1. technology and computing
2. consumer electronics
3. telephones
4. mobile phones
5. smart phones

On Generation and Applications of High-Density
Protein Microarrays
Ronald Sjöberg
Doctoral Thesis
Kungliga Tekniska Högskolan
Royal Institute of Technology
School of Biotechnology
Stockholm 2015
© Ronald Sjöberg 2015
Affinity Proteomics
Science for Life Laboratory
Division of Proteomics and nanobiotechnology
School of Biotechnology
KTH, Royal Institute of Technology
Tomtebodavägen 23A
171 65 Solna, Sweden
ISBN 978-91-7595-573-7
TRITA-BIO Report 2015:12
ISSN 1654-2312
Printed by US-AB
Abstract
Affinity proteomics has experienced rapid development over the last two decades and one of
the most promising platforms to emerge are the protein microarrays. The combination of
affinity reagents and miniaturisation enables assays for simultaneous high throughput and
sensitive protein analysis. Due to the combination of these desirable properties, a multitude
of protein array platforms for rapid and efficient study of proteomes and protein interactions
are in use today. Although the protein microarray field has more than two decades of history
to look back on the development of new protein microarray platforms continues to this day
and beyond.
In the paper I in this thesis, a microarray of eluates from dried blood spot samples collected
from neonates were designed and utilised for detection of complement factor 3 (C3)
deficiency. The data acquired from the microarrays platform were compared to C3 levels
obtained through enzyme-linked immunosorbant assay (ELISA), and the microarray assay
were found to separate the C3 deficient samples from the controls. The conclusion of this
investigation was that the microarray platform would be suitable for high-throughput
screening of C3 deficiency in neonates. Paper II outlines the work in developing a multiplex
platform for validation of affinity reagents. A set of 398 affinity binders, originating from five
research groups, were profiled against 432 antigens and representing both polyclonal rabbit
antibodies, monoclonal mouse antibodies, and recombinant single-chain variable fragments.
Approximately 50% of the binders were found to preferably recognise their intended target
while 10% of the binders did not generate any, or low, signals with their respective targets.
For paper III, a reverse phase array (RPPA) platform using fluorescence-based detection of IgA
deficiency in over 2.000 samples where validated on a label-free detection system and ELISA.
The data from the label-free platform and the RPPA were found agree well with each other
while data from ELISA did with neither of them. It was found that the label-free platform
proved to be well-suited for detection of IgA in serum. Paper IV describes one of the world’s
largest protein microarrays containing 21.120 recombinant protein fragments. We describe
some of the possible applications of these large-scale arrays, such as binding profiles for the
validation of antibodies with 11.520 and 21.120 recombinant proteins, as well as screening
for autoimmunity in human serum samples.
i
Populärvetenskaplig Sammanfattning
Allt levande, från virus och bakterier, till växter, människor, och andra djur, består till stor del
av något som kallas proteiner. Dessa proteiner sköter om alla de processer som sker i och
mellan celler som gör att cellerna och de vävnader de ingår i, samt vi som tänkande och
levande varelser, kan leva och föröka oss. När en individ blir sjuk, eller på något sätt förändras,
så är det möjligt att se det som en förändring i proteinernas produktion och samverkan. Detta
gör att det är av mycket stort intresse att kunna studera uttrycket av proteiner. Det finns dock
ett mycket stort antal olika proteiner, i människan uppskattar man det till att vara över
hundratusen, varav många proteiner är väldigt lika varandra och därmed svåra att åtskilja.
Detta innebär att studiet av proteiners uttryck och deras samspel kan vara mycket utmanande,
framförallt i biologiskt härledda vätskor såsom blod där nästan alla proteiner finns närvarande,
fastän med mycket stora koncentrationsskillnader.
Ett sätt att ändå göra detta är att använda proteinernas förkärlek för att interagera med
varandra, och då framför allt en viss typ av proteiner som kallas för antikroppar. Antikroppar
är grupp av proteiner som produceras av ryggradsdjur och vars uppgift i kroppen är att urskilja
och känna igen specifika proteiner även i väldigt komplexa proteinblandningar som t.ex. blod.
Genom att använda sig av antikroppar, och andra biomolekyler som fungerar på ungefär
samma sätt, så kan man då studera proteinsammansättningen även i dessa komplexa
biologiska prover.
Eftersom det finns så många proteiner vill man ofta studera många proteiner samtidigt och
ett sätt att gör det är genom att använda sig av så kallade mikromatriser. En matris består av
kända enheter ordnade i rutmönster med kända positioner och en mikromatris är följaktligen
en matris på mikrometerskala. Vad det innebär i praktiken är att man använder specialiserade
instrument för att placera väldigt små mängder av olika proteiner på fasta underlag i små
matriser och på detta sätt så kan man utföra analyser på tiotusentals proteiner eller prover
samtidigt.
När man går ner i skala till mikroskalan, och samtidigt upp antal samtidiga analyser till
tusentals, så får man dels både speciella problem och samtidigt helt nya möjligheter jämfört
med traditionella analysmetoder. Det är arbetet med att dels lösa dessa problem samt
utforskande av dessa fantastiska möjligheter som denna avhandling behandlar.
ii
This thesis will be defended on June 12 2015 at 10:00 in Inghe lecture hall in “Widerströmska huset”
(Tomtebodavägen 18A, Karolinska Institute, Solna Campus) for the degree of “Teknologie doktor” (Doctor of
Philosophy, PhD) in Biotechnology at KTH – Royal Institute of Technology.
Respondent:
Ronald Sjöberg, MSc in Biotechnology
Affinity Proteomics, Science for Life Laboratory
Division of Proteomics and Nanobiotechnology, School of Biotechnology
KTH - Royal Institute of Technology, Stockholm, Sweden
Faculty opponent:
Dr. Michael Taussig
Protein Technology Group, The Babraham Institute, Cambridge, UK
Cambridge Protein Arrays Ltd., Babraham Research Campus, Cambridge, UK
Evaluation committee:
Ass. Prof. Dr. Anders Andersson
Environmental Genomics, Science for Life Laboratory
Division of Gene Technology, School of Biotechnology
KTH - Royal Institute of Technology, Stockholm, Sweden
Ass. Prof. Dr. Sara Lind
Mass-Spectrometry Based Proteomics, Science for Life Laboratory
Analytical Chemistry, Department of Chemistry – BMC
Uppsala University, Uppsala, Sweden
Ass. Prof. Dr. Lukas Orre
Clinical Proteomics Mass Spectrometry, Science for Life Laboratory
Department of Oncology-Pathology, Cancer Proteomics
Karolinska Institute, Stockholm, Sweden
Chairman:
Prof. Dr. Joakim Lundeberg
National genomics Infrastructure, Science for Life Laboratory
Division of Gene Technology, School of Biotechnology
KTH – Royal Institute of Technology, Stockholm, Sweden
Main Supervisor:
Prof. Dr. Peter Nilsson
Affinity Proteomics, Science for Life Laboratory
Division of Proteomics and Nanobiotechnology, School of Biotechnology
KTH - Royal Institute of Technology, Stockholm, Sweden
Co-supervisor:
Ass. Prof. Dr. Jochen M Schwenk
Affinity Proteomics, Science for Life Laboratory
Division of Proteomics and Nanobiotechnology, School of Biotechnology
KTH - Royal Institute of Technology, Stockholm, Sweden
iii
List of Publications
The presented thesis is based on the following articles, referred to by their roman numerals.
All articles are included in the Appendix.
Article I
Magdalena Janzi, Ronald Sjöberg, Jinghong Wan, Björn Fischler, Ulrika von Döbeln, Lourdes
Isaac, Peter Nilsson, Lennart Hammarström, (2009), Screening for C3 Deficiency in Newborns
Using Microarrays, PLoS One, 4(4):e5321, doi: 10.1371/journal.pone.0005321.
Article II
Ronald Sjöberg*, Mårten Sundberg*, Anna Gundberg, Åsa Sivertsson, Jochen M. Schwenk,
Mathias Uhlén, Peter Nilsson, (2012), Validation of Affinity Reagents Using Antigen
Microarrays, New Biotechnology, 15;29(5):555-63. doi: 10.1016/j.nbt.2011.11.009.
* Equal contribution
Article III
Ronald Sjöberg, Lennart Hammarström, Peter Nilsson, (2012), Biosensor Based Protein
Profiling on Reverse Phase Serum Microarray, Journal of Proteomics & Bioinformatics, 5: 185189. doi: 10.4172/jpb.1000233.
Article IV
Ronald Sjöberg, Cecilia Mattsson, Eni Andersson, Cecilia Hellström, Heng Zhu, Mathias Uhlén,
Jochen M. Schwenk, Burcu Ayoglu, Peter Nilsson, Exploration of High-density Protein
Microarrays for Antibody Validation and Autoimmunity Profiling, (manuscript).
iv
Author’s Contribution to the Included Articles
The contribution of the author of this doctoral theses to the included articles is as follows:
Paper I
Experimental planning, development and production of the array platform, analysis of the
data obtained from the array platform and manuscript writing as co-responsible author.
Paper II
Study design, experimental planning and performing array-based laboratory work, data
analysis and manuscript writing as co-responsible author.
Paper III
Experimental planning, development and production of arrays for both array-based
platforms, performing all of the label-free array-based laboratory work, all data analysis, and
manuscript writing as main contributor.
Paper IV
Experimental planning, development and production of protein fragment arrays, all the
laboratory work and data analysis, and writing as the main contributor.
v
Related Articles
Affinity proteomics discovers decreased levels of AMFR in plasma from Osteoporosis patients.
Qundos U, Drobin K, Mattsson C, Hong MG, Sjöberg R, Forsström B, Solomon D, Uhlén M, Nilsson P,
Michaëlsson K, Schwenk JM. Proteomics Clin Appl. 2015 Feb 16. doi: 10.1002/prca.201400167.
A lateral flow protein microarray for rapid and sensitive antibody assays.
Gantelius J, Bass T, Sjöberg R, Nilsson P, Andersson-Svahn H. Int J Mol Sci. 2011;12(11):7748-59. doi:
10.3390/ijms12117748.
A roadmap to generate renewable protein binders to the human proteome.
Colwill K, Persson H, Jarvik NE, Wyrzucki A, Wojcik J, Koide A, Kossiakoff AA, Koide S, Sidhu S, Dyson
MR, Pershad K, Pavlovic JD, Karatt-Vellatt A, Schofield DJ, Kay BK, McCafferty J, Mersmann M, Meier
D, Mersmann J, Helmsing S, Hust M, Dübel S, Berkowicz S, Freemantle A, Spiegel M, Sawyer A, Layton
D, Nice E, Dai A, Rocks O, Williton K, Fellouse FA, Hersi K, Pawson T, Nilsson P, Sundberg M, Sjöberg R,
Sivertsson Å, Schwenk JM, Takanen JO, Hober S, Uhlén M, Dahlgren LG, Flores A, Johansson I, Weigelt
J, Crombet L, Loppnau P, Kozieradzki I, Cossar D, Arrowsmith CH, Edwards AM., Gräslund S. Nat
Methods. 2011 May 15;8(7):551-8. doi: 10.1038/nmeth.1607.
Validation of serum protein profiles by a dual antibody array approach.
Rimini R, Schwenk JM, Sundberg M, Sjöberg R, Klevebring D, Gry M, Uhlén M, Nilsson P. J Proteomics.
2009 Dec 1;73(2):252-66. doi: 10.1016/j.jprot.2009.09.009.
Selective IgA deficiency in early life: association to infections and allergic diseases during childhood.
Janzi M, Kull I, Sjöberg R, Wan J, Melén E, Bayat N, Ostblom E, Pan-Hammarström Q, Nilsson P,
Hammarström L. Clin Immunol. 2009 Oct;133(1):78-85. doi: 10.1016/j.clim.2009.05.014.
Anoctamin 2 protein as an autoimmune target in multiple sclerosis.
Burcu Ayoglu, Nicholas Mitsios, Ingrid Kockum, Mohsen Khademi, Björn Forsström, Ronald Sjöberg,
Johan Bredenberg, Izaura Lima, Eric Holmgren, Hans Grönlund, André Ortlieb Guerreiro Cacais, Nada
Abdelmagid, Mathias Uhlén, Tim Waterboer, Lars Alfredsson, Jan Mulder, Jochen M. Schwenk, Tomas
Olsson, Peter Nilsson (submitted).
Affinity proteomics in plasma within Hodgkin and diffuse large B-cell lymphoma identifies proteins
for disease identification.
Frauke Henjes, Claudia Fredolini, Kimi Drobin, Davide Tamburro,Mun-Gwan Hong, Björn Forsström,
Ronald Sjöberg, Elin Birgersson,Eni Andersson, Henrik Hjalgrim, Ingrid Glimelius, Janne Lehtiö, Jacob
Odeberg, Mathias Uhlén, Bengt Glimelius, Peter Nilsson, Karin E. Smedby and Jochen M. Schwenk (In
revision)
vi
Preface
Many years ago I was standing in a dusty construction site looking at a steel beam that I, then a
carpentry apprentice, and my team partner, the carpentry master, were supposed to fire-proof
according to set specifications. Not only did this require a specific set of fire-proofing goals to be
reached, but those goals had to be reached while still making the beam fit in to the space defined by
the blueprint. Due to the fact that those who makes blueprints quite often live in a different world
than those who has to make the blueprints become reality some problem solving was needed to
achieve the set goals. As my master were complaining about the impossibility of the task ahead I was
silently contemplating how construction work is mostly problem solving and how some people got
paid a lot more for solving problems, and without having to get dusty and dirty every day and retire
early from worn out backs, shoulders and knees. As I explained to my master how the work should be
done I was slowly starting to realise that maybe I should do something else in my life, maybe something
like engineering. At that time the construction business was in a slump and I kind of enjoyed the
downtime that was common when between jobs, as that gave me plenty of time to read interesting
books instead of breathing asbestos filled air in badly run construction sites. But nothing lasts forever
and once I myself became a carpentry master and when the construction business once more started
to get busy I went back to school to see what else there was to learn, and if I could get enough grades
to enter university to escape the harsh reality of manual labour. I already had my sights set on
engineering but I had not decided on whether it would be in mechanics or chemistry. But as I got a
lower grade in physics, due to the physics teacher getting a heart attack during test grading and my
test never being found by the substitute teacher, I decided that chemistry was the way to go.
After a few interesting years at university I finished my master thesis at the School of Biotechnology at
KTH and started working in what then was called the “PrEST Array group” where we were slowly trying
to build up the protein microarray platform for validation of Human Protein Atlas antibodies. This
supplied me with ample opportunities to do some interesting problem-solving besides the mindnumbing routine antibody dilutions, and although I was constantly planning to leave Stockholm and
return to my home town I never seemed to be able to leave. I guess the longer and darker winters and
my lack of interest in heavy industry, snowmobiles, winter sports, or hunting contributed to my
extended stay in Stockholm.
I quickly discovered that working with a technology that is so widely useful seem to always supply you
with new ways of being creative and to discover new amazing possibilities, especially if you are free to
improvise and to perform whatever test you wish without the need of immediate publishable results.
Over the years I acquired the knowledge and knowhow, which have now resulted in this short text, in
an almost organic way and only a small part of it could be fit between these book covers. The most
important knowledge that I have acquired since I started working with arrays is not really the technical
details concerning microarrays, but rather a way of thinking that I was never taught in school.
Unfortunately that is not easily translated into text, so general technical details of protein microarrays
will have to do. Owing to the diverse nature of the field of protein microarrays I can only describe a
small part of it. However, I have tried to include the three microarray variants that I have come in direct
contact with, but this is in no way a complete description. I hope that you will enjoy reading this book
just as much as I enjoyed writing it, because life should be about having fun.
vii
Table of Contents
Abstract .................................................................................................................................................... i
Populärvetenskaplig Sammanfattning .....................................................................................................ii
List of Publications...................................................................................................................................iv
Author’s Contribution to the Included Articles ........................................................................................v
Related Articles ....................................................................................................................................... vi
Preface.................................................................................................................................................... vii
Introduction ............................................................................................................................................ 1
From DNA to Proteomes ........................................................................................................................ 2
Proteins ............................................................................................................................................... 3
Proteome ............................................................................................................................................ 4
The Complexity of Proteomes ........................................................................................................ 4
Affinity Reagents .................................................................................................................................... 5
Antibodies ........................................................................................................................................... 6
Antibody Isotypes ........................................................................................................................... 7
Polyclonal Antibodies ..................................................................................................................... 8
Monoclonal Antibodies .................................................................................................................. 9
Antibody Derivatives .......................................................................................................................... 9
Other Affinity Reagents .................................................................................................................... 10
Proteomics ............................................................................................................................................ 11
Affinity Proteomics ........................................................................................................................... 11
Human Protein Atlas..................................................................................................................... 13
Protein Microarrays .............................................................................................................................. 15
Protein Microarray Formats ............................................................................................................. 17
Antigen Arrays .............................................................................................................................. 18
Capture Arrays .............................................................................................................................. 19
Reverse Phase Arrays ................................................................................................................... 21
Technical Considerations for Protein Microarrays .............................................................................. 22
Contact Printing ................................................................................................................................ 23
Non-contact Printing ........................................................................................................................ 25
In Situ Synthesis ............................................................................................................................ 26
Substrates ......................................................................................................................................... 27
Detection Principles .......................................................................................................................... 30
Label-based Assays ....................................................................................................................... 30
viii
Label-free Assays .......................................................................................................................... 32
Blocking and Background ................................................................................................................. 33
Local background .......................................................................................................................... 33
Global Background........................................................................................................................ 34
Multiplexing ...................................................................................................................................... 36
Image Analysis .................................................................................................................................. 38
Concluding Remarks ............................................................................................................................. 40
Present Investigations .......................................................................................................................... 43
Paper I - Screening for C3 Deficiency in Newborns Using Microarrays. ......................................... 43
Aims of the investigation.............................................................................................................. 43
Summary of findings ..................................................................................................................... 43
Results and conclusions ................................................................................................................ 44
Paper II - Validation of Affinity Reagents Using Antigen Microarrays............................................ 45
Aims of the investigation.............................................................................................................. 45
Summary of findings ..................................................................................................................... 46
Results and conclusions ................................................................................................................ 47
Paper III - Biosensor Based Protein Profiling on Reverse Phase Serum Microarray. ..................... 48
Aims of the investigation.............................................................................................................. 48
Summary of findings ..................................................................................................................... 48
Results and conclusions ................................................................................................................ 49
Paper IV - Exploration of High-density Protein Microarrays for Antibody Validation and
Autoimmunity Profiling. ................................................................................................................... 51
Aims of the investigation.............................................................................................................. 51
Summary of findings ..................................................................................................................... 51
Results and conclusions ................................................................................................................ 52
Future Perspectives .............................................................................................................................. 54
Acknowledgements .............................................................................................................................. 55
References ............................................................................................................................................ 56
ix
Introduction
Introduction
Most organism that we usually think about as “alive” consists of one or more cells. These cells
are composed of a lipidlayer encapsulating a biomolecular machinery and the information
describing how the biological machinery should be constructed and function.
