Schizophrenia Research Forum - A Catalyst for Creative Thinking

Live Discussion: Gene Expression Profiling of Postmortem Human Brain


Akira Sawa

Marquis Vawter

The venerable medical tradition of postmortem research would seem to be poised for a quantum leap in this era of molecular science, but old provisos and emergent issues must clearly be taken into account.

On 9 January 2008, Akira Sawa of Johns Hopkins University and Marquis Vawter of the University of California, Irvine, led an SRF online discussion of the fine art of generating and interpreting reliable gene expression data in psychiatric research. We invite you to read their background text below, along with an extensive comment by Vawter on several recent papers, particularly one in the Journal of Neuroscience Methods by Christine Miller and colleagues at Johns Hopkins University and the Stanley Medical Research Institute. We then ask that you share your ideas, warnings, horror stories, and other useful information in the form of a comment.

View Transcript of Live Discussion — Posted 12 March 2008

View Comments By:
Karoly Mirnics, Fuller Torrey — Posted 30 December 2007
Christine Miller — Posted 7 January 2008
Alberto Arregui — Posted 7 January 2008
Beth Thomas — Posted 7 January 2008
Sherry Leonard — Posted 7 January 2008
Kenji Hashimoto — Posted 7 January 2008
Tadafumi Kato, Kazuya Iwamoto — Posted 8 January 2008
Sinthuja Sivagnanasundaram, Duncan Sinclair, Cynthia Shannon Weickert — Posted 8 January 2008
Christine Miller — Posted 9 January 2008
Paul Harrison — Posted 9 January 2008
Christine Miller — Posted 9 January 2008
Barbara K. Lipska, Joel Kleinman — Posted 9 January 2008
Sinthuja Sivagnanasundaram — Posted 17 January 2008


Background Text
By Akira Sawa and Marquis Vawter

The purpose of this forum is to discuss factors that impact gene expression profiles in postmortem human brain in the context of microarray, quantitative PCR, in-situ hybridization, and other expression assessment methods.

The pathophysiology of several psychiatric disorders has been extensively studied using postmortem human brains. Enormous molecular information has been obtained from autopsied brains from schizophrenia patients through microarray and proteomic approaches. However, there are caveats to the use of this tissue as well: schizophrenia appears to be a disorder of neurodevelopmental origin; thus, major pathological events that affect neurodevelopment may be compensated for at later stages, such that they would not be seen or detected in autopsied brains. In addition, mechanisms and processes of brain maturation may be different between patients with schizophrenia and controls. More practically, confounding factors such as medication, smoking, and diet must be considered in studying autopsied brains. Gene expression profiles are affected by RNA quality, fixation, agonal factors, postmortem interval, as well as freeze-thaw effects of autopsied brains.

Thus, we plan to discuss methodological approaches for quality control in postmortem brain gene expression studies, as well as data interpretation, alternative approaches, and subject selection or matching. Our discussion will mainly focus on studies of gene expression profiles, since whole genome studies have been conducted. Similar strategies may also refer to studies targeting protein and peptide profiling.

Some provisional questions for discussion will be

1. What are the reliable parameter(s) for quality control of gene expression profiling with postmortem brains?

2. What are the appropriate strategy(ies) in data analysis and interpretation, considering possible confounding factors?

3. How can we address molecular changes associated with neurodevelopment in the autopsied brains? How can we address the question of possible differences in brain maturation and aging between patients and controls?

4. What types of tissues are to be considered as alternatives to autopsied brains?

5. What types of brain banks or collections are to be developed in the research community?

6. What kinds of data sharing systems are expected to promote research in the field?

Read Mark Vawter’s comment on several recent postmortem methods papers.


Transcript

Attendees/Participants

Tony Altar
Simge Aykan, Radboud University Netherlands
Jesper Brohede, Karolinska Intitutet, Sweden
Murray Cairns, Schizophrenia Research Institute, Sydney
Vibeke Catts, Brisbane, Australia
Doug Frost, University of Maryland
Carla Gallo, Universidad Cayetano Heredia, Lima, Peru
Sharon Graw, University of Colorado
Paul Harrison, University of Oxford
Vahram Haroutunian, Bronx VA Medical Center
Andreas Jeromin, Allen Institute for Brain Science, Seattle, WA
Tadafumi Kato, RIKEN Brain Science Institute
Joel E. Kleinman, NIMH
Tim Klempan, Douglas Hospital, Montreal
Barbara Lipska, NIMH
Christine Miller, Johns Hopkins University
Karoly Mirnics, Vanderbilt University
Paul Pavlidis, University of British Columbia in Vancouver
Silvana Sessarego, Universidad Peruana Cayetano Heredia, Lima, Peru
Akira Sawa, Johns Hopkins University
Cyndi Shannon Weickert, Schizophrenia Research Ins., University of New South Wales and POWMRI in Sydney Australia
Elaine Shen, Allen Institute for Brain Science
Sinthuja Sivagnanasundaram, Schizophrenia Research Laboratory, Australia
Beth Thomas, The Scripps Research Institute
Peter Thompson, University of Texas Health Science Center San Antonio
Hiroaki Tomita, Tohoku University
Mark Vawter, University of California, Irvine

Note: The transcript has been edited for clarity and accuracy.


Hakon Heimer
I'd like to introduce Akira Sawa of Johns Hopkins University and Mark Vawter of the University of California, Irvine, and I thank them for bringing us this idea and for their work to prepare the discussion.

Akira Sawa
Oh, so many people...wonderful! Hello and thanks for coming. We would like to structure this discussion into 20-minute segments. First, Mark will lead the session focusing on the importance of a gene expression profile database from control brains in which various confounding factors, such as PMI, and analytical methods will be discussed. In the second 20 minutes, I will lead the question of how neurodevelopmental disturbances underlying the pathology of schizophrenia can be studied in postmortem brains. We may need to discuss gene expression through development and aging. We may also need to discuss alternative approaches that can be combined with studies with postmortem brains. For the last 20 minutes, we wish to hear your opinions of which topics we need to discuss today. Do you have any ideas?

Mark Vawter
We can get started with ideas on controlling for gene expression confounds. We would like to discuss developing a database for postmortem gene expression using control subjects from different brain banks. (This addresses comments for ways to control artifacts and analysis; learn which genes are different due to confounds such as age, gender, PMI, pH, RNA quality. etc.) Our first question is, How does one evaluate gene expression confounds that occur in postmortem brain?

Karoly Mirnics
1) Clinical history, 2) pH, 3) 3'5' integrity, 4) yield/block would be pretty logical. Can we come up with something else that is reasonably objective?

Joel Kleinman/Barbara Lipska
Get large Ns and as much information as humanly possible!

Mark Vawter
Joel/Barbara, yes, “humanly” possible.

Paul Harrison
As always, Karoly's suggestions are spot on, as are Joel/Barbara's: we know basically which are the main factors (e.g., pH/RIN; see my comments on website), and now we have big enough brain series to finally nail the matter statistically. Of course, how we choose to interpret the results won't be so easy!

Elizabeth Thomas
Adding to Karoly's list: antipsychotic drug exposure.

Karoly Mirnics
Beth, do you mean tissue levels of antipsychotics or just history?

Elizabeth Thomas
Karoly, if tissue levels are available, that would be great. Questions: what do we think of recorded history? Do we believe them?

