SEARCH

SEARCH BY CITATION

Keywords:

  • biomarkers;
  • in vitro diagnostics;
  • personalized medicine;
  • post-translational modifications;
  • protein–protein interactions;
  • proximity ligation assay

Abstract

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

Abstract.  Blokzijl A, Friedman M, Pontén F, Landegren U (From the Department of Genetics and Pathology, Uppsala University, Uppsala, Sweden). Profiling protein expression and interactions: proximity ligation as a tool for personalized medicine. (Review) J Intern Med 2010; 268: 232–245.

The ability to detect very low levels of expressed proteins has enormous potential for early diagnostics and intervention at curable stages of disease. An extended range of targets such as interacting or post-translationally modified proteins can further improve the potential for diagnostics and patient stratification, and for monitoring response to treatment. These are critical building blocks for personalized treatment strategies to manage disease. The past few decades have seen a remarkably improved understanding of the molecular basis of disease in general, and of tumour formation and progression in particular. This accumulated knowledge creates opportunities to develop drugs that specifically target molecules or molecular complexes critical for survival and expansion of tumour cells. However, tumours are highly variable between patients, necessitating the development of diagnostic tools to individualize treatment through parallel analysis of sets of biomarkers.

The proximity ligation assay (PLA) can address many of the requirements for advanced molecular analysis. The method builds on the principle that recognition of target proteins by two, three or more antibodies can bring in proximity DNA strands attached to the antibodies. The DNA strands can then participate in ligation reactions, giving rise to molecules that are amplified for highly sensitive detection. PLA is particularly well suited for sensitive, specific and multiplexed analysis of protein expression, post-translational modifications and protein–protein interactions. The analysis of this extended range of biomarkers will prove critical for the development and implementation of personalized medicine.


Opportunities and challenges for protein diagnostics

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

New developments in molecular diagnostics provide hope for greatly improved opportunities to diagnose and distinguish forms of disease, and to select optimal therapy. Advances in DNA sequencing greatly reduce cost and increase speed of data acquisition. These new methods allow the genetic landscape of individual tumours to be recorded before and during treatment and can provide comprehensive measurements of epigenetic changes and of gene expression at the RNA level. However, protein analyses can provide important insights into disease states that cannot be readily accessed by DNA sequencing. The potential to observe disease processes anywhere in the body at early stages by protein measurements in blood samples is a very attractive concept. Similarly, molecular inspection of the signalling cascades that are targeted with expensive new cancer drugs could help to stratify patients according to expected responsiveness; this is becoming increasingly important both during drug development and in routine clinical care.

Opportunities for protein diagnostics, the topic of this article, have been hampered by a series of obstacles that render protein analyses considerably more difficult than the analyses of nucleic acids. Major challenges include the fact that the concentration of different cellular proteins can vary considerably [1] and differ by far more than billion-fold in serum or plasma [2], and the presence of multiple isoforms of proteins that are highly similar but with different roles. Moreover, the functions of proteins are generally exerted by complexes of interacting molecules, necessitating detection not only of individual molecules but also of these complexes.

Unlike the situation for DNA detection, there are no simple rules for the construction of affinity reagents for proteins, and accordingly suitable reagents can only be found using the immune systems of animals, or by searching through large molecular libraries in vitro. The requirements for detection of denatured proteins in western blot or in situ may differ from those for detecting native epitopes of proteins for example in plasma. Major projects such as the Human Protein Atlas (HPA; further discussed below) are conducted to develop and validate antibodies in a standardized way [3].

In this review, we will discuss the future diagnostic needs for an extended range of molecular analyses. Specifically, we will describe the potential of the proximity ligation assay (PLA) to provide information at the level of proteins beyond the capabilities of earlier methods. We discuss opportunities to expand the scope for diagnostics by detection of new classes of protein or protein complex biomarkers in blood and of using in situ PLA for imaging the expression of proteins, protein–protein interactions and post-translational modifications in tissues with subcellular resolution (Fig. 1).

image

Figure 1. Multiplex analysis with single-cell resolution has the potential to reveal altered expression levels, post-translational modifications and protein–protein interactions of several signalling pathways in parallel with response to treatment.

Download figure to PowerPoint

Detection of secreted biomarkers for diagnostics and evaluation of disease progression

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

Few of the many promising candidate biomarkers that are reported in the scientific literature go on to become part of clinical practice [4]. Many candidates fail at the validation stage, where strict requirements are applied for evaluating the specificity and sensitivity of the tests in distinguishing signs of health and disease. However, the availability of good biomarkers is fundamental in diagnostics and for patient follow-up.

Considerable efforts are being made worldwide to discover factors in blood that can be used as biomarkers for specific diseases, and to monitor disease progression and relapse. Biomarkers in body fluids have obvious advantages of accessibility. For example, minimally invasive investigations of specific protein profiles in serum or in circulating tumour cells could allow early diagnosis of primary tumours or metastases before they reach a size that permits visualization by radiological techniques. Factors can originate from the malignant cells, but noncancerous stromal cells adjacent to tumour cells can also be the source of secreted factors of use in diagnostics. This is exemplified by the secretion of stromal cell-derived factor 1 (SDF-1) from carcinoma-associated fibroblasts in breast tumours [5]. There is a need for methods of protein detection that are far more sensitive than the presently available techniques, to reveal the presence of proteins in plasma in minute quantities that may reflect pathological processes. It is currently possible to detect transcription factors such as p53 in serum using enzyme-linked immunosorbent assay (ELISA) [6]. Other transcription factors with more restricted expression profiles may also be present in plasma below the limits of detection of current analysis methods. Detection of these transcription factors in plasma has the potential to provide added value in the diagnosis of disease processes by identifying the tissue of origin of disease.

