Preview : Structure of G-CaMP2

10 12 2008

A high-resolution crystal structure of the genetically-encoded calcium indicator G-CaMP2 would aid in rational design of improved calcium indicators. Crystallization of G-CaMP2 was first reported here :

Crystallization and preliminary X-ray characterization of the genetically encoded fluorescent calcium indicator protein GCaMP2

M. M. Rodríguez Guilbe, E. C. Alfaro Malavé, J. Akerboom, J. S. Marvin, L. L. Looger and E. R. Schreiter

Fluorescent proteins and their engineered variants have played an important role in the study of biology. The genetically encoded calcium-indicator protein GCaMP2 comprises a circularly permuted fluorescent protein coupled to the calcium-binding protein calmodulin and a calmodulin target peptide, M13, derived from the intracellular calmodulin target myosin light-chain kinase and has been used to image calcium transients in vivo. To aid rational efforts to engineer improved variants of GCaMP2, this protein was crystallized in the calcium-saturated form. X-ray diffraction data were collected to 2.0 Å resolution. The crystals belong to space group C2, with unit-cell parameters a = 126.1, b = 47.1, c = 68.8 Å, [beta] = 100.5° and one GCaMP2 molecule in the asymmetric unit. The structure was phased by molecular replacement and refinement is currently under way.

High-resolution atomic structures and mutational analysis were presented at SfN 2008 (see this previous post)

However, today a competing group has published an independent report on a similar set of G-CaMP2 structures in Cell Structure.  More details to come…


Structural Basis for Calcium Sensing by GCaMP2

Qi Wang1,Bo Shui2,Michael I. Kotlikoff2andHolger Sondermann1,Go To Corresponding Author,

Genetically encoded Ca2+ indicators are important tools that enable the measurement of Ca2+ dynamics in a physiologically relevant context. GCaMP2, one ofthe most robust indicators, is a circularly permutated EGFP (cpEGFP)/M13/calmodulin (CaM) fusion protein that has been successfully used for studying Ca2+ fluxes invivo in the heart and vasculature of transgenic mice. Here we describe crystal structures of bright and dim states of GCaMP2 that reveala sophisticated molecular mechanism for Ca2+ sensing. In the bright state, CaM stabilizes the fluorophore in an ionized state similar to that observed in EGFP. Mutational analysis confirmed critical interactions between the fluorophore and elements of the fused peptides. Solution scattering studies indicate thatthe Ca2+-free form of GCaMP2 is a compact, predocked state, suggesting a molecular basis for the relatively rapid signaling kinetics reported for this indicator. These studies provide a structural basis for the rational design of improved Ca2+-sensitive probes.

Some interesting posters @ SfN

20 11 2008

Here’s a few posters that caught my eye at SfN.  Click the meeting planner for the full abstract

Optimizing two-photon activation of channelrhodopsin-2 for stimulation at cellular resolution


Spiral pattern of 2-photon excitation can drive neurons to spike.  A low NA objective helps. Need to do piezo-based Z-scanning if you use high NA, don’t with low NA.

In vivo two-photon imaging 1 mm deep into cortical brain tissue with novel microprism probe 


A cute method to image 1mm into cortex with 2-photon imaging. They used 2-6 month old mice. The just took a triangular prism whose hypotenuse was silvered and stuck it in the cortex. Then they internally reflected the beam off the prism and fired it sideways into cortex. Got good SNR to 300um lateral distance.  Some clippling of beam at edges of the prism gave somewhat inconsistent spatial resolution.

Self-complementary adeno-associated viral vectors for fast, efficient labeling of neurons and astrocytes in visual cortex in vivo


AAV is the way to go for expression of GECIs and ChR2 in vivo, but it takes a long time to express at high levels (2 weeks). They show that using a double stranded DNA version of AAV rather than single stranded gets protein expression up high much faster. Very high expression after one week. This is because the virus doesn’t need to take the time to make the second strand before expressing the protein.  See Xiao, X J. Virol 1998

Detection of single action potentials in vitro and in vivo with genetically-encoded Ca2+ sensors


Everything in the poster was in the Nature Methods paper.  Conversation reveled that YC3.60 works as well or better than D3cpv. Only have done up to whisker evoked stimulation, no imaging of spontaneous YC3.60 signals yet.