Vahram Haroutunian
Depends on the history. If history is derived from real medical or hospital records then okay; if it comes from self or informant reports, then not so good.

Christine Miller
Agreeing that pH can have (likely an indirect) effect on microarray output, and should be included in all analyses for confounding effect. What is your view about using results from low pH samples for transcripts that are resistant to pH, as long as the pHs are matched cases and controls?

Joel Kleinman/Barbara Lipska
No matching for pH is recommended for reasons that we have posted (see posted comments).

Paul Harrison
Christine, absolutely. If pH (or any other factor) doesn't happen to affect the transcript (or other variable) concerned, it seems unnecessary to omit the data.

Karoly Mirnics
To Christine Miller, no matching can replace quality samples—GIGO (garbage in, garbage out).

Elizabeth Thomas
Is there any evidence that pH changes with aging? Joel's comment had suggested this....

Paul Harrison
Beth, some studies find weak inverse correlations of pH with age. I guess it's probably mainly because old people more often die prolonged (agonal, hypoxic/ischemic) deaths.

Joel Kleinman/Barbara Lipska
Yes, age matters for pH and for gene expression. By the way, we are about to make public expression data from >200 subjects across life span from DLPFC (normals), including cause of death, PMI, RIN, age, etc. You will be able to judge for yourself.

Sinthuja Sivagnanasundaram
Elizabeth, have a look at Rae et al., 2003—pH and aging.

Sherry Leonard
What about cause of death? Has anyone looked at correlation with pH or gene expression profiles?

Mark Vawter
Sherry, we have done that for rapid vs. non-rapid death, and it certainly gives different profiles. But we don't have sufficient N for each cause of death.

Joel Kleinman/Barbara Lipska
Sherry, you will be able to use our database to look at this.

Vahram Haroutunian
Sherry, cause of death vis-à-vis agonal state is clearly one of the most critical variables affecting RNA integrity. There are a bunch of data on this (Stan et al., 2006; Johnston et al., 1997; Harrison et al., 1995).

Paul Harrison
Sherry, we found weak effects of mode of death (rated on a four-point scale of suddenness) with pH, admittedly on a small sample by current standards, but not predictive enough in our view to be of great value. It needs looking at properly, though.

Cyndi Shannon Weickert
Mark, are you talking about evaluating confounds beyond the typical ways of "matching," statistical regression, covariance, etc.?

Mark Vawter
Cyndi, the statistical analysis methods of a large database is open for discussion, but first having critical information in a large database to analyze would be a great first step.

Joel Kleinman/Barbara Lipska
First, I (Joel) have big problems with "matching"! We really have too little information to be able to match. This is why it is important to get individual data and not "matched pair" data.

Vahram Haroutunian
I agree with Joel regarding matching as a general approach. I think it is too easy for us to think that we know all of the relevant variables for matching.

Mark Vawter
Barbara/Joel, okay, if not matching, then putting a large number of controls unmatched so we can evaluate the things Paul Harrison referred to that we agree matter.

Karoly Mirnics
Joel, no matching for pH may have a rationale, but there must be some limit to what we are willing to accept, especially if the other parameters start looking bad.

Cyndi Shannon Weickert
I agree with Joel and Barbara that there is no way to match people on every possible variable, but matching for things that we know impact RNA, like pH and age, does not seem like a bad place to start to me, especially if one does not solely depend on matched pair statistics to interpret the data. In other words, the reasoning that since we cannot match them on everything means that we should not attempt to match the groups on anything does not make that much sense to me.

Mark Vawter
Karoly, there were multiple questions regarding the variables you mentioned earlier, and some had suggested that controls could be put together into a large database to establish some ways to analyze for these variables.

Karoly Mirnics
All, I would like to discuss a creation of a standard series of some 50+50 male female control brains by decade of life, and we would distribute this to everyone as a reference?

Vahram Haroutunian
Mark, given the lack of consensus or practice in how "controlness" is defined, I don't think that this is a great idea, unless the database becomes so huge that differences are washed out.

Cyndi Shannon Weickert
Karoly, if the database of 50 males and 50 females across the decades of life is created, should it be created fresh? And if so, what platform should be used?

Karoly Mirnics
Cindi, yes, I would do it from scratch, with the Exon 1 Affymetrix arrays.

Paul Pavlidis
Hi, in response to the idea about building a database: my group does bioinformatics, and we're working on tools that are designed to facilitate coordinated analysis across labs. Our system is called Gemma. The way it would work is a data sharing consortium would be created in Gemma allowing secure access to consortium members. Participating groups would submit their data from control samples along with the various parameters (age, drug status, etc). The data can then be jointly analyzed in various ways. I would love to hear how people react to the idea.

Mark Vawter
Paul, okay, great news for topic 1, control database. Don't we have enough microarray chips done on control to send these to a central database?

Paul Pavlidis
Right, the idea is to use what people already have on hand.

Mark Vawter
Harry, Joel, Paul, Karoly, Sherry, and others, couldn't we pool data such as Paul Pavlidis suggests?

Joel Kleinman/Barbara Lipska
Mark, how do you pool data from different platforms??

Mark Vawter
Joel, good question; it is not a trivial exercise. Cross-platform data could agree on large effects with large N.

Karoly Mirnics
Mark, the problem of the data across sites is a huge issue; you have experienced this across the three Pritzker sites.

Mark Vawter
Karoly, yes, three sites is a huge variable. The Stanley dataset also had 9-11 sites. The meta-analysis showed greater strength than the individual studies.

Paul Pavlidis
Joel, we use the sequences of the probes to match them across data sets. It is true that when there is "disagreement," this can be a cause. But this can be addressed (in part) using the sequences. And I do not advocate actually combining the data sets, rather, comparing the results.

Elizabeth Thomas
Is the goal of these "control" database(s) to have a standard reference for comparing samples to those from schizophrenics?

Paul Pavlidis
Elizabeth, I think the idea we have right now is simply to address confounds. Is there a common "pH signature," for example?

Mark Vawter
Paul Pavlidis, I would suggest that to start the database, you could use the SMRI controls; there are 12 microarray experiments on about 50 controls. The Harvard brain bank also has public data, and GEO is a source of control brain expression. Our group put in 100 cel files awhile back. Pritzker consortium also dropped in 1,200 cel files to GEO.

Paul Pavlidis
Mark, yes, we have actually loaded all of the available Harvard and GEO data. And we will be downloading the SMRI data “any day now.” So we can indeed begin.

Vahram Haroutunian
Mark, my concern with pooling data from multiple sources is that although the groups all have good internally consistent criteria for cases and controls, I am not convinced that these criteria are externally consistent.

Paul Harrison
For pooling across sites, I think a key step would be QC to check comparability of measures between labs; e.g., send chunks of the same brain to different labs for assessment of yield/RIN/housekeeping genes, etc.

Sherry Leonard
For any pooled database, we could certainly provide information on smoking effects on gene expression. Data on effects of neuroleptics would also be needed.

Christine Miller
Another consideration in combining sites is the differences in SNP frequency (stratification), which obviously have gene expression implications that can be seen within the (admittedly small) Stanley collection based on geographic location of the samples collected.

Elaine Shen
Question regarding the databases and samples: how do you control for anatomic region across samples, experiments, laboratories?

