Changes in the locus coeruleus during the course of Alzheimer's disease and their relationship to cortical pathology

In Alzheimer's disease (AD), the locus coeruleus (LC) undergoes early and extensive neuronal loss, preceded by abnormal intracellular tau aggregation, decades before the onset of clinical disease. Neuromelanin‐sensitive MRI has been proposed as a method to image these changes during life. Surprisingly, human post‐mortem studies have not examined how changes in LC during the course of the disease relate to cerebral pathology following the loss of the LC projection to the cortex.


INTRODUCTION
Alzheimer's disease (AD) is a neurodegenerative disorder currently affecting over 50 million people globally.The pathology includes gradual neuronal and synaptic loss, accumulation of extracellular amyloid-beta (Aβ), intracellular hyperphosphorylated (p)-tau and neuroinflammation.The locus coeruleus (LC) is a tiny brainstem nucleus that innervates the cerebral cortex with noradrenaline via its widespread projections, modulates sympathetic nervous system activity, regulates arousal [1] and the sleep-wake cycle [2][3][4] and plays a role in memory [5,6] and attention [7].The LC has been shown in numerous studies to undergo significant degeneration in AD, showing an average 60% reduction in cell number when compared with controls of a similar age [8][9][10].Cell loss occurs early in the disease course with neuronal loss observed at the prodromal AD or mild cognitive impairment (MCI) stage [11] and is progressive, correlating with both cognitive impairment [12] and disease severity [13].Additionally, ptau pathology is greater in the LC in those with MCI or AD compared with controls [14] and accumulates throughout the disease process [15,16].Furthermore, ptau has been observed in the LC at a very early age, prior to being observed in any other brain areas and in the absence of any cortical lesions [17,18].Therefore, the LC appears to be vulnerable to tau pathology decades before disease and symptom onset and prior to LC neuronal loss [16,19].Unsurprisingly, cortical noradrenaline is reduced in AD patients [20][21][22], likely contributing to neuroinflammation as noradrenaline has been shown to have an anti-inflammatory effect on microglia, suppressing proinflammatory cytokine secretion [23][24][25] and reducing oxidative stress [26].However, how LC neuronal loss relates to cortical inflammation as well as inflammatory changes within the LC throughout disease progression are unexplored in humans.While studies have looked at the effects of LC-NA dysfunction on cortical inflammation, inflammatory changes within the LC itself have not been examined.
A distinguishing feature of LC neurons is that they contain a high concentration of neuromelanin, a dark brown granular pigment that accumulates within the cytoplasm of the cell body with age [27].
Though the mechanisms behind neuromelanin synthesis are not fully understood, it is thought that excess dopamine/noradrenaline is oxidised by iron catalysts and stored within organelles as neuromelanin, therefore protecting the neurons from cellular damage caused via oxidative stress [28,29].It has been hypothesised that when LC neurons degenerate, the insoluble neuromelanin is released into the extracellular space where it provokes an inflammatory response, which may be associated with further neurodegeneration [30].While observations of the presence of extracellular neuromelanin in either the neuropil or within macrophages in the LC of AD samples have frequently been commented on, this has not been quantified or correlated with dementia severity [15,31,32].Here, we present a comprehensive analysis of the changes in the LC across AD progression and relate them to changes observed in the temporal cortex, an area receiving input from LC.Our observations will inform the interpretation of the neuromelanin-sensitive MRI method as a tool to stratify patients who may benefit from treatment targeting noradrenergic dysfunction.

Cases
Tissue from the pons containing the rostral/middle LC and the middle temporal gyrus (TL), an area typically affected by AD pathology that the LC projects to but from which the LC does not receive afferent projections, was provided by the South West Dementia Brain Bank
Sections were treated with hydrogen peroxide (H 2 O 2 ) to bleach the neuromelanin in order to distinguish it from the chromogen used to visualise the protein of interest as both are a similar brown colour.

