A simple proposal to make STAP protocols clearer
By Jun Seita
Detection of a GFP signal driven by the Oct3/4 promoter was extensively used to define STAP cells in the original STAP papers [1, 2] and similar approaches have been employed by many third-party attempts to reproduce STAP cells. On the other hand, it is well known that stressed and/or damaged cells harbor significant auto-fluorescence. This is particularly pertinent to the case of STAP cells, because the various stresses applied to cells may themselves induce cellular autofluorescence independent of Oct3/4-GFP reporter activation. Neither flow cytometry nor a fluorescent microscope can distinguish the origin of an observed fluorescent signal, e.g. GFP. Rather such equipment only measures the intensity of a photon within a certain range of wavelengths, as determined by light source and filters used. Distinguishing an authentic Oct3/4-GFP signal from auto-fluorescence is the critical first step to reproduce STAP cells. Thus, just following the original and updated protocol [3] provided by the authors is insufficient to lend confidence to the claim that low-pH treatment converts terminally differentiated cells into Oct3/4-GFP expressing cells.
As a stem cell biologist who totally relies on flow-cytometry technology, I’d like to propose one major addition to the protocol.
Please process cells isolated from wild-type mice in parallel with cells from Oct3/4-GFP transgenic mice all the way through the experiment and analyze them exactly side-by-side.
All other factors should be identically matched, namely background strain; age; gender of mice; low-pH treatment; and the subsequent culture process. This is the only negative control that can definitely distinguish an authentic Oct3/4-GFP signal from auto-fluorescence. This strategy named Fluorescence-minus-one (FMO) control is the most reliable way to define the boundary between positive and negative cells in modern flow-cytometry [4]. Non-stress-treated but cultured cells are not apropos for this purpose because the spectrum and intensity of auto-fluorescence varies based on the condition of the cells. Moreover, this is an addition, not a modification of the protocol, and thus the body of the original protocol remains intact.
The flow-cytometry data presented in Extended Data Figure 5g of the Nature letter [2] suggest that some controls relating to flow-cytometry might need to be revisited throughout the STAP studies and it might be worth determining other factors that need to be matched. Each sample should be analyzed using the exact same acquisition parameters on the flow-cytometer. The voltage setting of the photomultiplier tubes (which define acquisition sensitivity) and compensation parameters should be optimized and fixed before commencing analysis, and should not be adjusted between samples. Also the authors should declare what fluorochrome/fluorescent protein was used in each figure. Referring to the protocols [1-3] provided for what antibodies were employed in the original studies, there are 29 options for anti-CD90 from eBioscience, 22 options for anti-CD19 from Abcam, 4 options for anti-CD34 from Abcam, and 6 options for anti-Integrin alpha 7 from R&D Systems. The ambiguity over what exact antibodies were precisely used makes it challenging to reproduce the authors’ exact procedures. Also since many fluorochromes/fluorescent proteins have emission spectra that are wider than the filters for a particular detector, one always needs to deal with spillover of fluorescence into adjacent detectors by a method known as “compensation”. Perfectly compensating for spillover for certain complex combinations of distinct fluorescent emissions can be quite challenging, e.g. simultaneous analysis of GFP, PE and auto-fluorescence. Thus it is very important to know what combination of fluorochromes/fluorescent proteins were used to interpret the presented flow-cytometry plot in the figure. If one is not familiar with voltage setting, compensation, choice of combination of fluorochromes, or FMO, it is highly recommended to consult with flow-cytometry experts in your flow-cytometry core facility.
The addition of these simple, yet critically important steps and controls will produce far clearer data.
Author. Jun Seita, M.D., Ph.D.
Institute for Stem Cell Biology and Regenerative Medicine, Stanford University School of Medicine
References
1. Stimulus-triggered fate conversion of somatic cells into pluripotency.
Obokata H, Wakayama T, Sasai Y, Kojima K, Vacanti MP, Niwa H, Yamato M, Vacanti CA.
Nature. 2014 Jan 30;505(7485):641-7. doi: 10.1038/nature12968.
PMID: 24476887
2. Bidirectional developmental potential in reprogrammed cells with acquired pluripotency.
Obokata H, Sasai Y, Niwa H, Kadota M, Andrabi M, Takata N, Tokoro M, Terashita Y, Yonemura S, Vacanti CA, Wakayama T.
Nature. 2014 Jan 30;505(7485):676-80. doi: 10.1038/nature12969.
PMID: 24476891
3. Essential technical tips for STAP cell conversion culture from somatic cells
Haruko Obokata, Yoshiki Sasai & Hitoshi Niwa
Protocol Exchange (2014) doi:10.1038/protex.2014.008
Published online 5 March 2014
4. Modern flow cytometry: a practical approach.
Tung JW, Heydari K, Tirouvanziam R, Sahaf B, Parks DR, Herzenberg LA, Herzenberg LA.
Clin Lab Med. 2007 Sep;27(3):453-68, v. Review.
PMID: 17658402
Seems a healthy sign to observe constructive suggestions to any breakthrough type of findings by fellow scientists and “STAP” researchers need all kinds of advice to fill or excel the accuracy in data and fasten the process of reproducibility test. Though I am not able to follow these scientific discussion, it would be nice if Dr. Seita clarify the following point?
