Peter T. Nelson led this live discussion on 27 February 1998. Readers are invited to submit additional comments by using our Comments form at the bottom of the page.
Paper Under Discussion: Cheetham JE, Coleman PD, Chow N. Isolation of Single Immunohistochemically Identified Whole Neuronal Cell Bodies from Post-Mortem Human Brain for Simultaneous Anlaysis of Multiple Gene Expression. Journal of Neuroscience Methods 77 (1997): 43-48. Abstract
Live discussion held 27 February 1998.
Participants: Peter Nelson, Nienwen Chow, Paul Coleman, Kieran Breen, Huntington Potter.
Note: Transcript has been edited for clarity and accuracy.
Nienwen Chow-Paul Coleman: We are here - Chow and Coleman.
Peter Nelson: Hi there. I'm Pete Nelson. I guess most people haven't arrived yet. Maybe we could wait a minute or two before intros and discussion begins.
Nienwen Chow-Paul Coleman: Fine by us. BTW Pete, we just e-mailed our responses to your comments to June K. She suggested we do it this way, so she could append them to your comments. If you wish to discuss any of the points you raised on-line here, let's do so.
Peter Nelson: Maybe in around two minutes we should start, okay?
Nienwen Chow-Paul Coleman: OK. Questions? Points? Answers?
Peter Nelson: Perhaps we should begin with some introductions. My name is Peter Nelson. I'm an MD/PhD finishing my MD at U.Chicago. I'm interested in AD pathobiology.
Kieran Breen: Hi, I'm Kieran Breen, a lecturer from Dundee in Scotland. Peter Nelson: Hi Dr. Breen. I was very interested in the paper. One question that sprung to mind was if you are going to use this technique to look at gene expression in damaged cells, is it possible that only robust cells will survive the initial cell separation.
Nienwen Chow-Paul Coleman: Hi, I'm Nienwen Chow, a Ph.D.. I am working in Paul Coleman's group.
Peter Nelson: I had the same concern. Certainly, there'll be SOME greater frailty in injured cells, by definition.
Nienwen Chow-Paul Coleman: A number of questions have now been raised. With regard to the issue of frailty in sick cells, we are not sure what aspect of the procedure you are referring to. There are other methods of picking cells that are more recently available, such as laser capture micro dissection, that may get around your concerns.
Peter Nelson: But the point is, you're comparing cells of different healthfulness, in re: their RNA. Just by being unhealthy, certain RNAs might be degraded artifactually in a way that has not actual bearing on AD.
Nienwen Chow-Paul Coleman: There also was a private message about asking about more recent, unpublished results. That's OK too.
Peter Nelson: Well? Have you found anything interesting?
Nienwen Chow-Paul Coleman: Let us first address the issue of cells of different health. Is this not an issue with any method that looks at diseased vs. healthy tissue? We'll wait for an answer to this before going on to more recent results.
Peter Nelson: Well, especially one that so directly and dramatically manipulates the cells before analyzing them for something as fragile as RNA.
Kieran Breen: Of course it is, and it is one of the main problems in this research area. However, a technique such as IHC or ISH looks at the cells as they are, while single cell dissection may enrich for a given sub population, thus giving spurious results. Nienwen Chow-Paul Coleman: True, that is why we are exploring other methods of cell isolation. However, we would also emphasize that for selected messages we also confirm results by in situ hybridization.
Peter Nelson: Does it pretty much check out, in relation to ISH? Are there discrepancies?
Kieran Breen: Another question - do you think that it would be possible to differentiate between sick and ordinary cells following dissection? Do they maintain any sort of morphological changes (e.g. do tangle-bearing cells still have tangles?).
Nienwen Chow-Paul Coleman: For example, our aRNA data suggested no change of NF-M expression in AD, contrary to what has been previously demonstrated using Northern blotting. The result of no change in AD was confirmed with in situ, and we now suggest that the earlier studies showing decreased NF-M expression were a consequence of decreased numbers of neurons.
