Prospective longitudinal evaluation of cytokines in mild cognitive impairment due to AD and Lewy body disease

We conducted a prospective longitudinal study of plasma cytokines during the Mild Cognitive Impairment (MCI) stage of Lewy body disease and Alzheimer's disease, hypothesizing that cytokine levels would decrease over time and that this would be correlated with decline in cognition.

cultures and rodent models of AD. 5 Mouse models of AD have shown that non-steroidal anti-inflammatory drugs reduce the level of Aβ plaque load, 6 providing evidence that modifying peripheral inflammation in mice reduces AD pathology. Longitudinal studies have found plasma levels of inflammatory cytokines elevated up to 5 years before the clinical onset of dementia in AD 7 and we previously reported that, whilst IL-2, IL-4, IL-10 and IL-1β were increased at the mild cognitive impairment stage of AD (MCI-AD), they were not at the dementia stage. 8 Moreover, raised C-reactive protein (CRP) in midlife has been associated with a 3-fold increased risk of developing AD up to 25 years later. 9 There has been less research on inflammation in DLB but more work in the related alpha-synucleinopathy Parkinson's disease (PD).
We previously reported in a cross-sectional study significantly higher IL-1β, IL-4, IL-2 and IL-10 in DLB at the MCI stage (MCI-LB) but not at the dementia stage. 8 In addition, increased disease severity was associated with lower levels of IL-1β, IL-2 and IL-4 in both AD and DLB.
Reviews of the literature show PD patients have higher serum levels of inflammatory markers such as IL-1β, IL-6, IL-8, TNFα, IFNγ and RANTES 10,11 and studies from our centers have also reported that inflammatory markers predict disease progression in PD. 12 Numerous studies in PD show polymorphisms in several inflammation-related genes as being risk factors for the disease, including TREM2, IL1β, TNFα, LRRK2 and HLA-DR. 13,14 Two studies of PK11195 in DLB have reported increased microglial activation in cortical and subcortical areas 15,16 Alpha-synuclein has been shown to induce microglial activation in cell culture studies, 17 and in PD mouse models. 18 In summary, inflammation in AD and DLB may pre-date and possibly even play a role in precipitating the onset of these dementias, being present at prodromal and earlier stages but then diminishing with disease progression. However, no previous studies have prospectively examined cytokines in plasma in a longitudinal study in an MCI population or in LB disease. We therefore conducted a prospective longitudinal study in which subjects with MCI-LB and MCI-AD were re-assessed annually including repeating blood samples for analysis of peripheral inflammation. We hypothesized that in both MCI groups there would be a significant decrease in inflammatory markers over time from the elevated level reported previously 8 and that this decrease would correlate with severity of cognitive impairment, as measured by the Addenbrooke's Cognitive Examination-Revised (ACE-R).

| METHODS
Recruitment and baseline assessment has been detailed previously. 8,19 Briefly, patients aged 60 or older were recruited following assessment at local memory services and had a clinical diagnosis of MCI.
Additionally, patients reported one or more clinical symptoms sensitive, but non-specific, to Lewy Body (LB) disease (eg, mood changes, sleep disturbance, or autonomic symptoms), or service records indicated the presence of any core DLB features. Following consent, participants underwent a research level assessment involving a semi-structured interview, clinical assessment and neurological examination by the equivalent of a board-certified medical doctor (PCD). Cumulative Illness Rating Scale for Geriatrics (CIRS-G) were completed based on the clinical history. A detailed neuropsychological evaluation was also carried out as reported previously 20 which included the ACE-R. All subjects were offered dopaminergic imaging with FP-CIT SPECT at baseline, including those with MCI-AD. Images were randomized, coded, and then visually rated as normal/abnormal by an experienced consensus panel blind to clinical information and diagnosis as reported earlier 21 and incorporated into diagnoses.