Characterization of improved probes for the hybrid voltage sensor method of voltage imaging


A nice little sensor optimization poster.  They took the hVOS hybrid voltage sensor of dipicrylamine with membrane tethered GFP and improved it by changing the chromophore to Cerulean, and by using the “membrane-staple” strategy. Having membrane anchors on both the N and C-termini gave better quenching. Fast response, ~0.5ms, and 20% dF/F.

Crystal structure of the genetically encoded calcium indicator gcamp2


Jasper made crystal structures of G-CaMP2 in the apo and bound states.  Bound states crystalized as a heterodimer, but he was able to also crystalize the monomer. The structures show a pore to the chromophore in the apo state that is plugged in the Ca-bound state. Thus, the quenched apo state is due to solvent access to the chromophore.  This structural data should help rational design of better G-CaMP sensors.

Styryl dyes may inhibit synaptic release

19 11 2008

A new report in PNAS, Probing synaptic vesicle fusion by altering mechanical properties of the neuronal surface membrane, from Chuck Stevens’s lab, raises a serious concern about using styryl dyes to study release probability of synapses.  Styryl dyes, such as FM 1-43, partition into cell membranes and have been commonly used to measure synaptic release of vesicles in culture and brain slice.  The protocol is simple, bathe the neurons in dye, electrically stimulate to cause massive synaptic release and then dye uptake via vesicular endocytosis, wash off the dye, then observe the rate of destaining of the synapses following electrical stimulation.  This rate is directly related to the number of vesicle fusions during the final stimulation period. There is just one problem, Zhu and Stevens report that the presence of the dye reduces the release probability of the synapse.


FM 4-64 reduces synaptophysin-pHluorin response

FM 4-64 reduces synaptophysin-pHluorin response

They observed this by taking the FM dye measurements in neurons that expressed synaptophysin-pHluorin.  The fluoresence increase from the pHluorins was reduced in a [dye] dependent manner. A 15uM concentration of dye (not atypical for published experiments) reduced the pHluorin signal by 40%.  The dye had no effect on the calcium levels in the presynaptic terminals, indicating it was potentially due to an increased energetic cost of forming a fusion pore. Chuck then weaves together some basic principles with this data to make an estimate of the fusion pore size.  

While this paper may seem to cover a minor technical point, FM dyes have been used to make numerous inferences about presynaptic release properties, modes of vesicle recycling, the locus of LTP expression, and other basics of synaptic physiology. The thing that’s bugging me is if this effect is as prominent as advertised, how did people not notice the change in release probability with electrophysiological techniques?

Raw Data : Vesicular Release from Astrocytes, SynaptopHluorange

15 11 2008

When I was working on my Ph.D. thesis, I was trying to find some biological question to definitively answer with GluSnFR, my glutamate sensitive fluorescent reporter. One possibility was the study of glutamate release from astrocytes. Around that time, 2003/2004, there was increasing evidence that glutamate was not just scavenged by astrocytes, but was also released from astrocytic vesicles. It released in response to calcium elevations within the cell. Existing methods for measuring this release were somewhat crude, so it seemed a great test system for GluSnFR.

Unfortunately, since there seemed to be no specialized areas on the astrocyte where the vesicles fused, and the release rate was relatively slow, we were unable to detect glutamate release with GluSnFR. I thought this might be a problem of not knowing when and where to look. So my collaborator, Yongling Zhu, and I expressed pHluorins fused to VAMP or to synaptophysin in astrocyte cultures. When we looked at them under the microscope, they just looked green, no action…

But then we left the excitation light on for a few minutes. I happened to look back into the scope after they had been bathing in bright blue light and was astonished. I could directly see, by eye, spontaneous bursts of fluorescence across the cells. It was absolutely magnificent. The long application of light had bleached all of the surface expressed, bright pHluorins. But the pH-quenched pHluorins in the vesicles were resistant to bleaching. On this dimmer background, the fusion events were plain as day.