Proteomic approaches reveal large numbers of potential biomarkers [7]. The investigated molecules can be derived for example from conditioned media of tumour cell lines or from tumour tissue. Several approaches have been applied to detect these potential biomarkers and validate the findings. Recent studies have linked the profiles of proteins released in vitro from cells representing several different tumour types to protein expression profiles in their tissues of origin, as described for example in the HPA (http://www.proteinatlas.org/) [8]. The current version of the freely available HPA contains the protein profiles for over 8.800 unique human proteins, corresponding to approximately 42% of the estimated number of human protein coding genes [9], with access to protein expression data and underlying high-resolution images showing protein expression patterns in normal human tissues from 144 individuals and cancer tissues from 216 patients. The HPA database can be utilized as a starting point for in silico biomarker discovery [10]. Although a large fraction of human proteins are ubiquitously expressed, as shown in a recent global analysis of protein expression patterns [11], the identification of proteins that show cell type–specific expression is of utmost importance for finding diagnostic markers.

A particularly promising class of cancer biomarkers is factors that are normally secreted into the lumen of hollow organs such as the prostate, breast, intestine, lung or the gall and urinary bladders by polarized epithelial cells lining the cavity. An example of such a biomarker is the prostate-specific antigen that is released in high concentrations into seminal fluid. Upon tissue invasion by malignant cells through the basal membrane, molecules secreted from the malignant cells may no longer reach the lumen, but appear at increased levels in the blood [12].

Tissue biomarkers for differential diagnosis and treatment selection

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

Imaging of biomarkers in tissue from patients retrieved through surgery or biopsy can provide important information about the nature of disease processes and can form the basis for improved drug development and individualized treatment. With the emergence of target-specific drugs, it is increasingly important to individualize therapeutic decisions to identify the subpopulation of patients most likely to benefit from treatment and to avoid unnecessary costs and unwanted side effects associated with the therapy. A long list of target-specific drugs, including 10 monoclonal antibodies and 11 small molecule kinase inhibitors, has received approval for cancer treatments by the US Food and Drug Administration (FDA). One example of such a drug is cetuximab (Erbitux), which targets the epidermal growth factor receptor (EGFR) and is approved for the treatment of metastatic or advanced colorectal and head and neck cancers. However, targeted antibody therapies are not administered as a single treatment option but are usually combined with cytotoxic drugs or radiation therapy.

By focusing on key proteins that belong to signalling pathways important for tumour development, taking into account knowledge about frequently mutated genes along those pathways, it is possible to acquire valuable diagnostic information about activity states and susceptibility to targeted therapies. An example of where the identification of a subpopulation of patients that would benefit from target-specific treatment is of value is in colorectal cancer in which the EGFR is only upregulated in a subset of patients. To further complicate matters, only a subset of these patients will respond to EGFR-targeted therapies because of the mutations in the EGFR itself or in downstream signalling mediators such as KRAS or BRAF (so-called primary resistance) [13]. Detection of molecular pathways and protein interactions that become altered in response to treatment or owing to the progression of the disease (acquired resistance) is of particular importance as it can reveal resistance markers and identify targets for complementary treatment schedules that curtail the effect of the emergence of compensatory mutations.

To determine whether a patient undergoing treatment displays primary or acquired resistance to a target-specific drug, tissue from the tumour must be sampled and compared to normal tissue from the same patient. In a patient in whom the diagnosis has been confirmed through biopsy and who receives neo-adjuvant treatment prior to surgery, there is the possibility to compare the biopsy sample with the tissue obtained at surgery to detect acquired mutations. Another option is to re-biopsy after primary surgery to detect signs of resistance to adjuvant treatment. Fine-needle aspiration cytology is an alternative to large core-needle biopsy and offers the possibility to access tumours that are not readily available with standard biopsy equipment, allowing samples from almost any organ to be obtained. Although fine-needle aspiration cytology fails to preserve the histology of the sampled tissue, it produces minimal trauma compared with other methods and the risk of haemorrhage and infection is very low. Whilst the method has been extensively used for cytological morphology since the 1920s, it is now slated for an increased role as a means to access molecular patterns in diseased tissues, using rapidly developing molecular diagnostic procedures.

