Rapid warping of two-photon illumination wavefronts

A short paper in Optics Express looks interesting.  In A high speed wavefront determination method based on spatial frequency modulations for focusing light through random scattering media, Meng Cui presents a method for rapidly determining the optimal wavefront to ‘cancel out’ the scattering when 785nm light passes through turbid media.  In his example, a glass diffuser was used, but the clear goal for this work is to replace the glass with a brain.

To understand why this is so important for in vivo two-photon imaging, let’s review how 2-p imaging works. Light from a laser is focused to a point and swept across the field in a raster. The resulting fluorescence is of a different wavelength and can thus be filtered out from the excitation light. For each voxel, all the fluorescence that re-enters the objective is collected, regardless of its source.  The total amount of fluorescence collected for that timepoint in the sweep is assigned as the brightness of that voxel. Since the user knows where the laser was being aimed, scattering of fluorescence emission may reduce the brightness but will not blur the image.  However, scattering of the excitation light can dramatically reduce the excitation at the target voxel while increasing the off-target excitation of its neighbors. This causes a rapid increase in background fluorescence and blur at increasing brain depth.

The vasculature was labeled by injecting flourescein dextran into the circulatory stream. The light source was a regenerative amplifier. ‘‘0 mm’’ corresponds to the top of the brain. Left, XZ projection. Right, examples of XY projections. Note the increase in background fluo- rescence deeper than 600 mm in the brain due to out-of-focus 2PE. (Theer et al., 2003)

Previous reports work has shown that one can use adaptive optics to adjust the phase of the wavefront of the excitation light to correct for the scattering of the excitation.  However, determination of the optimal wavefront for a field of view took minutes, which could be problematic for imaging in an awake animal.  Any changes in the precise position of the brain might change the optimal wavefront.  Ideally, one would want a system that could optimize the wavefront every second, or even before every frame of acquisition (typically 4-8 Hz in a raster scan in vivo experiment)

Scattering in the brain warps two-photon excitation light, but adaptive optics can correct this.

I’ll let Meng Cui explain the technique in his own words

Elastic scattering is the dominant factor limiting the optical imaging depth in tissues. Take gray matter as an example, at 800 nm the scattering coefficient is 77 /cm and the absorption coefficient is 0.2 / cm. If there is a way to suppress scattering, the optical imaging depth could be greatly improved. Despite the apparent randomness, scattering is a deterministic process. A properly engineered wave can propagate inside scattering media and form a focus, a well understood phenomenon in the time reversal and optical phase conjugation (OPC) studies…

For applications on biological tissues, acquisition time on the order of one millisecond (ms) per degree of freedom is desired. Deformable mirrors can provide a high modulation speed. However the degrees of freedom are rather limited. A phase-only SLM can provide about one million degrees of freedom at a much lower modulation speed. In this work, I present a novel method, capable of providing as many degrees of freedom as a SLM with a data acquisition time of one ms per degree of freedom. The method was employed to focus light through a random scattering medium with a 400 ms total data acquisition time, ~three orders of magnitude faster than the previous report [25].

The essence of a COAT system is to phase modulate different input spatial modes while detecting the output signal from the target. To greatly improve the operation speed, the experiment requires a device that can provide fast phase modulation and can access a large number of spatial modes very quickly. To meet these two requirements, a pair of scanning Galvanometer mirrors was used to quickly visit different modes in the spatial frequency domain or k space, and a frequency shifted reference beam was provided for a heterodyne detection. The wavefront profile was first determined in k space and then transformed to the spatial domain. The spatial phase profile was displayed on a SLM to focus light onto the target. In such a design, the number of degrees of freedom is limited by the number of pixels on the SLM and the experiment speed is determined by the scanning mirror speed…

Compared to existing techniques, the reported method can provide both a high operation speed and a large number of degrees of freedom. In the current design, the operation speed is limited by the scanning mirror speed and the maximum number of degrees of freedom is limited by the SLM pixel number. In this demonstration, 400 spatial modes in k space were visited and the determined phase profile was displayed on the SLM. Depending on the scattering property of the media, more (up to 1920 x 1080) or less number of degrees of freedom can be used to optimize the focus quality and the operation speed.

