+1 Recommend
    • Review: found
    Is Open Access

    Review of 'The Spherical Nucleic Acids mRNA Detection Paradox'

    The Spherical Nucleic Acids mRNA Detection ParadoxCrossref
    An important evaluation of SmartFlare, but lacks quantification & many controls.
    Average rating:
        Rated 4 of 5.
    Level of importance:
        Rated 4 of 5.
    Level of validity:
        Rated 3 of 5.
    Level of completeness:
        Rated 5 of 5.
    Level of comprehensibility:
        Rated 3 of 5.
    Competing interests:

    Reviewed article

    • Record: found
    • Abstract: found
    • Article: found
    Is Open Access

    The Spherical Nucleic Acids mRNA Detection Paradox

    From the 1950s onwards, our understanding of the formation and intracellular trafficking of membrane vesicles was informed by experiments in which cells were exposed to gold nanoparticles and their uptake and localisation, studied by electron microscopy. In the last decade, building on progress in the synthesis of gold nanoparticles and their controlled functionalisation with a large variety of biomolecules (DNA, peptides, polysaccharides), new applications have been proposed, including the imaging and sensing of intracellular events. Yet, as already demonstrated in the 1950s, uptake of nanoparticles results in confinement within an intracellular vesicle which in principle should preclude sensing of cytosolic events. To study this apparent paradox, we focus on a commercially available nanoparticle probe that detects mRNA through the release of a fluorescently-labelled oligonucleotide (unquenching the fluorescence) in the presence of the target mRNA. Using electron, fluorescence and photothermal microscopy, we show that the probes remain in endocytic compartments and that they do not report on mRNA level. We suggest that the validation of any nanoparticle-based probes for intracellular sensing should include a quantitative and thorough demonstration that the probes can reach the cytosolic compartment.

      Review information

      Review text

      ( At the request of Raphael, I copy here the comments that I posted in my blog : https://greenfluorescentblog.wordpress.com/2015/11/18/new-data-on-smartflare-do-they-detect-mrna/ [slightly editted])

      Almost exactly a year ago, I wrote a post regarding my concerns with SmartFlare, supposedly a novel method for live imaging of RNA in cells.

      In a nutshell, SmartFlare are gold nanoparticles covered in oligos specific to a certain mRNA of interest. Supposedly, cells internalize these particles and, once the mRNA hybridize to the oligo, a complementary fluorecently labeled oligo is being unquenchhed and “flares”, indicating the present of said mRNA.

      You can read about my concerns in that older post, but apparently I wasn’t the only one concerned about their validity.

      Raphaël Lévy from U. of Liverpool (UK) was concerned as well. He endeavored into this open science project to try and answer his concerns. Here ar my comments:


      Raphael’s concerns were that these gold nanoparticles are maintained in endosomes and do not reach the “cytoplasm” where mRNAs reside. Since he assumes mRNAs are abent from endosomes, any fluorescent signal that we detect is an artefact.

      He used several imaging approaches to test his hypothesis: electron microscopy (EM), confocal microscopy, photothermal microscopy
       and immunofluoresence.

      Unfortunately, Raphael did not perform any assesment of mRNA levels, not by RT-PCR, nor FISH. I think that for a paper that is supposed to test the mRNA detection capacity of SmartFlare, one needs to confirm that the mRNA to be detected is present in this particular system, that it responds to the treatment as published and that it does or does not co-localize with the SmartFlare.  This is my main complaint about this paper, but not the only one.

      Here are some of my other concerns/problems with this paper:

      1. The paper starts by EM examination of gold particles uptake by the cells to examine particle size. Next, they use confcal microscopy to detect uptake. They mention that only ~1/4 of the cells uptake the particles. Is this common? Is it a matter of concentration? What is the range of uptake (particles/cell)?  Is it from 1 to 1000? Or do you see some cells with dozens of spots and all others have zero? [One should have the numbers from the EM studies] What would explain that?
      2. Based on the punctate images, they assume that these are endosomes. They then go through a series of imaging experiments to show that:
      • EM images which show the particles surounded by membranes. But I am not sure what I see. My advice – label all components in the image (plasma membrane, endosomes, mitochondria, etc…).
      • co-localization with fluorescent dextran (which, I assume, is taken into endosomes). I’m not sure I understant what that number represents. Is it  % overlap? The test says “rarely overlap” but to my eyes this is clearly not the case. It appears to fairly ovelap.
      • co-localization with transferrin-containing vesicles or lysosomes. Though it is clear that there’s no co-loc with the transferrin, to my eyes there’s high co-loc with lysosomes. Yet the text reads “little overlap”.
      • Anyway, I suggest having a higher mag or zoomed-in image for all, and a clearer expression of co-loc levels.
      • If the authors are right and its not endosomse or lysosomes, then what are these vesicles?  The authors do not suggest any alternative or ways to test that.