Key Points
• Quantification of neuromelanin in the locus coeruleus (LC) during the course of AD with assessment of LC changes in relation to AD pathology and neuroinflammation in the temporal lobe of 59 post-mortem human samples.
• Neuromelanin loss and ptau accumulation are early pathologies in the LC, before the neuroinflammatory reaction.
• The presence of extracellular neuromelanin deposits may reflect the selective vulnerability of LC neurons.
• Our findings support the use of neuromelanin-sensitive MRI methods as noninvasive tools to explore early AD changes.
F I G U R E 1 Quantification and illustrations in the locus coeruleus (LC) and temporal cortex of Aβ load (A), ptau load (B) and (C) spongiosis status (%) as a marker for cortical integrity during the course of Alzheimer's disease based on the Braak stage as a marker of disease severity.Immunostaining in LC was obtained after bleaching pretreatment.Asterisks in (A) denote LC cells that are only faintly visible by their cell membrane.The neuromelanin pigment, which is visible in Figure 2B, has been removed following bleaching leaving only the DAB reaction product.Counterstaining: haematoxylin.Scale bar = 50 μm.
the avidin-biotin-peroxidase complex method (Vectastain Elite, Vector Laboratories) with 3,3 0 diaminobenzidine as the chromogen and 0.05% H 2 O 2 as substrate (Vector Laboratories).All sections were counterstained with haematoxylin, dehydrated and mounted in Pertex (Histolab Products AB).The staining was performed in two batches with each batch containing cases from all Braak stages.All experiments included a negative control slide incubated in the absence of a primary antibody and a positive control tissue known to express the protein of interest.
In unbleached sections, neuromelanin in the LC was visualised using the Fontana-Masson stain kit (melanin stain, Abcam) used according to the manufacturer's protocol.Briefly, deparaffinised slides were placed in warm ammoniacal silver solution followed by incubation in gold chloride solution (0.2%), then sodium thiosulphate solution (5%) before the Nuclear Fast Red solution.Slides were dehydrated and mounted in Pertex.Neuromelanin pigment was identified as black by this method.A set of sections was stained with haematoxylin and eosin (H&E) to assess the tissue integrity.

Quantification
Stained sections were scanned using the automated slide scanner microscope Olympus VS110 v2.9.1 (Olympus America Inc).For the temporal gyrus, 30 regions of interest (ROIs) of 500 μm 2 were extracted from the grey matter of the adjacent sulcus with the Olympus VS-Desktop software.For the pons sections, the LC was identified on the H&E stain by the presence of large, pigmented cells.A singular square ROI of 1.67 mm 2 placed over the centre of the LC was sufficient to provide coverage of the LC in its entirety, allowing assessment of the whole structure.Quantitative image analysis was performed on the ROIs using ImageJ v1.53c, a Javabased imaged processing program (National Institutes of Health USA [36]).For each marker or stain (neuromelanin, H&E), an optimal threshold was selected to capture the staining.The area fraction measure function was used to calculate the percentage area labelled by the antibody or the stain expressed as protein load (%) for the marker.Quantification of images was performed blind to the Braak stage group.H&E images were analysed to identify vacuolation within the cortex (status spongiosis).The percentage area of residual tissue reflected the cortical tissue integrity, with higher status spongiosis corresponding to lower cortical integrity values (i.e., cortical neuropil and neuronal degeneration) [37].
LC neurons (either pigmented or unpigmented) were counted based on the presence of the cell body and with an area greater than 690 μm 2 , with size and shape also assessed.
For neuromelanin, several assessments were performed on the Fontana-Masson stain: the overall percentage area of stained neuromelanin area as described above, the amount of neuromelanin (%) per LC neuron and extraneuronal neuromelanin determined as the overall amount of neuromelanin minus the neuromelanin in all LC neurons.
The total protein concentration of the samples was determined using a Bicinchoninic Acid (BCA) assay (Pierce BCA Protein Assay Kit, Thermoscientific); 25 μL of brain homogenate was used for the proinflammatory and cytokine panels at 1:2 dilution and for the chemokine panel at 1:4 dilution.Each plate was read using a Meso Quickplex SQ120 according to the manufacturer's instructions.Protein levels for each analyte were obtained in pg/mL and normalised to the total protein concentration obtained by the BCA Assay.

Statistical analysis
Normality of the data was assessed for each variable using histograms, descriptive statistics and Q-Q plots.Baseline characteristics were tested for significant differences using analysis of variance (ANOVA) for age and post-mortem delay and Pearson χ 2 test for gender.For comparison between values across the three Braak groups, the ANOVA U-test with Tukey post-hoc test was used for normally distributed data and the Kruskal-Wallis test with Dunn's post-hoc test for nonparametric data.Pearson's or Spearman's rank correlations were used as appropriate to explore the relationship between the two brain areas.Bonferroni corrections were applied to adjust p values for the correlations and the MesoScale Discovery data to correct for multiple comparisons.All adjusted p values were adjusted with a threshold of 5% used for intergroup comparisons and 1% for correlations.
Statistical tests were performed using SPSS (IBM Statistical Package for Social Sciences v27) and graphs prepared using GraphPad Prism (Version 7.00 for Windows, GraphPad Software, La Jolla California USA, www.graphpad.com).

