High-resolution Imaging of Myeloperoxidase Activity Sensors in Human Cerebrovascular Disease

Progress in clinical development of magnetic resonance imaging (MRI) substrate-sensors of enzymatic activity has been slow partly due to the lack of human efficacy data. We report here a strategy that may serve as a shortcut from bench to bedside. We tested ultra high-resolution 7T MRI (µMRI) of human surgical histology sections in a 3-year IRB approved, HIPAA compliant study of surgically clipped brain aneurysms. µMRI was used for assessing the efficacy of MRI substrate-sensors that detect myeloperoxidase activity in inflammation. The efficacy of Gd-5HT-DOTAGA, a novel myeloperoxidase (MPO) imaging agent synthesized by using a highly stable gadolinium (III) chelate was tested both in tissue-like phantoms and in human samples. After treating histology sections with paramagnetic MPO substrate-sensors we observed relaxation time shortening and MPO activity-dependent MR signal enhancement. An increase of normalized MR signal generated by ultra-short echo time MR sequences was corroborated by MPO activity visualization by using a fluorescent MPO substrate. The results of µMRI of MPO activity associated with aneurysmal pathology and immunohistochemistry demonstrated active involvement of neutrophils and neutrophil NETs as a result of pro-inflammatory signalling in the vascular wall and in the perivascular space of brain aneurysms.

Magnetic resonance imaging (MRI) provides 3-D structural information about tissue volume as well as detailed physiological and anatomical information for thick tissue sections. Ultra-high-resolution MRI using small field-of-view acquisitions, i.e. micro-MRI (µMRI), can be used to obtain microscopic images within whole organs (e.g. MRI biopsy) reflecting differences in water relaxation times in various tissue components 1 . Such µMRI techniques have been used to study the developmental biology of mice 2,3 , reveal details of brain structure 4,5 , and show anatomical and functional changes of skeletal/connective tissue systems 6,7 . Moreover, development of specially designed radiofrequency coils has further allowed µMRI of sub-millimetre thick microscopic tissue sections 8,9 . Both non-specific 10 and pathology-specific MR contrast agents 5 have been evaluated for their ability to produce images with highly delineation of target areas and their compatibility with standard histology markers. However, the utility of µMRI as a surrogate for high-resolution contrast-assisted in vivo MR imaging has not been fully explored. Using animal models of human disease, we and others previously demonstrated that paramagnetic substrate-sensors of neutrophil myeloperoxidase (MPO) activity 11,12 are highly useful in imaging inflammation-linked pathologies of the cardiovascular system [13][14][15] and brain 16,17 . Generation of MPO-specific MR contrast is brought about via the enzymatic oxidation of MPO-reducing hydroxytryptamides. These paramagnetic hydroxytryptamides are small molecules, which undergo various oxidative transformations ultimately leading to the formation of dimers, oligomers and protein-bound forms 12,18 . Covalent binding to proteins results in the retention of paramagnetic chelated metals on cell surfaces and components of the extracellular matrix 19,20 , Fig. 1A. Here we report the utility of µMRI for detecting MPO activity in inflammatory lesions within the vascular wall of excised human brain aneurysms (hBA) and arteriovenous malformations (AVM). By combining µMRI with fluorescent probe-mediated imaging of MPO activity (Fig. 1B) and immunohistochemical corroboration we set forth to find out: (1) whether µMRI of enzyme-dependent vascular wall enhancement is feasible in the very small volumes of histology sections and; (2) whether such MR signal enhancement correlates with the distribution of MPO activity in the tissue.

Results
Histology of human tissue samples. The initial immunohistochemistry of unruptured hBA samples obtained as a result of surgical clipping revealed multiple IL-8 (CXCL8) cytokine positive cells infiltrating the vascular wall both along the adventitial and luminal sides of hBA ( Fig. 2A,C). Ineterleukin-6-positive cells were present as sparse infiltrates detectable only on the adventitial side of the vessels (Fig. 2B) while MCP-1 (CCL2) antigen was not detectable in any of the samples we tested. Smooth muscle cells in hyperplastic thickened areas of aneurysmal wall showed a strong activation of NF-kB with nuclei of cells binding an antibody against phosopho-p65 in all unruptured aneurysmal samples we processed (Fig. 2C).