The existence of cancer stem cells (CSCs) has received a great deal of attention and has been the focus of controversy in recent years. For some tumours, there is compelling evidence of the presence of tumour cells with stem cell characteristics [14]. CSCs are subpopulations within tumours that have self-renewing capacity and pluripotency and are generally highly resistant to both radiation and chemotherapeutics [15]. Recent studies have revealed increased expression of certain proteins in CSCs that are essential in resistance to treatment. The identification of protein–protein interaction networks that include these factors could identify novel pharmacological treatment targets. An example of such a key protein is Notch1 which is upregulated in colon cancer cells in response to chemotherapeutic treatment [16]. Inhibition of Notch1 by small interfering RNA (siRNA) improved chemosensitivity, whereas overexpression of the Notch-1 intracellular domain increased chemoresistance [16]. The importance of Notch1 for self-renewal and function of colon cancer-initiating cells have recently been described [17].

The emergence of resistance in therapy with the new generation of target-specific drugs for cancer treatment, along with the large degree of crosstalk between different signalling pathways and the robustness of many signalling networks because of their modularity, can help explain the often less than encouraging results with single target-specific drugs. Activation of parallel pathways has been shown to lead to resistance as in the case of amplification of MET resulting in resistance to EGFR-tyrosine kinase inhibitors (TKIs) [18]. To overcome the ability of tumour cells to compensate for cytostatic therapy via parallel redundant pathways or compensatory mutations, it is desirable to target several pathways simultaneously. Although such approaches meet with regulatory difficulties, there are several benefits of combination treatment. The effect of single acquired mutations at a target site that renders one drug nonfunctional would then be insufficient to mediate resistance in that tumour cell. Moreover, the dose of each individual compound could perhaps be reduced, to minimize unwanted side effects. Drug combination approaches include different therapeutic antibodies, different TKIs, a combination of antibody and TKI or an antibody/TKI plus irradiation or a cytotoxic drug. The multitude of ongoing clinical trials is exemplified by 14 different combination therapies for EGFR and vascular endothelial growth factor receptor inhibition in nonsmall cell lung cancer currently in phase II and III trials (reviewed by Pennell [19]). To better develop and evaluate combination treatment schedules, there is a need for methods that allow several protein–protein interactions or post-translational modifications to be monitored in parallel with single-cell resolution in biopsy material, as discussed below.

Current diagnostic tools

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

Measurement of protein expression in solution phase

Biomarkers in body fluids, such as blood or urine, are routinely measured in clinical chemistry laboratories. The first diagnostic test to measure the presence of protein as an indicator of human disease (an assay to measure the presence of albumin in urine as a marker for kidney disease) was developed as early as 1827 by Bright. Today more than 200 protein targets are measured in serum or plasma with an average of 1.5 new markers introduced per year during the last 15 years [20]. There are several different formats for diagnostic measurement but the gold standard of single protein detection is immunoassays.

The radioimmunoassay (RIA) was the first type of immunoassay [21]. RIA is a very sensitive method that is used for specific diagnostic tests of, for example, hormones and autoantibodies. However, the high costs and the requirement for specialized equipment to handle the radioligands prevent widespread use of this technique. Recent advances in mass spectrometry (MS), as discussed below, show great promise for multiplex biomarker detection, but this technique is currently used to determine only a few markers such as haemoglobin variants in haemoglobinopathies [22]. Protein detection assays in blood and other body fluids are currently dominated by sandwich ELISAs that provide the specificity afforded by dual recognition of target proteins by pairs of antibodies [23], and the convenience of read-out via chromophores produced in enzymatic reactions. In western blotting, another popular detection method, proteins are identified both via their electrophoretic mobility and via recognition by specific antibodies [24].

The immuno-polymerase chain reaction (PCR) technique was developed in the early 1990s [25] and uses antibodies coupled to DNA molecules instead of enzyme labelling. With the development of the quantitative PCR (qPCR) technology, amplification and detection of DNA could be performed within the same reaction. The exponential signal increase upon antigen detection achieved with PCR has resulted in an impressive decrease in the limit of detection compared with traditional ELISA. A disadvantage is that nonspecific binding of an antibody–oligonucleotide complex to surfaces also initiates amplification and hence reduces the signal-to-noise ratio.

Flow cytometry is a powerful method for rapid analysis of multiple cell characteristics of a large number of individual cells (thereby allowing good statistical accuracy), which is important especially in the case of heterogeneous samples. Flow cytometry is well established in medical diagnostics and different areas of clinical research (haematology, transplantation, stem cell research, tumour immunology, chemotherapy and genetics). Protein microarray technology is another powerful tool for studying multiple protein expression and protein–protein interactions with low sample consumption. Some of the problems for this technology have been to achieve high specificity and sensitive detection of low-abundant proteins.

Imaging of proteins in cells and tissues

Immunohistochemistry (IHC) has long been the gold standard method to study protein expression in tissue. In 1998, the first in situ prognostic marker, measured semi-quantitatively by IHC for human epidermal growth factor receptor 2 (HER2) in breast cancer, was approved by the FDA (the HercepTest, Genentech, Inc/DAKO Corp). The test was approved on the day the monoclonal antibody drug trastuzumab (Herceptin) was approved for treating metastatic breast cancer. This joint development of drug and companion diagnostic kit is becoming increasingly more common and is perhaps a necessity for the expensive molecularly targeted biological drugs to become established.