Demographics of the cases
The 59 cases were divided into three groups based on their Braak stage defined as 0-II, III-IV and V-VI.There was no significant difference in the proportion of males to females across the three groups as determined by the Pearson χ 2 test (χ 2 (2) = 2.179, p = 0.336).The ANOVA U-test showed a statistically significant difference in age at death between groups (F(2,56) = 3.339, p = 0.043), and post-hoc comparisons revealed that age at death was lower in Braak group V-VI when compared with Braak group III-IV (p = 0.050).No difference in age was detected between Braak group V-VI and 0-II and between Braak group III-IV and 0-II.There was no significant difference in post-mortem delay between Braak groups (H(2) = 3.118, p = 0.210) (Table S1).

Aβ and tau pathology
In the LC, Aβ pathology showed significant differences between Braak groups (F(2,51) = 5.826, p = 0.005), with a higher Aβ load in Braak group V-VI than in Braak group 0-II ( p = 0.004).There was no difference between Braak groups 0-II and III-VI or between Braak groups III-IV and V-VI.In the TL, significant differences were also detected between Braak groups (F(2,55) = 22.88, p < 0.001) with a higher Aβ load in Braak group III-IV compared with Braak group 0-II ( p = 0.006) and in Braak group V-VI compared with Braak group III-IV ( p = 0.002) (Figure 1A; Table 1).
In the LC, ptau load was significantly different between Braak groups (H(2) = 30.826,p < 0.001), with increased load in Braak group III-IV compared with Braak group 0-II ( p = 0.002) and in Braak group V-VI compared with Braak group 0-II (p < 0.001).In T A B L E 1 Summary of protein loads for each marker by Braak group analysed in the locus coeruleus and temporal lobe.1B; Table 1).
Expression of ptau in the LC and TL was associated with Braak stage in (r s = 0.777, p < 0.001; r s = 0.865, p < 0.001, respectively) (Table 2) and correlated in LC vs TL (r s = 0.694, p < 0.001) with higher ptau in LC at Braak group 0-II (p < 0.001), while ptau load was greater in the TL at Braak group III-IV ( p = 0.034) and Braak group V-VI ( p < 0.001) (Table 3).

Cortical integrity
Neuropil integrity, assessed as the inverse of status spongiosis, showed no difference between Braak groups in the LC (H(2) = 3.637, p = 0.162), while a significant difference was observed in the TL (H(2) = 7.937, p = 0.019) with reduced cortical integrity (i.e., increased status spongiosis) in Braak groups V-VI vs III-IV ( p = 0.018) (Figure 1C; Table 1).No association was observed with the Braak stage for either area (Table 2), but lower neuropil integrity in LC vs TL was observed at Braak stage 0-II ( p = 0.001) and III-IV ( p < 0.001; Table 3).

LC neurons and neuromelanin
LC neurons were identified on the Fontana-Masson stain by the presence of neuromelanin pigment, and only cells with a minimum area of 2000 pixels were included and compared across the Braak groups.
There was no difference in the size of the LC neurons as determined by the area between Braak groups (F(2,55) = 0.365, p = 0.696), nor was there a difference in the shape of the LC neurons as determined by their roundness across Braak groups (F(2,55) = 0.301, p = 0.741).
Iba1 expression was correlated with the Braak stage only in LC (r s = 0.467, p < 0.001; Table 2).Significantly more Iba1 load was detected in LC vs TL at Braak group III-IV ( p = 0.006) and Braak group V-VI (p < 0.001; Table 3).
For CD68, similar to HLA-DR, no significant difference was detected between Braak groups in LC or TL (H(2) = 0.892, p = 0.640 and H(2) = 1.430, p = 0.489, respectively) (Table 1).No correlation was detected with the Braak stage for either area or in LC vs TL (Tables 4 and 5).

