Final Report Summary - FLUODIAMON (Ultra-high resolution and ultra-sensitive fluorescence methods for ... diagnosis of early disease and disease progression in breast and prostate cancer)
Executive summary:
The survival chances of cancer patients can be dramatically increased by an early diagnosis. This also holds true for breast and prostate cancer, which are among the most frequent forms of cancer in Europe. However, with currently available methods, typically core needle biopsies, it is not possible to extract the necessary amount of sample material for a reliable diagnosis without running a high risk of severe side effects, not the least cancer cell seeding and risk of spread of the cancer to be diagnosed. As an alternative, fine-needle aspiration (FNA) cell sampling offers a patient friendly, minimally invasive, rapid and cost efficient tumour diagnostic procedure with minor side-effects. However, the amount of sample obtained by FNA is often sparse and consists of aspirated cells taken out of their tissue context. It is therefore often difficult to obtain a conclusive diagnosis based upon this sample material.
In this project, we have addressed this conflict between the amount of sample material needed to obtain conclusive diagnostic information and the severe side effects that can be caused by the diagnostic sampling itself. The overall objective of this project was to develop and validate a quantitative, minimally invasive diagnostic tool for early and conclusive detection, diagnosis and monitoring of disease progression of breast and prostate cancer. A methodology was developed making use of a combination of exciting recent advances in the field of fluorescence-based microscopy, for imaging of individual FNA-sampled cells. The methodology includes advances taking the spatial resolution far beyond the fundamental limits of optical resolution, the sensitivity down to an ultimate single-molecule level and multi-parameter detection schemes significantly increasing the fluorescence information by which these cellular images can be analysed. Apart from detecting and identifying specific protein tumour markers in the samples, tumour-specific spatial distribution patterns of the proteins within intact sample cells were also exploited. This represents a to-date almost unexploited dimension of diagnostic information. The novel optical methods were supported by state of the art affinity molecule biotechnology, fluorophore chemistry and proteomics-based biomarker identification, with the overall aim to extract a maximum amount of information out of the small amounts of sample material obtained by FNA.
Following extensive effort in optimising the sample preparation, handling and labelling and corresponding effort to optimise the performance of the fluorescence imaging techniques, first on cultured cells as models and then on cells from clinical FNA samples, a larger number of FNA samples were analysed with a combination of ultrahigh resolution fluorescence microscopy and imaging based on multi-parameter fluorescence detection. New bioinformatic tools were developed and used to analyse the acquired images and features and classifiers were established which in an objective manner could classify the imaged samples as normal or malignant. In parallel, progress in FNA sampling technology performed within the project led to further reduced risks for cancer cell seeding and increased the ability to obtain representative diagnostic material. Taken together, we can thus demonstrate an intact chain of accomplished steps, from reliable and patient-friendly FNA sampling with minimised risk of tumour cell dissemination, over robust and reproducible sample handling and preparation, specific targeting of selected proteins with fluorophore-labelled affinity molecules, characterisation of the FNA sampled cells by optical methods offering utmost sensitivity, specificity and spatial resolution and finally bioinformatic analyses to identify the differences in the image features necessary to distinguish normal from malignant cells. This represents a novel diagnostic concept, which can combine diagnostic reliability with safe, minimally invasive diagnostic sampling and thus enable a decisive improvement of the outcome of patients suffering from breast and prostate cancer.
Project context and objectives:
The survival chances of cancer patients can be dramatically increased by an early diagnosis. This particularly holds true for breast and prostate cancer, which are among the most frequent forms of cancer in Europe. With currently available methods, typically core-needle biopsies, it is not possible to extract the necessary amount of sample material for a reliable diagnosis without running a high risk of severe side effects.
On the other side, FNA based cell sampling technology is a patient friendly, minimally invasive, rapid and cost efficient tumour diagnostic procedure with negligible side-effects. However, despite the overall advantages of FNA-based diagnostics for breast and prostate tumours, there are three major drawbacks that prevent a general application of FNA in clinical routine:
1. FNA-based sampling in general only yields a sparse amount of diagnostic material, with aspirated cells in an extracellular fluid
2. FNA samples are frequently non-representative by containing too few cells of the suspicious lesion.
3. Even with a successful lesion biopsy it is often difficult to obtain a conclusive diagnosis based upon the frequently minimal cellular deviation of atypical or even fully transformed malignant cells compared with their normal counterparts.
For these reasons, core needle biopsy or surgical biopsy are to date still favoured over FNA for breast and prostate cancer diagnostics. However, due to their mechanical crudeness, they are highly traumatic and combined with significant risk for tumour cell seeding in the multiple needle tracts.
The overall objective of this project was to develop and validate a quantitative, minimally invasive diagnostic tool for early and conclusive detection, diagnosis and monitoring of disease and disease progression of breast and prostate cancer, with negligible sampling-related side-effects.
To increase the amount of information that can be extracted from the sample this project has taken benefit from the recent remarkable progress in fluorescence-based spectroscopy and imaging. In particular, we brought in exciting new concepts to increase the sensitivity and specificity of fluorescence detection and, not the least, to overcome the diffraction barrier in optical microscopy. Based on these concepts and methodologies, we could put in sight a development towards novel optical diagnostic methods unimaginable just a few years ago and to develop the means necessary to eliminate the conflict between the amount of sample material needed and the necessary diagnostic conclusiveness.
The overall strategy to be achieved within the FLUODIAMON project was thus to exploit the combination of the unique sensitivity, specificity and spatial resolution offered by the very latest developments in fluorescence microscopy and the minimal invasiveness of FNA based cytology. Based on the fluorescence methods the project aimed at developing and applying approaches where minute FNA samples could be subject to objective analyses on a molecular or subcellular level, thereby providing sufficient diagnostic information. The strategy of the project was in particular to analyse three main categories of proteins within the FNA-sampled cells: cytoskeletal proteins, receptor proteins located in the cellular membranes and proteins regulating the division cycles of the cells. Apart from the mere detection and identification of proteins of these categories in the sampled cells, tumour-specific spatial distributions of these proteins within the intact sample cells were also exploited.
To reach the overall goal of the project, efforts not only in the further development of fluorescence microscopy/spectroscopy and FNA sampling technology were needed. Following FNA-sampling, the FNA samples needed to be properly prepared; and diagnostically representative proteins in the cells needed to be identified and marked using specific affinity molecules, labelled with bright and stable fluorophore marker molecules. Extraction of diagnostically significant fluorescence information from the FNA-sampled cells needed to be optimised based on iterative efforts within optical imaging and sample preparation and labelling. Finally the acquired images of the FNA-sampled cells needed to be analysed with new bioinformatic means and finally validated towards clinical data.
Given the effort required spanning over a broad range of disciplines, the main objective of the project was also divided into a set of detailed objectives, reflecting the whole chain of steps needed to reach the final diagnostic procedure aimed for. The detailed objectives of the FLUODIAMON project were:
1. to improve spatial resolution of state of the art light microscopy in pathology by an order of magnitude
2. to improve the sensitivity of fluorescence-based imaging of FNA acquired cells to the ultimate single-molecule level
3. to take multi-parameter fluorescence imaging of individual FNA acquired cells to its extreme in terms of information content, largely based on photon statistical approaches and parameters extracted from non-linear effects
4. to develop standardised FNA-based sampling of suspected breast and prostate cancer lesions with negligible side-effects and with optimised needle visibility for ultra-sound guided needle positioning
5. to select already known molecular markers that will be compatible with the developed fluorescence imaging techniques and can be highly anticipated, when investigated by these techniques, to strongly correlate with malignant transformation and clinical tumour aggressiveness
6. to identify existing and develop new affinity molecules to these markers, which are highly specific and fluorophore labelled for optimised fluorescence readout properties
7. to refine bioinformatic evaluation and data processing to find the combination of fluorescence read-out parameters that most strongly correlate with the relevant clinical parameters and yield the strongest diagnostic reliability
8. to optimise the combination of techniques and procedures, to maximise sensitivity and specificity for FNA-based diagnostics of breast and prostate cancer, enabling a decisive improvement of the outcome for the patients suffering from these diseases.
Project results:
In the forefront of the project efforts was the further development and adaptation of cutting-edge optical methods so that they could extract sufficient diagnostic information out of very small amounts of diagnostic material. Reaching the overall objective of this project required interaction between a broad range of different disciplines, as well as significant progress within each of these disciplines. In particular, substantial effort and progress were needed within the following areas, covering all steps from the initial sampling step to the final evaluation of the acquired images and the validation of the whole diagnostic procedure:
1. sampling of suspect breast and prostate cancer lesions with minimal side effects by FNA
2. preparation and handling of the clinical FNA samples
3. selection of protein targets to be visualised by the optical methods in the cells in the FNA samples
4. selection and development of affinity binders
5. selection and development of fluorophore marker molecules which provide a maximum of sensitivity and specificity
6. development and optimisation of high-end fluorescence methods for diagnostics of sparse amount of cells sampled by FNA. Four different methodologies were considered in the project, namely ultrahigh resolution fluorescence microscopy - nanoscopy, or stimulated emission depletion (STED) microscopy -, multi-parameter detection imaging (MFDi), transient state (TRAST) imaging and two-photon excitation (TPX) spectroscopy.
7. development of bioinformatic tools for the diagnostic evaluation of the protein content and the subcellular features within the imaged FNA sampled cells.
Development of FNA sampling procedures of suspect breast and prostate cancer lesions
FNA based cell sampling technology is a patient friendly, minimally invasive, rapid and cost efficient tumour diagnostic procedure. Conventional methods side-effects are completely negligible with the FNA technique. In this project, the Stockholm-based small company Neodynamics AB has managed to improve the FNA sampling procedures in several aspects.
Firstly, a new FNA needle was developed with improved sample yield and surface-treated for considerably increased ultra-sound visibility. Sample yield (i.e. the amount of sample material obtained after needle aspiration) is important, given that FNA-based sampling in general only yields a very sparse and often also insufficient amount of cell sample material. Ultrasound visibility is important since upon sampling, when the needle is introduced in the tissue, the positioning of the needle prior to the aspiration relies on the guiding and visualisation via ultrasound. If the needle position is not positioned at the right place, the sample is obviously not obtained from the lesion region to be investigated. Final development on the needle design was performed resulting in two prototypes (outer diameter 0.7 mm) with and without enhanced ultra-sound visibility. Large scale clinical testing was carried out in order to evaluate their performance. The new needle achieved outstanding results when compared to needles commonly used today. It yielded three times more material than standard needles of the same diameter and two times more than needles with a diameter of 0.7 mm.
Secondly, needle additives to facilitate needle penetration, manoeuvrability and precision in positioning were developed. In the device developed by Neodynamics AB called 'Cytotest' mechanical energy is added to the sampling-needle by computer controlled oscillating movements. These oscillatory movements of the needle have been demonstrated to strongly facilitate tumour penetration, allow the direction of the needle to be changed to a larger extent upon penetration, as well as to increase the sampling phase efficacy. Also a rotating movement was added to the needle in order to mimic the way the needles are handled by trained cytologists. Moreover, the design of the bevel section of the needle has been adapted for optimal cutting behaviour together with the oscillatory and rotational movement schemes. 'Cytotest' has been used at two breast diagnostic centers in Stockholm (Sabbatbergs Breast Center and St Göran Hospitals Breast Unit) on more than 500 breast cancer patients with excellent results. Due to concerns from the operators regarding decrease of tactility upon sampling when using the oscillatory needles, an improved handle was designed, in collaboration with the design programme at Kalmar University, Sweden and successfully tested by clinical operators. The needle handle will be implemented into the 'Cytotest' instrument before large-scale production is initiated.
