Final Report Summary - BIOIMAGE-NMD (BIOIMAGE-Neuromuscular Diseases)
Executive Summary:
The overarching objective of the BIOIMAGE-NMD project was to deliver combined structural and molecular imaging biomarkers with proven utility for the detection of therapeutic effects in patients with rare neuromuscular diseases (NMD). The project addressed three major themes, divided across 8 academic Work Packages:
1) The first theme conducted a full translational development of diffusion MRI methodology optimised for muscle microstructure as a new putative biomarker for pathology in patients with NMD. Restricted self-diffusion behaviour of water within muscle tissue was observed in mouse muscle samples, indicating sensitivity of diffusion MRI to muscle microstructure. Studies used in vitro and in vivo scanning in wild type mice and the mdx model of DMD. Significant differences in diffusion were found between groups. Factors influencing the restriction of water diffusion were explored against histology in the same muscle samples and through development of a novel Monte-Carlo modelling platform for muscle architecture. While histological changes in muscle fibre size and size distribution were observed between wild type and dystrophic muscle, the over-riding factor defining water diffusion in muscle was determined to be the overall permeability of the system. Diffusion methods were translated to clinical scanners and new highly efficient methods to suppress lipid signal in the MRI scans were developed to deliver high quality, quantifiable scans in human subjects. Scan reproducibility was evaluated within and across sites in the Consortium. Finally, a first-in-man study in NMD was conducted in patients with Becker Muscular Dystrophy. Small, but significant differences were observed in patients.
2) The second theme aimed to provide a proof of principle that simultaneous MRI/MRSI can be used as a biomarker to monitor therapeutic efficacy in clinical trials in NMD, specifically in Duchenne Muscular Dystrophy (DMD). Clinical trials were led by our commercial partner BioMarin with 3 different antisense oligonucleotide (AON) treatments (named PRO051, PRO044 and PRO053, where the number indicates the exon within the DMD gene at which each is targeted), while additional patients were studied in a parallel natural history study. Quantitative imaging methods (Fat Fraction, T2 mapping, Phosphorus-31 MR spectroscopy) were incorporated into each study. Across these studies 75 patients underwent longitudinal imaging, obtaining a total of 220 scan datasets. The results across all studies confirmed that MRI derived fat fraction is a valid biomarker of disease severity and progression. Correlations were found between the loss of real muscle mass and functional outcome measures such as the 6-minute walk distance. In the largest patient subgroup (PRO051 treatment), a significant difference in rate of fat infiltration was observed between treated and non-treated patients, indicating effect of exon-skipping therapy. However, the diverse phenotypic expression of the disease as well as its non-linear progression on an individual basis are major confounds to establishing prognostic indices based on the imaging data. To enhance clinical trial capability, secure, web-based data transfer tools were developed for exchange of images between sites and the central analysis team. Novel analysis tools using semi-automated muscle segmentation were also created, speeding up image quantification by a factor of 7-17 times (equivalent to a time saving of ~6 hours per patient dataset).
3) The third theme developed a novel simultaneous Positron Emission Tomography (PET)/MRI technology directed towards advancing innovative drug development programmes for personalised medicines based on AON technology. We developed synthetic protocols to prepare [19F] labelled oligonucleotides as a first step to production of [18F] radiolabelled drugs for PET imaging. Labelled [19F]-AON could be produced in high milligram quantities with high purity (97%) allowing in vivo assessment of bioequivalence of the original and labelled AON compounds. The [19F]-AON or native AON were administered into mdx mice and ELISA analysis of tissue and plasma levels showed equivalent biodistribution patterns for labelled and unlabelled compounds, demonstrating that incorporation of the fluorine label did not alter the in vivo AON pharmacokinetic profile. Radiolabelling of AON with the positron emitting isotope [18F] was achieved on an automated radiosynthesis platform with reaction conditions and purification methods optimised to achieve the necessary activity/volume ratios using GMP compatible protocols. The first biodistribution studies using PET imaging of [18F]-AON were carried out in mdx and wild type mice using PET-MRI and demonstrated the principle of detecting this new class of fluorine-18 radiolabelled AON’s in vivo.
The BIOIMAGE-NMD project delivered significant advances and new technologies across all 3 themes, increasing our understanding of the role, benefits and challenges of imaging as outcome measures for clinical trials.
The BIOIMAGE-NMD consortium consisted of 2 SME companies, BioMarin and Consultants for Research in Imaging and Spectroscopy (CRIS) and 6 leading academic institutes, Newcastle University, the Leiden University Medical Centre, the Institute of Myology (Paris), University College London, Katholieke Universteit Leuven and Universita Cattolica Del Sacro Cuore (Rome) and was coordinated by Newcastle University.
Project Context and Objectives:
Duchenne Muscular Dystrophy (DMD) is a relatively rare but lethal genetic childhood disease which affects between 0.32 and 0.52 per 10,000 inhabitants. Worldwide, around 240,000 boys suffer from DMD. DMD is caused by mutations, often deletions, in the DMD gene that result in the disruption of the open reading frame leading to the loss of a loss of dystrophin protein expression. Dystrophin has an important role as a scaffolding within muscle fibres where it links the actin cytoskeleton with the extracellular matrix and absorbs force generated during muscle fibre contraction. Muscle fibres lacking dystrophin are damaged during normal activity and are unable to regenerate healthy tissue, eventually leading to progressive muscle weakness and wasting. While DMD is clinically well characterised it lacks any drug treatment which substantially alleviates the disease progression. One of the most promising emerging treatments for NMD, including DMD, is the use of exon skipping drugs, including Antisense Oligonucleotides (AON). AONs are a genuine example of personalised medicine, where subsets of patients are treated according to their specific genetic mutation. Despite the clear potential for AON technology, the ongoing clinical development of treatments for this type of rare NMD are impeded by 1) the low number of patients available with any given mutation, 2) the lack of objective and reliable outcome measures which span all phases of the disease, and 3) the current need to use invasive muscle biopsies to monitor therapeutic response.
Imaging techniques such as magnetic resonance imaging (MRI) and positron emission tomography (PET) offer the opportunity to non-invasively quantify many aspects of human tissue and may be useful as biomarkers for disease progression or outcome measures in clinical trials. In recent years it has been suggested that the development of clinical trial imaging biomarkers should be an integral part of the drug development pathway, such that imaging methods proven to determine whether a drug is reaching the desired tissue and having the expected therapeutic impact in vivo, are developed in parallel with the drug itself. This integrated approach would ensure that the necessary tools to quantify patient response are already available when a new compound enters clinical trial. This is particularly important in patients with rare diseases where large patient cohorts are not available to achieve statistical power using clinical endpoints alone. However, while this synergistic drug-biomarker development approach is well recognised, it has not been achieved for many of the treatments currently entering clinical trials.
In the context of muscle imaging biomarker development some methods are now ready to be tested in clinical trials. Additional techniques (such as diffusion MRI) may offer further value but need to be fully developed and validated before they can be considered for inclusion in forthcoming trials. An effective imaging biomarker must be objective, sensitive, specific and non-invasive. We suggest that no single imaging method will meet all of these criteria, but that a quantitative multi-measurement approach will provide a more complete and sensitive evaluation of muscle damage and repair. The BIOIMAGE-NMD programme therefore aimed todevelop a combined examination based on macro- and micro-structural MRI, the identification of molecular changes in tissue metabolism (through MR spectroscopic imaging, MRSI) and direct visualisation of the AON molecules in vivo using positron emission tomography (PET). The individual value of each measure was assessed in pre-clinical models and, where possible (given the relatively short timescale of the project), will be applied in patients to establish a proof of principle for clinical utility. These approaches can then be fused to form a truly integrated, simultaneous PET-MRI-MRSI protocol, for future use in patients with NMD.
The overarching objective of the BIOIMAGE-NMD project is to deliver combined structural and molecular imaging biomarkers with proven utility for the detection of therapeutic effects in patients with rare neuromuscular diseases (NMD). To achieve this goal, the programme addressed three specific objectives:
Objective 1: To develop a new generation of diffusion MRI methodology optimised for imaging in muscle.
BIOIMAGE-NMD sought to develop new MR based quantitative biomarkers of muscle fibre microstructure using diffusion contrast MRI. In these scans, microstructural sensitivity is conferred through diffusion-mediated interaction between the tissue water, the cell membranes and sub-cellular structures. Tissue injury or degeneration alters the restriction to diffusive water movement and diffusion weighted imaging (DWI) or diffusion tensor imaging (DTI) measurements can reveal and quantify this process, even though conventional radiological T1 or T2 weighted MRI may remain unchanged. DTI technology is used extensively in the brain and current attempts to apply diffusion imaging methods in muscle have largely been adapted from neuroimaging protocols, but have not been comprehensively optimised to account for the differences in cellular size between axons in the brain and muscle fibres or the presence of fatty tissue. Extensions to the DTI method exploiting diffusion mediated contrast have the potential to quantify muscle fibre sizes and detect changes induced by a drug treatment. Objective 1 was to develop these methods as additional new biomarkers of muscle damage. The methods would then be used to augment our state of the art simultaneous MRI / Magnetic Resonance Spectroscopic Imaging (MRSI) protocol for quantitative muscle imaging with the ultimate goal of applying these methods in clinical AON trials in DMD.
Objective 2: To provide a proof of principle that simultaneous MRI/MRSI can be used as a biomarker to monitor therapeutic efficacy in clinical trials in Duchenne Muscular Dystrophy (DMD).
Use of state of the art muscle MRI technologies was proposed to provide a simultaneous structural and molecular approach to follow AON treatment in clinical trials which are already underway or are being planned. MRI is widely available and offers powerful flexibility for non-invasive structural, physiological and molecular investigation. Skeletal muscle fibrosis and fat infiltration are common early features of most NMD due to the continued destruction-regeneration processes. In routine use MRI clearly demonstrates the fat infiltration but visual inspection of these qualitative images is not sensitive to the therapeutic effect of a drug between successive examinations. This may, however, be achieved in humans using quantitative MRI techniques that separate muscle fat and water components, allowing the computation of fat signal percentage maps and identifying inflammatory activity.
Molecular changes are also critical for the comprehensive evaluation of muscle damage. Phosphorus-31 MR spectroscopic imaging (31P-MRSI) is an endogenous molecular imaging technique which shows altered muscle energetics and muscle membrane stability in human dystrophic muscle. 31P-MRSI is a specialist technique but requires only an enhancement of the same scanner technology as used by MRI and so can be effectively combined with the quantitative imaging methods. Simultaneous use of quantitative MRI and MRSI therefore provides a clinically useful structural and metabolic assessment of muscle in patients with NMD, but these techniques have yet to be demonstrated as an effective tool to assess therapeutic intervention.
Objective 2 was therefore to establish a simultaneous MRI/MRSI protocol across the sites of the consortium to deliver a clinical proof of principle (PoP) demonstrating the objective benefit of these combined technologies based on their use in clinical trials of AON in patients with DMD. Objective 2 is closely linked to the outcomes of objective 1 whereby novel diffusion imaging based biomarker technology developed in the first theme would be incorporated within the clinical study as key milestones are reached.
Objective 3: To develop PET molecular imaging of AON in vivo as a technology to advance future innovative drug development programmes for personalised medicines.
We proposed new synthetic chemistry approaches for radio-labelling of the AON molecules themselves enabling PET molecular imaging. While instrumentation for simultaneous structural and molecular imaging by PET-MRI is now commercially available, PET imaging is fundamentally reliant on developing disease specific radiolabelled reporter ligands and, to date, no attempts have been made to develop such radiotracers targeted to disease activity or therapy in NMD. Objective 3 of BIOIMAGE-NMD aimed to make essential advances in the expansion of this technology by developing AON PET tracers to measure bio-distribution in vivo in animal models. Such PET/MRI tools would allow rapid assessment of tissue targeting, biodistribution and pharmacokinetics of AON in vivo, helping to accelerate the development of AON technology for DMD and broadening the applicability of AON technology in other neuromuscular and genetic diseases. In addition, as this technology is translated to humans (beyond BIOIMAGE-NMD) it will form a powerful tool to help define and explain the optimal treatment regimes, mechanism of action and drug delivery for this personalised medicine in humans.
The BIOIMAGE-NMD programme was divided into 8 scientific work packages to achieve these objectives, supported by 2 further work packages around dissemination and management/coordination. Individual objective for each work package were as follows:
WP1 - To implement, evaluate and optimise diffusion weighting MRI schemes to maximise sensitivity to pathological change in muscle and develop appropriate imaging analysis models to quantify changes in muscle fibre dimensions and distributions.
WP2 - To evaluate the sensitivity of muscle optimised diffusion imaging schemes to quantify muscle microstructure in mouse models for muscular dystrophy and define whether diffusion imaging derived biophysical metrics of muscle structure provide new detail in the assessment of NMD.
WP3 - To establish a robust conventional diffusion imaging sequence for use in skeletal muscle.
To develop faster scanning methods for T2 mapping, B1 mapping and Dixon imaging.
To translate and standardize optimised diffusion imaging schemes from pre-clinical to clinical protocols.
To determine whether conventional diffusion imaging provides added value in healthy controls over and above existing MRI methodology (Dixon, T2) and assess the effectiveness of the entire imaging protocol in Becker Muscular Dystrophy patients.
WP4 - To obtain MRI/MRSI data from patients recruited in 4 DMD clinical trials with different AONs and as a comparator group from patients recruited in a DMD natural history study, introducing improved imaging sequences and investigate novel contrasts from WP3 as they are validated.
WP5 - To evaluate the initial human diffusion imaging studies in comparison with the original MRI/MRSI protocol and determine whether diffusion metrics offer new biomarkers.
From this evaluation to decide whether the MRI/MRSI protocol should be updated to include a diffusion measurements in ongoing clinical trials.
WP6 - To centrally collect, curate and analyse the MR images and MRSI spectra generated by the multi-centre clinical trials in WP5 and subsequently identify the most relevant MRI and MRSI indices and their significance relative to other outcome measures.
To develop new processing algorithms for the treatment of the MR imaging data, that will reduce data processing time and improve the quality of the information extracted.
To develop a web-based software solution integrating data collection, quality control and processing dedicated to muscle analysis in multi-centric clinical trials.
WP7 – To develop a synthetic protocol for the preparation of a range of cold (19F)-labelled oligonucleotides, directly translatable for the production of 18F-radiolabelled drug product.
To determine the bioactivity of the cold labelled AONs in vitro and compare the bio-distribution and bioactivity of the cold labelled AON versus the unlabelled AON in mdx mice.