This information about what the cell is capable of is stored in genes encoded in what is called
the genome. The genome is, on a basic level, both the beginning and the end of the organism
as it would not exist without it, and once the cell has seized to exist new copies of the
information will have been transferred to new generations of the cell. Even though the
information regarding the structure and function of the cell is stored in its genome, what the
cell does, and how the information is ultimately expressed, is most accurately gleaned through
study of its biomolecular machinery. In the end, the activity and regulation of this
biomolecular machinery is dominated by what we call proteins.
What this means is that as biomolecular changes happen in a cell, it will be possible to detect
it as changes in the protein content of the cell. As cells and tissues grow, age, or become
afflicted by disease, the changes in protein composition and concentration will be an indicator
of what is happening inside the organism. This window into life and death makes the study of
protein content of cells, tissues, or bodily fluids very interesting to anyone who strives to
understand more about the biology of life.
If the goal is to measure the general protein content of a biological sample, a classical assay
such as a Bradford assay, SDS-gel assay, or enzyme-linked immunosorbent assay will probably
be able to deliver what is needed. But if the goal is to measure samples and proteins on a large
scale, and enter proteomics, the use of affinity proteomics in the form of protein microarrays
is a powerful choice. These protein microarrays, and their construction and use, will be the
subject of this thesis.
1
From DNA to Proteomes
From DNA to Proteomes
DNA, or “Deoxyribose Nucleic Acid”, constitutes the information storage of cells, and was first
discovered as an unknown substance in the nuclei of cells by Friedrich Miescher in 1869. The
following work on the characterisation of this substance found it to be composed of four basic
components, adenine, thymine, cytosine and guanine, and that the common structure and
role of DNA is a double-stranded polymer1 that stores the hereditary information in cells2.
These components are called ‘nucleotide bases’ and are arranged in stretches that contains
sub-sequences called ‘genes’, and each gene can code for one or several functional products
made out of ‘proteins’. In the sequences between these protein-encoding genes resides
regulatory elements that controls the expression genes and other non-protein functional
products3. The genetic information is expressed by transcription into RNA (RiboNucleic Acid),
where thymine is exchanged for uracil. RNA can fulfil several roles in a cell, one being the
transfer of information on protein structure from DNA into protein. Different prefixes or
suffixes are used depending on which role the transcribed RNA fulfils, and mRNA (messenger
RNA) serves the role of being the messenger between information stored in DNA and
expression of protein4. Proteins on the other hand uses twenty different molecules, all
belonging to a class of molecules called amino acids. This increase in variation increases the
number of possible combinations and offers much larger variability in the structure of proteins
then is possible with DNA and RNA.
‘The central dogma’ describes the transfer of information stored in DNA through transcription
to mRNA and translation into proteins. It was first postulated by Francis Crick in 1958 and is a
negative statement saying that “once (sequential) information has passed into protein it
cannot get out again”, or in a more simplified way, that transfer of information from protein
back to DNA does not exist. DNA and RNA can both be replicated and in some cases
information stored in RNA can flow backwards to DNA, but in no cases can information stored
in proteins flow back to either RNA or DNA5. However, the transcription of DNA is in many
ways regulated by proteins and so the transcription, translation, and regulation finally
becomes a circle. This circle is an important part of what we call ‘life’ and the study of this
circle is incorporated into the research field called ‘life-science’, of which protein microarrays
is a part.
2
From DNA to Proteomes
Proteins
Proteins were recognized as a special class of multicomponent molecules as early as mideighteen century. However, a more accurate description did not appear until 1838, when a
Dutch chemist named Gerardus Johannes Mulder could show that this specific class of
molecules was mainly constructed by six basic elements: carbon, hydrogen, amine, oxygen,
phosphorus, and sulphur. The term ‘protein’ originates from the Greek word ‘proteios’ which
means ‘primary’, and was coined not long after by his associate Jöns Jacob Berzelius as a
reference to it being of foremost importance for the living body.
Proteins comprise 20 amino acids and each amino acid consists of a carbon scaffold with
amine and carboxyl functional groups, and a side chain specific for each type of amino acid.
Stringed together as chains these amino acids form the primary protein structure. The amino
acids in these chains interact with each other through hydrogen bonds within the chain,
folding the protein into secondary structures such as helixes and sheets. In the end the total
amount of interactions comprising of hydrophobic interactions, salt bridges, hydrogen bonds,
and disulphide bonds within the amino acid chain will make the chain, and its secondary
structures, fold into a tertiary structure that gives the protein its overall shape. The folded
proteins can then form protein complexes by assembling into multimeric quaternary
structures. Proteins can also be modified by posttranslational modifications, such as altering
the length of the protein chain, adding functional groups or otherwise changing the structure
of the protein. The finished proteins then interact with each other both within and between
cells. These protein-protein interactions then provide physical structure for the cells, performs
enzymatic reactions, and relays signals within and between cells.
3
From DNA to Proteomes
Proteome
The proteome is the entire set of proteins expressed by a cell or organism at a specific time
and place. The origin of the word is a portmanteau of protein and genome, where genome has
its origin in the German word ‘genom’ which in turn originates from the Greek word ‘gene’,
meaning ’creation’ or ‘birth’. The word ‘proteome’ has been derived as such on the basis of
the proteins being expressed by the genome in the organism6, and the human genome
comprise approximately 20.000 genes that are annotated to encode for proteins.7 Most cells
are believed to express a large fraction of the protein-coding genes8 and significant correlation
between levels of transcription and expression have been observed.9 However, what
differentiates most organs and tissues in the body is that their cells differ from each other on
their levels of expression and post-transcriptional modifications of proteins. These differences
of expression levels and modifications is what gives otherwise similar cells different
phenotype and function, and this means that the proteome differs depending on the local
milieu. Accordingly, there can exist a multitude of proteomes in the same organism, even
though every cell in the organism has the same genome. This property of protein expression
and modification makes the study of proteomes of great interest in the field of life-sciences.
The Complexity of Proteomes
Many interesting applications of protein research involve measuring protein levels in complex
biofluids such as blood or tissue lysates. Samples derived from blood, i.e. serum and plasma,
are of high interest to life science researchers due to the fact that it is readily available for
sampling and comes into direct contact with almost all of the body. Any changes that occur
anywhere in the body should then be detectable in the circulating blood as a change in the
blood-proteome. These changes can be difficult to detect since proteins that have leaked out
of cells, due to disease or other damage, might be present in only a small fraction in the blood.
Finding and quantifying these interesting, but low abundant proteins in complex biological
samples can be a challenging task, since the protein concentration might range over ten orders
of magnitude.10
This complexity and large variation in abundance of proteins can cause interesting proteins to
become masked from detection by the noise generated by high-abundant protein species.
4
Affinity Reagents
One way of combating this complication of a complex proteome is to deplete complex samples
of the most prevalent proteins, or perform separation or fractionation of the constituent
proteins. This lowers the complexity and hopefully enables analysis of the low abundant
proteins species. Although this might seem like a suitable solution for single samples, it can
become labour intensive and impractical when sample collections in hundreds. An even more
problematic outcome that has to be considered is that it could introduce bias in the final
representation of the protein content, as has been shown for depletion.11 Extensive sample
preparation should therefore preferably be avoided if possible.
Affinity Reagents
Affinity reagents are compounds that bind specific substances and that are used to detect or
affect their specific targets in biochemical contexts. They can be derived from various sources
and comprise several different classes of molecules. The most commonly type of affinity
reagents are antibodies, but antibody-derived affinity reagents such as single-chain variable
fragments and Fab-fragments have also been developed and are of growing use. Other types
of affinity reagents can be based on other classes of proteins and peptides, or even
oligonucleotides. The type of backbone used for a reagent can greatly affect its chemical and
biological properties and it should thus be possible to design and produce reagents that fits
specific purposes. Even though a multitude of affinity reagents are available truly wellvalidated affinity reagents can be hard to come by12.
An ‘antigen’ is a substance that is capable of binding to antibodies produced by the immune
system in antibody-producing organisms. The word itself is an abbreviation of “antibody
generator” as some antigens can provoke an immune response that causes production of
antibodies. Although the origin of antigens are related to antibody production by a immune
system, the word is today used as a general catch-all for substances that are used as targets
for generating affinity reagents, regardless of whether those reagents are antibodies.
The production of antigens is often performed in cell cultures and often involves expressing
the protein as a fusion with another protein to aid in purification and recovery.13 This is not
always successful as the protein of interest may be insoluble or toxic to the host, causing
incomplete expression. To simplify the expression step, protein fragments is sometimes
5
Affinity Reagents
chosen to represent the full-length protein instead. However, properly folded full-length
proteins have been shown to result in a higher degree of conformation-specific reagents, 14
which can be desirable if the target protein retains its native conformation during the assay.
Antibodies
The concept of antibodies dates back to the beginning of humoral theory of immunity in 1890,
and still many think primarily of antibodies when they think of the concept of affinity reagents.
Antibodies are a class of protein commonly called immunoglobulins (Ig for short), and are part
of the immune system in all jawed vertebrates15.
Antibodies are produced in B-cells and are roughly Y-shaped molecules that are comprised of
two functional parts. One part on each ‘arm’ that recognises and binds to antigens, and a
second part that is responsible for binding to effector molecules and receptors present on
other cells in the immune system. These two parts of the antibody are generally referred to
as the ‘Fab-region’ and the ‘Fc-region’.
Figure 1: An antibody and its regions. “VH” and “VL” are variable regions of the heavy and light chains. These, here shown as
blue modules, will together with the constant region of the light chain (“CL”, light grey module) and the first constant region
of the heavy chain (“CH1”, top dark grey module) build-up the antigen binding part of the antibody, or the “Fab-region”. The
second and third constant chains of the heavy region, the four lower dark grey regions, will form the “Fc-region”, which
activates biological functions. These two regions, the recognition and the activation regions, are connected through the hinge
region, here shown in brown. Variations in the hinge region and the Fc-region generates different antibody subtypes.
6
Affinity Reagents
Both the Fab-region and the Fc-region are comprised of products from multiple gene
segments and some of these gene segments are highly variable on the gene level. As the Bcells matures from precursor cells the variable regions will rearrange to produce new versions
of these variable gene segments. The protein products of these variable gene segments are at
expression combined with the protein products of constant gene segments to form the
finished antibodies. These variable regions form the epitope binding parts on the Fab-region
of the antibody and the genetic rearrangement results in the generation of B-cells that
produce antibodies that recognizes different antigens16.
The variable regions of the antibody thus ensures that the antibodies can recognise the wide
variety of possible antigens that is needed to be detected for effective immunity. Antibodies
commonly do not recognize the whole antigen but only a small part of it, a so called epitope.
These epitopes can be linear stretches of the antigen, or conformational structural epitopes
consisting of different parts of the antigen that are close to each other in the threedimensional structure of the antigen. This flexibility in epitope recognition further increases
the recognition possibilities of antibodies, and while the variable Fab-regions rearranges
during maturation to recognize different antigens the Fc-regions remains constant allowing
for the effector functions to be retained. However, by removal of parts of the constant gene
segments during cell division different versions of the Fc-regions can be produced by different
B-cell clones while retaining the same Fc-region. This gives the immune system a certain
degree of modularity in the combination between antigen recognition and effector function,
and the combination of different effector functions to the same antigen recognition results in
different antibody ‘isotypes’.
Antibody Isotypes
Different effector functions for antibodies with the same epitope recognition is achieved by
combining the Fab-region with a few different Fc-regions that carry different effector
functions. These combinations are referred to as isotypes, of which there are five main classes,
IgM, IgG, IgE, IgA, and IgD, and all of these fulfil different functions in the immune system17.
IgM can be found in all vertebrate species, and as such is the evolutionary oldest of the five
main classes of antibodies. It is the first antibody isotype to be produced during immune
7
Affinity Reagents
response, and it can be found as a dimer on the outside of B–cells, or secreted as a pentamer
into the circulatory system. In its membrane-bound form it regulates B-cell response and
survival as a B-cell receptor while secreted IgM can be found as both natural IgM and immune
IgM. Natural IgM is constantly produced, and is mainly responsible for triggering clearance of
apoptotic cells. Immune IgM is produced as a response to pathogen exposure and can, upon
recognition of a foreign invader, activate a part of the immune system called the complement
system to kill the invader. The antigens for secreted IgM are often lipids or polysaccharides
and often show broad specificity with low affinity, but due to their pentameric structure, high
valency18, 19.
IgG is the most abundant immunoglobulin in the blood circulatory system, and appears after
multiple antigen challenges have led to maturation of the antibody response. It is secreted as
a monomer and can be divided into four subclasses due to small variations in the Fc-region.
These subclasses are called IgG1, IgG2, IgG3, and IgG4, and all differ in the flexibility of their
hinge region, thus giving them different effector functions. IgG induces activation of the
complement system, and activates phagocytes that ingests foreign or harmful material20.
IgE binds to receptors on the outside of cells in the immune system, and can upon antigen
recognition stimulate these cells to release inflammatory mediators. IgE mostly confers
immunity towards parasites, but also plays a big role in allergic reactions21. IgA is found mostly
in the mucosa where it functions as a first line defence in the protection from bacteria and
virus, but can also be found secreted in blood, and similarly to IgG it can be found as two
subclasses22, 23. IgD is mostly found on the surface of B–cells, where it is co-expressed with
IgM, and is believed to control B-cell activation and suppression 24.
Polyclonal Antibodies
B-cell clones with different versions of the variable region will bind different epitopes on the
same antigen which results in pools of antibodies with different clonality25. Collections of
these antibodies that originate from different B-cell clones are therefore called ‘polyclonal
antibodies’. Polyclonal antibodies are a common reagent used for affinity proteomics, and are
usually produced by immunizing a host animal with the antigen of interest, causing the
immune system of the host to produce antibodies against the antigen25. The strength of
polyclonal antibodies is their ability to recognize many different epitopes of the same antigen,
8
Affinity Reagents
both linear and structural26. This effectively means that multiple antibodies can bind to the
same antigen which increases the probability of detecting the antigen. This has been shown
to lower the limit of detection and to make them less sensitive to masking of epitopes, thus
making them useful in more applications.25, 27
As each host produces their own pool of polyclonal antibody batch to batch variation is
introduced, since the pool of antibodies extracted from one host animal is not exactly the
same as the pool extracted from another host animal. This means that the renewability of
polyclonal antibodies is limited and they are therefore usually used in vitro and not in vivo due
to the difficulties in determining the properties of changing pools of antibodies with different
clonalities.
Monoclonal Antibodies
Monoclonal antibodies originate from the same B-cell clone, and consequently bind the same
epitope of the same antigen. They are produced in hybridoma cells, a fused hybrid of spleen
cells from a host animal immunized with the antigen of interest, and myeloma cells 28, 29. The
result of the fusion is an immortal cell line that produces identical antibodies, and as long as
the cell line is kept alive there is a constantly renewable source for antibodies with the same
singular clonality. Monoclonal antibodies are labour-intensive and expensive to develop but,
once established, can be renewed indefinitely. The monoclonality makes it easier to
determine their epitope binding properties, as only one epitope has to be characterised.
However, since they are dependent on the accessibility of one single epitope they may also
have limited functionality and not work in certain applications if that specific epitope becomes
masked or otherwise inaccessible.
Antibody Derivatives
The modularity of antibodies provides the basis of affinity binders derived from chosen
segments of full-sized antibodies. Instead of producing whole antibodies, fragments of
antibodies, can be produced and utilised as affinity reagents. The two most common variants
are ‘fragment antigen binding’ (Fab) fragment and the ‘single-chain variable’ (ScFv) fragment.
The Fab fragments consists of the Fab-region of the antibody without the Fc-region, while ScFv
is only the variable region of antibodies. These smaller parts of full size antibodies can be
9
Affinity Reagents
recombinantly produced in large bacterial or phage libraries through random recombination
of the genes, which makes them cheaper to produce than monoclonal antibodies. Their small
size potentially allow them to penetrate deeper into tissue, and their recombinant production
enables improved affinity and possibility to fuse them to other molecules30, 31.
Other Affinity Reagents
The quest for designed affinity binders does not stop with the antibody derived fragments and
there are several affinity molecules based around other protein moieties or DNA/RNA
molecules. These are often referred to as scaffold binders or small-molecule binders. Scaffold
binders are based on stable molecular scaffolds that has an integrated affinity function that
can be varied through combinatorial engineering to produce different binding characteristics.
32
An example of a scaffold binder is the Affibody molecule, which is based on protein A that
comprises a three helix bundle where the amino acids in two of the helixes can be changed by
random mutation in combinatorial libraries or by directed engineering to produce different
binding characteristics.33 Other examples of scaffold-based or small-molecule binders are
Anticalins (derived from lipocalins)34, Adnectins (derived from fibronectins)35, DARPins
(ankyrin repeat proteins)36 and Aptamers (oligonucleotides)37. These scaffolds and small
molecules can be engineered to be stable under environmental conditions that would not be
suitable for full-length antibodies and similarly to antibody fragments they can also be
selected to high affinity and combined with other molecules.
10
Proteomics
Proteomics
The large-scale study of a field in biology is often referred to as –omics, such as proteomics for
proteome-wide studies of the entire dynamic library of proteins, and all their modifications,
localization, interactions and expression levels. The proteome-wide mapping of the entire set
of proteins in all their variants, which may number in the hundreds of thousands with all
possible combinations, represents a unique challenge38.
The fields of genomics and transcriptomics have been experiencing a rapid growth in
discoveries and research after mass-producing copies of a DNA strands became achievable
through PCR. Consequently, the ease of access to an almost unlimited number of copies laid
the foundation for the enormous success that the field of genomics has achieved so far.
However, for proteomics there is not any equivalent to PCR, and the production of proteins
often becomes a bottleneck due to this. Each protein needs to be recombinantly produced or
synthesized as short peptides and the methods available for this are often complicated and
entail many steps in which the production or purification can fail. Common problems during
recombinant protein production are failure in cloning of vectors, antigens are toxic to the cells,
and incorrect folding or posttranslational modifications. This means that the production of
proteins can be both costly and time-consuming and these limitations represent an impressive
challenge if studies are to be performed on true proteome-wide scale.
