latest Brainbow paper has some useful imaging advice

[George McNamara]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear listserv,

latest Brainbow paper has some useful imaging advice

http://www.nature.com/nmeth/journal/v10/n6/abs/nmeth.2450.html


        Improved tools for the Brainbow toolbox -pp540 - 547

Dawen Cai, Kimberly B Cohen, Tuanlian Luo, Jeff W Lichtman & Joshua R Sanes

doi:10.1038/nmeth.2450


I disagree with a couple of the authors imaging advice on page 544 (most
of the advice is very good):

Authors: "Fluorophores with overlap in the excitation or emission
spectra should be imaged sequentially rather than simultaneously to
minimize fluorescence cross-talk and thereby optimize color separation".

On a typical confocal microscope (ex. Leica SP5, SP8 or Zeiss LSM710,
780) there are several detectors (ex. 5 on SP5 or SP8) or detector array
(ex. LSM710 or 780). Therefore, fluorophores that excite well at a given
excitation wavelength should be imaged simultaneously. I also recommend
the latest detectors, ex., HyD for leica, GaAsP for Zeiss, photon
counting mode if available (and TCSPC lifetime if available).

In the "Image processing" paragraph on page 544, the authors suggest
(for confocal) "slower scanning or averaging of multiple scans".

Slower scanning is more likely to lead to photobleaching (though the
authors somewhat mitigate this by recommending low laser power). I
recommend resonant mode (confocal or multiphoton), for users who have a
resonant scanner.
I have previously suggested here on the listserv that acquiring X number
of confocal frames without averaging, and then calculating the median
for each pixel, is a much better approach than averaging. I encourage
the confocal vendors to implement this, or even better "result value
selection", based on the noise distribution of the instrument
(especially the detectors). Even better would be to acquire the raw
(R.S. mode) frames and clean up in the deconvolution algorithm.
       Speaking of detectors (and tto not leave out cameras): F. Huang,
... J. Bewersdorf have a nice article showing how well sCMOS can work in
single molecule localization microscopy ... doi:10.1038/nmeth.2488
current abstract link is
http://www.nature.com/nmeth/journal/vaop/ncurrent/abs/nmeth.2488.html

A lot of confocal microscope users are still unaware of simple (and
free) image processing methods to improve imaging data, such as PiMP,
http://jcs.biologists.org/content/125/9/2257.long  (I recommend for
63x/1.4 NA confocal, 30 frame acquisition [current plugin is for single
plane], filter 1.6, 16-bit output). Contact the authors for the ImageJ
plugin.



***

The Brainbow 3.x's "author's file" feature is at
http://www.nature.com/nmeth/journal/v10/n6/abs/nmeth.2487.html

Zebrafish fans and developmental biologists of all hues should find of
interest the same group's Zebrabow paper, PMID:

    23757414



***

The same issue of Nature Methods has several other microscope imaging
articles, including:

http://www.nature.com/nmeth/journal/v10/n6/abs/nmeth.2481.html
CLARITY for mapping the nervous system

http://www.nature.com/nmeth/journal/v10/n6/abs/nmeth.2477.html
Mapping brain circuitry with a light microscope


Imaging human connectomes at the macroscale
http://www.nature.com/nmeth/journal/v10/n6/abs/nmeth.2482.html




--



George McNamara, Ph.D.
Single Cells Analyst
L.J.N. Cooper Lab
University of Texas M.D. Anderson Cancer Center
Houston, TX 77054

[Martin Wessendorf-2]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hey, George--

On 6/23/2013 9:56 AM, George McNamara wrote:

> I disagree with a couple of the authors imaging advice on page 544 (most
> of the advice is very good):
>
> Authors: "Fluorophores with overlap in the excitation or emission
> spectra should be imaged sequentially rather than simultaneously to
> minimize fluorescence cross-talk and thereby optimize color separation".
>
> On a typical confocal microscope (ex. Leica SP5, SP8 or Zeiss LSM710,
> 780) there are several detectors (ex. 5 on SP5 or SP8) or detector array
> (ex. LSM710 or 780). Therefore, fluorophores that excite well at a given
> excitation wavelength should be imaged simultaneously. I also recommend
> the latest detectors, ex., HyD for leica, GaAsP for Zeiss, photon
> counting mode if available (and TCSPC lifetime if available).