Nienwen Chow-Paul Coleman: Tangled cells do still have tangles after picking. We emphasize that any cell property that can be revealed by IHC or any other procedure, such as Congo red, etc., that leaves the RNA intact can be applied to this technique.
Peter Nelson: Of course, this last point is of considerable interest....
Kieran Breen: This is good, because even if you enrich slightly for more healthy cells, you will, at the end of the day, be choosing the particular cell in which you are interested and you can confirm the cell by its morphology.
Peter Nelson: ...doesn't mean that it's not changed systematically and artifactually, though, Dr. Breen.
Nienwen Chow-Paul Coleman: With regard to more recent results, we do have a manuscript submitted in which we use the aRNA method, and our picking approach to distinguish pyramidal neurons from hippocampus as a function of disease state. Using a multi variate canonical analysis we are able to distinguish brains as to disease state very nicely.
Peter Nelson: Neat! Using the tangle/non-tangle criteria? Or AD vs. PD vs. control, etc.?
Nienwen Chow-Paul Coleman: It was AD vs. control. We put control in parens because although these were clinically control cases ( one lady living at home, paying bills, etc.) on neuropath exam, they were early Braak stages.
Nienwen Chow-Paul Coleman: Hi Hunt.
Huntington Potter: Hi. This is my first chat so I will probably listen for awhile.
Nienwen Chow-Paul Coleman: Please joins as you see fit.
Peter Nelson: Would it be gauche to ask for a preview of how the AD cases differed from controls in their hippocampal neuron RNAs?
Nienwen Chow-Paul Coleman: Long pause while we consider this.
Huntington Potter: One question is whether you are surprised to see ACT in neurons and if you also checked it in astrocytes.
Nienwen Chow-Paul Coleman: First, I would emphasize that the answers one gets depend on the panel of cDNAs used as probes. Given that, we find that the genes that contribute most heavily to the distinction include cyclin D1, an HSP, ACT, wee1.
Kieran Breen: Sorry, have to run. Thanks for the interesting discussion.
Peter Nelson: Cyclin D1 is a very interesting molecule, viz. Esiri's recent work. Peter Nelson: Bye Dr. Breen!
Nienwen Chow-Paul Coleman: Nice to chat with you. Welcome Jhenry.
Peter Nelson: Did you distinguish tangles from non-tangles in your more recent work? Nienwen Chow-Paul Coleman: Not in what we have submitted. This is ongoing now.
Peter Nelson: Is there anything you'd like to say about that later work?
Nienwen Chow-Paul Coleman: We agree that cyclin D1 is interesting, especially in view of the data of Steve Estus indicating its expression induced by APP. The Eseri work also do tie in here. The question is why is cyclin D1 being expressed.
Peter Nelson: What is the variability, from brain to brain, in how good the neurons are for this technique?
Peter Nelson: Oh, in re: the Esiri work, and the reason for the expression of cyclin D1. Perhaps your technique could be used to find the earliest signs of a diseased state, which causes other things to go haywire secondarily.
Nienwen Chow-Paul Coleman: Statistical analysis shows that the % contribution to variability are: 16% due to cell-to-cell, 4.4% brain to brain and 17.6% to disease.
Peter Nelson: ...And all the cells are definitely hippocampal pyramidal neurons of the same Sommer's Sector?
Nienwen Chow-Paul Coleman: Agree with Nelson last comment - that is one of the aims we have. Huntington Potter: I think that I got lost someplace and had to come back in. One further question I have is whether wee 1 is a common mRNA in the brain and whether it is restricted to neurons or glia.
Nienwen Chow-Paul Coleman: We believe that our cell sample included both CA1 and subiculum pyramids.. It is of note that end stage AD brains all clustered very tightly.
Peter Nelson: Dear Mssrs. Coleman, Chow, and Potter: it has been an interesting discussion, but I must go to a class. Thanks very much.
Huntington Potter: By the way. Hi Pete. Yes I remember you and your sheep eps.