Unfortunately, the green color overlapped with the emission of GluSnFR, so we couldn’t use it for a spatiotemporal marker of when and where to look for glutamate release. We tried using some ph-sensitive precursors to mOrange and mOrange2, developed by Nathan Shaner, but these seemed to block the release events. Since then, others have shown the functional relevance of glutamate release from astrocytes, and I turned the focus of GluSnFR measurements to synaptic spillover. This was one of the projects that was tantilizingly close, but got away. This movie of VAMP-pHluorin is almost five years old now, but it still looks cool… Enjoy!

If you are curious, this is what the Synaptophysin-mOrange looked like when we expressed it in hippocampal neuron cultures. Ammonium Chloride caused a massive fluorescence increase, by alkalizing the synaptic vesicles. Unfortunately, we never were able to see release via electrical stimulation. Details are in my thesis. Maybe someone else wants to give it a shot?

The great GECI shootout

21 07 2008

Dierk Reiff’s lab has done another head-to-head in vivo showdown between various GECIs and a synthetic dye. Their paper, Fluorescence changes of genetic calcium indicators and OGB-1 correlated with neural activity and calcium in vivo and in vitro, is very interesting and deserves a full write-up. I will present a detailed analysis of the paper in a future update.  For now, check the abstract.

Recent advance in the design of genetically encoded calcium indicators (GECIs) has further increased their potential fordirect measurements of activity in intact neural circuits. However, a quantitative analysis of their fluorescence changes ({Delta}Fin vivo and the relationship to the underlying neural activity and changes in intracellular calcium concentration ({Delta}[Ca2+]i) has not been given. We used two-photon microscopy, microinjection of synthetic Ca2+ dyes and in vivocalibration of Oregon-Green-BAPTA-1 (OGB-1) to estimate [Ca2+]i at rest and {Delta}[Ca2+]i at different action potential frequencies in presynaptic motoneuron boutons of transgenic Drosophila larvae. We calibrated {Delta}F of eight different GECIs in vivo to neural activity, {Delta}[Ca2+]i, and {Delta}F of purified GECI protein at similar {Delta}[Ca2+in vitro. Yellow Cameleon 3.60 (YC3.60), YC2.60, D3cpv, and TN-XL exhibited twofold higher maximum {Delta}F compared with YC3.3 and TN-L15 in vivo. Maximum {Delta}F of GCaMP2 and GCaMP1.6 were almost identical. Small {Delta}[Ca2+]i were reported best by YC3.60, D3cpv, and YC2.60. The kinetics of {Delta}[Ca2+]i was massively distorted by all GECIs, with YC2.60 showing the slowest kinetics, whereas TN-XL exhibited the fastest decay. Single spikes were only reported by OGB-1; all GECIswere blind for {Delta}[Ca2+]i associated with single action potentials. YC3.60 and D3cpv tentatively reported spike doublets. In vivo, the KD(dissociation constant) of all GECIs was shifted toward lower values, the Hill coefficient was changed, and the maximum {Delta}F was reduced. The latter could be attributed to resting [Ca2+]i and the optical filters of the equipment. These results suggest increased sensitivity of new GECIs but still slow on rates for calcium binding.

Deep & local Channelrhodopsin-2 two-photon activation

17 07 2008

An interesting paper on two-photon activation of channelrhodopsin-2 is out in Biophysical Journal. In In-depth activation of ChR2 sensitized excitable cells with high spatial resolution using two-photon excitation with near-IR laser microbeam, Mohanty et. al show cellular activation with a fast-scanning two-photon laser.

Action potential generation from Channelrhodopsin-2 with a two-photon beam has been difficult to achieve, presumably due to the small activation volume of the 2p spot. They show similar calcium transients in response to 2p stimulation as with one-photon stimulation. As depth increases, the one-photon response attenuates faster than the two-photon. Unfortunately, the supplemental info with  electrophysiology traces are not yet online.  Presumably, they are generating action potentials, but I’d like to see the raw data.  Interestingly, they also show calcium increases when the laser stays in once place.  This would imply that local depolarization causes local voltage-gated calcium channels to open, or that calcium is getting through the ChR2. I was under the impression that ChR2 has a low conductance for calcium, though this study by Caldwell et. al, in press for JBC, uses ChR2 specifically for its calcium permeability.