I'm confused about why you think this is bad advice.  If you have
coexpression of labels, it'll be much easier to determine whether or not
a particular cell is single-labeled or multiple-labeled if you use a
laser that doesn't excite all the candidates (or conversely, if you use
a barrier filter that doesn't pass all the candidates).  This is
particularly true when you have strong expression of one and weak
expression of the other.  Even with an detector array, there's a limit
to what you can ferret out (--unless you have infinite photons, of course!)

However, my confusion probably means I'm missing something so explain away!

Thanks--take care--

Martin Wessendorf


--
Martin Wessendorf, Ph.D.                   office: [phone]
Assoc Prof, Dept Neuroscience                 lab: [phone]
University of Minnesota             Preferred FAX: [phone]
6-145 Jackson Hall, 321 Church St. SE    Dept Fax: [phone]
Minneapolis, MN  55455                    e-mail: [hidden email]

[George McNamara]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Martin,

Sequential acquisition of fluorophores that can be excited at the same
wavelength is silly. On a multi-detector microscope like most point
scanning confocal microscopes, such as Leica SP5/SP8 (preferably with
HyD's) or Zeiss LSM780, the user has 5 (SP#) or up to 34 (LSM710 or 780,
2 standard PMTs, 32 channel spectral detector ... on 780 with GaAsP
spectral detector one would probably just use that). To take a simple
scenario, with the following three FPs,

mTFP1 ("Teal")
CY11.5 (Cyan-Venus high FRET; see S.S. Vogel papers for a series of CY
fusions with different FRET efficiencies)
mBeRFP (446ex, 615em), published in Yang et al 2013 PLoS One
http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0064849

one could sequentially excite at 458 nm (Argon ion laser), or 440 or 436
nm (other light sources), with single channel acquisition - as I quoted
previously. However, all three fluorophores (or even more with Vogel's
C-Y series) are going to emit and/or photobleach. So, if you have enough
detectors, best to acquire simultaneously.

***

One should pay attention to details when imaging, whether point scanning
confocal or widefield. For example, Krylova et al 2013 PLoS One
http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0063286
imaged four FP-FP dimers (in different cells) with single excitation:
mCherry-nls-mCherry
mRaspberry-nls-mRaspberry
mKate2-nls-mKate2
mPlum-nls-mPlum     (note: mPlum's are dim, they would have been better
off with mPlum-mPlum-mPlum(etc), Steve Vogel has gone up to Venus 6 [V6]).
My kudos to them for approximately doubling the brightness of each
protein by using tandem dimers (though tdTomato might have been a better
choice).

Krylova et al used a widefield microscope, 575/15 ex filter (593
dichroic) with 655/40 [635-675nm]  and 628/40 [608-648 nm] exciters.
Their figure 4 shows the two emission filters overlap in the range of
635-648 nm (bandpass interference filters are not absolutely vertical
cut on/off, but close enough). They would have been better with
non-overlapping emission bands (and two cameras would have been nice for
simultaneous acquisition). An optimal emission filter would be about
$150, a lot less than the publication charges for PLoS One or authors
time to work on this project. Ms. Vinita Popat, a summer student working
in our lab, calculated for me that 600-630 nm for the short wavelenghth
emission filter, and 630LP would be the optimal pair (under the
simplifying assumption that each FP is equally bright ... as noted
above, mPlum could have improved performance with a higher order multimer).

***

I recently (re)read a very nice paper from Richard Neher et al, showing
advantages to using two or more EXCITATION wavelengths - sequentially -
with the same emission bands, to obtain additional information (2x more
images, "multiple excitations greatly facilitate the decomposition").