As determined by our interdisciplinary cerebrovascular interventional imaging team, immunohistochemistry and immunofluorescent detection of specific neutrophil enzymatic markers in tissue sections demonstrated that regardless of hBA location, all unruptured hBA contained neutrophil-specific antigens, i.e. myeloperoxidase ( Fig. 3A-D) and elastase (Fig. 3D,F). Both enzymes could potentially be used as tracking markers of neutrophils. We observed the presence of intracellular and extracellular neutrophil MPO antigen in the majority of samples of clipped brain aneurysms. Neutrophils were localized primarily around the lumen after infiltrating into the vascular wall ( Fig. 3B-D), as well as within the walls of the adventitial neovessels 21,22 . Intracellular neutrophil elastase was also present and could be identified by using a single or dual staining (alongside with anti-MPO antibody, Fig. 3D,F). The vessel wall-infiltrated neutrophils showed more prominent anti-elastase staining than more peripheral MPO-positive neutrophils attached to the surface of the aneurysms (Fig. 3D). In the subluminal regions of the vessel walls we observed the formation of perivascular wall haemorrhages and neutrophil extracellular traps (NETs) suggesting terminal neutrophil activation in the areas of vascular pathology and microthrombosis (Fig. 3B, arrow). Furthermore, by using a fluorescent MPO substrate 5HT-Cy3 (5-hydroxytryptamide of the cyanine dye Cy3, a red fluorophore) we were able to perform visualization of MPO activity in surgical samples. The assessment of MPO activity as opposed to the presence of protein antigen alone was essential for determining whether it would be feasible to further process the samples for µMRI because MR signal depended on the presence of active MPO enzyme. Activation of 5HT-Cy3 by MPO results in production of reactive species that link to nearby molecules of which a substantial fraction are proteins of extracellular matrix and fibrin of thrombi. By incubating frozen sections with 5HT-Cy3 in the presence of hydrogen peroxide we observed MPO activity as bright red fluorescent signal associated with the cells infiltrating human AVM and hBA samples ( Fig. 4A-C). It should be noted that in the majority of processed human samples, MPO activity was positively identified in polymorphonuclear cells as well as extracellular deposits (Fig. 4C) suggesting the stability of extracellular enzyme resulting in preservation of enzyme activity. These cells had irregular shaped nuclei and were present in both unruptured hBA and AVM surgical samples. The cells stained positively not only for MPO activity with 5HT-Cy3 probe, but also expressed S100 calcium-binding proteins characteristic of granulocytes and some other leukocytes (Fig. 4D,E). Hydrogen peroxide is generated in the blood vessel wall as a result of superoxide anion dismutation. MPO is using hydrogen peroxide and Gd-5HT-DOTAGA (shown as an example) to generate oxidized products at protoporphyrin IX MPO catalytic sites. These products bind to nearby proteins. (B) Chemical formulas of Gd-5HT-DOTAGA and 5HT-Cy3. The simplified reaction mechanism between the hydroxytryptamide probes (R = Gd-DOTAGA or Cy3) and the oxidized form of MPO is shown below. The intermediates (free radicals) react with the proteins at tyrosine residues (pathway 1) or undergo dimerization/oligomerization (pathway 2) 19,49 .  Testing MPO substrates in vitro. To achieve MPO-specific µMRI imaging in histology sections we used two types of substrate-sensors: (1) Gd-bis-5HT-DTPA that was previously used by multiple research groups for imaging of MPO activity in vitro and in vivo [14][15][16] (2) novel MR imaging sensor Gd-5HT-DOTAGA. All tested MPO imaging sensors function as electron donors, which is an essential property required to reduce oxidized MPO to a catalytically active state. Initially we tested both substrates in parallel to investigate whether the substitution of the acyclic DTPA chelate for a macrocyclic DOTAGA results in the expected increase of stability and thus resistance to trans-metallation, i.e. the release of paramagnetic Gd(III) from the chelate in the presence of an excess of competing biologically relevant ions such as Zn(II) and phosphate. As we anticipated, in transmetallation test Gd-5HT-DOTAGA showed no statistically significant decrease of relaxation rate R1 (p > 0.05). A decrease in R1 would signify a conversion into insoluble form and disappearance of Gd(III) from the solvating water environment. In contrast, Gd-bis-5-HT-DTPA showed a time-dependent decrease of the initial R1 values consistent with the lower stability of this MPO substrate-sensor (i.e. greater propensity for trans-metallation, see Supplementary Figure S1A).