Affinity Proteomics
Affinity proteomics has become an important tool for studying the expression, localization,
functions, and interactions of proteins using large-scale and high-throughput methods. This is
owed to the possibility that is offered by affinity reagents to analyse complex proteomes
without the need for extensive sample workup12, 39. By employing affinity proteomics, an
increased understanding on how cells and organisms function can be reached, and new drug
targets or diagnostic markers for diseases can be discovered. Many different affinity-based
proteomic methods for achieving this exist, and some classical affinity-based methods are
ELISA and Western blot. Although originally considered low-throughput, an increased
automation of laboratory work have led to higher throughput while using these classical
11
Proteomics
methods12. However, this increase in automation still do not allow for the throughput
necessary to perform truly large-scale affinity proteomics and therefore large-scale affinity
proteomics methods are today dominated by various microarray formats such as tissue
microarrays40, protein microarrays41, bead arrays42 and lysate arrays43.
One limitation of affinity proteomics is in the access to well-characterised affinity reagents,
and this negatively effects how efficiently the information in the proteome can be mined. In
order to address this limitation, and to enable researchers to investigate more of the
proteome, several projects aiming to produce affinity reagents have been started14, 44-46. The
lack of characterisation generally expresses itself as a problem through off-target interactions,
which is when an affinity reagent interacts with more than the intended target. These offtarget interactions generally increase in numbers as the complexity of the sample increase,
due to the increase in possible interaction partners. An affinity reagent that have been
produced to be specific for a certain antigen might very well interact with other available
antigens, especially if they are available in higher concentration then the intended target. The
severity of this is easily understood if one considers a sandwich assay. A sandwich assay
comprises a capture molecule, a target molecule, and a detection molecule, and in the ideal
case the capture and detection molecules will not have the same off-target interactions. For
a binding to be detected in such an assay both the capture molecule and the detection
molecule must have found the target, and off-target interactions will thus not be detected.
However, off-target interactions can occur between different detection molecules as well as
between detection molecules and captured molecules if more than one pair of capture and
detection molecules are used, as an addition to the possible off-target interactions between
capture molecules and target molecules. Such an assay, of N targets, will have almost 4N2
possible interactions between capture, target, and detection molecules. For a 100-plex assay
this will then result in almost 40 000 different theoretical interaction pairs47. This complication
can to some extent be mitigated by employing single-binder assays, where the whole sample
is labelled, thus forgoing the need of a detection antibody. However, off-target interactions
can still be expected, especially when using polyclonal antibodies48. These off-target
interactions when multiplexing consequently becomes a problem that has been limiting some
areas in the field of affinity proteomics49-51.There is therefore a great need for well-
12
Proteomics
characterised affinity reagents that have been well validated for specificity in each assay
format that they are intended to be used in12.
Human Protein Atlas
One of the largest undertakings to generate affinity reagents for proteomic research has been
the Human Protein Atlas52 (HPA). The goal of the Human Protein Atlas has been to generate
high-quality antibodies towards all human proteins for a systematic exploration of the human
proteome. This is being achieved by employing high-throughput production of antibodies
against proteins derived from the approximately 20.000 protein encoding genes in the human
genome. These antibodies have been used to study the protein profiles of human tissues and
cells, as well as different body fluids53, 54 and have also been applied to non-human tissues55.
In order to achieve highly specific antibodies with preferably no off-target interactions a
strategy of using as unique part as possible of the targets for the generation of reagents have
been employed. This entails choosing protein fragments so that they represent a region of low
homology with less than 60% sequence identity similarity to any other human protein.
Transmembrane regions and signal peptides are also excluded, and no more than eight
identical amino acids within any ten amino acid stretch as compared to any other human
protein are allowed. The length of these fragments are on average around 80 amino acids,
which allows for generation of antibodies towards multiple possible epitopes. By targeting a
unique part each protein in this way it is possible to reduce the promiscuity of the antibodies
generated towards the protein, while retaining the flexibility of polyclonal antibodies and
ensuring that the antibodies will be applicable in to many different assays.
In HPA, the protein fragments are fused with a His6ABP-tag containing a six histidine part and
an albumin binding protein part. The histidine-tag allows for purification and detection with
tag-specific antibodies, and the albumin binding part increases the half-life of the antigen
during immunization. To obtain the polyclonal antibodies from the sera of the immunised
animals a two-step purification is used. First a depletion of tag-specific antibodies is performed
using affinity columns with the tag is immobilised. This is then is followed by purification of
the fragment-specific antibodies on a column with the protein fragment immobilised56-59. By
13
Proteomics
employing this dual affinity purification antibody-antigen pairs that can be utilised on a
multitude of different proteomic platforms within the HPA is obtained.
Figure 2: Distribution of fragment lengths. The distribution of fragment length of protein fragments produced as antigens for
antibody generation within the Human Protein Atlas. The protein fragments have a mean length of 81, and a median length
of 80 amino acids, and varies between 16 and 202 amino acids in length. The average weight of the protein fragments are 27
kDa, excluding the 18 kDa His6ABP affinity tag.
Currently the Human Protein Atlas contains a tissue atlas with immunohistochemical images
and information on human genes on both protein and mRNA level for all mayor organs. It also
comprises a cancer atlas for the most common forms of cancer, a cell line atlas with expression
profiles for a number of human cell lines, and a subcellular atlas with immunofluorescence
images for three cell lines52.
14
Protein Microarrays
Protein Microarrays
The first result of “microspot” in the PubMed database is from an article in Nature by Joseph
G. Feinberg that was published in 1961, “A ‘Microspot’ Test for Antigens and Antibodies”. In
this article a microscope cover-glass was coated with antiserum-agar, and microspots of
antigens were deposited using capillary tubes for detection of antigen-antibody precipitates60.
However, the birth of the first true microarray is usually dated to the work of Patrick O. Brown
et al. that was published in 1995 where they arrayed a set of 48 cDNAs from Arabidopsis
thaliana61. The concept of microarrays builds on the concept of the “ambient analyte
immunoassay” as described by Roger P. Ekins in 198962, and which effect on microarrays is
summarized in a review by Thomas O. Joos et. al63. In short, using a high concentration of
capture reagent in a small area results in highest possible local signal density, without affecting
the local concentration of analyte to any considerable degree. This means that the amount of
analyte captured directly reflects the concentration of the analyte in the sample, as the
measured signal is only dependent on the analyte concentration and the affinity constant of
the capture reagent. One major advantage of this is that the assay becomes independent of
sample volume, as long as there is enough analyte molecules present for the concentration to
not be affected in any significant way by the capture consumption. However, if the
concentration of the analyte is low while the affinity of the binder is high, equilibrium may
take unacceptable long due to constraints in mass transport64. The typical protein microarray
entail a first step of immobilising a protein, or a mix of proteins, on a solid phase substrate.
This is then followed by a primary interaction step where a liquid phase that contains an
interaction partner to one of the immobilised proteins is added. A second interaction step is
then often employed for detecting the primary interaction reagent, before a reporting step is
performed to quantify the signal from the reporter molecule.
15
Protein Microarrays
Figure 3: Molecular complex for detection of a target protein. Molecular complex of a mixed sample attached to a solid
phase substrate, containing a protein that is being recognised by a primary detection reagent, in this case an antibody. The
primary detection antibody is then recognised by a secondary detection reagent that carries a reporter molecule. This is a
basic protein microarrays setup and the generation of this complex follows the steps of arraying on the substrate, adding the
primary reagent, adding the secondary reagent, and finally detecting the reporter molecule.
Planar microarrays are manufactured by spotting small volumes of an analyte within a small
spatial area on a substrate such as a microscope slide to form grids of what is generally
referred to as ‘spots’ or ‘features’. These volumes are usually in the pico to nanoliter range,
and the distance between features and their diameter are usually in the micrometre range.
The low volumes, and the accompanying small spatial area that are needed for each feature,
means that large numbers of analytes can be arrayed and simultaneously analysed on one
single slide. This has predominantly been used for the analysis of gene expression by arraying
single stranded DNA or synthesized oligonucleotides and applying samples consisting of
labelled complementary single stranded DNA. However, the technique has also become
common for use in protein analysis65. The final result is often a false colour image of the array
with one or more colours representing different detection molecules, as can be seen in figure
4.
16
Protein Microarrays
Figure 4: A false-colour image of a protein microarray with 384 different proteins. Each feature in the array contains a
unique protein identity and is visualised using detection reagents specific for a small part of the protein that is common for all
unique protein sin the array. This results in a green signal showing the presence of proteins in each feature and aid in quality
control and image analysis. An antibody targeting one of the proteins have been added and can be detected as a red signal in
one of the features, in this case in the first column and third row of features.
Protein Microarray Formats
Not only protein-protein interactions can be analysed on protein microarrays. Protein
interactions with other possible interaction entities such as small molecules, liposomes and
carbohydrates, can also be investigated. The complexity of the analytes can vary depending
on if they are crude biological samples or pure reagents, and the multiplexing can lie on either
detecting many antigens in a few sample or few antigens in many samples. Protein
microarrays can consequently be used for a large number of applications and reagents. It also
means that the world of protein microarrays is a multidimensional world, which looks
different depending on the dimension that it is being projected on. Due to the variation in
possible microarray assays, variations in how to classify different protein microarrays assays
exists. However, the two main choices are to either refer to them based on what the goal of
the assay is or which phase that carries the analyte. The assays can either be analytical or
functional in their nature. The goal for analytical assays is to measure the concentration or
existence of an antigen in a sample, and to measure some biological activity for functional
assays. Therefore, one way of classifying protein microarrays is to refer to them as either
analytical protein microarrays or functional protein microarrays65-68. The other way of
classifying them is by referring to which phase, the solid substrate phase or the liquid phase
that the analyte is presented in. If the analyte is in the liquid phase of the assay it is regarded
17
Protein Microarrays
as a forward phase protein microarray and conversely as a reverse phase protein microarray
if the analyte is immobilized on the substrates solid phase69, 70.
I will here refer to protein microarrays as being either antigen arrays, where proteins are
arrayed either to be detected by affinity reagents targeted towards them or be investigated
for functional properties, or capture arrays, where affinity reagents targeting analytes are
arrayed. Alternatively I will refer to them as reverse phase arrays, which are then a more
complex case of the antigen array. This choice of nomenclature is due to the fact that capture
arrays are dependent on maintaining the structural integrity of the capture molecules in order
to preserve their function, while conformational integrity is not always necessary in antigen
arrays. Finally, the reverse phase arrays are unique in the complexity of the arrayed analytes
and as such represents a unique set of challenges.
Antigen Arrays
Antigen arrays can be used for a multitude of applications and assays due to the many
variations of assay that is possible. One early example is Michael Snyder et al. who generated
an array containing purified recombinantly produced gene-products from more than 90 % of
the S. cerevisiae genome, and investigated protein-protein interaction for calmodulin71. By
doing this they became the first group to show the power of antigen arrays in screening for
protein-protein interactions. Antigen arrays have since then been utilized for studying
posttranslational modifications and binding interactions in a multitude of applications72.
Antigen arrays can also be used for characterisation of antibodies by assessing their specificity,
accuracy, precision, and detection limit as performed by Brown et al73. Similarly this can also
be performed in a high throughput manner to validate the on-target interaction and profile
off-target interactions of affinity reagents74. Antigen arrays have also with great success been
applied for identification of targets for autoantibodies produced in autoimmune diseases and
immunodeficiency75. In this case the antigens, either known antigens for quantification of
autoantibodies or possible novel targets, are arrayed on a solid substrate and a biofluid
applied and analysed for its content of immunoglobulins76, 77. Similarly the immune response
towards pathogens can be measured by arraying antigens and detecting the presence of
antigen-specific immunoglobulins78.
18
Protein Microarrays
Production of peptide arrays using photolithography allows for ultra-high-density arrays,
which can cover the whole proteome even when overlapping peptides are used. One example
of this is the epitope mapping performed on ultra-dense peptide arrays comprising 2.1 million
6-mer -overlapping 12-mer peptides, which cover all of the human proteins. These arrays have
been successfully used for mapping the contribution of each amino acid to the binding of both
monoclonal and polyclonal antibodies, and to investigate specificity and cross-reactivity
across the whole epitome79.
While antigen arrays often are synonymous with protein or peptide arrays exceptions exists,
and one such that should be mentioned are the carbohydrate arrays. The study of
carbohydrates in cellular biology is a field of high interest to many research groups as
carbohydrates as biomolecules can be found both inside and outside of cells and fulfil various
biomolecular processes. However, due to the very large variety in branching and molecular
modifications of carbohydrates in biological systems, the cellular glycome is believed to be up
to 500.000 different structures. This large number of possible variations has made microarrays
an important technology for rapid analysis of carbohydrate-protein interactions80. The
production of carbohydrates for arrays are either done through chemical synthesis or through
natural sources. If the carbohydrates are large enough non-covalent adsorption through
electrostatic or hydrophobic interactions to a substrate is possible. However, if the
carbohydrates are small the resulting forces might be too weak to retain the carbohydrates
on the substrate during washing steps, and covalent attachment becomes necessary81, 82.
Carbohydrate arrays have been extensively used to investigate the carbohydrate biology and
its involvement in diseases, infection and cell-signalling80.
Capture Arrays
The variation in microarray capture assays, and the choice of technical approaches in labelling,
amplification, substrates and detection is large83. The main difference between antigen arrays
and capture arrays is that for capture arrays the arrayed capture molecules are generally
similar with active sites that need to accessible. In practise that means that immobilising
capture molecules, such as antibodies or Fab fragments, in an oriented approach with the
binding sites exposed to the liquid phase can result in up to ten-fold higher binding capacity
of the analyte in the samples, as compared to random immobilization84. The commonly used
19
Protein Microarrays
method to achieve oriented immobilisation with antibodies is the use of an intermediate
affinity protein, such as protein A or protein G, or to chemical modify the Fc-region84. Random
immobilisation of antibodies have been shown to cause conformational changes and loss of
function of the majority of the proteins85. This change in conformation could be assumed to
affect other types of proteins as well and therefore more emphasis has to be placed on
orientation of the arrayed proteins if the functionality of the proteins is important for the
assay. Further improvement to the retention of functionality can be seen in hydrogel-based
substrates, which often results in higher signal-to-noise ratios than flat 2D-surfaces86, 87, due
to higher binding capacity and improved hydration. An even further improvement of the
format is the bead-based assay42 that is performed in a liquid phase, thus maintaining the
hydration of the molecules and circumventing the loss of function due to denaturation.
While the common sandwich-based assay offers high sensitivity and specificity due to the dual
binding needed for readout, the sandwich-format is generally less compatible with
multiplexed array-assays, due to previously mentioned problem of off-target interactions
from detection reagents. The alternatives to using a sandwich assay is to label the whole
sample, and compare samples through included controls or a common reference, or do direct
comparison between samples. The reference design can be set up through a pool of equal
aliquots of all samples to be tested that is labelled using a different tag than the samples. This
pool can then be used as an internal normalization standard, similar to the commonly used
reference mix used for DNA microarray experiments11, 73, 86. This might seem as an attractive
approach for comparing different samples against each other as the relative abundance of
targets should be comparable through the reference pool. However, the sensitivity of such an
assay can be seriously hampered by competition for capture molecule binding sites that can
occur between sample and reference. This has led to widespread preference for the singlecolour approach where each sample is labelled with the same label, and compared directly or
through controls88. Although somewhat complex to set up, the capture arrays have become
an indispensable tool for biomarker discovery and have been applied in numerous disease
biomarker discovery platforms89.
20
Protein Microarrays
Reverse Phase Arrays
In a reverse phase assay the analytes are immobilized on the substrate and capture reagents
are applied in the solution phase. This approach allows for simultaneous analysis of up to tens
of thousands of analytes or samples but is sensitive to off-target interactions from detection
reagents. The reverse phase format has been extensively utilized for analysis of complex
samples such as tissue lysates90 and cell lines43, as well as for body fluids such as serum91 and
cerebral spinal fluid92. The goal of the analysis have often been the study of biomolecular
network pathways, but also for large-scale screening. Due to the complex nature of the
analytes or samples, the standard for reverse phase arrays is to array each sample in triplicates
with at least two different dilutions93. This increases the likelyhood that at least one dilution
will fall within the linear range of the detection method, even though the samples might have
variable concentration of the target molecule. However, this also limits the actual number of
samples that can be processed in one array, pushing it from thousands to hundreds, as
replicates and controls occupies spatial space.
Arraying and detecting targets in complex mixtures presents more challenges than samples
that contain only one or a few different proteins. The starting concentration of the samples
might be high and low-abundant target proteins might become masked by protein species of
higher abundance. The small volumes that are arrayed can result in few target molecules being
present in the feature if the starting concentration in the sample of the target is low. For the
lower abundant proteins the statistical chance of having the protein present in the spotted
volume consequently approaches zero even if no sample dilution is performed during
preparation. Low-abundant proteins might as a result be represented by only a few molecules
in each feature together with the complex mix of proteins of higher abundance. This results
in it becoming difficult to detect them among the unspecific background from off-target
interactions of the detection molecules, or by autofluorescence of the substrate. Another
challenge is that most substrates do not have the loading capacity to effectively retain all
proteins in a complex sample such as serum or plasma. The excess material can result in bleedoff of unbound material from the feature, which can cause an elevated background in the
vicinity of the feature94. Consequently, reverse phase assays often use high-capacity
substrates such as nitrocellulose95. The off-target interactions of detection molecules have to
21
Technical Considerations for Protein Microarrays
be addressed by choosing the detection molecules that have gone through extensive
validation as to their specificity and target-recognition, and autofluorescence can be
minimised by choosing a detection wavelength that does not overlap with the
autofluorescence of the substrate. The use of near-infrared scanners have therefore become
of prominent use in reverse phase array assays, owing to the fact the excitation and detection
wavelengths lies beyond that of autofluorescence of the popular nitrocellulose substrates96.
Technical Considerations for Protein Microarrays
Microarrays present a unique set of challenges due to the very minute volumes that are being
deposited and the small spatial scale of the features. Variables such as humidity and
temperature become important aspects when arraying such small amounts of liquid, and must
be taken into consideration to ensure that features are properly arrayed. The concentration
of arrayed proteins, buffer composition, and drying of the array greatly affects the
morphology and linear range of individual features. The better the alignment and morphology
of features can be controlled, the tighter they can be packed. A tighter packing increases the
achievable array density, which in turn increases the number of tests that can be carried out
simultaneously. Visualisation and quantification of the amount of protein immobilised in each
feature is often of great interest, both as a quality control and for normalisation purposes.
This can be achieved either by using a detection reagent that detects protein in general on
one array, and extrapolating the obtained results to other arrays generated during the same
production run. Another method is to produce the proteins as fusions to a tag that is
detectable using a tag-specific reagent. The tag can then be used to visualise or quantify the
total target content for each individual feature simultaneously as the analyte-specific test
through a dual-color approach. However, since a common tag might cause cross-reactivity
consideration must be made as to possible unwanted interactions towards the tag that might
result in increased background.
