Anti-Beta-Amyloid Passive Immunotherapy for Alzheimer’s Dementia and Amyloid Related Imaging Abnormalities (ARIA): What’s Next?

The National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA) sponsored a workshop on “Anti-Beta-Amyloid Passive Immunotherapy for Alzheimer’s Dementia and Amyloid Related Imaging Abnormalities (ARIA): What’s Next?”. Leading experts and researchers from around the world gathered on September 28-29, 2023, for this hybrid event. The workshop consisted of presentations and panel discussions with experts in the field, addressing both clinical and preclinical knowledge and future directions. Two anti-beta amyloid antibodies, aducanumab, and lecanemab, have received accelerated approval by the U.S. Food and Drug Administration (FDA) for the treatment of mild to moderate Alzheimer’s disease (AD), and Phase 3 trials for several additional antibodies are currently underway. During the clinical trials for this anti-beta amyloid immunotherapy, however, MRI abnormalities have been reported as an adverse treatment effect, termed “amyloid-related imaging abnormalities” (ARIA). Some patients have experienced severe side effects, including leaky blood-brain barrier (BBB), edema, brain hemorrhage, and even death. Therefore, it is critical to better understand the benefits and risks associated with this therapy and strategies to improve efficacy and safety.

Session 1: Clinical Overview

              There are two subtypes of ARIA: ARIA showing edema and effusion (ARIA-E) and ARIA showing microhemorrhages and superficial siderosis (ARIA-H). ARIA-E is likely due to changes in water flux at the BBB, while ARIA-H is likely due to damage to the walls of small arterioles. ARIA is associated with areas of dynamic changes in Aβ localization, especially clearance from vessel walls, and presents an issue with clinical applications of anti-beta amyloid immunotherapies. The two largest risk factors for ARIA-E are treatment dose and ApoE4 carrier status, but these alone are not predictive of which patients will develop ARIA. There are several challenges for exploring anti-amyloid immunotherapy as a treatment for AD, including: identifying the optimal time for antibody administration (secondary and primary prevention of ARIA), selection of patients based on imaging (amyloid PET, MRI), and molecular markers (amyloid and tau in CSF and plasma), risk assessment and management (ApoE4 carrier status, comorbidities), and duration and post-administration monitoring of immunotherapy. There are two prevailing mechanistic theories of ARIA. The first is that it is an inflammatory response to amyloid moving from plaques to the perivascular space, but the timing of ARIA-E weakens this theory as it occurs very early before much amyloid has been removed and does not increase over time. The second theory is that pulling amyloid out of smooth vessels directly changes the vessel structures, and this is supported by the timing of ARIA but discounted by the fact that TREM2 activation also can cause ARIA. There is also the possibility that ARIA is a marker of treatment response, as it is associated with greater amyloid reduction, but there is mixed evidence of reduction in neurodegeneration.

              An important aspect that may help in understanding ARIA is its similarity to cerebral amyloid angiopathy-related inflammation (CAA-ri), which is spontaneously occurring inflammation with similar imaging abnormalities and an association to amyloid in the brain. Around 80% of people with AD have CAA, which is driven by failure of amyloid clearance from blood vessels in the brain. Symptomatic intracerebral hemorrhage is rare in CAA but raises concerns about the use of anticoagulants or thrombolytics in the setting of anti-amyloid immunotherapy. One knowledge gap that exists for ARIA is whether damage to blood vessels is due to Aβ antibodies binding to preexisting CAA or due to antibodies leading to amyloid translocation to the vasculature and causing injury. Studies of immunotherapy in CAA could help to address these gaps.

              A key limitation to furthering our understanding of ARIA is the ability to measure ARIA-related changes in the brain more accurately, both post-mortem in neuropathological studies and in vivo in immunotherapy patients. Existing neuropathology findings in patients treated with anti-beta amyloid antibodies show increased CAA and related vasculopathy, but there is very little tissue available from immunotherapy patients to study. Since ARIA seems to be analogous to spontaneous CAA-ri, studying the much more widely available tissues with CAA-ri could help address many unanswered questions about ARIA pathology. As ARIA is defined as an imaging abnormality, good techniques for imaging the BBB are crucial for the diagnosis and monitoring of ARIA. Current MRI strategies to measure BBB leakiness including dynamic contrast enhancement with gadolinium for dynamic imaging and arterial spin labeling for proximal blood and water exchange, are not ideal for routine clinical use. MRI of fluid exchange is a new and rapidly evolving method that will continue to improve in sensitivity, resolution, and interpretability of data.

