Cerebrospinal Fluid Biomarkers for Cerebral Amyloid Angiopathy Diagnosis

Sporadic cerebral amyloid angiopathy (CAA) is an extremely common, well-defined, and clinically relevant small vessel disease of the brain, now recognized as a major cause of lobar intracerebral hemorrhage (ICH) in older individuals and a key vascular contributor to cognitive impairment, even in patients with presumed Alzheimer’s disease [1]. CAA is a complex progressive microvascular degenerative condition, with the neuropathological hallmark being various degrees of amyloid-β (Aβ) accumulation within the small arteries, arterioles, and capillaries of the cortex and leptomeninges [1]. The “big bang” of the CAA field has been the device and application of the clinico-radiological Boston criteria which allowed for CAA diagnosis during life in different clinical and research settings [2]. While core hemorrhagic structural brain MRI markers (including strictly lobar cerebral microbleeds and cortical superficial siderosis) within the framework of the Boston criteria have allowed us to make the CAA diagnosis with increasing certainty, these criteria are not without major limitations [2]. For instance, the above structural brain imaging markers associated with CAA are snapshots of irreversible measures of advanced disease and represent only the “tip of the iceberg” [3]. Some of these markers are not exclusive of CAA. Also, it is now becoming evident that CAA is not a single disease, but there are likely divergent pathophysiological phenotypes and clinical trajectories [1, 4]. Existing diagnostic criteria do not provide a straight-forward way of defining disease severity, predominant CAA phenotype, expected clinical trajectory or the contribution of CAA in other clinical settings not fulfilling the criteria, including Alzheimer’s disease or patients with another coexisting small vessel disease (e.g., mixed haemorrhages).

Cerebrospinal fluid (CSF) molecular biomarkers might offer an attractive complimentary approach for addressing some of these limitations in the CAA field by providing physiologic, dynamic measures of the underlying disease process. However, very sporadic studies and of limited sample size and quality have been published in this area [5]. The main focus of these studies has been on core CSF biomarkers of amyloid metabolism (Aβ₄₂, Aβ₄₀), neurodegeneration (t-tau), and tangle pathology (p-tau) in patients with symptomatic CAA [5]. One of the principal challenges, which might explain the lack of large scale studies on this otherwise attractive topic, is that most patients with Alzheimer’s disease have evidence of some CAA on autopsy, while a large proportion of otherwise healthy elderly persons have evidence of both Alzheimer’s disease and CAA neuropathologically [6]. In a recent meta-analysis of relevant studies, pilot data for a distinct core CSF profile for symptomatic CAA patients (5 CAA patient cohorts, n = 59 patients largely without dementia) emerged in comparison to either healthy controls or patients with Alzheimer’s disease (n = 94 and n = 158, respective) [5]. This CSF profile entailed the presence of both lower Aβ₄₂ and Aβ₄₀ levels when compared to controls, with Aβ₄₂ levels close to what is seen in patients with Alzheimer’s disease, but lower Aβ₄₀ levels. CSF t-tau in CAA showed intermediate levels relatively to healthy elderly controls and to patients with Alzheimer’s disease (i.e., higher than controls, but lower than Alzheimer’s disease) [5]. These findings were deemed preliminary and larger well-designed studies were suggested to validate the clinical use of CSF biomarkers in CAA.

The study by Grangeon et al. [7] in this issue provides a long-awaited validation on core CSF biomarkers in CAA. The authors present the largest cohort to-date, including 63 probable CAA patients according to the modified Boston criteria, and hence doubling the total sample size of available evidence on core CSF profile (t-Tau, and p-Tau, Aβ40 and Aβ₄₂) versus Alzheimer’s disease (with no MRI evidence of CAA) and versus controls (without any neurological conditions). In overall group comparisons, CAA versus Alzheimer’s disease patients showed lower levels of Aβ₄₀ (p = 0.001), t-tau (p = 0.008), and p-tau (p = 0.004), but quite similar Aβ₄₂ level (p = 0.07) [7]—findings in line with previous work. CAA had higher Aβ_42/40 ratios compared to Alzheimer’s disease.