| Clinical diagnosis
A three-person consensus clinical panel of experienced Board Certified old age psychiatrists (AJT, PCD, JPT) independently reviewed clinical notes taken from the baseline assessment and confirmed diagnoses of MCI according to NIA-AA criteria. 22 This was based on evidence of minimal functional impairment (independent living was maintained) and a CDR of 0 or 0.5, and the presence of subjective and objective cognitive decline. Anyone with dementia was excluded. To determine the etiology, the presence or absence of core LB symptoms were also rated by the panel, in accordance with the fourth consensus criteria for DLB. 23 Based on health service clinical notes and imaging results, those with possible significant vascular or frontotemporal

Key points
• Inflammatory cytokines were previously observed to be elevated in MCI relative to both healthy older adults and those with dementia • This was the case in both Lewy body and Alzheimer's diseases • In this follow-up study, the increased inflammation in MCI subsided as this progressed in severity • Inflammatory processes may represent a potential early therapeutic target in common neurodegenerative diseases etiologies, or parkinsonism pre-dating cognitive impairment by more than 1 year, were also excluded. Where possible, an informant was sought (spouse, friend, or family member) to provide additional information. FP-CIT findings were later incorporated into diagnoses but the panel decisions on MCI and symptoms were made blind to these findings.
Participants received a diagnosis of MCI with probable Alzheimer's disease (MCI-AD) when they had no core LB symptoms, a normal FP-CIT scan and evidence of decline which was characteristic of AD, with no evidence for another etiology, that is, they met the additional NIA-AA criterion of "etiology of MCI consistent with AD pathophysiologic process." 22 MCI with Lewy bodies (MCI-LB) was diagnosed when at least one core feature was present or the patient had an abnormal FP-CIT scan, in accordance with current consensus research criteria for the diagnosis of MCI-LB. 24 One participant did not consent to dopaminergic transporter imaging but had sufficient clinical LB symptomology for a MCI-LB diagnosis (two core features) without confirmatory biomarkers.

| Prospective evaluation
Participants were re-assessed every 12 months by a research nurse and/or doctor in a prospective longitudinal design, with the above assessment instruments repeated each year. Clinical assessments were undertaken blind to results of cytokine assessments. Core LB symptom presence and severity of cognitive impairment (MCI or dementia) were re-appraised at annual follow-ups by the clinical diagnostic panel and blood samples were repeated each year.

| Cytokine analyses
Baseline and repeat venous blood samples were taken from all patients using EDTA tubes, which were then centrifuged and the plasma removed. Samples were stored at −80 C until assays were performed. Cytokine assays were performed using the Meso Scale Discovery V-Plex Plus Proinflammatory Panel 1, which included IFNγ, IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13 and TNFα. Assays were performed at the Newcastle University BioScreening Core Facility according to the manufacturer's protocol, and samples were processed in triplicates. Samples which were under the limit of detection for a particular cytokine (concentration in pg/mL < 0.05 for IFN gamma, <0.03 for IL-13, <0.02 for IL-12p70 and <0.01 for all other cytokines) had cytokine levels which were low enough to be undistinguishable from background noise, and therefore these samples were treated as having "zero" levels of that cytokine.

| Statistical analyses
Longitudinal analyses were undertaken with R software packages lme4 and lmerTest. Repeated measures correlations were analyzed with the rmcorr package to account for non-independence of observations within individuals, and providing greater statistical power than alternative methods (eg, averaging across repeated observations, or assessing correlations at each time-point). Linear mixed-effects models assessed the effects of time (continuous) on cytokine levels, controlling for the effects of age, gender, concurrent illness (CIRS-G score), use of anti-inflammatory medications (non-steroidal antiinflammatory drugs or steroids), and MCI disease group. The effects of controlling for baseline cognitive function (ACE-R total score) and indices of local deprivation were also assessed, but their inclusion did not improve model fit in any cases. Improvements in model fit were assessed by likelihood ratio tests, with an alpha level of P < .05. Random intercept and time-slopes were included at the subject level, allowing for correlation between these where appropriate; fixed and random non-linear terms for time were also assessed, but did not improve fit over the linear-only term. Cytokine levels were log transformed as at baseline, and continuous covariates were mean-cen-

