UPDATED : UCSD Neuroscience Movies Back Online

Almost every year, the UCSD Neurosciences Graduate program makes a movie or performs some skits lampooning the faculty (and sometimes other students). These videos used to be hosted on my server in the Tsien Lab, but that machine came with me to DC. I’ve finally taken the time to re-encode them and upload them to Google Video. I also uncovered the DVD of the excellent 2003 movie “Les Lettres Perdues”. The video quality is not as good from this host as from a private server, but at least they will be universally accessible. Email me if you want a higher quality version.

UCSD Neuroscience Skits 2006

The Investigator - UCSD Neuroscience Movie 2005

Les Lettres Perdues - UCSD Neuroscience Movie 2003

Tsien Lab Baby

Journal Club : GFP Reconstitution Across Synaptic Partners (GRASP)

This week, I’m guest presenting a recent Neuron paper over at the Nature Network’s neuroscience journal club. The authors propose a clever new technique, Genetic Reconstitution Across Synaptic Partners, to track the locations of selected synaptic connections backed by an impressive set of in vivo proof-of-principal experiments in C. elegans. The gist of the strategy is to split a fluorescent marker into two non-functional components and then distribute each half on different sides of circuit’s connection. Only at synaptic connections would the two components be close enough to undergo trans-complementation and reconstitute a functional marker.

Check out the full presentation and join the discussion.

Optical imaging of neuronal glutamate release and spillover with GluSnFR

This post is difficult to craft. I’ve been struggling with whether to write an epic post describing the history of glutamate imaging, the major advances and players in the field and where I fit into it, or a simple post focused on my new paper. Since glutamate imaging is my field, I’ve got tons to say about it, but also there is probably no way to avoid significant personal bias in my account. So, I’ll go with the short form. For those interested in further reading, please check out these earlier reports, including our brief mention of neuronal glutamate measurements with GluSnFR prototypes, neuronal glutamate measurement with FLIPE and the optimization of FLIPE constructs from Wolf Frommer’s group, and the use of FLIPE’s in brain slice to look at broad patterns of glutamate release from the Huguenard group.

In this PNAS paper, Optical measurement of synaptic glutamate spillover and reuptake by linker optimized glutamate-sensitive fluorescent reporters, by Hires et al. from the group of Roger Tsien, the authors report on the optimization of GluSnFR, a genetically-encoded Glutamate Sensitive Fluorescent Reporter, and its application to the study of glutamate spillover. A cyan and yellow fluorescent protein bracket a glutamate periplasmic binding protein. Glutamate binding to the PBP causes a conformational change and a reduction in the amount of FRET between the fluorescent proteins. Glutamate concentration can be quantitatively determined by observing the ratio of the blue to yellow fluorescence. When fused to an extra-cellular membrane targeting motif and expressed in neurons, optical responses to synaptic glutamate release were detected.

FRET constructs are very fickle, with their response being very sensitive to how the sensor components are fused together. This paper clearly demonstrates this, as 176 linker combinations were screened for maximal ratio change and only one was far superior to all others. The optimized SuperGluSnFR showed a 44% ratio change between zero and saturating glutamate levels in Ringer’s solution, a 6.2-fold improvement over the original prototype. Importantly, the screen took place in a system, HEK cells with surface displayed GluSnFRs, that was physiologically similar to the neuronal system where the sensor was ultimately used. This ensured that the screen discovered useful improvements, rather than ones that worked great in the screening system, but did not express or respond well when expressed in neurons. Previous glutamate sensor optimization in bacteria lead to large responses in vitro that did not translate well when ported to surface-displayed plasmids. Note though that this optimized sensor, FLI⁸¹PE, has found use when bath applied to brain slice.

SuperGluSnFR was used to address questions about glutamate spillover. Under what conditions might glutamate spill beyond the synaptic cleft? How long does this spillover last, and what effects might it have? The paper makes the first direct, quantitative measurements of the timecourse of glutamate spillover. It shows that, at least in cell culture, spillover following burst stimulation can cause a significant glutamate transient along the entire dendritic surface, not just at the synaptic active zone. After a single action potential, spillover is insufficient to activate any extrasynaptic glutamate receptors. But, after a burst of stimulation, sub-micromolar glutamate levels persist long enough to activate extrasynaptic NMDA receptors. This could have a tremendous impact on dendritic computation, synaptic independence and heterosynaptic long term potentiation or depression.