Instruments utilising different technologies for production of microarrays have been available
on the market or presented in research articles. However, the principles used by the most
common of these instruments can be roughly divided into the two categories of contact
printing and non-contact printing. Contact printing involves making physical contact between
the substrate and pins so that material is transferred from the pin to the substrate. Non22
Technical Considerations for Protein Microarrays
contact printing includes a diverse group of systems where some utilises inkjet technology to
eject droplets through the air onto the substrate and other systems use photochemistry to
build up peptides on substrate surfaces. The liquid volumes that are deposited are often in
the nanoliter range and should produce features that are morphological homogenous and
simultaneously spatially discrete and high-density. A mentioned previously, this presents a
considerable challenge, owing to the small volumes and the great impact of even minute
changes in the ambient environment or sample composition.
Contact Printing
The most common way of producing arrays on a flat substrate, such as a microscope slide, is
by contact printing with pin-printing instruments. The reason for their widespread use is the
initial adoption for DNA microarray-production, and that the experience gained from DNA
microarrays were later initially applied to protein printing97.
Feature uniformity in pin-printing is affected by the viscosity of the solution, the substrate
surface properties such as planarity, and pin properties such as pin velocity and pin design.
Viscosity is in turn affected by temperature, where higher temperature results in lower
viscosity, which affects the volume deposited on the substrate. The change in volume
ultimately impacts size and morphology of the arrayed features. Another factor that impacts
feature size is the hydrophobicity of the substrate and solution, as deposited drops tend to
increase or decrease their contact with the substrate depending on the hydrophobicity of the
phases. Changing to a more hydrophobic buffer for the solution can subsequently result in a
larger feature area on a hydrophobic substrate, and a more hydrophilic substrate can
conversely lead to an increase in feature size97.
The most common pins used in pin-printing are solid pins, split pins, and quill pins. The solid
pins are loaded with solution by dipping the pin in the solution and allowing the capillary force
to retain a small volume at the tip of the pin. Solid pins have a low loading capacity and are
only useful for small scale arraying as a single loading can only array a few spots before being
depleted. The strength of the solid pins is that they are not very sensitive to the viscosity of
the solution, and consequently can be used for high viscosity solutions such as protein
samples. They are also not especially sensitive to dust and debris and can be cleaned fairy
23
Technical Considerations for Protein Microarrays
easily98. A modification of the solid pin is the “pin and ring” system that uses a large capillary
that is loaded with sample and a solid pin that passes through the meniscus of the capillary
before making contact with the substrate, thus increasing the capacity and requiring less
reloading of sample97.
Split pins are a further development of the solid pins and have a microchannel in the tip that
the solution is loaded into through capillary forces. They can accordingly produce more
features before reloading of the solution becomes necessary due to the larger volume. A
variation of the split pin is the quill pin which, together with the microchannel, contain a small
reservoir for further increase in loading capacity. Although these split and quill pins allow for
higher throughput they are also sensitive to dust and clogging, as well as dependent on lowviscosity solutions for successful deposition on the substrate. They are also dependent on high
humidity being maintained during printing to prevent drying of the samples in the pin97, 99.
When using split pins the feature size may also vary during arraying due to the changing
volume in the tip. The initial features gets a large volume deposited due to draining of the
excess solution on the outside of the pins, and become large or merge. The last features
produced becomes small or might not be arrayed at all due to completely depleted pins100, 101.
To prevent this, pre-printing on slides to drain the excess volume before continuing arraying
is generally necessary with split and quill pins. Due to the variation in feature size care should
be taken to keep track of the order of which the slides have been printed, to ensure that arrays
that are being used for a single project are as similar as possible. These pins are usually made
of metals such as stainless steel, tungsten, or titanium and the solid pins are relatively
insensitive to damage. However, split pins and quill pins can easily be deformed if the pins
touches the substrate with too much force97. A further limitation of pins is that proteins tend
to adsorb to metal surfaces102, making metal pins difficult to clean. Taking it together, these
limitations in the pin-based technology makes contact printing less desirable for producing
protein microarrays. However, pin-based printers are still in extensive use for protein
microarray production due to their legacy from the DNA microarray era.
24
Technical Considerations for Protein Microarrays
Non-contact Printing
Non-contact printing involves no physical contact between the arraying tool and the substrate
surface. Instead, droplets of the solution are contained in a reservoir and ejected through a
nozzle at some distance above the surface. This can be achieved in different ways103, but the
most common is by piezo actuation. In arraying-instruments that utilizes piezo actuation a
deformation of a small reservoir holding the solution causes a pressure change in the
reservoir, which in turn results in the ejection of a controlled amount of solution from a
nozzle104. One advantage of non-contact printing is that it allows for samples of higher
viscosity due to the use of nozzles. This makes it easier to array protein samples, which can be
of high-viscosity and difficult to array using pin-based instruments. Another advantage is the
lack physical contact with the surface of the substrate, as this removes the need to make the
printhead travel in the z-axis. Consequently drops can be dispensed “on the fly”, greatly
speeding up the deposition of droplets. In addition, this eliminates the risk of damaging the
substrate due to physical impact, which is something that otherwise can happen to sensitive
surfaced such as hydrogels and nitrocellulose. The comparably large reservoirs used in noncontact printing also means that more droplets can be deposited before reloading. This
improves feature to feature reproducibility, and ensures that the first and last feature in a
print run will have the same size.
Although non-contact print nozzles commonly are manufactured from other materials then
metals, adsorption of proteins can still occur. Additionally, deposition of salt from buffers and
samples can affect the nozzles ability to dispense droplets. This might result in features that
are missing or out of alignment, carry-over of sample between loadings, and small satellite
spots105.
A third way of producing protein arrays is by synthesizing the proteins or peptides directly on
the substrate such as by photolithic synthesis. The principle of the technique is to combine a
UV-light with photomasks or digital micromirrors that can be controlled to selectively activate
small areas of the array to deprotect amino acids with photo-sensitive protective groups.
These deprotected amino acids then become available for reaction with new amino acids
carrying photo-sensitive groups, and thus addition of amino acids to specific parts of the array
can be achieved in an iterative manner106. This can generate ultra-dense arrays of peptides
25
Technical Considerations for Protein Microarrays
with over 2 million unique peptides on a 2 cm2 area. Such an array would be enough to
represent all human genome-derived proteins as 18-mer polypeptides with as much as 12mer overlap79, 107.
In Situ Synthesis
Recombinant expression and purification of proteins can be expensive, time-consuming, and
fraught with hurdles such as poor expression, protein degradation and aggregation67, 108. An
alternative way is to express proteins directly on the substrate instead of performing
conventional expression and purifiaction109. This is usually referred to as in situ synthesis, or
cell-free protein array production. The two main methods for this are the “protein in situ
array”108, 110 and the “nucleic acid programmable protein array”110, 111, or in short; the PISA
and the NAPPA arrays. By generating the proteins of interest directly on the substrate, the
problem with solubility and functionality can to a large extent be alleviated.
For the PISA technology this is achieved by using the DNA sequence for the protein of interest,
together with a N- or C-terminal hexahistidine tag that allow for binding to an immobilisation
agent that is precoated on the substrate. These are then arrayed together with a cell-free
expression system that contains all the essential elements for transcription and translation.
The results is a coupled transcription and translation of the DNA, which is followed by
immediate capture of the protein on the surface of the substrate. After removal of the
expression system through washing only the captured, freshly synthesized proteins, should be
bound to the substrate surface. In NAPPA the DNA encodes the proteins as GST-tag fusions
and is arrayed together with a crosslinker, which immobilises the DNA on the substrate
surface, and an anti-GST capture antibody. This produces an array with colocalised DNA and
capture antibody, which can be stored and expressed on demand by covering the surface with
a cell-free expression system. After removal of the expression system an array of proteins
colocalised with their DNA and capture antibody remains112.
26
Technical Considerations for Protein Microarrays
Substrates
Microarrays results in a very visual representation of the acquired data, through false-colour
images of the slides and its features. These images can be difficult to interpret if the processes
involved in the immobilization of proteins and formation of features is not fully understood
by the analyst. One of the most typical problems occur due to nonuniformity of features in
the array, where the nonuniformity usually takes on the shape of rings. That thin liquid films
takes on ring-like appearances is well known, and the everyday example would be the coffee
ring effect that occur when particle-laden liquid is spilled and allowed to evaporate. The
general background to this phenomena is known to be the transport of material from the
inside of the droplet to the interface between gas and liquid, and the further transport
through capillary flow from the centre of the droplet to the pinned edges of the droplet.
Protein molecules tend to accumulate at the gas/liquid interface due to the fact that the
different hydrophobic and hydrophilic parts will have different preferences for being in either
the gas-phase or the liquid-phase113. This results in accumulation of protein at the interface,
which depletes the concentration within the droplet. The depletion then results in a lowering
of the concentration of proteins available for binding at the substrate surface. Further
diffusion of proteins along the interface to the boundary of the pinned edges of the droplet
depletes the centre of the gas/liquid interface, lowering the concentration at the interface
and resulting in further transport from the centre of the droplet114. In this way the
concentration available for binding to the substrate is continuously lowered, decreasing the
immobilisation efficiency115. This effect is evident regardless of whether the droplet is allowed
to dry or not, but delaying the drying will give the proteins more time to immobilise.
Consequently the use of liquids with low vapour-pressure, such as glycerol, can improve the
uniformity of the features116. Another solution is to outcompete the accumulation of proteins
at the gas/liquid interface by adding a surfactant117. This will displace the proteins from the
interface and increase the concentration at the substrate surface, which improves
immobilisation efficiency.
Microarray technology for proteins were originally adapted from DNA microarrays, but
proteins come with a completely different set of challenges owing to a greater variation in
their chemical properties. Due to the large variation in size, structure, charge etc. proteins will
27
Technical Considerations for Protein Microarrays
vary in performance on different substrates. This affects the usefulness of a substrate for
different assays and care must be taken when choosing the substrate. If the proteins are
immobilised in such a way that they change their properties through conformational changes
or that their active sites becomes inaccessible, then those proteins might not perform well if
the assay depends on their conformation. However, if the assay is independent of the
structure and orientation of the protein the immobilisation may have less importance for the
final result.
Many proteins are folded in their native state so that they display a hydrophilic exterior while
the hydrophobic parts are folded toward the centre of the proteins. If the proteins then are
immobilised on a hydrophobic substrate the result might be denaturation or even refolding of
the proteins. This could even cause the hydrophobic parts of a protein to be displayed on the
outside of the protein, while the hydrophilic parts folds inwards, inactivating the protein118.
One example of an assay that is greatly improved by retaining the conformation and correctly
orient the proteins is the antibody array, as previously mentioned. For antibody arrays the
orientation of the antibody molecules can be arranged by using protein A, protein G or the
recombinant protein A/G. These proteins bind to the Fc-region of antibodies and orient the
Fab-regions away from the substrate surface and owing to the increased accessibility of the
Fab-regions the amount of functional antibodies will be higher119.
Substrates can be divided into many different groups depending on their properties but the
two most easily understood is the division into different geometries and the division into
different coupling chemistries. The immobilisation chemistries can in general be divided in to
physical adsorption, affinity binding and covalent binding. When adsorption is used the
proteins attach to the substrate using nonspecific interactions such as hydrophobic
interactions, Van der Waals forces or electrostatic interactions. This immobilisation results in
random orientation of the immobilised proteins and can result in denaturing or other
structural changes, ultimately leading to loss of protein function. That means that those
immobilisation strategies that relies on adsorption might be less optimal for assays that
require functional proteins, such as the previously mentioned capture assays or functional
assays. Another disadvantage of physical adsorption is that adsorbed proteins may desorb
during assays, leading to loss of detection.
28
Technical Considerations for Protein Microarrays
An example of oriented affinity immobilising has already been given above in conjunction with
the discussion on orientation of antibodies and some other examples are also worth
mentioning. For instance, site specific biotinylation of proteins can be used to orient them on
a streptavidin-coated surface, and histidine–tagged recombinant proteins can be oriented by
arraying them on a substrate coated with metal complexes that bind to the histidine–tag. The
third mayor strategy behind the immobilisation of proteins in microarrays are the covalent
binding to reactive groups, such as aldehyde, epoxy, amino, or NHS ester (Nhydroxysuccinimide). These reactive groups bind through reaction with primary amine, thiol
and hydroxyl groups. The covalent attachment prevents the proteins from being removed
during washing but might cause them to denature, and subsequently care should be taken if
the arrays will be used for functional assays120.
As for geometries, the substrates can either have a two dimensional coating of a monolayer
of reactive groups on a flat support, or a three-dimensional structure such as membranes or
hydrogels. A substrate with a monolayer of reactive groups is comparatively easy to produce
and due to the low intrinsic autofluorescence of the substrate a low background can be
achieved. However, due to the limited geometry available for binding the amount of proteins
that can be immobilised is limited. If the amount of protein available in a droplet exceeds the
binding capacity of the surface the rest of the proteins will only be loosely associated with the
immobilised layer. This can cause increase in local background and/or abnormalities as tailing
of spots when the excess material is being spread out during blocking and washing stages. If a
higher amount of protein is to be arrayed a substrate that can be loaded with more protein is
of better use. This could be for instance a hydrogel-coated substrate, which comprises of
polymers that form a coating a few nm thick when dry but can swell to up to a few hundred
nm when hydrated121, 122. The hydrogel can then provide an environment that keeps the
proteins hydrated and stabilised while the attachment of proteins can be tailored using
reactive groups 123. Another substrate-coating that can accommodate even larger amounts of
proteins is nitrocellulose. Nitrocellulose covered substrates provide a high loading-capacity in
a microporous structure that can be around 10 µm thick. This nitrocellulose layer can
accommodate a large amount of protein and as such is suitable if the amount of material to
be immobilised in each feature is high. However, this increase in loading capacity also comes
29
Technical Considerations for Protein Microarrays
with a higher degree of autofluorescence, especially in the wavelengths commonly used for
fluorescent labels, and may require specialised detection instruments.
Detection Principles
In order to detect target proteins on a microarray an affinity molecule directed against the
target protein is applied to the array, a so called ’primary affinity reagent’. This affinity
molecule will then find and bind to its target. These affinity molecules can be labelled in such
way that they can be detected directly after binding. However, usually a labelled secondary
affinity reagent that is directed against the primary affinity reagent is applied after the binding
of the primary reagent, and this is then followed by detection of the attached label. This way
of detecting the interaction between an affinity reagent and its antigen is called label-based
detection. The reason for the use of two affinity reagents is that it makes it possible to detect
different primary affinity reagents without the need to do multiple labelling experiments, and
the possibility to buy off-the-shelf commercially available secondary detection reagents.
Another advantage is that using a secondary affinity reagent means that the label will not
interfere with the binding site of the primary affinity reagent, if the binding site of the
secondary reagent is known. It is also possible to do label-free detection of the affinity
reagents where some innate quantity of the reagent itself is measured. This makes it possible
to avoid many of the difficulties that are part of labelling molecules and detecting different
kinds of labels. However, label-free assays comes with their own sets of limitations and
problems124.
Label-based Assays
Label-based arrays can be based on radioactive molecules, fluorescent molecules,
nanoparticles, quantum dots or biological barcodes. However, the most commonly used
method is to use a fluorescently labelled secondary affinity reagent to detect the primary
affinity reagent. One example is to use a primary antibody raised in one animal and a labelled
secondary antibody, directed against the Fc-fragment of antibodies from the same animal as
the primary antibody, raised in a another animal. This then ensures that no interference with
the variable Fab-region of the primary antibodies occurs.
30
Technical Considerations for Protein Microarrays
Dual colours enable detection of two targets at the same time. As previously mentioned a tag
attached to a recombinant protein can be detected in one colour-channel, and serve as a
confirmation of the presence of the protein on the array, while the detection of specific
proteins is performed using the other colour-channel. Using more complex detection methods
more colours can be detected on the same array, and subsequently a higher multiplexing can
be achieved. Unfortunately the different fluorescent molecules tend to have emission spectra
that overlap to some extent. What this means that the higher the level of multiplexing the
more difficult it might become to find molecules that have good separation between their
respective emission spectra. Consequently, an increase in background can be expected as
different molecules adds to the measured signal due to spectral overlapping125, as well as the
previously mentioned off-target interaction increase.
The most commonly used fluorescent molecules are Cy5 and Cy3. These cyanine fluorophores
have been known to be sensitive to photobleaching and quenching, leading to lowered
fluorescent intensity and a bias in the representation of the dyes126, 127. An alternative is the
Alexa dyes that show lower dye bias due to their higher resistance to both photobleaching
and quenching, which should result in higher correlation between the two channels. However,
conflicting results when comparing fluorescent molecules have been found127-129, indicating
that the differences between these two popular dyes might be of minor extent. Other
examples of fluorescent detection molecules are rhodamine, texas red, fluorescein, and
quantum dots130, 131.
One limitation with label-based detection is the amount of fluorescent molecules that can be
attached to each detection molecule without adversely affecting its recognition capabilities.
In order to increase this number, and thus achieve a higher signal and detect proteins of lower
abundance, an amplification method such as rolling circle amplification can be applied. Rolling
circle amplification is an enzymatic amplification approach using an oligonucleotide primer, a
circular single-stranded DNA, DNA-polymerase and nucleotides. The primer is attached to an
affinity molecule that bind to the desired antigen. The circular DNA hybridizes to the primer,
and the DNA-polymerase starts replicating the DNA using the free nucleotides. This creates a
long strand of single-stranded DNA complementary to the circular DNA. Primers coupled to a
fluorophore, or other detection molecules, can then bind to the long single-stranded DNA132.
31
Technical Considerations for Protein Microarrays
This not only means that it is possible to attach multiple detection molecules to one affinity
molecule to amplify the signal, but also that a higher degree of multiplexing can be
achieved133.
Label-free Assays
In label-free assays an inherent property of the target itself, such as the mass of the molecule,
is measured. This is a desirable property of an assay as it ensures that the measured signal is
a direct result of the target molecule. In comparison to using labelled detection molecules the
label-free assay would not be subjected to off-target interactions towards other detection
molecules, and not be dependent on laborious labelling of the molecules.