| RESULTS
Seventy-seven MCI patients (21 MCI-AD, 56 MCI-LB) completed the baseline assessment and provided blood samples, as reported previously. 19 Of these, 56 (21 MCI-AD, 35 MCI-LB) completed at least one follow-up assessment and provided blood samples. Those who were not available for repeated samples had a significantly lower baseline cognitive function (ACE-R) score (t[33] = −2.854, P = .007), but did not significantly differ in their baseline cytokine levels. Increased retention of AD likely reflected the more severe course of DLB. 25,26 Differential diagnoses were stable over this time (ie, there were no cases which changed from MCI-LB to MCI-AD), but of those included who provided two or more samples, 10 MCI-AD (48%) and 13 MCI-LB (37%) converted to dementia over the course of assessment. The baseline characteristics of these 56 patients are provided in Table 1. MCI-LB were slightly younger and as expected there were sex differences between the disease groups (more females in AD group) and minor differences in neuropsychiatric symptoms (higher in the MCI-LB patients), with scores being modest and consistent with early stage disease. The two groups did not differ in their levels of motor impairment as assessed with the UPDRS-III, though in the full cohort as previously described including those lost to follow-up, MCI-LB had significantly higher scores in this measure. 21 As impairments were "rated as seen," higher UPDRS-III scores in MCI-AD most likely reflect non-parkinsonian age-related motor impairments; this is supported by a moderate association between increasing age and UPDRS-III scores  Table 2. As with the larger baseline group, 8 there were no significant differences in baseline cytokine levels between the diagnostic groups.

| Prospective evaluation of cytokines in MCI
We repeated at least one annual review for each patient with the lon-  Figure 1 shows the change in the remaining eight cytokines over this period and Table 3 shows the findings from the linear mixed effects models. Six cytokines (IFNγ, IL-1β, IL-2, IL-4, IL-6 and IL-10) showed highly significant decreases over time. Four of these (IL-1β, IL-2, IL-4 and IL-10) had been significantly higher than controls at baseline. 8 IFNγ and IL-6 were not significantly different from controls at baseline but there was high variability in these cytokines,

| Correlations of cytokines with decrease in cognition
Examining our second hypothesis, in the overall group we found sig- and 0.57) between worsening of cognitive impairment (decrease in ACE-R) and decrease in cytokine levels for IFNγ, IL-1β, IL-2, IL-4 and IL-10 (see Table 4). IL-8 had a weak non-significant correlation in the opposite direction (increasing cytokine level associated with lower cognition) and there was no correlation for IL-6 and TNFα.
Regarding other potential relationships with cytokine changes, there were no differences between the disease groups in any of the changes in cytokine levels nor were there significant effects due to age or sex or general illness burden (as measured by the CIRS-G) or use of anti-inflammatory drugs. There was also no correlation between change in cytokine levels and severity of parkinsonism (measured using UPDRS).
The data that support the findings of this study are available by request through Dementias Platform UK at https://portal. dementiasplatform.uk/DAMatrix. 27

