Abstract
Keywords
INTRODUCTION
Alzheimer’s disease (AD) is the most common cause of dementia and represents an unprecedented public health problem. Although much of the underlying biological mechanisms leading to AD are still unclear, the amyloid hypothesis has tremendously influenced research conducted in both academic and pharmaceutical settings [1]. Since the accumulation of amyloid-β(Aβ) is generally believed to play a causative role in AD, the development of disease-modifying drugs aimed at reducing the Aβlevels in the brain has been a priority for many pharmaceutical companies. A number of reports have shown that active and passive Aβimmunotherapeutic approaches are effective in reducing brain Aβ in animal [2–6] and in clinical studies [7, 8]. Bapineuzumab, a humanized form of the murine monoclonal antibody 3D6, recognizes the N-terminus of Aβand binds all forms of Aβ(e.g., monomer, prefibrillar aggregates, and plaques) [9]. Bapineuzumab (and 3D6) is believed to clear Aβcentrally through both microglial phagocytosis [5] and neutralization of soluble aggregates [9]. In several Aβimmunotherapy clinical trials, abnormalities have been detected by brain MR imaging [10, 11]—so called amyloid-related imaging abnormalities (ARIA) [12]. ARIA is an acronym that describes a spectrum of MRI findings including vasogenic edema and sulcal effusions (ARIA-E) and microhemorrhages (ARIA-H) and were first observed during a Phase 1 study of bapineuzumab [13] and in a number of other clinical trials with amyloid-modifying agents [14]. Interestingly, these imaging abnormalities are often clinically asymptomatic, and ARIA-E appears as a transient phenomenon [15]. Further, treatment-related ARIA seems to occur during the first weeks of therapy and the risk of ARIA decreases over time [15]. Hence, ARIA represents an intriguing phenomenon of which the clinical relevance is not yet fully understood. However, from a safety perspective, ARIA cannot be neglected and has had a profound impact on the design of clinical immunotherapy trials. Nonetheless, although both PET imaging and histopathological studies have demonstrated that immunotherapies may reduce the amyloid burden, the exact mechanisms giving rise to ARIA remain to be fully elucidated. It has been hypothesized that the removal of vascular Aβupon immunotherapy leads to a transient disruption of the blood-brain barrier (BBB) that is eventually repaired upon chronic treatment [16, 17]. Conversely, in several animal studies, Aβimmunotherapy has been associated with increased vascular amyloid deposition accompanied by increased microhemorrhages, a likely correlate to ARIA-H in humans [18]. This increased vascular Aβmay be due to an overload of an already reduced perivascular clearance pathway. Aβclearance mechanisms are comprehensively reviewed elsewhere [19, 20]. To our knowledge, no studies have addressed the time course of such leakage events and the appearance of putative transient BBB disruptions that allow fluid, but not cells to extravasate, which would be considered equivalent of ARIA-E. In humans, ARIA-E is often anatomically associated with white matter, likely due to the prolonged accumulation of fluid along fiber tracks. The lack of extensive white matter as well as the lack of gyri and sulci in the brains of rodents [21] limit the use of T2 weighted/fluid attenuation inversion recovery (FLAIR) sequences to monitor fluid extravasation. To better understand the relationship between Aβimmunotherapy and ARIA, three in vivo MRI studies were performed to study BBB impairment in a PDAPP mouse model after anti-Aβimmunization using 3D6. PDAPP mice express high levels of a familial mutation of hAPP (V717F) and progressively develop both parenchymal and vascular amyloid deposition in a region-specific manner [22]. Various histological studies in the PDAPP mouse model have shown that 3D6 treatment clears amyloid from the vascular walls with a corresponding compromise of the integrity of the vascular wall, which can result in microhemorrhages, and subsequent hemosiderin deposition [16]. In our study, to facilitate the monitoring of transient disruptions of BBB integrity that might lead to extravasation of fluids, we used a series of repeated gadolinium (Gd-DOTA)-enhanced T1-weighted MRI scans and a T1 mapping model [23–26] where the changes in MR signal and the estimates of the T1-relaxation time (due to the leakage of the contrast agent) can be used both to localize BBB disruption and to assess the degree of BBB permeability.