Tissue microarray technology, which enables parallel analyses of large numbers of small tissue samples on the same slide [26], in combination with IHC techniques, has now become routine and the two methodologies are indispensable for diagnostics. Although IHC provides valuable diagnostic and prognostic information for patient management, the method is only semi-quantitative, and evaluation is time-consuming and requires skill. Considerable effort is being devoted to developing automated image analysis tools for improved quantification in IHC and to standardize analysis between different laboratories [27–29]. Quantitative image analysis tools have been an important part of the effort to standardize HER2 testing in women with breast cancer [30], and better concordance has been reported between fluorescence in situ hybridization and IHC when using computer-assisted image analysis. Although of great value for imaging protein expression, IHC depends on binding by single antibodies, unlike the sandwich assays commonly used for detecting proteins in solution, and there is therefore an inherent risk of nonspecific binding. New imaging methods are emerging that provide increased specificity by employing dual (or more) molecular recognition events. One example is the proximity ligation technology (discussed in detail below).

Imaging using MS is a method that may enable analysis with preserved spatial distribution of target proteins in tissues without the need for staining with specific affinity reagents [31]. Matrix-assisted laser desorption/ionization imaging MS for tissue sections was first reported in 1997 [32] and since then has been used to examine the variable distribution of protein biomarkers across tissues, and to discover and identify biomarkers. A major focus in MS-assisted protein imaging has been to improve the molecular classification of cancer grade, examine tumour margins and help to predict clinical outcome [33]. Potentially, many hundreds of proteins could be imaged from a single scan. However, the technology has several limitations that need to be overcome, such as the currently modest resolution (10–100 μm) and the difficulty in identification and validation of biomarkers (reviewed by Seeley and Caprioli [33]).

Benefits of multiplex analysis in retrospective biobanks and in clinical settings

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

Analysis of large sets of patient material stored in biobanks will be needed to provide reference correlations between treatment history and profiles of protein expression and protein interaction networks on the one hand and disease outcome on the other. Because of the limited amount of individual biobank samples of plasma or tumour tissue, there is a need for sensitive, parallel analysis of large sets of proteins with low sample consumption. Multiplex measurements also reduce labour and consumable costs and provide valuable internal controls for improved precision.

Highly multiplexed proteomic techniques are dominated by MS [34]. Although MS is an excellent technique for identification of proteins, it nonetheless has its limitations. First, the narrow dynamic range makes it impossible to comprehensively analyse proteins in, for example, plasma without any fractionation steps. Second, there is a risk of incorrect identification of peptides derived from the target protein and a requirement for extensive bioinformatic analysis of the results. A third limitation of MS lies in the quantification of proteins. Recent technological developments by Anderson and Aebersold and their respective co-workers using multiple reaction monitoring have enabled quantitative MS with improved sensitivity and range [34, 35].

There are other multiplex formats in addition to MS, such as planar capture arrays and bead-based assays. Antibody arrays have the potential to detect a large number of proteins simultaneously. Recently, 810 cancer-related antibodies were used for dual-colour protein detection in serum, plasma and urine samples with a reported sensitivity in the picomolar to femtomolar range [36]. Assays dependent on binding by single-affinity reagents have limitations in specificity compared to dual recognition techniques, but it has proven difficult to perform large sets of sandwich assays in parallel because of increased risks of cross-reactivity between the many pairs of antibodies that are used. As biomarkers in the blood are present from very low to very high concentration [20], it is important that assays have a large dynamic range.

In situ multiplex analysis of tissues is still limited because of the paucity of different chromogenic dyes or fluorophores that can be used with good resolution. This limitation can be partly overcome by using sequential tissue sections or by removing and re-applying signal detection reagents.

Detection beyond protein expression

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

Post-translational modifications

In cancer, cellular processes such as proliferation, apoptosis, migration and differentiation are common responses to activation and/or blocking of upstream signalling pathways. Understanding the activation of pathways could therefore provide valuable diagnostic information about the state of the tissues. Post-translational modifications are important regulators of many pathways. For example, protein phosphorylation is important in many physiological processes, and it is often deregulated under pathological conditions. Tyrosine kinases are involved in important signalling pathways (e.g. PI3K and MAPK). Phosphorylation of proteins in these pathways initiates cascades of protein phosphorylation leading to many different cellular activities, and pharmacological inhibition is an important form of targeted therapy. A number of TKIs have been approved for cancer treatment. To avoid unnecessary and expensive treatment with possible unwanted side effects, it is important to develop in vitro diagnostic tests that can determine the phosphorylation status in individual patients. Other post-translational modifications regulating the course of cancers include ubiquitination and sumoylation, which labels proteins for proteasomal degradation, and acetylation and glycosylation which also influence protein function. Phosphorylation state-specific antibodies are widely used for in situ studies of protein phosphorylation.