Associations between the different markers in LC and temporal cortex
We then explored the possible relationships between the different markers in LC and TL.In LC (Table 4), Aβ and ptau loads were related (r s = 0.457, p < 0.001).Iba1 was the only microglial marker with associations as follows: with higher Aβ (r s = 0.409, p = 0.004), ptau (r s = 0.647, p < 0.001) and extracellular neuromelanin (r s = 0.453, p < 0.001).
The temporal lobe had only one association between Aβ and ptau load (r s = 0.748, p < 0.001), as expected (Table 5).
Of note, no association was observed between the age at death, post-mortem delay or dementia duration with the different markers in LC and temporal cortex (data not shown).
In LC, significant differences were observed in (Table S3): In the TL (Table S4), of the 30 analytes examined, differences between Braak groups were detected only for IL-15 (H(2) = 7.461, p = 0.024), which saw a significant increase in Braak group V-VI compared with Braak group 0-II ( p = 0.002).There was no significant difference between Braak groups 0-II and III-IV or III-IV and V-VI.

DISCUSSION
While studies have demonstrated that the LC undergoes degeneration in AD [11][12][13], with ptau accumulating first in the axons and then in the LC neuronal bodies before being detected cortically [17], the associations with disease progression, LC projections to the cerebral cortex and the neuroinflammatory response have not been explored.
Our findings showed the loss of LC neurons to be associated with AD progression, with the neuronal number being relatively preserved in the earlier Braak group (0-II), and dramatically reduced after Braak stage III-IV, at the time of and not before the appearance of the cognitive symptoms, usually defined as prodromal AD.On average, there was a 23% decrease in neuronal cell number between controls and those with prodromal AD (Braak group III-IV) and a further 65% decrease in neurons between those with prodromal AD and those with AD, in line with previous research [11,12].A corresponding decrease in DβH expression was observed, associated with the number of LC neurons.Although DβH expression detected noradrenergic activity in both axons and cell bodies, this observation suggests that decreased DβH expression is driven by the reduction in LC neurons.
TH expression was also decreased in Braak group V-VI, but there was no difference between group III-IV and V-VI.Thus, DβH expression was more tightly correlated with the number of LC neurons and is proposed to be the preferred marker to identify noradrenergic neurons in the LC.Neuromelanin expression was decreased early in the disease process, while the amount of neuromelanin per neuron remained unchanged until the late stage of the disease, suggesting that the neuromelanin reduction might be driven by the loss of LC neurons and not by the loss of pigment within the neurons.However, LC neuronal loss was very severe in the late stage of the disease.
Therefore, the neuromelanin loss might have been initially higher than or somehow preceded cell loss, inferring that the most heavily pigmented LC neurons may degenerate first and that, at the later stages of disease, the surviving LC neurons appear to contain less pigment.
tangles in the cerebrum [17,19].Tau pathology significantly increased in the LC with disease progression and was already observed at greater amounts in prodromal AD (Braak group III-IV) [39], as previously published [14,16].Of note, tau pathology in the LC continued to increase despite the loss of LC neurons, suggesting that the surviving LC neurons and axons contain high amounts of tau.However, from Braak stage III-IV onwards, more ptau was detected in the temporal cortex than in the LC, potentially due to the loss of LC neurons.
The cortical integrity assessment showed neuropil degeneration only in the later stages of the disease in the temporal cortex, as expected.Indeed, cortical atrophy is an important pathological and imaging feature of AD, but this is usually not detected until symptoms are pronounced [40].On the other hand, AD severity had no association with neuropil degeneration in the LC.This was perhaps surprising as it was predicted that neuronal loss occurring in the LC would be associated with neuropil degeneration, but it may reflect the fact that LC axonal projections and associated synapses are in the cerebral cortex.
In the LC, Iba1 expression, a marker of microglial homeostasis [34], was increased with disease progression and associated with accumulation of ptau and extraneuronal neuromelanin, but not with LC neuronal loss.Of note, Iba1 expression was significantly higher in LC in the late stage of AD, while significant increases of ptau and extraneuronal neuromelanin were observed from an early stage (Braak stage III-IV).This implies that ptau deposition and loss of neuromelanin by LC neurons may precede the neuroinflammatory response.The expression of HLA-DR or CD68 was not changed, and thus the microglial changes in the LC appear to be towards a physiological (Iba1) rather than a pathological phenotype (CD68, HLA-DR).In the cortex, we did not detect changes in CD68 expression, as previously reported on larger cohorts [35,41].Nevertheless, our investigation of two different regions in the same brain, both affected by the pathology, highlights the complexity and heterogeneity of the microglial response in humans [42,43].
Assessment of extraneuronal neuromelanin showed an association with the Braak stage.Although qualitative observations of extraneuronal neuromelanin deposits have been reported [15,31,32], this has never been quantified.It was hypothesised that as LC neurons degenerate, insoluble neuromelanin would remain in the extracellular spaces where it will likely have a toxic effect on nearby neurons, if not readily phagocytosed or degraded [44], consistent with the negative association between extraneuronal neuromelanin and LC neuronal number that we report.In vitro experiments suggested that extraneuronal neuromelanin deposits induce an inflammatory reaction induce chemotaxis recruiting immune cells to inflammatory sites, these findings also advocate for dysfunctional microglia as previously discussed.Yet this observation was not replicated in the temporal cortex, also emphasising the neuroanatomical heterogeneity of the inflammatory response.One limitation is that we were unable to solely isolate the LC from the tissue provided and so had to homogenise tissue that included not only the LC but also the surrounding pons.Therefore, the results included concentrations of markers from both the LC and surrounding pons, which may have affected the findings.
Very few studies have explored the LC in living patients using the neuromelanin-sensitive MRI, an imaging technique developed to explore LC changes [47].The LC signal on NM-MRI has been observed to correlate with the LC volume measured post-mortem [48,49] and thus is considered a reliable method to explore the LC signal in vivo.Interestingly, our study could explain some of the observations reported.In one of the largest case-control studies [50], LC signal was reduced in mild AD participants mirroring our observation of decreased neuromelanin in those with Braak stage III-IV, rather than a loss of LC neurons, thus supporting the use of neuromelanin MRI method as a useful noninvasive tool to explore LC signal as an early biomarker in AD.

