A F F Y M E T R I X
                                                                                                                VOLUME 4 ■ ISSUE 2
                                                                                                                          AMB REPRINT

www.microarraybulletin.com                                                                                                    Summer 2008

EURASNET Consortium Evaluates the Performance of
Different Microarrays for Studying Alternative Splicing
Chris Smith of Cambridge University, Tyson Clark of Affymetrix and Melissa Cline of UC Santa Cruz discuss
EURASNET’s approach for comparing commercial microarrays for investigating alternative splicing

By Stacey Ryder

Cambridge, UK, August 1, 2008 — A             very well-annotated exons to exons for          algorithm that normalizes exon signal
European consortium of nearly 40 labo-        which there is only a low level of confi-       to the level of gene expression. Exon-
ratories has conducted a study which          dence in the prediction. These arrays pro-      level signal intensities reflect a combi-
identified GeneChip® Human Exon 1.0           vide the necessary power for discovering        nation of the number of transcripts
ST Arrays as an effective tool for studying   new alternative splicing events.”               made from a gene and the frequency in
alternative splicing—a process that results      After the data was collected by the          which a particular exon appears in
in multiple isoforms of a single gene, and    EURASNET team, scientists at                    those transcripts. Included in the
a major source of protein diversity.          Affymetrix grouped the resulting hits           ordered list of hits were 32 high-confi-
    The study compared the ability of         by confidence levels using a simple             dence splicing events. Smith’s group
microarray platforms to investigate
alternative splicing in transcripts regu-
lated by the splicing factor polypyrimi-       Chris Smith, Ph.D.,                 is a
dine tract binding (PTB) protein, and its
                                                professor in the Department of
paralog, nPTB. Researchers hope that
uncovering the mechanisms which reg-            Biochemistry at the University of
ulate alternative splicing will lead to a       Cambridge. For the past 15 years, his
better understanding of human disease.
                                                group has focused on understanding
    The consortium, dubbed EURAS-
NET (European Commission-funded                 the molecular mechanisms behind
Network of Excellence), found that              alternative mRNA splicing. Dr. Smith
GeneChip Human Exon 1.0 ST Arrays
                                                is also a member of the EURASNET
outperformed other splice-sensitive
array platforms in a study which investi-       consortium, a group of almost 40
gated alternative splicing in cancer            laboratories across Europe that is
(HeLa) cells. GeneChip Exon Arrays,
which include approximately 1.4 million         combining resources to research

probe sets, allow for complementary             alternative splicing on a larger scale.
high-resolution gene expression and
                                                Dr. Smith and his laboratory were
alternative splicing analysis. By using
these genome-wide arrays, EURASNET              recently involved in a EURASNET
researchers were readily able to detect         study to determine the most appro-
and confirm a total of 38 splicing events.
                                                priate commercial microarray platform
    “Exon arrays are global,” said
Cambridge University’s Dr. Chris Smith,         for investigating alternative splicing.
a EURASNET member whose lab
designed the study. “They feature a huge
number of probe sets…ranging from the