A central question has been whether analyses of protein phosphorylation can offer nonredundant information compared to measurement of protein expression. Many studies, focusing primarily on the epidermal growth factor receptor (ErbB) family, have investigated the correlation between phosphorylation status and protein expression, and their relative clinical value as prognostic or predictive markers or markers of therapeutic effects (reviewed by Mandell [37]). No significant correlation was found in a majority of the studies and the question remains as to what extent analyses of protein phosphorylation can provide additional information. Another example of the potential diagnostic value of phosphorylations is in breast cancer where the oestrogen receptor (ER) status is an important biomarker for prognosis and response to treatment. However, only approximately 50% of patients with tumours that are ER+ respond to endocrine therapy with the drug tamoxifen [38]. Murphy and co-workers have shown in a cohort of more than 300 patients that the level of phosphorylation at position 118 of the ERα (site P-S118) was associated with a significantly better clinical outcome for patients on tamoxifen treatment [39]. Multiple phosphorylated forms of ERα could be found in different breast cancer tissue samples, and it is possible that profiling of these individual phosphorylation events carries prognostic importance to select subgroups of patients that would benefit from endocrine therapy with better precision. An interesting example of phosphorylation resulting in a dramatic effect on protein function concerns a single phosphorylation by the serine-threonine kinase AKT on the ephrin (EPH) receptor A2 (EPHA2), which converts the protein from a tumour suppressor to a tumour promoter [40]. Eph receptors and their interacting ligands (ephrins) are cell surface molecules involved in bidirectional cell communication. The members of this receptor family can contribute to either tumour growth or inhibition, depending on the context.

One limitation for current assays has been the difficulty of quantifying the staining intensity and subcellular localization of phosphorylation patterns. Another obstacle is the validation of phospho-specific antibodies, as such antibodies frequently fail to correctly identify the intended protein. Finally, yet another consideration is the relative stability of protein phosphorylation in patient samples. Standardized procedures will be required for handling patient material from surgical excision to tissue fixation if phosphorylation status is to be used to provide a basis for prognostication of response to treatment. Some laboratories have established very good collaboration with the clinic by assigning personnel to collect and fix tissue during surgery. Standardized collection of patient material is important both in clinical routine and for retrospective studies of biobank material.

Protein–protein interactions

The human interactome has been estimated to consist of approximately 130 000 interactions between pairs of proteins, of which the majority remain undiscovered [41]. So how would one choose which protein interactions to study? A first approach to select candidate proteins is often to search datasets of mRNA expression to identify genes with altered expression in the disease of interest. Much information on published protein interactions is compiled in databases such as HiMAP [42], Corum [43] and the Protein Reference Database [44]. Some of the interactions described in these databases are derived from yeast two-hybrid data or co-immunoprecipitation experiments of proteins after transfection and overexpression, and should be interpreted with some caution. Not all protein pairs that interact when overexpressed in for example yeast cells are present in the same cellular compartment under physiological conditions. There is a continuous flow of reports of new interactions, but more than 85% of published human protein–protein interactions have only been reported in a single publication [45]. To evaluate the quality of the data and determine whether it is valid in the cell type studied, it is important to verify whether both proteins are expressed in the tissue of interest and whether they co-localize to the same subcellular compartment. There are several theoretical clues that can provide further evidence for a true interaction. Evolutionary comparison of the two interacting surfaces can give some insight. If the proteins have co-evolved, and especially if the interacting surfaces have co-evolved, this strengthens the possibility of a true interaction of functional significance [46].

There are several ways in which protein–protein interactions in tissue can be studied. The VeraTag assay [47, 48] was developed to take into account not only protein expression but also to make detection of protein–protein interactions possible. This assay uses pairs of antibodies that are brought in proximity upon binding to the target protein or protein complex. Thereby a fluorescent tag can be released from one of the antibodies by a cleaving function conjugated to the other antibody. The released tags are separated and detected as fluorescence peaks by capillary electrophoresis. The VeraTag assay has been used for quantification of protein expression of the cancer biomarker HER2 in breast cancer tissue samples [48]. The results from this study correlated well with measurement of protein expression by microscopy, and the VeraTag assay showed the additional advantage of an extended dynamic range [48]. The relevance of measurements of protein–protein interactions of endogenously expressed members of the ErbB family in breast cancer cell lines or formalin-fixed paraffin-embedded metastatic breast cancer tissue samples after trastuzumab treatment was investigated using the VeraTag assay [47, 49]. A strong correlation was found between expression levels of HER2 and HER2:HER2 dimers [47], and this study was followed by a more in-depth analysis of EGFR homodimerization and HER2:EGFR heterodimerization [49]. It still remains to be seen whether measurements of interactions between HER2 and other receptors in the family may have additional clinical relevance.

A further example of where enhanced detection reactions are needed to predict responses to therapy involves KRAS mutations, which are common in human tumours such as colorectal cancer. Current treatment strategies for colorectal cancer rely on surgical removal of the primary tumour in combination with pre- and/or postsurgery radiation and pharmacological treatment. In addition to chemotherapeutics, a new class of target-specific drugs are now in clinical use, including antibodies directed against EGFR, which is commonly expressed by tumour cells. However, the emergence of resistance seriously hinders development of target-specific drugs. Mutations that modify the KRAS protein, or further downstream BRAF, render monotherapy with the anti-EGFR antibody cetuximab relatively ineffective [50]. Currently, the KRAS gene is routinely used to investigate mutations known to occur in tumours from patients with colorectal cancer. Potentially, a protein level assay could reveal the altered activity state of KRAS or BRAF proteins, and the distribution of the mutant forms in tumours.