Thirdly, Neodynamics AB has succeeded with the development of a fully functional anti-seeding instrument as well as a specialised anti-seeding needle. The anti-seeding instrument consists of a signal processor, an amplifier as well as a patient unit. The anti-seeding needle tip acts as an electrode whereas the shaft is isolated. When radio frequency pulses are given, a current is applied between a dispersive electrode, for instance located over the chest of the patient when sampling from the breast and the needle electrode. Thereby, the tissue around the open needle tip is heating up in a much localised fashion where the current density is highest. Starting with tests on pork livers provided the knowledge to successively improve the used pulses and their properties, an ethical permit was then acquired in order to commence with clinical studies. From the tests on patients undergoing the sampling procedure based on the anti-seeding procedure the results have been positive in every aspect. No or very little additional discomfort is experienced by the patients and the analysed samples have not been denaturised. To estimate the extent of local tumour cell-dissemination upon FNA sampling, partners Neodynamics AB and KI (Wiksell and Auer) investigated the blood droplet(s) that spontaneously penetrates the skin orifice after FNA needle retraction. They found that there is a high frequency of these blood droplets contained tumour cells. From this finding it is reasonable to assume that upon needle sampling, cancer cells can also enter the circulating blood system of the patient and may finally also cause distant metastasis. Interestingly, applying the developed anti-seeding procedure significantly reduced the overall amount of blood and exudate passing the skin orifice after retraction of the FNA needle. Moreover, a test series on patients indicated that there were no viable suspended tumour cells in the blood droplets leaking out of the skin orifice after needle retraction.
Taken together, the development on FNA sampling techniques is likely to lead to a dramatic improvement regarding sampling related risks, sample reproducibility, quality and volume of collected cells as well as regarding safety against cell tumour dissemination. These developments also give the necessary prerequisites, covering the first steps in the diagnostic procedure developed in the project.
Preparation and handling of the clinical FNA samples
Once a patient sample is safely and reliably obtained by the FNA sampling procedures, the subsequent preparation and handling of the cellular material need to be optimised and properly adapted to the overall diagnostic procedure. In the project, the optimisation of the sample handling and preparation had to incorporate many aspects:
1. the sample preparation should not compromise the specificity of the affinity molecules against their protein targets in the cells
2. the recorded fluorescence intensity from the fluorophore marker molecules (labelled onto the affinity molecules) versus the background level in the imaging had to be maintained, as well as the stability, reliability and reproducibility of the recorded fluorescence parameters
3. the cellular morphology and specific cellular features should not be severely affected by the sample handling and preparation
4. the preparation should yield a proper adhesion of cells to glass cover slides and preparation formats compatible with transport
5. the preparation protocols should be sufficiently simple, so that they could be performed at clinical sites with typically limited resources and infrastructure for cell handling.
Taken all these aspects together favoured a procedure, where immediately after sampling the patient material was subject to a three-step procedure consisting of a blood cell lysis step, a fixation and a cytospin step, thereby preparing the sample for subsequent sending, labelling and imaging.
Selection of protein targets
In general, the cure of cancer diseases strongly depends upon early and detailed diagnosis of type and stage of the malignant disease, typically requiring a molecular characterisation of the malignancy. Consequently, the aim of the project was to further develop and apply analyses of the minute FNA samples on a molecular or subcellular level. Partners Karolinska Institutet (KI) and Lübeck University (LUEBECK) have previously identified a substantial number of proteins in freshly taken clinical samples from tumours of the breast, prostate and others. By comparing cancer type specific protein expression with that of corresponding normal tissue, single proteins or groups of proteins were identified, highly correlated to malignant transformation and clinical tumour aggressiveness.
In this project, we wanted to extend the protein characterisation beyond the plain presence/absence, or rather deviations in the amount of these proteins in the FNA sampled cells or in the extracellular sample material. Given the sensitivity and spatial resolution offered by our fluorescence methods, we also wanted to map the spatial distributions of selected proteins within the FNA sampled cells as objective parameters specific for malignant transformations. The project focussed on diagnostically monitoring three different categories of proteins: cytoskeletal proteins, membrane proteins and cell cycle regulating proteins.
The specific proteins selected to demonstrate the diagnostic procedure and the diagnostic feasibility of the microscopic techniques were found from extensive literature searches and among our already identified tumour markers. The final condensed list of protein biomarkers for breast and prostate cancer were then selected according to the following parameters:
1. presence or absence of protein markers in disease as compared to controls
2. potential of changes in spatio-temporal behaviour and interaction patterns in diseased stages
3. changes in the sub-cellular location of the protein upon disease.
In the project, the diagnostic procedure was eventually narrowed down on analysing the abundance, interactions and/or the spatial distribution patterns of the following proteins in the FNA sampled cells: vimentin and tubulin (cytoskeletal proteins), IGF1R, HER1, HER2 (membrane proteins), Cyclin A, Cyclin E (cell cycle regulating proteins).
The monitoring of the presence and spatial distribution patterns in individual cells of these specifically selected proteins proved to be a successful strategy.
Selection and development of affinity binders to the protein targets
The implementation of the novel and advanced fluorescence-based imaging techniques for the diagnostic procedure, as aimed for in the project, requires that the specific proteins that were selected to be analysed also can be properly detected. In this project the strategy was to use affinity binders that bind specifically to the proteins.
Three different classes of affinity probes were used. Apart from conventional antibody molecules, available via commercial suppliers, affinity binders of two different classes were developed. At the biotechnology department of the Royal Institute of Technology, Stockholm (KTH), new affinity binders were developed based on an existing protein library technology platform denoted 'affibody binding proteins'. These affinity reagents, Affibodies, are characterised by a small sized scaffold, about 2 nm in diameter and consisting of 58 amino acids. Thirteen of these amino acids, making up a binding surface area of approximately 800 Å2, can be combinatorially randomised. Since there are 20 different amino acids in the proteins of humans, there are in principle different combinations for the set of 13 amino acids at the binding surface and thus considerable potential to optimise the affinity and selectivity of the Affibodies towards specific target proteins. At Academisch Ziekenhuis Leiden (LUMC), heavy chain antibodies (HCAbs), were developed for the project. HCAbs represent a class of antibodies specific for camels. The HCAb fragments responsible for the target binding are significantly smaller than conventional antibodies, represent the smallest naturally occurring protein-based affinity binders known to date and are frequently used as binders within biomedical research, diagnosis and therapy.
To provide the necessary affinity binders to the project, an inventory of available affinity reagents to the selected protein targets was performed and finalised and an affinity protein reagent bank was established for these targets. It should be borne in mind that even if there are one or several affinity binders reported to be specifically binding to a certain protein with high affinity, both the binders and the targets are proteins and their conformations and their binding properties may change dramatically depending on the environmental conditions. Following identification of protein targets, available affinity reagents to these proteins therefore had to be further investigated with respect to their binding properties under conditions close to those expected in the FNA sampled cells. In these investigations, the reagents were either directly or indirectly labelled. The investigations of the binders were performed by conventional fluorescence-based confocal microscopy on cells from three selected cell lines representing normal cells and two different aggressive forms of breast cancer. As a result of these investigations, a list of affinity binders to the prioritised list of protein targets could be identified.
In parallel to these efforts, work was performed both at LUMC and KTH towards development of novel affibody and HCAb fragment binders to an interesting breast cancer marker candidate, the protein SATB1. This work resulted in the identification and validation of several new affibody and HCAb fragment variants. These were produced recombinantly, subsequently labelled by various principles and found to selectively recognize the SATB1 target in cell samples. Data from binding studies suggest that different epitopes are preferred by the two classes, opening up for a possible increased specificity in obtained localisation signals. The new binders were used either separately or in combination to evaluate the presence and subcellular localisation of the SATB1 target in different cell types. Although SATB1 was eventually not selected among the proteins showing the largest diagnostic potential in the diagnostic procedures of the project, the possible alternative use of the developed binders for detection of targets present in cell-lysates was demonstrated and shows the versatility and value of having such selective binders to targets of interest.
Taken together, a set of affinity binders could be identified for the selected target proteins, well functional also under the conditions relevant for FNA sampled cells. The project has further shown that novel technologies for development of novel affinity proteins of different nature have the potential to be very valuable for future diagnostic applications.
Selection and development of fluorophores
Both high-resolution microscopy (nanoscopy) and high-sensitivity, multi-parameter, fluorescence-based characterisation of proteins in cells critically depends on the availability of suitable fluorophores, carrying specific chemical groups so that they can be coupled to relevant affinity proteins or to other molecules of interest. Besides spectral properties suited to the optical instrumentation, the most important properties of the label are high fluorescence quantum yield and high photostability. The University of Siegen (UNISI) has a vast experience in the area of fluorophore development and was the main partner responsible for the fluorophore selection and optimisation.
Commercially available fluorescent labels were evaluated with respect to their fluorescence properties and tested more in detail by the various spectroscopic and imaging techniques in the project. Apart from fluorescence quantum yield and photostability, a critical aspect investigated was the stability and reliability of the fluorescence parameters recorded from the fluorophores. Based on these investigations, a set of fluorophore labels could be identified as appropriate for the project and were subsequently used in the developed imaging procedures.
In addition to testing and selection of commercially available fluorophores for the project, new chromophores were also synthesised. In particular, new hydrophilic and water-soluble fluorophores were designed and synthesised, specifically suitable for the labelling of affinity molecules such as antibodies and affibodies. The newly synthesised fluorophores displayed minimised dye aggregation and very low unspecific binding to cell-based specimens, yet exceptional properties in terms of photostability and brightness, as required for the ultrasensitive and high-resolution microscopy techniques in the project.
Taken together, highly photostable and bright fluorophores are indispensable for successful fluorescence spectroscopy/imaging in general and in particular for the cutting-edge fluorescence-based methodologies developed in this project.
Development and optimisation of high-end fluorescence methods for diagnostics of FNA-sampled cells
In the project, several fluorescence-based methods were developed and finally evaluated for use in a procedure to diagnostically classify minute amounts of FNA-sampled material. The strategy was to provide a combined procedure, with the complementary information obtainable from super-resolution imaging on the one side and new, extremely sensitive and specific fluorescence-based optical imaging techniques on the other, forming a sound basis for an extensive analysis of FNA aspirates and a final highly reliable diagnostic classification.
Along this strategy, four different fluorescence-based methods were developed and modified towards diagnostic use in this context.
1. STED microscopy: In this project, we have set as a goal to introduce ultrahigh-resolution microscopy for diagnostic purposes and the further development and adaptation of STED microscopy, or nanoscopy, for diagnostic evaluation of FNA-sampled cells has been a major activity. A STED microscope design based on a novel supercontinuum light source was implemented at MPIBPC during the first year of the project and at KTH during the second year. Following its publication, the microscope has received considerable attention and has now been duplicated in several labs throughout Europe. The experience gained in the construction of the supercontinuum based STED system has emphasised the key role of the light source for the further improvement and dissemination of STED microscopy. The developed STED instruments provide all the necessary functionality, including dual-colour ultrahigh resolution capabilities, clearly representing two of the most significant readout parameters for our diagnostic procedure. In terms of robustness and usability they are close to commercial STED instruments currently on the market, but offer higher resolution and true dual-colour capabilities. Moreover, the cost of the instruments is almost an order of magnitude lower than that of current commercial instruments. These instruments may thus serve as a blueprint for a future commercial instrument. Following establishment of the STED instruments, proof-of-concept studies on cultured cancer cells were performed.