Cyndi Shannon Weickert
Elaine, one way to begin to approach consistency in dissections across laboratories would be to provide a map and a photograph of the brain blocks before and after dissection.

Karoly Mirnics
Paul P, probe regions may mean different splice variants.

Joel Kleinman/Barbara Lipska
To all, it occurred to us during analyses of our expression data from microarrays that we also need to think about multiple transcript issue; one transcript may be affected in a very different way (by age, etc.) than other transcripts.

Paul Pavlidis
Karoly, we can track how each probe maps to known transcripts. This is indeed complicated by unknown transcript variety abundance, and also that there can be unknown transcripts.

Karoly Mirnics
Paul P, you may not have enough bioinformatics information to know the real alternate splicing that occurs in the tissue.

Paul Pavlidis
Karoly, yes, this is what I meant by “unknown transcripts”: there are unknowns. This means that disagreements among data sets could be difficult to address. But I think we can accomplish a good deal.

Karoly Mirnics
Paul, Mark, I am still worried about the patchwork of lumping all together. Previous experience shows that it does not work the best.

Paul Harrison
Karoly, I agree: unknown and uncharacterized splice variants (including multiple promoter usage and brain-specific isoforms) are a major issue.

Mark Vawter
Karoly, Paul H., with the sequence for each probe known on the commercial platforms, I think it would be possible to determine which part of the gene is interrogated. We won't know the splice variants unless we specifically try to measure each one. Is there an array that will do this?

Paul Pavlidis
Karoly, just one more comment. I don't advocate lumping the data together, but facilitating comparisons and meta-analyses can be done without that.

Karoly Mirnics
Paul P., you will still be limited with the small power of each data set (e.g., we are still screwed. ;-( )

Paul Pavlidis
Karoly, the reason to push for a meta-analysis is precisely to address the weakness of any individual data set.

Vahram Haroutunian
Agree with Karoly and Paul P. on both sets of last comments. Comparison of different control groups, especially with known demographics such as race, sex, and age would be very informative. Yes, I think the ideas here need to be flushed out much more.

Paul Pavlidis
Karoly, I agree that it would be ideal to have a single gigantic control data set, but even then “independent” confirmation by smaller studies would be beneficial. Anyway, I agree we need to discuss this more elsewhere.

Akira Sawa
Dear all, it sounds like we have substantial consensus in the need to build a control brain database in which many confounding factors are addressed. Regarding each factor, it seems we still have debates (it would be exciting to spend another 2-3 hours, but we have only 35 minutes left). Mark, how about summarizing the discussion of these topics into what kind of common control database may be possibly built?

Joel Kleinman/Barbara Lipska
I am not sure that we really have any consensus on building the control database. Before anyone can reach any consensus there has to be a well-defined question.

Mark Vawter
Joel, yes, it may need some time to develop, given the feedback here today.

Karoly Mirnics
Organizers, this is a good first meeting, but we will need more time on each of these topics if we are to arrive somewhere. Would it make sense to have a follow-up, with a bunch of concrete suggestions on how to do it, and then move toward specifics?

Sherry Leonard
Could we have a meeting at Biological Psychiatry?

Joel Kleinman/Barbara Lipska
Sherry, yes, we should.

Mark Vawter
Sherry, yes a meeting in person at SOBP.

Cyndi Shannon Weickert
I would like to meet face-to-face at SOBP; it beats getting up at 4 a.m.!

Mark Vawter
Paul Pavlidis, thanks for the suggestion. I hope you can join us at SOBP.

Paul Pavlidis
Just checked my calendar: looks good. :)

Joel Kleinman/Barbara Lipska
Barbara and I are going to release our database in a user-friendly format. I don't know what else you want us to say or do.

Karoly Mirnics
Joel, we are salivating.

Joel Kleinman/Barbara Lipska
Karoly, hungry?

Karoly Mirnics
YUUUUUUUUPPPPP. Yummmy data!

Akira Sawa
Joel and Barbara, great! Thank you!

Mark Vawter
Joel/Barbara, thanks much.

Akira Sawa
Dear all, the first half has been very productive in exchanging our ideas on this important topic openly. This is just an important start, not to get a superficial conclusion now. How about moving to the next topic? In the second 20 minutes, I will lead the discussion of how neurodevelopmental disturbances underlying the pathology of schizophrenia can be studied in postmortem brains. We may need to discuss gene expression through development and aging. We may also need to discuss alternative approaches that can be combined with studies with postmortem brains.

Mark Vawter
Paul P., when Joel/Barbara’s data are available, that will be a tremendous resource, but Joel, please tell us more.

Akira Sawa
Joel and Barbara, as far as I know, your group has obtained very interesting expression data on genetic susceptibility factors for schizophrenia during development and aging?

Joel Kleinman/Barbara Lipska
Akira, developmental expression data are important and they will be accompanied by extensive genotyping data.

Mark Vawter
Joel, what data would you make available?

Joel Kleinman/Barbara Lipska
Barbara won't let me type anymore today. From Barbara now: just wait a few more weeks/months.

Sherry Leonard
Joel/Barbara, will you have cause of death expression data, not just agonal state rank?

Joel Kleinman/Barbara Lipska
Sherry, yes, all detailed information.

Tadafumi Kato
Akira, we don't know the exact brain region causative for schizophrenia and bipolar. How can we identify the region to examine? Most of the data in the Stanley database deal with frontal cortex. However, is it the crucial region? I think it is impossible to find the region by postmortem brain studies only.

Karoly Mirnics
Akira, Kato, I strongly believe that SCH and BP are systemic brain disorders, and we will find some different and some similar disturbances across virtually all brain regions in schizophrenia.

Joel Kleinman/Barbara Lipska
Our focus has been DLPFC. We will do hippocampus next for all of our normals and 200 patients with schizophrenia, bipolar disorder, and MDD.

Cyndi Shannon Weickert
I think that there is now ample evidence to suggest that the molecular pathology in schizophrenia may be cortex-wide; that gives us many regions of the brain to examine. Should there be a focus beyond DLPFC?

Sherry Leonard
Cyndi, I feel that the hippocampus needs more study in schizophrenia, perhaps as subregions.

Cyndi Shannon Weickert
One thing that is important to understanding how schizophrenia develops—beyond just looking at how individual SNPs or haplotypes impact brain development—is to build a mechanistic understanding of the risk genes so that genetic polymorphisms that may impact development of the normal human brain can be grouped more conceptually. Also, I would like to point out that other groups have microarray gene expression across human brain development and that Maree Webster and I had a poster at Davos in 2006 on the gene expression changes in the lead genetic candidates. Of the 13 genes examined, only dysbindin and RGS4 show a protracted increase in expression during development and take a decade to reach adult levels. The majority of schizophrenia susceptibility genes show a decrease in expression with increasing age. The high expression of these genes in the first postnatal year suggests that they may be playing a prominent role in brain development. We hope to make this data public as well.

Mark Vawter
Joel/Barbara, aren't you working on SNPs vs. gene expression vs. development?

Joel Kleinman/Barbara Lipska
Mark, yes.

Vahram Haroutunian
I have not been able to clearly identify for myself how gene expression data during development should/can be interpreted in the context of changing numbers, sizes, and types of neurons, glia, etc. Do we know what the relevant expression denominators are? Are conventional denominators used in arrays applicable to brains that are undergoing such huge changes, especially during the first couple of decades, let alone fetal tissue?

