Reader Comments

Post a new comment on this article

mCerulean3 is less photostable than mTurquoise in living cells

Posted by Gadella on 27 May 2011 at 14:15 GMT

After careful comparative analysis we were unable to reproduce the major claims made in this paper. On the contrary, our results indicate that there is no advantage of mCerulean3 over mTurquoise, and that the photostability of mCerulean3 in mammalian cells is inferior to mTurquoise and even worse as compared to ECFP. For the comparison we used a mCerulean3 clone obtain from the second author of the paper. The correctness of the construct was confirmed by sequencing.

We found that under widefield illumination conditions (436 nm Hg lamp), mCerulean3 when expressed in HeLa, N1E or NIH3T3 cells was > 10x more sensitive to bleaching as compared to mTurquoise. A movie in which the photostability of mTurquoise and mCerulean3 are compared (under exactly identical conditions, excitation power of 1.6 W/cm2) can be found here ( Moreover, no recovery of mTurquoise fluorescence was observed after interruption of bleaching, indicating that mTurquoise does not show significant reversible photobleaching in vivo. The decreased photostability of mCerulean3 makes it less suitable for FLIM-FRET applications because upon (partial) bleaching, its fluorescence lifetime decreases more than that of mTurquoise, possibly leading to false-positive FRET results by FLIM.

The large discrepancy between the data published in this PLOS One paper by Markwardt et al. and our own repeated observations most likely are due to the artificial environment for testing photostability by Markwardt et al. The authors used purified protein attached to HiTrap beads mounted in antifade medium (Prolong Gold from Invitrogen). We did not repeat these measurements but we feel that living cells are the more relevant environment for testing the photostability of the fluorescent proteins.

We could not confirm the claim that mCerulean3 is brighter than mTurquoise (table 1). According to Table 1, the authors suggest that the extinction coefficient of mCerulean3 is 40,000 M-1 cm-1. This value is too high. Also the tabulated extinction coefficient of mCerulean is too high. We think the problem is that these extinction coefficients were obtained using calculated protein concentrations instead of experimentally determined protein concentrations (see reference 5 of the paper). Our experimental determined extinction coefficient of FPLC purified recombinant mCerulean, mCerulean3 and mTurquoise are all around 30,000-33,000 M-1 cm-1. Correct values for the extinction coefficients were published by us and others (see ref 21, 26 and 27). Our spectroscopic analysis revealed that the extinction coefficient and quantum yield of mTurquoise and mCerulean3 are quite similar. This is confirmed by our quantitative brightness analysis in cells and by single molecule FCS analyses (see ref 26 for methodology) that reveal a similar brightness of mCerulean3 and mTurquoise (the latter being ~ 5% brighter in mammalian cells). These results are expected since mCerulean3 carries the same T65S mutation that is responsible for the brightness enhancement in mTurquoise (see ref 26).

No competing interests declared.

RE: mCerulean3 is less photostable than mTurquoise in living cells

mrizzo replied to Gadella on 27 May 2011 at 17:59 GMT

This is an interesting post from Dr. Gadella on his comparison with mTurquoise (developed by his laboratory) and mCerulean3 (developed in mine) in mammalian cells. As the data he refers to is unpublished, I can only provide clarification on our own unpublished observations.

1. Regarding the comparison of Dr. Gadella’s protein and ours, we do find that mCerulean3 and mTurquoise are both good and useful, and do have very similar brightness by any measure (our calculations showed mCerulean3 to have an edge by basically the same difference as Citrine and Venus, or less than a fifth of a photon, to put it in perspective), and that both proteins can be bleached with enough power. Bleaching is a hard thing to quantify, particularly in cells, where things like focus shifts, changes in cell shape and cell health can affect the observed fluorescence. The movie that Dr. Gadella has generously posted shows this elegantly. The mCerulean3 cells change shape during the course of the experiment. Is this a change in focus (as I would suggest), or shape, or simply due to decreased fluorescence around the edges from bleaching (as Dr. Gadella argued)? Or a combination of all 3 (likely)? It is hard to know definitively. This was the reason we chose to examine bleaching in recombinant preparations and it is my belief that this is best way to understand the molecular behavior. As for cells, which is protein is better? Those of us who make FP-fusions regularly know that this is at present, a decision that is best made empirically. For example, some of our constructs work better with Venus, some with Citrine. What I can share is that in our hands mCerulean3 has worked quite well for the 30+ fusions we have made, and that the FRET variance goes down considerably compared to previous CFPs (even mTurquoise).