2. TRAST imaging: The information content of the nanoscopy images can be complemented by also encoding spectroscopic information. In this project, a concept for TRAST microscopy developed at KTH and opening for the possibility to monitor transient photo-induced non-fluorescent states of fluorescent molecules on a massive parallel scale and at arbitrary concentrations, was further developed towards diagnostic applications. This concept can add considerable sensitivity to existing fluorescence microscopic readouts, providing also an additional dimension of fluorescence information that has previously been almost fully overlooked in fluorescence microscopy. The TRAST approach was demonstrated as a tool to image and spatially resolve the metabolism within individual live cells and to detect differences in metabolic activity typically occurring in cancer cells.
3. MFDi: In order to be able to extract sufficient diagnostic information out of sparse, or even individual FNA-sampled cells, it is necessary not only to monitor species of one target protein in the cells, but of several different proteins at the same time. HHU has developed a new general strategy based on MFD to register and quantitatively analyse fluorescence images. In MFDi pulsed excitation with typically at least two different lasers is used for generation of fluorescence in the sample. Following optimisation of instrument hardware, software as well as sample preparation protocols, final MFDi measurement routines could be established on cultured cells, allowing detection and identification of multiple significant molecular or cytohistological features in the cells. This formed the basis for the subsequent application of MFDi on multiply stained FNA patient samples. This also demonstrates that the developed acquisition and analysis procedures match the prerequisites for further adoption of the MFDi technique as a common diagnostic tool of cellular samples.
4. TPX technology for ultrasensitive detection and identification of pathological protein profiles from cytological aspiration material: The TPX technology, developed at the Laboratory of Biophysics (UTURKU), is extremely well suited to measure concentrations of target molecules from minute samples with high precision and sensitivity. The strategy was to develop the TPX technology to monitor the prescence of nonlocalised proteins and their concentrations in extracellular aspiration material, providing complementary information to that obtained from the fluorescence-based imaging methods outlined above. Development within the project has demonstrated that the assay should be readily integrable to the FNA sampling protocol of the project. A small aliquote of the original sample could be set aside for TPX analysis, providing complementary diagnostic information to that obtained from the cellular imaging.
Combined diagnostic procedure performed on FNA-sampled material
In the instrument development, the different modalities were developed with the consideration to be used either in a parallel, simultaneous approach within one integrated instrument, or the combination of techniques would be achieved by a sequential use of the techniques on the FNA samples.
Several setups were established in the project enabling a parallel use of techniques. It was realised that although the FNA samples gave only a limited amount of diagnostic material, they still contained a sufficient number of cells to be divided on several glass slides. Given this possibility, it was concluded that the diagnostically most significant combination of accessible parameters could be obtained by performing different imaging modalities separately and sequentially. Thereby, the conditions for each of the measurement techniques can be optimised independent of each other and one can avoid the compromises that follow from a combined imaging with two or more of the modalities, on the same cells and at the same time.
A sample handling and preparation protocol was defined taking into account that FNA samples are far more complex and heterogeneous than standardised cells from cultured cell lines and may contain blood cells, fatty tissue and other contaminations. Extensive effort was required to yield optimal fluorescence characteristics and to ensure that the cellular handling does not compromise any of the most significant readout parameters of the different procedures used to image the cells. The optimisation of the sample handling and preparation had to incorporate many aspects, including proper specificity of several affinity molecules against their molecular targets, fluorescence intensity versus background, stability, reliability and reproducibility of the recorded parameters, maintenance of cellular morphology and specific cellular features, proper adhesion of cells to glass cover slides and preparation formats compatible with transport, as well as sufficient ease of preparation at clinical sites with typically limited resources and infrastructure for cell handling. Taken all these aspects together favoured a procedure including fixation of the sampled cells for stability and reproducibility reasons. The advantages of using fixed samples more than well compensated for the additional information potentially retrievable from unfixed, or even live, cells. To include fixation in the preparation of the FNA samples essentially also defined STED and MFDi as the combination of techniques most suitable for the diagnostic characterisation of the FNA-sampled cells.
With the FNA sample handling and preparation properly optimised, with a combination of MFDi and Nanoscopy identified as the diagnostically most useful combination of imaging methods providing uniquely retrievable diagnostic information and finally with a set of parameters and targets identified from the initial studies by STED and MFDi on cultured cells and then on cells from a few FNA samples as likely targets to provide high diagnostic power, tests were finally performed on a larger set of clinical FNA samples. Clinical FNA sampling from patients with suspect cancer lesions was performed in Stockholm and a few times also in Uppsala. Samples were obtained from suspect cancer lesions both from the breast and from the prostate, but with the major part of the samples obtained from the breast. After reserving a major part from each FNA sample for the routine diagnostic assessments, the remaining part of the samples were directly taken for preparation, distributed on up to 12 different glass slides and then passed on for imaging, locally to KTH and to HHU and MPIBPC. From the FNA samples, high-resolution images and spectroscopic data were acquired at KTH, HHU and MPIBPC at four different platforms using the different and complementary techniques.
Taken together, samples from 97 patients were altogether sampled at the different imaging sites as described above. Several features indicative of malignant development could be observed in both the MFDi and STED images already by some visual inspection. However, for a more quantitative analysis all image data from the FNA samples were sent to partner UH for bioinformatic analysis and verification.
Bioinformatic tools for diagnostic evaluation and final clinical verification
The modern imaging technologies developed and used in this project are capable of measuring intracellular molecules with a spatial resolution unimaginable by most researchers in the field 10 years ago. The high-resolution and high-content of the images requires advanced computational methodologies (bioinformatics) to translate the image data into knowledge of clinical benefit. In the project, novel and accurate microscope technologies were developed and applied to FNA-sampled cells from suspect breast and prostate cancer lesions. The major challenge at this point was the development and implementation of computational methods capable of pre-processing and analysing the image data systematically and fast. Before this project there were no published computational methods for pre-processing and analysing STED and MFDi images at the scale required in clinical settings.
The computational analysis of image data and integration of the extracted information from the images was done by Systems Biology Laboratory at University of Helsinki. Firstly, this project resulted in 1 486 STED and MFDi images from 63 patients. By using a collection of microfeature detection methods, we were able to detect hundreds to thousands of feature locations from a single image. Each of these locations, in turn, provided several quantitative values that were used in a prediction step. In order to be able to analyze such a large amount of data, we used a computational framework called Anduril (Freely available at http://www.anduril.org online. Analysis, scripts and results can be found at http://csbl.fimm.fi/pub/FLUODIAMON online) that allowed us to perform all mathematical operations in a systematic fashion. After developing the management system for systematic use of the images, we developed novel pre-processing methods for MFDi and STED images.
The main objective of the bioinformatics analysis was to identify markers that predict whether the cells obtained with FNA were normal or cancerous cells. The development of computational methods for cutting edge, high-resolution and multiparameter microscopy has resulted in freely available software for analysing large amounts of images in a systematised fashion, novel methods that are widely usable in various experimental settings and microscope techniques and clinically important predictors that can aid in breast cancer patient diagnosis.
Overall diagnostic procedure and its validation
Combining the efforts and progress within the different areas as outlined above, an overall diagnostic procedure could be established in line with the initial intentions. An intact chain of accomplished steps, from reliable and patient-friendly FNA sampling with minimised risk of tumour cell dissemination, over robust and reproducible sample handling and preparation, specific targeting of selected proteins with fluorophore-labelled affinity molecules, characterisation of the FNA sampled cells by optical methods offering utmost sensitivity, specificity and spatial resolution and finally bioinformatic analyses to fish out the differences in the image features necessary to distinguish normal from malignant cells, has lead to the demonstration of a novel diagnostic concept. This concept can combine diagnostic reliability with safe, minimally invasive diagnostic sampling and thus enable a decisive improvement of the outcome of patients suffering from breast and prostate cancer.
For validation, the data from the imaging configurations for the different patient FNA samples were compared to the diagnoses set for these patients clinically (collected and compiled by KI). The comparison showed that we managed to create six independent classifiers capable to detect cancer from the different image data in minimally 55 % of the cases and maximally in 80 % of the cases. It can thus be concluded that each of the six independent diagnostic classifiers can yield significant predictive strength. As a next step, in order to further improve the classification accuracy, several classifiers can be combined. Our early results confirm that a clear improvement in accuracy can be obtained, though this need be tested more thoroughly with more patient samples in the near future.
Complementary clinical validation was also focussed on the collection, preparation and analysis of clinical material of formalin-fixed and paraffin-embedded (FFPE) benign and malignant breast cancer specimens as well as of normal adjacent breast tissue. Together with experienced histopathologists, these specimens were evaluated and the most representative areas were used for generating a tissue microarray (TMA) that allows for standardised high-throughput immunohistochemistry analysis and marker validation.
Taking these validation paths together, the developed classifiers for the images from the patient FNA samples have been directly validated against the set clinical diagnoses for the same patients and the relevance of the selected markers have been verified on a large number of relevant patient material. The diagnostic procedures have been validated and shown to provide an attractive strategy for future clinical diagnostic schemes.
Potential impact
The FLUODIAMON project is highly interdisciplinary. The expertise needed for this project comprises clinical cytology, cancer proteomics, molecular biotechnology, fluorophore chemistry, fluorescence microscopy, nanotechnology, optics, solid state detector technology, data processing and bioinformatics. The level and width of competence required cannot be found on a national level but requires a European initiative. In particular, the project has demonstrated that by applying analyses based on novel state of the art optical microscopy techniques, even quite small amounts of diagnostic material is sufficient for reliable diagnoses of breast and prostate cancer. Thereby, the amount of diagnostic material obtainable via minimally invasive FNA is sufficient and more invasive core-needle or surgical biopsies can be avoided.
FNA techniques
The FLUODIAMON has supported three major advancements in FNA sampling technology. Firstly, a new FNA needle has been developed with a higher sample yield and improved surface treatment for higher ultrasound visibility. Secondly, by adding mechanical energy to the needle during the sampling the penetration of the needle has been considerably improved. Thirdly, an anti-seeding procedure has been established and clinically tested.
These advancements are likely to have a large impact on future diagnostics. The new FNA sampling technology contribute with considerable improvements on several of the most important aspects of diagnostic sampling: increased amount of sample material with the same degree of invasiveness, higher precision of positioning and penetrability is likely to increase the fraction of obtained samples that indeed are from the region of interest and not the least the risk of tumour cell dissemination is strongly reduced. These features make the FNA sampling technology very attractive for breast and prostate cancer diagnostics. However, it is also worth pointing out that the needle development can also to a large part be applied on core-needles as well. For the case core-needle would still be needed, mechanical oscillations and anti-seeding procedures can strongly improve the features in a similar way also for these larger diameter needles. The developed needle technology is also not only limited to breast and prostate cancer diagnostics. There is a range of different forms of cancer, where invasive sampling is to be avoided, but where FNA harnessed with the new technology can be considered as a possible way for diagnostic sampling. In general, the needle development within this project can have a considerable impact on all forms of diagnostic sampling, of cancer or infectious diseases, where risk of spread, accessibility, or risk of damage of vital tissues is an issue. Finally, the technology for diagnostic needle sampling also seems to have implications for treatment.