WP8 - To generate Proof of Principle for the use of in vivo PET imaging of radiolabelled AONs as a non-invasive tool to visualize the pharmacokinetics of AON after systemic administration in mdx mice.
To establish the correlation between AON biodistribution data from PET imaging and exon skipping efficacy in diverse skeletal muscle group in mdx mice.
To establish the correlation between AON biodistribution data from PET imaging with MRI based muscle structural data in mdx mice.
Project Results:
The BIOIMAGE-NMD consortium was formed to address significant challenges in evaluating new therapeutic treatments in clinical trials in rare neuromuscular diseases. Sensitive and validated outcome measures and biomarkers of drug action are required which can be effectively applied in small cohort groups, particularly in children. Imaging methodologies were already identified as important non-invasive tools, but further evaluation of existing techniques was required. Further, magnetic resonance based approaches were identified which had not been optimised for muscle pathology, but which offered exciting potential as novel outcome measures. The BIOIMAGE-NMD programme addressed the clinical trial potential of advanced and quantitative imaging through an ambitious and complete translational development of methodologies from in vitro and early pre-clinical testing, through implementation in the clinical environment and evaluation in healthy subjects, first-in-man observational studies in neuromuscular disease and ultimately adoption into clinical trials of exon-skipping in boys with Duchenne muscular dystrophy. This translational development theme ran alongside evaluation of existing imaging methodology which was incorporated into the trial protocols of the earliest AON clinical trials to assess proof of principle of imaging biomarkers. Finally, the programme sought to develop new molecular imaging methods through new synthetic chemistry approaches to radiolabel the AON molecules for imaging via positron emission tomography (PET).
To achieve these objectives BIOIMAGE-NMD assembled a Consortium combining SME expertise in drug development with academic expertise in experimental (MRI) and theoretical physics, imaging science, chemistry, neurophysiology and clinical neurology. The commercial focus for SMEs are significantly different to those in academia, with pressures derived from business performance. By their nature, SMEs are responsive to their commercial environment, and highly successful SMEs will always be at risk of take-over or merger by larger organisations. During the lifetime of the BIOIMAGE-NMD programme there were two significant events with our main SME partner, but through careful management of the work plan these did not impede progress. Early in the programme, the clinical trial (outside of BIOIMAGE-NMD) of the first AON from our SME partner Prosensa was halted due to lower than expected clinical efficacy. The clinical trials at which the imaging developments within BIOIMAGE-NMD were targeted were therefore delayed while the data from this first trial were analysed in detail. Subsequently, after the next phase of trials had commenced Prosensa was the subject of a corporate acquisition by BioMarin who took over the role within the programme, but after a further 12 months determined to halt all therapeutic development and clinical trials for neuromuscular disease groups. Through careful management and the ongoing support from BioMarin, these impact of these factors on the overall outcome of the BIOIMAGE-NMD programme was minimised.
The programme worked towards 10 deliverables, with an associated 46 milestones across the full set of 10 work packages and achieved the vast majority of these targets. Below we describe in detail the main results from the scientific Work Packages along with key Deliverables and Milestones.
No new foreground was generated during the lifetime of the project.
Work Packages – in detail
WP1 - Optimising diffusion imaging for detection of changes in muscle tissue microstructure
Summary of Achievements
Diffusion imaging methods for muscle MRI were developed based on the stimulated-echo (STE) preparation methodology and evaluated in the hindlimb muscles of wild type and mdx mice (a transgenic model for the clinical condition Duchenne Muscular Dystrophy). Comprehensive measurements were made in vivo, repeated in the same tissue post mortem and compared to histological findings in the same tissue. The STE sequence allowed flexible variation in the diffusion time, which is a key sequence parameter influencing the probability of interaction between diffusing water molecules and the local tissue microstructure. Diffusion times were studied between short (20-50ms, as normally used in brain investigations) and very long (750ms). The imaging data demonstrated the expected restricted diffusion behaviour, indicating that the sequence was sensitive to muscle microstructure. Highly significant differences in water diffusivity (apparent diffusion coefficient, ADC) were observed between the animal groups with the greatest differences found with much longer diffusion times than are typically used in clinical scanners.
Computational modelling was developed to investigate the sensitivity of diffusion MRI to key features of pathological muscle. Monte-Carlo simulations of diffusing spins in hierarchical structures were implemented, including explicit modelling of the effect of diffusion-attenuating pulse sequences. Muscle tissue was modelled as parallel cylinders with radii drawn from gamma distributions fitted to histological data of mouse forelimb muscle tissue derived from the literature and our preclinical studies. Each cylinder contains a population of random close-packed cylinders with smaller, but also gamma-distributed, radii as a model of hierarchical muscle tissue structure covering six orders of magnitude in size. This is the most detailed models of muscle tissue simulated in the diffusion MRI literature. Simulations were then used to synthesise diffusion measurements over a broad range of scan parameters in five separate microstructural scenarios:
- histology of healthy tissue,
- atrophy, where all internal fibres are reduced in size,
- dystrophic tissue, using the muscle fibre radius distribution derived from our mdx mice,
- increased membrane permeability,
- atrophy where a fixed percentage of fibres are removed.
These data demonstrated that permeability is the dominant effect in changing the form of the diffusion decay curve. It also found that although all scenarios considered could in principle be detected at clinically-feasible SNR, the effects of each change were difficult to disambiguate when considered simultaneously.
Combining the knowledge derived from the computational modelling with the experimental observations of diffusion change in mice, we note that quantitative estimation of our target outcome parameter (muscle fibre radius, or a metric of muscle fibre radii distribution) is extremely challenging. While the diffusion signal contains important radiological contrast which does reflect differences in microstructural environment, this contrast is not invertible to reveal detailed geometric parameters. Diffusion contrast is potentially a unique marker which is not exploited in conventional approaches and may however be used as a biomarker in its own right.
Key Deliverables and Milestones
D1.1: “A new state-of-the-art protocol for quantitative diffusion imaging in muscle”, was documented using the simulation software and data fitting approaches to define a parameter space (diffusion time and diffusion b-values) which provided an efficient diffusion scan acquisition protocol optimised for muscle imaging.
Development of the muscle water diffusion simulation framework was a major milestone for WP1. This software has been incorporated into the Camino Diffusion MRI toolkit which is available as open-source code and scripts for running the simulations will be made available to the scientific community on request.
WP2 - Evaluation of diffusion imaging methods in mouse models for muscular dystrophy
Summary of Achievements
The STE diffusion imaging methods developed for mouse muscle MRI in WP1 were established on small animal MRI scanners at the two preclinical imaging sites in the Consortium (UNEW, LUMC). Scanner vendor differences in MRI sequences were overcome and scanner performance was harmonised to yield excellent reproducibility of quantitative diffusion data from test objects and subsequently in vivo. Diffusion imaging was then used to study restricted water self-diffusion behaviour in mouse hindlimb muscles in a range of in vivo studies and animal models. Scanning evaluated the effect of disease through comparisons of wild type and transgenic models of neuromuscular diseases (mdx, mdx-xist, ± exercise intervention to increase muscle pathology) and of ageing (7-8 weeks, 22 weeks and 44 weeks), with imaging data compared to histological assessment of muscle fibre size distribution and permeability.
Diffusion weighted imaging, data showed small but significantly increased diffusivity at long diffusion times (>100ms) in young mdx animals compared to wild type (8-16 weeks of age), while in studies of older animals (22 and 44 weeks) there was a more marked increase in diffusivity in the mdx groups. Conversely, pathological assessment of the muscles in these animals revealed that the mean muscle fibre size reduced with increasing age in the mdx animals. In impermeable fibres, such a reduction in fibre radius would cause greater restriction to diffusion movement and hence reduced diffusivity, contrary to our experimental observations. The observed increase in diffusivity is most likely explained by an increase in permeability of the muscle tissue to water diffusion, which was supported by observation of greater intra-fibre uptake of Evans Blue dye in the mdx muscles. Measurements using diffusion tensor imaging also suggested elevated diffusion in dystrophic muscle, but magnitude of the effect was more subtle.
Muscle T2 mapping was also performed in the same animals, as this parameter is already used in clinical investigations and is also sensitive to water mobility. Elevated T2 values were found in dystrophic muscle (mdx) with a greater dispersion in values than in wild type muscle, however no significant relationship was found between T2 and diffusion metrics, suggesting that diffusion and T2 measurements have different sensitivities to muscle pathology. We conclude that both T2 and diffusion are potentially useful biomarkers for muscle damage in animal models of neuromuscular diseases.
A novel framework for fitting advanced models to the diffusion MRI data has also been developed based on anomalous diffusion modelling. This is an entirely new C++ framework which fits models derived from Fractional Diffusion to diffusion-weighted data. To achieve a fitting approach which does not require users to specify initial starting conditions for fitting, a method was developed fitting a chain of models of increasing complexity, each of which provides the initial parameter guesses for the next model such that all initial parameter guesses are derived from the input data. The fitting code framework was shared with project partners for application to the multi-site preclinical) and clinical (WP3) acquisitions. We have applied this anomalous diffusion approach for the first time in muscle imaging and created and demonstrated that this analysis model is sensitive to changes in muscle fibre distribution when tested against synthetic, in vitro and in vivo data.
Key Deliverables and Milestones
D2.1: Report on the value of using diffusion imaging to assess muscle pathology, in comparison to T2 mapping methods in mouse models of NMD.
The advanced data fitting framework will be made available as open-source code o the wider research community once the findings from the BIOIMAGE-NMD studies have been published.
WP3 - Evaluation of diffusion MRI methods for assessing muscle pathology in NMD
Summary of Achievements
This work package addressed a large number of technical challenges around the translational step between animal and human imaging and potential to use our new diffusion methods in clinical trials.
As our target clinical trials were in children with Duchenne muscular dystrophy there was a need to shorten the imaging protocol both to improve compliance (leading to higher quality data) and to allow protocol time to include the new diffusion measurements. Our conventional clinical trial imaging protocol used 3 independent scan acquisitions to quantify muscle fat fraction and muscle water T2. Fat fraction is extracted from the Dixon acquisition, while muscle water T2 quantification required both a multi-spin-echo (mSE) acquisition and an associated B1 map to apply corrections for imperfections in the MRI RF excitation pulse slice profile and uniformity across the body. Using advanced reconstruction (extended phase graph, EPG) of the mSE data combined with biexponential modelling of the signal decay function, we demonstrated that fat fraction and T2 could both be extracted from the mSE data alone. Importantly, there was no correlation between T2 and fat fraction values using this method which demonstrates independence of the extracted quantified values. We evaluated this approach in a range of patients with varying NMD in comparison with the standard 3 scan approach and established that the EPG fitting derived fat fraction correlated well with Dixon derived measurement (≈1.5% bias). The new approach eliminates the need for separate Dixon and B1 mapping sequences and reduced the protocol time by ~20 minutes.
Establishing a robust diffusion imaging sequence for use in skeletal muscle required addressing the
challenge posed by fat infiltration which is a major feature in skeletal muscle of patients with NMD. Incomplete fat suppression in imaging voxels containing both fat and water, or voxels where the fat signal is misregistered due to chemical shift displacement both result in inaccurate diffusion quantitation. A new and highly effective method was developed to suppress fat by combining frequency selective suppression pulses for the large lipid signals (~1.5ppm) with a Dixon approach for the smaller olefinic fat signal. The new method was successfully implemented as part of a STE diffusion sequence on clinical MR systems from two of the main vendors. The full fat suppressed STE diffusion sequence was used with a diffusion tensor encoding scheme and intra- and inter-site performance was evaluated across 3 sites of the Consortium using measurements in the same cohort (n=5) of healthy controls travelling between scanners. A post-processing pipeline was established with algorithms to separate the water signal from any residual fat, eliminate spatial image distortion and calculate pixelwise diffusion metrics in muscle. Within a site mean diffusivity (MD) data had good reliability (maximum SD across the group of <10%), but there were differences in values between sites of 15-20%, (present as a systematic offset in all diffusion values). Analysis of the diffusion data using conventional diffusion metrics showed that tissue fractional anisotropy (FA) values were overestimated at low signal to noise ratio (SNR). A minimum SNR of 7 was determined as the threshold which constrained controls FA and MD results within a 5% confidence interval.
The final phase in this work package was to evaluate the new diffusion methodology in muscular dystrophy patients. A protocol combining both standard (spin-echo) and advanced (STE with two diffusion times, 100 ms and 300 ms) diffusion sequences was used in 10 patients with Becker Muscular dystrophy (BMD). Although fat infiltration was suppressed by the new scheme, the residual signal from remaining muscle water was still reduced simply due to volume loss of muscle within each imaging pixel, leading to lower SNR. In BMD patient muscles our SNR threshold for robust FA calculation (SNR>7) was not fulfilled in 50% of ROIs, corresponding to the muscles with highest fatty infiltration. Of all muscles sampled within the leg, the tibialis anterior was the only one to meet the SNR criterion. Here FA was significantly higher in muscle of BMD patients (n=5) as compared to healthy controls (n=4) (p=0.016). MD values were not significantly different in BMD compared to control in any muscle but tended to be lower in soleus (p=0.057). These results show that it is feasible to perform STE-DTI in BMD patients, and thereby to take advantage of extended mixing times for muscle cell specific applications. This study highlighted two other significant observations; SNR cannot be determined from scan parameters alone, but depends on the degree of disease related muscle loss and secondly the trend to disease related decrease in MD is in opposition to the changes seen in animal models. The origin of the changes in human muscle are unclear but interpretation based on our animal data indicate they are a complex balance between smaller fibre sizes, possibly in competition with increased membrane damage and also may involve development of fibrosis. Our overall interpretation of the BMD study is that in muscles such as the TA which is relatively spared by fatty infiltration, we anticipate that the use of mixing times of 100ms and above will allow early detection of changes in DTI metrics.
Key Deliverables and Milestones
D3.1: Report on the new State of the Art (comparison of conventional and advanced methods) for assessment of human muscle by diffusion MRI.
In this WP we achieved the major milestone of first-in-man assessment of diffusion imaging in NMD patients.
WP4 - Proof of principle of MRI/MRSI biomarkers in clinical trials in DMD
Summary of Achievements
The conventional MRI and MRSI protocol was successfully implemented in all clinical trials within the BIOIMAGE-NMD project and a leg holder was sconstructed for use at each site. Study-specific imaging manuals were written and sites were trained in the use of the protocol before being validated to take part in the clinical trials through quality control data in a phantom and subsequently in healthy volunteers. To underwrite data quality during the longitudinal investigations all sites were required to rescan phantoms every 6 months and healthy volunteers every 2 years.