Nienwen Chow-Paul Coleman: With regard to Hunt PotterÕs question, we see wee1 in both AD and control neurons. We have not looked at glia yet.
Peter Nelson: Thanks, Hunt!
Nienwen Chow-Paul Coleman: Good to meet you Peter Nelson.
Peter Nelson: Bye! Thanks!
Huntington Potter: Curious. One thing Carmela and I saw some time ago was that ACT antibodies sometimes stained neuronal cell bodies.
Nienwen Chow-Paul Coleman: Well, I guess that our data confirm that.
Huntington Potter: No I am happy good luck with the new exps. over and out.
Nienwen Chow-Paul Coleman: Goodbye.
By Peter T. Nelson
Cheetham and colleagues describe a novel and ingenious way of isolating brain cells and assaying for multiple RNA transcripts, in which context one wishes to elucidate information about the RNA in a given cell in comparison to the RNA of the cell's neighbors. This technique has special application to neurodegenerative disease. The following discussion includes my thoughts about the methods employed; figure-by-figure analysis of results; and some discussion on the paper and some of the uses of this new technology.
This is primarily a methods paper. The method for isolating and analyzing a single cell's RNA is indicated on the flow chart of Figure 1. The procedure is straightforward, appealingly so. This is also, of course, a preliminary paper, and the authors will in follow-up studies describe the application of the technique in more rigorous detail. At the present stage, however, I am left with some questions:
Question 1) After a postmortem interval of up to 16 hrs, the brain is dissected; 1mm blocks are minced out; cells are physically smeared onto an adhesive surface after trypsin treatment; immunocytochemistry is performed in which huge antibody proteins have access to the cell cytoplasm; and the cells are plucked away with a pipet. Is it not possible, after all of this dramatic manipulation, that RNA could flow out of the one cell and into another, rendering future PCR inaccurate?
Response by Nienwen Chow: During tissue processing, RNA from broken cells would be degraded. That is why acridine orange staining on tissue smear (Figure 2A) does not show a general orange background. After fixation, we believe that RNA are retained within the cells. If RNA leaking from fixed cells is a problem, in situ hybridization and in situ RT-PCR would not be possible.
Question 2) What effect of postmortem interval was noted?
Response by Nienwen Chow: The maximum postmortem interval for our tissues is 16 hours (average is 6 hrs.), and all of the brains we have taken worked. However, it only shows the presence of RNA from tissues with various postmortem delay, but nothing about the quality of RNA. We never studied the effect of postmortem delay on individual RNA species. For information on that, please see the paper by Lukiw et al., (Int. J. Neurosci., 1990, 55:81-88.).
Question 3) What percentage of the brains that were analyzed produced viable RNA (only one was shown)?
Response by Nienwen Chow: 100 percent.
Question 4) How quantitative are these particular dot-blot hybridizations?
Response by Nienwen Chow: If radioactive probes are used and the hybridization intensities are detected by Phosphimager, within a wide range of aRNA abundance, the hybridization signal is linearly related to the concentation of aRNA. In a recently submitted manuscript, we extended the same study shown in Figure to 30 different cDNAs and showed the same results.
Question 5) It would have been interesting to have included a different cell population, e.g. GFAP- immunoreactive astrocytes. Doing so would have demonstrated the flexibility of the technique and allowed a comparison of the two cell populations' RNA.
Figure 2. Acridine orange and NF-H IHC reveal that on the (separate) slide smears are cells that contain RNA and are positively stained for NF-H. One is given the impression that these trypsin-treated cells are very close to one another, and that as they were smeared on the slide they rolled over onto one another. If a (very large) antibody protein could penetrate the cell membrane, couldn't RNAs have transferred too? And if one elected to pluck with a pipet one of the neurons in Figure 2B, might one be in danger of having contamination from the other two nearby neurons? That being said, it is impressive and attractive to have such isolated and characterized cells as are being demonstrated here.
Response by Nienwen Chow: Again, there is a concern about RNA leakage. We have addressed that in Methods 1).