The authors went a step further and defined three distinct CSF profiles in probable CAA patients [7]: (a) a profile similar to Alzheimer’s disease patients (i.e., decreased CSF Aβ₄₂ level associated with elevated p-tau and/or tau level), which was the most common profile (50.8%); (b) isolated decreased Aβ₄₂ level, without any CSF tau alterations (34.9%); and (c) normal Aβ₄₂ level with or without increased t-tau or p-tau (found in only 9 patients) [7]. Of note, these 3 CAA groups were not different in demographic, clinical, cognitive, and MRI characteristics, suggesting that the CSF profile might capture pathophysiological processed beyond the clinical-MRI phenotypes of CAA. When they compared Aβ₄₀ level between these CAA patient groups based on CSF profiles, no difference was found between group (a) and (b), while higher Aβ₄₀ level was found in group (c) compared to other profiles [7]. In other words, the authors demonstrated that Aβ₄₀ level was selectively reduced in all three groups of probable CAA patients, despite the potential for concomitant Alzheimer’s disease pathology (captured for example by group (a)). Aβ₄₀ level have high diagnostic potential for identifying CAA and discriminating it from Alzheimer’s disease across the spectrum of potential CAA phenotypes (from pure CAA to CAA-Alzheimer’s disease). The level of Aβ₄₀ did not seem to be affected by the CAA severity as captured on brain MRI (lobar cerebral microbleeds and cortical siderosis burden) [7], indicating that MRI and CSF biomarkers capture different pathophysiological facets of the disease process. I find the authors’ approach robust and novel in terms of potential future clinical applicability. The reduced level of CSF Aβ₄₀ in CAA is thought to be secondary to “selective trapping” of this species in the microvascular networks, in contrast with Alzheimer’s disease, where only the Aβ₄₂ species is found “trapped” in the parenchyma.

Despite some inherent limitations of the study, the most notable being the retrospective design and hence the inclusion of a “convenience” sample of patients, not standardized MRI blood-sensitive sequences protocols etc. [7], I would argue that this type of selection bias would have resulted in attenuated CSF differences, not in clear between-groups differences as noted. There is now increased focus and some preliminary data on other amyloid species and non-amyloid biomarkers that might potentially discriminate cerebrovascular Aβ pathology (aka CAA) from parenchymal Aβ deposition (Alzheimer’s disease): Apolipoprotein D (lower in CAA) [8], Aβ₃₈ (lower in CAA) [9], neuroleukin (elevated in Alzheimer’s disease) [10], among others. One can envision a future CSF biomarkers panel that could facilitate the diagnosis and monitoring of CAA in certain clinical settings.

How should the practicing clinician apply the current CSF data for CAA diagnosis? Despite that MRI-based Boston criteria appear to be the most reliable diagnostic approach for CAA, there is likely a role for incorporating CSF makers in certain clinical settings. In borderline scenarios where the Boston criteria are not fulfilled (e.g., patients with another coexisting small vessel disease, mixed haemorrhages, etc.), but there is still significant concern for underlying CAA that might have important implications for treatment decisions, one could obtain CSF to look for low Aβ₄₀. Since Aβ₄₀ if often not commercially available, another reasonable approach would be to assess Aβ₄₂—intermediate Aβ₄₂ levels (lower that the general population, but higher than expected in Alzheimer’s disease), especially in the absence of significantly increased p-tau [5] in a patient without dementia, would make a CAA diagnosis possible in the appropriate clinical setting.

The findings by Grangeon and colleagues [7] have relevance for clinical diagnosis and research in the CAA and Alzheimer’s disease fields. The current study [7] is an important one in the expanding universe of CAA biomarkers. It coincides with a new era of rediscovered interest in CSF and plasma CAA biomarkers in larger sample sizes compared to what we have been seeing in the past. Further well-designed studies are obviously needed to explore how core CSF markers could be practically integrated to the Boston criteria, though such studies might not be feasible given that the gold-standard for a definite CAA diagnosis in such validation studies remains whole brain autopsy. As a next step to fuel rapid evidence synthesis in the field, independent of the precise CSF analytical method used and of different cohorts, an updated meta-analysis of fluid biomarkers in CAA is currently ongoing [5].