Using a stepwise position scanning, the method achieves an operation speed of one ms (400 μs transition time + 600 μs recording time) per spatial mode, ~three orders of magnitude faster than the previous report. Using a continuous position scanning and a faster position scanner such as resonant scanning mirrors, polygon mirror scanners, or acousto-optic deflectors, the operation speed can be potentially increased by at least one order of magnitude. It is anticipated that the reported technique will find a broad range of applications in biomedical deep tissue imaging.

UPDATE : DIADEM Final Results

The DIADEM automated neuronal reconstruction contest has finished.  Accurate, fast, and high-resolution automated neuron reconstruction is of vital importance to cracking the mystery of how neural circuits perform. Even with perfect knowledge of the firing patterns of every cell in a circuit, our understanding of how these patterns are produced and how the information is processed would be quite limited.  True understanding requires knowledge of the precise wiring diagram.  This prize is a good first step towards bringing awareness of this tricky problem to the world’s best computer scientists.

$75,000 in prize money was to go to the group that was able to produce high-quality reconstructions of neuronal structures at least 20x faster than by-hand reconstructions.  In the finals, the fastest speed achieved was 10X the by-hand method. Some groups were hindered by slight variances in the source data formatting, which normally isn’t a big deal unless you only have 20 minutes to produce as much reconstruction as possible…

Since no group was able to beat the hard floor, but substantial progress was made, the money was distributed amongst these finalists.

Badrinath Roysam Team, $25,000
“for the better overall generality of their program in producing robust reconstructions by integration of human and machines interactions.”

Armen Stepanyants Team, $25,000
“for the better overall biological results in the spirit of pure automation.”

Eugene Myers Team, $15,000
“for the excellent quality and strength of their algorithm.”

German Gonzalez Team, $10,000
“for their deeper potential, more original approach, and ultimate scalability of their proposed solution.”

Deniz Erdogmus Team
“for elevating themselves above the current state of automated reconstructions…with a deep understanding of the technical and scientific problems.”

Congrats to the placing teams.

Cameleon-Nanos : High Affinity GECIs

Takeharu Nagai’s lab has published in Nature Methods, Spontaneous network activity visualized by ultrasensitive Ca²⁺ indicators, yellow Cameleon-Nano, demonstrating a new set of calcium indicators based on yellow cameleon. Back when he was still Take-san, Take’s ability to churn out and manually screen hundreds of cameleon variants was impressive and inspiring. With high-throughput GECI pipelines now ramping up at Janelia, the idea of laboriously screening 200 variations on a theme (be it cameleons or GluSnFRs), seems a bit archaic. However, this paper is a good example of the progress that can still be made by understanding the needed sensor parameters and fiddling with the primary amino acid structure in a relatively low-throughput way. Take-sensei’s results are another example of the pramatic rule in protein design, “when in doubt, tinker with the linker.”

The cameleon-nano family achieves greater apparent calcium affinity than YC2.60-4.60, reaching levels of up to 15nM.  They did this by increasing the flexibility of the linker by extending the standard Tsien/Miyawaki/Baird Gly-Gly-Ser linker with additional glycines.  In this case, the longer the linker between the CaM and M13 segments, the greater the apparent affinity. Interestingly, improvement by increasing linker flexibility is precisely the opposite the advice Atsushi and Take gave me for achieving high ratio changes with FRET reporters.  Back at RIKEN in 2002, they suggested I use short, stiff linkers to restrict the rotational freedom of the fluorescent pairs.  Then one could find orientations where relative rotation of dipole moment gave much greater FRET changes than would be expected from changes in FP distance alone. Take and Atsushi’s big YC2.60/3.60 paper strongly supported this idea!  However, as our understanding of the ideal parameters of calcium sensor’s for in vivo imaging has grown, development directions have adjusted.

Cameleon-Nanos achieve higher signal/noise for sparse action potentials at the expense of linearity.  Like Fluo-4, the signal saturates at relatively low AP frequencies.  I think the absolute affinities measured for this family (15, 30, 50 and 140nM) should be considered very rough estimates. They extrapolated these values from stopped-flow binding experiments, because

Although we would like to measure the koff of YC2.60 and its high affinity variants such as YC-Nano15, we could not do it because it was very difficult to precisely control free Ca2+ concentration at around few tens of nM as far as we used EGTA (Kd for Ca2+ = 151 nM in 0.1 M ionic strength, pH 7.2 at 25 oC). For this purpose, much stronger Ca2+ chelator with a smaller Kd value was required. However there is no such Ca2+ chelator available now.