MATERIAL AND METHODS
Animal model
The PDAPP mouse model overexpresses the human amyloid precursor protein (hAPP), which successfully recapitulates several neuropathological features characteristic of AD. Many of the histological, biochemical, and structural alterations present in the PDAPP mouse closely resemble the changes found in the human AD brain [22]. Aβdeposition seen in the PDAPP mouse brain is age-dependent and region-specific. By 12 months of age, amyloid deposition is noticeable throughout the hippocampus and in the frontal region of the cortex. Between 12 and 16 months of age, an accelerated deposition is observed [27]. Female PDAPP and wild type (WT) mice (age of 12 and 16 months) were received from Taconic Biosciences (Hudson, NY, US). Female PDAPP mice were used for practical reasons, to allow the housing of multiple animals per cage (n = 8 per cage). All experimental procedures were performed in accordance with European guidelines for the care and use of laboratory animals and were approved by the Committee on Animal Care and Use at the University of Antwerp, Belgium (Ethical Dossier number:2012-33).
Experimental study design
In a preliminary study, we wanted to explore the feasibility of MRI to monitor vascular dynamics and BBB integrity. PDAPP (n = 11) and WT (n = 11) mice (18 months old) were used. To investigate changes in BBB integrity associated with anti-Aβimmunotherapy, three independent longitudinal MRI studies were designed (Fig. 1). PDAPP mice were treated with saline (_S) or the antibody 3D6 (_T), which was administered intraperitoneally, once weekly (3 mg kg- 1, 0.5 ml). When treatment coincided with the MRI experiments, antibody was administered immediately after the imaging session. WT controls received weekly injections of saline (0.5 ml). In Study 1, PDAPP and WT mice (12 months old) were treated for 12 weeks (study 1 –n = 8 PDAPP_T;n = 8 PDAPP_S;n = 8 WT_S) with T1 MRI imaging at baseline, week 1, 4, and 12. In Study 2, PDAPP mice (12 months old) were treated for 5 weeks (study 2 - n = 12 PDAPP_T;n = 11 PDAPP_S;n = 8 WT_S). During this particular study, the dose of 3D6 was increased from 3 to 10 mg kg- 1 in week 4 and 5 to attempt to increase the number of events. In Study 3 (n = 11 PDAPP_T), PDAPP mice (16 months old) were treated for 4 weeks. For Studies 2 and 3, MRI was performed on a weekly basis (Fig. 1).
Magnetic resonance imaging
Animal preparation
All imaging experiments were performed on spontaneously breathing mice under isoflurane (Isoflo®, Abbot Laboratories Ltd.) anesthesia (induction 3%—maintenance 1.8%), administered in a gaseous mixture of 30%O2 and 70%N2. Respiration rate, monitored with a small animal respiration pad (MR-compatible Small Animal Monitoring and Gating System, SA instruments, Inc.), was maintained within normal physiological ranges. Rectal temperature was maintained at 37.0±0.2°C using a feedback coupled warm air system (MR-compatible Small Animal Heating System, SA Instruments, Inc.). To immobilize the head in a reproducible flat-skull position during the MRI experiments, mice were secured in an MRI compatible mouse stereotactic device. The head was held by a nose cone—including tooth bar—used for anesthetic gas delivery and blunt earplugs. The tail vein was cannulated with a 26-gauge needle (BD Vasculon Plus, Helsingborg, Sweden) for subsequent contrast agent injection. An actively decoupled surface array (2×2) receiver coil was positioned on top of the head. Homogeneous radiofrequency (RF) excitation was achieved using a proton volume resonator.