As most proteins are likely to have several interaction partners at any given moment, it would be useful to simultaneously detect interactions between a ‘hub’ protein and several of its binding partners in biopsy material, to better reflect a shift in protein interactions. Such analyses could serve to pinpoint disease processes and help to assess responses to treatment to ensure selection of optimal drug combinations and dosages. To use this information, maps of protein–protein interactions would need to be established for different tumour types and improved diagnostic methods would be required for clinical applications.

Advanced protein analyses via PLAs

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

As discussed above, there is a need for protein detection methods with improved performance. Assays with greater sensitivity of detection compared to current techniques, and with enhanced ability to reveal modifications and interactions, preferably in multiplex, could lead to analyses of an increased range of biomarkers that reflect disease and predict drug response. PLA technology allows configuration of assays for highly specific recognition of proteins and protein complexes, and the results of the detection reactions are encoded as amplifiable DNA strands for sensitive multiplex detection.

The assays build on the principle that recognition of target proteins by two, three or more affinity probes (typically antibodies) brings into proximity DNA strands that are attached to the affinity reagents [51]. These DNA strands can then participate in ligation reactions, giving rise to molecules that are amplified by methods such as PCR [51] or rolling circle amplification (RCA) [52] for solution-phase and localized detection reactions, respectively (Fig. 2). Since its introduction in 2002, PLA technology has been employed for detection of proteins in complex biological material and in tissue samples. The first PLA was performed as a homogenous assay, where all components are present in solution during measurement and no washes are required. The cytokine platelet-derived growth factor (PDGF)-BB was detected by PLA in human serum with about 1000-fold greater sensitivity compared to a commercially available ELISA for the same protein [51]. This study was followed by others demonstrating the suitability of PLA for sensitive detection of a variety of proteins [53, 54] and infectious agents [55]. The method can also be performed with a first capture step on a solid support before pairs of oligonucleotide-modified antibodies are added. This form of the assay allows analyses of larger sample volumes, even ones containing material that might interfere with the reactions, as reactions can be washed before proceeding to the ligation and amplification steps [56, 57].

image

Figure 2. Solution-phase and in situ proximity ligation. (a) Detection of proteins and complexes in blood and other solution-phase samples with read-out via quantitative PCR. Antibodies with attached oligonucleotides having either a free 5′ or 3′ end can bind to the protein complex. Upon proximal binding, the pairs of oligonucleotides can hybridize to a connector oligonucleotide, guiding their joining by ligation. The ligation products of the proximity probes are amplified and detected by PCR with real-time detection, providing a measure of the amount of detected target proteins. (b) Detection of endogenous protein complexes in cells and tissues with localized detection via rolling circle amplification (RCA). When two antibodies bind the same protein molecule or a pair of interacting proteins, then oligonucleotides attached to these antibodies can guide the joining of two subsequently added linear oligonucleotides to form a covalently joined circular structure by enzymatic DNA ligation. The circular DNA strand is then copied in an RCA process initiated using one of the antibody-bound oligonucleotides serving as a primer. The RCA product, including hundreds of complements of the DNA circle, bundles up in a submicron spot, easily detected after hybridization of fluorescence-labelled oligonucleotides that are complementary to a tag sequence in the RCA product. The RCA products can be analysed by microscopy or using flow cytometry.

Download figure to PowerPoint

Measuring proteins in solution phase

The proximity ligation mechanism presents a number of advantages over conventional methods for protein analysis in samples such as blood and other body fluids. The unique opportunity to design assays that depend on target recognition by three [56, 57] or more antibodies means that highly specific detection can be achieved in a manner analogous to that of nested PCR with three or four primers used for highly specific DNA amplification. This aspect is critical to investigate proteins that are present at trace levels in, for example, the blood or cerebrospinal fluid and could enable analyses of a new generation of protein biomarkers. The multiple recognition reactions also allow detection of complexes of interacting proteins or of post-translational modification of proteins (e.g. by phosphorylation or glycosylation) to better characterize detected proteins, again extending the range of possible protein targets for bioassays.

So-called prostasomes represent a class of potential biomarker that can become accessible to analysis by taking advantage of the possibility of investigating several epitopes using PLA. Prostasomes are membrane-coated protein complexes that are normally released by the prostate epithelium into seminal fluid [58], but that may be present at increased levels in blood in cases of infiltrative growth of malignified epithelial cell. By designing a variant of the PLA technique that requires simultaneous recognition of proteins present in prostasomes by five antibodies, Kamali-Moghaddam and co-workers were, for the first time, able to detect these complexes in blood. They observed significantly increased levels in prostate cancer patients with markedly atypical cells, compared to healthy controls (Kamali-Moghaddam et al., submitted).

The ability of PLA to represent detected proteins as strings of nucleotides also offers further advantages, in addition to the opportunity to amplify signals. The nucleotide tags can also be used to encode DNA motifs that identify the detected proteins in multiplex detection reactions with parallel read-out using powerful methods for DNA analysis. This means that individual samples can be investigated with respect to large sets of proteins whose expression levels can be directly compared, reducing consumption of precious sample material and providing increased precision. The DNA ligation step also enables detection reactions to be restricted to the appropriate pairs or trios of detection reagents for particular targeted proteins. This is in contrast to conventional sandwich assays in which the risk of cross-reactions rises rapidly with increased numbers of targeted proteins, severely limiting opportunities for multiplexed assays. In a recent report, a 7-plex homogenous phase PLA was used to analyse plasma biomarkers with minimal cross-reactivity [59]. Multiplex PLA has also been developed for solid phase assay using a dual-tag microarray format, and proof of principle was shown by detection of vascular endothelial growth factor (VEGF) [60], where capture of VEGF in solution was followed by PCR amplification, restriction and capture on a dual-tag microarray. The detection was then carried out through amplification by RCA and fluorescence read-out.