The two most common label-free methods for measuring proteins are surface plasmon
resonance (SPR) and mass spectrometry134, 135. The basic function of a mass spectrometer is
to measure proteins or peptides by ionising and accelerating them to a common velocity and
measuring their deflection in a magnetic field. The resulting deflection depends on their mass
and their charge and the output can be compare to databases with known or theoretical
output. This potentially allow for identification of proteins and their proteoforms but often
requires preparation of samples through enzymatic digestion or protein separation to reach
high sensitivity136. In SPR the change in angle of reflection that occurs as mass changes when
proteins interact with each other on metal surfaces is used to determine when a binding event
happens. By measuring this change over time full kinetic information of the binding event can
be extracted. Although SPR has been used for detection of interaction between proteins for
quite some time there have not been any SPR-platform developed that can be used in a true
high-throughput microarray format. Some microarray applications of SPR have been
produced, but the number of possible simultaneous analytes have usually remained limited
to a few hundred135 although arrays of thousands of features are possible137.
32
Technical Considerations for Protein Microarrays
Blocking and Background
One of the factors that has to be addressed when generating a protocol for any type of protein
microarray is the issue of “background”. Background is caused by detected off-target signals
by the sample, the detection reagents, the substrate, or the contents of the buffer used for
dilution of either sample or targets. In order to address any kind of background the origin of
the background and the causing agent must first be defined, since it can originate from a
number of different sources.
Local background
The local background is commonly considered to be the signal in the reporter channel that
originates from the area immediately surrounding the feature, and there are multiple
contributing factors to this local background. One possible contributor to local background
are satellite spots, which are small spots surrounding the feature that originates from spraying
due to incorrect viscosity when using non-contact printers. Another contributor is smearing
or tails that originates from excess material being smeared from moist spots when liquid is
applied to the array. A third contributor is autofluorescence form the substrate itself and the
fourth is interactions between proteins in the liquid phase and the substrate surface
surrounding the features. These different factors has to be addressed separately; satellite
spots must be addressed by using the correct viscosity in the print buffer, smearing and tailing
must be addressed by ensuring the features are properly dried before any liquid is applied,
and autofluorescence must be addressed by using substrates and detection methods that are
mutually compatible. However, the possibly most difficult local background to address is offtarget interactions between the substrate and the sample or detection reagents. On some
substrates the local background can be addressed through chemical inactivation of the
reactive groups by adding organic chemicals that deactivates the reactive groups.
Alternatively, a well-defined protein, or mix of proteins, can be added and allowed to bind to
the substrate to form a uniform layer94, 138. The goal of inactivating or blocking the substrate
in this way is to reduce off-target interactions towards the substrate to weak interactions and
to achieve a uniform spread. Some care should be applied when deciding on which strategy
that will be used for blocking, especially if the inactivation is performed with a protein. It
should preferably be a substance that can be assumed to have weak interactions with the
33
Technical Considerations for Protein Microarrays
contents of the type of sample that will be applied. As an example bovine serum albumin is
often a suitable blocking protein if the sample that will be applied in the liquid phase contains
antibodies that is to be detected. Since serum albumin is the most prominent protein in serum
of mammals most antibodies could be expected to show low general reactivity towards serum
albumin. Addition of bovine serum albumin results in a monolayer of protein forming on the
substrate surface, and this then quenches unreacted groups and limits the interactions of
other proteins in the liquid phase.139 Adding more than one protein, such as a mixture of
bovine serum albumin and fat-free powdered milk or gelatine can further reduce the
background, but might cause an increase in off-target binding to the substrate due to the
increase in complexity of the blocking agent. This local background can become a problem if
it is high or uneven, but it often does not affect the detection of on-target binding to any
greater extent and adverse effects is limited to making it difficult for image analysis software
to detect the boundaries of features. However, if the local background is high or non-uniform
it can affect the foreground data and thus requiring that it is subtracted from the data. Most
two-dimensional or hydrogel substrates will result in an off-target adsorption that is too low
or uniform to adversely affect the resulting data, as long as the foreground signal is of
acceptable quality. If that is the case then applying downstream background subtraction to
the dataset can add, rather than remove noise, and care should be taken as to how adjustment
for local background is performed. Thicker substrates such as nitrocellulose can often result
in considerably higher local background due to higher autofluorescence from the substrate140.
Applying blocking procedures for thicker substrates can result in lower local background also
for these substrates. However, it is far more efficient to choose detection wavelengths that lie
beyond the autofluorescence of the material, such as near-infrared detection for
nitrocellulose as the detected signal then lies beyond the autofluorescence of the substrate.
Global Background
So far only the background caused by the substrate has been taken into consideration, but a
more difficult problem is often the background from off-target interactions between the
proteins in the liquid phase and the features. This is originates from off-target interactions
between proteins in the sample or the detection reagents and proteins in the features, and is
commonly related to the concentration and complexity of the liquid phase. As an example
Brown et al. found that this background fluorescence is dependent on the total amount and
34
Technical Considerations for Protein Microarrays
the complexity of the protein mixture in the liquid phase. Consequently they found that high
overall concentration tends to reduce the precision in the measurements73. To address this
they suggested fractionation of samples to reduce the liquid phase complexity. However, as
mentioned previously depletion, separation or fractionation of samples are less desirable due
to decreasing reproducibility11 and is not suitable for high-throughput analysis. Although it can
be true that increased complexity increases both local and global background, addition of
proteins to a complex sample can also reduce background signals. This is commonly utilised
by adding the same blocking agent to the sample as were used for blocking the substrate. This
blocks the off-target adsorption of the proteins in the liquid phase before it is added to the
array. Analogous to that, by adding more proteins off-target interactions between proteins in
the liquid phase and features can be blocked. One such example is the global background that
arise from some serum samples when applied to an antigen array where the features carries
a six-histidine group. This group can be found in the Epstein-Barr virus that is often found in
humans, and many individuals can produce antibodies towards this peptide-group. By adding
this six-histidine group to the serum sample most of the antibodies towards that group will
interact with the six-histidine in solution rather than the six-histidine immobilized in the
features. The same principle can be applied to detection reagents; if the detection reagent is
a labelled goat antibody then adding un-labelled goat antibodies to the sample can prevent
weak background interactions between sample and reporter. The global background can also
to some extent be alleviated by diluting the sample more if the foreground is saturated. This
will reduce the detected signals from both the on-target foreground and the off-target global
background, and as the foreground enters the linear range of the detection method the global
background will approach the lower detection limit. However, when the global background
reaches the lower detection limit, any further dilution might only serve to increase noise in
the global background, as the off-target background approaches the noise level that is
introduced through other sources. This makes it difficult to assess if the background is due to
off-target interactions or introduced trough other sources. This is illustrated in figure 5 where the
relation between foreground and global background is moved between off-target noise in high
antibody concentrations and noise from other sources for low antibody concentration.
35
Technical Considerations for Protein Microarrays
Figure 5: 3D plots of two antibody dilutions. The first plot shows the reported fluorescent signal intensity for six dilutions of
two antibodies in separate subarrays. Their on-target interaction are denoted as red peaks, off-target interactions that are
less or equal to two standard deviations of the global background are denoted as green peaks, and off-target interactions
above two standard deviations are denoted as yellow peaks. Each dilution were performed in separate wells and the
separation is represented by the white areas between the subarrays. The plot to the right shows the same data, except as
relative to the on-target peak for each subarray.
Multiplexing
When using microarrays, one of the choices that has to be made is if the degree of multiplexing
should emphasise features or analytes. One single feature physically occupies a small area on
a slide and hypothetically a much larger number of features should be possible to array then
what is practically possible. This is due to technical challenges in arraying that can cause the
distance between the features can become too small, with merging of features as a result. The
distance needed between features depend on the precision of the instrument that is used for
generating the arrays, the buffer composition of the features, hydrophobicity of the substrate,
the lattice arrangement, temperature , humidity etc. Since not all of these variables can be
controlled for, a distance that is large enough to compensate for these variables has to be
chosen. As an example, features with a 100 µm diameter arrayed in a regular lattice might
need a centre to centre distance of approximately 180 µm to prevent the features from
merging. The smaller this distance between features are the greater the risk of features
merging and the greater the distance is the smaller the risk becomes. By increasing the
precision of arraying instruments, and by controlling the ambient conditions and sample
viscosity, variation in placement can be reduced and packing-density of features can be
increased.
It might be assumed that the optimal choice is to maximise the useable area on each substrate
and acquire the maximum amount of data-points per sample. This might be desirable if the
goal is to perform and undirected approach for novel discoveries of protein interactions or
36
Technical Considerations for Protein Microarrays
biological markers for diseases, such as for autoimmunity or protein profiling. However, it
comes at a great cost as it requires access to expendable reagents whose relevance might not
be of great importance in the proposed experiment, as well as requiring the use of more
substrate. The addition of more unique features do not necessarily result in a one to one
increase in meaningful foreground signals, which results in diminishing returns on the cost
incurred from acquiring or producing the arrays. The alternative to this undirected approach,
is to use a planned directed approach. Employing a directed approach entails applying a
hypothesis as to which targets that might be reactive in the specific experimental context and
acquiring those targets for arraying. By this the spatial size of the arrays can be kept smaller,
and higher throughput per substrate can be reach through the use of subarrays and
compartmentalisation. The sparsity of data that can occur for very large arrays is shown in figure 6
and illustrates the diminishing returns that can be a result of an undirected approach. If the
interactions shown in red in figure 6 would have been thought to have relevance in the disease setting,
and a directed approach comprising of these proteins had been performed, the cost incurred for the
analysis could have been minimised. This is unfortunately not always not possible and as such these
large-scale arrays represent an important tool for discovering unexpected autoimmune interactions.
Figure 6: 3D representation of the all 21,120 values in one array. The data originates from autoimmunity profiling of a human
serum sample. Any data below or equal two standard deviations is coloured in green while data points above two but below
or equal to ten standard deviations is coloured in yellow. Data above ten standard deviations is coloured in red.
37
Technical Considerations for Protein Microarrays
The added benefit of analysing more analytes per substrate is that variation and cost per
experiment can be reduced simultaneously. This change in throughput can be illustrated by
considering that a session with an arraying instrument might produce 100 microscope slides
with one array comprising 57.000 features, usable for in total 100 analytes. Alternatively, it
can produce 100 slides with 24 arrays with 1.500 features for a total of 2.400 analytes. The
throughput can thus be varied between features or analytes through compartmentalisation.
Compartmentalisation can be achieved through the use of multiwell hybridisation cassettes
or microscope slide holders and commercially available holders can be found with up to 192
compartments per slide. While such a large number of compartments allow for a large number
of samples to be simultaneously analysed, the arrays will by necessity be very small. If larger
arrays is needed holders suitable for automation could be a more effective choice. A holder
for four microarray slides with 24 compartments per slide results in a 96-well standard format
for implementation of high-throughput applications using standard pipetting robots. In the
end the choice often stands between reaching high-throughput in the dimension of either
analytes or features, and compensating for the lack of throughput in the other dimension
through increased cost and time.
Image Analysis
Detection of fluorescent labels is usually conducted in a microarray scanner that use lasers to
excite the fluorescent labels and a detector to detect the emitted light. The output of the
instrument are false-colour images from each channel that can be merged into one final
image. A scan of a full microscope slide can comprise millions of data points, most of which
are of little interest as they are derived from the empty slide around the features. The number
of data points depend on the resolution of the scanner, which is often adjustable between 5,
10, 20, and 40 µm. At a resolution of 10 µm each data point will be represented by a 10x10
µm square, or “pixel”. As a result, at a resolution of 10 µm a circular feature with a diameter
of 100 µm will be comprised of approximately 80 pixels, i.e. 80 measurements. If an increase
in measurements is desired and the resolution is increased to 5 µm, then the measurements
made in each feature will quadruplicate to over 300.
38
Technical Considerations for Protein Microarrays
Superficially it may seem like the best choice of resolution would be the highest resolution
possible but this is not always the case. As the resolution doubles the number of pixels
quadruplicates regardless of whether the pixels are of interest. This will then result in longer
scan times and a quadruplicating of the size of the image files. For light use this might not
matter much as the increase in time and file size for a few scans might not become much of a
problem. However, in large-scale projects that produce large arrays and numerous images the
time spent on scanning and transferring files, as well as the size of storage space for large
image files, makes it impractical to use higher resolutions. From this follows that the
resolution that should be chosen is dependent on the resolution that is needed for a particular
application. For most microarray application the 10 µm resolution will be suitable since the
feature diameter usually will be between 100 – 150 µm, meaning that there will be
approximately 80 to 180 pixels per feature. However, some applications, such as arrays
produced through lithography, can produce features at the nanometre scale and
consequently benefit from, or demand, high-resolution scans.
Due to the large overhead of pixels originating from outside the array or in between features,
extraction of data is preceded by defining the areas occupied by each individual feature.
Thankfully this is to large extent achieved through software that uses algorithms to identify
the location of the features. The automatic identification is followed by manual input for those
features that are misalignment, ill-defined, missing or otherwise need attention. Once the
features have been localised the software will separate the values from the features from the
values immediately surrounding the feature. The values can then be exported to a more easily
handled text-file format. Simple mathematical calculations such as standard deviation for each
feature can often be performed and manually or automatically added notes on feature quality
can be added.
Once the data have been extracted and saved as more easily addressable files the question of
how to evaluate the data arises. Depending on the experiment, extensive statistics may have
to be performed in order to correct for experimental variation and abnormalities. However,
no “catch-all” method for analysis of protein microarrays data exists due to the diverse assay
methodologies that are possible for protein microarrays. Reference designs similar to what
has been used for DNA arrays have been successfully implemented for protein arrays141, 142.
39
Concluding Remarks
Although, one drawback with applying statistical methods developed for DNA arrays is that
they often assume a normal distribution that is equally distributed in both up and down
regulation of expression, and this is rarely the case for protein data. However, the basic steps
of filtering bad features, performing background correction, and normalising within and
between arrays are similar between DNA and protein arrays. An alternative to the reference
design is to normalise for the amount of targets in each feature143, which is an attractive
alternative if the proteins have been expressed as fusions to a common tag. A special case are
the reverse phase arrays that often utilises dilution curves, and sometimes suffer from large
spatial variations due to their thicker substrates. As a result some analysis methods tailored
specifically for reverse phase arrays have been developed
144-146.
All in all, many protein
microarray assays require their own tailored strategy for data analysis depending on the type
of assay.
Concluding Remarks
So how does all this tie together? To an outside observer protein microarrays may seem like
a simple solution to the problem of conducting fast and efficient analysis without using large
volumes of precious samples or unsurmountable amounts of time. In reality the protein
microarrays are a complex collection of different experimental methods whose common
ground is the use of proteins in a “capture molecule – analyte molecule” setup, and the
attachment of either of these molecules to a solid substrate in predetermined positions. As
might have become apparent in the preceding text, this basic setup quickly branches out into
multiple platforms that differ from each other in regards to arraying instruments, substrates,
buffers, and detection methods. The final type of assay depends then on what kind of capture
and analyte molecules that are being used and the capability of the laboratory performing the
analysis. The final assays can roughly be divided into three groups. The group of assays where
the orientation and functionality of the immobilised reagent is of importance for the success
of the experiment. The group of assays where the orientation and functionality of the
immobilised reagent is not of vital importance for the success of the experiment. The third
group of assays is then be where the immobilised reagent is a complex mixture of high
concentration. The most obvious examples of these would then be the antibody array, the
protein fragment or peptide array, and the reverse phase array. To be able to perform any of
these three general types of experiments all special equipment that is needed is an instrument
40
Concluding Remarks
for producing the arrays, a scanner, and a healthy dose of curiosity and love for a good
challenge. In the four papers described under “Present Investigations” I have detailed some
of the work that has gone into understanding some of these technologies and the
development of our own platforms for arraying and analysing reagents.
In paper I and III are examples of developing and discovering the possibilities and limitations
of the reverse phase format while using the instrumentation at hand, rather than the
instrumentation that would have been optimal for the task. In paper I we were limited by the
equipment we had at hand and, as all good engineers and scientist do, did the best we could
with the equipment at our disposal and the final result met our goal. However, after paper I
there were still unresolved questions in regard to accuracy and robustness of our reverse
phase platform. Those questions were then answered to some extent during the work that
laid the foundation for paper II. What these two paper show was that it was possible to do
high-throughput screening of blood-derived samples with high reproducibility, for proteins
that could be considered high and medium abundant. They also showed the impact and
importance of being able to perform all these tests simultaneously, as the reproducibility
within and between experiments were more robust for the array-assays compared to the
“gold standard” ELISA. Today a new reverse phase platform is being developed using more
suitable instrumentation, and we are taking another step in the evolution of our microarray
platforms analytical space.
In paper II and III some of the work that we have performed on the development and scaleup of our protein microarray platform for high-throughput validation and profiling of affinity
reagents and biofluids is detailed. There is a remarkable difference between repeatedly
performing a routine analysis and performing analysis using unknown reagents, especially
when the work has to be performed in a high-throughput manner. This was the main issue in
paper II as we performed validation experiments of hundreds of reagents with unknown
properties in short time. By applying the knowledge we had previously gained by the routine
analysis of HPA antibodies, as to what a successful analysis could be expected to look like, we
were able to develop and implement several new protocols using different buffers and
detection reagents. This also meant that we got confirmation that our routine analysis
41
Concluding Remarks
performed within HPA was a valid analysis method also for other types of affinity reagents,
and that the setup was functional in a broader affinity reagent context.
The arrays that are routinely produced for the validation of HPA antibodies have also found
their use in other applications, such as screening for targets for autoantibodies in autoimmune
disease contexts. One of the difficulties encountered when trying to analyse a sample with
entirely unknown interaction partners on microarrays containing few possible targets, is the
problem with identifying a functional experiment if there are no targets that results in clearly
defined signals in the array. In such a scenario one might be inclined to believe that the
experimental protocol is at fault, and either abandon the project or start unnecessary
optimisation attempts. The need for larger arrays for screening is what drove the development
of the large-scale and high-density protein fragment array detailed in paper IV. The
combination of an arraying buffer with low vapour pressure and a high-speed arraying
instrument lead to the possibility to array 384-microtiterplates with protein fragments in a
fast and efficient manner, while minimising evaporation and sample consumption. We could
thus store plates full of protein fragments that had been produced over a long period of time
and accrue a large library of protein fragments for construction of large-scale protein
microarrays. These microarrays that comprises tens of thousands of features have since
proven to be useful for both validation and profiling of HPA antibodies, as well as
autoimmunity profiling of human serum. The background behind these four papers are more
thoroughly described in the chapter Present Investigations and the articles themselves are
included in the appendix.
42
Present Investigations
Present Investigations
Paper I - Screening for C3 Deficiency in Newborns Using
Microarrays.
Aims of the investigation
Complement factor 3 (C3) deficiency is an inherited defect of the immune system and might
result in organ damage or death due to severe infections. It is caused by mutations in the C3
gene resulting in absence or diminished levels of C3, or a dysfunctional variant of C3. The
disorder is rare, and is commonly sorted under the larger group of inherited defects of the
immune system that is called “primary immunodeficiency disorders”. The deficiency is mostly
noted in infancy due to life-threatening infections, and neonatal screening for detection of C3
deficiency before the onset of severe infection would be desirable for administration of
prophylactic treatment to reduce of adverse effects.