Imaging protocols were then used in the AON intervention clinical trials in DMD patients alongside the routine clinical and functional evaluations. All imaging data in DMD patients was acquired using the unmodified conventional imaging protocol defined at the outset of the BIOIMAGE-NMD programme. To preserve data consistency in ongoing clinical trials, the revision of trial protocols to introduce fast scanning or diffusion imaging developments (WP3) were only ever planned to coincide with new clinical trials. Commercial and clinical factors during the lifetime of the BIOIMAGE-NMD programme led to a departure from the original clinical trial phasing such that availability of more advanced imaging methods did not map to the revised clinical trial schedules.
A total of 52 patients enrolled in the clinical trials were scanned using the conventional protocol. Scans were collected in the initial PRO053 and PRO044 trials. The anticipated global roll out phase for PRO053 did not take place so MRI data was added to the PRO051-DMD115501 extension study as part of the original agreed global roll out to sites beyond the Consortium. The PRO051-DMD115501 trial included scans at baseline, 6, 12 and 18 months, while for PRO044, baseline and 6-month MRI exams were acquired. For PRO053, baseline scans and several follow-up visits (between 3 and 30 months) were acquired. As a comparator group, natural history data (PRO-DMD-01 study) was obtained in a further 23 subjects who were scanned at baseline, 6, 12 and 18 months. A total of 220 patient data sets were acquired across all studies and were available to evaluate the value of imaging as outcome measures in WP6.
Key Deliverables and Milestones
D4.1: Status update on the imaging data collected in the clinical trials and natural history study.
WP5 - Decision on the use of novel imaging methods in the PoP clinical trial
Summary of Achievements
This short work package evaluated the scientific evidence in relation to the performance of the novel imaging methodology developed in work packages 1 to 3. This WP took the form of a data review and discussion between all partners of the Consortium to debate the merits of the new methodologies and arrive at a consensus on their value for clinical trial evaluation. Specifically, for the fast scanning methods developed in WP3, the criteria required evidence that the new methods could be introduced to replace the conventional approaches without introducing bias or systematic errors in the quantified data. The methods should also be appropriate for clinical use and must be achievable on the network of systems across the clinical trial sites. The EPG and biexponential analysis approach of the mSE scan data was considered to be a significant step forward which met both of these criteria. As the method relies solely on post-processing it could be applied, with appropriate preparatory work on test object and control subject scans, to data from all sites. The scientific work underpinning the method demonstrated low bias.
The advanced diffusion methodology was considered to be exciting and to provide interesting and potentially valuable new indices of muscle pathology. The opposing observations between animal data (WP1 and WP2, where histological correlates were available) and BMD patients (WP3, where no histology was available) did not provide a full understanding of the sensitivity of the measures in human subjects. Further work is clearly required to elucidate the complete biophysics of diffusion processes in healthy and diseased muscle. The diffusion methodology was therefore not considered sufficiently mature for clinical trial use.
Key Deliverables and Milestones
D5.2: Report detailing the consensus decision on the performance and quantitative compatibility of new fast scan T2, B1 or Dixon methods with original state of the art methods in the MRI/MRSI protocol.
A summary of the consensus statement in Deliverable Report D5.2 is as follows:
1) Diffusion imaging methods show promise as a tool to detect alterations in muscle microstructure with disease progression. At the current time there are discrepancies between the observations in animal models and human subjects and further work is required to define an underpinning theoretical model which reconciles the experimental data. At the current time diffusion imaging is not recommended as a putative outcome measure.
2) T2 mapping is a valuable measurement to detect the presence of muscle edema associated with inflammation. T2 does not itself provide a marker for disease progression and can be modified acutely by treatments such as steroids. The precise value of T2 measurements in clinical trials depends on the nature of the treatment and the specific disease being studied.
3) Neuromuscular imaging protocols seeking to measure muscle water T2 should use EPG analysis of the T2 data, which eliminates the need for B1 map data collection and reduces the time burden on patients, particularly in studies involving children.
WP6 - MRI and MRSI data analysis
Summary of Achievements
This work package addressed the many challenges of centralised analysis of clinical trial imaging data, specifically
- Developing a method for secure file transfer from trial sites to the central team,
- Developing time-efficient, semi-automated analysis tools for robust muscle segmentation
- Performing the analysis of the clinical trial data acquired in WP4.
We report our achievements for each of these tasks below:
A) A web-based tool for secure file transfer
We developed a client/server web-based application allowing the imaging centres to transfer the data in an easy way and to perform quality control checks at site level (such as verification of subject IDs, correct labelling of the data, completion of necessary NMR data). The processes are automated and fast at both ends (imaging centres and central data analysis centre). At the level of the central reader, quality control of the data can be performed (such as accurate Dixon reconstruction) and segmentation of the muscles is integrated using the semi-automated algorithm developed in this work package. Processing of the data can be performed in batch. Great care was given to achieve GCP compliancy, which guided the development of several features (data tracking, encrypted archiving, secured login, code versioning). A first version of this web-based service is being tested at the time of writing this report. The full-scale deployment is not yet activated for a lack of maturity, but the developing partner, CRIS is aiming for this last feature to be ready by the first quarter of 2018.
B) A semi-automated tool for muscle segmentation and statistical analysis
The extraction of muscle fat fraction and water T2 values requires definition of region of interests (ROI) delineating each muscle at each slice of the imaging dataset which is a major bottleneck in clinical trial data analysis. To address this challenge, a semi-automated, muscle segmentation algorithm was developed. As the main challenge to automatic segmentation is detection of muscle boundaries in the present of high levels of fat infiltration the algorithm was validated on scans from the lower-legs of healthy volunteers and Duchenne patients. The algorithm was compared with the gold-standard manual approach using similarity metrics: volumetric similarity (1 – absolute volume difference divided by sum of compared volumes), Dice coefficient (overlap measurement), Hausdorff distance (maximal distance between any points on each of the segmented volumes).
Manual segmentation times on the volunteer cohort were 35 min per muscle on average (x12 muscles = 7 hours per subject). The mean segmentation time with the new tool was 2 min, yielding an acceleration factor of 17.5. On the Duchenne patients, the acceleration factor was less (mean 7 mins) due to the smaller number of slices to segment. On the muscle groups of the healthy volunteers, the mean volume similarity was 0.96 mean Dice coefficient was 0.90 and the mean Hausdorff distance was 10.75mm. On the Duchenne patients, similar to better results were found showing that the fatty infiltrations did not hinder the algorithm performance.
C) Analysis of the clinical trial data acquired in WP4
The analysis protocol required that all subject scans were quantified to derive the following MRI outcome measures:
- fat fraction in each individual muscle and muscle groups,
- cross-sectional area and contractile cross-sectional area in muscle groups,
- water T2 mapping in muscle.
Datasets from expert sites where 31P-MRSI data was available were quantified to obtain spectral indices of tissue energetic and metabolic state (pH, PCr, Pi and PDE levels) in gastrocnemius muscle.
The 220 patient data sets included Dixon, MSME (T2), and B1 sequences for the lower limbs (both legs and thighs). 31P MRS data was obtained in boys enrolled in the PRO053 study scanned in Paris, Newcastle and Leiden, from boys enrolled in the PRO044 study scanned in Leiden and in the natural history (PRO-DMD-01) also scanned in Leiden. A total of 25 31P MRS exams were obtained for PRO053, 6 exams for PRO044 and 8 exams for the natural history study. Clinical and functional data acquired by the consortium partner BioMarin (formerly Prosensa) within the main clinical trial datasets was made available during the final 3 months of the programme to allow analysis alongside the imaging data. The clinical data provided the genetic mutation of all subjects in the natural history study, most of which were exon-44, exon-45, exon-52 or exon-53 skippable with only one of the subjects being exon-51 skippable. Whenever possible, quantitative findings from the MRI/SI data were correlated to functional data (particularly the 6-minute walk distance as one of the main standard clinical measures accepted as a primary outcome measure for clinical trials in NMD patients).
Data from the natural history study (PRO-DMD-01) and the PRO051-DMD115501 clinical trial comprised the largest part of the MRI/S data within the BIOIMAGE-NMD project, enabling us to perform a comparative analysis between the two patient groups and examine the longitudinal data for possible therapeutic effect in the treated group. After quality checks, datasets from 3 of the 6 sites in the PRO051-DMD115501 were excluded due to incomplete (missing paired datasets between baseline and follow-up) or poor data quality, leaving 24 patient datasets with 82 imaging examinations. All data from the natural history study was included (23 patient datasets with 63 scans). The observations we draw from these patient datasets are as follows:
i) Observations relating to Fat Fraction and Treatment Effect
The trajectories of fat fraction change over time were examined for each muscle and considering progression as a function of patient age. While each patient was only scanned over a short duration, the range of ages at enrolment spanned 5 to 15 years such that the combined trajectory plots across all subjects could be used to observe the overall shape of the progression, which typically followed a sigmoidal curve. When looking at the trajectories in more detail we observed that patients of the same age have significantly different amounts of fatty infiltration at their baseline scans. Moreover, when looking at two patients of the same age and with almost the same amount of fat infiltration at baseline, we saw a different degree of fat fraction increase in the following year (i.e. there were different slopes to the individual subject trajectories). Despite these clear differences between individual subjects, groupwise comparison of the average fat fraction increases at 1 year intervals between a carefully matched subpopulation of treated (DMD115501) and untreated (natural history) subjects showed a significant 51-exon-skipping effect, with treated boys showing a significantly smaller increase in fat fraction than the untreated boys (p<0.01).
In the PRO053 and PRO044 studies subject numbers were small (n=9 in both groups) with typically only the baseline scan and 1 or 2 follow-up scans. The PRO044 subjects were significantly older at study inclusion than subjects in the other trials, had lower 6-minute-walk-distance values (related to their older age) but showed no real fat fraction changes after 6 months. Two of the PRO053 subjects showed a steep increase in fat fraction over the course of two years whereas the other subjects had stable values across the course of the study. These two subjects were untreated whereas the others received treatment. Their respective 6-minute-walk distances decreased accordingly. Subject numbers were too small to allow meaningful statistical analysis or to make any conclusion beyond a qualitative description.
ii) Observations relating to muscle water T2
Muscle water T2 reflects disease activity and is a sensitive yet unspecific biomarker of the inflammatory/necrotic/oedematous processes. Muscle water T2 is also sensitive to steroid treatment.
Contrary to fat fraction, water T2 is prone to fluctuations (high water T2 values in active disease while low water T2 in more passive phases of the disease), making its interpretation more complex. The threshold for determining a ‘pathological’ water T2 value depends on site-specific elements such as acquisition parameters and field strength, requiring normalization of the data. To achieve a robust assessment of patient data we restricted analysis to data from a single site where both treated and untreated subjects were scanned (thus same scanner, same field strength, same acquisition parameters). We observed generally lower water T2 values (only significant in leg muscles) in the DMD115501 subjects compared to the natural history subjects. Water T2 values also seemed to increase after 1 year in treated patients as compared to the untreated patients.
iii) Observations relating to Fat Fraction and Disease Progression
Six minute walk distance data was available for subjects at baseline and the first 6 month follow-up assessment. Across patients in all trials, there was a clear inverse relationship between 6MWD and muscle fat fraction. For example, in the hamstring muscle subjects with low fat fraction (<10%) had 6MWD of 350-525m, while subjects with fat fraction >10% had much reduced 6MWD. From the longitudinal data, the general trend was observed that increasing fat fraction was associated with decline in 6MWD in the individual subject, while stable fat fraction was associated with little change in 6MWD. Some individuals did show increased 6MWD at the follow-up time point which may reflect the influence of subject motivation on functional performance, age related effects (increasing muscle bulk with maturation balanced against disease progression) as well as possible treatment effect.
iv) Observations of metabolic effects by 31P-MRSI
Phosphorus-31 MRS data were available in patients scanned within the main consortium sites which included subjects in the PRO053 and PRO044 studies. These patient showed increased pH and PDE values compared to controls.
To summarise the findings from the BIOIMAGE-NMD Proof of Principle clinical trial studies, we have confirmed that fat fraction as derived from Dixon MRI is a valid biomarker of disease severity. Clear correlations were found between the loss of real muscle mass (as determined by MRI) and functional outcome measures such as the 6-minute walk distance. A comparison of fat fraction between AON treated PRO051 patients with natural history data did illustrate a treatment effect. However, based on the available data, fat fraction does not seem to be predictive of disease evolution. The diverse phenotypic expression of the disease as well as its non-linear progression on an individual basis, hamper the establishment of prognostic indices based on a single fat fraction determination, even when age is taken into account. Results of other MRI/S biomarkers such as water T2 and 31P MRS indices are even more difficult to interpret because of multiple reasons. Muscle water T2 reflects the disease activity and therefore needs short-term evaluation of its fluctuations on a subject-to-subject basis in order to determine its discriminant power when comparing pre- and post-treatment. 31P MRS results were scarce due to the limited number of sites who were able to acquire these kind of data and more data is needed in DMD.
Key Deliverables and Milestones
D6.10: Summary & conclusions of the analysis of the different outcome measures generated during the study (M42)
WP7 - Selection of the prosthetic group and labelling strategy for 18F radiolabelling of AON
Summary of Achievements
We have developed synthetic protocols to prepare [19F] labelled oligonucleotides as a first step in production of [18F] radiolabelled drugs for PET imaging. To couple the fluorine-18 radiolabelled prosthetic group to the antisense oligonucleotide (AON) it was necessary to introduce a reactive functionality into the structure which is achieved by incorporation of an appropriate ‘modifier’ at the end of the conventional AON production process. The aminooxy moiety is typically used due to the selectivity, and efficiency of the process. Aminooxy modified AON was therefore prepared and then two different prosthetic group strategies were explored for ‘tagging’ the modified AON to create a [19F]-AON. These different prosthetic groups led to very different stabilities and yields. The most efficient group was selected for further development. Multiple production runs of the [19F]-AON were carried out demonstrating the robustness of the process. Labelled [19F]-AON could be produced in high milligram quantities with high purity (97%) allowing in vivo assessment of bioequivalence of the original and labelled AON compounds.
The [19F]-AON or native AON were administered into mdx mice using a protocol defined to establish the relevant biological data on a timescale to match the PET study which is limited by the fluorine-18 half-life. Mice were given a single injection and tissues (skeletal muscle groups, heart, liver, kidney and spleen) were isolated 4 hours post-treatment after perfusion of the mice. Plasma samples were also acquired throughout the 4 hour period. AON levels in plasma and tissues were quantified by an ELISA. A comparable biodistribution pattern was observed for both the labelled and unlabelled compounds, which demonstrated that incorporation of the fluorine label into the AON did not alter AON pharmacokinetic profile.