There are two significant limitations to the conclusions of this paper. First is that the experiments were done in dissociated hippocampal culture at room temperature. Glutamate transporters are faster at physiological temperatures, and the geometry of the neuropil in vivo might reduce the impact of spillover. Secondly, there is no electrophysiology to directly support the NMDAR activation assertion. Hopefully, some other group will pick up this thread and do more rigorous testing. GluSnFR imaging in acute brain slice should be easy enough using in utero electroporation techniques, and many, many labs have the electrophysiology experience needed. I’m tempted to do the experiments myself, but there is simply no time!

If anyone would like to try SuperGluSnFR for their own work, send me an email and I’ll be happy to send out an aliquot!

Large-scale model of mammalian thalamocortical systems

I’ll be the first to admit that my limited focus area in neuroscience, the levels of molecules and cells, biases me to attend to the trees, while missing the patterns in the whole forest. Many of our best recording and imaging methods, single-unit electrophysiology, fluorescence imaging and multi-unit extracellular arrays give us access to only a very tiny piece of the brain at any given moment. Yet endogenous neural activity is dependent on powerful (and subtle) interactions between geographically distant regions of the brain. Whole-brain measurement techniques, such as functional magnetic resonance imaging, magnetoencephalography, and diffusion tensor imgaging, can measure these interactions, but tend to have poor temporal-spatial resolution. Therefore, in order to understand the gestalt of the brain, it is useful to examine models that integrate our knowledge of single cell morphology and activity patterns, local circuitry, and distant connectivity.

The complexity of any brain model is limited by the computational power available. With 10^11 neurons and 10^15 synapses in the brain, each with a wide degree of possible synaptic strength, the computational power required to precisely model the entire brain is currently unavailable (though the Blue Brain project is trying). But maybe we don’t need every neuron. Maybe we can cull this 10^11 number down to something more manageable, a million, that proportionally represent a large number of the neuronal classes of the cortex. In this PNAS paper, Large-scale model of mammalian thalamocortical systems by Eugene Izhikevich and (Nobel laureate) Gerald Edelman, they demonstrate that many endogenous-like patterns of brain activity spontaneous emerge from just this sort of reduced model.

The model reduces the neuron count, leaves out many deep brain structures, and simplifies spike generation, but leaves intact important long-range connections, cellular morphology, dendritic spike initiation, synaptic plasticity and learning rules and dopaminergic modulation. With a sufficient ‘priming of the pump’ by allowing random miniPSPs to percolate through the network, the wiring organizes itself such that spontaneous activity patterns (delta, beta and gamma oscillations, wave propagation, anticorrelated clustering) emerge that are reminescent of in vivo patterns of activity. They also show the system is chaotic; a single spike added or subtracted evolves a totally different activity pattern after half a second, evoking the specter of determinism and the illusion of free will.

These are still early days, as they have not integrated any sensory input into the system and leave out many important brain structures. Nor do they make any testable predictions with the model. Nevertheless, the model appears to be readily extensible, and as greater understanding of brain regions and greater computing power become available, the power of the model may dramatically increase. For now, this paper is a tantalizing reminder of why many of us were originally attracted to the study of neuroscience, the quest to understand how our brain’s activity creates consciousness, and a context in which to place the little trees we struggle to understand.

Props to Neurochannels for the link.

Blog Roundup

Here’s a quick overview of some posts that got my attention in the last month…

Neurodudes has a brief writeup of video-rate superresolution imaging from Stefan Hell’s group. I don’t have access to Science Express PDF’s through our institutional subscription (how much must they be charging for that?), making a full writeup impossible. But you can at least check out the abstract and supporting info here. Video-Rate Far-Field Optical Nanoscopy Dissects Synaptic Vesicle Movement (Westphal et al.). The optical resolution isn’t quite as good as PALM or STORM, but the speed of acquisition is fantastic, permitting its use on dynamic living processes.

Eric Thomson from Neurochannels has posted a detailed Journal Club style review at Nature Network of the paper Spatiotemporal Dynamics of Cortical Sensorimotor Integration in Behaving Mice (Ferezou et al.) from Carl Petersen’s group. Using voltage-sensitive dyes they show the timing and spreading of activity from the sensory to motor cortex, following whisker stimulation in awake mice.

Biosingularity reports on results from Susumu Tonegawa’s group published in Science Express as Transgenic Inhibition of Synaptic Transmission Reveals Role of CA3 Output in Hippocampal Learning (Nakashiba et al.). They use a novel method, doxycycline-inhibited circuit exocytosis-knockdown (DICE-K), to transiently and selectively shut down the tri-synaptic pathway of the hippocampus (ER->DG->CA3->CA1->ER), while leaving the monosynaptic pathway (ER->CA1->ER) intact. These mice can still learn incrementally, but one-trial contextual learning and pattern completion recall is wiped out.

