Functional Genome Analysis  (B070)
Deutsches Krebsforschungszentrum, Im Neuenheimer Feld 580
D-69120 Heidelberg, Germany.




  Affinity Proteomics  -  Antibody Microarrays
   Antibody Microarrays

Biomedical Studies
   Detailed protocols
  Single-molecule detection

   Directed stability improvement
   Pancreatic cancer (PDAC)

     PDAC microenvironment

   Effective binder selection

     PDAC cell secretome

   Sample preparation

   Affinomics consortium
   Gastric cancer

     PDAC cell proteome

   Mass transport   Affinity Proteomics Group
   Binder sequence validation
   Bladder cancer recurrence


As an immediate consequence of large-scale genomic sequencing, strong interest has emerged in analysing the function of the DNA-encoded information on a similarly global scale. However, many aspects of modulation and regulation of cellular activity cannot be investigated at the level of nucleic acids but require an analysis of the proteome. Post-transcriptional control of protein translation, post-translational modifications as well as protein degradation by proteolysis, for example, have profound effects at the functional level. Estimations suggest that there are more than 200 types of protein modification. Its proportion and importance is reflected by the fact that 5% to 10% of mammalian genes encode for proteins that modify other proteins.
The complexity in the human proteome is expected to range from a hundred thousand to several million different protein molecules. With respect to data interpretation, the situation is additionally complicated by the facts that not for every protein of multicellular organisms the function is known and that there may be different functions dependent on structure variations or interacting partners. In addition, the dynamic range of protein expression is very large indeed.
Various mass spectrometry based processes exist for a powerful analysis of proteins of an organism or tissue. Also, assays such as yeast-two-hybrid analyses in all their facets permit global studies for the identification of interaction partners. Nevertheless, many other, possibly even more powerful methods are prerequisite to approaching the world of protein analysis in a manner similar to what is already possible for studies at the level of nucleic acids, and beyond.
Protein microarrays have an enormous potential of developing into a tool that will allow global characterisation of molecule mixtures at the protein level. Knowledge of genomic sequences and transcriptional profiles do not suffice for a reliable description of actual protein expression, let alone an analysis of protein structures and biochemical activities or a quantitative examination of protein-protein interactions. This kind of information, however, is crucial for a molecular understanding the biology of cells, tissues or whole organisms and has a broad biotechnical and medical potential. Utilising recently developed processes, we perfom such analyses on a large-scale with nevertheless high reproducibility, a near-single-molecule sensitivity, and an accuracy that is as good as or even superior to ELISA-based assays.

Antibody microarrays:
Utilising antibody microarrays, we pursue the analysis of studying variations in actual protein abundance, isoform occurrence and other structural variations, in particular in cells, tissues and body liquids. Basic technical processes were studied in much detail, such as appropriate surfaces, the effect of kinetics and mass transport and labelling procedures as well as many other aspects in order to establish a working system. Detailed protocols are available that allow reproducible and reliable analysis of expression variations on complex protein extracts from tissues, cells or body liquids down to attomolar concentrations. Antibody generation and selection was and is performed in collaborations with company partners as well as EU-funded initiatives that aim at the creation of well-characterised and specific antibodies or other binders (e.g., Affinomics). In addition, improvements of the preparation of protein extracts proved crucial for success. The current set-up was and is used in various projects, frequently combining the information on protein levels with other data. Also, quantification of the results is performed, either by actual counting of individual molecules or by an analysis of dissociation parameters.

Protein microarrays:
We utilise protein microarrays containing (mostly full-length) molecules for the investigation of protein-interactions in a quantitative manner. Microarray production is done by in situ synthesis by an transcription/translation process on the microarrays, starting from full-length cDNAs. Protein interaction of all kinds as well as the influence of co-factors such as small molecules can be studied this way. The largest protein array produced so far contains some 14,000 individual proteins. The current set-up was and is used in various projects for actual measurements, frequently combining the information on protein levels with other data.
Personalised proteomics:
In a recently finished technical development, we added to the in situ protein production a process that allows to present on the microarray the proteins in exactly the conformation as they occur in tissues or other samples of individual patients, reflecting all mutations or splice variants that are specific for the particular sample. Thereby, particularly the effects of individual variations on protein interaction - with other proteins, nucleic acids or samll compounds - can be studied in a quantitative manner.