                                                                                           A MB INTERVIEW REPRINT ■ SUMM E R 2 0 0 8
                                                                 consortium wanted to come up with a         from the gene and how often a particu-
                                                                 model experiment that could be used to      lar exon is included in those transcripts.
                                                                 compare different array platforms.              To tease out splicing information,
                                                                     We had been looking at the splicing     we use methods that normalize out
                                                                 factor, PTB protein, for a number of        those transcriptional differences to
                                                                 years. We had performed experiments         observe individual probe sets relative to
                                                                 in which we knocked down PTB and its        an estimate of the gene expression
                                                                 paralog, nPTB, in HeLa cells and then       level. Changes in the gene-level-nor-
                                                                 used quantitative proteomics to identify    malized intensities suggest a change in
                                                                 up- and down-regulated spots. Our           exon inclusion rates and thus differen-
                                                                 hypothesis was that we should be able       tial splicing.
                                                                 to pick up examples where alternative           The confidence levels were assigned
                                                                 splicing had switched. So far we’ve         based on a combination of the statisti-
    Tyson Clark, Ph.D.,                is a member               identified 25 PTB-regulated events by       cal significance of the change in inclu-
                                                                 this approach.                              sion rate and an examination of the
    of the Product Development Group at
                                                                     The consortium decided our experi-      genomic context by observing the
    Affymetrix, Inc. He received his Ph.D. from                  ment would be a good model experi-          underlying probe set intensities super-
    the University of California at Santa Cruz,                  ment because we were using a                imposed onto the genome.
                                                                 completely independent technology               Smith: The high-confidence predic-
    where his work focused largely on alterna-
                                                                 platform to look for changes in alterna-    tions that Tyson gave us included 32
    tive splicing. As a graduate student, he was                 tive splicing, and we were investigating    events. We validated those at about 95
    involved in some of the first microarray
                                                                 a splicing factor for which a large         percent level. The results were much
                                                                 number of targets were known.               better than I had expected.
    experiments to study alternative splicing.                       Cline: What array platform did you
                                                                 test? What were its advantages?             Understanding alternative splicing
used reverse transcription followed by                               Smith: We quickly focused on the        Cline: When you observe an alterna-
real-time PCR to validate 95 percent of                          Human Exon 1.0 ST Array from                tive splicing event, how can you start
those events.                                                    Affymetrix, as during the validation        putting that into a functional context?
   Dr. Smith and Tyson Clark, of                                 exercise, it became evident to us that it       Smith: We are not all that interested
Affymetrix’ Product Development                                  was out-performing the alternatives.        in identifying individual alternative
group, recently spoke to Melissa Cline,                              In addition, exon arrays have the       splicing events. We want to have a large
a Senior Informatics Postdoctoral                                major advantage of being global. They       group of co-regulated events, then get
Fellow at the University of California at                        feature a huge number of probe sets—        into the informatics of trying to deci-
Santa Cruz, to discuss the study and                             approximately 1.4 million—ranging           pher how the gene sequence makes one
future directions for alternative splicing                       from the very well-annotated exons to       particular exon responsive to a particu-
research. The three discussed:                                   exons for which there is only a low level   lar splicing regulator.
  • Advantages of the GeneChip                                   of confidence in the prediction. These          This was one of the reasons our
     Human Exon 1.0 ST Array                                     arrays provide the necessary power for      experiment was used as the model
  • Understanding the functional rele-                           discovering new alternative splicing        experiment. We know a lot about the
     vance of alternative splicing                               events.                                     optimal binding sequences for this pro-
  • Future directions of alternative                                 My lab carried out the experiments      tein. So when it comes to unraveling
     splicing research                                           and supplied biological triplicate sam-     direct from indirect effects, we can look
                                                                 ples of both the knockdown and the          at the genomic context and whether a
Study design                                                     control for array analysis without          binding site is likely to be there.
Cline: Dr. Smith, why was EURAS-                                 revealing what the experiment was,              If you conducted an experiment like
NET interested in comparing microar-                             except that it was a knockdown of a         ours, but didn’t know what the binding
rays, and how was your laboratory’s                              splicing factor.                            site was, once you had a set of co-reg-
experiment chosen as the model?                                      Cline: Dr. Clark, you assembled a       ulated events, you could do the infor-
   Smith: I am involved in one of the                            set of predictions that were grouped by     matics and try to identify motif
EURASNET work programs that                                      confidence level. What did you use to       enrichment in and around those co-
focuses on finding global methods for                            assess the confidence?                      regulated exons.
analyzing alternative splicing. Within                               Clark: On the exon array, the signal        In fact, one of the other EURAS-
that program, there is a project aimed at                        intensity for any given exon is the com-    NET members who has been collabo-
identifying the best microarray platform                         bination of transcription and splicing—     rating with us has already made some
for investigating alternative splicing. The                      how many transcripts are being made         surprising discoveries.