Key Deliverables and Milestones
D7.1: In vitro/in vivo bioactivity data of the two ‘cold’ labelled materials
WP8 - Evaluating PET/MRI imaging protocols in mdx mouse models
Summary of Achievements
In this second work package associated with radiolabelling of the AON molecule, the methodology developed using cold [19F] labelling was modified to use the positron emitting isotope [18F] to create the [18F]-AON compound for use in PET imaging studies. The automated preparation of the [18F]-AON was achieved using a Modular Radiosynthesis platform. Reaction conditions and purification methods were optimised to achieve activity/volume ratios suitable for pre-clinical use. GMP compatible purification and QC protocol have been developed for the [18F]-AON using both high pressure liquid chromatography (HPLC) methods and single phase extraction (SPE) methods to ensure the methodology can be translated in the future (beyond BIOIMAGE-NMD) to typical commercial automated radiochemistry platforms used by clinical PET facilities. Residual solvent analysis demonstrated levels that are below the limits required for clinical use.
The first biodistribution studies using PET imaging [18F]-AON were carried out in a small number of mdx and wild type mice. Scanning was performed using PET-MRI and demonstrated the principle of detecting this new class of fluorine-18 radiolabelled AON’s in vivo. The use of a PET-MR scanner for this study has also highlighted that both PET and MR data can be obtained from a single subject. The MRI acquisition however only considered standard structural imaging and the other advanced quantitative MR methods developed in work packages 1-3 were not collected. The relationship between the PET and MRI methodologies remains to be developed. The fundamental concept of PET/MR studies for the evaluation of AON therapies has been demonstrated and as such this milestone has been completed.
Key Deliverables and Milestones
D8.1: Report on application of PET/MRI imaging to correlate AON levels with muscle structure
This report was the only deliverable which was not achieved across the full programme. Work to deliver this activity was not completed within the timeframe of the project due to 2 main factors, both around staffing at the two main partners in this activity. The key post-doctoral research worker who was instrumental in establishing the PET labelling methodology was recruited to another institution and departed the programme with 3 months of work remaining, leaving insufficent time to find a replacement to complete the work before the end of funding. In parallel, the commercial decision by BioMarin to halt clinical trials of the AON agents fed through to changes in their programme of work around AON development limiting activity within the second partner who were to conduct the ELISA analysis alongside the PET imaging. In combination, these factors prevented the correlation analysis which was the topic of the Deliverable Report. While the full report could not be delivered, the Periodic Report for the final (3rd) period does describe significant advances and proof of principle for the technology.
WP9 - Dissemination
Summary of Achievements
Initial dissemination activities were focussed on the building of the website and intranet as well as establishing publication guidelines. From the outset, the Project Office worked closely both with consortium partners, in particular BioMarin and their continued contact with patients and patient organisations to raise awareness of the project. As the project has progressed, dissemination activities have reflected the increased research activities in the different work packages, resulting in a substantial numbers of publications and conference articles relating to BIOIMAGE-NMD.
Alongside publications and abstracts, the BIOIMAGE-NMD consortium was involved in the organisation and hosting of an international conference, held in London on 26th May 2016, entitled “'Making outcomes work' - Stakeholder Workshop on outcome measure development and implementation for DMD”. The coordinator from Newcastle University, along with many representatives from the BIOIMAGE-NMD consortium partners including WP leads from WP1, WP2, WP3, WP5 and WP6 attended. In January 2017, BIOIMAGE-NMD also took part in the international conference entitled “'Making outcomes work' - Stakeholder Workshop on outcome measure development and implementation for DMD”. The coordinator from Newcastle University, along with many representatives from the BIOIMAGE-NMD consortium partners including WP leads from WP1, WP2, WP3 and WP6. WP Leads, including the coordinator of BIOIMAGE-NMD, were also invited to present on BIOIMAGE-NMD at the Newcastle University Institute of Genetic Medicine Seminar Series.
Key Deliverables and Milestones
D9.1: Plan for the dissemination and Use of Foreground
The PDUF was created at the start of the project and the Deliverable was submitted on time. The PDUF was kept updated throughout the lifetime of the project.
WP10 - Project Management and Co-ordination
Summary of Achievements
Overall, the project has been managed well, with 6-monthly Partner meetings as well as 3 monthly Scientific Committee teleconferences. The Project Office worked closely with all partners to ensure progress of key deliverables and milestones and has worked closely with the EU Scientific and Legal Officers to coordinate and resolve complicated issues relating to project partner engagement in Work Package delivery. These issues were solved satisfactorily and have led to an amendment of the original DOW with distribution of budget and additional and/or amended milestones.
Financial reporting was coordinated via the Newcastle University EU Finance Office and all partners have followed EU guidelines on submitting the relevant (audited) reports when required.
Key Deliverables and Milestones
D10.1: Final reports will be delivered to the Commission within 60 days after the end of the project
All periodic reports as well as the Final Report were delivered on time to the Commission.
Potential Impact:
- Quantitative Imaging as Biomarkers for Clinical Trials in Rare Diseases:
Neuromuscular diseases include a range of highly heterogeneous, frequently devastating rare and ultra-rare conditions which affect both sexes and all age groups across the world. A conservative estimate of the overall prevalence of NMD indicates that 1 in 3500 of the population may be expected to have a disabling inherited NMD presenting in childhood or later in life. In the context of BIOIMAGE-NMD we have focussed on patients with Duchenne muscular dystrophy which is characterised by severe, progressive and irreversible loss of muscle tissue and function and affects 1 in 5000 male births. Affected children are typically diagnosed at about four or five years of age, when it becomes apparent that their motor abilities are restricted. Without treatment most require the use of a wheelchair around the age of 9.5 years. Extreme muscle weakness results in orthopaedic problems and patients also suffer respiratory and cardiac complications. Death often occurs by the late teens or early adulthood, mostly due to respiratory or cardiac failure. The goal of existing management is to minimise the impact of these complications as there are no curative treatments for any NMD. The costs of NMD conditions, in terms of health economics, is enormous. It is estimated that in 2010 across 30 European countries, €8 billion was either spent on health care of NMD or lost due to reduced productivity (Gustavsson et al., Eur. Neuropsychopharmacology, 21, 718, 2011).
Currently, many companies and academics are trying to develop different types of therapies for these rare diseases, including the exon-skipping AON therapeutics studied within BIOIMAGE-NMD and phosphorodiamidate morpholino oligomers (PMO). However, they all face the same continuing need for sensitive and validated clinical trial end points. The 6 minute walking distance (6MWD) test remains the main accepted primary end point in NMD clinical trials. As secondary outcomes the North Star Ambulatory Assessment (NSAA), measurements of the muscle strength using hand held myometry and timed tests are used. However, it is not possible to demonstrate a statistically significant result with the 6MWD test on the small numbers of patients available in rare disease clinical trials. Within the BIOIMAGE-NMD consortium we have spearheaded the application of non-invasive imaging technologies as biomarkers which can meet this unmet need. BIOIMAGE-NMD has developed an array of tools based around MRI and PET imaging which could in the future eliminate the need for muscle biopsies in a patient population where muscle function is already compromised. In contrast to biopsies, imaging biomarkers will also allow us to look at a much wider range of muscle types in the patients and assess muscle types that normally are not prone to biopsies such as the muscles of the neck. All of our imaging biomarker developments have been targeted at quantitative measures, with the aim to provide maximal sensitivity to detect disease progression and changes induced by therapy.
The work within BIOIMAGE-NMD has confirmed that fat fraction as derived from Dixon MRI is a valid biomarker of disease severity and disease progression. Changes in muscle fat content measured by imaging map strongly to the reduced functional outcome measures such as the clinically accepted 6MWD. While the 6MWD provides an overall measure of ambulatory function in the patient, the imaging measures provide a much more detailed assessment of the muscle-by-muscle progression of the disease. Importantly, while BIOIMAGE-NMD has concentrated on muscle imaging in the thigh and leg, the Dixon and T2 mapping methodologies are entirely applicable in the upper limbs. Thus, while the 6MWD loses meaning in patients who have lost ambulation, the imaging based measures offer the opportunity to monitor progression and response to intervention through upper limb assessments. Non-ambulatory NMD patients do not then need to be excluded from clinical trials based on the design of the outcome measure assessments.
Our ability to achieve the objective of proof of principle use of MRI methods as biomarkers to monitor therapeutic efficacy in clinical trials in neuromuscular diseases has been slowed by a combination of factors. Firstly, the complexity and diversity of phenotypic expression of the disease as well as its non-linear progression on an individual basis, hamper the establishment of prognostic indices based on single fat fraction determination, even when age is taken into account. Secondly, the clinical efficacy of the first generation AON therapeutics which were studied as the clinical trial exemplars for detection of effect were lower than anticipated (a factor which led to the early closure of trials and withdrawal of these compounds from licencing applications to the regulators). While efficacy was difficult to detect using the standard clinical outcome measures, a detailed analysis of the imaging data did demonstrate differences in fat fraction progression between AON treated PRO051 patients and untreated patients in the parallel natural history study. Clearly, further work is required to understand disease progression in patients with NMD and how to select or stratify patients for clinical trials. The data from BIOIMAGE-NMD adds to our current level of understanding and continues to support the potential of imaging biomarkers. We are convinced by our own data that imaging biomarkers do offer sensitivity to detect changes induced by therapy much earlier than the conventional outcome measures, which would fill a real gap in the development of new therapies in (rare) neuromuscular diseases. Indeed, the experience and knowledge gained from the clinical trial imaging in BIOIMAGE-NMD has already shaped trial protocols being pursued by members of the consortium with other SME partners. As next generation AONs, or alternative approaches (exon-skipping or beyond) come through the drug development pipeline, imaging biomarkers will continue to play an important role in patient assessment.
- Developing advanced tools to enable clinical trials with imaging biomarker endpoints
At the commencement, and throughout the lifetime of BIOIMAGE-NMD there has been much commercial interest and activity around using imaging to monitor progression and potential therapeutic response in NMD trials (both within our consortium partners and from other SMEs and researchers in this space). A major challenge to delivery of trial data analysis has been the sheer depth of information which the imaging investigation encapsulates. Extracting muscle-by-muscle data on tissue properties such as fat fraction was traditionally performed manually, taking typically one day of expert analysis time per subject scan timepoint. The cost-effectiveness of conducting image analysis was therefore an important adjunct factor in commercial decisions to include imaging in clinical trials. Through the work performed by the SME partner CRIS, a new semi-automated tissue segmentation tool was created which reduces analysis time from hours to minutes, is effective even in datasets with extensive fat infiltration as are found in NMD patients and which will produce reliable quantification without the potential for user bias associated with manual assessment. This tool is a significant step forward in providing a cost effective pathway for clinical trial analysis. The technology resides within the European SME and represents a tangible competitive advantage.
- Enabling Positron Emission Tomography for rapid biodistribution studies of AONs
We have demonstrated for the first time that positron emitting radiolabels can be attached to AON molecules without affecting the biodistribution or exon skipping efficacy of these molecules in vivo. This is a major tool which can be used to advance new AON molecules both in the drug development environment and in the subsequent translation to clinical trials in NMD and in wider clinical applications. A major factor in extending the use of AONs in rare diseases is the challenge of demonstrating effect in only small numbers of subjects. PET imaging offers the highest sensitivity of all in vivo imaging methodologies and so through the developments made in BIOIMAGE-NMD provides an important tool for future work.
Firstly, through PET imaging of AON, the biodistribution can be rapidly determined in vivo during preclinical testing, simultaneously observing all organs of the animal. The non-invasive nature of the measurement allows repeat investigation in the same animal without the need for sacrifice to determine tissue levels which offers an increase in statistical power over invasive studies which must use independent animals for each evaluated time point. We expect that this technology should enhance drug development for second (and later) generation AON molecules.
Secondly, when translating AON studies from preclinical work into early phase human studies, the use of PET labelling offers a tool to ensure that the drug is reaching the same target in patients as is expected from the preclinical work. As AON molecules enter the muscle cell through areas of muscle membrane damage, the potential combination of diffusion MRI, which we have shown in animal models to be highly sensitive to permeability, with PET studies would provide important data.
Thirdly, PET imaging offers a tool to assess and optimise AON administration routes, both in animal studies and in human trials. The acute fate of the AON, rates of distribution across the body and clearance can be measured directly through dynamic PET scanning (limited by the half-life of the 18F isotope) in the individual patient. This approach can eventually feed through into development of personalised treatment regimes based on monitoring of the drug distribution in each patient.
- Dissemination activities
The BIOIMAGE-NMD Consortium has engaged in dissemination of the project findings through a wide array of routes, including presentations at specialist international conferences in the field of imaging, radiochemistry, neurology and neuromuscular diseases.
A Global Workshop on the use of biomarkers in NMD clinical trials was hosted at the British Library on the 26th May 2016 by Andrew Blamire (Coordinator, BIOIMAGE-NMD) and Volker Straub (Coordinator, SCOPE-NMD). The objective of the workshop was to bring together representatives of all the Consortium partners (BIOIMAGE-NMD and the FP7 funded SCOPE-NMD), the European Commission, EMA and patient organisations to review the recent FP7 funded work on the use of imaging and other biomarkers for clinical trials and specifically to discuss the progress on outcome measures for evaluation of patients with DMD. All key stakeholders were represented and there was a wide ranging debate on making outcomes work in the future for this disease. The coordinator from Newcastle University, along with many representatives from the BIOIMAGE-NMD consortium partners including WP leads from WP1, WP2, WP3, WP5 and WP6 attended. In January 2017, BIOIMAGE-NMD also took part in the international conference entitled “'Making outcomes work' - Stakeholder Workshop on outcome measure development and implementation for DMD”. The coordinator from Newcastle University, along with many representatives from the BIOIMAGE-NMD consortium partners including WP leads from WP1, WP2, WP3 and WP6. WP Leads, including the coordinator of BIOIMAGE-NMD, were also invited to present on BIOIMAGE-NMD at the Newcastle University Institute of Genetic Medicine Seminar Series.
Presentation of the work from BIOIMAGE-NMD will also be showcased through invited presentations by the Consortium Coordinator and several of the work-package leaders at the First International Conference on Imaging in Neuromuscular Disease, to be held in Berlin in November 2017.