Mustafa et al. (2017) Oncotarget 8, 11963-11976. pdf icon

Syafrizayanti et al. (2017) Sci. Rep. 7, 39756. pdf icon
Schmidt et al. (2011) J. Prot. Res. 10, 1316. pdf icon

Kamhieh-Milz et al. (2016) J. Proteomics 150, 74-85. pdf icon
Schröder et al. (2011) Protein Micoarrays - Meth. Mol. Biol., 203. pdf icon

Bakdash et al. (2016) Cancer Res. 76, 4332-4346. pdf icon
Alhamdani & Hoheisel (2011) Mol. Anal. & Genome Disc., Wiley, 219.

Loeffler et al. (2016) Nature Comm. 7, 11844. pdf icon
Sill et al. (2010) BMC Bioinformatics 11, 556. pdf icon

Sill et al. (2016) Microarrays 5, 19. pdf icon
Alhamdani et al. (2010) Proteomics 10, 3203. pdf icon

Kibat et al. (2016) New Biotechnol. 33, 574-581. pdf icon
Schröder et al. (2010) Antibody Engineer., Vol. 2, Springer, 429. pdf icon

Bal et al. (2016) Br. J. Haematology 4, 602-615. pdf icon
Alhamdani et al. (2010) J. Prot. Res. 9, 963. pdf icon

Nijaguna et al. (2015) J. Proteomics 128, 251-261. pdf icon
Schröder et al. (2010) Mol. Cell. Prot. 9, 1271. pdf icon

Mock et al. (2015) Oncotarget 6, 13579-13590. pdf icon
Gloriam et al. (2010) Mol. Cell. Prot. 9, 1. pdf icon

Betzen et al. (2015) Proteomics Clin. Appl. 9, 342. pdf icon
Alhamdani et al. (2009) Genome Med. 1, 68. pdf icon

Bradbury et al. (2015) Nature 518, 27. pdf icon
Börner et al. (2009) BioTechniques 46, 297. pdf icon

Hoheisel (2014) labor&more 10/14, 10. pdf icon
Taussig et al. (2007) Nature Meth. 4, 13. pdf icon

Srinivasan et al. (2014) Proteomics 14, 1333. pdf icon
Kusnezow et al. (2007) Proteomics 7, 1786. pdf icon

Syafrizayanti et al. (2014) Exp. Rev. Prot. 11, 107.
pdf icon
Kusnezow et al. (2006) Mol. Cell. Prot. 5, 1681. pdf icon

Marzoq et al. (2013) J. Biol. Chem. 288, 32517. pdf icon
Angenendt et al. (2006) Mol. Cell. Prot. 5, 1658. pdf icon

Lueong et al. (2013) J. Prot. Bioinf. 07, 004. pdf icon
Kusnezow et al. (2006) Proteomics 6, 794. pdf icon

Schröder et al. (2013) Proteomics Clin. Appl. 7, 802. pdf icon
Kersten et al. (2005) Expert Rev. Proteomics 2, 499. pdf icon

Hoheisel et al. (2013) Proteomics Clin. Appl. 7, 8. pdf icon
Kusnezow & Hoheisel .(2003) J. Mol. Recognit. 16, 165. pdf icon

Alhamdani et al. (2012) J. Proteomics 75, 3747. pdf icon
Kusnezow et al. (2003) Proteomics 3, 254. pdf icon

Friedrich et al. (2011) Proteomics 11, 3757. pdf icon
Kusnezow & Hoheisel (2002) BioTechniques 33, 14. pdf icon

AUC values of the 189 individual serum markers. Analysis by Receiver Operating Characteristic (ROC) curves was performed for all identified serum protein markers individually. Panel A shows the result calculated from the training set; the respective AUC values are shown, ranging from 55.2% to 96.0%. In panel B, the AUC values are shown as calculated for the individual marker molecules in the test set. For presentation, the order of the markers along the x-axis was kept as in panel A, highlighting the limited degree of reproducibility for individual markers.
Comparison of the tumour cell secretome and patient sera for an accurate serum-based diagnosis of pancratic ductal adenocarcinoma