Journal Club Content

I’m going to try adding an additional type of content. Notes from journal clubs I attend. These will be more of a quick data dump format than a strict write-up, and topics will be broader than the specific field of brain imaging. The text formatting will be normal text = presented material, italics = audience interjections, bold = presenter responses. Journal Club entries can be accessed specifically by the top navigation bar. Hopefully this can expand our audience and perhaps promote additional discussion.

The first Journal Club discussion is here :

Sparse representation of sounds in the unanesthetized auditory cortex.

Hromádka T, Deweese MR, Zador AM.

How do neuronal populations in the auditory cortex represent acoustic stimuli? Although sound-evoked neural responses in the anesthetized auditory cortex are mainly transient, recent experiments in the unanesthetized preparation have emphasized subpopulations with other response properties. To quantify the relative contributions of these different subpopulations in the awake preparation, we have estimated the representation of sounds across the neuronal population using a representative ensemble of stimuli. We used cell-attached recording with a glass electrode, a method for which single-unit isolation does not depend on neuronal activity, to quantify the fraction of neurons engaged by acoustic stimuli (tones, frequency modulated sweeps, white-noise bursts, and natural stimuli) in the primary auditory cortex of awake head-fixed rats. We find that the population response is sparse, with stimuli typically eliciting high firing rates (>20 spikes/second) in less than 5% of neurons at any instant. Some neurons had very low spontaneous firing rates (<0.01 spikes/second). At the other extreme, some neurons had driven rates in excess of 50 spikes/second. Interestingly, the overall population response was well described by a lognormal distribution, rather than the exponential distribution that is often reported. Our results represent, to our knowledge, the first quantitative evidence for sparse representations of sounds in the unanesthetized auditory cortex. Our results are compatible with a model in which most neurons are silent much of the time, and in which representations are composed of small dynamic subsets of highly active neurons.

3D and Multicolor Superresolution Imaging

Progress in superresolution imaging is still moving very quickly. Here are two more great papers in the field.

First, Huang et al. from Xiaowei Zhuang’s group published a Science paper that moves superresolution imaging into three dimensions. Previously, STORM and PALM techniques were most useful for thin sections where the z-axis depth is well-constrained. Breaking the diffraction limit in the z-dimension was thought to possibly require recording from multiple angles, standing wave TIRF or optical lattice microscopy. Instead, the authors simply inserted a weak cylindrical mirror in between the imaging lens and the objective. This distorted the shape of the point spread function in the x- and y-dimensions, dependent on the z-axis distance from the focal plane. By examining the shape of each photoactivated molecule’s ‘photon cloud’, they were able to unambiguously assign a z-axis depth. This was a simple and clever way to map a third dimension of information on top of the two they were recording.

Due to increasing point spread widths with greater depth, the localization accuracy decreases with distance from the focal plane. Therefore, they only examined structures within a 500nm window around the focal depth. Z-scanning the focal plane could increase the depth range, though this might waste signal by photobleaching out of focus fluorophores. However, this is less of a concern in the STORM vs. PALM approach as the cyanine dyes used for STORM can be cycled on many times, while the Eos-FP used in PALM permanently bleaches. Of course, if a dye molecule moves position between on-cycles, this will degrade the effective resolution of the STORM approach.

PALM proponents also have a new paper out. Shroff et al. from Eric Betzig’s group show an alternative method of dual-color superresolution imaging. They co-express genes labeled with photoactivatable tandem dimer EosFP and with reversibly photoswitchable Dronpa or PS-CFP. The EosFP-tagged molecules are first photoactivated (405nm illumination), localized (561nm) and bleached. This process photoactivates a signficant population of the Dronpa or PS-CFP molecules. After all EosFP has been bleached, the activated second label is switched back to the dark state (Dronpa), or photobleached (PS-CFP) (488nm). The remaining second label can then be specifically photoactivated, localized and bleached.

A major advantage of this dual-color PALM technique over Zhuang or Hell’s two-color photoswitching approach is that all the fluorescent reagents are genetically encoded rather than antibody labeled. This permits more precise localization of the label to the target of interest. It also allows greater label packing density and more mild fixation. A disadvantage is that genetic overexpression could cause mislocalization of the target or artificial aggregation due to residual dimerization tendencies of the fluorescent tags. However, unnatural aggregation can also be induced with antibody labeling. Perhaps adaptation of Don Arnold’s FP tagged intrabodies could address this concern.