2        S U M M E R 2 0 0 8 ■ AMB I N T ERV I EW R EP R I N T
   For instance, he has found significant     used rather than a more straightforward          standing of alternative splicing?
enrichment of particular motifs within        path of just turning off transcription               Clark: Analysis methods for
the exons. Before this discovery, people      of that gene. It’s interesting to imagine        extracting alternative splicing informa-
thought the binding sites of this protein     how these sorts of different mecha-              tion from microarray data are still in
                                                                                               their infancy. Being able to run one of
                                                                                               these experiments and use an analysis
“It’s interesting to imagine how these sorts of different                                      method similar to what we are able to
                                                                                               do today for studying gene expression
  mechanisms evolved. Obviously there is some reason                                           would be great. This would give us a list
                                                                                               of hits that we are confident represents
                                                                                               real alternative splicing events. Then,
  why the cell does it, but that is still a mystery.”                                          being able to take that information and
                                                                                               figure out which isoforms are being
were usually in the upstream intron or in     nisms evolved. Obviously there is some           expressed from the gene would be ter-
the flanking intron. But having this large    reason why the cell does it, but that is         rific. The methods for doing this are
data set, we can clearly see that the exons   still a mystery.                                 still in development.
themselves are enriched in binding sites          Smith: My take on that would be                  Smith: Ideally, you want an array
for this protein.                             that if you’re an RNA binding protein            platform to generate a list of predic-
    Again, the rather bold, long-term         or a gene for an RNA binding protein,            tions of alternative splicing events in
aim here is to be able to use the genom-      the best way you can control yourself            which you have high confidence but
ic sequence, to determine the tissue          or other members of your family or               that is also as inclusive as possible. The
specificity of a particular splicing event    neighbors is at a post-transcriptional           data analysis procedure that Tyson
or whether the event is inducible in          level. You are not going to be much              implemented was ideal in the first
response to particular signaling path-        good at binding to DNA and affecting             instance, because it generated a “high-
ways or affected by particular drugs.         transcription, but it makes sense to             confidence” set of alternative splicing
    I think we are a long way from that.      feedback at the RNA level.                       predictions with a very low false posi-
But by having a very reliable array plat-         It goes beyond splicing. There are a         tive rate. However, at the moment there
form that can quantitatively measure          lot of RNA binding proteins that auto-
changes in splicing, we can either look       regulate. Poly-A binding protein binds to
for tissue-specific changes or changes        a poly-A tract in its 5’ UTR and it halts
in response to knocking out a factor or       translation. I think feedback regulation is
adding a drug. Then we hope to go             widely used in many biological contexts.
back and do the informatics in order to           Cline: What advice would each of
determine the code in the primary             you give to researchers who are just
sequence that makes particular exons          starting to study alternative splicing
sensitive to these changes.                   using microarrays?
    Cline: Some researchers say that              Clark: One microarray experiment
alternative splicing is really nothing more   is going to generate a lot of data. While
than another mechanism to turn off            the experiment itself takes just a couple
protein production. Would either of you       of days, you might spend months or
care to comment on this assessment?           years following up on the data. With
    Smith: It’s true that some splicing       that in mind, carefully planning your                 Melissa Cline, Ph.D.,                 is a
events are on and off switches. But           experiment upfront will save you lots of
                                                                                                   senior informatics postdoctoral fellow in the
there are many examples of where that         time in the long run.
is not the case. Just recently there have         For example, adding biological repli-            laboratory of Manny Ares at the University
been papers on the Drosophila Down            cates can improve your ability to detect             of California at Santa Cruz. She previously
syndrome cell adhesion molecule,              alternative splicing with higher confi-
                                                                                                   worked as a postdoctoral researcher on the
DSCAM. It has more than 38,000 iso-           dence. Also, a clear understanding of
forms. These papers show that each            the limitations of your analysis method              Cytoscape Project at the Pasteur Institute
isoform seems to have a functionally          will make it easier to identify potential            and was a staff scientist at Affymetrix, Inc.,
important role.                               false positives.
                                                                                                   where she helped develop a prototype chip
    Clark: In terms of evolution, it’s
interesting to think about why a mech-        Looking to the future                                set for the GeneChip® Human Exon Array.
anism that involves splicing for turning      Cline: What technological develop-
off expression of a protein would be          ments might help advance our under-

