" = keep submerged.
* fix (i.e. formaldehyde), permeabilize if needed.
* immunofluorescence (i.e. http://www.nano-tag.com 2ndary nanobodies with each mouse mAb) ... can include DAPI and/or other counterstains here (example: fluorescent phalloidin).
* wash extensively (but quickly).
* "drip on" some Prolong Glass with imaging dish tilted, so that it forces aqueous media away ... pipet out the "run off", drop more (but not too much $) Prolong Glass ... goal is ~100% Prolong Glass, ~0% aqueous.
* allow to "cure" 30+ hours, in the dark, at room temperature, no lid, large volume of air (i.e. not small sealed box) to let volatiles escape. ... Probably simplest to go closer to 48 hours (and would be nice to be consistent in experiments).
20180803Fri ... connecting to our file server from Windows (win 7).
* ask George for the name of our file server - and please do not give out the name or IP address of our server.
* you are welcome to set up your own 'share drive to transfer your files (and can we please have 42 Terabytes of space on yours?).
Some Windows PC's are able to see our file server. Some are not. Today we -- "we" being 99% Jim Potter and 1% GM -- were able to trouble shoot the network acess issue.
(i) windows PC (win7 or ideally win10)
(ii) plugged into the JHU SOM network (Ethernet cable).
(iii) you have administrator privileges on the PC (does not need to be 'Administrator' login name).
1. Use JHARS to connect to JHU network (if you have not already done so) https://jhars.nts.jhu.edu/
1a. after setting up (or confirming) JHARS, probably useful to power off the PC, wait a few seconds, then power up and log in.
2. enable all the items in the Local Area Connection Properties dialog box (it is probably ok to enable more, but at minimum you need IPv6 and IPv4 and probably more).
3. using CMD prompt -> IPconfig / all (2nd screen shot below) ... see that DHCP Server 10p15.76.226
Windows Start menu ... cmd ... ipconfig / all
==> DHCP Server 10.15.176.226
==> Subnet mask 255.255.255.0 (if this is not correct, you may not be able to see JHU network at all!).
January 15, 2019 (20190115U) new book and eBook:
Basic Confocal Microscopy second edition
W. Gray (Jay) Jerome, Robert L. Price 2018
Chapter 1 includes:
Our Ten Commandments of confocal imaging are as follows.
1. The Perfect Microscope and the Perfect Microscopist Do Not Exist
2. Confocal Microscopy Is More Than a Confocal Microscope
3. During Specimen Processing the Integrity of the Specimen Must Be Maintained as Much as Possible
4. Photons Are Your Friends and Signal-to-Noise Ratio (SNR) Is King (GM note: and Queen and President and Premier ... and Dean, milliDean, microDean, nanoDean)
5. Quantification of Fluorescence in a Confocal Micrograph Is a Challenge and at Best Is Only Semiquantitative
6. Scientific Digital Imaging and Normal Digital Imaging (Family Photography) Are Not the Same
7. Your Image Is Your Data: Garbage in Will Result in Garbage Out
8. The Resolution and Bit Depth Present in a Digital Image Are a One-Way Street
9. The JPEG (Joint Photographic Experts Group) Image File Format Is EVIL but Useful
10. Storage Media Is Essentially Free and Infinite
* JHU staff can download eBook PDF for 'free' (that is, JHU has a subscription to publisher's content ... which comes out of NIH and other grants indirect overhead).
* Springer ofters MyCopy softcover edition $24.99 see "Buy" button at top right of https://link.springer.com/book/10.1007%2F978-3-319-97454-5
#1. reminds me of my high school's National Honor Society slogan (which I'm paraphrasing here): "Those of you who think you're perfect, amuse those of us who are". (our NHS T-shirt had the original slogan and Mr. Wampole's face ... Mr. Wampolewas the advanced math teacher in addition to neing NHS chapter advisor).
#5. fluorescence intensity is 'at best only semiquantitative' ... this is sad & true
* (I blame it on 30+ years of researchers not requiring 'good intensity quantitation' of manufacturers AND manufacturers not making quantitation easy and reasonably priced).
* I suggest ACCM's Leica SP8 confocal microscope HyD detectors in photon counting mode enables users to come close to quantitation. There are still issues of laser performance (most lasers fluctuate in power), Z-drift and XY-drift (not a big deal for sngle focus plane, single field measurements), and specimen refractive index induced issues (if any mismatch in R.I., then Z affects intensity -- see Staudt and Hell "TDE" paper and Olympus silicone oil graph).