Pulse shaping for 2-photon signal enhancement

Gains in signal to noise ratios of organic dyes and genetically encoded indicators often come in modest steps following screening of large numbers of compounds or clones. Improvements are usually specific to individual chromophores, leading to the pigeonholing of development efforts on a small handful of indicators that have already undergone systemic optimization (i.e. cameleons, G-CaMP and troponin-based GECIs). Indicator photobleaching imposes strict limits on the amount of information which can be extracted by optical indicators. Improvement of specific indicators and their constituents is a worthy and necessary goal, but more generalizable improvements can be made by changing the nature of the illumination source. A series of papers from a variety of groups has shown that careful manipulation of the structure of pulse laser illumination can produce dramatic improvements in signal/noise and photobleaching during non-linear (two-photon) imaging. This is generalizable to numerous optical indicators. Reduction of photobleaching and photoinduced tissue damage will be essential for continuous optical monitoring of sparse neural activity.

 My first encounter with these techniques came in 2003 during a lab presentation by Atsushi Miyawaki, who showed intriguing results with two-photon illumination of GFP. Kawano et al. shined ultra-short (28 femtosecond) pulses from a Ti-Sapphire laser on a plate of immobilized GFP. Due to the uncertainty principal, these ultra-short pulse durations cause a broad spread (~100nm) in the frequency of the laser pulse. They then actively modulated the phase of different frequency bands of the pulse. The interesting part is that they coupled this modulation to a feedback genetic algorithm that sought to increase the ratio of the GFP fluorescence to the intensity of the laser input. Over several hundred iterations of modulation, the system learned how to dramatically increase the output fluorescence over input power by tuning the phase of the frequency components. Using these optimally shaped pulses, they reduced the photobleaching rate by a factor of four! This was an impressive result, but it is unclear how useful this technique would be to live samples, in their heterogeneous aqueous environments. The tuning parameters might be less stable across a non-uniform sample.

The above paper raises intriguing questions on the nature of two-photon excited states and photobleaching. GFP and other fluorescent proteins have multiple bleaching modes, some permanent, some dark or UV-reversible. A more clear understanding of the photochemistry of bleaching could lead to improved illumination pulse design that could keep the chromophore away from these undesired states.

Early last year, Stenfan Hell’s group demonstrated a dramatic reduction in one and two-photon photobleaching by avoiding recurrent excitation of the GFP chromophore when it was in a dark absorbing state.Standard two-photon imaging procedure is to illuminate with a Ti-Sapphire laser pulsed at 80MHz, with a interpulse gap of 12.5ns. This gap is five times longer then the 2.4ns fluorescent lifetime of EGFP, giving the chromophore plenty of time to emit a photon and decay from the excited singlet S1 state. But is the singlet state the precursor to most photobleaching? Donnert et al. varied the pulse rate from 40 to 0.5MHz and discovered that photobleaching was dramatically reduced at the lower pulse rates, especially below 1MHz (1us interpulse interval). Under one photon illumination, total photons extracted from GFP before bleaching was increased 20-fold, while the rhodamine dye Atto532 increased by 8-fold. This suggests that the primary precursor to photobleaching is not the S1 state but is due to photon absorbtion during a dark triplet state T1, which has a relatively long lifetime of ~1us. Don’t illuminate during this state, and prevent most photobleaching! Under two photon illumination (800nm), even greater reductions in photobleaching of 25 and 20-fold respectively took place. This is particularly important because the much higher illumination power used in 2p excitation normally causes a dramatic, non-linear increase in the rate of photobleaching over 1p imaging in the region of focus.

What is special about the T1 triplet state that makes it more prone to causing photobleaching? Is it simply the longer lifetime gives a greater opportunity for an additional photon to hit, jumping to T2 and inducing a photochemical breakdown of the chromophore or the surrounding residues? Does T1 have a broader range of vibrational energies that can more easily engage the variety of photobleaching reactions than S1? How does the photobleaching rate of S2 compare to T2?

Despite the impressive reduction in photobleaching, and hence the greater S/N for a given bleach rate, Hell’s approach has a major drawback. Slowing the pulse train down ~100-fold also slows acquisition time down 100-fold. Therefore this technique is most useful for fixed specimens. Real-time, high resolution imaging of dynamic processes would be seriously degraded. There are work-arounds, such as wide-field pulsed illumination, rapid laser sweeping or multipoint parallel illumination, but these require additional technical development to make them feasible for the average, or even well-above average investigator.