              Another limitation facing the field is a lack of good biomarkers for ARIA, including inflammatory imaging and more sensitive peripheral biomarkers. Biomarkers could be used to track progression and predict the risk of ARIA. The ability to predict risk could make prevention trials a feasible approach. There is so much variability in patients that can influence vessel health and there is overlapping pathology of CAA and AD, making the discovery of reliable biomarkers very difficult.

              Existing data on ARIA come from ongoing clinical trials. Radiographic severity of ARIA is not predictive of symptom severity and most ARIA events are asymptomatic. ARIA does seem to be related to amyloid burden and rate of amyloid removal, but there is no relationship between baseline amyloid load and ARIA. ApoE4 carrier status has emerged as a risk factor, with a higher incidence of ARIA-E, but not ARIA-H, in E4 carriers. Isolated ARIA-H occurs at a similar incidence in placebo and anti-beta-amyloid-treated patients. The proportion of people who develop ARIA is only a subset of the people who will develop dementia. There is a need to delineate the phenotype that goes with the outcomes and weigh the risks of ARIA with the benefits of immunotherapy, with timing of treatment an important factor.

Session 2 – Basic Mechanisms Overview

Several key topics were addressed, shedding light on the fundamental processes related to CAA, microvascular disease, and neurovascular dysfunction. Cerebral microvascular disease is characterized by highly variable arteriolar pathology with prominent smooth muscle cell changes, combined with a limited variety of parenchymal injuries. Cerebral microbleeds/microhemorrhages can be primary or secondary. Primary microbleeds are well-established for CAA and are due to vessel wall injury. Secondary microbleeds manifest as hemorrhagic microinfarction. Endothelial erythrophagocytosis, where red blood cells stall in capillaries and become taken up by endothelial cells, induces microglial response and microhemorrhages, possibly leading to CAA/AD.

Neurovascular regulation is important for overall vascular health and the regulation of blood flow. Neurovascular dysfunction may be a biomarker of CAA, which is marked by progressive amyloid accumulation and vascular damage. Aβ and ApoE4 disrupt neurovascular regulation, causing progressive vascular damage and reduced Aβ clearance via oxidative stress. Border-associated macrophages, which are the major source of this oxidative stress, may contribute to ARIA and could be a therapeutic target in CAA and ARIA.

BBB Breakdown in anti-beta-amyloid passive immunotherapy was discussed, as well as the phenomenon of microhemorrhages in mouse models related to immunotherapy for CAA. Advanced age and advanced pathology have a role in the induction of these microhemorrhages, and other mechanisms of microhemorrhage induction are being studied. Anti-beta amyloid antibodies lead to reduced expression of genes related to metabolism and increased expression of genes related to immune regulation and cytokines. Matrix metalloproteinases (MMPs) have been implicated in microhemorrhages and engagement of the Fc portion of the antibodies seems to play a crucial role in microhemorrhage.

The panel discussion covered the potential use of blood-based biomarkers to identify patients who can benefit from treatments, the role of meningeal lymphatics in clearance pathways, and the need for better preclinical models to provide insights for clinical trials. The importance of differentiating mechanisms for BBB opening, fluid movement in edema, and the molecular mechanisms underlying MRI imaging was also emphasized.

Session 3: Future Research Ideas

After the discussion on what we know about ARIA, the focus turned to future preclinical and clinical research ideas, including the exploration of animal models, strategies to protect the blood-brain barrier, and alternative targets for the prevention of vascular damage.  