Mark Vawter
Harry, what about looking for genes that are turned on or off during development?

Joel Kleinman/Barbara Lipska
Harry, We agree that relative to what age group or standard is an issue. We have created a standard reference with multiple brain regions and ages to compare and contrast any individual sample.

Akira Sawa
Harry, if we are interested in information on cell types, laser-captured microdissection (LCM) may be useful, which we frequently use to address this question. Do you agree with me?

Vahram Haroutunian
Akira, yes, I agree about LCM, but it is still a question for me whether "housekeeping" genes in an adult neuron are "housekeeping" in the same way in fetal. We know that this is not the case for many genes.

Joel Kleinman/Barbara Lipska
Harry, we do not use housekeeping genes for comparing in our developmental array studies.

Paul Harrison
Having Joel/Barbara's goldmine of developmental data will be great; it will no doubt show lots of interesting genes with interesting ontogenic trajectories of various kinds, but to interpret clearly (and to prove or disprove anything vis-à-vis schizophrenia), we will need complementary approaches (as well as the SNP data).

Tadafumi Kato
Mark, to know the consequences of development, study of DNA methylation will provide useful information in addition to gene expression.

Mark Vawter
Tadafumi, yes, the same DNA provides different messages in different brain regions, so studying methylation by region by gene will be informative, yet challenging. Are there any high-throughput methods for DNA methylation assays?

Tadafumi Kato
Mark, we are developing and other groups have already published microarray-based methylation analysis.

Cyndi Shannon Weickert
Tadafumi, I agree, and people are starting to look at this (Akbarian and others), so finding a way to add other types of biological information on a large data set would be good (as a second step), like uploading DNA methylation or proteomics data, but how would all this be managed? It seems to me that something like this (back to the database idea) would have important implications for many aspects of medicine beyond schizophrenia.

Tadafumi Kato
Cyndi, the only way to have meaningful information is to integrate everything: expression, SNP, CNV, and methylation.

Mark Vawter
Tadafumi, yes, systems biology (integrating everything in a large comprehensive view)!

Paul Harrison
Tadafumi and others, as methylation is both temporally and regionally (if not cell-type) specific in human brain (e.g., Ladd-Acosta et al., 2007), it won't be trivial to do this.

Akira Sawa
Integrating everything...important. Please do include time course in these parameters.

James Chao
Perhaps network analysis, too, with gene coexpression.

Tadafumi Kato
Akira, and tissue heterogeneity should also be included, using dissection or other methods.

Elizabeth Thomas
Regarding developmental changes, if we consider that development (i.e., myelination) continues into the late twenties, looking at subjects early in illness would be important. Our microarray findings (not yet published) suggest that the most dramatic changes in gene expression occur early in illness (less than 5 years from diagnosis, subjects all 18-25 years old).

Cyndi Shannon Weickert
Elizabeth, looking at changes in gene expression proximal to the onset will be very valuable! Good work.

Elizabeth Thomas
Cyndi, thanks.

Mark Vawter
Joel/Barbara, are there any young schizophrenia subjects or high-risk brains that could be studied?

Joel Kleinman/Barbara Lipska
It is unclear to me why we would have young patients with schizophrenia in a postmortem study. Given the importance of genotype, there is no substitute for having numbers as large as possible for the index and control groups—this is far more important.

Mark Vawter
Joel/Barbara, hmmm, since genetics is only one part, it seems that the early brain changes could be seen independent of genetic risk.

Joel Kleinman/Barbara Lipska
What early brain changes are you referring to?

Mark Vawter
Joel/Barbara, it is an open hypothesis: are there brain changes in young subjects at risk for schizophrenia? Some early imaging data suggest brain region size differences. (Early unpublished differences in cases vs. controls presented at the 2007 ICOSR meeting in Colorado Springs by John Gilmore from UNC.)

Akira Sawa
I believe that data of gene expression from young people (0-20 years old) should be an important reference in comparison with that from adults. We do not have to get data from young-onset schizophrenia patients, but it may be important to know age-dependent gene expression change in order to properly interpret data from aged schizophrenia patients and controls. Don't you think so?

Joel Kleinman/Barbara Lipska
Akira, yes, we agree completely!

Paul Harrison
Joel/Barbara are right about sample size being crucial; once we get into the hundreds of subjects, it becomes realistic to examine contributions of genotype, stress, age, etc.

Cyndi Shannon Weickert
I think one thing that having young patients with schizophrenia helps with is sorting out what David Lewis et al. referred to in their latest review paper as the four "C’s" (Lewis and González-Burgos, 2008). If you can look at gene expression early enough in the disease, you may be able to start to better order the events of the pathology and to understand what the consequences may be in molecular terms as they "unfold." Studying genotype is one way to go, but we all know that many "risk" genes are found in people that never go on to develop schizophrenia, so there is no better substitute for studying schizophrenia than the schizophrenia brain itself.

Karoly Mirnics
Paul, in this climate of competition (and not collaboration, regardless what NIH is trying to sell), will we ever arrive at hundreds of samples?

Paul Harrison
Karoly, not in the U.K.; that's for sure!

Mark Vawter
Karoly to Paul, I think meeting at SOBP to discuss putting together hundreds of samples could also be helpful.

Karoly Mirnics
Mark, that would be the Holly Grail of this research.

Joel Kleinman/Barbara Lipska
Collaboration only occurs when it is in the interest of both or all the collaborators.

Akira Sawa
By the way, we have only 3 more minutes. We now have 47 colleagues in this roundtable!! We still have a lot to talk about. Maybe different occasions at SOBP over wine/beer. Today's chat should be an important template for future scientific discussions. Dear all, do you have something to add in the last 2-3 minutes?

Hakon Heimer
The room will stay open as long as you like if you want to "rush the speaker" or gather in small groups, but as Akira said, now is that time for final statements, since some people have to leave for lunch, dinner, or go back to bed in the other hemisphere.

Cyndi Shannon Weickert
Hakon, how can anyone sleep with all this excitement!

Tadafumi Kato
Genetic studies are going from common SNPs to multiple rare variant studies. The postmortem brain studies may also go to the same direction. The cause may differ patient to patient.

Joel Kleinman/Barbara Lipska
Okay, my last comment of the day. Toxicology on brains of patients with schizophrenia has cost us $590/case. WGA chips cost about the same.

Karoly Mirnics
A new wrench: my biggest issue still remains disease homogeneity. Two schizophrenics may be very different from each other, and we have no power to uncover the substratifications. Yet, virtually all statistics we use assume that all schizophrenics are suffering from a similar disturbance (I am guilty of this as charged, too).

Joel Kleinman/Barbara Lipska
Karoly, absolutely correct. This is part of the reason that genotypes are crucial.

Christine Miller
Karoly, although two schizophrenics are very different from each other, they also can be very different within the same family. Presumably, the genetic component within the same family is quite similar....

Karoly Mirnics
Christine, in postmortem you will have no family members, but unrelated subjects.

Christine Miller
Karoly, yes, but my point was that disease heterogeneity can exist even within a largely similar genetic (and similar gene expression) background. Timing of gene expression may be key.