Imaging proteins and their interaction partners in cells and tissues

For purposes of microscopic imaging, the requirement for dual recognition by pairs of antibodies in PLA can both enhance specificity and also allow detection of interacting protein pairs or secondary modifications in a manner that has not been possible previously. Accordingly, in situ PLA is suitable to reveal an extended range of molecular features in cells and tissues compared to standard IHC. The pairwise binding by two detection reagents with attached oligonucleotides during in situ PLA guides the formation of a circular DNA strand, which can then be replicated into a bundle of DNA just at the resolution of light microscopy (Fig. 2b). Each of these DNA bundles is capable of binding hundreds of fluorescently labelled oligonucleotides complementary to the replicated DNA strand, producing brightly fluorescent spots. This localized amplified detection reaction presents a number of advantages. First, it allows the detection of individual target molecules or molecular complexes, and software has been developed that enables automated enumeration of detection signals for objective, digital measurement of staining patterns by counting spots rather than estimating staining intensity [61]. The strong fluorescence also means that background fluorescence from tissues becomes a lesser although still significant problem. Second, it reduces the impact of nonspecific binding by detection reagents, allowing sensitive detection. Fluorescence read-out permits multiplex analysis using different excitation wavelengths. The possibility of studying several proteins and protein–protein interactions in parallel by performing multiplex analysis is very appealing for reducing workload, reagent consumption and costs. Finally, it also means that more information can be gathered from limited biological samples (e.g. from biobanks) and allows cross-referencing between different cellular processes at single-cell resolution. In addition to microscopic analysis, in situ PLA studies can also be performed via flow cytometric analysis to generate better statistical data, as illustrated by the analysis of ErbB interactions [62]. As an alternative to fluorescence detection, enzymatic PLA read-out (e.g. via peroxidase depositing coloured products at the site of binding) avoids considerations of autofluorescence and allows conventional counterstaining for improved histological analysis in routine pathology. In situ PLA with enzymatic read-out produces images similar to those of IHC, but quantitative information is recovered for stained tissue sections using computer-based image analysis tools to quantify the RCA products [63].

In situ PLA has been used to study stimulation-dependent phosphorylation of the cytokine receptor, PDGF receptor β (PDGFRβ) [64] (Fig. 3a). This is a transmembrane protein tyrosine kinase receptor that dimerizes upon ligand binding and becomes autophosphorylated at several sites in the C-terminal intracellular domain. Signalling through the receptor leads to activation of several pathways that promote cell proliferation, motility and survival. The mechanism has been implicated in human tumours. Studies of these cellular processes have demonstrated the importance of spatial and temporal control of PDGFR phosphorylation [65, 66]. PLA-based quantification and identification of the subcellular localization of phosphorylated PDGFRβ using image analysis software has advantages over normal immunofluorescence by offering increased selectivity for distinguishing the phosphorylated forms of the receptor [64]. In situ PLA has also been used to study the perturbation between different protein complexes involved in various cancers exemplified by (i) inhibition of the interaction between PDGFRβ and the PI3-kinase, an interaction partner in the downstream cell signalling, upon treatment of primary cells with the TKI imatinib followed by ligand stimulation [67] and (ii) blocking the interaction of the transcription factor and oncoprotein c-Myc and its obligatory partner protein Max in the presence of a low molecular weight compound [52] (Fig. 3b). In situ PLA has recently been used to study mucin glycoproteins (Fig. 3c), as altered expression of carbohydrate and peptide moieties of mucin glycoproteins has been considered a possible molecular marker in premalignant and overtly malignant lesions. The coincidence of the protein and its carbohydrate modification could be demonstrated through the requirement in PLA for multiple binding events. This carbohydrate-modified protein is rarely observed in normal tissue but is frequently present in several types of cancer [68]. This is a first step towards more extensive assays for alterations of glycoproteins that may serve as diagnostic or prognostic tools.

image

Figure 3. Detection of protein–protein interactions and phosphorylation in cell lines and patient tissue samples. (a) Detection of phosphorylation of the platelet-derived growth factor receptor β (PDGFRβ). Human fibroblast BJ hTert cells were untreated (i) or stimulated with 100 ng mL−1 PDGF-BB (ii). The phosphorylation status of PDGFRβ was detected by in situ proximity ligation assay (PLA) using rabbit antibodies directed against PDGFRβ and a pan-phospho-tyrosine-specific mouse monoclonal antibody, followed by a pair of anti-rabbit and anti-mouse immunoglobulin-specific antibodies with attached oligonucleotides. (b) Perturbation of heterodimerization between the c-Myc and Max proteins. Normal human fibroblasts were treated for 6 h with vehicle (control) (iii) or with the low molecular weight compound 10058 (iv). This was followed by the addition of antibodies directed against the c-Myc and Max proteins, with attached oligonucleotides for analysis by in situ PLA. After treatment with the inhibitor, far fewer detection signals were observed, demonstrating that the c-Myc/Max interaction had been significantly inhibited by this known inhibitor of c-Myc–Max interactions compared to the control (P < 0.01). (c) Gastric mucosa with intestinal metaplasia that stains for MUC2 and sialyl-Tn in goblet cells. PLA shows that the MUC2 protein is carrying the sialyl-Tn carbohydrate. Normal gastric mucosa, negative for MUC2 and sialyl-Tn, is also observed.