Animal models that recapitulate disease are an invaluable resource for understanding disease mechanisms and the development of therapeutics. There were presentations on preclinical models for CAA and ARIA in animals such as rodents and non-human primates. There are several species that, like humans, naturally develop CAA, including bears, cats, dogs, and squirrel monkeys. The pros and cons of rodent models of CAA and ARIA were discussed. As models for CAA, mice with familial Alzheimer’s disease (FAD)-related mutations in amyloid precursor protein (APP) have been well-characterized and exhibit CAA which can be increased by comorbid conditions or additional genetic modulation. However, they exhibit parenchymal and cerebral vascular amyloid, CAA develops secondary to parenchymal amyloid plaque pathology, and they rely on non-physiological overexpression of APP with FAD mutations. Transgenic mice with CAA-related mutations in APP have also been well-characterized and, depending on the mutated gene, can present with CAA-related microhemorrhages, CAA type 1, or CAA type 2. However, the CAA results from mutations and is not naturally occurring, microhemorrhage can be inconsistent, and the CAA pathology only develops from a neuronal source of amyloid beta.

Transgenic rats may be a more promising model than mice for CAA. They are phylogenetically closer to humans than mice, have larger brains which facilitates higher-resolution neuroimaging, and can be sampled for CSF. rTg-DI, a rat model that expresses human APP with Dutch Iowa mutations, is good for studying CAA-type 1 (capillary amyloidosis) and CAA-ri. They exhibit increased microgliosis, microhemorrhage, vessel occlusions, and changes in coagulation and complement pathways even in the absence of anti-amyloid monoclonal antibodies. Another rat model is rTg-D (Dutch mutation), which mimics CAA-Type 2 (larger vessels) and has more muted CAA-ri and microhemorrhage. As with the mouse models, these rat models have the drawback of CAA pathology that develops from a neuronal sole source of amyloid beta and familial CAA mutations in the amyloid beta peptide.

Non-human primates (NHPs) are also being used to model CAA and ARIA. Amyloid beta is nearly identical across primates, aged primates naturally develop Aβ deposition, and CAA is universal in aged NHPs. Both effects and side-effects of immunotherapy can be modeled in primates and immunoprevention strategies can be assessed in primates. While primate models offer biological relevance to humans, relatively large brains, and abundant fluids for biomarkers, they also present challenges, including difficulties in imaging Aβ and a long-life stage prior to Aβ deposition. Spontaneous ARIA has been found in an aged squirrel monkey with CAA in the future transgenic marmosets could be developed with an accelerated timeline of disease. NHPs represent a promising model for both CAA and ARIA.

BBB injury is a hallmark of ARIA, and therefore protecting the BBB is essential in combating the negative effects of immunotherapy. The role of fibrin in BBB maintenance was discussed. In AD, fibrinogen binds to component receptor 3 (CR3) on microglia and macrophages, which leads to inflammation. Fibrin deposition in AD, CAA, and ARIA is correlated with activated microglia.  A potential strategy for addressing this is to block the fibrin-CR3 interaction while leaving the soluble form of fibrin intact. The causative role of fibrin in neurodegeneration is supported by genetic and pharmacological evidence, with fibrin being linked to synapse loss and cognitive impairment in AD mice. Blocking fibrin-CD11b (a component of CR3) microglia signaling was shown to prevent behavioral alterations in AD mice, suggesting its potential as a therapeutic target.

An additional strategy proposed for the prevention of vascular damage was targeting medin, another form of vascular amyloid that is present in vessels of CAA patients’ brains. Mouse studies have shown that knockout of medin results in less CAA and microhemorrhages, while overexpression of medin increases CAA. The medin gene (MFG-E8) is elevated in AD patients and loss- of- function mutations in MFG-E8 are protective against cardiovascular disease. Ongoing work in this area includes the development of anti-medin monoclonal antibodies to try to prevent or reduce CAA and possibly mitigate ARIA.

A final strategy for addressing BBB injury is using activated protein C (APC) to protect the endothelial barrier and pericytes. APC is an anti-coagulant with cytoprotective properties and has been shown to reduce hemorrhages in several stroke models. A mutated form of APC, 3K3A-APC selectively eliminates the anti-coagulant activity of APC while preserving cell-signaling and cytoprotective activity and may be a good therapeutic strategy, especially in individuals with ApoE4 alleles. People with ApoE4 alleles have more cyclophilin A in their pericytes, leading to an upregulation in MMP9 and endothelial cell disruption. 3K3A-APC has been shown to reduce hemorrhage, improve neurological outcomes, and have BBB-repairing and anti-inflammatory effects. A proposal was made to test 3K3A-APC for ARIA and its potential to reduce ARIA-E and ARIA-H, taking advantage of its BBB-stabilizing, anti-hemorrhagic, and anti-inflammatory effects.