I’m not sure what to make of the first paper. Are they really able to fire action potentials with two-photon stimulation, at depth?  Or are the calcium traces they are seeing simply the result of localized calcium flux.  I’ll followup once the Supplemental Data becomes available.  Still worth a look if this is the sort of thing you are interested in.

Sensing salty currents with Mermaids

16 07 2008

A new genetically-encoded voltage sensor paper is out from a friend and former mentor of mine, Atsushi Miyawaki. One memorable moment when working in his lab during the RIKEN summer program of 2002 was when Atsushi took me into his office and whipped out a custom green laser pointer. These had been banned in Japan, as fans would shine their powerful light into the eyes of pitchers and batters at baseball games. Atsushi was really proud of his. He smiled and then started sweeping the light point over the rocks in his fishtank. Each ‘rock’ was actually coral his lab had collected from fluorescent protein hunting trips, and each glowed a different color when the green light hit it. He has been putting these novel discoveries to good use.

In Improving membrane voltage measurements using FRET with new fluorescent proteins, Tsutsui et. al take two fluorescent proteins discovered and engineered by the Miyawaki lab, mUKG and mKOk, and graft them onto the Ci-VSP scaffold used in VSFP2.1 (also developed at RIKEN).  The green and orange fluorescent proteins undergo significant FRET transfer which is voltage dependent.  They get 40% dR/R per 100mV with a 2 component association rate of around 10 and 200ms. Unsurprisingly, the kinetics speed up at physiological temperatures to 5-20ms on and off.  They are able to pick up single pseudo-action potentials in Neuro2A cells, though the response is highly filtered. They are also able to see very clear spontaneous waves of potential change in cardiomyocytes (23% dR/R) and single spikes in cultured neurons (2% dR/R for 1AP). They dub this voltage sensor “Mermaid”.

The authors state that they used the new FPs due to their improved photostability and especially pH resistance. 

Additionally, because Aequorea GFP variants are pH-sensitive, and neuronal activity causes considerable acidification, the responses of sensors to depolarization in intact neurons may be overwhelmed by sustained changes resulting from acidification.

Granted that mOrange2 is pretty pH-sensitive, but I’m not sure this is a real issue, or a potential issue to justify using their new FPs.  From the spectra of mUKG vs. EGFP, it would seem that EGFP’s 10nm further redshifted emission would be a superior FRET pair for mKOk.  It smells like there may be a bit of bundling of various independent projects into this paper.  However, they do make a good point that this pair will have a different preferred dipole orientation than existing FRET pairs, which could lead to improved performance in some constructs.  

Things I’m still wondering :

  • Have they tried using the improved VSFP3.1 scaffold? This was shown to be much faster than 2.1.  I suspect the mUKG is not as tolerant to C-terminal truncation than CFP and GFP.  
  • What about using EGFP as the donor?  Could you then use the VSFP3.1 scaffold?
  • Is there a rapid non-FRET quenching of the donor upon depolarization as seen in VSFP3.1?
  • Why is the single wavelength fluorescence increasing in both channels in figure 2d?  Is there some photoactivation going on?
  • I’d love to see a head to head comparison of VSFP3.1 and Mermaid under identical conditions. Also responses in brain slice at physiological temperatures.

Voltage sensitive imaging powering up

8 07 2008

I’m starting to come around on voltage imaging. I haven’t been a fan of it for a number of reasons.

  • The response sizes suck.  Classic dyes and genetically encoded systems get a few percent fluorescence change at best. 
  • The response speeds suck. Measuring continuous current injections from -100mV to +150mV is not very interesting.  Action potentials are interesting.  But they are fast.
  • Toxicity. The dyes kill neurons, or strongly perturb their electrical properties.