We had a previously established platform for screening of C3 deficiency in serum in a reverse
phase protein array (RPPA) format, and we now aimed to extend that platform to eluates from
dried blood spot samples (DBSS).The reason for using DBSS were that it is are routinely
collected from newborns at birth and are already widely used for neonatal screening. The
results that we acquired from the RPPA format were compared to values acquired through
ELISA, for confirmation of relevance of the findings for both serum and DBSS eluates.
Summary of findings
Two C3 deficient serum samples and twelve control serum samples were analysed for their C3
levels by both RPPA and ELISA. No C3 were detected for the C3 deficient samples in either of
the platforms, while both platforms detected C3 in the control samples. Two C3 deficient DBSS
eluates and 278 control DBSS eluates were arrayed together with the serum samples. No C3
were detected in either of the deficient DBSS samples while C3 were readily detected in most
of the control DBSS samples. Some of the control DBSS samples showed low levels of C3 on
the RPPA as compared to ELISA and were reprinted six times for a total of up to 47 replicates.
The protein content of these six discrepant samples were then determined by SDS-PAGE and
43
Present Investigations
compared to normal controls. Some low molecular weight bands present in the discrepant
samples were noted, suggesting degradation of C3.
Results and conclusions
The correlation between RPPA and ELISA was generally low. However, the RPPA showed
considerably lower coefficient of variation (15±6%) compared to the ELISA (29±13%). This
might be explained by the lack of batch effects in the RPPA format, but still did not explain the
consistently low RPPA values for six of the controls. Both the RPPA and the ELISA used
polyclonal anti-C3 antibodies for detection of C3 so it remains to be determined why the ELISA
would more readily detect degraded C3, in combination with a generally higher variation.
Unfortunately this question could not be fully answered and remained unanswered. Overall
the RPPA method developed for DBSS proved to be able to detect deficiencies, although it
seemed sensitive to false positives, making it useful as a screening method for C3 deficiency.
However, the RPPA would then need to be followed up by validation of identified deficiencies
through an alternative platform.
Figure 7: Data correlation between RPPA and ELISA assays for C3 levels in DBSS. The correlation between serum microarray
intensities and ELISA values for the Swedish C3 deficient patient (green circle) and 269 controls. Six samples (marked in red)
show low intensities on the microarrays when compared to C3 (g/l). The results for the low intensity samples are based on 6
printings and a total of 20-47 replicates.
44
Present Investigations
Paper II - Validation of Affinity Reagents Using Antigen Microarrays.
Aims of the investigation
This investigation was initiated in the context of the SH2-consortium14, which were an
international program to generate a complete set of affinity reagents to SH2-containing
human proteins. The consortium comprised several research groups developing different
types of affinity reagents towards the 110 known human SH2-containing proteins encoded in
the human genome. There were two sources for production of antigens representing the SH2containing proteins, the Structural Genomics Consortium147 (SGC) that produced the SH2domains and the Human Protein Atlas52 (HPA) that produced protein fragments. The affinity
reagents towards these antigens were generated by five research groups, as well as multiple
companies, and comprised antigen purified polyclonal rabbit antibodies, mouse monoclonal
antibodies and recombinant single-chain variable fragments (ScFv). In total 64 SH2-domain
containing proteins were represented by 105 antigens, and 398 affinity reagents towards
these antigens were generated.
Due to the varying origin and production methods of the antigens and affinity reagents, a
versatile high-throughput protein microarray setup that could be used for validation of
reagents specificity had to be established. We had previously developed the protein
microarray for validation of antibodies produced within HPA, and now broadened the
applicability of this platform to comprise more antigens and other methods of detection. To
achieve this, the 105 SH2-antigens were supplemented with 301 non-SH2 protein fragments
and arrayed these antigens in 14 identical subarrays per slide, and performed optimisation for
each set of binders to establish working protocols. This encompassed establishing suitable
dilutions of reagents and compatible buffers.
45
Present Investigations
Figure 8: Design of the SH2-based antigen microarrays. (a) The layout and distribution of the 432 spotted antigens. The red
spots corresponds to SH2 related HPA antigens and the blue spots corresponds to non-SH2 HPA antigens. The green spots
indicate where the SGC SH2-domains are spotted. (b) Quality assurance of the produced antigen microarrays is performed
with a chicken anti-His6ABP antibody, visualising the presence of all HPA produced antigens.
Summary of findings
Despite the complexity of performing high-throughput validation of a diverse set of affinity
reagents, we could identify reagents from all providers towards all SH2-domain antigens.
Using our microarrays we could determine that approximately 50% of the affinity reagents
had binding profiles where the most prominent signal was from their intended target, and
only approximately 10% had low or no interaction with their intended target while displaying
reactivity with other antigens.
46
Present Investigations
Figure 9: Schematic view of the results from validation of a monoclonal antibody towards protein GRB2. The y-axis in the
histogram shows the relative signal intensity from each arrayed antigen. The x-axis contains all antigens in the order that they
were arrayed with the recombinant protein fragments produced in the HPA to the left as single replicates and the domains
produced by SGC to the right in duplicates. The signal from the HPA GRB2 antigens is shown as green bars, and the signals
from the SGC domain is shown as red bars. Off-targets interactions are shown as black bars. In the sequence plot displayed
below the histogram the full length protein is drawn as a horizontal bar coloured in grey with the endpoints of the SH2-domain
marked with blue vertical bars. The HPA protein fragments are drawn as green horizontal bars and the SGC SH2-domain as a
red horizontal bar. All protein fragments have their first and last amino acid written next to them, as corresponding to the
full-length protein. This example shows that the monoclonal antibody that were produced against the SGC produced SH2domain of GRB2 recognises the proper target, and that the epitope is likely to reside in the 50 amino acid stretch between
amino acid 109 and 159.
Results and conclusions
Our investigation were triggered by the need of a high-throughput platform for the validation
of on-target interactions of a large number of diverse affinity reagents from multiple
providers. At first glance this task seemed daunting as we recognised the need for combining
a high-throughput workflow with multiple optimisations for reagents that we had little to no
previous experience with. We approached this task by using our established platform as a
springboard to launch optimisation assays that were needed to accommodate the different
reagents. We also recognised the need for previous experience of microarray method
development and image analysis to interpret results and adjust protocols accordingly.
Approximately 600 tests including optimisations, were done for the 398 affinity reagents. This
means that less than two test per reagent were needed. This clearly showed the strength of
protein microarrays as validation tools for affinity reagent production.
47
Present Investigations
Paper III - Biosensor Based Protein Profiling on Reverse Phase
Serum Microarray.
Aims of the investigation
This investigation was an extension of our previous work using RPPAs. As mentioned in
conjunction with the paper I, we had noticed a discrepancy between the results obtained using
the RPPA and the ELISA. We had now produced a new RPPA with almost 2.500 serum samples
for investigating the feasibility of screening for IgA-deficiency using this platform. We once
again noticed a noticeable discrepancy between the data obtained through the RPPA and the
ELISA. We had been offered the opportunity to evaluate the feasibility of using a novel
instrument for our work and to develop new assays for the platform. We therefore decided
to bring this investigation into this biosensor based microarray platform, to confirm the
previous results. The biosensor based platform utilised surface plasmon resonance (SPR) to
measure the difference in refraction that occurs as the mass changes when a ligand binds to
a protein on a metal surface. The output after data acquisition are full binding curves that
allow for extraction of interaction kinetics. The instrument that we been given access to had
the capability to measure 400 binding curves simultaneous for six consecutive ligand injections
for up to a total of 2.400 binding curves per chip.
Summary of findings
A total of 2.423 serum samples were arrayed on glass slides and analysed for their IgA content.
182 of those samples were then re-arrayed on glass slides and arrayed on sensor chips for the
SPR measurement. This set of 182 samples comprised 28 samples that had been identified as
IgA deficient using both the RPPA and the ELISA, 100 samples that showed large discrepancy
between the RPPA and the ELISA, and 54 samples that were randomly chosen from the sample
set. The within-slide and between-slide reproducibility of the glass slides were investigated as
well as the reproducibility between printings. For the SPR chips the between-chip
reproducibility was also investigated. The correlation between the three platforms, RPPA,
ELISA, SPR, were also assessed.
48
Present Investigations
Figure 10: Reproducibility of the label-based RPPA platform. A) Plotting the spot replicates within each slide in a correlation
plot shows a correlation of R = 0,98 between spot replicates. B) Plotting two slide replicates shows a correlation coefficient of
R = 0,98 also between slides, albeit with a lower correlation for samples with higher target concentration. C) Plotting the 182
samples that were reprinted and reanalysed on the fluorescence based arrays shows a correlation coefficient of R = 0,9. This
indicates that the spotting procedure was robust and with acceptable reproducibility both within each print run as well as
between print runs.
Results and conclusions
No replicates were performed on the ELISA platform unless a sample showed a large
discrepancy between the RPPA and the ELISA, due to the time and sample consuming nature
of the ELISA. Consequently, no proper investigation into the reproducibility of the ELISA could
be included in the final paper. The reproducibility within the RPPA platform and the SPR
platform, and between the two array platforms, proved to be excellent. A Spearman
correlation over 0,9 was obtained for the within platform comparison, and just below 0,9 for
the between platform comparison. The ELISA showed less agreement with a correlation of
only 0,66 (ELISA vs. RPPA) and 0,76 (ELISA vs. SPR), and this was shown by the correlation plots
as seen in figure 11. We had previously suspected that the discrepancy between the RPPA and
the ELISA was driven by the ELISA data, and the new platform seemed to confirm this. The
coefficient of variation for any of the platforms were not included in the final article due to
the lack of appropriate replicates for the ELISA platform. However, for those samples that
there were replicates performed on the ELISA the CV was 50%, while for the RPPA and the SPR
they were 17% and 5% respectively. Many of the samples that were replicated on the ELISA
were samples that were chosen to be reanalysed due to their discrepancy with their RPPA
result. Therefore the observed CV were not be deemed fully representative of the platform.
49
Present Investigations
Figure 11: Correlation between the RPPA-platforms and the ELISA. A) The two microarray-based platforms show acceptable
agreement with a correlation coefficient of R = 0,89. B) A correlation plot between fluorescence-based array and ELISA show
less agreement with a correlation coefficient of R = 0,66. C) The correlation plot between SPR-based array and ELISA also show
less agreement between platforms then the two array-based platforms with a correlation coefficient of R = 0,76.
The SPR platform proved to generate data with CV < 5% and with a possibility to extract full
binding kinetics if necessary. It was consequently deemed to be useful for measuring IgA in
serum in a RPPA-like format. As a side note to that is that we previously investigated the
instruments applicability for detection of proteins in serum in a capture array format.
However, this resulted in little success due to unsurmountable difficulties with clogging of the
fluidics system in the instrument. After several failed attempts it was revealed that it was a
known problem for the instrument and was due to the high protein concentration in serum.
Another drawback with the instrument was the extraordinarily large sample volume of 1,6 ml
that were needed for injection, which precluded the use of the instrument for many
applications. The instrument was later withdrawn from the market as other instruments with
similar function that were less sample demanding became more widespread.
50
Present Investigations
Paper IV - Exploration of High-density Protein Microarrays for
Antibody Validation and Autoimmunity Profiling.
Aims of the investigation
All antibodies generated within the Human Protein Atlas (HPA) are validated for their ontarget binding with protein microarrays that comprise of their cognate recombinant protein
fragment and 383 other fragments, which are currently in the HPA antibody production
pipeline. There are new batches of 384 fragments arrayed regularly for continuous testing of
cognate antibodies. The 384-well microplates that contains the protein fragments can be
stored and reused for production of new microarrays, and multiple batches can thus be
combined to produce larger arrays. We took advantage of this possibility and produced a set
of arrays containing 11.520 protein fragments with two subarrays per slide, and a set of arrays
containing 21.120 protein fragments in one subarray per slide. This undertaking was initiated
for the advancement of our protein microarray platform and for investigating the relevance
of the classifications that are assigned to each antibody during validation in the HPA.
Summary of findings
We profiled 48 polyclonal antibodies on arrays with 384 and 11.520 protein fragments, and
10 of those antibodies were further analysed on arrays with 21.120 protein fragments. Nine
serum samples from patients with secondary progressive multiple sclerosis were also analysed
for their autoimmunity reactivity profile on the arrays with 11.520 protein fragments, and two
of those were further profiled on the arrays with 21.120 protein fragments. Two of the
antibodies and two of the serum samples were finally profiled on a commercially available
protein microarray platform comprising over 17.000 full-length proteins. The 48 polyclonal
antibodies that were profiled on arrays with 384 and 11.520 protein fragments showed
increasing spread of off-target interactions on 11.520 protein fragments for antibodies with an
increasing number of off-target interactions for 384 protein fragments, as seen in figure 12A. This
indicates that the routine validation performed on protein fragment arrays in the HPA is able to detect
antibodies with a high number of off-target interactions towards complex samples. A subset of those
antibodies that showed one single off-target interaction >40% of the on-target interaction, and a
subset of the group antibodies that showed no off-target interactions, were further analysed on the
51
Present Investigations
arrays with 21.120 protein fragments, as seen in figure 12B . This revealed that the antibodies with a
single strong off-target interaction does not necessarily result in a large increase of off-target
interaction as compared to those antibodies that showed no off-target interaction on 384 protein
fragments. Consequently, those antibodies with only one strong interaction on 384 protein fragments
may not be necessary to fail in the validation step. One example that illustrates the increasing amount
of off-target interactions that could be expected from an antibody as it is exposed to more possible
epitopes is shown in figure 13, with barplots from 384, 11.520 and 21.120 protein fragments and the
full-length protein microarray. The lack of off-target interactions towards full length proteins implies
that epitopes detected in the protein fragments might be hidden in the full-length folded proteins,
thus making the antibody more specific then what is implied on the protein fragment arrays.
Figure 12: The number of off-target interactions with a relative signal >15% of the on-target interaction. A) On the x-axis is
the number of off-target interactions (> 15% of on-target) on arrays with 384 protein fragments plotted in groups of zero to
more than three interactions and with an additional group for those that showed on strong (>40% of the on-target
interaction). On the y-axis is the number of off-target interactions on arrays with 11.520 protein fragments. B) The green
connected dots are five antibodies that had no off-target interactions on 384 protein fragments and the red connected dots
are five antibodies with one off-target interaction >40% of the on-target interaction on 384 protein fragments.
Results and conclusions
The assembly of 384-well microplates for production of arrays for validation of antibodies
enabled the generation of large-scale arrays with 21.120 protein fragments. These large-scale
arrays can be further expanded and hold the potential to become an important addition to
our toolbox for profiling both affinity reagents biofluids. Using these arrays we could show
that the validation performed in the HPA carries relevance when considering large collections
of antigens. The production of these arrays also hints at the feasibility of constructing large
antibody arrays using the HPA antibodies as these antibodies are also stored in microplates,
and in a buffer suitable for direct arraying. We also had the opportunity to compare our results
from both antibodies and autoimmunity samples to a full-length protein microarray. That
showed that the full-length protein microarrays useful for validation of on-target binding and
52
Present Investigations
screening of interaction partners of antibodies, and that they add value to our autoimmunity
screening.
Figure 13: The first three barplots shows the binding profile for one antibody on arrays with 384, 11.520, and 21.120 protein
fragments. The initial binding profile on 384 protein fragments revealed a signal off-target interaction. As the number of
protein fragments in the array increased the number of identified off-target interactions increases as well with the correct
antigen marked by red bars and the initial off-target interaction marked in black. At the bottom is the same antibody tested
on full length protein microarrays containing approximately 17.000 full length proteins where only one strong off-target
interaction was detected.
53
Future Perspectives
Future Perspectives
The work in this thesis provides details to the work that constitute the foundation of the
protein microarray platform. The platform have undergone a steady evolution from what we
today consider small-scale arrays towards more, bigger, faster, and better. And the work will
of course not end now.
An antigen array comprising of almost all antigens produced within the HPA will soon be in
the making. This will give us unprecedented possibilities for profiling of reagents and samples,
with over 40.000 unique recombinant protein fragments as targets. To achieve this, new
arraying procedures and new protocols for analysis will have to be developed.
On the reverse phase array platform we aim to array a complete biobank of approximately
12.000 serum samples for investigation of higher abundant protein levels looking across
various diseases. This would then come to represent maybe the largest RPPA ever produced
so far, and will enable new and exciting research venues for the protein microarray platform.
Last, but definitely not least, looms the possibility of arraying one of the world’s largest
collections of antibodies. As the Human Protein Atlas on November 6th 2014 summarised its
knowledge of the human proteome it comprised proteome analysis of almost 17.000 unique
proteins that have been analysed using approximately 24.000 antibodies. Thousands of these
antibodies, together with many more that have not yet been published, are currently
accessible from plates that has previously been available for the validation. One can envision
a transfer of aliquots of these antibodies into plates suitable for arraying that will enable the
generation of capture arrays with tens of thousands of antibodies. This will then allow for
profiling of protein levels for thousands of proteins simultaneously on a single slide, which will
open up entirely new possibilities for proteomic research.
The field of affinity proteomics is set to continue growing and evolve and I hope to continue
as a part of these efforts that advances it into the future by discovering new uses and by
developing new applications for protein microarrays.
54
Acknowledgements
Acknowledgements
Peter, min huvudhandledare som har gett mig frihet att vara kreativ och testa det jag vill
testa utan att ifrågasätta vad alla plottar på buffertar, morfologier, och bakgrundssignaler
egentligen gör för nytta.
Jochen, min bi-handledare, för fruktbara diskussioner om ditten och datten och för
rödpennan för korrigering av mitt slarviga skrivande.
Mathias, för att du har samlat alla dessa människor under proteinatlasets paraply
Alla i HPA och AlbaNova som jag har jobbat med under dessa år.
PAPP, från att bara vara PA så har vi blivit PAPP, och ju fler desto roligare.
Burcu, som har stått ut med mig som kontorsgranne och med mitt konstanta klagande över
dåliga studenter, dåliga labb-rapporter, dåliga artiklar, dåliga experiment och vad helst annat
jag har hittat att gnälla över.
Mårten, som utgjorde PA-trion med Peter och mig i början av min tid här, och som jag har
spätt många antikroppar med och förundrat mig över så många konstiga printningar
tillsammans med.
Mamma Siv, en lång och krokig väg har det varit och utan ditt stöd så hade jag inte tagit mig
hit.
55
References
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
56
Watson, J.D. & Crick, F.H. Molecular structure of nucleic acids; a structure for deoxyribose
nucleic acid. Nature 171, 737-738 (1953).