The first panel discussion in this session addressed the importance of understanding dosing paradigms, blood-based biomarkers, and clinical manifestations of ARIA. The clinical spectrum of ARIA-E and ARIA-H, particularly in ApoE4 carriers, was discussed. The panel emphasized risk stratification, dose titration, and the need for a standardized interpretation of ARIA. They also highlighted the significance of metabolic testing for brain activity and the potential effects of edema and inflammation on symptoms. Almost all AD patients have amyloid angiopathy, which is the single greatest risk factor for ARIA.  We need to see if edema or inflammation is causing symptoms.

The interaction between anti-amyloid immunotherapy and anti-thrombotic agents is under investigation. The use of anticoagulants in trials for certain anti-amyloid drugs has shown increased risk for ARIA. Additionally, the number of previous microbleeds matters. More microbleeds increase the risk for ICH. The risk of thrombolysis and anti-amyloid treatment for individuals with CAA is under scrutiny. Concerns are raised regarding e4/e4 homozygous patients and those with more than 10 microbleeds. ARIA occurs with similar frequency with and without antithrombotic use.

In order to optimize the safety of anti-amyloid therapy, some strategies include careful baseline MRIs, ApoE testing, and risk-benefit discussions. There is a need for imaging biomarkers to assess the risk of ARIA in anti-amyloid recipients. The right outcome measures and the timing of these assessments are crucial. There is also a need for better detection of underlying CAA. Antiplatelet use seems to be safe. Currently, the emergency management of patients with little data and no guidelines is a challenge.

Risk factors for ARIA include ApoE4, existing microbleeds, overt cerebrovascular disease, and existing cerebrovascular risk factors like hypertension and diabetes.  The severity of ARIA does not seem to be directly linked to the degree of amyloid removal in the brain. Potential strategies to reduce ARIA risks include treating patients with low amyloid burden, dosing regimen adjustments, and improving understanding of cerebrovascular risk factors. There is a need to find how to maximize the number of people eligible to receive AAMA and still reduce ARIA rate.

Anti-amyloid auto-antibodies play a significant role in the rare and aggressive inflammatory syndrome CAA-ri which shares significant similarities with ARIA-E. They may even be the same, reflecting exaggerated drug engagement in removing Abeta plaques. ARIA-E seems to have a strong relationship with ARIA-H as 50-70% of ARIA-H occurred after ARIA-E in patients while the rate of isolated ARIA-H (without ARIA-E) is similar between treatment groups and placebo controls (~10%). Based on this, prevention and therapeutic management of ARIA-E may help prevent ARIA-H. This is also comparable with the ARIA Paradox Model which posits that the disassembly of plaques is causally linked to a paradoxical increase in CAA due to the exaggerated mobilization of Abeta, and neuroinflammation that may culminate in CAA-related complications, like ARIA-H (biomarker data would help evaluate this model). Based on TSPO PET binding increasing with ARIA-E, it was suggested that microglial activation plays a role in ARIA. There was also a suggestion that the relationship between anti-Abeta autoantibodies in cerebrospinal fluid and ARIA may be a potential biomarker.

Emerging targets for immunotherapy include blocking the C1s complement pathway, due to the potential role of the complement system in anti-amyloid mAb-related ARIA, as well as the use of anti-inflammatory and neuroprotective drugs like Semaglutide and Liraglutide which improve BBB and vascular health, reduce inflammation and improve cognition. Some outstanding questions include if current mAbs are removing AD plaques at the expense of increased CAA pathology, if this is a transient or long-lasting effect, and if preventative corticosteroid therapy is effective in at-risk patients.