List of Websites:
www.bioimage-nmd.eu
The overarching objective of the BIOIMAGE-NMD project was to deliver combined structural and molecular imaging biomarkers with proven utility for the detection of therapeutic effects in patients with rare neuromuscular diseases (NMD). The project addressed three major themes, divided across 8 academic Work Packages:
1) The first theme conducted a full translational development of diffusion MRI methodology optimised for muscle microstructure as a new putative biomarker for pathology in patients with NMD. Restricted self-diffusion behaviour of water within muscle tissue was observed in mouse muscle samples, indicating sensitivity of diffusion MRI to muscle microstructure. Studies used in vitro and in vivo scanning in wild type mice and the mdx model of DMD. Significant differences in diffusion were found between groups. Factors influencing the restriction of water diffusion were explored against histology in the same muscle samples and through development of a novel Monte-Carlo modelling platform for muscle architecture. While histological changes in muscle fibre size and size distribution were observed between wild type and dystrophic muscle, the over-riding factor defining water diffusion in muscle was determined to be the overall permeability of the system. Diffusion methods were translated to clinical scanners and new highly efficient methods to suppress lipid signal in the MRI scans were developed to deliver high quality, quantifiable scans in human subjects. Scan reproducibility was evaluated within and across sites in the Consortium. Finally, a first-in-man study in NMD was conducted in patients with Becker Muscular Dystrophy. Small, but significant differences were observed in patients.
2) The second theme aimed to provide a proof of principle that simultaneous MRI/MRSI can be used as a biomarker to monitor therapeutic efficacy in clinical trials in NMD, specifically in Duchenne Muscular Dystrophy (DMD). Clinical trials were led by our commercial partner BioMarin with 3 different antisense oligonucleotide (AON) treatments (named PRO051, PRO044 and PRO053, where the number indicates the exon within the DMD gene at which each is targeted), while additional patients were studied in a parallel natural history study. Quantitative imaging methods (Fat Fraction, T2 mapping, Phosphorus-31 MR spectroscopy) were incorporated into each study. Across these studies 75 patients underwent longitudinal imaging, obtaining a total of 220 scan datasets. The results across all studies confirmed that MRI derived fat fraction is a valid biomarker of disease severity and progression. Correlations were found between the loss of real muscle mass and functional outcome measures such as the 6-minute walk distance. In the largest patient subgroup (PRO051 treatment), a significant difference in rate of fat infiltration was observed between treated and non-treated patients, indicating effect of exon-skipping therapy. However, the diverse phenotypic expression of the disease as well as its non-linear progression on an individual basis are major confounds to establishing prognostic indices based on the imaging data. To enhance clinical trial capability, secure, web-based data transfer tools were developed for exchange of images between sites and the central analysis team. Novel analysis tools using semi-automated muscle segmentation were also created, speeding up image quantification by a factor of 7-17 times (equivalent to a time saving of ~6 hours per patient dataset).
3) The third theme developed a novel simultaneous Positron Emission Tomography (PET)/MRI technology directed towards advancing innovative drug development programmes for personalised medicines based on AON technology. We developed synthetic protocols to prepare [19F] labelled oligonucleotides as a first step to production of [18F] radiolabelled drugs for PET imaging. Labelled [19F]-AON could be produced in high milligram quantities with high purity (97%) allowing in vivo assessment of bioequivalence of the original and labelled AON compounds. The [19F]-AON or native AON were administered into mdx mice and ELISA analysis of tissue and plasma levels showed equivalent biodistribution patterns for labelled and unlabelled compounds, demonstrating that incorporation of the fluorine label did not alter the in vivo AON pharmacokinetic profile. Radiolabelling of AON with the positron emitting isotope [18F] was achieved on an automated radiosynthesis platform with reaction conditions and purification methods optimised to achieve the necessary activity/volume ratios using GMP compatible protocols. The first biodistribution studies using PET imaging of [18F]-AON were carried out in mdx and wild type mice using PET-MRI and demonstrated the principle of detecting this new class of fluorine-18 radiolabelled AON’s in vivo.
The BIOIMAGE-NMD project delivered significant advances and new technologies across all 3 themes, increasing our understanding of the role, benefits and challenges of imaging as outcome measures for clinical trials.
The BIOIMAGE-NMD consortium consisted of 2 SME companies, BioMarin and Consultants for Research in Imaging and Spectroscopy (CRIS) and 6 leading academic institutes, Newcastle University, the Leiden University Medical Centre, the Institute of Myology (Paris), University College London, Katholieke Universteit Leuven and Universita Cattolica Del Sacro Cuore (Rome) and was coordinated by Newcastle University.
Project Context and Objectives:
Duchenne Muscular Dystrophy (DMD) is a relatively rare but lethal genetic childhood disease which affects between 0.32 and 0.52 per 10,000 inhabitants. Worldwide, around 240,000 boys suffer from DMD. DMD is caused by mutations, often deletions, in the DMD gene that result in the disruption of the open reading frame leading to the loss of a loss of dystrophin protein expression. Dystrophin has an important role as a scaffolding within muscle fibres where it links the actin cytoskeleton with the extracellular matrix and absorbs force generated during muscle fibre contraction. Muscle fibres lacking dystrophin are damaged during normal activity and are unable to regenerate healthy tissue, eventually leading to progressive muscle weakness and wasting. While DMD is clinically well characterised it lacks any drug treatment which substantially alleviates the disease progression. One of the most promising emerging treatments for NMD, including DMD, is the use of exon skipping drugs, including Antisense Oligonucleotides (AON). AONs are a genuine example of personalised medicine, where subsets of patients are treated according to their specific genetic mutation. Despite the clear potential for AON technology, the ongoing clinical development of treatments for this type of rare NMD are impeded by 1) the low number of patients available with any given mutation, 2) the lack of objective and reliable outcome measures which span all phases of the disease, and 3) the current need to use invasive muscle biopsies to monitor therapeutic response.
Imaging techniques such as magnetic resonance imaging (MRI) and positron emission tomography (PET) offer the opportunity to non-invasively quantify many aspects of human tissue and may be useful as biomarkers for disease progression or outcome measures in clinical trials. In recent years it has been suggested that the development of clinical trial imaging biomarkers should be an integral part of the drug development pathway, such that imaging methods proven to determine whether a drug is reaching the desired tissue and having the expected therapeutic impact in vivo, are developed in parallel with the drug itself. This integrated approach would ensure that the necessary tools to quantify patient response are already available when a new compound enters clinical trial. This is particularly important in patients with rare diseases where large patient cohorts are not available to achieve statistical power using clinical endpoints alone. However, while this synergistic drug-biomarker development approach is well recognised, it has not been achieved for many of the treatments currently entering clinical trials.
In the context of muscle imaging biomarker development some methods are now ready to be tested in clinical trials. Additional techniques (such as diffusion MRI) may offer further value but need to be fully developed and validated before they can be considered for inclusion in forthcoming trials. An effective imaging biomarker must be objective, sensitive, specific and non-invasive. We suggest that no single imaging method will meet all of these criteria, but that a quantitative multi-measurement approach will provide a more complete and sensitive evaluation of muscle damage and repair. The BIOIMAGE-NMD programme therefore aimed todevelop a combined examination based on macro- and micro-structural MRI, the identification of molecular changes in tissue metabolism (through MR spectroscopic imaging, MRSI) and direct visualisation of the AON molecules in vivo using positron emission tomography (PET). The individual value of each measure was assessed in pre-clinical models and, where possible (given the relatively short timescale of the project), will be applied in patients to establish a proof of principle for clinical utility. These approaches can then be fused to form a truly integrated, simultaneous PET-MRI-MRSI protocol, for future use in patients with NMD.
The overarching objective of the BIOIMAGE-NMD project is to deliver combined structural and molecular imaging biomarkers with proven utility for the detection of therapeutic effects in patients with rare neuromuscular diseases (NMD). To achieve this goal, the programme addressed three specific objectives:
Objective 1: To develop a new generation of diffusion MRI methodology optimised for imaging in muscle.
BIOIMAGE-NMD sought to develop new MR based quantitative biomarkers of muscle fibre microstructure using diffusion contrast MRI. In these scans, microstructural sensitivity is conferred through diffusion-mediated interaction between the tissue water, the cell membranes and sub-cellular structures. Tissue injury or degeneration alters the restriction to diffusive water movement and diffusion weighted imaging (DWI) or diffusion tensor imaging (DTI) measurements can reveal and quantify this process, even though conventional radiological T1 or T2 weighted MRI may remain unchanged. DTI technology is used extensively in the brain and current attempts to apply diffusion imaging methods in muscle have largely been adapted from neuroimaging protocols, but have not been comprehensively optimised to account for the differences in cellular size between axons in the brain and muscle fibres or the presence of fatty tissue. Extensions to the DTI method exploiting diffusion mediated contrast have the potential to quantify muscle fibre sizes and detect changes induced by a drug treatment. Objective 1 was to develop these methods as additional new biomarkers of muscle damage. The methods would then be used to augment our state of the art simultaneous MRI / Magnetic Resonance Spectroscopic Imaging (MRSI) protocol for quantitative muscle imaging with the ultimate goal of applying these methods in clinical AON trials in DMD.
Objective 2: To provide a proof of principle that simultaneous MRI/MRSI can be used as a biomarker to monitor therapeutic efficacy in clinical trials in Duchenne Muscular Dystrophy (DMD).
Use of state of the art muscle MRI technologies was proposed to provide a simultaneous structural and molecular approach to follow AON treatment in clinical trials which are already underway or are being planned. MRI is widely available and offers powerful flexibility for non-invasive structural, physiological and molecular investigation. Skeletal muscle fibrosis and fat infiltration are common early features of most NMD due to the continued destruction-regeneration processes. In routine use MRI clearly demonstrates the fat infiltration but visual inspection of these qualitative images is not sensitive to the therapeutic effect of a drug between successive examinations. This may, however, be achieved in humans using quantitative MRI techniques that separate muscle fat and water components, allowing the computation of fat signal percentage maps and identifying inflammatory activity.
Molecular changes are also critical for the comprehensive evaluation of muscle damage. Phosphorus-31 MR spectroscopic imaging (31P-MRSI) is an endogenous molecular imaging technique which shows altered muscle energetics and muscle membrane stability in human dystrophic muscle. 31P-MRSI is a specialist technique but requires only an enhancement of the same scanner technology as used by MRI and so can be effectively combined with the quantitative imaging methods. Simultaneous use of quantitative MRI and MRSI therefore provides a clinically useful structural and metabolic assessment of muscle in patients with NMD, but these techniques have yet to be demonstrated as an effective tool to assess therapeutic intervention.
Objective 2 was therefore to establish a simultaneous MRI/MRSI protocol across the sites of the consortium to deliver a clinical proof of principle (PoP) demonstrating the objective benefit of these combined technologies based on their use in clinical trials of AON in patients with DMD. Objective 2 is closely linked to the outcomes of objective 1 whereby novel diffusion imaging based biomarker technology developed in the first theme would be incorporated within the clinical study as key milestones are reached.
Objective 3: To develop PET molecular imaging of AON in vivo as a technology to advance future innovative drug development programmes for personalised medicines.
We proposed new synthetic chemistry approaches for radio-labelling of the AON molecules themselves enabling PET molecular imaging. While instrumentation for simultaneous structural and molecular imaging by PET-MRI is now commercially available, PET imaging is fundamentally reliant on developing disease specific radiolabelled reporter ligands and, to date, no attempts have been made to develop such radiotracers targeted to disease activity or therapy in NMD. Objective 3 of BIOIMAGE-NMD aimed to make essential advances in the expansion of this technology by developing AON PET tracers to measure bio-distribution in vivo in animal models. Such PET/MRI tools would allow rapid assessment of tissue targeting, biodistribution and pharmacokinetics of AON in vivo, helping to accelerate the development of AON technology for DMD and broadening the applicability of AON technology in other neuromuscular and genetic diseases. In addition, as this technology is translated to humans (beyond BIOIMAGE-NMD) it will form a powerful tool to help define and explain the optimal treatment regimes, mechanism of action and drug delivery for this personalised medicine in humans.
The BIOIMAGE-NMD programme was divided into 8 scientific work packages to achieve these objectives, supported by 2 further work packages around dissemination and management/coordination. Individual objective for each work package were as follows:
WP1 - To implement, evaluate and optimise diffusion weighting MRI schemes to maximise sensitivity to pathological change in muscle and develop appropriate imaging analysis models to quantify changes in muscle fibre dimensions and distributions.
WP2 - To evaluate the sensitivity of muscle optimised diffusion imaging schemes to quantify muscle microstructure in mouse models for muscular dystrophy and define whether diffusion imaging derived biophysical metrics of muscle structure provide new detail in the assessment of NMD.
WP3 - To establish a robust conventional diffusion imaging sequence for use in skeletal muscle.
To develop faster scanning methods for T2 mapping, B1 mapping and Dixon imaging.
To translate and standardize optimised diffusion imaging schemes from pre-clinical to clinical protocols.
To determine whether conventional diffusion imaging provides added value in healthy controls over and above existing MRI methodology (Dixon, T2) and assess the effectiveness of the entire imaging protocol in Becker Muscular Dystrophy patients.
WP4 - To obtain MRI/MRSI data from patients recruited in 4 DMD clinical trials with different AONs and as a comparator group from patients recruited in a DMD natural history study, introducing improved imaging sequences and investigate novel contrasts from WP3 as they are validated.
WP5 - To evaluate the initial human diffusion imaging studies in comparison with the original MRI/MRSI protocol and determine whether diffusion metrics offer new biomarkers.
From this evaluation to decide whether the MRI/MRSI protocol should be updated to include a diffusion measurements in ongoing clinical trials.
WP6 - To centrally collect, curate and analyse the MR images and MRSI spectra generated by the multi-centre clinical trials in WP5 and subsequently identify the most relevant MRI and MRSI indices and their significance relative to other outcome measures.
To develop new processing algorithms for the treatment of the MR imaging data, that will reduce data processing time and improve the quality of the information extracted.
To develop a web-based software solution integrating data collection, quality control and processing dedicated to muscle analysis in multi-centric clinical trials.
WP7 – To develop a synthetic protocol for the preparation of a range of cold (19F)-labelled oligonucleotides, directly translatable for the production of 18F-radiolabelled drug product.
To determine the bioactivity of the cold labelled AONs in vitro and compare the bio-distribution and bioactivity of the cold labelled AON versus the unlabelled AON in mdx mice.