We further tested whether the MPO-mediated catalysis reaction leads to a conversion of both macrocyclic and acyclic linear Gd(III) chelating substrates into products with higher Gd(III) molar relaxivity, which underlie the enhancement of MR signal in T1-weighted images. The effects of MPO reaction on the relaxivity of the initial substrates were analysed at three different field strengths. At low magnetic field strength (0.47T, 20 MHz)  Figure S2). Both substrates (Gd-bis-5HT-DTPA and Gd-5HT-DOTAGA) resulted in a similar, approximately 1.5-1.7 fold increase of molar transverse relaxivity after MPO and hydrogen peroxide were added to reaction mixtures ( Figure S1B). The increase was even more pronounced in the case of Gd-5HT-DOTAGA/MPO reactions run in the presence of human serum albumin (2.5-fold increase of longitudinal relaxivity) while Gd-bis-5HT-DTPA showed a lower relative increase. An increase of r1 due to MPO-catalysed reaction of Gd-5HT-DOTAGA was also seen at the higher clinical field strength of 1.4T. Unlike the measurements performed at lower magnetic field strengths (i.e. 0.47 and 1.5T), the measurements performed at 7T did not show any detectable Gd(III) relaxivity change when MPO reaction products were compared to the initial substrate ( Figure S1B). Measurements of T2 and T2* of Gd-5HT-DOTAGA products of the reaction with MPO pointed to a strong increase of transverse relaxation effects, i.e. a sharp increase of r2/r1 ratio when compared to low field/lower resonance frequency environments.
Micro-MRI and histology corroboration. The verification of MPO activity in human tissue sections by a fluorescent 5HT-Cy3 substrate (Fig. 4) and subsequent testing of paramagnetic tryptamides ( Figure S1) suggested the feasibility of detecting MPO paramagnetic substrate-sensors in human tissue using µMRI histology. We first validated the sensitivity of our µMRI system by performing parallel fluorescent and MR imaging of specially designed tissue-like phantoms. These phantoms were designed to obtain a model tissue containing MPO within thick and thin histological sections. Phantoms were made by moulding standard MPO solutions with known specific activity and concentration into separate cylindrically shaped gels by adding a polyacrylamide-stabilized Matrigel/agarose matrix. This approach (1) enabled imaging of samples containing various concentrations of MPO simultaneously, (2) allowed diffusion of MPO substrate-sensors into the matrix of phantom sections during co-incubations; (3) allowed testing of various MPO substrate-sensors simultaneously and (4) allowed measurement of water relaxation times after subsequent retention of MPO reaction products in the matrix of phantom sections. The range of MPO concentrations tested was chosen to emulate the range previously measured by our group in surgical samples of human aneurysmal tissue, i.e. approximately 200 U of MPO activity per mg-tissue in MPO-positive aneurysms 21 . Measurements of fluorescence intensity demonstrated that the average fluorescence intensity of MPO sample sections increased in order of increasing concentrations of MPO in the matrix (Fig. 5A,B); r1 and r2 molar relaxivities along with mean MR signal intensities were measured following 7T µMR acquisitions of embedded MPO samples. Results showed a near-linear correlation of MR R1 and R2 relaxation rate values (r 2 = 0.76-0.78, p = 0.04-0.05) with increasing MPO concentrations (0.1-1.0 U MPO/µl) in section volumes of the phantom (Fig. 5C,D) and demonstrated the sensitivity of our µMRI imaging system.
Next, by using the same µMRI acquisition set up we set out to investigate the utility and sensitivity of µMRI for the detection of inflammation in human brain tissue as evidenced by the presence of MPO activity and neutrophil infiltration in human brain tissue. Initially we performed imaging of sections prepared from a sample of human brain arteriovenous malformation (AVM) because it provided a larger sample volume and thus better coil-filling factor, which enables faster image acquisition and optimization (Fig. 6). Despite the endogenous bright contrast due to white matter present in the field of view, the imaging of AVM showed the presence of enhancement around the irregular vascular luminae (shown by arrows, Fig. 6A). This pattern of T1w-SE sequence generated signal  (Fig. 6A), which correlated with the areas that were strongly reactive with anti-MPO antibody on histology sections (Fig. 6B).