Vahram Haroutunian
Christine, agree. I was struck by a comment that Joe Coyle made to me yesterday. Schizophrenia does not breed true. The offspring of two phenotypically similar schizophrenics may have/often have a very different syndromic or symptomatic profile.

Karoly Mirnics
C. Miller, you may be right, but there is so much we cannot control—age of death, medication, lifestyle, genetic makeup—and this makes two schizophrenic brains so different that they may show very few commonalities in a gene expression pattern. This is a question that plagues us all. Joel's genotyping may get around the genetic issue, but what to do with the rest?

Sherry Leonard
Chris Miller and Karoly, there are huge effects of stress on gene expression, and stress could be very different for each subject. Also agree that methylation is an important parameter to think about, as is miRNA.

Cyndi Shannon Weickert
Does anyone know how many postmortem brains are even available around the world, if the brain banks were to cooperate with this notion?

Karoly Mirnics
Cindy, good ones or bad ones:)))

Vahram Haroutunian
Cyndi, an update of Paul Harrison's paper a few years ago: doing a meta-analysis on brain weight in schizophrenia (Harrison et al., 2003) could help identify the numbers of cases.

Cyndi Shannon Weickert
Thanks, Vahram, I will have a look. G'day everyone and I vote for having the meeting in Sydney! Cheers!

Helga Smith
Karoly, maybe if you ask the families that have multiple members with schizophrenia via places like schizophrenia.com you could get better/younger samples after suicides?

Karoly Mirnics
Helga, we tried many things, but the reality is that most of the banks grow by four to five good schizophrenic brains a year.

Hakon Heimer
Karoly, that's a disappointing statistic. Perhaps SRF can get NARSAD/NAMI more involved.

Helga Smith
Karoly, I have read all parent message boards for 10 years and this problem of getting brains has never been discussed. The parents would love to help.

Karoly Mirnics
Helga, believe it or not, the availability of the high-quality brains is the most limiting factor in our research. Thank you all; I am looking forward to a great meeting at SOBP.

Joel Kleinman/Barbara Lipska
Who will organize something for SOBP?

Sherry Leonard
Can Akira and Mark organize something for SOBP?

Mark Vawter
Joel, I would offer to do that with Akira and others. As a reminder, please e-mail nico@schizophreniaforum.org for an interest in meeting/discussing topics 1, 2 further.

Paul Harrison
Good session! Here's to a meeting at SOBP. Bye.

Akira Sawa
Mark and Joel, great!

Vahram Haroutunian
Joel, I think that an NIMH-organized meeting/working group around this topic would be better than a few minutes at SOBP.

Tadafumi Kato
Thank you, everyone; very hot session!

Christine Miller
Thanks to everyone!

Joel Kleinman/Barbara Lipska
Okay. If the group wants NIMH, I will try to set it up. Otherwise, we can try for SOBP or a small meeting at my home.

Sherry Leonard
Joel, I hope you have a big house!

Joel Kleinman/Barbara Lipska
Yes, I have a big house. Barbara and I are signing off for now.

Mark Vawter
All, I see names on the right of many others who didn't get a word in. Thanks for coming today. We will send out a transcript...and Akira and I will arrange a follow-up.

Cyndi Shannon Weickert
Just kidding, of course. I prefer SOBP, not NIMH! Should we meet a day early to have more time?

Akira Sawa
Dear all, thank you very much for coming!

Hakon Heimer
A big round of applause for Akira and Mark.

Joel Kleinman/Barbara Lipska
Hands clapping.

Sherry Leonard
Many thanks.

Karoly Mirnics
Organizers, great job.

Elaine Shen
Thanks to the organizers. I've found this to be very interesting and hope to get involved more in future discussions.

Mark Vawter
Elaine, I think we will need to try this again.

Tony Altar
Mark, thanks for the offer to send out the transcripts for this meeting, and keep me in mind for the next one.

Mark Vawter
Tony, sure, we didn't really even address medications. I think that will be important as part of understanding the pathology.

Andreas Jeromin
Great forum. I would like to see some other issues discussed as well.

Tony Altar
Andreas, what other issues did you have in mind?

Andreas Jeromin
Inter-laboratory comparison of RNA and tissue QC, for example, need for imaging. Histology as part of the acquisition.

Hakon Heimer
Tony, Andreas, please write a comment at the discussion page on any other issues.

Mark Vawter
All, thanks again for your comments, and please post more at the Forum. Hakon, I will be logging off. Great job hosting this forum.

Hakon Heimer
Well, I think we're done, Mark. A good day's work and it's only morning in California.

Akira Sawa
Hakon, Nico, Mark, and friends, I will be logging off. Thank you very much!

Hakon Heimer
Akira, thank you for bringing this lively crowd to SRF.

Comments on Online Discussion
Comment by:  Karoly Mirnics, SRF AdvisorFuller Torrey
Submitted 30 December 2007
Posted 30 December 2007

Summary of Workshop Regarding Future Use of Stanley Brain Collection Tissue
(30 November 2007, Bethesda, Maryland)

There appears to be good consensus among researchers working with Stanley Brain Collection tissue on what is needed for future brain collections. Much has been learned from the existing collections and so there is a lot of expertise available. Thanks to the postmortem research, we have identified processes and molecular mechanisms that are critically involved in schizophrenia, and there is an emerging consensus about some of the most important findings. However much more is to be done.

The complexity of the anatomical brain structures must be respected. The cellular phenotypes making up the brain are different, and the disease process will affect them differently. Thus, we must find ways to respect the microanatomy in our experiments. There appears to be a consensus that laser-capture techniques using single cells is the most promising wave of the future. The problems are that it is difficult, labor intensive, and expensive. Other promising techniques are stereology, protein arrays, assessment of splice variants and mass spectrometry imaging.

Many participants cautioned that we should not assume that we know which parts of the brain are most important to study in schizophrenia and bipolar disorder despite the concentration of work in two or three areas. Schizophrenia appears to be the disease of multiple brain regions, and we must expand our experiments beyond the prefrontal cortex, temporal cortex, anterior cingulate and hippocampus.

Methods of extracting RNA from fixed tissue, even tissue fixed for many years, are likely to improve markedly, though some limitations of postmortem tissue may remain. However, the researchers were optimistic that archival, previously fixed tissue may become a valuable source of RNA in the near future.

There was a consensus that existing brain collections should not be combined, because of methodological differences, although a virtual online collection was discussed as a way to identify new cohorts, e.g., all cases with schizophrenia who died within two years of original diagnosis. Furthermore, the researchers favored replication of findings across multiple cohorts over combining samples from different cohorts into a single group of samples/experimental series.

There was much discussion regarding the importance of doing SNPs, copy number variations, etc., and making the data available on a public database. There is a strong consensus that all data should be made available online, potentially in raw data format, so that many other researchers can identify and follow up leads, especially ones that may be initially cryptic in the dataset. Investigating the same samples with many different tools (e.g. expression arrays, SNP chips, epigenetic modifications, etc) may be a very powerful strategy, especially if all those data are properly integrated into an easy-to-use database.

Stanley Medical Research Institute is continuing to support neuropathology research as a priority research area. Details are available at the institute website.