OK, voltage-sensitive imaging isn’t totally useless, for example see Carl Petersen’s recent paper on Spatiotemporal Dynamics of Cortical Sensorimotor Integration in Behaving Mice (2007). But if the above problems could be solved, then voltage sensitive imaging would be a strong competitor to calcium imaging for the non-invasive, high-resolution monitoring of patterns of network activity. There has been considerable progress ameliorating these problems in the past few years, much of it by a consortium of labs (Isacoff, Knöpfel, Bezanilla, Miesenböck, and others) focused on these issues. (Umlaut’s apparently help in this field).

First, let’s look at a minor breakthrough for the fully genetically-encoded strategy.  In Engineering and Characterization of an Enhanced Fluorescent Protein Voltag Sensor (2007), The Knöpfel group tagged the recently discovered voltage sensitive phosphotase (Ci-VSP) with CFP and YFP FRET pairs in place of the phosphotase domain. This tagged protein expressed at the membrane much more efficiently than previous genetically encoded voltage sensors based on potassium channel subunits. By injecting physiological voltage changes and averaging 50-90 traces, they were able to pull out a few percent ratio change from a brief series of action potentials. Single spikes were resolvable.  Although this sensor (VSFP2.1) was pretty slow (tau > 10ms), this new substrate looked promising for future sensor development.

They have since sped the response up.  In Engineering of a Genetically Encodable Fluorescent Voltage Sensor Exploiting Fast Ci-VSP Voltage-Sensing Movements (2008 ), they determined that the gating motion of the voltage sensing component was very fast (~1ms), while the fluorescence change was slow (~100ms). So they did what any good FRET tinkerer would do, chop away at the linkers between FP components.  The sensor response improved, and they noticed that there was a disconnect between the speed of the CFP and YFP responses. Not only was CFP decreasing from enhanced FRET, it was being directly quenched by interactions with the lipid membrane. Chopping off the YFP from the the construct then dramatically increased the speed of the CFP quench. This improved sensor, VSFP3.1 has an activation time constant of 1.3ms, though it’s response magnitude is still quite small (a few % dF/F).

A hybrid approach to measuring electrical activity in genetically specified neurons (2005) has a much greater response magnitude. Pancho Bezanilla’s group exploited the rapid, voltage-dependent translocation of the small molecule quencher dipicrylamine (DPA) through the plasma membrane to change the fluorescence of membrane-teathered GFP in a voltage-dependent manner. Responses of the hybrid voltage sensor (hVOS) were relatively large (34% per 100mV) and fast (0.5ms). Single action potentials were detectable without averaging.  However, since DPA is a charged molecule, it significantly increased the capacitance of the membrane. The levels of DPA required to see large responses inhibited action potentials and were intolerable to neurons.

Last month in Rational Optimization and Imaging In Vivo of a Genetically Encoded Optical Voltage Reporter (2008 ), Sjulson and Miesenböck reported optimized parameters for the hVOS approach. They built a quantitative model of the quenching effects of DPA on membrane-teathered GFP.  The quenching is limited by the distance the DPA can approach the chromophore of GFP.  Only the closest DPA molecule to the chromophore significantly contributes to a GFP’s quenching. After lots of pretty heat maps and graphs, the model tells them to chop off the tail of EGFP to bring the C-terminal tethering sequence closer to chromophore. I should note that an 11 amino acid C-terminal truncation of ECFP has improved the response of a tremendous number of FRET reporters and has been standard practice for the last 8 years. By shortening the linker they manage to triple the response size. I’d suggest, if they haven’t already, to lop off another six amino acids (end the EGFP with …LEFVTAA) and see if works.  EGFP and ECFP usually tolerate it.

Using this optimized reporter, they are able to reduce DPA concentrations to levels that are usable in vivo, at least for a few minutes. They record fast optical responses to electrical activity in the Drosophila antennal lobe using 2uM DPA.  But after a few minutes, the DPA loaded neurons become strongly inhibited.


The bottom line? Voltage-sensitive imaging has seen big progress in the last few years, but still has a long way to go to gently record single APs in a dish or in vivo. Or does it?  I’m hearing whispers that a different group has developed a synthetic dye technique that is getting >10% dF/F to single APs with millisecond response times. Is it the real deal? Watch this space…

SLICK labeling and new FPs

1 07 2008

There is a nice writeup of the single-neuron labeling with inducible Cre-mediated knockout (SLICK) paper from Guoping Feng‘s lab over at the Alzheimer’s Research forum. The method simultaneously knocks out a gene in a small number of cells, while highlighting the knocked-out cells with a cytosolic fluorescent protein. In a comment to the Schizophrenia Research Forum, Joseph Gogos points out a similar technique his lab published last year in Current Biology.