      3.  To test if SmartFlare responds to changing mRNA levels, they treated the cells with a drug which, supposedly, induces an increase of VEGF mRNA levels (VEGF is the mRNA tested here).

      • First of all, the authors do not show that VEGF is indeed expressed in these cells (HeLa) under their growth conditions.
      • Second, they do not show that these cells respond to the drug as previously published. They quote 3 papers which showed a range of 2-10 fold increase (none of the cells that were examined were HeLa cells), each paper uses a different concentration of this drug. (Raphael does not mention the concentration and duration of treatment). These are important controls.
      • Third, we know today that not all cells in the culture are equal. Some respond faster, some respond to higher levels than others. Since only 1/4 of the cells internalize SmartFlare (with even or odd distribution?), it is possible that, by chance, most of the cells (how many were imaged?) that Raphael examined are those which responded poorly to the drug.
      • The best way would be to image SmartFlare with FISH in these cells. This wil give you a) expression levels of VEGF; b) is there a correlation between expresion level and SmartFlare fluorescence; and c) enable co-localization analysis of the mRNA and SmartFlare fluorescence (is the mRNA in these vesicles?).

      4. There is another option why there’s no increase in SmartFlare signal upon increase in mRNA levels – the system is already saturated. This is a matter of mathematics: on the one side, how many mRNA molecules/cell (dozens? hundreds? thousands?). On the other hand – how many particles/cell and how many oligos/particle? Do these numbers match?

      • I would do the reciprocal experiment – to reduce (even to zero) the VEGF mRNA level (e.g. by siRNA, knock/out). Then, see if SmartFlare fluorescence is reduced, in correlation to mRNA levels in a particular cell.

      5. Throughout the text, Raphael assumes that mRNA molecules are absent from endosomes. Is that proven? Since there’s a whole field of study that shows mRNAs in exosomes (extra-cellular vesicles derived from the multivesicular body), I can assume that at least some mRNAs are indeed encapsulated in vesicles inside cells. This, of course, needs to be tested on a case by case basis (VEGF in this case).


      A minor point:

      In the introduction (and then the discussion) Raphael mentions in vitro and in situ methods to detect mRNAs, then goes on to say how great it is if we had a system to visualize mRNA in live cells. Well, we do, more than one. We have MS2-like systems that allow live tracking of mRNAs at the single molecule level. There are also GFP-mimic RNA aptamers (Spinach, Broccoli, RNA-Mango) which enable visualization of RNA in live cells. So SmartFlare is only the 3rd option.


      Overall, the only thing that was convincing, to me, was that SmartFlares are encapsulated in some kind of vesicles. I am not concinved that they do not detect mRNA inside.




      We thank Gal for his critical review. We have improved the manuscript accordingly, adding an important control experiment and several discussion points. Version 2 has been submitted and is currently in the type-setting stage.

      In his original review Gal wrote: "Overall, the only thing that was convincing, to me, was that SmartFlares are encapsulated in some kind of vesicles."

      The authors are delighted as this was the key message of the study.


      Point 1: Heterogeneity of uptake

      Regarding cell-to-cell heterogeneity in the SmartFlare experiments, the best imaging technique to characterize heterogeneity is optical microscopy at it offers a larger field of view than electron microscopy. In most studies, including our own, using electron microscopy to extract general conclusions about the distribution of uptake would be mistaken because only a few cells are analysed and each sections represents a small percentage of the cell volume. For this reason, we use electron microscopy only to observe the intracellular localisation.

      The observation of heterogeneous uptake, in our experience of working with nanoparticles, is not common (doi:10.1021/nn9006994 ; 10.1021/nn300868z]. If particles are well dispersed in the medium they should interact with all cells and there should be some uptake in all cells, with some level of cell-to-cell variability due to cell size and cell cycle variations (among others). What we observe here is however more extreme than would normally be expected. This is a robust and reproducible observation - it was observed with all three conditions in all of our experiments. We can also see it in the few publications where SmartFlares were used and the microscopy images are not saturated (for example doi:10.18632/oncotarget.2744).

      One possible explanation is that there is some degree of interparticle association before they interact with the cells so that some cells encounter preformed groups of particles while others do not see much. Obviously this is a little speculative. Providing a solid explanation for this observation is beyond the scope of the study, but the observation itself is very relevant since these massive differences of uptake would certainly bias, for example, FACS results, especially when SmartFlares are used to sort heterogenous populations of cells based on mRNA levels (for example doi:10.1016/j.ymeth.2015.04.022).


      Point 2: Endosomal localisation

      We have decided not to annotate the EM images as suggested. This is mainly for two reasons:

      1. Without the addition of immunogold labelling of compartments for validated markers (say of mitochondria, golgi &c.), we cannot identify a compartment with any certainty, doing so from observation feels disingenuous.
      2. Any overlay on the image will (even to a small extent) obfuscate the data. We therefore are disinclined to add labels which may alter the assessment of the data.