Is there any related solution that can be easily applied to imaging live, dynamic cells? In this month’s Nature Methods, Na et al. from Eric Betzig’s group present an ‘exciting’ approach. They note that with conventional two photon illumination, most of the available laser power is wasted, intentionally blocked to reduce photodamage. As alluded to above, photodamage increases non-linearly with illumination intensity, making two-photon illumination methods particularly harmful. The authors demonstrate that this damage increases proportional to intensity to the ~2.4 power. To try to attenuate this effect, they used a series of mirrors to split the single ultra-short intense pulse (140 femtosecond) in half, and half again and again and again… They end up with 128 pulses of 1/128 the full intensity nearly evenly spaced every 37 picoseconds. This entire pulse group has a duration of 12ns, and with a 80MHz pulse cycle (12.5ns inter pulse interval) illuminates the chromophore with a steady stream of relatively dim pulsed light.

This split pulse illumination dramatically enhances acquisition speed and signal/noise. Images acquired at a rate of 0.4us/pixel with splitting look far more clear than those acquired at 25.6us/pixel without splitting. Photobleaching is reduced by a factor of four and acute photodamage is also reduced. Additional splitting may be possible and further improve the photobleaching attenuation. Importantly, they demonstrate this technique with GFP in fixed brain slices and in live worms and imagine dynamic responses with a calcium dye in living hippocampal slices. This technique appears to let you eat your cake and have it too. The implementation of the pulse splitting is modular, appears relatively simple to those with customizable two-photon instruments and works with existing Ti-Sapphire lasers. I anticipate rapid adoption by serious imaging labs.

Each of the above advances attacks the problem of indicator photobleaching by a different approach, and each focuses on a different aspect of the photochemistry. Theoretically, one could even combine all three for maximum photon collection efficiency before photobleaching, though this would also require combining the drawbacks of each. Photobleaching will continue to be a major concern in the imaging of dynamic processes, particularly when the signal is not synchronized with the onset of image acquisition. These techniques show substantial progress towards alleviating this concern, and I’m heartened to see a number of excellent labs are focusing so much energy on it.

A final question : How will each of these techniques affect acceptor photobleaching (I’m looking at you Citrine and Venus) in FRET imaging experiments? Do the same processes apply when the excitation is coming from a FRET donor?

We are back!

Time to get this blog rolling again…

School and work related things kept me from posting for a while, but now I’m in the groove in a new position and can start doing more frequent updates.

Since the last post I’ve got a paper in press with Yongling Zhu and Roger Tsien on optical imaging of glutamate with genetically-encoded reporters. Brainwindows will review the field of glutamate imaging once the paper is available online (any day now…). I also finished my thesis on Design, Development and Use of Genetically-Encoded Fluorescent Reporters of Neuronal Activity and got my Ph.D. from UCSD’s Neurosciences program. Drove across the country and started a post-doc at Janelia Farm with Loren Looger and Karel Svoboda. I’m hopeful our work will make a significant positive impact on the usefulness of genetically-encoded optical indicators in vivo.

-Andrew

UCSD vs. MIT SFN Party Smackdown

The Society for Neuroscience conference starts today in America’s Finest City (San Diego). The question on everyone’s mind is, who is going to throw the best party? Sure there are plenty of themed mixers and socials, but few really stay interesting for long.

The past few years, the Picower Center for Learning and Memory at MIT has consistently had the biggest bash, really peaking in 2006 at the eye-popping Atlanta mega-club Compound. With a big open bar tab that unfortunately gets drained within an hour, and an open invitation, these are always packed with people early on, go strong till last call, and feature plenty of Neuroscience ’star power’. This year, the party starts Monday at 9pm at Deco’s on 5th Ave. in the Gaslamp. Get there early, as Deco’s is a relatively small place.

Nature and Neuron each throw lower-key parties, with the best hors d’oeuvres and are definitely the place to do serious science/business networking. Security is pretty loose, as long as you let the door know that you know that the party is for Nature or Neuron. When and where these parties might be in San Diego is under intense investigation by BrainWindows staff.

The most exclusive of all are the mysterious Emory parties, where you better bring the printout of your personalized invitation email if you want to get in.

This year, there is a new group that is trying to dethrone the PCLM as hosts of the biggest event. UCSD Neurosciences is hosting an open-invite, open-bar event this Sunday at Aubergine, at 4th & Island in the Gaslamp. The bar tab opens at 9pm, and if the PCLM parties are any guide, I would get there at 9. Bring friends!

Who will impress the community the most? PCLM has a five year reputation, and the experience of Earl Miller and Susumu Tonegawa behind it. But UCSD knows San Diego, and it’s grad-student run social committee has held numerous, very successful local events. As a soon-to-be alum of both UCSD and PCLM, I’m looking forward to finding out who does it best. See you there!