Also in the writeup is coverage of the new fluorescent protein variants from the Tsien Lab.  These include mOrange2 made by Nathan Shaner, which is a much more photostable version of mOrange. This should immediately replace mOrange in most constructs.  Also of note is TagRFP-T from Michael Lin and his trusty undergraduate assistant Michael McKeown. Tag-T is an extremely photostable derivative of the Evrogen protein TagRFP. Tag-T was discovered by screening Tag mutants in bacterial colonies on a solar simulator. Toxicity in sensitive cells (in vivo neurons) hasn’t been fully determined yet, but in vitro these new FPs all look great. Now I wish they would make a super-bleach resistant Citrine for my FRET constructs.

Giving synapses a ‘born on’ label

30 06 2008

Memories are thought to be encoded by the patterns of synaptic connections in the brain. Learning can either delete or change the strength of existing synapses, or add new synapses. Following a learning process, how can we tell which synapses were added to encode this new memory?  

One strategy is to make a timelapse movie of the synapses.  In mice, this can be accomplished by installing a cortical window on the skull, and imaging the changes in structure of GFP labelled neurons. However, this is technically demanding, only works with sparsely labeled neurons, and accesses only a small subset of the neurons which may be involved in the learning process.  

Ideally, one could have a tag which can discriminate between synapses existing before learning takes place, and new ones generated after learning has occurred. Whole brain regions could then be examined at a single timepoint to see where new synapses were added. In a large step towards that goal, Michael Lin et. al, from the lab of Roger Tsien, report TimeSTAMP, a genetic label for newly synthesized protein.

The authors engineered the NS3 protease from the hepatitis C virus (HCV) to cleave itself at just the right pace. They then fuse tags (fluorescent proteins or epitopes) before and after the cleavage site. This fusion is then tagged to the end of a protein of interest. Shortly after synthesis, the protein cleaves off the C-terminal tag, but the N-terminal is left on. This cleavage is inhibitable by a variety of small molecule blockers. In the presence of the blocker, the C-terminal tag stays on. By controlling when drug is applied, they can selectively label a set of proteins of a particular age with the tags.

The choice of NS3 protease was very clever, as it is a favorite drug target of biotech and pharma companies.  Many inhibitors of this protein have been synthesized, exhaustively characterized in vitro and in clinical trials. This work is a great example of the standard research flow going in reverse; a basic-science project from an academic lab is actually benefitting from pharma company research. Stability, bioavailablity and toxicity have already been worked out.  One of the biggest impediments is actually getting ahold of these compounds. Companies with their survival hanging on the clinical success of a single small molecule inhibitor are understandably reluctant to hand out stocks for academic research. Note the roller coaster stock price of Vertex following results of its NS3 protease inhibitor (VX-950) trials. 

The authors use PSD-95 tagged to TimeSTAMP as a proxy marker of synaptic age. In neuronal culture, they show that newly synthesized synapses have a C-tag / N-tag ratio of about twice as large as old synapses.

They extend the technique to whole fruit fly brains, showing a very heterogeneous distribution of CaMKII synthesis across Kenyon cells in different areas of the mushroom body.

So far TimeSTAMP has not been shown to work in mice. Mice were not included in the paper due to the long generation time for transgenics. Given the good signal to noise and the large number of possible inhibitor molecules, I think this technique could be quite powerful in mammalian systems. It’s big advantage would be to label large populations of neurons or synapses in diverse brain regions, including those inaccessible to two-photon microscopy. TimeSTAMP’s success in labeling new synapses in the intact brain will be dependent on finding a protein to tag at the synapse with low turnover over the course of a learning experiment. Though PSD-95 appears to be a reasonable marker in culture, others have shown a higher rate of turnover in vivo, making in unsuitable for a synaptic marker.