Magnetic resonance imaging: T1 weighted MRI
The experiments were performed on a Bruker Pharmascan 7T imaging system with horizontal bore (Bruker, Ettlingen, Germany). For each animal, the following protocols were used after standard spectrometer adjustments (coil tune/match, RF gain calibration, shim, scout image acquisition). To ensure uniform slice positioning, 2-dimensional coronal, axial, and sagittal T2-weighted images were acquired for each individual animal. Single slice positioning was performed in the axial plane, between bregma -1.22 mm and 2.18 mm. Following the slice positioning, serial sets of T1-weighted images were continuously obtained, before (baseline T1, pre-Gd) and every 2 min up to 30 min after (post-Gd) Gd-DOTA injection (0.2 mmol kg- 1;Dotarem®;Guerbet). Gd-DOTA was delivered manually and by the same experimenter. T1-weighted images were acquired using a multiple inversion-recovery echo-planar imaging (IR-EPI) sequence with varying TIs: 28, 800, 1000, 1400, 4000, 8000 ms. Following parameters were used: BW: 300 kHz;TR/TE: 10 s/20.4 ms;image matrix (128×128) (acquisition matrix: 98×84, zero filled read-out and partial FT phase);FOV (20×20) mm2;slice thickness: 1 mm;number of averages: 2. The temporal resolution of this technique is 2 min.
For each time point before and after contrast administration, the T1-values were estimated on a pixel-by-pixel basis. The measured signal intensities (SI) of the six T1-weighted images of the IR-EPI scans were fitted to the different TI according to the 3-parameter T1-relaxation of the spin system after an inversion pulse: (SI(TI) = ½A-B exp(-TI/T1)½+bias, Paravision 5.1). Manually drawn regions of interest (ROIs) of the left and right cortex, bilateral hippocampus and thalamus were defined by a single researcher based on a standard mouse brain atlas (Paxinos) and mean T1 values were extracted for each of the individual ROIs and all the time points from the repeated scans (Fig. 2). For each ROI, the mean baseline relaxation time is obtained from the average T1-value of the 5 pre-Gd scans and the change of T1 (ΔTI) versus this baseline is calculated for every time point.
Brain tissue preparation
Mice where perfused transcardially with saline, followed by approximately 20 ml of ice-cold 4%paraformaldehyde in 0.1 M PBS (pH 7.4). The whole mouse head was post-fixed overnight in 4%paraformaldehyde at 4°C prior to transfer to PBS. Brains were then processed and double-labeled for amyloid burden and microhemorrhages. The sections were co-labeled with biotinylated 3D6 and a modified Perls’Prussian blue iron reaction for hemosiderin (NeuroSciences Associates, Knoxville, TN).
RESULTS
T1 mapping
BBB integrity was assessed in vivo using Gd-DOTA-enhanced MRI. Body weight was determined weekly to monitor general wellbeing of the mice. Overall, body weight remained constant in both PDAPP and WT animals (data not shown).
In a preliminary experiment, to assess the feasibility of our approach and prior to the administration of any treatment, we compared the impact of the Gd-DOTA injections on aged PDAPP and WT mice (18 months old). In the cortex, a greater T1 drop was observed in PDAPP compared to the WT mice. Since amyloid is known to accumulate in the cortical vessels, these findings suggested that small BBB disruptions in old, untreated PDAPP mice, below the limit of visual detection, were greater than in WT animals, even before treatment is considered (Fig. 3).
Next, to investigate whether the effects seen both at the individual animal level and group level were due to true leakage of the Gd-DOTA into the brain parenchyma and not related to vascular effects, the muscle tissue of the jaw, where the vasculature is naturally leaky, was included as a positive control [28]. In all the groups (PDAPP_T, PDAPP_S, and WT_S), a massive drop in T1 relaxation values (≈900 ms) in the muscle was observed immediately after Gd-DOTA injection. Note that on T1 weighted images, influx of Gd-DOTA was as a signal enhancement, however, after calculation of the quantitative T1 maps, Gd-DOTA leakage into the brain was seen as a decrease in T1 relaxation values. Following the steep initial decline, T1 relaxation values gradually recovered in the direction of the baseline values. In the muscle, no difference was observed between the different scanning time points (weeks). The signal in the muscle remained unchanged and was not affected by 3D6 treatment (Fig. 4).
