Low plasma neurofilament light levels associated with raised cortical microglial activation suggest inflammation acts to protect prodromal Alzheimer’s disease

Neurofilaments are found in the cytoplasm of neurons and act to transport nutrients and other molecules along axons. The subunit proteins constituting neurofilaments include neurofilament light (NfL), medium and heavy (L, M and H) proteins based on their molecular weight [1]. The neurofilaments control axonal diameter, and it has been reported that cerebrospinal fluid (CSF) levels of NfL provide a non-specific marker of axonal degeneration. Patients with amyotrophic lateral sclerosis where pyramidal tract involvement is present show greatly raised CSF NfL levels, while those with only peripheral motor neuron disease (progressive muscular atrophy) show smaller elevations [2, 3]. Multiple sclerosis cases in the chronic progressive phase of their disease also show raised CSF NfL, but those who have early relapsing remitting disease show only modest elevations [4]. Raised levels of CSF NfL have also been reported in Alzheimer’s disease (AD) and Huntington’s disease [2, 5, 6]. No relationship was found between cognitive status and CSF NfL levels in AD patients, though disability of Huntington’s disease cases, when rated with the UHDRS, correlated with CSF NfL levels [6].

More recently, an ultrasensitive digital array for enzyme-linked immunosorbent assay (ELISA) of proteins including NfL has been developed [7], which allows reliable detection of single molecules of NfL in plasma or serum. NfL plasma levels were found to correlate strongly with those measured in CSF. A large number of studies have now shown that blood NfL is a robust biomarker of neuronal injury, irrespective of the cause [8]. In the context of AD, serum NfL has been found to mark the onset of neurodegeneration in subjects at risk for familial disease [9]. A recent study found increased plasma NfL in AD patients compared to nondemented controls [10]. A larger study involving the Alzheimer’s Disease Neuroimaging Initiative (ADNI-1) cohort reported raised plasma NfL in sporadic mild cognitive impairment (MCI) and Alzheimer cases [11]. The plasma NfL levels correlated with reduced CSF β-amyloid 1–42 and raised CSF tau in the MCI cohort. At baseline, high plasma NfL levels in the MCI and AD cohort were associated with worse Mini-Mental State Examination (MMSE) and ADAS-COG 11 scores and subjects had larger ventricular and smaller hippocampal volumes and thinner cortices.

Microglia are part of the intrinsic immune defence of the brain but normally exist in a resting state monitoring the local constituents of brain extracellular fluid [12]. In AD, there is abnormal formation of extracellular plaques of β-amyloid protein fibrils and intraneuronal tangles of hyperphosphorylated tau. This characteristic pathology of AD is associated with an inflammatory response which involves activated microglia surrounding plaques and degenerating neurons [13]. Activated, but not resting, microglia express the translocator protein (TSPO) on the outer membranes of their mitochondria [14]. ¹¹C-PK11195 positron emission tomography (PET) is an in vivo marker of the TSPO expressed by activated microglia. Recently, we reported that 80% of β-amyloid-positive MCI cases showed associated active inflammation and raised TSPO [15]. However, whether the inflammation was playing a protective or cidal role in the disease was unclear. It has been reported in established AD that levels of temporoparietal cortical microglial activation correlate inversely with cognitive ratings on the MMSE [16]. This suggests that the microglial phenotype may be toxic in these subjects. However, microglia are known to also express protective phenotypes releasing growth factors, clearing amyloid fibrils, and remodelling damaged synapses. Here, we report the presence of an inverse association of plasma NfL levels, acquired from cases with either prodromal (β-amyloid-positive MCI) or early clinical β-amyloid-positive AD, with their levels of cortical microglial activation measured with ¹¹C-(R)-PK11195 PET [17]. These findings are corroborated by measurements of atrophy and microstructural integrity using magnetic resonance imaging (MRI).

Twenty-seven MCI/AD subjects (8 females/19 males) were included in the analyses. The subjects had a mean age of 73.6 years (± 6.1), mean MMSE score of 25.9 (± 3.0), and a mean plasma NfL level of 27.9 pg/mL (± 10.7). One AD subject had no DWI sequence due to claustrophobia, so MD analyses only include 26 subjects.