Affinity molecule development
In the FLUODIAMON project, a set of affinity binders were identified for a selected group of target proteins, well functional also under the conditions relevant for FNA sampled cells. The project has shown that novel technologies for development of novel affinity proteins of different nature can be very valuable for future diagnostic applications. Compared to conventional, full-sized antibody binding reagents, the alternative binder protein classes used can provide advantages in terms of size, the preference to and selectivity of where on the target protein they bind. The two groups of affinity binders developed in the project are considerably smaller than conventional antibodies. The small size of the marker molecules make them very valuable reagents for the optical techniques, where the high resolution offered may even make the actual size of the binder an issue. The small size may also allow binding to sites on the proteins not accessible by larger binders due to steric reasons. Since at least affibodies can be designed such that the position of the fluorophore marker molecule(s) can be specified to desired locations at the affibodies, also the fluorescence properties can be modulated and optimised via the design of the affibodies. This controlled fluorescent labelling, including both number and position of fluorophores, was taken benefit of in the project, where the fluorescence properties of the samples are very important for the overall performance of the high-end fluorescence imaging methods used.
Taken together, the affinity molecule development in the project has demonstrated the feasibility of using novel class of affinity molecules for high-end optical imaging and diagnostics and that the specific properties of the affinity binders can offer specific advantages compared to conventional antibody binders. The combined use of high-end fluorescence microscopy and these novel groups of affinity binders will have applicability for a wide range of cellular imaging, extending well beyond breast and prostate diagnostics.
Fluorophore development
In the FLUODIAMON project the access to high-performance fluorophores has been critical to the development of the diagnostic imaging procedures. The demands on the fluorophore marker molecules are for both high-resolution microscopy and high-sensitivity, multi-parameter, fluorescence-based characterisation of proteins considerably higher than in more conventional fluorescence imaging and spectroscopy. Besides carrying specific chemical groups so that the fluorophores can be coupled to relevant affinity proteins or to other molecules of interest and spectral properties suited to the optical instrumentation, the most important properties of the labels are high fluorescence quantum yield and high photostability. Our project has demonstrated the feasibility and strength of using high-end fluorescence imaging methods for breast and prostate cancer diagnostics. However, the applicability of these imaging methods goes far beyond this particular application. No matter the application, the performance rely on the use of appropriate fluorophores and the developed fluorophores in this project and the experience coming from the development and use of these, will be highly useful in a range of applications in diagnostics, cell biology and biosciences in general.
Cancer proteomics
In a sense, the project merged knowledge about tumour-specific markers with the use of highly sensitive, high resolution imaging methods by which the markers can be detected and identified down to a single molecule level and by which their molecular status can be analysed in the context of individual sample cells. In particular, the project has shown for a limited number of selected protein targets that apart from plain deviations in the amount of the proteins in the FNA sampled cells, the diagnostic strength can be considerably increased by further extending the protein characterisation to include also high-resolution spatial distribution and high-sensitivity expression level profiling of several proteins in the individual cells at the same time. The impact this project will have in the context of protein biomarkers thus essentially lies in the notion that not only expression levels of certain individual proteins matters. Our results suggest new aspects to consider when identifying new tumour markers in the future.
Development and optimisation of high-end fluorescence methods for diagnostics of FNA-sampled cells
For the diagnosis of breast and prostate cancer today's standard procedures for diagnostic sampling are core needle or surgical biopsy. However, due to their mechanical crudeness, these procedures are highly traumatic and encumbered with a significant risk for tumour cell seeding in the needle tracts. In contrast, FNA offers a minimally invasive, extremely patient-friendly extraction of cellular material. Four different methodologies were developed in the project and although only two of those were combined in the final characterisation of the FNA sampled cells, their results will have an impact on future cellular imaging and diagnostics.
1. STED microscopy: The development of the STED technique within the project has lead to the first demonstration of its use for cancer diagnostics, but also paves the way for future applications of ultrahigh resolution microscopy in cell biology and biomedicine in general.
2. TRAST imaging: In the project, several realisations of the technique in various setups have shown its broad applicability and the ability to retrieve information coupled to low frequency molecular encounters or metabolic states have been demonstrated. Since the metabolic state and the regulation of the metabolism is one of the first things to be influenced in cells upon disease and toxic influence, the ability to monitor the metabolic state and the metabolic regulation of cells is also likely to open for applications in toxicology, drug tests and several other applications in fundamental cell biology.
3. MFDi: In this project, a new general strategy based on MFD to register and quantitatively analyse fluorescence images was further developed for imaging purposes (MFDi). The developed MFDi technology has been demonstrated as a useful diagnostic imaging method and a similar use of MFDi for subcellular diagnostics can most likely be extended to a broader range of diseases in the future. MFDi could be broadly used in cell biology in general. Parts of the hardware and the concepts for fluorescence data acquisition and processing included in MFDi can also be used outside of the context of MFDi. Partner B&H has developed and adapted several instrument components as well as concepts of this kind for commercial use in this project.
4. TPX technology: The TPX-technology, originally developed by the UTURKU group and commercialised by Arctic Diagnostics Ltd is extremely well suited to measure concentrations of target molecules from minute samples with high precision and sensitivity. The assay development within the project has demonstrated that the assay should be readily integrable to the FNA sampling protocol of the project. In addition, the assay development within the project has also set the ground for developments of the TPX technology towards a new method for detection of MRSA and the development of TPX assay for bacterial counting have further paved the way towards quantitative and sensitive TPX assays of cellular matrices.
Bioinformatics
Bioinformatic tools were developed to analyze the large number of images taken by STED and MFDi on FNA-sampled cells from suspect breast and prostate cancer lesions. The high-resolution and high-content of the images required advanced computational methodologies (bioinformatics) to translate the image data into clinically useful information. Before this project there were no published computational methods for pre-processing and analysing STED and MFDi images at the scale required in clinical settings.
With the goal to identify markers that predict whether the imaged FNA-sampled cells were normal or cancerous, partner UH developed computational methods based on strong prediction methods developed in the machine learning community. The efforts resulted in clear improvement in diagnosing patients with the FNA procedure. The development of computational methods for cutting edge, high-resolution and multiparameter microscopy has resulted in:
1. freely available software for analysing large amounts of images in a systematised fashion
2. novel methods that are widely usable in various experimental settings and microscope techniques and
3. clinically important predictors that can aid in breast cancer patient diagnosis.
In the project, in particular the study of microfeatures has emerged as a novel powerful approach in the analysis of images, opening new possibilities in all studies, where the resolution is large enough for imaging subcellular features. The outcome of the bioinformatic analysis demonstrates the power of multiple marker and/or ultrahigh resolution spatial distribution pattern analyses of proteins on a subcellular level as a new path for early, sensitive cancer detection, but the developed image analysis tools also have a more general applicability.
Overall diagnostic procedure
In this project, we have developed and successfully demonstrated a joint methodology for objective, quantitative sub-cellular diagnosis of early breast and prostate cancer disease. By applying analyses based on novel state of the art optical microscopy techniques much more information can be extracted and even quite small amounts of diagnostic material is sufficient for reliable diagnoses of breast and prostate cancer. Thereby, the amount of diagnostic material obtainable via minimally invasive FNA is sufficient and more invasive core-needle or surgical biopsies can be avoided. The potential impact following from this successful development of the methodology cannot be overestimated:
1. it can strongly contribute to an increased cure and survival for breast and prostate cancer patients
2. it should be able to provide valuable support guiding decisions whether therapeutic interventions are to be undertaken and if so, when and to what extent
3. it should minimise the risk of cancer cell seeding, patient discomfort and other side effects such as pain and infections
Outside of breast and prostate cancer diagnostics and progression monitoring, the methodology should be applicable also to other diagnostic fields facing the same principal conflict between minimal invasiveness and the information content needed for a clear diagnosis.
As a major, overall result of this project, we have obtained a diagnostic methodology for breast and prostate cancer that combines diagnostic reliability with minimal invasiveness and a negligible risk for sampling-related side effects, in particular cancer cell seeding and spread. This combination is impossible to obtain with other presently available methodologies. The ability to obtain conclusive information with minimal invasiveness and without risk of cancer cell seeding and spread also enables safe and reliable repetitive sampling and thereby quantitative monitoring, analysis and prediction of disease progression over time, e. g. for follow-up of pre-malignant lesions or for evaluation of response to therapy. The developed methodology can therefore also be expected to form the basis for important diagnostic tools supporting and guiding possible therapeutic interventions. It is also reasonable to predict that the proposed methodology would offer significantly improved and profoundly new possibilities for monitoring and analysing, on a subcellular level, molecular mechanisms for tumour development and effects of various therapies. The project results thus clearly can be expected to contribute to the development of new therapies and pharmaceuticals and in particular therapies based on molecular and cellular mechanisms.
In the forefront of the potential impact that this project may bring about is the fact that breast and prostate cancer belong to the most frequently occurring malignancies in women and men, respectively, in the western world. When possible, detection of these diseases at the earliest stage results in a 15 year survival rate of more than 90 %. In contrast, survival of breast cancer patients exhibiting advanced stages has not improved during the last 20 years. Thus, for improving survival of patients suffering from breast and prostate cancer, the most important task is to detect, diagnose and treat malignancies as early as possible. The benefit to patients that patient-friendly and secure sampling methods for breast and prostate cancer would bring about, offering the necessary diagnostic sensitivity, specificity and reliability, is enormous.
Furthermore, this project improves one of the most important quality aspects in the diagnostics and treatment of patients with breast and prostate cancer - that sample extraction may cause severe side effects and even worsen the outcome. At this stage, we have developed, demonstrated and clinically verified the methodology. Already at this stage, the procedure is objective and with additional refinements of our combined procedure it is likely that the diagnostic sensitivity and specificity can be even improved compared to today's standards.
Dissemination activities and exploitation of results
The activities within FLUODIAMON have throughout the project generated a continuous flow of dissemination activities.
In total FLUODIAMON have to-date resulted in 24 peer-reviewed articles, all published in well recognised scientific journals. Ten additional publications are either already submitted or soon to be submitted. Moreover, seven PhD theses and four MSc theses have to date been achieved with support from the project. In addition, several PhD students will graduate in the next one to two years and their PhD theses are also to a large extent a fruit of and supported by the project.
Among dissemination activities one can also mention several open access software packages for classification and image analysis, interviews in newspapers and TV, exhibitions and well over 50 invited oral presentations and posters at international conferences and workshops.
On the commercial side, 10 specific patents or patent applications, related to needle sampling technology, fluorophores and detector and imaging technology, have been either generated or have directly profited from the activities and results within the project. A corresponding development of several different commercial products within these areas has also commenced as a result of the project.
In conjunction to the last consortium meeting, a workshop was arranged in Stockholm on 2 December 2011 ('Ultra-high resolution and ultra-sensitive fluorescence methods for objective sub-cellular diagnosis of early disease and disease progression in breast and prostate cancer'). This workshop was very well attended and brought researchers from the full range of disciplines covered within the FLUODIAMON project. Finally, last but not least, the progress within the project now allows us to disseminate the results also to a wider audience. From a patient perspective, considerable benefits have been demonstrated. A first seminar with breast cancer patients has been arranged by partner KI to disseminate these results. Given the positive and encouraging results from the large-scale assessment and clinical verification, a larger meeting with breast cancer patients and clinicians is already planned together with the major patient organisations in Sweden during the autumn 2012 and similar activities will also be planned in Germany.
Taken together, the FLUODIAMON projects has yielded a very broad range of results, which has been disseminated scientifically, on the commercial side and not the least due to the large impact on future cancer diagnostics, towards patient organisations and clinicians.