In 12-month-old PDAPP and WT mice, at baseline, no differences in T1 values were observed. After i.v. injection of Gd-DOTA, a smaller decrease in T1 relaxation time (≈50 ms) in the brain was observed. This was in contrast to the large drop observed in the muscle. At week 0 (pre-treatment), no differences were observed between the groups (PDAPP_T, PDAPP_S, and WT_S). The average ΔT1 for the PDAPP_T group (animals with and without active BBB disruption) doubled to –100 ms while the other groups remained at ≈–50 ms. The magnitude of the drop in T1 normalized afterwards (Fig. 4).
Frequency and timing of gadolinium-enhanced signals
To document fully the observed treatment effects, bilateral ROIs of individual animals were generated for the cortex, hippocampus, and thalamus. Only the cortical ROIs demonstrated a significant impact from the Gd-DOTA, so only the cortical T1 versus time plots were generated. The overall effect observed in animals that received the 3D6 treatment could be ascribed to a subset of individual animals that showed markedly larger post-Gd T1 drops (referred to as ‘events’). Visual inspection of the T1-weighted MRI images at TI = 1400 ms revealed a significant increase in signal intensity after Gd-DOTA injection in the meningeal and midline area in a subset of the 3D6 treated animals. This corresponded to the areas known to first accumulate vascular Aβdeposits [16, 17] and is consistent with the hypothesis that BBB dysfunction is related to the removal of vascular amyloid (Fig. 5). The increase in signal intensity was unilateral in a majority of the cases (Fig. 6, Table 1).
In the first study, a significant larger drop in T1 in the cortex was observed in 2 PDAPP mice treated with 3D6 at week 1. After 4 weeks of treatment, one of the 2 PDAPP mice still displayed some BBB impairment. No Gd-DOTA leakage was found after 12 weeks of treatment.
In the second study, the imaging frequency was increased to a weekly scanning protocol. After 2 weeks of treatment, a significantly larger drop in T1 relaxation values was observed in 1 animal. After 3 weeks of 3D6 treatment, the same animal and an additional animal displayed a large drop in T1 relaxation (Fig. 6). To enhance the occurrence of leakages, the dose of 3D6 was increased to 10 mg/kg for the last 2 weeks (weeks 3 and 4). After the dose increase, three mice experienced somewhat smaller ΔT1 changes of the order of 150 ms –200 ms. There was 1 mouse that produced an event seen on the raw T1 weighted images with a ΔT1 < 100 ms.
Due to the temporal frequency of the MRI imaging and uncertainty regarding the evolution of a BBB event, it is unclear whether the smaller, minor events really have a lower magnitude peak or have been imaged before or after the peak enhancement effect. In the saline-treated PDAPP group, a small decrease of approximately 50 ms in T1 values compared to baseline measurements (week 0) was observed in the cortex at all time points. There was no treatment effect in the WT group.
In a third study, 16-month-old PDAPP mice were treated with 3D6 and sacrificed once an event was recorded. After 1 week of treatment, 1 animal showed a major event. Immediately after scanning, the animal was sacrificed for histological analyses. Two additional animals with major events were identified after 2 weeks of treatment. These animals were also sacrificed for histological analyses.
An overview of the different studies is represented in Fig. 8. Only cortical ROIs of individual animals were displayed. After treatment with 3D6 in 12- and 16-month-old PDAPP, several mice showed BBB disruption.
Histopathological analyses
Immunotherapy with the 3D6 antibody increased the occurrence of microhemorrhages (hemosiderin deposits) in leptomeningeal vessels of PDAPP mice. All the animals that showed BBB leakage by MRI were positive for Perls’Prussian blue staining by histology (Fig. 7). Hemosiderin deposition and Gd-DOTA leakage co-localized at the level of the leptomeninges.