The second panel discussion in this session touched upon patient selection, clinical study designs, real-world data collection, blood-based biomarkers, long-term outcomes, ARIA management, and diagnostic tools for ARIA. It was reiterated that it is still unclear whether ARIA is due to the binding of preexisting vascular amyloid or the enhanced clearance of plaque amyloid on its way out of the brain (getting stuck in the vasculature). The removal of vascular amyloid may allow for better vasoreactivity (e.g., CBF) which in turn may lead to cognitive improvement/stabilization. Better models of CAA/ARIA are needed although mice, rats, and squirrel monkeys seem to be providing information so far.  5XE4 mice develop plaques and CAA – at 8 months of age, ~70% of the amyloid in their brain is in the blood vessels (CAA). We still need more animal models with different ApoE genotypes, including the protective ApoE genotype to test and develop new treatments.

The need for biomarkers to detect CAA was highlighted as existing methods are challenging (MRI-SWI, arterial spin, DWI). It was noted that more advanced MRI methods may be able to provide more accurate and early diagnosis of ARIA, play a role in differentiating between symptomatic and asymptomatic ARIA patients, and even provide mechanistic insights. MMP3 and MMP10 (and potentially sC5b-9) may be good blood-based biomarkers for CAA and may be used to predict ARIA. We need biomarkers to identify the risk of serious ARIA and clinical trials to look at how patient selection may help separate benefits from the risks of Abs therapy. Such studies would help identify why ApoE4 is a risk factor for ARIA-E and ARIA-H and individuals who develop ARIA should be further studied through cognitive testing, identification of potential biomarkers, and use of different dosing regimens and dosing optimization. It would be important to understand if we can dose through ARIA and mild symptoms. Long-term patient outcomes with imaging, lab, and cognitive data should be shared and collected in registries with harmonized data to make comparisons, even across countries. For example, it would be interesting to look at people of African ancestry with ApoE4 genotype who do not have the same elevated risk of AD as Caucasians and if they are also at increased risk of ARIA. Other required work includes the separation of radiographic and symptomatic ARIA. Several mitigation strategies were discussed including combining anti-amyloid mAbs with semaglutide, anti-C1s mAb, or anti-APOE mAb (to poorly lipidated APO found in plaques and CAA).

Scientific Gaps and Priorities

Participants identified gaps in knowledge ranging from mechanistic to clinical, methodological challenges, and urgently needed clinical guidelines.

Some gaps in knowledge and areas for further study included identifying the mechanism of amyloid delivery to vessels, identifying the role of ApoE4 at the vessels where it may change the inflammatory response and lead to hemorrhaging, understanding the role and nature of CAA vs CAA-related inflammation in ARIA and if immunosuppression would be helpful, understanding if some individuals experience greater immune response/inflammation induced by Abs or if their BBB is more susceptible to damage, and exploring the role of cerebrovascular factors such as BBB disruption and the relationship between anti-thrombotic therapy and ARIA. The NVU and reactivity of the vasculature were identified as a hypothesis to look at as a mechanism for protective effect. In order to address such gaps, the development and use of animal models (including larger animals and non-human primates), such as knock-ins and next-generation models were thought to be valuable. There was also an expressed interest in PET ligands to identify CAA, PET and MRI imaging for detecting and differentiating ARIA subtypes, and biomarkers for ARIA susceptibility and vasoreactivity, especially in the blood but also in the CSF and plasma. It was noted that utilizing existing datasets for identifying blood-based biomarkers may be a good strategy.

From a clinical perspective, some identified gaps include integration with anticoagulation and thrombolysis, a need for more neuropathology of ARIA cases, looking at the correlation between the number of injections/dosing and ARIA risk, and a study of the long-term impact of symptomatic vs asymptomatic ARIA on disease progression and cognitive outcomes. There is a need for biomarkers and criteria to stratify patients based on risk for ARIA and significant inflammatory reactions (e.g., ApoE genotype, age, amyloid load) as prolonged imaging protocols are challenging and may not be clinically feasible. Participants expressed interest in efforts for data sharing and collaborations to establish registries with imaging, lab, and cognitive data. Similarly, the development of guidelines and emergency care protocols for ARIA cases, including software for standardized readouts and clinician support was requested. Finally, alongside efforts to address ARIA, it was noted that consideration and effort towards non-amyloid therapies should not be eliminated with dual vaccines for treating amyloid-related diseases and long-term active immunization for preventing ARIA also being viable paths to explore.