Figure 3. A northern blot of aRNA transcripts from nine cells of one human brain. Some of the transcripts are quite large. One wonders if all the brains had RNA this good...
Response by Nienwen Chow: Yes.
Figure 4. Dot-blot hybridization of aRNA from a single cell using aRNA of different concentrations. This is the crux of the paper. Here they have cDNA clones blotted to nylon membranes being washed in a medium of radiolabelled aRNA from isolated neurons. There are present numerous transcripts in the aRNA; furthermore, the blots demonstrate linear increase in staining intensity for the individual transcripts. Why such a low level of expression? Is the number of actin transcripts in a nerve cell so low in comparison to, say, cyclin D1?
Response by Nienwen Chow: We want to make a correction here. The actin cDNA used in the study is actually an actin-bundling protein clone. A mistake was made by the commercial vender of this cDNA clone. However, it remains a fact that the hybridization intensities for some RNA transcripts with medium abundance are very low. We noted that the intensity has something to do with the sequence and the length of the cDNA used. The efficiency of amplification may also be different for different transcripts, depending on the length of the poly A tail. Other factors may also come into play here.
Table 1. The result of "RNA profiling" from nine neurons of a single control brain with five different cDNA dot-blot reverse probes. It seemed from the table that there were some neurons that had relatively higher values (e.g. cell #'s 1,6,9) and some had relatively low values (e.g. cell #'s 2,3,8), and the basic proportions of one transcript to the other was relatively constant (a1-ACT>GAPDH>cyclin D1=nestin). Is this finding consistent with previously described quantities of these transcripts in nerve cells?
Response by Nienwen Chow: In order to compare data across sasmples, one needs to normalize the values to an internal control. In this case, we did not think any cDNA could serve as a good internal control. In a recently submitted manuscript, we proposed to use the average of all measurements (cDNAs) as an internal control.
Cheetham and colleagues here introduce a new way of analyzing RNA from individual characterized neurons isolated from the human brain. The strength of the experimental technique is, firstly, that it appears to work sometimes (nothing to be taken for granted!). Secondly, it offers the chance to compare "sick" cells to cells nearby that don't appear afflicted. This has special application to diseases such as Alzheimer's, Parkinson's, Pick's, progressive supranuclear palsy, Lewy Body Disease, inclusion body myositis, and others, in which there are obvious manifestations within "sick" cells (various inclusion bodies); e.g. how does a tangle-bearing cell differ from a non-tangle-bearing cell re: mRNA transcripts? Lastly, when you have transcribed aRNA in a test-tube from a single cell, it allows for all sorts of manipulation (e.g. analysis of splicing changes) for that given cell that other techniques, such as in situ hybridization, cannot approach. This technique surely holds great promise--that is the central message of the paper.
It seems premature to discuss this technique with a deeply critical eye, because the present paper does not seem a deep demonstration of it. If the article purports to demonstrate producing RNA from isolated characterized neurons, it has yet to truly describe the quality of the RNA or fully dispel concerns of contamination between neurons. It would be nice to see the RNA produced from many cell types; from many brains; and from the brains of patients with disease. It would be nice to see a more complex description of the quality of RNA that was produced. It would be nice to have an explanation of why only six of 53 neurons produced PCR- quality RNA.
Response by Nienwen Chow: We do not feel that the method needs to be limited to cells with inclusion bodies as is implied by Dr. Nelson. Any characteristic that can be detected by immunohistochemistry may be used to select cells. Detection of cell properties of interest need not be limited to immunohistochemistry, but could encompass any method that retains the integrity of the RNA.
"It would be nice to see the RNA produced from many cell types . . ."
Response by Nienwen Chow: In our ms. recently submitted we limit ourselves to one cell type, but 5 brains representing different disease states.
"It would be nice to have an explanation of why only 6 of 53 neurons produced PCR quality RNA"
Response by Nienwen Chow: This PCR was done with primers for G3PDH and now we use primers for 18s ribosomal RNA and the rate of positives is roughly 50 percent or higher. This seems to imply an issue of sensitivity of PCR rather than the quality of the RNA.