DISCUSSION
The present in vivo studies are the first to longitudinally examine the integrity of BBB following chronic treatment with the anti-Aβantibody 3D6. To advance our understanding of the relationship between ARIA and the pathophysiology of AD, we combined Gd-DOTA enhanced T1 MRI with histopathology to evaluate BBB integrity in PDAPP mice upon anti-Aβimmunotherapy treatment. Gd-positive signals were observed in a subset of 3D6 treated mice but not in WT or saline-treated PDAPP. Further, the temporal relationship of these BBB impairments was documented with a resolution of weekly scans. These consecutive scans indicated that the individual BBB lesions occurred early in the treatment and appeared to resolve spontaneously within a couple of days. Additionally, the risk of developing individual events seemed to diminish over a period of a few weeks.
Assessment of BBB integrity: T1 MRI
Histological analyzes for the presence of albumin and other plasma proteins in postmortem brains have provided substantial information about BBB impairment in AD and animal models (for review see [29]). However, this evidence is mainly indirect and there is no information on the exact spatio-temporal characteristics of BBB impairment. In vivo detection of BBB impairment in AD has gained increased attention as these lesions might potentially contribute to cognitive dysfunction [30]. Spontaneous microbleeds in patients have been detected using T2*-weighted imaging (based on the detection of hemosiderin substances that are paramagnetic and cause an inhomogeneity in the magnetic field). However, the relationship to histologically detected microhemorrhages is not always straightforward [31]. Following intravenous administration of superparamagnetic iron oxide particles, Beckman and colleagues [32] showed the feasibility of detecting cerebral amyloid angiopathy-related microvascular lesions in APP23 transgenic mice and demonstrated that MRI has the sensitivity to noninvasively monitor the development of vascular pathology in transgenic mice. However, while ARIA-H has been reported in transgenic mouse models, to date only one publication described spontaneous appearance of ARIA-H and ARIA-E in old aged transgenic mice [33]. Additionally, until now, no well-established protocols were available to model treatment related ARIA. Gd-based MRI has been used in clinical as well as research settings to evaluate altered BBB permeability in diseases (e.g., multiple sclerosis [34] and cancer [35, 36] or injury (e.g., stroke)) [37]. After a bolus injection of Gd into the vein, the blood level of Gd rises rapidly, creating a gradient across the capillary endothelial membrane. In regions with relatively free capillary permeability, the contrast agent will leak across the vessel wall and begins to accumulate in the perivascular interstitial fluid (e.g., muscle). In the brain, the intact BBB will prevent leakage of contrast material. However, when the contrast agent leaks from the blood compartment to the brain parenchyma due to BBB disruption, the altered relaxation characteristics will be visible by T1 MRI. Although Gd T1 MRI has been used to probe BBB impairment in many diseases, it has not been applied extensively in AD research (for review, see [38]). Recently, using an advanced dynamic contrast-enhanced MRI protocol, Montagne and colleagues [39] showed an age-dependent BBB breakdown in the hippocampus. In addition, the BBB breakdown was more pronounced in MCI patients. Moreover, BBB disruptions using Dynamic contrast-enhanced (DCE) MRI have been shown in vascular cognitive impairment patients [40]. However, so far, there is no convincing evidence of BBB disruption with this contrast-agent in AD, presumably due to the rapid restoration of the vascular wall. In the present study, we found that T1 relaxation values decreased significantly in the meningeal and cortical midline areas of PDAPP mice treated with 3D6. Histological analyses showed that all the animals that displayed BBB leakage visible by MRI were positive for hemosiderin by histology. Previously, it was shown that anti-Aβantibody treatment in mouse models has been associated with increased incidence of vascular microhemorrhages, detected by hemosiderin staining [6, 17]. Zago and colleagues [16] reported histological alterations in cerebral vessels of PDAPP mice after 3D6 treatment. They showed that much of the vascular Aβclearance occurred within the