Project website: http://www.biomolphysics.kth.se/fluodiamon
Contact details of the coordinator:
Prof. Jerker Widengren
KTH, Dept Applied Physics / Exp Biomol Physics
Albanova Univ Center, 106 91 Stockholm
Telephone: +46-855-378030
Fax: +46-855-378216
Email: jwideng@kth.se
The survival chances of cancer patients can be dramatically increased by an early diagnosis. This also holds true for breast and prostate cancer, which are among the most frequent forms of cancer in Europe. However, with currently available methods, typically core needle biopsies, it is not possible to extract the necessary amount of sample material for a reliable diagnosis without running a high risk of severe side effects, not the least cancer cell seeding and risk of spread of the cancer to be diagnosed. As an alternative, fine-needle aspiration (FNA) cell sampling offers a patient friendly, minimally invasive, rapid and cost efficient tumour diagnostic procedure with minor side-effects. However, the amount of sample obtained by FNA is often sparse and consists of aspirated cells taken out of their tissue context. It is therefore often difficult to obtain a conclusive diagnosis based upon this sample material.
In this project, we have addressed this conflict between the amount of sample material needed to obtain conclusive diagnostic information and the severe side effects that can be caused by the diagnostic sampling itself. The overall objective of this project was to develop and validate a quantitative, minimally invasive diagnostic tool for early and conclusive detection, diagnosis and monitoring of disease progression of breast and prostate cancer. A methodology was developed making use of a combination of exciting recent advances in the field of fluorescence-based microscopy, for imaging of individual FNA-sampled cells. The methodology includes advances taking the spatial resolution far beyond the fundamental limits of optical resolution, the sensitivity down to an ultimate single-molecule level and multi-parameter detection schemes significantly increasing the fluorescence information by which these cellular images can be analysed. Apart from detecting and identifying specific protein tumour markers in the samples, tumour-specific spatial distribution patterns of the proteins within intact sample cells were also exploited. This represents a to-date almost unexploited dimension of diagnostic information. The novel optical methods were supported by state of the art affinity molecule biotechnology, fluorophore chemistry and proteomics-based biomarker identification, with the overall aim to extract a maximum amount of information out of the small amounts of sample material obtained by FNA.
Following extensive effort in optimising the sample preparation, handling and labelling and corresponding effort to optimise the performance of the fluorescence imaging techniques, first on cultured cells as models and then on cells from clinical FNA samples, a larger number of FNA samples were analysed with a combination of ultrahigh resolution fluorescence microscopy and imaging based on multi-parameter fluorescence detection. New bioinformatic tools were developed and used to analyse the acquired images and features and classifiers were established which in an objective manner could classify the imaged samples as normal or malignant. In parallel, progress in FNA sampling technology performed within the project led to further reduced risks for cancer cell seeding and increased the ability to obtain representative diagnostic material. Taken together, we can thus demonstrate an intact chain of accomplished steps, from reliable and patient-friendly FNA sampling with minimised risk of tumour cell dissemination, over robust and reproducible sample handling and preparation, specific targeting of selected proteins with fluorophore-labelled affinity molecules, characterisation of the FNA sampled cells by optical methods offering utmost sensitivity, specificity and spatial resolution and finally bioinformatic analyses to identify the differences in the image features necessary to distinguish normal from malignant cells. This represents a novel diagnostic concept, which can combine diagnostic reliability with safe, minimally invasive diagnostic sampling and thus enable a decisive improvement of the outcome of patients suffering from breast and prostate cancer.
Project context and objectives:
The survival chances of cancer patients can be dramatically increased by an early diagnosis. This particularly holds true for breast and prostate cancer, which are among the most frequent forms of cancer in Europe. With currently available methods, typically core-needle biopsies, it is not possible to extract the necessary amount of sample material for a reliable diagnosis without running a high risk of severe side effects.
On the other side, FNA based cell sampling technology is a patient friendly, minimally invasive, rapid and cost efficient tumour diagnostic procedure with negligible side-effects. However, despite the overall advantages of FNA-based diagnostics for breast and prostate tumours, there are three major drawbacks that prevent a general application of FNA in clinical routine:
1. FNA-based sampling in general only yields a sparse amount of diagnostic material, with aspirated cells in an extracellular fluid
2. FNA samples are frequently non-representative by containing too few cells of the suspicious lesion.
3. Even with a successful lesion biopsy it is often difficult to obtain a conclusive diagnosis based upon the frequently minimal cellular deviation of atypical or even fully transformed malignant cells compared with their normal counterparts.
For these reasons, core needle biopsy or surgical biopsy are to date still favoured over FNA for breast and prostate cancer diagnostics. However, due to their mechanical crudeness, they are highly traumatic and combined with significant risk for tumour cell seeding in the multiple needle tracts.
The overall objective of this project was to develop and validate a quantitative, minimally invasive diagnostic tool for early and conclusive detection, diagnosis and monitoring of disease and disease progression of breast and prostate cancer, with negligible sampling-related side-effects.
To increase the amount of information that can be extracted from the sample this project has taken benefit from the recent remarkable progress in fluorescence-based spectroscopy and imaging. In particular, we brought in exciting new concepts to increase the sensitivity and specificity of fluorescence detection and, not the least, to overcome the diffraction barrier in optical microscopy. Based on these concepts and methodologies, we could put in sight a development towards novel optical diagnostic methods unimaginable just a few years ago and to develop the means necessary to eliminate the conflict between the amount of sample material needed and the necessary diagnostic conclusiveness.
The overall strategy to be achieved within the FLUODIAMON project was thus to exploit the combination of the unique sensitivity, specificity and spatial resolution offered by the very latest developments in fluorescence microscopy and the minimal invasiveness of FNA based cytology. Based on the fluorescence methods the project aimed at developing and applying approaches where minute FNA samples could be subject to objective analyses on a molecular or subcellular level, thereby providing sufficient diagnostic information. The strategy of the project was in particular to analyse three main categories of proteins within the FNA-sampled cells: cytoskeletal proteins, receptor proteins located in the cellular membranes and proteins regulating the division cycles of the cells. Apart from the mere detection and identification of proteins of these categories in the sampled cells, tumour-specific spatial distributions of these proteins within the intact sample cells were also exploited.
To reach the overall goal of the project, efforts not only in the further development of fluorescence microscopy/spectroscopy and FNA sampling technology were needed. Following FNA-sampling, the FNA samples needed to be properly prepared; and diagnostically representative proteins in the cells needed to be identified and marked using specific affinity molecules, labelled with bright and stable fluorophore marker molecules. Extraction of diagnostically significant fluorescence information from the FNA-sampled cells needed to be optimised based on iterative efforts within optical imaging and sample preparation and labelling. Finally the acquired images of the FNA-sampled cells needed to be analysed with new bioinformatic means and finally validated towards clinical data.
Given the effort required spanning over a broad range of disciplines, the main objective of the project was also divided into a set of detailed objectives, reflecting the whole chain of steps needed to reach the final diagnostic procedure aimed for. The detailed objectives of the FLUODIAMON project were:
1. to improve spatial resolution of state of the art light microscopy in pathology by an order of magnitude
2. to improve the sensitivity of fluorescence-based imaging of FNA acquired cells to the ultimate single-molecule level
3. to take multi-parameter fluorescence imaging of individual FNA acquired cells to its extreme in terms of information content, largely based on photon statistical approaches and parameters extracted from non-linear effects
4. to develop standardised FNA-based sampling of suspected breast and prostate cancer lesions with negligible side-effects and with optimised needle visibility for ultra-sound guided needle positioning
5. to select already known molecular markers that will be compatible with the developed fluorescence imaging techniques and can be highly anticipated, when investigated by these techniques, to strongly correlate with malignant transformation and clinical tumour aggressiveness
6. to identify existing and develop new affinity molecules to these markers, which are highly specific and fluorophore labelled for optimised fluorescence readout properties
7. to refine bioinformatic evaluation and data processing to find the combination of fluorescence read-out parameters that most strongly correlate with the relevant clinical parameters and yield the strongest diagnostic reliability
8. to optimise the combination of techniques and procedures, to maximise sensitivity and specificity for FNA-based diagnostics of breast and prostate cancer, enabling a decisive improvement of the outcome for the patients suffering from these diseases.
Project results:
In the forefront of the project efforts was the further development and adaptation of cutting-edge optical methods so that they could extract sufficient diagnostic information out of very small amounts of diagnostic material. Reaching the overall objective of this project required interaction between a broad range of different disciplines, as well as significant progress within each of these disciplines. In particular, substantial effort and progress were needed within the following areas, covering all steps from the initial sampling step to the final evaluation of the acquired images and the validation of the whole diagnostic procedure:
1. sampling of suspect breast and prostate cancer lesions with minimal side effects by FNA
2. preparation and handling of the clinical FNA samples
3. selection of protein targets to be visualised by the optical methods in the cells in the FNA samples
4. selection and development of affinity binders
5. selection and development of fluorophore marker molecules which provide a maximum of sensitivity and specificity
6. development and optimisation of high-end fluorescence methods for diagnostics of sparse amount of cells sampled by FNA. Four different methodologies were considered in the project, namely ultrahigh resolution fluorescence microscopy - nanoscopy, or stimulated emission depletion (STED) microscopy -, multi-parameter detection imaging (MFDi), transient state (TRAST) imaging and two-photon excitation (TPX) spectroscopy.
7. development of bioinformatic tools for the diagnostic evaluation of the protein content and the subcellular features within the imaged FNA sampled cells.
Development of FNA sampling procedures of suspect breast and prostate cancer lesions
FNA based cell sampling technology is a patient friendly, minimally invasive, rapid and cost efficient tumour diagnostic procedure. Conventional methods side-effects are completely negligible with the FNA technique. In this project, the Stockholm-based small company Neodynamics AB has managed to improve the FNA sampling procedures in several aspects.
Firstly, a new FNA needle was developed with improved sample yield and surface-treated for considerably increased ultra-sound visibility. Sample yield (i.e. the amount of sample material obtained after needle aspiration) is important, given that FNA-based sampling in general only yields a very sparse and often also insufficient amount of cell sample material. Ultrasound visibility is important since upon sampling, when the needle is introduced in the tissue, the positioning of the needle prior to the aspiration relies on the guiding and visualisation via ultrasound. If the needle position is not positioned at the right place, the sample is obviously not obtained from the lesion region to be investigated. Final development on the needle design was performed resulting in two prototypes (outer diameter 0.7 mm) with and without enhanced ultra-sound visibility. Large scale clinical testing was carried out in order to evaluate their performance. The new needle achieved outstanding results when compared to needles commonly used today. It yielded three times more material than standard needles of the same diameter and two times more than needles with a diameter of 0.7 mm.
Secondly, needle additives to facilitate needle penetration, manoeuvrability and precision in positioning were developed. In the device developed by Neodynamics AB called 'Cytotest' mechanical energy is added to the sampling-needle by computer controlled oscillating movements. These oscillatory movements of the needle have been demonstrated to strongly facilitate tumour penetration, allow the direction of the needle to be changed to a larger extent upon penetration, as well as to increase the sampling phase efficacy. Also a rotating movement was added to the needle in order to mimic the way the needles are handled by trained cytologists. Moreover, the design of the bevel section of the needle has been adapted for optimal cutting behaviour together with the oscillatory and rotational movement schemes. 'Cytotest' has been used at two breast diagnostic centers in Stockholm (Sabbatbergs Breast Center and St Göran Hospitals Breast Unit) on more than 500 breast cancer patients with excellent results. Due to concerns from the operators regarding decrease of tactility upon sampling when using the oscillatory needles, an improved handle was designed, in collaboration with the design programme at Kalmar University, Sweden and successfully tested by clinical operators. The needle handle will be implemented into the 'Cytotest' instrument before large-scale production is initiated.