Plasma NfL correlated inversely with PK11195 BP_ND levels in areas of frontal, parietal, precuneus, occipital, and sensorimotor cortices in the 27 MCI/AD cases (Fig. 1). MD correlated positively with plasma NfL levels in areas of the temporal lobes and the cingulate cortex (Fig. 2a). Levels of MD correlated inversely with PK11195 BP_ND primarily in parietal areas (Fig. 2b) and overlapped with areas of PK11195–NfL inverse correlation. It should be noted that not all subjects had whole-brain coverage for the DWI sequence, which led to lack of statistical inference in the most superior and inferior parts of the cerebrum (shown as dark grey areas on the figures involving MD). Finally, cortical thickness correlated inversely with plasma NfL levels in a single cluster in the left frontal lobe (Fig. 2c). There were no correlations between cortical thickness and PK11195 binding. Additionally, no correlations were detected between MMSE ratings and plasma NfL levels (ρ = − 0.057, P = 0.78) or cortical PK11195 binding, or between CDR-SOB scores and plasma NfL levels (ρ = 0.257, P = 0.20) or PK11195 binding.

This paper is the first to report in vivo an inverse association of plasma NfL concentrations with levels of cortical microglial activation imaged with PET in patients with prodromal or early mild AD confirmed by the presence of raised cortical β-amyloid deposition detected with ¹¹C-PiB PET. Since plasma NfL levels are now a well-established biomarker for the extent of axonal damage, these results suggest that cortical microglial activation in early Aβ-positive individuals may initially play a protective role.

Plasma NfL levels have been shown to correlate strongly with those in CSF [7]. A recent study by Ashton et al. found that plasma NfL concentrations correlated with both severity of post-mortem neurofibrillary tangle pathology and nerve cell loss from neurodegeneration [29]. We have previously found PK11195 binding to be associated with amyloid load in prodromal AD [15]. In that series, it was noted that amyloid-positive MCI/AD subjects with absent or very low tau load still presented with high PK11195 binding [21]. This finding emphasises that β-amyloid aggregation can trigger an inflammatory brain response in vivo independent and ahead of the presence of tau tangles.

Activated microglia can be seen surrounding amyloid plaques and degenerating neurons containing tau tangles in Alzheimer brain slices with immunohistochemical stains [30]. The role of these microglia is still being debated. It has been suggested that initially microglia may be protective, attempting to clear toxic amyloid oligomers in early disease and releasing nerve growth factors. Subsequently, they fail to achieve this function and their levels decline. A later wave of microglia is then activated as tau tangles form and their load rises and these microglia express a toxic phenotype releasing cytokines which kill neurons and their axons so driving disease progression [17]. PK11195 PET is unable to discriminate between protective and cytotoxic activated microglia as both these microglial phenotypes express TSPO. If the majority of activated microglia in our MCI and early AD cohort exhibit a protective phenotype, then this could explain our observation of an inverse correlation between cortical microglial load and plasma NfL levels.

Surface-based cortical statistics showed that areas where brain inflammation levels correlated with plasma NfL levels in our MCI/AD group were localised in frontal, cingulate, parietal, precuneus, occipital, and sensorimotor cortices. The first four areas are particularly targeted by amyloid deposition [31], while occipital and primary sensorimotor cortices usually are relatively spared. Pyramidal cells in layers 3 and 5 of association cortical areas project to the brainstem and spinal cord, and hence, it is most likely that degeneration of these axons is responsible for the leakage of NfLs into the plasma.

We also found cortical areas where there was an inverse correlation between MD and PK11195 binding. These areas overlapped with areas of NfL–PK11195 inverse association, but were smaller in extent. As both NfL and MD are considered surrogate markers of axon loss, this result could corroborate a model of activated microglia in areas of intact microstructure. MD correlated positively with NfL in areas associated with early involvement in AD. Specifically, the left medial temporal lobe, the right lateral temporal lobe, and the cingulate cortex showed this association. These areas may represent a later disease stage where microglia are no longer protective and neurodegeneration has started. We found only a few small areas of inverse correlation between plasma NfL levels and cortical thickness. Cortical thickness is not sensitive to the initial neurodegeneration on an individual basis, and with the small sample size studied here; the statistical power is probably insufficient to pick up strong correlations for cortical thickness measurements.

While other studies have found association between cognitive function, e.g. MMSE and NfL [32], we found no associations between cognitive function (MMSE, CDR-SOB) and neither plasma NfL levels or cortical PK11195 binding. This could be due to our small sample size and the fact that it consist of MCI/early AD cases.

We have found that in early and prodromal Alzheimer patients raised plasma levels of NfL protein, a marker of axonal neurodegeneration, are inversely correlated with raised levels of PK11195 binding, a marker of microglial activation / neuroinflammation, in cortical areas. As raised plasma NfL reflect active neurodegeneration, our finding supports the view that inflammation initially plays a protective role in AD. Both these modalities have the potential to provide useful biomarkers in future studies on the efficacy of neuroprotective agents. The inverse correlation of levels of raised activated microglia in early AD with reduced plasma levels of NfL could suggest that initially boosting microglia activation, possibly with immunotherapy targeting abnormal protein aggregations, in early disease could be protective, but this hypothesis needs to be tested in further studies.