A total of two ruptured and six unruptured hBA were further subjected to imaging with µMR after obtaining 50-80 µm sections (Figs 7 and 8). An ultrashort TE (UTE) pulse sequence was chosen to decrease the T2* effects at 7T that interfered with Gd-mediated longitudinal water relaxation time shortening (Fig. 7A). In a relatively large hBA (Fig. 7). we observed two major areas of MR signal enhancement, which correlated with the presence of adventitial angiogenesis (Fig. 7, asterisk) and luminal micro haemorrhage (Fig. 7, arrows), both of which contained MPO activity as confirmed by using a fluorescent substrate of MPO (Fig. 7C). Overall, the same trend was observed in the samples of ruptured as well as unruptured hBA. In general, fluorescent substrate localized in tissue areas containing MPO activity which matched very well with the areas of elevated MR signal determined by applying UTE pulse sequence or T1w-SE pulse sequences (Figs 7 and 8). Due to the extensive infiltration of neutrophils and NETs in the luminal clots, the MPO activity was easily detectable in the samples of ruptured hBA (Fig. 8). For example, the matrix of the clot appeared bright on T1w images (Fig. 8A) after treating with paramagnetic MPO substrate-sensor. The same area showed MPO activity resulting in red fluorescence after incubating with 5HT-Cy3, and the level of fluorescence decreased dramatically after adding the irreversible inhibitor of MPO (Fig. 8G).

Discussion
In this study, we used a µMRI histology setup for determining the distribution of MPO, a molecular marker of active inflammation within the microarchitecture of very small surgically excised fragments of hBA. Molecular imaging of this essential enzyme's catalytic activity has been made possible by linking of an MR detectable paramagnetic metal (chelated Gd(III) to a substrate of the MPO enzyme (5HT). Upon diffusion into the tissue, the Gd(III)-conjugated 5HT substrate is oxidized by MPO, and is then retained at the site primarily as a result of binding to proteins. As a result, the areas with higher water relaxation rates are generated within the areas of high MPO activity. Two versions of chelated Gd(III) conjugated to 5HT were tested in this study: an acyclic DTPA chelate (Gd-bis-5HT-DTPA) and a macrocyclic DOTA chelate (Gd-5HT-DOTAGA). The distinction between linear and macrocyclic Gd(III) chelates is clinically significant owing to the higher safety profile of macrocyclic formulations 23,24 . Gd-5HT-DOTAGA demonstrated a higher MPO-mediated molar longitudinal relaxivity (r1) increase due to a lower initial relaxivity of non-reacted Gd-5HT-DOTAGA at 0.47-1.4T. The differences between the relaxivities of non-modified versus MPO-oxidized Gd-5HT-DOTAGA substrate-sensor are what give rise to the changes in MR signal intensity between pre-and post-contrast µMRI images and thus the potential clinical utility of this MPO sensor.
We anticipated that testing of MPO substrate-sensors at clinically relevant (1.5 and 3T) field strengths would eventually be needed to obtain a clearer picture of how compounds will perform in a more standard clinical setting. The higher field strength of 7T has some advantages for MR imaging, i.e. the overall higher tissue signal-to-noise ratio essential for the imaging of extremely small histology sections. High-field MRI also presents obstacles such as the increase in the ratio of r2/r1 compared to lower field strengths, which results in lower post-contrast signal enhancement for a given dose of Gd(III)-based MRI contrast agent. Those limitations can be partially overcome with optimization of pulse sequences such an ultrashort echo time (UTE) sequence, which maximizes signal from tissues with exceedingly short transverse relaxation times. Other factors must also be considered in interpreting µMR imaging results acquired in this study. Since the thickness of sections intended for µMRI exceeded standard histology section thickness by a factor of 10, the registration of the two image sets is not always 1:1 and this imposes limitations on the interpretation of signal overlap between the two visualization modalities. However, this does not prevent registering of µMR images with fluorescence images of MPO activity distribution in the tissue (Fig. 7). Results presented here and obtained in hBA imaging using MPO substrate-sensors confirmed previous results showing positive correlation between normalized MPO enzymatic activity localized in aneurysmal microstructure, and the presence of neutrophil infiltration as determined by immunohistochemistry and vascular wall inflammation/hBA instability 21,25,26 . The presence of MPO activity in the blood of patients, mainly as a component of neutrophil NETs 27 brought about by the cells of innate immune Cerebral aneurysms are characterized histologically by chronic inflammation and degenerative changes to the vessel wall which may contribute to aneurysm pathogenesis and rupture [29][30][31][32] . Numerous inflammatory cytokines such as interleukins 33,34 , monocyte chemoattractant protein 1 (MCP1) 35,36 , as well as activation of NF-kB 30,37 have been shown mechanistically to contribute to the pathogenesis of cerebral aneurysms. The picture emerging from results presented here, as well as numerous studies obtained in animal models and humans, shows the pathology of aneurysm development and rupture is complex and that multiple signalling cascades occurring in a range of vascular cell types and the extracellular matrix are involved (reviewed in 38 ). Vascular wall remodelling may occur early in response to chronic vascular wall stress resulting from haemodynamic abnormalities. Ensuing endothelial damage may trigger the up-regulation of transcription factor NF-kB in smooth muscle cells and recruitment of inflammatory cells releasing IL-1β, IL-8, TNFα and other pro-inflammatory mediators 39,40 . We confirmed the presence of many IL-8-positive cells that infiltrate hBA wall potentially in response to downstream signalling associated with NF-kB activation (Fig. 2). Interleukin-8 (CXCL8) is a potent neutrophil chemotactic factor 41 additionally promoting angiogenesis suggesting active participation of neutrophils in aneurysmal wall remodelling 42 . The demonstrated use of µMRI for the detection of MPO activity in situ in non-fixed human tissue sections has applicability to the ever-increasing number of investigations that seek to understand the role of molecular and nanoparticle-assisted imaging in detecting the development and progression of brain vascular pathology 22,43 . The potential of µMRI as a preclinical and translational imaging modality is promising but is still in the early stages of being fully validated. However, we anticipate that high-resolution µMRI has a potential of becoming widely applicable in the development of new tissue-specific contrast agents and sensors involving direct interaction with the components of surgically excised human tissue samples. We foresee broad application of this technique in research directed at testing specific activity of novel MR enzyme-specific substrates in realistic in vitro model systems.