WP8 - To generate Proof of Principle for the use of in vivo PET imaging of radiolabelled AONs as a non-invasive tool to visualize the pharmacokinetics of AON after systemic administration in mdx mice.
To establish the correlation between AON biodistribution data from PET imaging and exon skipping efficacy in diverse skeletal muscle group in mdx mice.
To establish the correlation between AON biodistribution data from PET imaging with MRI based muscle structural data in mdx mice.
Project Results:
The BIOIMAGE-NMD consortium was formed to address significant challenges in evaluating new therapeutic treatments in clinical trials in rare neuromuscular diseases. Sensitive and validated outcome measures and biomarkers of drug action are required which can be effectively applied in small cohort groups, particularly in children. Imaging methodologies were already identified as important non-invasive tools, but further evaluation of existing techniques was required. Further, magnetic resonance based approaches were identified which had not been optimised for muscle pathology, but which offered exciting potential as novel outcome measures. The BIOIMAGE-NMD programme addressed the clinical trial potential of advanced and quantitative imaging through an ambitious and complete translational development of methodologies from in vitro and early pre-clinical testing, through implementation in the clinical environment and evaluation in healthy subjects, first-in-man observational studies in neuromuscular disease and ultimately adoption into clinical trials of exon-skipping in boys with Duchenne muscular dystrophy. This translational development theme ran alongside evaluation of existing imaging methodology which was incorporated into the trial protocols of the earliest AON clinical trials to assess proof of principle of imaging biomarkers. Finally, the programme sought to develop new molecular imaging methods through new synthetic chemistry approaches to radiolabel the AON molecules for imaging via positron emission tomography (PET).
To achieve these objectives BIOIMAGE-NMD assembled a Consortium combining SME expertise in drug development with academic expertise in experimental (MRI) and theoretical physics, imaging science, chemistry, neurophysiology and clinical neurology. The commercial focus for SMEs are significantly different to those in academia, with pressures derived from business performance. By their nature, SMEs are responsive to their commercial environment, and highly successful SMEs will always be at risk of take-over or merger by larger organisations. During the lifetime of the BIOIMAGE-NMD programme there were two significant events with our main SME partner, but through careful management of the work plan these did not impede progress. Early in the programme, the clinical trial (outside of BIOIMAGE-NMD) of the first AON from our SME partner Prosensa was halted due to lower than expected clinical efficacy. The clinical trials at which the imaging developments within BIOIMAGE-NMD were targeted were therefore delayed while the data from this first trial were analysed in detail. Subsequently, after the next phase of trials had commenced Prosensa was the subject of a corporate acquisition by BioMarin who took over the role within the programme, but after a further 12 months determined to halt all therapeutic development and clinical trials for neuromuscular disease groups. Through careful management and the ongoing support from BioMarin, these impact of these factors on the overall outcome of the BIOIMAGE-NMD programme was minimised.
The programme worked towards 10 deliverables, with an associated 46 milestones across the full set of 10 work packages and achieved the vast majority of these targets. Below we describe in detail the main results from the scientific Work Packages along with key Deliverables and Milestones.
No new foreground was generated during the lifetime of the project.
Work Packages – in detail
WP1 - Optimising diffusion imaging for detection of changes in muscle tissue microstructure
Summary of Achievements
Diffusion imaging methods for muscle MRI were developed based on the stimulated-echo (STE) preparation methodology and evaluated in the hindlimb muscles of wild type and mdx mice (a transgenic model for the clinical condition Duchenne Muscular Dystrophy). Comprehensive measurements were made in vivo, repeated in the same tissue post mortem and compared to histological findings in the same tissue. The STE sequence allowed flexible variation in the diffusion time, which is a key sequence parameter influencing the probability of interaction between diffusing water molecules and the local tissue microstructure. Diffusion times were studied between short (20-50ms, as normally used in brain investigations) and very long (750ms). The imaging data demonstrated the expected restricted diffusion behaviour, indicating that the sequence was sensitive to muscle microstructure. Highly significant differences in water diffusivity (apparent diffusion coefficient, ADC) were observed between the animal groups with the greatest differences found with much longer diffusion times than are typically used in clinical scanners.
Computational modelling was developed to investigate the sensitivity of diffusion MRI to key features of pathological muscle. Monte-Carlo simulations of diffusing spins in hierarchical structures were implemented, including explicit modelling of the effect of diffusion-attenuating pulse sequences. Muscle tissue was modelled as parallel cylinders with radii drawn from gamma distributions fitted to histological data of mouse forelimb muscle tissue derived from the literature and our preclinical studies. Each cylinder contains a population of random close-packed cylinders with smaller, but also gamma-distributed, radii as a model of hierarchical muscle tissue structure covering six orders of magnitude in size. This is the most detailed models of muscle tissue simulated in the diffusion MRI literature. Simulations were then used to synthesise diffusion measurements over a broad range of scan parameters in five separate microstructural scenarios:
- histology of healthy tissue,
- atrophy, where all internal fibres are reduced in size,
- dystrophic tissue, using the muscle fibre radius distribution derived from our mdx mice,
- increased membrane permeability,
- atrophy where a fixed percentage of fibres are removed.
These data demonstrated that permeability is the dominant effect in changing the form of the diffusion decay curve. It also found that although all scenarios considered could in principle be detected at clinically-feasible SNR, the effects of each change were difficult to disambiguate when considered simultaneously.
Combining the knowledge derived from the computational modelling with the experimental observations of diffusion change in mice, we note that quantitative estimation of our target outcome parameter (muscle fibre radius, or a metric of muscle fibre radii distribution) is extremely challenging. While the diffusion signal contains important radiological contrast which does reflect differences in microstructural environment, this contrast is not invertible to reveal detailed geometric parameters. Diffusion contrast is potentially a unique marker which is not exploited in conventional approaches and may however be used as a biomarker in its own right.
Key Deliverables and Milestones
D1.1: “A new state-of-the-art protocol for quantitative diffusion imaging in muscle”, was documented using the simulation software and data fitting approaches to define a parameter space (diffusion time and diffusion b-values) which provided an efficient diffusion scan acquisition protocol optimised for muscle imaging.
Development of the muscle water diffusion simulation framework was a major milestone for WP1. This software has been incorporated into the Camino Diffusion MRI toolkit which is available as open-source code and scripts for running the simulations will be made available to the scientific community on request.
WP2 - Evaluation of diffusion imaging methods in mouse models for muscular dystrophy
Summary of Achievements
The STE diffusion imaging methods developed for mouse muscle MRI in WP1 were established on small animal MRI scanners at the two preclinical imaging sites in the Consortium (UNEW, LUMC). Scanner vendor differences in MRI sequences were overcome and scanner performance was harmonised to yield excellent reproducibility of quantitative diffusion data from test objects and subsequently in vivo. Diffusion imaging was then used to study restricted water self-diffusion behaviour in mouse hindlimb muscles in a range of in vivo studies and animal models. Scanning evaluated the effect of disease through comparisons of wild type and transgenic models of neuromuscular diseases (mdx, mdx-xist, ± exercise intervention to increase muscle pathology) and of ageing (7-8 weeks, 22 weeks and 44 weeks), with imaging data compared to histological assessment of muscle fibre size distribution and permeability.
Diffusion weighted imaging, data showed small but significantly increased diffusivity at long diffusion times (>100ms) in young mdx animals compared to wild type (8-16 weeks of age), while in studies of older animals (22 and 44 weeks) there was a more marked increase in diffusivity in the mdx groups. Conversely, pathological assessment of the muscles in these animals revealed that the mean muscle fibre size reduced with increasing age in the mdx animals. In impermeable fibres, such a reduction in fibre radius would cause greater restriction to diffusion movement and hence reduced diffusivity, contrary to our experimental observations. The observed increase in diffusivity is most likely explained by an increase in permeability of the muscle tissue to water diffusion, which was supported by observation of greater intra-fibre uptake of Evans Blue dye in the mdx muscles. Measurements using diffusion tensor imaging also suggested elevated diffusion in dystrophic muscle, but magnitude of the effect was more subtle.
Muscle T2 mapping was also performed in the same animals, as this parameter is already used in clinical investigations and is also sensitive to water mobility. Elevated T2 values were found in dystrophic muscle (mdx) with a greater dispersion in values than in wild type muscle, however no significant relationship was found between T2 and diffusion metrics, suggesting that diffusion and T2 measurements have different sensitivities to muscle pathology. We conclude that both T2 and diffusion are potentially useful biomarkers for muscle damage in animal models of neuromuscular diseases.
A novel framework for fitting advanced models to the diffusion MRI data has also been developed based on anomalous diffusion modelling. This is an entirely new C++ framework which fits models derived from Fractional Diffusion to diffusion-weighted data. To achieve a fitting approach which does not require users to specify initial starting conditions for fitting, a method was developed fitting a chain of models of increasing complexity, each of which provides the initial parameter guesses for the next model such that all initial parameter guesses are derived from the input data. The fitting code framework was shared with project partners for application to the multi-site preclinical) and clinical (WP3) acquisitions. We have applied this anomalous diffusion approach for the first time in muscle imaging and created and demonstrated that this analysis model is sensitive to changes in muscle fibre distribution when tested against synthetic, in vitro and in vivo data.
Key Deliverables and Milestones
D2.1: Report on the value of using diffusion imaging to assess muscle pathology, in comparison to T2 mapping methods in mouse models of NMD.
The advanced data fitting framework will be made available as open-source code o the wider research community once the findings from the BIOIMAGE-NMD studies have been published.
WP3 - Evaluation of diffusion MRI methods for assessing muscle pathology in NMD
Summary of Achievements
This work package addressed a large number of technical challenges around the translational step between animal and human imaging and potential to use our new diffusion methods in clinical trials.
As our target clinical trials were in children with Duchenne muscular dystrophy there was a need to shorten the imaging protocol both to improve compliance (leading to higher quality data) and to allow protocol time to include the new diffusion measurements. Our conventional clinical trial imaging protocol used 3 independent scan acquisitions to quantify muscle fat fraction and muscle water T2. Fat fraction is extracted from the Dixon acquisition, while muscle water T2 quantification required both a multi-spin-echo (mSE) acquisition and an associated B1 map to apply corrections for imperfections in the MRI RF excitation pulse slice profile and uniformity across the body. Using advanced reconstruction (extended phase graph, EPG) of the mSE data combined with biexponential modelling of the signal decay function, we demonstrated that fat fraction and T2 could both be extracted from the mSE data alone. Importantly, there was no correlation between T2 and fat fraction values using this method which demonstrates independence of the extracted quantified values. We evaluated this approach in a range of patients with varying NMD in comparison with the standard 3 scan approach and established that the EPG fitting derived fat fraction correlated well with Dixon derived measurement (≈1.5% bias). The new approach eliminates the need for separate Dixon and B1 mapping sequences and reduced the protocol time by ~20 minutes.
Establishing a robust diffusion imaging sequence for use in skeletal muscle required addressing the
challenge posed by fat infiltration which is a major feature in skeletal muscle of patients with NMD. Incomplete fat suppression in imaging voxels containing both fat and water, or voxels where the fat signal is misregistered due to chemical shift displacement both result in inaccurate diffusion quantitation. A new and highly effective method was developed to suppress fat by combining frequency selective suppression pulses for the large lipid signals (~1.5ppm) with a Dixon approach for the smaller olefinic fat signal. The new method was successfully implemented as part of a STE diffusion sequence on clinical MR systems from two of the main vendors. The full fat suppressed STE diffusion sequence was used with a diffusion tensor encoding scheme and intra- and inter-site performance was evaluated across 3 sites of the Consortium using measurements in the same cohort (n=5) of healthy controls travelling between scanners. A post-processing pipeline was established with algorithms to separate the water signal from any residual fat, eliminate spatial image distortion and calculate pixelwise diffusion metrics in muscle. Within a site mean diffusivity (MD) data had good reliability (maximum SD across the group of <10%), but there were differences in values between sites of 15-20%, (present as a systematic offset in all diffusion values). Analysis of the diffusion data using conventional diffusion metrics showed that tissue fractional anisotropy (FA) values were overestimated at low signal to noise ratio (SNR). A minimum SNR of 7 was determined as the threshold which constrained controls FA and MD results within a 5% confidence interval.
The final phase in this work package was to evaluate the new diffusion methodology in muscular dystrophy patients. A protocol combining both standard (spin-echo) and advanced (STE with two diffusion times, 100 ms and 300 ms) diffusion sequences was used in 10 patients with Becker Muscular dystrophy (BMD). Although fat infiltration was suppressed by the new scheme, the residual signal from remaining muscle water was still reduced simply due to volume loss of muscle within each imaging pixel, leading to lower SNR. In BMD patient muscles our SNR threshold for robust FA calculation (SNR>7) was not fulfilled in 50% of ROIs, corresponding to the muscles with highest fatty infiltration. Of all muscles sampled within the leg, the tibialis anterior was the only one to meet the SNR criterion. Here FA was significantly higher in muscle of BMD patients (n=5) as compared to healthy controls (n=4) (p=0.016). MD values were not significantly different in BMD compared to control in any muscle but tended to be lower in soleus (p=0.057). These results show that it is feasible to perform STE-DTI in BMD patients, and thereby to take advantage of extended mixing times for muscle cell specific applications. This study highlighted two other significant observations; SNR cannot be determined from scan parameters alone, but depends on the degree of disease related muscle loss and secondly the trend to disease related decrease in MD is in opposition to the changes seen in animal models. The origin of the changes in human muscle are unclear but interpretation based on our animal data indicate they are a complex balance between smaller fibre sizes, possibly in competition with increased membrane damage and also may involve development of fibrosis. Our overall interpretation of the BMD study is that in muscles such as the TA which is relatively spared by fatty infiltration, we anticipate that the use of mixing times of 100ms and above will allow early detection of changes in DTI metrics.
Key Deliverables and Milestones
D3.1: Report on the new State of the Art (comparison of conventional and advanced methods) for assessment of human muscle by diffusion MRI.
In this WP we achieved the major milestone of first-in-man assessment of diffusion imaging in NMD patients.
WP4 - Proof of principle of MRI/MRSI biomarkers in clinical trials in DMD
Summary of Achievements
The conventional MRI and MRSI protocol was successfully implemented in all clinical trials within the BIOIMAGE-NMD project and a leg holder was sconstructed for use at each site. Study-specific imaging manuals were written and sites were trained in the use of the protocol before being validated to take part in the clinical trials through quality control data in a phantom and subsequently in healthy volunteers. To underwrite data quality during the longitudinal investigations all sites were required to rescan phantoms every 6 months and healthy volunteers every 2 years.