Pancreatic cancer is the currently most lethal malignancy. Toward an accurate diagnosis of the disease in body liquids, we studied the protein composition of the secretomes of 16 primary and established cell lines of pancreatic ductal adenocarcinoma (PDAC). Compared to the secretome of non-tumorous cells, 112 proteins exhibited significantly different abundances. Functionally, the proteins were associated with PDAC features, such as decreased apoptosis, better cell survival and immune cell regulation.
The result was compared to profiles obtained from 164 serum samples from two independent cohorts – a training and a test set – of patients with PDAC or chronic pancreatitis and healthy donors. Eight of the 112 secretome proteins exhibited similar variations in their abundance in the serum profile specific for PDAC patients, which was composed of altogether 189 proteins.
The 8 markers shared by secretome and serum yielded a 95.1% accuracy of distinguishing PDAC from healthy in a Receiver Operating Characteristic curve analysis, while any number of serum-only markers produced substantially less accurate results. Utility of the identified markers was confirmed by classical enzyme linked immunosorbent assays (ELISAs). The study highlights the value of cell secretome analysis as a means of defining reliable serum biomarkers.
Mustafa et al. (2017) Oncotarget 8, 11963-11976. pdf icon

Utilisation of antibody microarrays for the selection of specific and informative antibodies from library binders of unknown quality

Many diagnostic and therapeutic concepts require antibodies of high specificity. Recombinant binder libraries and related selection approaches allow the isolation of antibodies against almost every target of interest. Nevertheless, it cannot be guaranteed that selected antibodies perform well and interact specifically enough with analytes unless an elaborate characterisation is performed. Here, we present an approach to shorten this process by combining the selection of suitable antibodies with the identification of informative target molecules by means of antibody microarrays, thereby reducing the effort of antibody characterisation by concentrating on relevant molecules.
In a pilot scheme, a library of 456 single-chain variable fragment (scFv) binders to 134 antigens was used. They were arranged in a microarray format and incubated with the protein content of clinical tissue samples isolated from pancreatic ductal adenocarcinoma and healthy pancreas, as well as recurrent and non-recurrent bladder tumours. We observed significant variation in the expression of the E3 ubiquitin-protein ligase (CHFR) as well as the glutamate receptor interacting protein 2 (GRIP2), for example, always with more than one of the scFvs binding to these targets. Only the relevant antibodies were then characterised further on antigen microarrays and by surface plasmon resonance experiments so as to select the most specific and highest affinity antibodies. These binders were in turn used to confirm a microarray result by immunohistochemistry analysis.

Kibat et al. (2016) New Biotechnol. 33, 574-581. pdf icon

Scheme of the selection and characterisation process for a highly effective isolation of both specific and informative antibodies (or other binders) from a library of uncharacterised molecules.

Protein profiling of gastric cancer and neighbouring control tissues

Protein profiling was performed on gastric cancer tissue samples. Sixteen pairs of postoperative gastric adenocarcinomas and adjacent non-cancerous control tissues were analysed on microarrays that contain 813 antibodies targeting 724 proteins. Only 17 proteins were found to be differentially regulated, much fewer molecules than the number identified when comparing tumour to healthy control tissues. Insulin-like growth factor-binding protein 7 (IGFBP7), S100 calcium binding protein A9 (S100A9), interleukin-10 (IL-10) and mucin 6 (MUC6) exhibited the most profound variations. For an evaluation of the proteins’ capacity for discriminating gastric cancer, a Receiver Operating Characteristic curve analysis was performed. For confirmation, immunohistological analyses were done on samples prepared from another cohort of patients with gastric cancer.

Figure legend: Typical results of immunohistochemical analyses. Identified marker molecules were validated by immunohistochemistry on an independent set of tumour tissues and stomach samples from donors who had no cancer. Dark brown colour is indicative of the presence of the respective proteins.

Sill et al. (2016) Microarrays 5, 19. pdf icon

map of partner locations
Affinomics: Proteome binders for characterisation of human proteome function; generation, validation, application
       logo EU FP7