McCarty, M. & Avery, O.T. Studies on the Chemical Nature of the Substance Inducing
Transformation of Pneumococcal Types : Iii. An Improved Method for the Isolation of the
Transforming Substance and Its Application to Pneumococcus Types Ii, Iii, and Vi. J Exp Med
83, 97-104 (1946).
Gerstein, M.B., Bruce, C., Rozowsky, J.S., Zheng, D., Du, J., Korbel, J.O., Emanuelsson, O.,
Zhang, Z.D., Weissman, S. & Snyder, M. What is a gene, post-ENCODE? History and updated
definition. Genome Res 17, 669-681 (2007).
Morris, K.V. & Mattick, J.S. The rise of regulatory RNA. Nat Rev Genet 15, 423-437 (2014).
Crick, F. Central dogma of molecular biology. Nature 227, 561-563 (1970).
Wilkins, M.R., Pasquali, C., Appel, R.D., Ou, K., Golaz, O., Sanchez, J.C., Yan, J.X., Gooley, A.A.,
Hughes, G., Humphery-Smith, I., Williams, K.L. & Hochstrasser, D.F. From proteins to
proteomes: large scale protein identification by two-dimensional electrophoresis and amino
acid analysis. Bio/technology 14, 61-65 (1996).
Consortium, E.P. An integrated encyclopedia of DNA elements in the human genome. Nature
489, 57-74 (2012).
Ponten, F., Gry, M., Fagerberg, L., Lundberg, E., Asplund, A., Berglund, L., Oksvold, P.,
Bjorling, E., Hober, S., Kampf, C., Navani, S., Nilsson, P., Ottosson, J., Persson, A., Wernerus,
H., Wester, K. & Uhlen, M. A global view of protein expression in human cells, tissues, and
organs. Mol Syst Biol 5, 337 (2009).
Gry, M., Rimini, R., Stromberg, S., Asplund, A., Ponten, F., Uhlen, M. & Nilsson, P. Correlations
between RNA and protein expression profiles in 23 human cell lines. BMC Genomics 10, 365
(2009).
Anderson, N.L. & Anderson, N.G. The human plasma proteome: history, character, and
diagnostic prospects. Mol Cell Proteomics 1, 845-867 (2002).
Schroder, C., Jacob, A., Tonack, S., Radon, T.P., Sill, M., Zucknick, M., Ruffer, S., Costello, E.,
Neoptolemos, J.P., Crnogorac-Jurcevic, T., Bauer, A., Fellenberg, K. & Hoheisel, J.D. Dual-color
proteomic profiling of complex samples with a microarray of 810 cancer-related antibodies.
Mol Cell Proteomics 9, 1271-1280 (2010).
Stoevesandt, O. & Taussig, M.J. Affinity proteomics: the role of specific binding reagents in
human proteome analysis. Expert Rev Proteomics 9, 401-414 (2012).
Larsson, M., Brundell, E., Nordfors, L., Hoog, C., Uhlen, M. & Stahl, S. A general bacterial
expression system for functional analysis of cDNA-encoded proteins. Protein Expr Purif 7,
447-457 (1996).
Colwill, K. & Graslund, S. A roadmap to generate renewable protein binders to the human
proteome. Nat Methods 8, 551-558 (2011).
Litman, G.W., Rast, J.P., Shamblott, M.J., Haire, R.N., Hulst, M., Roess, W., Litman, R.T., HindsFrey, K.R., Zilch, A. & Amemiya, C.T. Phylogenetic diversification of immunoglobulin genes
and the antibody repertoire. Molecular Biology and Evolution 10, 60-72 (1993).
Market, E. & Papavasiliou, F.N. V(D)J recombination and the evolution of the adaptive
immune system. PLoS Biol 1, E16 (2003).
Woof, J.M. & Burton, D.R. Human antibody-Fc receptor interactions illuminated by crystal
structures. Nat Rev Immunol 4, 89-99 (2004).
Ehrenstein, M.R. & Notley, C.A. The importance of natural IgM: scavenger, protector and
regulator. Nat Rev Immunol 10, 778-786 (2010).
References
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
Vollmers, H.P. & Brandlein, S. Natural IgM antibodies: from parias to parvenus. Histol
Histopathol 21, 1355-1366 (2006).
Vidarsson, G., Dekkers, G. & Rispens, T. IgG subclasses and allotypes: from structure to
effector functions. Front Immunol 5, 520 (2014).
Novak, N., Kraft, S. & Bieber, T. IgE receptors. Current Opinion in Immunology 13, 721-726
(2001).
Macpherson, A.J., McCoy, K.D., Johansen, F.E. & Brandtzaeg, P. The immune geography of
IgA induction and function. Mucosal Immunol 1, 11-22 (2008).
Delacroix, D.L., Dive, C., Rambaud, J.C. & Vaerman, J.P. IgA subclasses in various secretions
and in serum. Immunology 47, 383-385 (1982).
Edholm, E.-S., Bengten, E. & Wilson, M. Insights into the function of IgD. Developmental &
Comparative Immunology 35, 1309-1316 (2011).
Lipman, N.S., Jackson, L.R., Trudel, L.J. & Weis-Garcia, F. Monoclonal versus polyclonal
antibodies: distinguishing characteristics, applications, and information resources. ILAR
journal / National Research Council, Institute of Laboratory Animal Resources 46, 258-268
(2005).
Rockberg, J., Schwenk, J.M. & Uhlen, M. Discovery of epitopes for targeting the human
epidermal growth factor receptor 2 (HER2) with antibodies. Mol Oncol 3, 238-247 (2009).
Ehrlich, P.H., Moyle, W.R., Moustafa, Z.A. & Canfield, R.E. Mixing two monoclonal antibodies
yields enhanced affinity for antigen. J Immunol 128, 2709-2713 (1982).
Milstein, C. The hybridoma revolution: an offshoot of basic research. BioEssays 21, 966-973
(1999).
Kohler, G. & Milstein, C. Continuous cultures of fused cells secreting antibody of predefined
specificity. Nature 256, 495-497 (1975).
Hudson, P.J. Recombinant antibody fragments. Current Opinion in Biotechnology 9, 395-402
(1998).
Nelson, A.L. Antibody fragments: hope and hype. mAbs 2, 77-83 (2010).
Nygren, P.A. & Skerra, A. Binding proteins from alternative scaffolds. J Immunol Methods
290, 3-28 (2004).
Nord, K., Gunneriusson, E., Ringdahl, J., Stahl, S., Uhlen, M. & Nygren, P.A. Binding proteins
selected from combinatorial libraries of an alpha-helical bacterial receptor domain. Nat
Biotechnol 15, 772-777 (1997).
Schlehuber, S. & Skerra, A. Anticalins as an alternative to antibody technology. Expert opinion
on biological therapy 5, 1453-1462 (2005).
Koide, A., Bailey, C.W., Huang, X. & Koide, S. The fibronectin type III domain as a scaffold for
novel binding proteins. J Mol Biol 284, 1141-1151 (1998).
Stumpp, M.T., Binz, H.K. & Amstutz, P. DARPins: a new generation of protein therapeutics.
Drug discovery today 13, 695-701 (2008).
Proske, D., Blank, M., Buhmann, R. & Resch, A. Aptamers--basic research, drug development,
and clinical applications. Appl Microbiol Biotechnol 69, 367-374 (2005).
Kenyon, G.L., DeMarini, D.M., Fuchs, E., Galas, D.J., Kirsch, J.F., Leyh, T.S., Moos, W.H.,
Petsko, G.A., Ringe, D., Rubin, G.M., Sheahan, L.C. & National Research Council Steering, C.
Defining the mandate of proteomics in the post-genomics era: workshop report. Mol Cell
Proteomics 1, 763-780 (2002).
Uhlen, M. Affinity as a tool in life science. Biotechniques 44, 649-654 (2008).
Uhlen, M. & Ponten, F. Antibody-based proteomics for human tissue profiling. Mol Cell
Proteomics 4, 384-393 (2005).
Ayoglu, B., Haggmark, A., Neiman, M., Igel, U., Uhlen, M., Schwenk, J.M. & Nilsson, P.
Systematic antibody and antigen-based proteomic profiling with microarrays. Expert Rev Mol
Diagn 11, 219-234 (2011).
57
References
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58
Schwenk, J.M., Gry, M., Rimini, R., Uhlen, M. & Nilsson, P. Antibody suspension bead arrays
within serum proteomics. J Proteome Res 7, 3168-3179 (2008).
Nishizuka, S., Charboneau, L., Young, L., Major, S., Reinhold, W.C., Waltham, M., KourosMehr, H., Bussey, K.J., Lee, J.K., Espina, V., Munson, P.J., Petricoin, E., 3rd, Liotta, L.A. &
Weinstein, J.N. Proteomic profiling of the NCI-60 cancer cell lines using new high-density
reverse-phase lysate microarrays. Proc Natl Acad Sci U S A 100, 14229-14234 (2003).
Stoevesandt, O. & Taussig, M.J. European and international collaboration in affinity
proteomics. N Biotechnol 29, 511-514 (2012).
Taussig, M.J., Stoevesandt, O., Borrebaeck, C.A., Bradbury, A.R., Cahill, D., Cambillau, C., de
Daruvar, A., Dubel, S., Eichler, J., Frank, R., Gibson, T.J., Gloriam, D., Gold, L., Herberg, F.W.,
Hermjakob, H., Hoheisel, J.D., Joos, T.O., Kallioniemi, O., Koegl, M., Konthur, Z., Korn, B.,
Kremmer, E., Krobitsch, S., Landegren, U., van der Maarel, S., McCafferty, J., Muyldermans,
S., Nygren, P.A., Palcy, S., Pluckthun, A., Polic, B., Przybylski, M., Saviranta, P., Sawyer, A.,
Sherman, D.J., Skerra, A., Templin, M., Ueffing, M. & Uhlen, M. ProteomeBinders: planning a
European resource of affinity reagents for analysis of the human proteome. Nat Methods 4,
13-17 (2007).
Haab, B.B., Paulovich, A.G., Anderson, N.L., Clark, A.M., Downing, G.J., Hermjakob, H., Labaer,
J. & Uhlen, M. A reagent resource to identify proteins and peptides of interest for the cancer
community: a workshop report. Mol Cell Proteomics 5, 1996-2007 (2006).
Juncker, D., Bergeron, S., Laforte, V. & Li, H. Cross-reactivity in antibody microarrays and
multiplexed sandwich assays: shedding light on the dark side of multiplexing. Current opinion
in chemical biology 18, 29-37 (2014).
Schwenk, J.M., Igel, U., Neiman, M., Langen, H., Becker, C., Bjartell, A., Ponten, F., Wiklund,
F., Gronberg, H., Nilsson, P. & Uhlen, M. Toward next generation plasma profiling via heatinduced epitope retrieval and array-based assays. Mol Cell Proteomics 9, 2497-2507 (2010).
MacBeath, G. Protein microarrays and proteomics. Nat Genet 32 Suppl, 526-532 (2002).
Poetz, O., Henzler, T., Hartmann, M., Kazmaier, C., Templin, M.F., Herget, T. & Joos, T.O.
Sequential multiplex analyte capturing for phosphoprotein profiling. Mol Cell Proteomics 9,
2474-2481 (2010).
Landegren, U., Vanelid, J., Hammond, M., Nong, R.Y., Wu, D., Ulleras, E. & KamaliMoghaddam, M. Opportunities for sensitive plasma proteome analysis. Anal Chem 84, 18241830 (2012).
Uhlen, M., Fagerberg, L., Hallstrom, B.M., Lindskog, C., Oksvold, P., Mardinoglu, A.,
Sivertsson, A., Kampf, C., Sjostedt, E., Asplund, A., Olsson, I., Edlund, K., Lundberg, E., Navani,
S., Szigyarto, C.A., Odeberg, J., Djureinovic, D., Takanen, J.O., Hober, S., Alm, T., Edqvist, P.H.,
Berling, H., Tegel, H., Mulder, J., Rockberg, J., Nilsson, P., Schwenk, J.M., Hamsten, M., von
Feilitzen, K., Forsberg, M., Persson, L., Johansson, F., Zwahlen, M., von Heijne, G., Nielsen, J.
& Ponten, F. Proteomics. Tissue-based map of the human proteome. Science 347, 1260419
(2015).
Uhlen, M., Oksvold, P., Fagerberg, L., Lundberg, E., Jonasson, K., Forsberg, M., Zwahlen, M.,
Kampf, C., Wester, K., Hober, S., Wernerus, H., Bjorling, L. & Ponten, F. Towards a
knowledge-based Human Protein Atlas. Nat Biotechnol 28, 1248-1250 (2010).
Ponten, F., Schwenk, J.M., Asplund, A. & Edqvist, P.H. The Human Protein Atlas as a
proteomic resource for biomarker discovery. J Intern Med 270, 428-446 (2011).
Mulder, J., Bjorling, E., Jonasson, K., Wernerus, H., Hober, S., Hokfelt, T. & Uhlen, M. Tissue
profiling of the mammalian central nervous system using human antibody-based proteomics.
Mol Cell Proteomics 8, 1612-1622 (2009).
Berglund, L., Bjorling, E., Jonasson, K., Rockberg, J., Fagerberg, L., Al-Khalili Szigyarto, C.,
Sivertsson, A. & Uhlen, M. A whole-genome bioinformatics approach to selection of antigens
for systematic antibody generation. Proteomics 8, 2832-2839 (2008).
Berglund, L., Bjorling, E., Oksvold, P., Fagerberg, L., Asplund, A., Szigyarto, C.A., Persson, A.,
Ottosson, J., Wernerus, H., Nilsson, P., Lundberg, E., Sivertsson, A., Navani, S., Wester, K.,
References
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
Kampf, C., Hober, S., Ponten, F. & Uhlen, M. A genecentric Human Protein Atlas for
expression profiles based on antibodies. Mol Cell Proteomics 7, 2019-2027 (2008).
Berglund, L., Andrade, J., Odeberg, J. & Uhlen, M. The epitope space of the human proteome.
Protein Sci 17, 606-613 (2008).
Larsson, M., Gräslund, S., Yuan, L., Brundell, E., Uhlén, M., Höög, C. & Ståhl, S. Highthroughput protein expression of cDNA products as a tool in functional genomics. Journal of
Biotechnology 80, 143-157 (2000).
Feinberg, J.G. A 'microspot' test for antigens and antibodies. Nature 192, 985-986 (1961).
Schena, M., Shalon, D., Davis, R.W. & Brown, P.O. Quantitative monitoring of gene
expression patterns with a complementary DNA microarray. Science 270, 467-470 (1995).
Ekins, R.P. Multi-analyte immunoassay. J Pharm Biomed Anal 7, 155-168 (1989).
Templin, M.F., Stoll, D., Schrenk, M., Traub, P.C., Vohringer, C.F. & Joos, T.O. Protein
microarray technology. Trends Biotechnol 20, 160-166 (2002).
Kusnezow, W., Syagailo, Y.V., Ruffer, S., Baudenstiel, N., Gauer, C., Hoheisel, J.D., Wild, D. &
Goychuk, I. Optimal design of microarray immunoassays to compensate for kinetic
limitations: theory and experiment. Mol Cell Proteomics 5, 1681-1696 (2006).
Kodadek, T. Protein microarrays: prospects and problems. Chemistry & biology 8, 105-115
(2001).
Zhu, H. & Snyder, M. Protein chip technology. Current opinion in chemical biology 7, 55-63
(2003).
LaBaer, J. & Ramachandran, N. Protein microarrays as tools for functional proteomics.
Current opinion in chemical biology 9, 14-19 (2005).
Phizicky, E., Bastiaens, P.I., Zhu, H., Snyder, M. & Fields, S. Protein analysis on a proteomic
scale. Nature 422, 208-215 (2003).
Paweletz, C.P., Charboneau, L., Bichsel, V.E., Simone, N.L., Chen, T., Gillespie, J.W., EmmertBuck, M.R., Roth, M.J., Petricoin, I.E. & Liotta, L.A. Reverse phase protein microarrays which
capture disease progression show activation of pro-survival pathways at the cancer invasion
front. Oncogene 20, 1981-1989 (2001).
Sheehan, K.M., Calvert, V.S., Kay, E.W., Lu, Y., Fishman, D., Espina, V., Aquino, J., Speer, R.,
Araujo, R., Mills, G.B., Liotta, L.A., Petricoin, E.F., 3rd & Wulfkuhle, J.D. Use of reverse phase
protein microarrays and reference standard development for molecular network analysis of
metastatic ovarian carcinoma. Mol Cell Proteomics 4, 346-355 (2005).
Zhu, H., Bilgin, M., Bangham, R., Hall, D., Casamayor, A., Bertone, P., Lan, N., Jansen, R.,
Bidlingmaier, S., Houfek, T., Mitchell, T., Miller, P., Dean, R.A., Gerstein, M. & Snyder, M.
Global analysis of protein activities using proteome chips. Science 293, 2101-2105 (2001).
Merbl, Y. & Kirschner, M.W. Protein microarrays for genome-wide posttranslational
modification analysis. Wiley Interdiscip Rev Syst Biol Med 3, 347-356 (2011).
Haab, B.B., Dunham, M.J. & Brown, P.O. Protein microarrays for highly parallel detection and
quantitation of specific proteins and antibodies in complex solutions. Genome Biol 2,
RESEARCH0004 (2001).
Sjoberg, R., Sundberg, M., Gundberg, A., Sivertsson, A., Schwenk, J.M., Uhlen, M. & Nilsson,
P. Validation of affinity reagents using antigen microarrays. N Biotechnol 29, 555-563 (2012).
Rosenberg, J.M. & Utz, P.J. Protein microarrays: a new tool for the study of autoantibodies in
immunodeficiency. Frontiers in Immunology 6 (2015).
Joos, T.O., Schrenk, M., Hopfl, P., Kroger, K., Chowdhury, U., Stoll, D., Schorner, D., Durr, M.,
Herick, K., Rupp, S., Sohn, K. & Hammerle, H. A microarray enzyme-linked immunosorbent
assay for autoimmune diagnostics. Electrophoresis 21, 2641-2650 (2000).
Haggmark, A., Hamsten, C., Wiklundh, E., Lindskog, C., Mattsson, C., Andersson, E., Lundberg,
I.E., Gronlund, H., Schwenk, J.M., Eklund, A., Grunewald, J. & Nilsson, P. Proteomic profiling
reveals autoimmune targets in sarcoidosis. Am J Respir Crit Care Med 191, 574-583 (2015).
59
References
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
60
Robinson, W.H. Antigen arrays for antibody profiling. Current opinion in chemical biology 10,
67-72 (2006).