I’m not sure why they didn’t just use the higher affinity, Mg++ insensitive, chelator BAPTA to make the Kd measurements the right way, with a linear regression of log-log fluorescence/concentration values.  Due to instrument dead time, and the high affinity, I didn’t like stopped-flow based Kd measurements in the early GCaMP papers, and I don’t like them now.  Also, the apparent calcium Kd will be highly dependent on solution ionic strength and [Mg++] which is unreported. Despite these quibbles, which are important only inasmuch as they give insight into the mechanism of improvement and the direction of future development, the cameleon-nano family looks promising for mammalian brain imaging.  I still wonder if, assuming the reported Kd values are relevant in vivo, YC2.60 would be the best of the bunch, since cortical neurons have a resting Kd of ~50nM, which implies that a single AP transient of say 200nM free [Ca++] increase would push the calcium levels right up into the sweet-spot of YC2.60’s sensitivity.

This is all the more interesting given the recent results in YC3.60 imaging from Maz Hasan’s group.  Previously, he had shown that transgenic YC animals were pretty bad for imaging.  However, AAV-mediated gene delivery of YC3.60 has significantly improved the responses of the YC family. I’m not sure if they are really up to GCaMP3 levels under identical in vivo conditions, but they might have better long-term protein stability (or that might depend on which viral serotype is used.) What about cameleon-nanos, what about YC2.60?

Cell Cycle Visualization in Development

Atsushi Miyawaki’s lab has developed a series of neat tools for visualizing cell cycle progress.

For zebrafish, the zFucci system consists of two fluorescent proteins, mKO2 and mAG, that are fused to Cdt1 and geminin genes.  Cell cycle- regulated proteolysis of these fusion proteins causes each cell to display orange fluorescence in G₁ phase nuclei and green fluorescence in both the nucleus and cytoplasm of S/G₂/M phase cells.

Video of cell cycle transitions in culture. Click for the video.

The last time I saw Atsushi give a talk, he showed an incredible time lapse video from the zebrafish cleavage stage that I haven’t been able to find online.  However, here is a video from later in development of the zebrafish that is still pretty remarkable.

Development of a zebrafish visualized by zFucci. Click to see the video.

This two component system has been adapted for watching the transition from neural stem cells to differentiated neurons in living mice. The Color-Timer system uses double transgenics with the fluorescent protein KOr fused to nestin and EGFP fused to doublecortin.  In this system, neural stem cells fluoresce orange, while newly differentiated neurons fluoresce green.

The cerebral cortex of an E14.5 double Tg mouse embryo of nestin/KOr was time-lapse imaged. Click for video

Sugiyama, M., Sakaue-Sawano, A., Iimura, T., Fukami, K., Kitaguchi, T., Kawakami, K., Okamoto, H., Higashijima, S., & Miyawaki, A. (2009). Illuminating cell-cycle progression in the developing zebrafish embryo Proceedings of the National Academy of Sciences, 106 (49), 20812-20817 DOI: 10.1073/pnas.0906464106

Kanki, H., Shimabukuro, M., Miyawaki, A., & Okano, H. (2010). “Color Timer” mice: visualization of neuronal differentiation with fluorescent proteins Molecular Brain, 3 (1) DOI: 10.1186/1756-6606-3-5

Three Cheers for GCaMP : Optogenetic Brain Reading

Three papers are out online in Nature Methods that show big improvements in calcium imaging with genetically encoded sensors.  They are are based on the fluorescence intensity indicator, GCaMP.   GCaMP, first developed by Junichi Nakai, consists of a GFP that has been circularly permuted so that the N and C termini are fused and new termini are made in the middle of the protein.  Fused to one terminus is calmodulin and the other is a peptide, M13, that calmodulin (CaM) binds to in the presence of calcium. The name is supposed to look like GFP with a CaM inserted into it, G-CaM-P.  Normally the GFP is dim, as there is a hole from the outside of its barrel into the chromophore.  Upon binding calcium, this hole is plugged and fluorescence increases.