A few pitfalls of the technique come to mind:
Comment:1) RNA degradation seems inevitable. Some messages will be degraded and others won't; hence, any conclusion drawn using this technique may have systematic errors.
Response by Nienwen Chow: True. Answered in part above. In addition, comparison of "diseased" and "normal" cells from the same region of the same brain obviates this issue to some extent.
Comment:2) Let us consider the example of studying the differences between "normal" and "tangle-bearing" cells' RNAs. Let us say that a certain message is lessened in the tangle-bearing cells. Perhaps this difference would only be due to a difference in how the different cells react to postmortem artifact. Even if tangle-bearing cells have an increase in a message, that too could be due to some artifact of the technique. (Parenthetically, it should be added that an animal model of neurofibrillary changes, such as the aged sheep, could come in handy in this situation).
Response by Nienwen Chow: True. This problem also applies to in situ hybridization, immunohistochemistry, etc. Use of tangle and non-tangle neurons from aged sheep is a good suggestion, with the caveat that the mechanism giving rise to NFT in sheep may differ from that producing NFT in humans with AD.
Comment:3) Another critique of the paper is that in situ hybridization allows for quantitation of messages in cells. Dr. Coleman and colleagues have been fruitfully comparing messages from tangle- and non-tangle-bearing cells for years.
That being said, the advantages of having individual cells' RNA for comparison of pathological cells vs. normal cells, and of cells from pathological vs. normal brains, are great, and the implications probably surpass the imagination of this journal club reviewer. The work of Cheetham and colleagues is of great interest to anyone in the field and one may anticipate work of great significance being produced soon from that laboratory.
Response by Nienwen Chow: We are not sure in which way this is a critique. Studying one message at a time with in situ hybridization demonstrated to us dramatically that expression of some messages increases in NFT neurons, and decreases for other messages. It seemed to us that characterizing profiles could be accomplished more efficiently with the aRNA method than with the one message at a time of in situ hybridization. We continue to use both methods and, in fact, for selected messages confirm results by both methods. For example, our aRNA data have suggested that NF-M is unchanged in AD neurons. This was confirmed by in situ hybridization study in our lab. We attribute the previously reported decreases in NF-M expression demonstrated by Northern blots to loss of neurons.
- Isolation of Single Immunohistochemically Identified Whole Neuronal Cell Bodies from Post-Mortem Human Brain for Simultaneous Anlaysis of Multiple Gene Expression
- Cheetham JE, Coleman PD, Chow N. Isolation of single immunohistochemically identified whole neuronal cell bodies from post-mortem human brain for simultaneous analysis of multiple gene expression. J Neurosci Methods. 1997 Nov 7;77(1):43-8. PubMed.
- Hardy J, Israël A. Alzheimer's disease. In search of gamma-secretase. Nature. 1999 Apr 8;398(6727):466-7. PubMed.
- Wolfe MS, Xia W, Ostaszewski BL, Diehl TS, Kimberly WT, Selkoe DJ. Two transmembrane aspartates in presenilin-1 required for presenilin endoproteolysis and gamma-secretase activity. Nature. 1999 Apr 8;398(6727):513-7. PubMed.
- De Strooper B, Annaert W, Cupers P, Saftig P, Craessaerts K, Mumm JS, Schroeter EH, Schrijvers V, Wolfe MS, Ray WJ, Goate A, Kopan R. A presenilin-1-dependent gamma-secretase-like protease mediates release of Notch intracellular domain. Nature. 1999 Apr 8;398(6727):518-22. PubMed.
- Struhl G, Greenwald I. Presenilin is required for activity and nuclear access of Notch in Drosophila. Nature. 1999 Apr 8;398(6727):522-5. PubMed.
- Ye Y, Lukinova N, Fortini ME. Neurogenic phenotypes and altered Notch processing in Drosophila Presenilin mutants. Nature. 1999 Apr 8;398(6727):525-9. PubMed.