Forsstrom, B., Axnas, B.B., Stengele, K.P., Buhler, J., Albert, T.J., Richmond, T.A., Hu, F.J.,
Nilsson, P., Hudson, E.P., Rockberg, J. & Uhlen, M. Proteome-wide epitope mapping of
antibodies using ultra-dense peptide arrays. Mol Cell Proteomics 13, 1585-1597 (2014).
Park, S., Gildersleeve, J.C., Blixt, O. & Shin, I. Carbohydrate microarrays. Chemical Society
Reviews 42, 4310-4326 (2013).
Wang, D., Liu, S., Trummer, B.J., Deng, C. & Wang, A. Carbohydrate microarrays for the
recognition of cross-reactive molecular markers of microbes and host cells. Nat Biotechnol
20, 275-281 (2002).
Pei, Z., Yu, H., Theurer, M., Walden, A., Nilsson, P., Yan, M. & Ramstrom, O. Photogenerated
carbohydrate microarrays. Chembiochem 8, 166-168 (2007).
Haab, B.B. Methods and applications of antibody microarrays in cancer research. Proteomics
3, 2116-2122 (2003).
Peluso, P., Wilson, D.S., Do, D., Tran, H., Venkatasubbaiah, M., Quincy, D., Heidecker, B.,
Poindexter, K., Tolani, N., Phelan, M., Witte, K., Jung, L.S., Wagner, P. & Nock, S. Optimizing
antibody immobilization strategies for the construction of protein microarrays. Anal Biochem
312, 113-124 (2003).
Butler, J.E., Ni, L., Brown, W.R., Joshi, K.S., Chang, J., Rosenberg, B. & Voss, E.W., Jr. The
immunochemistry of sandwich ELISAs--VI. Greater than 90% of monoclonal and 75% of
polyclonal anti-fluorescyl capture antibodies (CAbs) are denatured by passive adsorption.
Molecular immunology 30, 1165-1175 (1993).
Miller, J.C., Zhou, H., Kwekel, J., Cavallo, R., Burke, J., Butler, E.B., Teh, B.S. & Haab, B.B.
Antibody microarray profiling of human prostate cancer sera: antibody screening and
identification of potential biomarkers. Proteomics 3, 56-63 (2003).
Rimini, R., Schwenk, J.M., Sundberg, M., Sjöberg, R., Klevebring, D., Gry, M., Uhlén, M. &
Nilsson, P. Validation of serum protein profiles by a dual antibody array approach. Journal of
Proteomics 73, 252-266 (2009).
Wingren, C. & Borrebaeck, C.A. Antibody microarray analysis of directly labelled complex
proteomes. Curr Opin Biotechnol 19, 55-61 (2008).
Wilson, J.J., Burgess, R., Mao, Y.-Q., Luo, S., Tang, H., Jones, V.S., Weisheng, B., Huang, R.-Y.,
Chen, X. & Huang, R.-P. in Advances in Clinical Chemistry (Elsevier.
Sonntag, J., Schluter, K., Bernhardt, S. & Korf, U. Subtyping of breast cancer using reverse
phase protein arrays. Expert Rev Proteomics 11, 757-770 (2014).
Janzi, M., Kull, I., Sjoberg, R., Wan, J., Melen, E., Bayat, N., Ostblom, E., Pan-Hammarstrom,
Q., Nilsson, P. & Hammarstrom, L. Selective IgA deficiency in early life: association to
infections and allergic diseases during childhood. Clin Immunol 133, 78-85 (2009).
Gyorgy, A.B., Walker, J., Wingo, D., Eidelman, O., Pollard, H.B., Molnar, A. & Agoston, D.V.
Reverse phase protein microarray technology in traumatic brain injury. J Neurosci Methods
192, 96-101 (2010).
Akbani, R., Becker, K.F., Carragher, N., Goldstein, T., de Koning, L., Korf, U., Liotta, L., Mills,
G.B., Nishizuka, S.S., Pawlak, M., Petricoin, E.F., 3rd, Pollard, H.B., Serrels, B. & Zhu, J.
Realizing the promise of reverse phase protein arrays for clinical, translational, and basic
research: a workshop report: the RPPA (Reverse Phase Protein Array) society. Mol Cell
Proteomics 13, 1625-1643 (2014).
Taylor, S., Smith, S., Windle, B. & Guiseppi-Elie, A. Impact of surface chemistry and blocking
strategies on DNA microarrays. Nucleic Acids Res 31, e87 (2003).
VanMeter, A., Signore, M., Pierobon, M., Espina, V., Liotta, L.A. & Petricoin, E.F., 3rd Reversephase protein microarrays: application to biomarker discovery and translational medicine.
Expert Rev Mol Diagn 7, 625-633 (2007).
References
96.
97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
Calvert, V., Tang, Y., Boveia, V., Wulfkuhle, J., Schutz-Geschwender, A., Michael Olive, D.,
Liotta, L. & Petricoin, E., III Development of multiplexed protein profiling and detection using
near infrared detection of reverse-phase protein microarrays. Clin Proteom 1, 81-89 (2004).
Barbulovic-Nad, I., Lucente, M., Sun, Y., Zhang, M., Wheeler, A.R. & Bussmann, M. Biomicroarray fabrication techniques--a review. Critical reviews in biotechnology 26, 237-259
(2006).
Weibel, C. “The Spotting Accelerator™”, Customizable Head Assembly for Advanced
Microarraying. Journal of the Association for Laboratory Automation 7, 89-94 (2002).
McQuain, M.K., Seale, K., Peek, J., Levy, S. & Haselton, F.R. Effects of relative humidity and
buffer additives on the contact printing of microarrays by quill pins. Analytical Biochemistry
320, 281-291 (2003).
Austin, J. & Holway, A.H. Contact printing of protein microarrays. Methods Mol Biol 785, 379394 (2011).
Fang, Y. Air stability of supported lipid membrane spots. Chemical Physics Letters 512, 258262 (2011).
Nakanishi, K., Sakiyama, T. & Imamura, K. On the adsorption of proteins on solid surfaces, a
common but very complicated phenomenon. Journal of bioscience and bioengineering 91,
233-244 (2001).
Romanov, V., Davidoff, S.N., Miles, A.R., Grainger, D.W., Gale, B.K. & Brooks, B.D. A critical
comparison of protein microarray fabrication technologies. Analyst 139, 1303-1326 (2014).
McWilliam, I., Chong Kwan, M. & Hall, D. Inkjet printing for the production of protein
microarrays. Methods Mol Biol 785, 345-361 (2011).
Hartmann, M., Sjodahl, J., Stjernstrom, M., Redeby, J., Joos, T. & Roeraade, J. Non-contact
protein microarray fabrication using a procedure based on liquid bridge formation. Anal
Bioanal Chem 393, 591-598 (2009).
Fodor, S.P., Read, J.L., Pirrung, M.C., Stryer, L., Lu, A.T. & Solas, D. Light-directed, spatially
addressable parallel chemical synthesis. Science 251, 767-773 (1991).
Buus, S., Rockberg, J., Forsstrom, B., Nilsson, P., Uhlen, M. & Schafer-Nielsen, C. Highresolution mapping of linear antibody epitopes using ultrahigh-density peptide microarrays.
Mol Cell Proteomics 11, 1790-1800 (2012).
He, M. & Taussig, M.J. Single step generation of protein arrays from DNA by cell-free
expression and in situ immobilisation (PISA method). Nucleic Acids Res 29, E73-73 (2001).
Schmidt, R., Cook, E.A., Kastelic, D., Taussig, M.J. & Stoevesandt, O. Optimised 'on demand'
protein arraying from DNA by cell free expression with the 'DNA to Protein Array' (DAPA)
technology. J Proteomics 88, 141-148 (2013).
He, M., Stoevesandt, O. & Taussig, M.J. In situ synthesis of protein arrays. Curr Opin
Biotechnol 19, 4-9 (2008).
Ramachandran, N., Hainsworth, E., Bhullar, B., Eisenstein, S., Rosen, B., Lau, A.Y., Walter, J.C.
& LaBaer, J. Self-assembling protein microarrays. Science 305, 86-90 (2004).
Ramachandran, N., Raphael, J.V., Hainsworth, E., Demirkan, G., Fuentes, M.G., Rolfs, A., Hu,
Y. & LaBaer, J. Next-generation high-density self-assembling functional protein arrays. Nat
Methods 5, 535-538 (2008).
Clark, D.C., Mackie, A.R., Wilde, P.J. & Wilson, D.R. Differences in the Structure and Dynamics
of the Adsorbed Layers in Protein-Stabilized Model Foams and Emulsions. Faraday Discuss
98, 253-262 (1994).
Deegan, R.D., Bakajin, O., Dupont, T.F., Huber, G., Nagel, S.R. & Witten, T.A. Capillary flow as
the cause of ring stains from dried liquid drops. Nature 389, 827-829 (1997).
Deng, Y., Zhu, X.Y., Kienlen, T. & Guo, A. Transport at the air/water interface is the reason for
rings in protein microarrays. J Am Chem Soc 128, 2768-2769 (2006).
61
References
116.
117.
118.
119.
120.
121.
122.
123.
124.
125.
126.
127.
128.
129.
130.
131.
132.
133.
134.
62
Olle, E.W., Messamore, J., Deogracias, M.P., McClintock, S.D., Anderson, T.D. & Johnson, K.J.
Comparison of antibody array substrates and the use of glycerol to normalize spot
morphology. Experimental and Molecular Pathology 79, 206-209 (2005).
Mackie, A.R., Gunning, A.P., Wilde, P.J. & Morris, V.J. Orogenic displacement of protein from
the air/water interface by competitive adsorption. J Colloid Interf Sci 210, 157-166 (1999).
Angenendt, P., Glokler, J., Sobek, J., Lehrach, H. & Cahill, D.J. Next generation of protein
microarray support materials: evaluation for protein and antibody microarray applications. J
Chromatogr A 1009, 97-104 (2003).
Danczyk, R., Krieder, B., North, A., Webster, T., HogenEsch, H. & Rundell, A. Comparison of
antibody functionality using different immobilization methods. Biotechnol Bioeng 84, 215223 (2003).
Sauer, U. Impact of substrates for probe immobilization. Methods Mol Biol 785, 363-378
(2011).
Derwinska, K., Gheber, L.A. & Preininger, C. A comparative analysis of polyurethane hydrogel
for immobilization of IgG on chips. Anal Chim Acta 592, 132-138 (2007).
Harbers, G.M., Emoto, K., Greef, C., Metzger, S.W., Woodward, H.N., Mascali, J.J., Grainger,
D.W. & Lochhead, M.J. A functionalized poly(ethylene glycol)-based bioassay surface
chemistry that facilitates bio-immobilization and inhibits non-specific protein, bacterial, and
mammalian cell adhesion. Chemistry of materials : a publication of the American Chemical
Society 19, 4405-4414 (2007).
Sobek, J., Aquino, C., Weigel, W. & Schlapbach, R. Drop drying on surfaces determines
chemical reactivity - the specific case of immobilization of oligonucleotides on microarrays.
BMC Biophys 6, 8 (2013).
Ray, S., Mehta, G. & Srivastava, S. Label-free detection techniques for protein microarrays:
prospects, merits and challenges. Proteomics 10, 731-748 (2010).
Forster, T., Costa, Y., Roy, D., Cooke, H.J. & Maratou, K. Triple-target microarray experiments:
a novel experimental strategy. BMC Genomics 5, 13 (2004).
Fare, T.L., Coffey, E.M., Dai, H., He, Y.D., Kessler, D.A., Kilian, K.A., Koch, J.E., LeProust, E.,
Marton, M.J., Meyer, M.R., Stoughton, R.B., Tokiwa, G.Y. & Wang, Y. Effects of atmospheric
ozone on microarray data quality. Anal Chem 75, 4672-4675 (2003).
Cox, W.G., Beaudet, M.P., Agnew, J.Y. & Ruth, J.L. Possible sources of dye-related signal
correlation bias in two-color DNA microarray assays. Anal Biochem 331, 243-254 (2004).
Staal, Y.C., van Herwijnen, M.H., van Schooten, F.J. & van Delft, J.H. Application of four dyes
in gene expression analyses by microarrays. BMC Genomics 6, 101 (2005).
Berlier, J.E., Rothe, A., Buller, G., Bradford, J., Gray, D.R., Filanoski, B.J., Telford, W.G., Yue, S.,
Liu, J., Cheung, C.Y., Chang, W., Hirsch, J.D., Beechem, J.M., Haugland, R.P. & Haugland, R.P.
Quantitative comparison of long-wavelength Alexa Fluor dyes to Cy dyes: fluorescence of the
dyes and their bioconjugates. J Histochem Cytochem 51, 1699-1712 (2003).
Chan, W.C., Maxwell, D.J., Gao, X., Bailey, R.E., Han, M. & Nie, S. Luminescent quantum dots
for multiplexed biological detection and imaging. Curr Opin Biotechnol 13, 40-46 (2002).
Espina, V., Woodhouse, E.C., Wulfkuhle, J., Asmussen, H.D., Petricoin, E.F., 3rd & Liotta, L.A.
Protein microarray detection strategies: focus on direct detection technologies. J Immunol
Methods 290, 121-133 (2004).
Schweitzer, B., Wiltshire, S., Lambert, J., O'Malley, S., Kukanskis, K., Zhu, Z., Kingsmore, S.F.,
Lizardi, P.M. & Ward, D.C. Immunoassays with rolling circle DNA amplification: a versatile
platform for ultrasensitive antigen detection. Proc Natl Acad Sci U S A 97, 10113-10119
(2000).
Schweitzer, B., Roberts, S., Grimwade, B., Shao, W., Wang, M., Fu, Q., Shu, Q., Laroche, I.,
Zhou, Z., Tchernev, V.T., Christiansen, J., Velleca, M. & Kingsmore, S.F. Multiplexed protein
profiling on microarrays by rolling-circle amplification. Nat Biotechnol 20, 359-365 (2002).
Sang, S., Wang, Y., Feng, Q., Wei, Y., Ji, J. & Zhang, W. Progress of new label-free techniques
for biosensors: a review. Critical reviews in biotechnology, 1-17 (2015).
References
135.
136.
137.
138.
139.
140.
141.
142.
143.
144.
145.
146.
147.
Scarano, S., Mascini, M., Turner, A.P. & Minunni, M. Surface plasmon resonance imaging for
affinity-based biosensors. Biosens Bioelectron 25, 957-966 (2010).
Scherl, A. Clinical protein mass spectrometry. Methods.
Boozer, C., Kim, G., Cong, S., Guan, H. & Londergan, T. Looking towards label-free
biomolecular interaction analysis in a high-throughput format: a review of new surface
plasmon resonance technologies. Current Opinion in Biotechnology 17, 400-405 (2006).
Frederix, F., Bonroy, K., Reekmans, G., Laureyn, W., Campitelli, A., Abramov, M.A., Dehaen,
W. & Maes, G. Reduced nonspecific adsorption on covalently immobilized protein surfaces
using poly(ethylene oxide) containing blocking agents. Journal of biochemical and biophysical
methods 58, 67-74 (2004).
MacBeath, G. & Schreiber, S.L. Printing proteins as microarrays for high-throughput function
determination. Science 289, 1760-1763 (2000).
Walter, J.-G., Stahl, F., Reck, M., Praulich, I., Nataf, Y., Hollas, M., Pflanz, K., Melzner, D.,
Shoham, Y. & Scheper, T. Protein microarrays: Reduced autofluorescence and improved LOD.
Engineering in Life Sciences 10, 103-108 (2010).
Sill, M., Schroder, C., Hoheisel, J.D., Benner, A. & Zucknick, M. Assessment and optimisation
of normalisation methods for dual-colour antibody microarrays. BMC Bioinformatics 11, 556
(2010).
Eckel-Passow, J.E., Hoering, A., Therneau, T.M. & Ghobrial, I. Experimental design and
analysis of antibody microarrays: applying methods from cDNA arrays. Cancer Res 65, 29852989 (2005).
Olle, E.W., Sreekumar, A., Warner, R.L., McClintock, S.D., Chinnaiyan, A.M., Bleavins, M.R.,
Anderson, T.D. & Johnson, K.J. Development of an internally controlled antibody microarray.
Mol Cell Proteomics 4, 1664-1672 (2005).
Zhang, L., Wei, Q., Mao, L., Liu, W., Mills, G.B. & Coombes, K. Serial dilution curve: a new
method for analysis of reverse phase protein array data. Bioinformatics 25, 650-654 (2009).
Anderson, T., Wulfkuhle, J., Liotta, L., Winslow, R.L. & Petricoin, E., 3rd Improved
reproducibility of reverse-phase protein microarrays using array microenvironment
normalization. Proteomics 9, 5562-5566 (2009).
Kaushik, P., Molinelli, E.J., Miller, M.L., Wang, W., Korkut, A., Liu, W., Ju, Z., Lu, Y., Mills, G. &
Sander, C. Spatial normalization of reverse phase protein array data. PLoS One 9, e97213
(2014).
Structural Genomics, C., China Structural Genomics, C., Northeast Structural Genomics, C.,
Graslund, S., Nordlund, P., Weigelt, J., Hallberg, B.M., Bray, J., Gileadi, O., Knapp, S.,
Oppermann, U., Arrowsmith, C., Hui, R., Ming, J., dhe-Paganon, S., Park, H.W., Savchenko, A.,
Yee, A., Edwards, A., Vincentelli, R., Cambillau, C., Kim, R., Kim, S.H., Rao, Z., Shi, Y.,
Terwilliger, T.C., Kim, C.Y., Hung, L.W., Waldo, G.S., Peleg, Y., Albeck, S., Unger, T., Dym, O.,
Prilusky, J., Sussman, J.L., Stevens, R.C., Lesley, S.A., Wilson, I.A., Joachimiak, A., Collart, F.,
Dementieva, I., Donnelly, M.I., Eschenfeldt, W.H., Kim, Y., Stols, L., Wu, R., Zhou, M., Burley,
S.K., Emtage, J.S., Sauder, J.M., Thompson, D., Bain, K., Luz, J., Gheyi, T., Zhang, F., Atwell, S.,
Almo, S.C., Bonanno, J.B., Fiser, A., Swaminathan, S., Studier, F.W., Chance, M.R., Sali, A.,
Acton, T.B., Xiao, R., Zhao, L., Ma, L.C., Hunt, J.F., Tong, L., Cunningham, K., Inouye, M.,
Anderson, S., Janjua, H., Shastry, R., Ho, C.K., Wang, D., Wang, H., Jiang, M., Montelione, G.T.,
Stuart, D.I., Owens, R.J., Daenke, S., Schutz, A., Heinemann, U., Yokoyama, S., Bussow, K. &
Gunsalus, K.C. Protein production and purification. Nat Methods 5, 135-146 (2008).
63