2. With regards the extinction coefficient, I stand by our numbers (which were from our own data, not from the literature, using experimentally determined protein concentrations calibrated against recombinant GFP). I respectfully disagree with Dr. Gadella about the source of the inconsistencies. In my view, the variance in the numbers may be from unfolded protein contaminating the protein batch, hence lowering the calculated extinction coefficient value. We get much better protein folding with the T5 system than we ever did with the T7 system used by Dr. Gadella’s lab and others. I suggest that the values obtained from proteins prepared using the T7 system are too low, and that the methods referenced in the comment to do not account for the presence of unfolded protein in the absorbance measure.

3. With regards to “The decreased photostability of mCerulean3 makes it less suitable for FLIM-FRET applications because upon (partial) bleaching, its fluorescence lifetime decreases more than that of mTurquoise, possibly leading to false-positive FRET results by FLIM.”, I am not sure why this should be the case for a properly controlled experiment, in terms of leading towards a false-positive and not just increased variance in the lifetime. To the extent that the mCerulean3 lifetimes measured in cells is variable, we found the statistical variance to be very similar between the two proteins.

No competing interests declared.

RE: RE: mCerulean3 is less photostable than mTurquoise in living cells

mrizzo replied to mrizzo on 09 Jun 2011 at 23:42 GMT

Upon further investigation, we have not been able to reproduce the findings of Gadella. In COS7 cells at 37 degrees, mCerulean3 bleaches less than 5% over 200 images (~25 s of continuous illumination, n=8).

No competing interests declared.

mCerulean3 is less photostable and less bright than mTurquoise in living cells

Gadella replied to mrizzo on 22 Mar 2012 at 14:42 GMT

All statements made in our comment from May 11th 2011 can now be verified quantitatively from our new published study. We refer to the Nature Communications website:

In this paper we show that the extent of photobleaching in live cells of mTurquoise is 3±1% and of mCerulean3 is 39±2 % after 160 s wide-field illumination at 1.6 W/cm2 (436 nm Hg lamp). Hence under these conditions in cells, Cerulean3 is 13x less photostable. Similar results are obtained in 3T3 cells, HeLa cells and N1E-115 cells. In addition, we show that the mTurquoise quantum yield is 0.84±0.02 whereas mCerulean3 quantum yield is 0.80±0.01. Furthermore, we demonstrate that extinction coefficients of mTurquoise and Cerulean3 are equal (same absorption spectra) and 30,000 cm-1 M-1.

The increased brightness of mTurquoise versus Cerulean3 was confirmed with 3 independent experimental approaches: extinction coefficient x quantum yield; single molecule molecular brightness with FCS experiments; and in vivo detected relative brightness upon quantitative coexpression with YFP in cells.

In the above publication we also describe a new even brighter cyan fluorescent protein mTurquoise2 with a quantum yield of 93% . In cells mTurquoise2 displays 20% increased brightness as compared to mTurquoise. We demonstrate that mTurquoise2 is the preferred CFP donor to YFP in FRET studies. Because the crystal structures of the Turquoise proteins have been solved, we now understand the structural basis for the improved spectroscopic properties.

No competing interests declared.

RE: mCerulean3 is less photostable and less bright than mTurquoise in living cells

mrizzo replied to Gadella on 01 May 2012 at 22:17 GMT

To clarify,

1. The CFP described by Dorus and company was not obtained from the corresponding PI's laboratory. Notably absent from the referenced manuscript is how mCerulean3 was obtained or constructed.  I cannot verify that the data provided by Dorus's lab is from a protein 100% identical to the one characterized in the referenced manuscript.

2. We stand behind our published QY of 0.88. This value was independently reproduced in 2 different laboratory settings (at Vanderbilt and Maryland)

3. The photoswitching behaviors of turquoise and mCerulean3 were also independently reproduced in 2 different laboratories (at Maryland and Florida)

4. The lifetime properties of mCerulean3 were independently verified in two laboratories (at Maryland and Indiana).

5. The marginal improvement provided by the I146F Turquoise2 mutation was not observed in our Cerulean CFPs. In fact, we reported a marginal decrease in fluorescence (see Table S1). Our discovery of this mutation was overlooked by the referenced manuscript.

No competing interests declared.