| DISCUSSION
Our primary hypotheses based on previous literature, that there would be a decline in plasma cytokine levels in people with MCI due to AD and DLB over time and that such a decrease would correlate with the accompanying decline in cognition, were both supported by our analyses. These findings suggest there may be a therapeutic window for use of anti-inflammatory drugs in these two major neurodegenerative diseases including this MCI stage of these illnesses.
Both AD and DLB have cholinergic deficits and respond well to cholinesterase inhibitors and both show modest benefits with memantine. 28 However, there are no disease modifying treatments available for either disease, with a large number of trials having failed to demonstrate benefits. 29,30 Identification of new therapies for these devastating diseases is a major priority and targeting neuroinflammation is an attractive alternative therapeutic approach. Previous evidence from genetics in AD, 2 and Lewy body disease 14 and peripheral measures of inflammation in milder AD 1 and LBD 10 supports a role for inflammatory processes in both diseases. We previously reported that whilst by the dementia stage cytokine levels had returned to the level of healthy controls, they were elevated at the MCI stage in both diseases. 8,19 This is consistent with other evidence that inflammatory markers appear to be increased years before the development of AD dementia, 9 though no studies seem to have assessed MCI-AD. To our knowledge there have been no previous prospective studies of inflammation in DLB, and the only previous prospective studies in PD have been from our centers in people without cognitive impairment, which have supported a role for peripheral inflammation in disease progression in PD. 12 In addition, a CSF study in PD which did not include anyone with dementia but found increasing levels of the inflammatory marker YKL-40, a marker of macrophages, which correlated with cognitive decline. 31 These patients are difficult to compare with those in our study and likely represent a different phase of LB disease since they had had many years of motor LB disease at study entry.
The few PET studies of neuroinflammation have used microglial markers and found evidence for inflammation in the brain in AD 3,4 and DLB. 15,16 In the latter study cerebral inflammation was higher in mild compared with moderate/severe DLB and the microglial activation was also found to be positively associated with cognitive performance. 16  There is now increasing focus on targeting inflammation as a treatment for AD and our findings suggest such treatments could be extended to DLB. Several epidemiological studies have demonstrated the protective effects of non-steroidal anti-inflammatory drugs in reducing incidence of AD 34,35 and a more recent study showed that users of disease modifying anti-rheumatic drugs for rheumatoid arthritis were at reduced risk of dementia (hazard ratio of 0.60). 36 However, several randomized controlled trials have failed to show any significant benefits for anti-inflammatory drugs. [37][38][39] Intriguingly, post-hoc analysis of the ADAPT trial revealed that treatment of asymptomatic individuals did reduce AD incidence, but only with at least 2 to 3 years of treatment. 40 Such results fit with our study findings indicating that a therapeutic window for such treatments may exist prior to later disease at the dementia stage. This likely begins before the MCI stage but MCI currently represents the most realistic window of therapeutic opportunity during which patients might be identified for such treatment. Such potential benefits from anti-inflammatory medication might also be targeted at earlier pre-symptomatic stages of disease but for practical purposes, since MCI is now widely diagnosed in memory services, such patients are a more realistic group to choose for clinical trials.
Our study benefits from thorough clinical evaluation and detailed diagnostic assessments of all the patients and the prospective annual re-evaluation and diagnostic re-assessment of all patients with followup up to 3 years from baseline. Repeat samples were not available from either the cognitively healthy or dementia comparison groups.
Controls would not be expected to change and since the dementia cytokine levels were already low then they had limited scope for decline but comparison with these in future assessments would clarify how quickly MCI cytokine levels subside to the levels observed in dementia. A clear limitation is the modest numbers in each group, though the highly significant nature of our findings and consistencies between cytokines does not suggest a lack of power for our primary outcomes. We were also not able to adequately assess two of our cytokines (IL-13 and IL-12p70) due to difficulties in accurate detection of these molecules and as expected people with more cognitive impairment at baseline and with LB disease, which is a more severe illness, 25,26 were less likely to complete follow-up assessments; 38% of MCI-LB from baseline were unavailable to provide repeat samples.
These did not differ in their baseline inflammation, but their subsequent inflammatory trajectories are unknown, and could differ from those who were observed, a limitation of this sample. Furthermore, whilst our findings measuring cytokines do not indicate differences in inflammation between AD and DLB, there may nonetheless be real differences in the immune-inflammatory response between these two diseases which may be identified by focusing on other aspects of this complex phenomenon, for example, using flow cytometry to measure the profile of T-cell lymphocytes.
While there is no reason to believe that these biomarkers pro- With sufficient observations and timespan, different trajectory directions and shapes may be identifiable with a data-driven longitudinal clustering method, for example, latent class mixed modeling.
In summary, we found strong evidence for a decrease in peripheral inflammation in both AD and LBD from elevated levels of cytokines at the MCI stage towards healthy control levels typical of the dementia stages of these diseases, findings which have important implications for the design, conduct and outcome measures of future anti-dementia trials involving putative anti-inflammatory agents.