CONCLUSIONS
As AD progresses, LC neuronal number decreases while Aβ, ptau and extraneuronal neuromelanin deposition appear, which is accompanied by reactive microglia at the end stage of the disease.Our observations highlight that ptau and extraneuronal neuromelanin deposition are the earliest pathologies to occur in the LC, before the neuroinflammatory reaction, and that the extracellular neuromelanin deposits identified early in the disease may reflect and influence the selective vulnerability of the LC neurons in AD.Our findings support the use of neuromelanin-MRI as a sensitive technique to identify early changes in LC during life.

(
Bristol) from 59 donors.Cases were divided into three groups based on cortical Braak staging of ptau pathology: Braak stage 0-II (n = 19), Braak stage III-IV (n = 20) and Braak stage V-VI (n = 20).Cases with significant cerebrovascular disease were excluded from the selection.Baseline characteristics are displayed in Table S1.Formalin-fixed paraffin-embedded (FFPE) tissue was used for immunohistochemistry, and frozen tissue, available from the same area for 52 of the 59 cases (Braak 0-II n = 19, Braak III-IV n = 18, Braak V-VI n = 15), was used to examine 30 inflammatory proteins, cytokines and chemokines using MesoScale Discovery multiplex assays.

[ 45 ,
46], which might explain the association between Iba1 and extraneuronal neuromelanin, with microglia being recruited in the attempt to remove the extraneuronal neuromelanin.The absence of increased phagocytic activity by microglia in response to the presence of extraneuronal neuromelanin suggests impaired microglial function.Of note, the relationship between extraneuronal neuromelanin with Aβ and ptau in the LC supports the link between the death of LC neurons and the development of AD pathology.The exploration of the inflammatory environment in the LC showed decreased expression of chemokines, MIP 1β, MDC, TARC and IL-8, in the late disease stage.As these chemokines typically Association with the Braak stage in the locus coeruleus and temporal cortex.TL, a significant difference was observed between Braak groups (H(2) = 38.834,p < 0.001), with a higher ptau load in Braak group III-IV compared with Braak group 0-II ( p = 0.006) and in Braak Note: Adjusted significant p < 0.05 (bold).pvaluescalculated using either ANOVA U-test or Kruskal-Wallis test followed by the appropriate post-hoc test.Data presented as mean ±standard deviation for normally distributed data or median [Interquartile range] for nonparametric data.Abbreviations: DβH, dopamine β-hydroxylase; LC, locus coeruleus; TH, tyrosine hydroxylase; TL, temporal lobe.T A B L E 2Note: Significant p < 0.001 in bold.Spearman factor (r s ), p value according to the data distribution.Abbreviations: DβH, dopamine β-hydroxylase; n/a, nonapplicable; TH, tyrosine hydroxylase.the group V-VI compared with Braak group III-IV ( p = 0.007) (Figure