Download figure to PowerPoint

High-content screening by in situ PLA and microscopy can allow molecularly targeted evaluation of cellular responses to drugs by investigating protein–protein interactions or modifications during the course of drug development, or for selection of the appropriate therapy in individual patients. Leuchowius and co-workers illustrated the possibility of screening protein interactions and modifications in a high-content format with PLA [67]. In their study, a library of kinase inhibitors was investigated for effects on the phosphorylation of PDGFRβ and interactions with PI 3-kinase in human primary fibroblasts, and they were able to identify 13 out of 86 compounds as hits. The known PDGFR inhibitors imatinib, sorafenib and sunitinib, which are in current clinical use, showed strong inhibition of PDGFRβ phosphorylation (Fig. 4).

image

Figure 4. In situ proximity ligation assay-based screening of a library of kinase inhibitors to find compounds that inhibits the phosphorylation of the platelet-derived growth factor receptor β (PDGFRβ) in stimulated human primary fibroblasts. The clinically used kinase inhibitors imatinib, sorafenib and sunitinib showed strong inhibition of PDGFRβ phosphorylation (figure is adapted from Leuchowius et al. [67].

Download figure to PowerPoint

As discussed above, the emergence of resistance towards monotherapy approaches with the new generation of target-specific drugs for cancer treatment has signalled the need to develop combination treatment strategies, to simultaneously target a specific protein or dysregulated pathway with two or more types of drugs. The detection and quantification in situ of altered protein expression and multiple protein–protein interaction signatures in parallel and with single-cell resolution has the potential to identify subpopulations of cells that are prone to particular resistance patterns. Furthermore, this could illuminate dysregulated pathways that may contribute to the capacity of a tumour to cause metastatic spread, and it can help to identify appropriate drugs for further treatment to eradicate these cells. Simultaneous detection of two or more protein–protein interactions during in situ analysis could allow selection of effective derivatives from an original compound with minimal side effects on other pathways. A multiplex assay could also assist the development of therapies with suitable drug combinations to target redundant parallel pathways in the same cell.

Conclusions

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

Progress in molecular understanding of disease processes in combination with powerful new analytical techniques may greatly extend the range of molecular features that can be evaluated as biomarkers, either in body fluids or distributed in tissues. A rapidly growing interest in molecular diagnostics is also motivated by the need to stratify patient groups according to their expected response to expensive new therapies targeted at specific molecular mechanisms. The concept of companion diagnostic tests to accompany new drugs during the different phases of drug development is being adopted by major pharmaceutical companies. These same tests will probably follow the new therapies into clinical practice.

The proximity ligation technology discussed here represents a new analytical mechanism whereby specificity of detection can be adjusted for improved performance, and high sensitivity can be achieved via DNA amplification. We have illustrated how the method can be used to measure new classes of plasma biomarkers and to image the distribution of activated signalling proteins in cell lines and tissues. An important line of development for the PLA technique is to allow large sets of proteins and protein complexes to be assessed in multiplex (e.g. in blood samples), and for parallel imaging of signal transduction cascades in health and disease. This will permit comprehensive sensing of cellular signalling events in research, diagnostics and drug development. A likely further development of advanced protein diagnostic tests as discussed herein is the adaptation for simple and rapid analyses at the point of care. This area of medical diagnostics is expected to grow very rapidly and may have important healthcare benefits for western medicine as well as for those in underdeveloped countries. Improved molecular processes will also bring complex diagnostic tasks within reach of analysis at the doctor’s surgery, for detection of early-stage disease and prompt selection of optimal therapy.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References

Work in our laboratory is supported by the Knut and Alice Wallenberg Foundation, the Swedish Research Council, the European Community’s 6th and 7th Framework Programs and the Swedish research agencies Vinnova and the Strategic Foundation. We thank Irene Weibrecht, Karl-Johan Leuchowius, Tim Conze and Ola Söderberg (Department of Genetics & Pathology, Uppsala University, Sweden) for help with illustrations.

References

  1. Top of page
  2. Abstract
  3. Opportunities and challenges for protein diagnostics
  4. Detection of secreted biomarkers for diagnostics and evaluation of disease progression
  5. Tissue biomarkers for differential diagnosis and treatment selection
  6. Current diagnostic tools
  7. Benefits of multiplex analysis in retrospective biobanks and in clinical settings
  8. Detection beyond protein expression
  9. Advanced protein analyses via PLAs
  10. Conclusions
  11. Conflict of interest statement
  12. Acknowledgements
  13. References