    Neher RA, Mitkovski M, Kirchhoff F, Neher E, Theis FJ, Zeug A 2009
    Blind source separation techniques for the decomposition of multiply
    labeled fluorescence images. </pubmed/19413985> Biophys J. 96:
    3791-800. doi: 10.1016/j.bpj.2008.10.068.

    PMID: 19413985    
    http://www.cell.com/biophysj/retrieve/pii/S0006349509000927

    Methods of blind source separation are used in many contexts to
    separate composite data sets according to their sources. Multiply
    labeled fluorescence microscopy images represent such sets, in which
    the sources are the individual labels. Their distributions are the
    quantities of interest and have to be extracted from the images.
    This is often challenging, since the recorded emission spectra of
    fluorescent dyes are environment- and instrument-specific. We have
    developed a nonnegative matrix factorization (NMF) algorithm to
    detect and separate spectrally distinct components of multiply
    labeled fluorescence images. It operates on spectrally resolved
    images and delivers both the emission spectra of the identified
    components and images of their abundance. We tested the proposed
    method using biological samples labeled with up to four spectrally
    overlapping fluorescent labels. In most cases, NMF accurately
    decomposed the images into contributions of individual dyes.
    However, the solutions are not unique when spectra overlap strongly
    or when images are diffuse in their structure. To arrive at
    satisfactory results in such cases, we extended NMF to incorporate
    preexisting qualitative knowledge about spectra and label
    distributions. We show how data acquired through excitations at two
    or three different wavelengths can be integrated and that multiple
    excitations greatly facilitate the decomposition. By allowing
    reliable decomposition in cases where the spectra of the individual
    labels are not known or are known only inaccurately, the proposed
    algorithms greatly extend the range of questions that can be
    addressed with quantitative microscopy.

ImageJ plugins at      
http://www.mh-hannover.de/cellneurophys/poissonNMF/NMF/

I hope vendors start including this capability. I do wish Prof. Neher
had chosen a less mathematically smelly word than decomposition - one
reason I prefer to call this spectral unmixing.

Enjoy,

George




On 6/23/2013 2:23 PM, Martin Wessendorf wrote:

> Hey, George--
>
> On 6/23/2013 9:56 AM, George McNamara wrote:
>
>> I disagree with a couple of the authors imaging advice on page 544 (most
>> of the advice is very good):
>>
>> Authors: "Fluorophores with overlap in the excitation or emission
>> spectra should be imaged sequentially rather than simultaneously to
>> minimize fluorescence cross-talk and thereby optimize color separation".
>>
>> On a typical confocal microscope (ex. Leica SP5, SP8 or Zeiss LSM710,
>> 780) there are several detectors (ex. 5 on SP5 or SP8) or detector array
>> (ex. LSM710 or 780). Therefore, fluorophores that excite well at a given
>> excitation wavelength should be imaged simultaneously. I also recommend
>> the latest detectors, ex., HyD for leica, GaAsP for Zeiss, photon
>> counting mode if available (and TCSPC lifetime if available).
>
> I'm confused about why you think this is bad advice.  If you have
> coexpression of labels, it'll be much easier to determine whether or
> not a particular cell is single-labeled or multiple-labeled if you use
> a laser that doesn't excite all the candidates (or conversely, if you
> use a barrier filter that doesn't pass all the candidates).  This is
> particularly true when you have strong expression of one and weak
> expression of the other.  Even with an detector array, there's a limit
> to what you can ferret out (--unless you have infinite photons, of
> course!)
>
> However, my confusion probably means I'm missing something so explain
> away!
>
> Thanks--take care--
>
> Martin Wessendorf
>
>

--



George McNamara, Ph.D.
Single Cells Analyst
L.J.N. Cooper Lab
University of Texas M.D. Anderson Cancer Center
Houston, TX 77054