The Affinomics programme aims to leverage existing efforts in Europe to generate large-scale resources of validated protein-binding molecules (binders) as affinity reagents for characterisation of the human proteome and to apply them in comprehensive structural and functional analyses of protein expression, interactions and complexes. The project was preceded by the ProteomeBinders and AffinityProteome consortia.
Proteome targets will be focused on five categories of inter-related human proteins involved in signal transduction, cell regulation and cancer, namely protein kinases, SH2 domain-containing proteins, protein tyrosine phosphatases, proteins somatically mutated in cancers and candidate cancer biomarkers. Binders to about 1000 protein targets will be made over the course of the programme.
A high throughput, coordinated production pipeline for antigens and binders will be established. Target antigens will be expressed in three forms, as folded full-length proteins or domains, as large peptide fragments (PrESTs) based on low homology to other human proteins and as small peptides, in some cases phosphorylated. Binder types to be generated include affinity-purified polyclonal antibodies, monoclonal antibodies, recombinant antibody fragments and non-immunoglobulin scaffolds.
An important aspect will be the development of highly efficient next generation recombinant selection methods, based on phage, cell and ribosome display, capable of producing high quality binders at greater throughput and lower cost than hitherto. Systems and procedures for thorough binder validation and quality control will be established. The affinity reagents will be applied in advanced innovative and sensitive technologies for specific detection of target proteins and interacting protein complexes in cells, tissues and fluids, for improved understanding of protein function and new classes of diagnostic assays.
For more detailed information, click on the map of the consortium.
Sequence validation of binders
Antibodies and other binder types are crucial for any proteome analysis. Although many tens of thousands of antibodies are available from commercial sources and academic institutions (see antibodypedia, for example), quality and reproducibility of analyses performed with these molecules vary substantially. Antibodies exhibit huge differences in specificity and affinity, including binders that target the same protein. Worse, even binders that are supposedly from the same source showed significant lot-to-lot variation.
One reason for variation could be the actual assay; an antibody perfoming very well on Western blots may not meet the quality standards of pull-down experiments or microarray analyses, and vice versa. Another point, however, that is contributing significantly to low reproducibility is a lack of commonly accepted performance parameters and tests for their definition This is made worse by the fact that currently most binders are ill-described. Consequently, one cannot be sure, if a binder is exactly the molecule that was used in assays before.
Initiated by Andrew Bradbury, 101 scientists of the Affinomics consortium and beyond have had some thoughts about the matter. A simple solution to the problem of how to make sure that everybody is really using the very same binder molecules in their experiments would be a sequence verification of each binder. While relatively easily achievable for recombinant binders and monoclonal antibodies, it is more difficult to establish for polyclonal binders, for example. A more important obstacle for an implementation of such a scheme, however, could be the reluctance of antibody producers to share the sequences of their molecules, since this would make available their good binders to the entire scientific (and commercial) community for free.

Bradbury et al. (2015) Nature 518, 27-29.  pdf icon

Prediction of recurrence of non muscle-invasive bladder cancer

About 70% of newly diagnosed cases of bladder cancer are low-stage, low-grade, non muscle-invasive. Standard treatment is transurethral resection. About 60% of the tumours will recur, however, and in part progress to become invasive. Therefore, surveillance cystoscopy is performed after resection. In the USA and Europe alone, about 54,000 new patients per year undergo repeated cystoscopies over several years, who do not experience recurrence. Analysing in a pilot study resected tumours from patients with and without local recurrence after a period of five years, we identified 255 proteins with significantly differential abundance. Most are involved in the regulation and execution of apoptosis and cell proliferation. A multivariate classifier was constructed based on 20 proteins. It facilitates the prediction of recurrence with a sensitivity of 80% and a specificity of 100%. As a measure of overall accuracy, the area under the curve (AUC) value was found to be 91%. After validation, such a signature could help to adjust the treatment scheme and the rigidity of surveillance. This could dramatically reduce surveillance cost and simultaneously improve the patients’ outcome substantially.

Figure legend: Strong repression of the TGF-beta signalling pathway in recurrent cancer. Affected proteins are labelled in green or red if their expression was lower or higher, respectively, in recurrent rather than non-recurrent cancer.

Srinivasan et al. (2014) Proteomics 14, 1333. pdf icon

images of different incubations
Detailed protocols for expression profiling by antibody microarrays

        logo BMBF                    logo NGFN
As a multiplexing technique, antibody microarrays facilitate the highly parallel detection of hundreds of different analytes from very small sample volumes of only few microliters. This is combined with a high sensitivity in the picomolar to femtomolar range, which is similar to the sensitivity of ELISA, the gold standard for protein quantification. In order to obtain such sensitivities in a robust and reproducible manner for sets of several hundreds of analytes, it is essential to optimise the experimental layout, sample handling, labelling and incubation as well as data processing steps.
Based on earlier work, we continuously developed the processing of microarrays and protein samples. In the publications listed below, we present in detail our antibody microarray protocols for multiplexed expression profiling studies, which permit the analysis of the abundance of very many proteins in plasma, urine, cell and tissue samples.

Schröder et al. (2010) Antibody Engineering, Vol. 2, SpringerVerlag, 429-445. pdf icon

Schröder et al. (2010) Mol. Cell. Prot. 9, 1271-1280. pdf icon

Alhamdani et al. (2010) Proteomics 10, 3203-3207. pdf icon

Schröder et al. (2011) Protein Micoarrays - Meth. Mol. Biol., Springer, 203-221.


DKFZ Homepage
Top of Page