Thirdly, Neodynamics AB has succeeded with the development of a fully functional anti-seeding instrument as well as a specialised anti-seeding needle. The anti-seeding instrument consists of a signal processor, an amplifier as well as a patient unit. The anti-seeding needle tip acts as an electrode whereas the shaft is isolated. When radio frequency pulses are given, a current is applied between a dispersive electrode, for instance located over the chest of the patient when sampling from the breast and the needle electrode. Thereby, the tissue around the open needle tip is heating up in a much localised fashion where the current density is highest. Starting with tests on pork livers provided the knowledge to successively improve the used pulses and their properties, an ethical permit was then acquired in order to commence with clinical studies. From the tests on patients undergoing the sampling procedure based on the anti-seeding procedure the results have been positive in every aspect. No or very little additional discomfort is experienced by the patients and the analysed samples have not been denaturised. To estimate the extent of local tumour cell-dissemination upon FNA sampling, partners Neodynamics AB and KI (Wiksell and Auer) investigated the blood droplet(s) that spontaneously penetrates the skin orifice after FNA needle retraction. They found that there is a high frequency of these blood droplets contained tumour cells. From this finding it is reasonable to assume that upon needle sampling, cancer cells can also enter the circulating blood system of the patient and may finally also cause distant metastasis. Interestingly, applying the developed anti-seeding procedure significantly reduced the overall amount of blood and exudate passing the skin orifice after retraction of the FNA needle. Moreover, a test series on patients indicated that there were no viable suspended tumour cells in the blood droplets leaking out of the skin orifice after needle retraction.
Taken together, the development on FNA sampling techniques is likely to lead to a dramatic improvement regarding sampling related risks, sample reproducibility, quality and volume of collected cells as well as regarding safety against cell tumour dissemination. These developments also give the necessary prerequisites, covering the first steps in the diagnostic procedure developed in the project.
Preparation and handling of the clinical FNA samples
Once a patient sample is safely and reliably obtained by the FNA sampling procedures, the subsequent preparation and handling of the cellular material need to be optimised and properly adapted to the overall diagnostic procedure. In the project, the optimisation of the sample handling and preparation had to incorporate many aspects:
1. the sample preparation should not compromise the specificity of the affinity molecules against their protein targets in the cells
2. the recorded fluorescence intensity from the fluorophore marker molecules (labelled onto the affinity molecules) versus the background level in the imaging had to be maintained, as well as the stability, reliability and reproducibility of the recorded fluorescence parameters
3. the cellular morphology and specific cellular features should not be severely affected by the sample handling and preparation
4. the preparation should yield a proper adhesion of cells to glass cover slides and preparation formats compatible with transport
5. the preparation protocols should be sufficiently simple, so that they could be performed at clinical sites with typically limited resources and infrastructure for cell handling.
Taken all these aspects together favoured a procedure, where immediately after sampling the patient material was subject to a three-step procedure consisting of a blood cell lysis step, a fixation and a cytospin step, thereby preparing the sample for subsequent sending, labelling and imaging.
Selection of protein targets
In general, the cure of cancer diseases strongly depends upon early and detailed diagnosis of type and stage of the malignant disease, typically requiring a molecular characterisation of the malignancy. Consequently, the aim of the project was to further develop and apply analyses of the minute FNA samples on a molecular or subcellular level. Partners Karolinska Institutet (KI) and Lübeck University (LUEBECK) have previously identified a substantial number of proteins in freshly taken clinical samples from tumours of the breast, prostate and others. By comparing cancer type specific protein expression with that of corresponding normal tissue, single proteins or groups of proteins were identified, highly correlated to malignant transformation and clinical tumour aggressiveness.
In this project, we wanted to extend the protein characterisation beyond the plain presence/absence, or rather deviations in the amount of these proteins in the FNA sampled cells or in the extracellular sample material. Given the sensitivity and spatial resolution offered by our fluorescence methods, we also wanted to map the spatial distributions of selected proteins within the FNA sampled cells as objective parameters specific for malignant transformations. The project focussed on diagnostically monitoring three different categories of proteins: cytoskeletal proteins, membrane proteins and cell cycle regulating proteins.
The specific proteins selected to demonstrate the diagnostic procedure and the diagnostic feasibility of the microscopic techniques were found from extensive literature searches and among our already identified tumour markers. The final condensed list of protein biomarkers for breast and prostate cancer were then selected according to the following parameters:
1. presence or absence of protein markers in disease as compared to controls
2. potential of changes in spatio-temporal behaviour and interaction patterns in diseased stages
3. changes in the sub-cellular location of the protein upon disease.
In the project, the diagnostic procedure was eventually narrowed down on analysing the abundance, interactions and/or the spatial distribution patterns of the following proteins in the FNA sampled cells: vimentin and tubulin (cytoskeletal proteins), IGF1R, HER1, HER2 (membrane proteins), Cyclin A, Cyclin E (cell cycle regulating proteins).
The monitoring of the presence and spatial distribution patterns in individual cells of these specifically selected proteins proved to be a successful strategy.
Selection and development of affinity binders to the protein targets
The implementation of the novel and advanced fluorescence-based imaging techniques for the diagnostic procedure, as aimed for in the project, requires that the specific proteins that were selected to be analysed also can be properly detected. In this project the strategy was to use affinity binders that bind specifically to the proteins.
Three different classes of affinity probes were used. Apart from conventional antibody molecules, available via commercial suppliers, affinity binders of two different classes were developed. At the biotechnology department of the Royal Institute of Technology, Stockholm (KTH), new affinity binders were developed based on an existing protein library technology platform denoted 'affibody binding proteins'. These affinity reagents, Affibodies, are characterised by a small sized scaffold, about 2 nm in diameter and consisting of 58 amino acids. Thirteen of these amino acids, making up a binding surface area of approximately 800 Å2, can be combinatorially randomised. Since there are 20 different amino acids in the proteins of humans, there are in principle different combinations for the set of 13 amino acids at the binding surface and thus considerable potential to optimise the affinity and selectivity of the Affibodies towards specific target proteins. At Academisch Ziekenhuis Leiden (LUMC), heavy chain antibodies (HCAbs), were developed for the project. HCAbs represent a class of antibodies specific for camels. The HCAb fragments responsible for the target binding are significantly smaller than conventional antibodies, represent the smallest naturally occurring protein-based affinity binders known to date and are frequently used as binders within biomedical research, diagnosis and therapy.
To provide the necessary affinity binders to the project, an inventory of available affinity reagents to the selected protein targets was performed and finalised and an affinity protein reagent bank was established for these targets. It should be borne in mind that even if there are one or several affinity binders reported to be specifically binding to a certain protein with high affinity, both the binders and the targets are proteins and their conformations and their binding properties may change dramatically depending on the environmental conditions. Following identification of protein targets, available affinity reagents to these proteins therefore had to be further investigated with respect to their binding properties under conditions close to those expected in the FNA sampled cells. In these investigations, the reagents were either directly or indirectly labelled. The investigations of the binders were performed by conventional fluorescence-based confocal microscopy on cells from three selected cell lines representing normal cells and two different aggressive forms of breast cancer. As a result of these investigations, a list of affinity binders to the prioritised list of protein targets could be identified.
In parallel to these efforts, work was performed both at LUMC and KTH towards development of novel affibody and HCAb fragment binders to an interesting breast cancer marker candidate, the protein SATB1. This work resulted in the identification and validation of several new affibody and HCAb fragment variants. These were produced recombinantly, subsequently labelled by various principles and found to selectively recognize the SATB1 target in cell samples. Data from binding studies suggest that different epitopes are preferred by the two classes, opening up for a possible increased specificity in obtained localisation signals. The new binders were used either separately or in combination to evaluate the presence and subcellular localisation of the SATB1 target in different cell types. Although SATB1 was eventually not selected among the proteins showing the largest diagnostic potential in the diagnostic procedures of the project, the possible alternative use of the developed binders for detection of targets present in cell-lysates was demonstrated and shows the versatility and value of having such selective binders to targets of interest.
Taken together, a set of affinity binders could be identified for the selected target proteins, well functional also under the conditions relevant for FNA sampled cells. The project has further shown that novel technologies for development of novel affinity proteins of different nature have the potential to be very valuable for future diagnostic applications.
Selection and development of fluorophores
Both high-resolution microscopy (nanoscopy) and high-sensitivity, multi-parameter, fluorescence-based characterisation of proteins in cells critically depends on the availability of suitable fluorophores, carrying specific chemical groups so that they can be coupled to relevant affinity proteins or to other molecules of interest. Besides spectral properties suited to the optical instrumentation, the most important properties of the label are high fluorescence quantum yield and high photostability. The University of Siegen (UNISI) has a vast experience in the area of fluorophore development and was the main partner responsible for the fluorophore selection and optimisation.
Commercially available fluorescent labels were evaluated with respect to their fluorescence properties and tested more in detail by the various spectroscopic and imaging techniques in the project. Apart from fluorescence quantum yield and photostability, a critical aspect investigated was the stability and reliability of the fluorescence parameters recorded from the fluorophores. Based on these investigations, a set of fluorophore labels could be identified as appropriate for the project and were subsequently used in the developed imaging procedures.
In addition to testing and selection of commercially available fluorophores for the project, new chromophores were also synthesised. In particular, new hydrophilic and water-soluble fluorophores were designed and synthesised, specifically suitable for the labelling of affinity molecules such as antibodies and affibodies. The newly synthesised fluorophores displayed minimised dye aggregation and very low unspecific binding to cell-based specimens, yet exceptional properties in terms of photostability and brightness, as required for the ultrasensitive and high-resolution microscopy techniques in the project.
Taken together, highly photostable and bright fluorophores are indispensable for successful fluorescence spectroscopy/imaging in general and in particular for the cutting-edge fluorescence-based methodologies developed in this project.
Development and optimisation of high-end fluorescence methods for diagnostics of FNA-sampled cells
In the project, several fluorescence-based methods were developed and finally evaluated for use in a procedure to diagnostically classify minute amounts of FNA-sampled material. The strategy was to provide a combined procedure, with the complementary information obtainable from super-resolution imaging on the one side and new, extremely sensitive and specific fluorescence-based optical imaging techniques on the other, forming a sound basis for an extensive analysis of FNA aspirates and a final highly reliable diagnostic classification.
Along this strategy, four different fluorescence-based methods were developed and modified towards diagnostic use in this context.
1. STED microscopy: In this project, we have set as a goal to introduce ultrahigh-resolution microscopy for diagnostic purposes and the further development and adaptation of STED microscopy, or nanoscopy, for diagnostic evaluation of FNA-sampled cells has been a major activity. A STED microscope design based on a novel supercontinuum light source was implemented at MPIBPC during the first year of the project and at KTH during the second year. Following its publication, the microscope has received considerable attention and has now been duplicated in several labs throughout Europe. The experience gained in the construction of the supercontinuum based STED system has emphasised the key role of the light source for the further improvement and dissemination of STED microscopy. The developed STED instruments provide all the necessary functionality, including dual-colour ultrahigh resolution capabilities, clearly representing two of the most significant readout parameters for our diagnostic procedure. In terms of robustness and usability they are close to commercial STED instruments currently on the market, but offer higher resolution and true dual-colour capabilities. Moreover, the cost of the instruments is almost an order of magnitude lower than that of current commercial instruments. These instruments may thus serve as a blueprint for a future commercial instrument. Following establishment of the STED instruments, proof-of-concept studies on cultured cancer cells were performed.