Imaging protocols were then used in the AON intervention clinical trials in DMD patients alongside the routine clinical and functional evaluations. All imaging data in DMD patients was acquired using the unmodified conventional imaging protocol defined at the outset of the BIOIMAGE-NMD programme. To preserve data consistency in ongoing clinical trials, the revision of trial protocols to introduce fast scanning or diffusion imaging developments (WP3) were only ever planned to coincide with new clinical trials. Commercial and clinical factors during the lifetime of the BIOIMAGE-NMD programme led to a departure from the original clinical trial phasing such that availability of more advanced imaging methods did not map to the revised clinical trial schedules.
A total of 52 patients enrolled in the clinical trials were scanned using the conventional protocol. Scans were collected in the initial PRO053 and PRO044 trials. The anticipated global roll out phase for PRO053 did not take place so MRI data was added to the PRO051-DMD115501 extension study as part of the original agreed global roll out to sites beyond the Consortium. The PRO051-DMD115501 trial included scans at baseline, 6, 12 and 18 months, while for PRO044, baseline and 6-month MRI exams were acquired. For PRO053, baseline scans and several follow-up visits (between 3 and 30 months) were acquired. As a comparator group, natural history data (PRO-DMD-01 study) was obtained in a further 23 subjects who were scanned at baseline, 6, 12 and 18 months. A total of 220 patient data sets were acquired across all studies and were available to evaluate the value of imaging as outcome measures in WP6.
Key Deliverables and Milestones
D4.1: Status update on the imaging data collected in the clinical trials and natural history study.
WP5 - Decision on the use of novel imaging methods in the PoP clinical trial
Summary of Achievements
This short work package evaluated the scientific evidence in relation to the performance of the novel imaging methodology developed in work packages 1 to 3. This WP took the form of a data review and discussion between all partners of the Consortium to debate the merits of the new methodologies and arrive at a consensus on their value for clinical trial evaluation. Specifically, for the fast scanning methods developed in WP3, the criteria required evidence that the new methods could be introduced to replace the conventional approaches without introducing bias or systematic errors in the quantified data. The methods should also be appropriate for clinical use and must be achievable on the network of systems across the clinical trial sites. The EPG and biexponential analysis approach of the mSE scan data was considered to be a significant step forward which met both of these criteria. As the method relies solely on post-processing it could be applied, with appropriate preparatory work on test object and control subject scans, to data from all sites. The scientific work underpinning the method demonstrated low bias.
The advanced diffusion methodology was considered to be exciting and to provide interesting and potentially valuable new indices of muscle pathology. The opposing observations between animal data (WP1 and WP2, where histological correlates were available) and BMD patients (WP3, where no histology was available) did not provide a full understanding of the sensitivity of the measures in human subjects. Further work is clearly required to elucidate the complete biophysics of diffusion processes in healthy and diseased muscle. The diffusion methodology was therefore not considered sufficiently mature for clinical trial use.
Key Deliverables and Milestones
D5.2: Report detailing the consensus decision on the performance and quantitative compatibility of new fast scan T2, B1 or Dixon methods with original state of the art methods in the MRI/MRSI protocol.
A summary of the consensus statement in Deliverable Report D5.2 is as follows:
1) Diffusion imaging methods show promise as a tool to detect alterations in muscle microstructure with disease progression. At the current time there are discrepancies between the observations in animal models and human subjects and further work is required to define an underpinning theoretical model which reconciles the experimental data. At the current time diffusion imaging is not recommended as a putative outcome measure.
2) T2 mapping is a valuable measurement to detect the presence of muscle edema associated with inflammation. T2 does not itself provide a marker for disease progression and can be modified acutely by treatments such as steroids. The precise value of T2 measurements in clinical trials depends on the nature of the treatment and the specific disease being studied.
3) Neuromuscular imaging protocols seeking to measure muscle water T2 should use EPG analysis of the T2 data, which eliminates the need for B1 map data collection and reduces the time burden on patients, particularly in studies involving children.
WP6 - MRI and MRSI data analysis
Summary of Achievements
This work package addressed the many challenges of centralised analysis of clinical trial imaging data, specifically
- Developing a method for secure file transfer from trial sites to the central team,
- Developing time-efficient, semi-automated analysis tools for robust muscle segmentation
- Performing the analysis of the clinical trial data acquired in WP4.
We report our achievements for each of these tasks below:
A) A web-based tool for secure file transfer
We developed a client/server web-based application allowing the imaging centres to transfer the data in an easy way and to perform quality control checks at site level (such as verification of subject IDs, correct labelling of the data, completion of necessary NMR data). The processes are automated and fast at both ends (imaging centres and central data analysis centre). At the level of the central reader, quality control of the data can be performed (such as accurate Dixon reconstruction) and segmentation of the muscles is integrated using the semi-automated algorithm developed in this work package. Processing of the data can be performed in batch. Great care was given to achieve GCP compliancy, which guided the development of several features (data tracking, encrypted archiving, secured login, code versioning). A first version of this web-based service is being tested at the time of writing this report. The full-scale deployment is not yet activated for a lack of maturity, but the developing partner, CRIS is aiming for this last feature to be ready by the first quarter of 2018.
B) A semi-automated tool for muscle segmentation and statistical analysis
The extraction of muscle fat fraction and water T2 values requires definition of region of interests (ROI) delineating each muscle at each slice of the imaging dataset which is a major bottleneck in clinical trial data analysis. To address this challenge, a semi-automated, muscle segmentation algorithm was developed. As the main challenge to automatic segmentation is detection of muscle boundaries in the present of high levels of fat infiltration the algorithm was validated on scans from the lower-legs of healthy volunteers and Duchenne patients. The algorithm was compared with the gold-standard manual approach using similarity metrics: volumetric similarity (1 – absolute volume difference divided by sum of compared volumes), Dice coefficient (overlap measurement), Hausdorff distance (maximal distance between any points on each of the segmented volumes).
Manual segmentation times on the volunteer cohort were 35 min per muscle on average (x12 muscles = 7 hours per subject). The mean segmentation time with the new tool was 2 min, yielding an acceleration factor of 17.5. On the Duchenne patients, the acceleration factor was less (mean 7 mins) due to the smaller number of slices to segment. On the muscle groups of the healthy volunteers, the mean volume similarity was 0.96 mean Dice coefficient was 0.90 and the mean Hausdorff distance was 10.75mm. On the Duchenne patients, similar to better results were found showing that the fatty infiltrations did not hinder the algorithm performance.
C) Analysis of the clinical trial data acquired in WP4
The analysis protocol required that all subject scans were quantified to derive the following MRI outcome measures:
- fat fraction in each individual muscle and muscle groups,
- cross-sectional area and contractile cross-sectional area in muscle groups,
- water T2 mapping in muscle.
Datasets from expert sites where 31P-MRSI data was available were quantified to obtain spectral indices of tissue energetic and metabolic state (pH, PCr, Pi and PDE levels) in gastrocnemius muscle.
The 220 patient data sets included Dixon, MSME (T2), and B1 sequences for the lower limbs (both legs and thighs). 31P MRS data was obtained in boys enrolled in the PRO053 study scanned in Paris, Newcastle and Leiden, from boys enrolled in the PRO044 study scanned in Leiden and in the natural history (PRO-DMD-01) also scanned in Leiden. A total of 25 31P MRS exams were obtained for PRO053, 6 exams for PRO044 and 8 exams for the natural history study. Clinical and functional data acquired by the consortium partner BioMarin (formerly Prosensa) within the main clinical trial datasets was made available during the final 3 months of the programme to allow analysis alongside the imaging data. The clinical data provided the genetic mutation of all subjects in the natural history study, most of which were exon-44, exon-45, exon-52 or exon-53 skippable with only one of the subjects being exon-51 skippable. Whenever possible, quantitative findings from the MRI/SI data were correlated to functional data (particularly the 6-minute walk distance as one of the main standard clinical measures accepted as a primary outcome measure for clinical trials in NMD patients).
Data from the natural history study (PRO-DMD-01) and the PRO051-DMD115501 clinical trial comprised the largest part of the MRI/S data within the BIOIMAGE-NMD project, enabling us to perform a comparative analysis between the two patient groups and examine the longitudinal data for possible therapeutic effect in the treated group. After quality checks, datasets from 3 of the 6 sites in the PRO051-DMD115501 were excluded due to incomplete (missing paired datasets between baseline and follow-up) or poor data quality, leaving 24 patient datasets with 82 imaging examinations. All data from the natural history study was included (23 patient datasets with 63 scans). The observations we draw from these patient datasets are as follows:
i) Observations relating to Fat Fraction and Treatment Effect
The trajectories of fat fraction change over time were examined for each muscle and considering progression as a function of patient age. While each patient was only scanned over a short duration, the range of ages at enrolment spanned 5 to 15 years such that the combined trajectory plots across all subjects could be used to observe the overall shape of the progression, which typically followed a sigmoidal curve. When looking at the trajectories in more detail we observed that patients of the same age have significantly different amounts of fatty infiltration at their baseline scans. Moreover, when looking at two patients of the same age and with almost the same amount of fat infiltration at baseline, we saw a different degree of fat fraction increase in the following year (i.e. there were different slopes to the individual subject trajectories). Despite these clear differences between individual subjects, groupwise comparison of the average fat fraction increases at 1 year intervals between a carefully matched subpopulation of treated (DMD115501) and untreated (natural history) subjects showed a significant 51-exon-skipping effect, with treated boys showing a significantly smaller increase in fat fraction than the untreated boys (p<0.01).
In the PRO053 and PRO044 studies subject numbers were small (n=9 in both groups) with typically only the baseline scan and 1 or 2 follow-up scans. The PRO044 subjects were significantly older at study inclusion than subjects in the other trials, had lower 6-minute-walk-distance values (related to their older age) but showed no real fat fraction changes after 6 months. Two of the PRO053 subjects showed a steep increase in fat fraction over the course of two years whereas the other subjects had stable values across the course of the study. These two subjects were untreated whereas the others received treatment. Their respective 6-minute-walk distances decreased accordingly. Subject numbers were too small to allow meaningful statistical analysis or to make any conclusion beyond a qualitative description.
ii) Observations relating to muscle water T2
Muscle water T2 reflects disease activity and is a sensitive yet unspecific biomarker of the inflammatory/necrotic/oedematous processes. Muscle water T2 is also sensitive to steroid treatment.
Contrary to fat fraction, water T2 is prone to fluctuations (high water T2 values in active disease while low water T2 in more passive phases of the disease), making its interpretation more complex. The threshold for determining a ‘pathological’ water T2 value depends on site-specific elements such as acquisition parameters and field strength, requiring normalization of the data. To achieve a robust assessment of patient data we restricted analysis to data from a single site where both treated and untreated subjects were scanned (thus same scanner, same field strength, same acquisition parameters). We observed generally lower water T2 values (only significant in leg muscles) in the DMD115501 subjects compared to the natural history subjects. Water T2 values also seemed to increase after 1 year in treated patients as compared to the untreated patients.
iii) Observations relating to Fat Fraction and Disease Progression
Six minute walk distance data was available for subjects at baseline and the first 6 month follow-up assessment. Across patients in all trials, there was a clear inverse relationship between 6MWD and muscle fat fraction. For example, in the hamstring muscle subjects with low fat fraction (<10%) had 6MWD of 350-525m, while subjects with fat fraction >10% had much reduced 6MWD. From the longitudinal data, the general trend was observed that increasing fat fraction was associated with decline in 6MWD in the individual subject, while stable fat fraction was associated with little change in 6MWD. Some individuals did show increased 6MWD at the follow-up time point which may reflect the influence of subject motivation on functional performance, age related effects (increasing muscle bulk with maturation balanced against disease progression) as well as possible treatment effect.
iv) Observations of metabolic effects by 31P-MRSI
Phosphorus-31 MRS data were available in patients scanned within the main consortium sites which included subjects in the PRO053 and PRO044 studies. These patient showed increased pH and PDE values compared to controls.
To summarise the findings from the BIOIMAGE-NMD Proof of Principle clinical trial studies, we have confirmed that fat fraction as derived from Dixon MRI is a valid biomarker of disease severity. Clear correlations were found between the loss of real muscle mass (as determined by MRI) and functional outcome measures such as the 6-minute walk distance. A comparison of fat fraction between AON treated PRO051 patients with natural history data did illustrate a treatment effect. However, based on the available data, fat fraction does not seem to be predictive of disease evolution. The diverse phenotypic expression of the disease as well as its non-linear progression on an individual basis, hamper the establishment of prognostic indices based on a single fat fraction determination, even when age is taken into account. Results of other MRI/S biomarkers such as water T2 and 31P MRS indices are even more difficult to interpret because of multiple reasons. Muscle water T2 reflects the disease activity and therefore needs short-term evaluation of its fluctuations on a subject-to-subject basis in order to determine its discriminant power when comparing pre- and post-treatment. 31P MRS results were scarce due to the limited number of sites who were able to acquire these kind of data and more data is needed in DMD.
Key Deliverables and Milestones
D6.10: Summary & conclusions of the analysis of the different outcome measures generated during the study (M42)
WP7 - Selection of the prosthetic group and labelling strategy for 18F radiolabelling of AON
Summary of Achievements
We have developed synthetic protocols to prepare [19F] labelled oligonucleotides as a first step in production of [18F] radiolabelled drugs for PET imaging. To couple the fluorine-18 radiolabelled prosthetic group to the antisense oligonucleotide (AON) it was necessary to introduce a reactive functionality into the structure which is achieved by incorporation of an appropriate ‘modifier’ at the end of the conventional AON production process. The aminooxy moiety is typically used due to the selectivity, and efficiency of the process. Aminooxy modified AON was therefore prepared and then two different prosthetic group strategies were explored for ‘tagging’ the modified AON to create a [19F]-AON. These different prosthetic groups led to very different stabilities and yields. The most efficient group was selected for further development. Multiple production runs of the [19F]-AON were carried out demonstrating the robustness of the process. Labelled [19F]-AON could be produced in high milligram quantities with high purity (97%) allowing in vivo assessment of bioequivalence of the original and labelled AON compounds.
The [19F]-AON or native AON were administered into mdx mice using a protocol defined to establish the relevant biological data on a timescale to match the PET study which is limited by the fluorine-18 half-life. Mice were given a single injection and tissues (skeletal muscle groups, heart, liver, kidney and spleen) were isolated 4 hours post-treatment after perfusion of the mice. Plasma samples were also acquired throughout the 4 hour period. AON levels in plasma and tissues were quantified by an ELISA. A comparable biodistribution pattern was observed for both the labelled and unlabelled compounds, which demonstrated that incorporation of the fluorine label into the AON did not alter AON pharmacokinetic profile.