The first paper, A genetically encoded reporter of synaptic activity in vivo, from Leon Lagnado’s group, targets GCaMP2 to the outer surface of synaptic vesicles. This localization allows the fluorescence signal to be confined to the presynaptic terminal, where calcium fluxes in response to action potentials are high.  This targeting improves the response magnitude of GCaMP2 and permits the optical recording of synaptic inputs into whatever region of the brain one looks at.  They demonstrate the technique in live zebrafish.

In the second paper, Optical interrogation of neural circuits in Caenorhabditis elegans, from Sharad Ramanathan’s group, GCaMP2 has been combined with Channelrhodopsin-2 to perform functional circuit mapping in the worm.   Since the worm’s structural wiring diagram has been essentially solved, functional data could say much about how “thick” the wires between each cell are.  Unfortunately, with GCaMP2, the responses are too slow and weak to distinguish direct from indirect connections.

Finally, we have published a paper, Imaging neural activity in worms, flies and mice with improved GCaMP calcium indicators, describing the improved GCaMP3.  This indicator has between 2-10x better signal to noise than GCaMP2, D3cpv and TN-XXL, depending on the system you are using.  It’s kinetics are faster and it is more photostable than FRET indicators, and the responses are huge.  When expressed in motor cortex of the mouse, neuronal activity is easily seen directly in the raw data.  Furthermore, the sensor can be expressed stably for months, making it a potential tool for observing how learning reshapes the patterns of activity in the cortex.

Imaging of mouse motor cortex (M1) expressing the genetically-encoded calcium indicator GCaMP3 through a cortical window. After 72 days of GCaMP3 expression, large fluorescence transients can be seen in many neurons that are highly correlated with mouse running.

GCaMP3 is not perfect. It cannot reliably detect single action potential in vivo in mammals, though I doubt that any existing GECI can. Work continues on future generations of GCaMP that may achieve 100% fidelity in optical reading of the bits in the brain. However, there is considerable evidence from a number of groups that have been beta-testing the sensor, including the Tank lab of “quake mouse” fame, that it is a significant leap forward and unlocks much of the fantastic and fantasized potential of genetically-encoded calcium indicators.

Comparison of fluorescence changes in response to trains of action potentials in acute cortical slices.

I will try to post a more complete writeup of GCaMP3 for Brain Windows soon, with an unbiased eye to its strengths and weaknesses.  We worked very hard to carefully characterize this sensor’s effects on cellular and circuit properties.  If you have any questions about GCaMP3, please post them to the comments.

For further info about strategies for GECI use and optimization, check out our previous paper, Reporting neural activity with genetically encoded calcium indicators in Brain Cell Biology.

The official press release from HHMI regarding GCaMP3 is available here.

Automated ROI analysis for calcium imaging

One of the most time consuming and frustrating tasks associated with fluorescence imaging in the brain is picking out your regions of interest.  Which pixels do you include in as part of the cell and which are part of the surrounding neuropil?  Often, the answer is not obvious, and even with painstaking selections you can make errors.  Eran Mukamel et. al, from Mark Schnitzer‘s lab just published this Neurotechnique Automated Analysis of Cellular Signals from Large-Scale Calcium Imaging Data that aims to simplify and improve the results of ROI selection. 

The authors used a multistage approach to identify and quantify the calcium-dependent fluorescence changes of imaged neurons. First, they used principal component analysis to identify the components of the image that were likely calcium signal related and which were noise.  The sparse nature of the calcium response (calcium transients are brief and spatially confined) helped the separation from the noise. They threw the noise away.  Then they used independent component analysis to pick out which components of the calcium signal changed in a manner independent from other pieces of the signal.  These likely represent individual cells. Using this output, they performed auto-segmentation of the image into numerous individual neurons or processes and measured the fluorescence change in those regions.  In simulations of data, it resulted in superior data fidelity over hand drawing ROIs.  They also validated it with real in vivo calcium imaging.

 

Automated Cell Sorting Identifies Neuronal and Glial Ca2+ Dynamics from Large-Scale Two-Photon Imaging Data

 

Whether its neuronal imaging, high-speed motion tracking or multielectrode recordings, tremendously large data sets are currently being generated in systems neuroscience. It is simply impossible for a single post-doc to crunch all of her data without major automated computational techniques.  In calcium imaging, the resources that have been poured into the development and release of powerful new tools requires an equal effort on the data analysis end to maximize the value of this technique.  The automated algorithms presented in this paper look very promising and we will definitely be checking them out in the near future.