first 12 weeks of treatment (86%reduction). Hemosiderin deposits were detected in leptomeningeal vessels as early as 7 weeks after treatment. It was suggested that leptomeningeal vessels containing vascular amyloid are predisposed to microhemorrhages during the initial phases of Aβclearance by immunotherapy [16]. In the present study, a subset of PDAPP mice displayed a relatively steep drop in T1 values, already after 1-2 weeks of 3D6 treatment. Histological analyses showed that all the animals that displayed BBB leakage visible by MRI were positive for hemosiderin by histology and these leakages seemed to occur in areas with apparent reduction of Aβplaques. Interestingly, Bapineuzumab-treated subjects with ARIA-E showed greater amyloid reduction by PET imaging at week 71 [41] suggesting that BBB leakage may lead to local increase of antibody concentration and higher rate of plaque clearance. We only observed BBB leakages in a subset of 3D6-treated animals and not in WT or untreated PDAPP mice. Moreover, BBB leakage events occurred early in the treatment, already detectable after 1 week of immunization in a subset of animals. In addition, they seemed to be transient, and thus mimicking the pattern of the ARIA events observed in a subset of clinical trial patients [42]. Although the consequences of BBB disruption can lead to dysregulated molecular and ionic flux across the damaged BBB that can potentially culminate in neuroinflammation and neurodegeneration [30, 43], most cases of ARIA in the bapineuzumab and other anti-amyloid immunotherapy trials reported to date were clinically asymptomatic [15]. Symptoms, when observed, appear to be transient in nature.
Limitations of the study
The purpose of the present study was to develop a tool to visualize and quantify the extent of BBB impairment upon immunotherapy in an animal model of AD. However, these BBB disruptions likely follow the sparse nature of vascular amyloid deposition in animal models and occur focally with complex spatial patterns. ROI based analyses hampered the setting of a specific threshold on the T1 maps because of an averaging of the T1 values across the entire ROI, which dilutes the signal. In this respect, a voxel by voxel quantification method would be required to capture the full extent of the BBB impairment, but given the substantial variation already present in the ROI-averaged results, it is questionable whether voxel-based analysis would increase the sensitivity or specificity. Thus both visual inspections and ROI analyses are necessary to reliably detect BBB impairment. The exact pathophysiological changes that underlie the increase in vascular permeability have yet to be fully addressed. Our results indicate that MR imaging methods with contrast agents facilitate the exploration of vascular events with unprecedented spatio-temporal resolution in animal models of Aβdeposition. Additional histological and biochemical studies might shed some light on the underlying pathologies of both ARIA-H and ARIA-E.
In conclusion, with regard to the currently reported clinical experience with ARIA-H and ARIA-E in clinical trials and the preclinical work with PDAPP mice, our findings support the hypothesis that treatment with 3D6 leads to transient leakage from amyloid-laden vessels. These observations are important to increase our understanding of ARIA observed in AD patients upon Aβlowering therapies. In addition, the current study has provided valuable insights on the time course of vascular alterations during immunization treatment, which will be useful to guide further exploration of the nature of ARIA. Finally, the developed methodology may be useful for the development of novel anti-Aβagents and/or optimal immunotherapy dosing strategies that minimize vascular responses.
Footnotes
ACKNOWLEDGMENTS
This study was sponsored by Janssen Alzheimer Immunotherapy R&D. Wagner Zago was an employee of Janssen Alzheimer Immunotherapy R&D at the beginning of study and is now employee of Prothena Biosciences Inc. This research was also supported by the European Union’s Seventh Framework Programme under grant agreement number 278850 (INMiND). The Bruker Pharmascan 7T system (Bruker, Ettlingen, Germany) was purchased through Hercules foundation funding (Belgium) under the promoter-ship of Prof. dr. Annemie Van der Linden. Caroline Guglielmetti is holder of an IWT PhD fellowship (agency for innovation by Science and Technology).