Key Deliverables and Milestones
D7.1: In vitro/in vivo bioactivity data of the two ‘cold’ labelled materials
WP8 - Evaluating PET/MRI imaging protocols in mdx mouse models
Summary of Achievements
In this second work package associated with radiolabelling of the AON molecule, the methodology developed using cold [19F] labelling was modified to use the positron emitting isotope [18F] to create the [18F]-AON compound for use in PET imaging studies. The automated preparation of the [18F]-AON was achieved using a Modular Radiosynthesis platform. Reaction conditions and purification methods were optimised to achieve activity/volume ratios suitable for pre-clinical use. GMP compatible purification and QC protocol have been developed for the [18F]-AON using both high pressure liquid chromatography (HPLC) methods and single phase extraction (SPE) methods to ensure the methodology can be translated in the future (beyond BIOIMAGE-NMD) to typical commercial automated radiochemistry platforms used by clinical PET facilities. Residual solvent analysis demonstrated levels that are below the limits required for clinical use.
The first biodistribution studies using PET imaging [18F]-AON were carried out in a small number of mdx and wild type mice. Scanning was performed using PET-MRI and demonstrated the principle of detecting this new class of fluorine-18 radiolabelled AON’s in vivo. The use of a PET-MR scanner for this study has also highlighted that both PET and MR data can be obtained from a single subject. The MRI acquisition however only considered standard structural imaging and the other advanced quantitative MR methods developed in work packages 1-3 were not collected. The relationship between the PET and MRI methodologies remains to be developed. The fundamental concept of PET/MR studies for the evaluation of AON therapies has been demonstrated and as such this milestone has been completed.
Key Deliverables and Milestones
D8.1: Report on application of PET/MRI imaging to correlate AON levels with muscle structure
This report was the only deliverable which was not achieved across the full programme. Work to deliver this activity was not completed within the timeframe of the project due to 2 main factors, both around staffing at the two main partners in this activity. The key post-doctoral research worker who was instrumental in establishing the PET labelling methodology was recruited to another institution and departed the programme with 3 months of work remaining, leaving insufficent time to find a replacement to complete the work before the end of funding. In parallel, the commercial decision by BioMarin to halt clinical trials of the AON agents fed through to changes in their programme of work around AON development limiting activity within the second partner who were to conduct the ELISA analysis alongside the PET imaging. In combination, these factors prevented the correlation analysis which was the topic of the Deliverable Report. While the full report could not be delivered, the Periodic Report for the final (3rd) period does describe significant advances and proof of principle for the technology.
WP9 - Dissemination
Summary of Achievements
Initial dissemination activities were focussed on the building of the website and intranet as well as establishing publication guidelines. From the outset, the Project Office worked closely both with consortium partners, in particular BioMarin and their continued contact with patients and patient organisations to raise awareness of the project. As the project has progressed, dissemination activities have reflected the increased research activities in the different work packages, resulting in a substantial numbers of publications and conference articles relating to BIOIMAGE-NMD.
Alongside publications and abstracts, the BIOIMAGE-NMD consortium was involved in the organisation and hosting of an international conference, held in London on 26th May 2016, entitled “'Making outcomes work' - Stakeholder Workshop on outcome measure development and implementation for DMD”. The coordinator from Newcastle University, along with many representatives from the BIOIMAGE-NMD consortium partners including WP leads from WP1, WP2, WP3, WP5 and WP6 attended. In January 2017, BIOIMAGE-NMD also took part in the international conference entitled “'Making outcomes work' - Stakeholder Workshop on outcome measure development and implementation for DMD”. The coordinator from Newcastle University, along with many representatives from the BIOIMAGE-NMD consortium partners including WP leads from WP1, WP2, WP3 and WP6. WP Leads, including the coordinator of BIOIMAGE-NMD, were also invited to present on BIOIMAGE-NMD at the Newcastle University Institute of Genetic Medicine Seminar Series.
Key Deliverables and Milestones
D9.1: Plan for the dissemination and Use of Foreground
The PDUF was created at the start of the project and the Deliverable was submitted on time. The PDUF was kept updated throughout the lifetime of the project.
WP10 - Project Management and Co-ordination
Summary of Achievements
Overall, the project has been managed well, with 6-monthly Partner meetings as well as 3 monthly Scientific Committee teleconferences. The Project Office worked closely with all partners to ensure progress of key deliverables and milestones and has worked closely with the EU Scientific and Legal Officers to coordinate and resolve complicated issues relating to project partner engagement in Work Package delivery. These issues were solved satisfactorily and have led to an amendment of the original DOW with distribution of budget and additional and/or amended milestones.
Financial reporting was coordinated via the Newcastle University EU Finance Office and all partners have followed EU guidelines on submitting the relevant (audited) reports when required.
Key Deliverables and Milestones
D10.1: Final reports will be delivered to the Commission within 60 days after the end of the project
All periodic reports as well as the Final Report were delivered on time to the Commission.
Potential Impact:
- Quantitative Imaging as Biomarkers for Clinical Trials in Rare Diseases:
Neuromuscular diseases include a range of highly heterogeneous, frequently devastating rare and ultra-rare conditions which affect both sexes and all age groups across the world. A conservative estimate of the overall prevalence of NMD indicates that 1 in 3500 of the population may be expected to have a disabling inherited NMD presenting in childhood or later in life. In the context of BIOIMAGE-NMD we have focussed on patients with Duchenne muscular dystrophy which is characterised by severe, progressive and irreversible loss of muscle tissue and function and affects 1 in 5000 male births. Affected children are typically diagnosed at about four or five years of age, when it becomes apparent that their motor abilities are restricted. Without treatment most require the use of a wheelchair around the age of 9.5 years. Extreme muscle weakness results in orthopaedic problems and patients also suffer respiratory and cardiac complications. Death often occurs by the late teens or early adulthood, mostly due to respiratory or cardiac failure. The goal of existing management is to minimise the impact of these complications as there are no curative treatments for any NMD. The costs of NMD conditions, in terms of health economics, is enormous. It is estimated that in 2010 across 30 European countries, €8 billion was either spent on health care of NMD or lost due to reduced productivity (Gustavsson et al., Eur. Neuropsychopharmacology, 21, 718, 2011).
Currently, many companies and academics are trying to develop different types of therapies for these rare diseases, including the exon-skipping AON therapeutics studied within BIOIMAGE-NMD and phosphorodiamidate morpholino oligomers (PMO). However, they all face the same continuing need for sensitive and validated clinical trial end points. The 6 minute walking distance (6MWD) test remains the main accepted primary end point in NMD clinical trials. As secondary outcomes the North Star Ambulatory Assessment (NSAA), measurements of the muscle strength using hand held myometry and timed tests are used. However, it is not possible to demonstrate a statistically significant result with the 6MWD test on the small numbers of patients available in rare disease clinical trials. Within the BIOIMAGE-NMD consortium we have spearheaded the application of non-invasive imaging technologies as biomarkers which can meet this unmet need. BIOIMAGE-NMD has developed an array of tools based around MRI and PET imaging which could in the future eliminate the need for muscle biopsies in a patient population where muscle function is already compromised. In contrast to biopsies, imaging biomarkers will also allow us to look at a much wider range of muscle types in the patients and assess muscle types that normally are not prone to biopsies such as the muscles of the neck. All of our imaging biomarker developments have been targeted at quantitative measures, with the aim to provide maximal sensitivity to detect disease progression and changes induced by therapy.
The work within BIOIMAGE-NMD has confirmed that fat fraction as derived from Dixon MRI is a valid biomarker of disease severity and disease progression. Changes in muscle fat content measured by imaging map strongly to the reduced functional outcome measures such as the clinically accepted 6MWD. While the 6MWD provides an overall measure of ambulatory function in the patient, the imaging measures provide a much more detailed assessment of the muscle-by-muscle progression of the disease. Importantly, while BIOIMAGE-NMD has concentrated on muscle imaging in the thigh and leg, the Dixon and T2 mapping methodologies are entirely applicable in the upper limbs. Thus, while the 6MWD loses meaning in patients who have lost ambulation, the imaging based measures offer the opportunity to monitor progression and response to intervention through upper limb assessments. Non-ambulatory NMD patients do not then need to be excluded from clinical trials based on the design of the outcome measure assessments.
Our ability to achieve the objective of proof of principle use of MRI methods as biomarkers to monitor therapeutic efficacy in clinical trials in neuromuscular diseases has been slowed by a combination of factors. Firstly, the complexity and diversity of phenotypic expression of the disease as well as its non-linear progression on an individual basis, hamper the establishment of prognostic indices based on single fat fraction determination, even when age is taken into account. Secondly, the clinical efficacy of the first generation AON therapeutics which were studied as the clinical trial exemplars for detection of effect were lower than anticipated (a factor which led to the early closure of trials and withdrawal of these compounds from licencing applications to the regulators). While efficacy was difficult to detect using the standard clinical outcome measures, a detailed analysis of the imaging data did demonstrate differences in fat fraction progression between AON treated PRO051 patients and untreated patients in the parallel natural history study. Clearly, further work is required to understand disease progression in patients with NMD and how to select or stratify patients for clinical trials. The data from BIOIMAGE-NMD adds to our current level of understanding and continues to support the potential of imaging biomarkers. We are convinced by our own data that imaging biomarkers do offer sensitivity to detect changes induced by therapy much earlier than the conventional outcome measures, which would fill a real gap in the development of new therapies in (rare) neuromuscular diseases. Indeed, the experience and knowledge gained from the clinical trial imaging in BIOIMAGE-NMD has already shaped trial protocols being pursued by members of the consortium with other SME partners. As next generation AONs, or alternative approaches (exon-skipping or beyond) come through the drug development pipeline, imaging biomarkers will continue to play an important role in patient assessment.
- Developing advanced tools to enable clinical trials with imaging biomarker endpoints
At the commencement, and throughout the lifetime of BIOIMAGE-NMD there has been much commercial interest and activity around using imaging to monitor progression and potential therapeutic response in NMD trials (both within our consortium partners and from other SMEs and researchers in this space). A major challenge to delivery of trial data analysis has been the sheer depth of information which the imaging investigation encapsulates. Extracting muscle-by-muscle data on tissue properties such as fat fraction was traditionally performed manually, taking typically one day of expert analysis time per subject scan timepoint. The cost-effectiveness of conducting image analysis was therefore an important adjunct factor in commercial decisions to include imaging in clinical trials. Through the work performed by the SME partner CRIS, a new semi-automated tissue segmentation tool was created which reduces analysis time from hours to minutes, is effective even in datasets with extensive fat infiltration as are found in NMD patients and which will produce reliable quantification without the potential for user bias associated with manual assessment. This tool is a significant step forward in providing a cost effective pathway for clinical trial analysis. The technology resides within the European SME and represents a tangible competitive advantage.
- Enabling Positron Emission Tomography for rapid biodistribution studies of AONs
We have demonstrated for the first time that positron emitting radiolabels can be attached to AON molecules without affecting the biodistribution or exon skipping efficacy of these molecules in vivo. This is a major tool which can be used to advance new AON molecules both in the drug development environment and in the subsequent translation to clinical trials in NMD and in wider clinical applications. A major factor in extending the use of AONs in rare diseases is the challenge of demonstrating effect in only small numbers of subjects. PET imaging offers the highest sensitivity of all in vivo imaging methodologies and so through the developments made in BIOIMAGE-NMD provides an important tool for future work.
Firstly, through PET imaging of AON, the biodistribution can be rapidly determined in vivo during preclinical testing, simultaneously observing all organs of the animal. The non-invasive nature of the measurement allows repeat investigation in the same animal without the need for sacrifice to determine tissue levels which offers an increase in statistical power over invasive studies which must use independent animals for each evaluated time point. We expect that this technology should enhance drug development for second (and later) generation AON molecules.
Secondly, when translating AON studies from preclinical work into early phase human studies, the use of PET labelling offers a tool to ensure that the drug is reaching the same target in patients as is expected from the preclinical work. As AON molecules enter the muscle cell through areas of muscle membrane damage, the potential combination of diffusion MRI, which we have shown in animal models to be highly sensitive to permeability, with PET studies would provide important data.
Thirdly, PET imaging offers a tool to assess and optimise AON administration routes, both in animal studies and in human trials. The acute fate of the AON, rates of distribution across the body and clearance can be measured directly through dynamic PET scanning (limited by the half-life of the 18F isotope) in the individual patient. This approach can eventually feed through into development of personalised treatment regimes based on monitoring of the drug distribution in each patient.
- Dissemination activities
The BIOIMAGE-NMD Consortium has engaged in dissemination of the project findings through a wide array of routes, including presentations at specialist international conferences in the field of imaging, radiochemistry, neurology and neuromuscular diseases.
A Global Workshop on the use of biomarkers in NMD clinical trials was hosted at the British Library on the 26th May 2016 by Andrew Blamire (Coordinator, BIOIMAGE-NMD) and Volker Straub (Coordinator, SCOPE-NMD). The objective of the workshop was to bring together representatives of all the Consortium partners (BIOIMAGE-NMD and the FP7 funded SCOPE-NMD), the European Commission, EMA and patient organisations to review the recent FP7 funded work on the use of imaging and other biomarkers for clinical trials and specifically to discuss the progress on outcome measures for evaluation of patients with DMD. All key stakeholders were represented and there was a wide ranging debate on making outcomes work in the future for this disease. The coordinator from Newcastle University, along with many representatives from the BIOIMAGE-NMD consortium partners including WP leads from WP1, WP2, WP3, WP5 and WP6 attended. In January 2017, BIOIMAGE-NMD also took part in the international conference entitled “'Making outcomes work' - Stakeholder Workshop on outcome measure development and implementation for DMD”. The coordinator from Newcastle University, along with many representatives from the BIOIMAGE-NMD consortium partners including WP leads from WP1, WP2, WP3 and WP6. WP Leads, including the coordinator of BIOIMAGE-NMD, were also invited to present on BIOIMAGE-NMD at the Newcastle University Institute of Genetic Medicine Seminar Series.
Presentation of the work from BIOIMAGE-NMD will also be showcased through invited presentations by the Consortium Coordinator and several of the work-package leaders at the First International Conference on Imaging in Neuromuscular Disease, to be held in Berlin in November 2017.
List of Websites:
www.bioimage-nmd.eu