* Fast FLIM and - simpler and less expensive to get going and less data deluge - "fast photon counting" (FPC) can be implemented on any PMT or Hybrid or APD based point scanning confocal microscope. Re: Becker&Hickl fast FLIM or ISS FastFLIM" (and either would be less expensive to add to our FV3000RS confocal microscope than buying a new fully loaded Leica SP8 Falcon Fast FLIM ... bonus: Wolfgang Becker correctly disses Leica's featuring 'fast lifetime contrast' (FALCON) over fast TCSPC data).
Grey and Price 2018 Table 1.1:
Pawley J (2000) The 39 steps: A cautionary tale of quantitative 3-D fluorescence microscopy. BioTechniques 28:884–888
Leica Microsystems - THUNDER Imager Tour at JHU SOM 3/3019
Leica - LIGHTNING and THUNDER
Leica THUNDER Tour - See Through the Haze with Computational Clearing
Johns Hopkins University, School of Medicine, Ross Fluorescence Imaging Center
* Summary: alternative to 'optical clearing' (which requires fixation and chemical clearing) for your fluorescent specimens.
* Demonstrations March 12-15, 2019. Two instruments:
1. "Model organisms": Leica M205, 5x 0.5NA objective lens ... large field of view.
2. "Live cells": Leica DMi8 inverted microscope, environmental controls (37 C, 5%CO2) available, full set of objective lenses.
* Seminar: THUNDER Imagers – Decode 3D Biology in Real-time.
Computational Clearing – available exclusively on Leica Microsystems THUNDER Imagers – offers groundbreaking ease of use, throughput, speed and sensitivity for 3D tissue, live cell and model organism imaging. Unparalleled image quality from stereo, upright and inverted live cell microscopes, with no special sample preparation needed.
THUNDER Imager - how it works (pdf) ... two imaging systems platforms.
LIGHTNING info (web link below - pdf download at bottom of that page) ... 'adaptive deconvolution' (GPU enabled).
20190712Fri - JHU Data Repository (Repositories)
==> JHU now (2019) uses Microsoft OneDrive for single user backup and sharing, ex. a graduate student or postdoc can share specific OneDrive folders with JHU colleagues and P.I. (not sharable outside JHU) Each JHU employee or student gets 5 Terabytes (5 Tb) OneDrive storage for free. Additional space should be arrangable through I.T.
For published data:
Craedl - Collaborative Research Administration Environment & Data Library
Craedl is hosted by MARCC Maryland Advanced Research Computing Center https://www.marcc.jhu.edu
JHU Sheridan Libraries
* Data Services https://archive.data.jhu.edu/
* more info from https://guides.library.jhu.edu/dataservices/publish-and-share/how-to
Find a Repository
Research data can be archived into the JHU Data Archive. Research data is made discoverable and publicly-accessible, assigned a unique, persistent identifier (such as DOI) for accurate citation by others, and managed to ensure that they are usable in the future. JHU Data Archive is available to all JHU researchers, including graduate students. Archiving services for projects under 1 TB are FREE. For archiving larger datasets, please contact us to discuss fees that may apply. More information about data sharing and archiving is available on our Archiving page.
University Archives and Records Management
The Ferdinand Hamburger University Archives preserves the history of the Johns Hopkins University in print and digital form, including the privately-held records of JHU faculty.
Selecting a Repository for Data Deposit
Tips and set of questions researchers can use in determining whether a particular research data repository will work for their circumstances.
DSpace preserves digital materials generated related to Johns Hopkins research. It is also a place for Electronic Theses and Dissertations (ETD) for Johns Hopkins students (If you need help with the thesis/dissertation submission, you can find the instruction here, including an 1-hour recorded ETD workshop video).
Inter-university Consortium for Political and Social Research (ICPSR)
One of the largest archives of datasets in the world, the Inter-university Consortium for Political and Social Research, (ICPSR) has a vast collection of social, political, and behavioral science datasets. These datasets can be used in a multitude of fields such as sociology, political science, history, business, public health, economics, and education. You might consider archiving with ICPSR if your work falls into one of these categories or other related subjects and/or if you are looking for an archive that will support sensitive, restricted use data.
Registry of Research Data Repositories (re3data.org)
A global registry of data repositories that covers research data repositories from different academic disciplines. Use re3data.org to search for data repositories in a specified discipline and download and/or deposit data there.