View all comments by Karoly Mirnics
View all comments by Fuller TorreyComment by:  Christine Miller
Submitted 7 January 2008
Posted 7 January 2008

My view is that postmortem studies of the affected organ (the brain, in this case) will always be important and that the focus should be on expanding and improving the repertoire of available techniques. Data derived from a variety of methods help to form the most complete picture, and the significance attributed to any given set of results must be evaluated and re-evaluated based on current knowledge about the limitations of the methods used.

I have two points to make: one about improving our understanding of postmortem effects and the other about variability in sectioning.

1. The first relates to animal models of postmortem factors that affect mRNA, protein, and metabolite measurements from brain tissue. To my knowledge, the animal studies that have been published do not mimic the postmortem conditions that are in effect for deceased humans. Specifically, the effects of postmortem interval have been modeled after the brain has been removed from the cranium, not before, whereas the PMI referred to in the human studies relates to the time between death and the removal of the brain from the cranium. Why might this distinction be important? A seminal study by the lab of Sherry Leonard (Leonard et al., 1993) indicated that although pre-autopsy PMI was unrelated to mRNA recovery, the interval between autopsy and freezing was important. What mechanism could account for this difference? As many enzymes retain their activity postmortem, it is possible that severing of axons may initiate processes that affect mRNA and protein levels. Thus, animal models should measure PMI as the time between death and removal of the brain from the cranium. Different modes of death should also be addressed, in particular because cervical dislocation involves severing of axons.

2. The second concern relates to the issue of sectioning variability. The advantage of macrodissection versus laser-capture microdissection is that without prior knowledge of neuron-to-neuron or astrocytic variation in expression, sampling a larger area is likely to result in more representative mean values than sampling a smaller area. The exception is variability in the proportion of white/grey matter in a sample chunk, which can vary substantially at the macroscopic level. One means of controlling for this variability is to measure the expression of genes more specific to white matter. An example of such a gene (abundantly expressed) is PLP (myelin proteolipid protein). For quantitative Western work, effective antibodies are commercially available for this protein.

References:

Leonard S, Logel J, Luthman D, Casanova M, Kirch D, Freedman R. Biological stability of mRNA isolated from human postmortem brain collections. Biol Psychiatry. 1993;33(6):456-66. Abstract

View all comments by Christine MillerComment by:  Alberto Arregui
Submitted 7 January 2008
Posted 7 January 2008

In research projects involving postmortem material, extreme care needs to be applied in formulating conclusions. The long-term effects of neuroleptics, even in patients who have died without use of meds for many months, on brain substances is unknown. The known effect of "protracted" illnesses on brain transmitters is also present. I participated in studies done at the MRC Neurochemical Unit in Cambridge in the late 1970s (e.g., Arregui et al., 1979; Arregui et al., 1980; Mackay et al., 1982), and many findings of altered peptides were left unpublished because it was very difficult to find appropriate explanations for the findings. Good luck.

References:

Arregui A, Mackay AV, Iversen LL, Spokes EG. Reduction of angiotensin-converting enzyme in substantia nigra in early-onset schizophrenia. N Engl J Med. 1979 Mar 1;300(9):502-3. Abstract

Arregui A, Mackay AV, Spokes EG, Iversen LL. Reduced activity of angiotensin-converting enzyme in basal ganglia in early onset schizophrenia. Psychol Med. 1980 May 1;10(2):307-13. Abstract

Mackay AV, Iversen LL, Rossor M, Spokes E, Bird E, Arregui A, Creese I, Synder SH. Increased brain dopamine and dopamine receptors in schizophrenia. Arch Gen Psychiatry. 1982 Sep 1;39(9):991-7. Abstract

View all comments by Alberto ArreguiComment by:  Beth Thomas
Submitted 7 January 2008
Posted 7 January 2008

I have a question about whether there is any consensus over how to best address the potential influences of lifetime antipsychotic drug exposure on gene expression findings. This is certainly an important confounding factor in any gene expression study using postmortem samples from schizophrenic subjects. Three approaches have been used in the literature, although each has its weakness.

1. Comparisons of gene expression levels from drug-treated schizophrenic subjects to those who had been “medication-free” for various periods of time. The doubt here is that, after a presumed lifetime of antipsychotic drug treatment, would a period of 3-6 months “drug-free” be long enough to reverse the chronic effects of drugs?

2. Correlations between gene expression levels and recorded drug doses. This seems logical; however, it is known that many patients are noncompliant with their reported medications, and sometimes drug traces or metabolites are found in subjects that are different from what is listed in the medical records.

3. Measurements of antipsychotic drug influences in rodents. This approach offers a controlled means to evaluate the effects of specific antipsychotic drugs at various doses and durations. Nonetheless, drug effects may not be similar in rodents due to differences in brain structure and complexity, and the fact that rodents lack underlying pathological insult(s) causing the disease itself.

Must we then consider that any gene expression alteration detected in postmortem tissue is possibly a drug effect?

View all comments by Beth ThomasComment by:  Sherry Leonard
Submitted 7 January 2008
Posted 7 January 2008

The topics for discussion are all important! Would like to add one more: cause of death. Some years ago we did in vitro translation and PAGE from mRNA isolated from control subjects, matched for age, sex and ethnicity, who had died from disparate causes. The protein profiles were VERY different. Since that time we also match samples for cause of death if possible. Wondering if anyone else has thought about this issue.

What is really needed is a database with expression profiles from controls with different causes of death, but that is dreaming!

View all comments by Sherry LeonardComment by:  Kenji Hashimoto
Submitted 7 January 2008
Posted 7 January 2008

We believe that postmortem human brain samples are critical for examining molecular changes associated with the pathophysiology of schizophrenia. However, there are a number of confounding factors, including the postmortem interval (PMI). A number of studies have used autopsied brain samples for which there were no significant differences between the mean PMI values, but there were wide ranges (e.g., ~72 hrs for Stanley Brain Samples) of PMI. For example, levels of amino acids, including glutamate, in the brain are clearly affected by PMI. In a study of brain amino acids in postmortem samples from schizophrenia patients, we first normalized their time-course changes to simple equations in rodents, which were then applied to the studies of human autopsied brain samples (Hashimoto et al., 2007). Therefore, the use of simple equations using animals may be necessary to determine the substances affected by confounding factors.

Schizophrenia is a neurodevelopmental disorder. Therefore, we may not see or detect evidence of the disorder in autopsied brains since major pathological events that could affect neurodevelopment may be compensated for at later stages. As alternatives to autopsied brains, it is important to measure samples (e.g., blood, CSF) of intact human patients. For example, we reported decreased levels of D-serine in serum or CSF of schizophrenic patients, supporting the NMDA receptor hypofunction hypothesis in schizophrenia (Hashimoto et al., 2003; Hashimoto et al., 2005). It is also of great interest to measure the levels of endogenous substances at an early stage to predict the onset of schizophrenia.

As an alternative technique, brain imaging techniques including PET and MRS are suitable techniques to detect in vivo changes of receptors, enzymes, and endogenous substances in intact human brain. Thus, I believe that the use of brain imaging techniques as well as the autopsied brain samples is necessary to study the pathophysiology of schizophrenia.