                                                                                            A MB INTERVIEW REPRINT ■ SUMM E R 2 0 0 8               3
is a trade-off. The need to minimize the                         biology to try to affect splicing or to             A F F Y M E T R I X
false positives in a highly curated data set
also leads to false negatives. We are
                                                                 find small compounds that would free
                                                                 spliceosomes at particular steps or to              MICROARRAY
losing a lot of information along the way.
If you are looking for events where there
                                                                 look for small molecules that can affect
                                                                 particular splicing patterns and might
is overlapping regulation from individual                        be developed as therapeutics.
                                                                                                                                Editorial Staff
factors, you may find that you complete-                             There is also a program to look at
                                                                                                                     Rachel Steger, Founding Editor
ly miss the overlap among the false neg-                         the interaction between transcription
atives. So there is plenty of scope for                          and alternative splicing. That is clearly
                                                                                                                     Bronwyn Barnett, Managing Editor
refining the analysis procedures to                              very important. It’ll be interesting to
reduce the extent of the trade-off.                              see to what extents regulation at the
                                                                                                                     Sven Brown, Senior Editor
    Cline: What future plans does                                transcription level and elongation and
                                                                                                                     Stephen Antonopoulos, Photo Editor
EURASNET have for this technology?                               chromatin modification are the direct
                                                                                                                     Stacey Ryder, Associate Editor
    Smith: As a result of this pilot                             effectors of alternative splicing. There
                                                                                                                     Sabrina Banwait, Associate Editor
experiment, I expect to see more groups                          are clear examples where that occurs.
                                                                                                                     Marva Maida, Graphic Designer
interested in using similar approaches to                            It’s very likely that array technology
                                                                                                                     Jim Butler, Graphic Designer
look at targets of splicing factors. In                          will be applicable to some of these
fact, one of the other groups has already                        other programs, too.
published an exon array analysis of                                  Cline: Dr. Clark, how do you antic-
hnRNP-L-regulated alternative splicing                           ipate alternative splicing technology
events. We would also like to start look-                        being used in the next few years?
ing at co-regulation—looking for events                              Clark: There are so many different        association of mis-splicing or changes
that are regulated by combinations of                            potential uses for it. There is a lot of      in alternative splicing being either a
factors. It will be interesting to see how                       evidence that mistakes in alternative         direct cause of disease or associated
well exon arrays can investigate this in                         splicing can lead to human disease. I         with particular diseases. EURASNET
view of the trade-off we spoke about,                            think a better understanding of that is       has a program dedicated to mis-regulat-
because this type of experiment would                            in our future, as well as just a more gen-    ed alternative splicing and disease.
need an analysis approach with a low                             eral understanding of how alternative         Having a platform that is able to take a
level of false negatives.                                        splicing is regulated. A lot of the work      complete look at the splicing landscape
    We’ve been talking about a program                           that Chris is doing is going to take us       between normal and disease samples
to look at more global methods for                               down that path.                               would be very useful, and I am sure we
analyzing alternative splicing. There are                            Smith: That’s a very good point.          will see many developments in this area
various other programs using chemical                            There is increasing evidence of the           in years to come.

                                                            F O R M O R E I N F O R M AT I O N
    Contacts                                                     Center for Molecular Biology of RNA           ■   Didier Auboeuf -
    ■ Chris Smith, Ph.D.                                         University of California at Santa Cruz            www.eurasnet.info/didier_auboeuf.shtml
    Department of Biochemistry                                   Santa Cruz, CA 95064                          ■   Albrecht Bindereif -
    University of Cambridge                                      cline@biology.ucsc.edu                            www.eurasnet.info/albrecht_bindereif.shtml
    80 Tennis Court Road
    Cambridge                                                    Companies                                     Further Reading
    CB2 1GA                                                      ■ Affymetrix, Inc. – www.affymetrix.com       ■ Spellman R., Llorian M., Smith C. W.

    United Kingdom                                               ■ Partek – www.partek.com                     Crossregulation and functional redundancy
    cwjs1@mole.bio.cam.ac.uk                                                                                   between the splicing regulator PTB and its par-
                                                                 Organizations                                 alogs nPTB and ROD1. Molecular Cell 27(3):420-
    ■  Tyson Clark, Ph.D.                                        ■ EURASNET – www.eurasnet.info                34 (2007).
    Senior Scientist                                             ■ Cambridge University – www.cam.ac.uk        ■ Spellman R., Smith C. W. Novel modes of

    Affymetrix, Inc.                                                                                           splicing repression by PTB. Trends in Biochemical
    3420 Central Expressway                                      People                                        Sciences 31(2):73-6 (2006).
    Santa Clara, CA 95051                                        ■ Gil Ast - www.eurasnet.info/gil_ast.shtml   ■ Hung, L.-H., Heiner, et al. Diverse roles of

    tyson_clark@affymetrix.com                                   ■ Ben Blencowe -                              hnRNP L in mammalian mRNA processing: a
                                                                     www.utoronto.ca/intron/blencowe.html      combined microarray and RNAi analysis. RNA
    ■ Melissa Cline, Ph.D.                                       ■   Juan Valcárcel -                          (14) 284-296 (2008).
    Senior Informatics Postdoctoral Fellow                           www.eurasnet.info/juan_valcarcel.shtml    ■ Hattori D., et al. Dscam diversity is essential for

    Department of Molecular, Cell and                            ■   Stefan Stamm -                            neuronal wiring and self-recognition. Nature
    Developmental Biology                                            www.eurasnet.info/stefan_stamm.shtml      13;449(7159):223-7 (2007).

4        S U M M E R 2 0 0 8 ■ AMB I N T ERV I EW R EP R I N T

To top