      We have clarified the use of Manders' colocalisation coefficient in the text and Methods. This number is a mean value of all fields studied. As we include only one image in the figure, this single figure may not be represented by the chosen field. We have adapted the wording in the text to clarify.

      We spent considerable time debating whether to show high magnification or large field images. We concluded that high magnification further distorts the perceived colocalisation and would not be as representative. Conversely, low magnificaion images do not capture the key finding of the paper which we feel is the punctate distribution. In any case, all of the raw data are available for further inspection which makes the choice of illustration less critical.

      The compartment in which the SmartFlares reside is indeed a fascinating question. We cite in our paper a study on this very topic from the Mirkin group suggesting that the majority of SNAs reside in Late Endosomes at 24h (doi:10.1021/ja503010a). Given the suspected (although largely unconfirmed) idea that uptake is through caveolae-mediated endocytosis, another realistic option is that SmartFlares reside in Caveosomes, which would not accumulate fluid-phase markers such as dextran (see for example doi:10.1038/35074539 Figure 4). We feel however, that a more thorough investigation of this topic is beyond the scope of this study.


      Point 3: Confirming VEGF mRNA levels

      Whilst we do not measure absolute VEGF RNA or protein levels, we use standard growth conditions and defer to the many published studies as well as the Human Protein Atlas which both strongly support the existance of VEGF RNA and protein in these cells.

      We were remiss in not including the conditions for DMOG treatment. This has been included in the revised version. Furthermore, we now include an additional control where we show by qPCR in three independent replicates a 19 fold increase in VEGF mRNA in the presense of DMOG (using the same cells and under identical growth conditions to those in which the previous experiments were done).

      It should be noted that SmartFlares are not supposed to be imaging the localisation of mRNA but only the cellular mRNA level. This is because the fluorescence increase corresponds (in principle) to an event where the fluorescent probe is released from the particle while the mRNA binds to it. Therefore the detected probe is not bound to the mRNA. We also remark that FISH was not conducted in any of the articles introducing the SmartFlare (Nano-Flare) nor its successor, the Sticky-Flare technology (doi:10.1073/pnas.1510581112). It would have been particularly relevant for the latter as those are indeed supposed to image mRNA localisation; see our (rejected) letter to PNAS published as a PrePrint (doi:10.1101/029447) for more details.


      Point 4: SmartFlare maths

      We agree that it is informative to do some basic number crunching of this sensing problem. A simple attempt is provided below and further details can be found in a related blog post. Here are the key numbers as we see them.

      Based on the concentration of the SmartFlares, we can estimate the number of SmartFlares per cell to be of the order of 75,000. This would correspond to ~ 3,750,000 oligo probes per cell assuming 50 oligo probe per smartflare (doi:10.1021/nl072471q).

      The copy number of any specific mRNA per cell depends on sequence, cell types, signalling events etc, but typically it ranges from a few copies to a few thousands of copies (doi:10.1101/gr.110882.110). Therefore, our estimate above indicates an excess of oligo probes of at least three orders of magnitude over the most abundant mRNA.

      If just 0.1% of these probes would bind their target, it would block 3,750 mRNA presumably resulting in some degree of silencing. Both the manufacturer and Mirkin’s group report that there is no silencing effect in the conditions of these experiments. Therefore, at least 99.9% of the SFs do not bind their target mRNA.

      Seferos et al (doi:10.1021/ja0776529) show that in the absence of release of the probe, fluorescence value of ~30% of the total value after release is observed (doi:10.1021/ja0776529, Figure 1b). This is presumably due to a non-complete quenching of the fluorescence. For the SFs to work, we would therefore have to detect a variation of less than 0.1% over a background of ~30%.


      Point 5: mRNA in the Endosomes/Exosomes

      We recognise the body of literature on mRNA and microRNA in exosomes, however we feel this is irrelevant to the current study for the following reasons:

      • The product is sold as a means to quantify cytosolic mRNA following endosomal escape.
      • As you rightly point out, exosomes are thought to form from invagination of late endosomes to form multivesicular bodies which then fuse with the plasma membrane, releasing the intraluminal vesicles. The topology of this scheme would not expose the mRNA to the lumen of the multivesicular body (the possible location of internalised SmartFlares). In order to work, this would require endosomal escape by SmartFlares, retention of the reporter strands in the cytosol, inclusion into the intraluminal vesicles and the subsequent interaction with the mRNA releasing the reporter strand. This idea employs many more assumptions than are necessary to explain our findings.

      Minor point
      We have edited the introduction to include reference to these technologies.

      2016-03-15 16:05 UTC

      Comment on this review