References:

Hashimoto K, Sawa A, Iyo M. Increased levels of glutamate in brains from patients with mood disorders. Biol Psychiatry. 2007 Dec 1;62(11):1310-6. Abstract

Hashimoto K, Fukushima T, Shimizu E, Komatsu N, Watanabe H, Shinoda N, Nakazato M, Kumakiri C, Okada S, Hasegawa H, Imai K, Iyo M. Decreased serum levels of D-serine in patients with schizophrenia: evidence in support of the N-methyl-D-aspartate receptor hypofunction hypothesis of schizophrenia. Arch Gen Psychiatry. 2003 Jun ;60(6):572-6. Abstract

Hashimoto K, Engberg G, Shimizu E, Nordin C, Lindström LH, Iyo M. Reduced D-serine to total serine ratio in the cerebrospinal fluid of drug naive schizophrenic patients. Prog Neuropsychopharmacol Biol Psychiatry. 2005 Jun ;29(5):767-9. Abstract

View all comments by Kenji HashimotoComment by:  Tadafumi KatoKazuya Iwamoto
Submitted 7 January 2008
Posted 8 January 2008

Thanks to the endeavors of Dr. Torrey and all the other contributors of the Stanley Medical Research Institute, postmortem brain study has been opened to the research community worldwide. Gene expression profiles have been analyzed in the Stanley brain bank samples by several laboratories including ours and the data are now open to the public. However, the description on the results shown in the database and the content of the papers from individual groups are not always consistent, partly because of the difference in the philosophy of how the confounding factors should be controlled for.

The paper by Dr. Miller and colleagues dealt with this important issue of confounding factors affecting gene expression profiles in postmortem brains using the largest numbers of brain samples to date (Weis et al., 2006). They reported that sample pH did not affect the RNA quality. This is well in accordance with our previous report showing that there was only a weak relationship between pH and call rate by GeneChip (Iwamoto et al., 2006). Based on this finding, Weis et al. proposed that samples should be assessed mainly by RNA integrity number (RIN), which is calculated from the electropherogram of RNA. Samples should not be excluded due to other reasons such as low pH and long postmortem interval. These factors affect the expression of subgroups of genes, but do not impair the RNA quality in general.

Considering the difficulties in collecting brain samples of patients with mental disorders, it is obvious that as many brain samples as possible should be subject to analysis. It is also true, however, that the observed disease-related changes might be caused by confounding factors, because the cause of death is usually quite different between patients and controls.

This issue has become a matter of debate also in the study of bipolar disorder. It is still unknown whether the global downregulation of mitochondria-related genes in the postmortem brains of bipolar disorder patients (Konradi et al., 2004) is due to low sample pH due to difference of cause of death (Li et al., 2004; Tomita et al., 2004; Iwamoto et al., 2005), or due to decreased pH reflecting the pathology of bipolar disorder itself (Sun et al., 2006).

Several other systematic changes of gene expression, such as those related to oligodendrocyte-related genes, have been observed in the postmortem brains of patients with mental disorders. Such systematic gene expression changes observed in mental disorder patients can be due to either pathophysiology itself or confounding factors.

In any case, such discussion has arisen after the analysis of many samples including those with low pH or longer PMI. Thus, the proposal by Dr. Miller and colleagues that all samples with good RIN should be analyzed is quite reasonable as the starting point of postmortem brain study.

In carefully investigating the molecular basis of altered gene expression in mental disorders, examination of other parameters using different experimental techniques would be useful. In this regard, the analysis of granular cell layer necrosis performed by Dr, Miller and colleagues is an interesting approach. Integrated analysis using genomic and epigenomic profiles together with gene expression profiles will also be needed to further investigate the nature of gene expression changes in the postmortem brains in mental disorders.

References:

Weis S, Llenos IC, Dulay JR, Elashoff M, Martínez-Murillo F, Miller CL. Quality control for microarray analysis of human brain samples: The impact of postmortem factors, RNA characteristics, and histopathology. J Neurosci Methods. 2007 Sep 30;165(2):198-209. Abstract

Iwamoto K, Bundo M, Ueda J, Kato T. Expression of ribosomal subunit genes increased coordinately with postmortem interval in human brain. Mol Psychiatry. 2006 Dec;11(12):1067-9. Abstract

Konradi C, Eaton M, MacDonald ML, Walsh J, Benes FM, Heckers S. Molecular evidence for mitochondrial dysfunction in bipolar disorder. Arch Gen Psychiatry. 2004 Mar;61(3):300-8. Abstract

Li JZ, Vawter MP, Walsh DM, Tomita H, Evans SJ, Choudary PV, Lopez JF, Avelar A, Shokoohi V, Chung T, Mesarwi O, Jones EG, Watson SJ, Akil H, Bunney WE Jr, Myers RM. Systematic changes in gene expression in postmortem human brains associated with tissue pH and terminal medical conditions. Hum Mol Genet. 2004 Mar 15;13(6):609-16. Abstract

Tomita H, Vawter MP, Walsh DM, Evans SJ, Choudary PV, Li J, Overman KM, Atz ME, Myers RM, Jones EG, Watson SJ, Akil H, Bunney WE Jr. Effect of agonal and postmortem factors on gene expression profile: quality control in microarray analyses of postmortem human brain. Biol Psychiatry. 2004 Feb 15;55(4):346-52. Abstract

Iwamoto K, Bundo M, Kato T. Altered expression of mitochondria-related genes in postmortem brains of patients with bipolar disorder or schizophrenia, as revealed by large-scale DNA microarray analysis. Hum Mol Genet. 2005 Jan 15;14(2):241-53. Abstract

Sun X, Wang JF, Tseng M, Young LT. Downregulation in components of the mitochondrial electron transport chain in the postmortem frontal cortex of subjects with bipolar disorder. J Psychiatry Neurosci. 2006 May;31(3):189-96. Abstract

View all comments by Tadafumi Kato
View all comments by Kazuya IwamotoComment by:  Sinthuja SivagnanasundaramDuncan SinclairCynthia Shannon Weickert (SRF Advisor)
Submitted 8 January 2008
Posted 8 January 2008

A Few Technical Comments About Evaluating RNA Quality From Human Brain

Since the measurement of RNA quality using several outputs from the Agilent 2100 Bioanalyser is considered a key tool in evaluating RNA quality from the human brain, we would like to point out just a few experimental factors that may impact the determination of RNA quality using the Bioanalyser: 1) the effect of heating the RNA samples prior to loading on the RNA 6000 Nano Chip, 2) the effect of drying down the samples, and 3) the effect of running the RNA through a purification column.

We will comment on each of these points based on our experience in the Schizophrenia Research Laboratory in Sydney Australia.

1) Although heating the total RNA sample before running the RNA 6000 Nano Chip is recommended by the manufacturer, it can result in lower RIN values (note that in both Lipska et al., 2006 and in Weis et al., 2007 the RNA samples were not heated prior to loading).

2) In the Weis et al. paper they found that drying down the RNA improved the RNA quality. In our laboratory, we did not find an improvement in the RNA quality after drying the samples down at either room temperature (RT) or at 60oC (Fig. 1).

Figure 1. Electropherogram of total RNA before and after drying down samples at RT.

The RIN values of RNA derived from total RNA from the sample of human frontal cortex before drying down and after did not differ significantly (measured 5-6 times in each condition; Fig. 2). We found that the characteristics of the peaks did seem to change after drying down as found by Weis et al. 2007, but in our case the 18S peak was more pronounced, significantly decreasing the 28S/18S ratio (p <0.01; Fig. 1).

Figure 2. Comparison of RIN values of total RNA before and after drying down at RT.