2. TRAST imaging: The information content of the nanoscopy images can be complemented by also encoding spectroscopic information. In this project, a concept for TRAST microscopy developed at KTH and opening for the possibility to monitor transient photo-induced non-fluorescent states of fluorescent molecules on a massive parallel scale and at arbitrary concentrations, was further developed towards diagnostic applications. This concept can add considerable sensitivity to existing fluorescence microscopic readouts, providing also an additional dimension of fluorescence information that has previously been almost fully overlooked in fluorescence microscopy. The TRAST approach was demonstrated as a tool to image and spatially resolve the metabolism within individual live cells and to detect differences in metabolic activity typically occurring in cancer cells.
3. MFDi: In order to be able to extract sufficient diagnostic information out of sparse, or even individual FNA-sampled cells, it is necessary not only to monitor species of one target protein in the cells, but of several different proteins at the same time. HHU has developed a new general strategy based on MFD to register and quantitatively analyse fluorescence images. In MFDi pulsed excitation with typically at least two different lasers is used for generation of fluorescence in the sample. Following optimisation of instrument hardware, software as well as sample preparation protocols, final MFDi measurement routines could be established on cultured cells, allowing detection and identification of multiple significant molecular or cytohistological features in the cells. This formed the basis for the subsequent application of MFDi on multiply stained FNA patient samples. This also demonstrates that the developed acquisition and analysis procedures match the prerequisites for further adoption of the MFDi technique as a common diagnostic tool of cellular samples.
4. TPX technology for ultrasensitive detection and identification of pathological protein profiles from cytological aspiration material: The TPX technology, developed at the Laboratory of Biophysics (UTURKU), is extremely well suited to measure concentrations of target molecules from minute samples with high precision and sensitivity. The strategy was to develop the TPX technology to monitor the prescence of nonlocalised proteins and their concentrations in extracellular aspiration material, providing complementary information to that obtained from the fluorescence-based imaging methods outlined above. Development within the project has demonstrated that the assay should be readily integrable to the FNA sampling protocol of the project. A small aliquote of the original sample could be set aside for TPX analysis, providing complementary diagnostic information to that obtained from the cellular imaging.
Combined diagnostic procedure performed on FNA-sampled material
In the instrument development, the different modalities were developed with the consideration to be used either in a parallel, simultaneous approach within one integrated instrument, or the combination of techniques would be achieved by a sequential use of the techniques on the FNA samples.
Several setups were established in the project enabling a parallel use of techniques. It was realised that although the FNA samples gave only a limited amount of diagnostic material, they still contained a sufficient number of cells to be divided on several glass slides. Given this possibility, it was concluded that the diagnostically most significant combination of accessible parameters could be obtained by performing different imaging modalities separately and sequentially. Thereby, the conditions for each of the measurement techniques can be optimised independent of each other and one can avoid the compromises that follow from a combined imaging with two or more of the modalities, on the same cells and at the same time.
A sample handling and preparation protocol was defined taking into account that FNA samples are far more complex and heterogeneous than standardised cells from cultured cell lines and may contain blood cells, fatty tissue and other contaminations. Extensive effort was required to yield optimal fluorescence characteristics and to ensure that the cellular handling does not compromise any of the most significant readout parameters of the different procedures used to image the cells. The optimisation of the sample handling and preparation had to incorporate many aspects, including proper specificity of several affinity molecules against their molecular targets, fluorescence intensity versus background, stability, reliability and reproducibility of the recorded parameters, maintenance of cellular morphology and specific cellular features, proper adhesion of cells to glass cover slides and preparation formats compatible with transport, as well as sufficient ease of preparation at clinical sites with typically limited resources and infrastructure for cell handling. Taken all these aspects together favoured a procedure including fixation of the sampled cells for stability and reproducibility reasons. The advantages of using fixed samples more than well compensated for the additional information potentially retrievable from unfixed, or even live, cells. To include fixation in the preparation of the FNA samples essentially also defined STED and MFDi as the combination of techniques most suitable for the diagnostic characterisation of the FNA-sampled cells.
With the FNA sample handling and preparation properly optimised, with a combination of MFDi and Nanoscopy identified as the diagnostically most useful combination of imaging methods providing uniquely retrievable diagnostic information and finally with a set of parameters and targets identified from the initial studies by STED and MFDi on cultured cells and then on cells from a few FNA samples as likely targets to provide high diagnostic power, tests were finally performed on a larger set of clinical FNA samples. Clinical FNA sampling from patients with suspect cancer lesions was performed in Stockholm and a few times also in Uppsala. Samples were obtained from suspect cancer lesions both from the breast and from the prostate, but with the major part of the samples obtained from the breast. After reserving a major part from each FNA sample for the routine diagnostic assessments, the remaining part of the samples were directly taken for preparation, distributed on up to 12 different glass slides and then passed on for imaging, locally to KTH and to HHU and MPIBPC. From the FNA samples, high-resolution images and spectroscopic data were acquired at KTH, HHU and MPIBPC at four different platforms using the different and complementary techniques.
Taken together, samples from 97 patients were altogether sampled at the different imaging sites as described above. Several features indicative of malignant development could be observed in both the MFDi and STED images already by some visual inspection. However, for a more quantitative analysis all image data from the FNA samples were sent to partner UH for bioinformatic analysis and verification.
Bioinformatic tools for diagnostic evaluation and final clinical verification
The modern imaging technologies developed and used in this project are capable of measuring intracellular molecules with a spatial resolution unimaginable by most researchers in the field 10 years ago. The high-resolution and high-content of the images requires advanced computational methodologies (bioinformatics) to translate the image data into knowledge of clinical benefit. In the project, novel and accurate microscope technologies were developed and applied to FNA-sampled cells from suspect breast and prostate cancer lesions. The major challenge at this point was the development and implementation of computational methods capable of pre-processing and analysing the image data systematically and fast. Before this project there were no published computational methods for pre-processing and analysing STED and MFDi images at the scale required in clinical settings.
The computational analysis of image data and integration of the extracted information from the images was done by Systems Biology Laboratory at University of Helsinki. Firstly, this project resulted in 1 486 STED and MFDi images from 63 patients. By using a collection of microfeature detection methods, we were able to detect hundreds to thousands of feature locations from a single image. Each of these locations, in turn, provided several quantitative values that were used in a prediction step. In order to be able to analyze such a large amount of data, we used a computational framework called Anduril (Freely available at http://www.anduril.org online. Analysis, scripts and results can be found at http://csbl.fimm.fi/pub/FLUODIAMON online) that allowed us to perform all mathematical operations in a systematic fashion. After developing the management system for systematic use of the images, we developed novel pre-processing methods for MFDi and STED images.
The main objective of the bioinformatics analysis was to identify markers that predict whether the cells obtained with FNA were normal or cancerous cells. The development of computational methods for cutting edge, high-resolution and multiparameter microscopy has resulted in freely available software for analysing large amounts of images in a systematised fashion, novel methods that are widely usable in various experimental settings and microscope techniques and clinically important predictors that can aid in breast cancer patient diagnosis.
Overall diagnostic procedure and its validation
Combining the efforts and progress within the different areas as outlined above, an overall diagnostic procedure could be established in line with the initial intentions. An intact chain of accomplished steps, from reliable and patient-friendly FNA sampling with minimised risk of tumour cell dissemination, over robust and reproducible sample handling and preparation, specific targeting of selected proteins with fluorophore-labelled affinity molecules, characterisation of the FNA sampled cells by optical methods offering utmost sensitivity, specificity and spatial resolution and finally bioinformatic analyses to fish out the differences in the image features necessary to distinguish normal from malignant cells, has lead to the demonstration of a novel diagnostic concept. This concept can combine diagnostic reliability with safe, minimally invasive diagnostic sampling and thus enable a decisive improvement of the outcome of patients suffering from breast and prostate cancer.
For validation, the data from the imaging configurations for the different patient FNA samples were compared to the diagnoses set for these patients clinically (collected and compiled by KI). The comparison showed that we managed to create six independent classifiers capable to detect cancer from the different image data in minimally 55 % of the cases and maximally in 80 % of the cases. It can thus be concluded that each of the six independent diagnostic classifiers can yield significant predictive strength. As a next step, in order to further improve the classification accuracy, several classifiers can be combined. Our early results confirm that a clear improvement in accuracy can be obtained, though this need be tested more thoroughly with more patient samples in the near future.
Complementary clinical validation was also focussed on the collection, preparation and analysis of clinical material of formalin-fixed and paraffin-embedded (FFPE) benign and malignant breast cancer specimens as well as of normal adjacent breast tissue. Together with experienced histopathologists, these specimens were evaluated and the most representative areas were used for generating a tissue microarray (TMA) that allows for standardised high-throughput immunohistochemistry analysis and marker validation.
Taking these validation paths together, the developed classifiers for the images from the patient FNA samples have been directly validated against the set clinical diagnoses for the same patients and the relevance of the selected markers have been verified on a large number of relevant patient material. The diagnostic procedures have been validated and shown to provide an attractive strategy for future clinical diagnostic schemes.
Potential impact
The FLUODIAMON project is highly interdisciplinary. The expertise needed for this project comprises clinical cytology, cancer proteomics, molecular biotechnology, fluorophore chemistry, fluorescence microscopy, nanotechnology, optics, solid state detector technology, data processing and bioinformatics. The level and width of competence required cannot be found on a national level but requires a European initiative. In particular, the project has demonstrated that by applying analyses based on novel state of the art optical microscopy techniques, even quite small amounts of diagnostic material is sufficient for reliable diagnoses of breast and prostate cancer. Thereby, the amount of diagnostic material obtainable via minimally invasive FNA is sufficient and more invasive core-needle or surgical biopsies can be avoided.
FNA techniques
The FLUODIAMON has supported three major advancements in FNA sampling technology. Firstly, a new FNA needle has been developed with a higher sample yield and improved surface treatment for higher ultrasound visibility. Secondly, by adding mechanical energy to the needle during the sampling the penetration of the needle has been considerably improved. Thirdly, an anti-seeding procedure has been established and clinically tested.
These advancements are likely to have a large impact on future diagnostics. The new FNA sampling technology contribute with considerable improvements on several of the most important aspects of diagnostic sampling: increased amount of sample material with the same degree of invasiveness, higher precision of positioning and penetrability is likely to increase the fraction of obtained samples that indeed are from the region of interest and not the least the risk of tumour cell dissemination is strongly reduced. These features make the FNA sampling technology very attractive for breast and prostate cancer diagnostics. However, it is also worth pointing out that the needle development can also to a large part be applied on core-needles as well. For the case core-needle would still be needed, mechanical oscillations and anti-seeding procedures can strongly improve the features in a similar way also for these larger diameter needles. The developed needle technology is also not only limited to breast and prostate cancer diagnostics. There is a range of different forms of cancer, where invasive sampling is to be avoided, but where FNA harnessed with the new technology can be considered as a possible way for diagnostic sampling. In general, the needle development within this project can have a considerable impact on all forms of diagnostic sampling, of cancer or infectious diseases, where risk of spread, accessibility, or risk of damage of vital tissues is an issue. Finally, the technology for diagnostic needle sampling also seems to have implications for treatment.