Methods
This prospective study involving human pathology samples was approved by the University of Massachusetts Medical School (UMMS) Institutional Review Board (IRB) and patient's written informed consent was required before surgery. HIPAA compliance was enforced throughout the study. The following inclusion criteria were applied: included were adult female and male patients (ages 18-79 years old) with symptomatic or asymptomatic, ruptured or unruptured saccular cerebral artery aneurysms for which microsurgical clipping has been deemed the most appropriate course of treatment. The following exclusion criteria were applied: traumatic or mycotic aneurysms, aneurysms in patients with concurrent adult polycystic kidney disease, aneurysms in patients with concurrent degenerative connective tissue disorders including cystic medial necrosis, fibromuscular dysplasia, Marfan syndrome, Turner's syndrome, and giant cell and Takayasu's arteritis, patients within a vulnerable population, including foetuses, neonates, pregnant women, patients unable to provide informed consent, institutionalized individuals and prisoners. A total of n = 10 surgically resected samples of human saccular aneurysms were included in the study (n = 7 female, n = 3 male, 2 ruptured, 8 unruptured, median age -57 years; see Table 1). One Myeloperoxidase substrate synthesis and testing. Fluorescent substrate 5HT-Cy3, (5-hydroxytryptamide of Cy3) was synthesized by using a method described in 44 . Briefly, 5HT-Cy3 was synthesized by reacting 1.1 molar excess of 5-hydroxytryptamine with mono N-hydroxysuccinimide ester of Cy3 (GE Healthcare Bios-Sciences Corp., Piscataway NJ) in DMF in the presence of triethylamine under argon. The reaction product was precipitated using 10 volumes of acetone:diethyl ether (1:2 by vol), dried under argon and purified by using C18 HPLC (Discovery C18 column, Sigma, St. Louis MO) and a gradient of acetonitrile in 0.05 M triethylammonium acetate, pH 7.0.
SCIeNtIfIC RePoRtS | (2018) 8:7687 | DOI:10.1038/s41598-018-25804-y MR microscopy. Thick sections of human brain vascular pathology specimens (Table 1, 80-100 µm thick) were treated as above in the presence of MPO imaging substrates (0.5 mM in PBS) for 30 minutes in the absence or in the presence of MPO inhibitors before being imaged with µMRI using a T1w-SE sequence. Direct µMRI of tissue sections was performed using a set of homebuilt histological coils tuned to operate at the Larmor frequency corresponding to 300 MHz and interfaced to a 200 mm horizontal bore 7 Tesla Bruker µMRI system (Bruker BioSpin) equipped with an actively shielded gradient coil insert 750-mT/m gradient strength, 100-μs rise time 9 . Depending on the sample (40-µm to 100-µm) and the histology coil used, the MRI in-plane spatial resolution ranged from 57-µm to 100-µm acquired within the overnight scan. Pulse sequences used: 2D T1w-GE with TE/ TR 3.2/100 ms; 2D multi-echo T2*w-GE TE/ES/TR 4.9/4.2/100 ms.
Statistics. Statistical analysis was performed by using Prism 7 (GraphPad Software Inc.) Results are expressed as mean ± SD. The analysis of means and assignment of statistical significance was be performed by using two-tailed unpaired t-test with Welch's correction.
Data availability. All data generated or analysed during this study are included in this published article (and its Supplementary Information files).