Haven’t the authors concluded that the low acid stress generated “STAP” cells in combination of the “Detection of a GFP signal driven by the Oct3/4 promoter” and the development of various organ tissues with the implant of the cultivated “STAP” cells into mouse, probably more emphasis on the latter?
I agree that all the experiments require rigorous control, but am not sure if your suggestion is critical for the study in question. Extended Fig. 5g of the Nature letter is not even about STAP cells. The authors also used live fluorescence imaging, immunostaining, and qPCR to detect Oct4 after the acid treatment. Don’t see much merit on questioning about it. I saw a couple of anonymous posts reporting Oct4 expression after an acid treatment by Western. Oct4 is upregulated in T cells by cytokines, too (PMID: 23207483). I guess in vitro results of Obokata et al.’s are in line of this and are not a big surprise.
While I agree that an FMO is the best control here, couldn’t auto fluorescence be accounted for by looking on other channels in the original dataset (IE. a “dump channel”)? In other words, cells that are auto fluorescent would be brighter than the mean on pretty much every channel whereas cells that are actually GFP positive should have the same intensity as any other cell on say an FL4 channel where there will be no meaningful spectral bleed-through from a GFP signal.
There are two possibilities: autofluorescence during cell death and non-specific gene expression during the unusual metabolic environment (which can accidentally drive Oct4-GFP expression).
Some autofluorescence can be distinguished from GFP signal when excitation wavelength is properly controlled. Most sources of metabolic autofluorescence (FAD, NADH, heme) are excited at shorter wavelength than GFP, although emission spectrum can overlap, or have a distinct double peaked spectrum. Autofluorescence from denatured protein will have a broader spectral distribution but is expected to be weaker.
One way to get away from these issues is to put the Oct3/4 promoter to the popular pGL vector and do the same thing with bioluminescence instead of fluorescence. This alternative approach is so simple that I am sure people in the field have done it and I won’t be surprised if there are Oct3/4-luc transgenic lines.
The second question, to which Jun did not address, is whether or not the potential non-specific gene expression under the low pH condition accidentally drives Oct4-GFP expression. The HCl treatment can disrupt metabolic conditions inside the cell and lead to uncontrolled gene expression. I think the insertion is not targeted to a specific locus in the experiment and it is possible that GFP expression is a consequence of this random transcription. A simple experiment to test this is to use some other GFP expressing strains not driven by Oct4 promoter and see if increased GFP expression can be seen.
Forgive my ignorant nitpicking, but as you know the “intensity of a photon” is by definition unity, it being the indivisible quantum of light….
Thanks. Does “intensity of photons” make sense?
Light is, of course, a bunch of photons. But the intensity of light is the total number of photons so we have a problem with different dimensions (units). A single photon carries energy, not intensity, and the energy is related only to its wavelength.
As the previous poster said, “intensity of light” or “number of photons”. Some people say “intensity of photons” which is kind of inaccurate but it might pass. I guess in biological fluorescence microscopy, the photomultipliers are used either in “counting mode” (if the number of photons incident per unit time is low enough so that the electronics can distinguish individual “pings” corresponding to photon hits), or far more pervasively in “analog mode” if the rate is so high that individual photons can no longer be resolved, and there is simply a analog-to-digital converter that measures the continuous current per detector channel. I can see that this might be the reason why you might for technical reasons use the word “intensity” instead of “number”. In your Japanese translation, a photon doesn’t have a wavelength range; each photon has a specific wavelength, whereas a collection of photons (emitted from your cells?) might be distributed over a range of different wavelengths. If some of the photons have wavelengths that are either too blue or too red, they will be detected by neighboring photomultiplier channels that are equipped with bandpass filters that should supposedly observe a different fluorescence marker (i.e., the spurious “spill over” effect that you mention). All these problems might be averted by simply using light 光 instead of photons. Sorry for the nitpicking.
This is interesting, thanks. Dr. Seita provided references for everything except the autofluorescence under stress posit. I was not aware this is a serious issue when analyzing stressed or manipulated cells so I was interested in knowing more. I did a quick pubmed search and came across these two references which both indicate that stress (predominantly oxidative, which I would assume is conferred by low pH environment) decreases autofluorescence.
1. http://www.ncbi.nlm.nih.gov/pubmed/11770840
2. http://www.ncbi.nlm.nih.gov/pubmed/15270590
Am I missing/misinterpreting something? Could you please provide a source to explain how stress induces autofluorescence? That seems to a be a critical point in this argument and would be interesting to learn more about this.
best,
Stacven
As I stated in the text, the spectrum and intensity of auto-fluorescence varies based on the condition of the cells. Thus you cannot guess from other conditions.
I suppose I should clarify what I was asking. To assert something to the effect of.. “it is well known that stressed and/or damaged cells harbor significant auto-fluorescence.” indicates strongly positive that stress induces autofluorescence (ie. increase in FITC signal in stressed vs. non stressed cells). Thus, the assertion is not subjective but rather based on some evidence. Can you provide any evidence that this is the case? Is there a scenario in which stress induces autofluorescence? Because I could not find one and it would be interesting to read evidence to support your argument.