Affinity molecule development
In the FLUODIAMON project, a set of affinity binders were identified for a selected group of target proteins, well functional also under the conditions relevant for FNA sampled cells. The project has shown that novel technologies for development of novel affinity proteins of different nature can be very valuable for future diagnostic applications. Compared to conventional, full-sized antibody binding reagents, the alternative binder protein classes used can provide advantages in terms of size, the preference to and selectivity of where on the target protein they bind. The two groups of affinity binders developed in the project are considerably smaller than conventional antibodies. The small size of the marker molecules make them very valuable reagents for the optical techniques, where the high resolution offered may even make the actual size of the binder an issue. The small size may also allow binding to sites on the proteins not accessible by larger binders due to steric reasons. Since at least affibodies can be designed such that the position of the fluorophore marker molecule(s) can be specified to desired locations at the affibodies, also the fluorescence properties can be modulated and optimised via the design of the affibodies. This controlled fluorescent labelling, including both number and position of fluorophores, was taken benefit of in the project, where the fluorescence properties of the samples are very important for the overall performance of the high-end fluorescence imaging methods used.
Taken together, the affinity molecule development in the project has demonstrated the feasibility of using novel class of affinity molecules for high-end optical imaging and diagnostics and that the specific properties of the affinity binders can offer specific advantages compared to conventional antibody binders. The combined use of high-end fluorescence microscopy and these novel groups of affinity binders will have applicability for a wide range of cellular imaging, extending well beyond breast and prostate diagnostics.
Fluorophore development
In the FLUODIAMON project the access to high-performance fluorophores has been critical to the development of the diagnostic imaging procedures. The demands on the fluorophore marker molecules are for both high-resolution microscopy and high-sensitivity, multi-parameter, fluorescence-based characterisation of proteins considerably higher than in more conventional fluorescence imaging and spectroscopy. Besides carrying specific chemical groups so that the fluorophores can be coupled to relevant affinity proteins or to other molecules of interest and spectral properties suited to the optical instrumentation, the most important properties of the labels are high fluorescence quantum yield and high photostability. Our project has demonstrated the feasibility and strength of using high-end fluorescence imaging methods for breast and prostate cancer diagnostics. However, the applicability of these imaging methods goes far beyond this particular application. No matter the application, the performance rely on the use of appropriate fluorophores and the developed fluorophores in this project and the experience coming from the development and use of these, will be highly useful in a range of applications in diagnostics, cell biology and biosciences in general.
Cancer proteomics
In a sense, the project merged knowledge about tumour-specific markers with the use of highly sensitive, high resolution imaging methods by which the markers can be detected and identified down to a single molecule level and by which their molecular status can be analysed in the context of individual sample cells. In particular, the project has shown for a limited number of selected protein targets that apart from plain deviations in the amount of the proteins in the FNA sampled cells, the diagnostic strength can be considerably increased by further extending the protein characterisation to include also high-resolution spatial distribution and high-sensitivity expression level profiling of several proteins in the individual cells at the same time. The impact this project will have in the context of protein biomarkers thus essentially lies in the notion that not only expression levels of certain individual proteins matters. Our results suggest new aspects to consider when identifying new tumour markers in the future.
Development and optimisation of high-end fluorescence methods for diagnostics of FNA-sampled cells
For the diagnosis of breast and prostate cancer today's standard procedures for diagnostic sampling are core needle or surgical biopsy. However, due to their mechanical crudeness, these procedures are highly traumatic and encumbered with a significant risk for tumour cell seeding in the needle tracts. In contrast, FNA offers a minimally invasive, extremely patient-friendly extraction of cellular material. Four different methodologies were developed in the project and although only two of those were combined in the final characterisation of the FNA sampled cells, their results will have an impact on future cellular imaging and diagnostics.
1. STED microscopy: The development of the STED technique within the project has lead to the first demonstration of its use for cancer diagnostics, but also paves the way for future applications of ultrahigh resolution microscopy in cell biology and biomedicine in general.
2. TRAST imaging: In the project, several realisations of the technique in various setups have shown its broad applicability and the ability to retrieve information coupled to low frequency molecular encounters or metabolic states have been demonstrated. Since the metabolic state and the regulation of the metabolism is one of the first things to be influenced in cells upon disease and toxic influence, the ability to monitor the metabolic state and the metabolic regulation of cells is also likely to open for applications in toxicology, drug tests and several other applications in fundamental cell biology.
3. MFDi: In this project, a new general strategy based on MFD to register and quantitatively analyse fluorescence images was further developed for imaging purposes (MFDi). The developed MFDi technology has been demonstrated as a useful diagnostic imaging method and a similar use of MFDi for subcellular diagnostics can most likely be extended to a broader range of diseases in the future. MFDi could be broadly used in cell biology in general. Parts of the hardware and the concepts for fluorescence data acquisition and processing included in MFDi can also be used outside of the context of MFDi. Partner B&H has developed and adapted several instrument components as well as concepts of this kind for commercial use in this project.
4. TPX technology: The TPX-technology, originally developed by the UTURKU group and commercialised by Arctic Diagnostics Ltd is extremely well suited to measure concentrations of target molecules from minute samples with high precision and sensitivity. The assay development within the project has demonstrated that the assay should be readily integrable to the FNA sampling protocol of the project. In addition, the assay development within the project has also set the ground for developments of the TPX technology towards a new method for detection of MRSA and the development of TPX assay for bacterial counting have further paved the way towards quantitative and sensitive TPX assays of cellular matrices.
Bioinformatics
Bioinformatic tools were developed to analyze the large number of images taken by STED and MFDi on FNA-sampled cells from suspect breast and prostate cancer lesions. The high-resolution and high-content of the images required advanced computational methodologies (bioinformatics) to translate the image data into clinically useful information. Before this project there were no published computational methods for pre-processing and analysing STED and MFDi images at the scale required in clinical settings.
With the goal to identify markers that predict whether the imaged FNA-sampled cells were normal or cancerous, partner UH developed computational methods based on strong prediction methods developed in the machine learning community. The efforts resulted in clear improvement in diagnosing patients with the FNA procedure. The development of computational methods for cutting edge, high-resolution and multiparameter microscopy has resulted in:
1. freely available software for analysing large amounts of images in a systematised fashion
2. novel methods that are widely usable in various experimental settings and microscope techniques and
3. clinically important predictors that can aid in breast cancer patient diagnosis.
In the project, in particular the study of microfeatures has emerged as a novel powerful approach in the analysis of images, opening new possibilities in all studies, where the resolution is large enough for imaging subcellular features. The outcome of the bioinformatic analysis demonstrates the power of multiple marker and/or ultrahigh resolution spatial distribution pattern analyses of proteins on a subcellular level as a new path for early, sensitive cancer detection, but the developed image analysis tools also have a more general applicability.
Overall diagnostic procedure
In this project, we have developed and successfully demonstrated a joint methodology for objective, quantitative sub-cellular diagnosis of early breast and prostate cancer disease. By applying analyses based on novel state of the art optical microscopy techniques much more information can be extracted and even quite small amounts of diagnostic material is sufficient for reliable diagnoses of breast and prostate cancer. Thereby, the amount of diagnostic material obtainable via minimally invasive FNA is sufficient and more invasive core-needle or surgical biopsies can be avoided. The potential impact following from this successful development of the methodology cannot be overestimated:
1. it can strongly contribute to an increased cure and survival for breast and prostate cancer patients
2. it should be able to provide valuable support guiding decisions whether therapeutic interventions are to be undertaken and if so, when and to what extent
3. it should minimise the risk of cancer cell seeding, patient discomfort and other side effects such as pain and infections
Outside of breast and prostate cancer diagnostics and progression monitoring, the methodology should be applicable also to other diagnostic fields facing the same principal conflict between minimal invasiveness and the information content needed for a clear diagnosis.
As a major, overall result of this project, we have obtained a diagnostic methodology for breast and prostate cancer that combines diagnostic reliability with minimal invasiveness and a negligible risk for sampling-related side effects, in particular cancer cell seeding and spread. This combination is impossible to obtain with other presently available methodologies. The ability to obtain conclusive information with minimal invasiveness and without risk of cancer cell seeding and spread also enables safe and reliable repetitive sampling and thereby quantitative monitoring, analysis and prediction of disease progression over time, e. g. for follow-up of pre-malignant lesions or for evaluation of response to therapy. The developed methodology can therefore also be expected to form the basis for important diagnostic tools supporting and guiding possible therapeutic interventions. It is also reasonable to predict that the proposed methodology would offer significantly improved and profoundly new possibilities for monitoring and analysing, on a subcellular level, molecular mechanisms for tumour development and effects of various therapies. The project results thus clearly can be expected to contribute to the development of new therapies and pharmaceuticals and in particular therapies based on molecular and cellular mechanisms.
In the forefront of the potential impact that this project may bring about is the fact that breast and prostate cancer belong to the most frequently occurring malignancies in women and men, respectively, in the western world. When possible, detection of these diseases at the earliest stage results in a 15 year survival rate of more than 90 %. In contrast, survival of breast cancer patients exhibiting advanced stages has not improved during the last 20 years. Thus, for improving survival of patients suffering from breast and prostate cancer, the most important task is to detect, diagnose and treat malignancies as early as possible. The benefit to patients that patient-friendly and secure sampling methods for breast and prostate cancer would bring about, offering the necessary diagnostic sensitivity, specificity and reliability, is enormous.
Furthermore, this project improves one of the most important quality aspects in the diagnostics and treatment of patients with breast and prostate cancer - that sample extraction may cause severe side effects and even worsen the outcome. At this stage, we have developed, demonstrated and clinically verified the methodology. Already at this stage, the procedure is objective and with additional refinements of our combined procedure it is likely that the diagnostic sensitivity and specificity can be even improved compared to today's standards.
Dissemination activities and exploitation of results
The activities within FLUODIAMON have throughout the project generated a continuous flow of dissemination activities.
In total FLUODIAMON have to-date resulted in 24 peer-reviewed articles, all published in well recognised scientific journals. Ten additional publications are either already submitted or soon to be submitted. Moreover, seven PhD theses and four MSc theses have to date been achieved with support from the project. In addition, several PhD students will graduate in the next one to two years and their PhD theses are also to a large extent a fruit of and supported by the project.
Among dissemination activities one can also mention several open access software packages for classification and image analysis, interviews in newspapers and TV, exhibitions and well over 50 invited oral presentations and posters at international conferences and workshops.
On the commercial side, 10 specific patents or patent applications, related to needle sampling technology, fluorophores and detector and imaging technology, have been either generated or have directly profited from the activities and results within the project. A corresponding development of several different commercial products within these areas has also commenced as a result of the project.
In conjunction to the last consortium meeting, a workshop was arranged in Stockholm on 2 December 2011 ('Ultra-high resolution and ultra-sensitive fluorescence methods for objective sub-cellular diagnosis of early disease and disease progression in breast and prostate cancer'). This workshop was very well attended and brought researchers from the full range of disciplines covered within the FLUODIAMON project. Finally, last but not least, the progress within the project now allows us to disseminate the results also to a wider audience. From a patient perspective, considerable benefits have been demonstrated. A first seminar with breast cancer patients has been arranged by partner KI to disseminate these results. Given the positive and encouraging results from the large-scale assessment and clinical verification, a larger meeting with breast cancer patients and clinicians is already planned together with the major patient organisations in Sweden during the autumn 2012 and similar activities will also be planned in Germany.
Taken together, the FLUODIAMON projects has yielded a very broad range of results, which has been disseminated scientifically, on the commercial side and not the least due to the large impact on future cancer diagnostics, towards patient organisations and clinicians.
Project website: http://www.biomolphysics.kth.se/fluodiamon
Contact details of the coordinator:
Prof. Jerker Widengren
KTH, Dept Applied Physics / Exp Biomol Physics
Albanova Univ Center, 106 91 Stockholm
Telephone: +46-855-378030
Fax: +46-855-378216
Email: jwideng@kth.se