3) Also, our electropherogram of total RNA (Fig. 1) has a more pronounced 5S peak than that reported by Weis et al. We have found that running the total RNA through a QIAGEN column almost eliminates the RNA band traveling at the 5S range (data not shown). Also, running total RNA through a column can change the RIN values typically increasing RIN values somewhat artificially.

In order to make sure that we are comparing “apples to apples” across studies, we suggest that people should report RIN values both before and after column purification. Also, we point out that drying down RNA samples may not always lead to improvements in RNA quality.

View all comments by Sinthuja Sivagnanasundaram
View all comments by Duncan Sinclair
View all comments by Cynthia Shannon WeickertComment by:  Christine Miller
Submitted 9 January 2008
Posted 9 January 2008

Reply to Cynthia Shannon Weickert and colleagues:

Thanks for sharing your experience with concentrating RNA. It is important to clarify that our paper (Weis et al., 2007) did not assert that concentrating the RNA actually improved its quality for microarray analysis (i.e. degraded RNA was not restored to its original intact state). Rather, the point was that this was another example of RNA handling procedures that can affect the electropherogram without actually improving RNA quality.

I am curious as to the starting RNA used to test the effect of concentrating via drying down. Was it solvent extracted RNA (eg. with Trizol)? This RNA likely exists in a highly folded state due to the alcohol precipitation step during extraction, and much less additional folding would be expected to occur when concentrated by drying. The extent of folding should affect migration of the 28S and 18S bands in the nanochip. Column-extracted RNA, in contrast, has been denatured during the column binding step (i.e. folding largely eliminated in mRNA, and possibly somewhat affected in 28S and 18S RNA as well). This changes its electrophoretic mobility and probably accounts for its greater stability as compared to Trizol-extracted RNA (Miller and Yolken, 2003), by preventing autocatalysis (and degradation by protein enzymes as well) that is enhanced by the folded state (Cech's work is a good reference for this). In the basic science literature, folding in mRNA is measured by A260/280 which, in the absence of protein, is higher for column-extracted RNA than in solvent-extracted RNA. Column-extracted RNA would therefore be more capable of undergoing an increase in the folded state during concentration. Folding would be predicted to occur based on the increase in di-valent salt and the exclusion of intra-molecular water during concentration.

I have heard that column extracted RNA and solvent extracted RNA apparently do give different results in microarray analysis. The wash step in column extraction will typically remove fragments <150 to 200 bp, and species larger than 10,000 bp are unlikely to bind well. Unless you are interested in very large or very small mRNAs, it is my view that the column-extraction methods are better because they exclude degraded mRNA and result in a more stable product for storage purposes. In addition, many endogenous antisense species exist as short fragments, which can interfere with RT-PCR for the sense strand, but may be less of a problem for microarray analysis.

References:
Miller CL and Yolken RH, 2003, Methods to optimize the generation of cDNA from postmortem human brain tissue. Brain Res Protocols 10: 156-167.

View all comments by Christine MillerComment by:  Paul Harrison
Submitted 9 January 2008
Posted 9 January 2008

Several of the above comments pertain to the use of pH and/or RIN to predict/control for variation in mRNA quality in postmortem brains. Could I also raise the issue of utility and reliability: the pH of a given brain is pretty consistent between one brain area and another, and has a high test-retest reproducibility (and temporal stability). By contrast, we find RIN shows more variation in all these respects (as also shown by Lipksa et al., 2006, and other recent papers), presumably due to the technical factors discussed by Miller and Weickert, as well as to biological factors. Probably as a consequence, our experience is that (depending on the experimental dataset) pH sometimes does better than RIN in explaining variance in mRNA signals between brains, sometimes RIN does better, and sometimes but not always the combination is better than either alone. I think there are two implications of this:

1. Although RIN is a valuable index, it has its limitations and it should not be viewed uncritically or automatically as the gold standard in the field—or by implication that pH is now inferior or redundant.

2. We should not set criteria as to a “minimum RIN” required for a study to be acceptable—rather, we should (a) encourage authors to report in greater detail how (and when) RIN was measured; and (b) state clearly how they determined whether RIN affected the variables being assessed, and on this empirical basis to justify the inclusion/exclusion of brains—and the statistical analysis employed. I would suggest that the same pragmatic approach should be taken to maximum postmortem intervals, or any other potential confounder.

View all comments by Paul HarrisonComment by:  Christine Miller
Submitted 9 January 2008
Posted 9 January 2008

Clarification to my last commment: a higher 260/280 corresponds to less folding of the mRNA.

View all comments by Christine MillerComment by:  Barbara K. LipskaJoel Kleinman
Submitted 9 January 2008
Posted 9 January 2008

We have previously suggested that accurate assessment of multiple confounding factors and their inclusion as regressors in the analysis is critical for obtaining reliable and accurate quantification of mRNA expression (Lipska et al., 2006). One of these confounding factors is pH. Insofar as lower pH has been associated with decreased mRNA expression in postmortem human brain, decreased pH in schizophrenia may represent an important potential confound in comparisons between patients and controls. We are now showing that decreased pH is related to increased concentration of lactic acid (Halim et al., in press).

However, in contrast to the previous notion that an increase in lactic acid represents evidence for primary metabolic abnormalities in schizophrenia, we hypothesized that this increase is secondary to prior antipsychotic treatment. We have tested this by first demonstrating that lactate levels in the cerebellum of patients with schizophrenia are increased relative to control subjects. Second, we have shown that there is an excellent correlation between lactate levels in the cerebellum and pH, and that this correlation is particularly strong in patients. Third, we have shown in rats that chronic haloperidol and clozapine increase lactic acid concentration in the frontal cortex relative to vehicle. These data suggest that lactate increases in postmortem human brain of patients with schizophrenia are associated with decreased pH and that these changes are likely related to antipsychotic treatment rather than a primary metabolic abnormality in the prefrontal cortex of patients with schizophrenia. Moreover, the effect of antipsychotics on lactate levels may appear particularly exaggerated in elderly patients with schizophrenia.

References:

Halim ND, Lipska BK, Hyde TM, Deep-Soboslay A, Herman M, Thakar J, Verma A, Kleinman JE. Increased lactate Levels and reduced pH in postmortem brains of schizophrenics: Medication confounds. J Neurosci Meth. In press.

Lipska BK, Deep-Soboslay A, Shannon Weickert C, Hyde TM, Martin CE, Herman MM, Kleinman JM. Critical factors in gene expression in postmortem human brain: focus on studies in schizophrenia. Biol Psychiatry. 2006;60:650-658. Abstract

View all comments by Barbara K. Lipska
View all comments by Joel KleinmanComment by:  Sinthuja Sivagnanasundaram
Submitted 17 January 2008
Posted 17 January 2008

Reply to Comment by Christine Miller
Thanks for the clarification. We did use Trizol for our RNA extraction and we plan to use this total RNA primarily for RT-PCR. We agree that different extraction protocols will result in different measures of RNA quality, and we suggest that details of the RNA isolation, purification, and amplification methods should be clearly stated in publications. Our measure of the A260/280 ratio for Trizol extracted RNA before and after purification was on average 1.89 and 2.05, respectively, which suggests that column purification improves the A260/280 ratio.

View all comments by Sinthuja Sivagnanasundaram