key: cord-0282929-3e9iv1j9
authors: Elkind, M. S.; Moon, Y.; Rundek, T.; Wright, C.; CHeung, K.; Sacco, R. L.; Hornig, M.
title: Immune Markers Are Associated with Cognitive Performance in a Multiethnic Cohort: the Northern Manhattan Study
date: 2021-01-20
journal: nan
DOI: 10.1101/2021.01.18.21250068
sha: 6a538335924c5f683214005ab65d8f81494cf028
doc_id: 282929
cord_uid: 3e9iv1j9

OBJECTIVE: To determine whether immune protein panels add significant information to correlates of cognition. BACKGROUND: Immune mechanisms in vascular cognitive impairment and dementia are incompletely characterized. DESIGN/METHODS: A subsample of the prospective Northern Manhattan Study underwent detailed neuropsychological testing. Cognitive scores were converted into Z-scores and categorized into four domains (memory, language, processing speed, and executive function) based on factor analysis. Blood samples were analyzed using a 60-plex immunoassay. We used least absolute shrinkage and selection operator (LASSO) procedures to select markers and their interactions independently associated with cognitive scores. Linear regression models assessed cross-sectional associations of known correlates of cognition with cognitive scores, and assessed model fit before and after addition of LASSO-selected immune markers. RESULTS: Among 1179 participants (mean age 70+8.9 years, 60% women, 68% Hispanic), inclusion of LASSO-selected immune markers improved model fit above age, education, and other risk factors (p for likelihood ratio test<0.005 for all domains). C-C Motif Chemokine Ligand 11 (CCL 11, eotaxin), C-X-C Motif Chemokine Ligand 9 (CXCL9), hepatocyte growth factor (HGF), and serpin E1 (plasminogen activator inhibitor-1) were associated with each of the domains and with overall cognitive function. Immune marker effects were comparable to conventional risk factors: for executive function, each standard deviation (SD) increase in CCL11 was associated with an effect equivalent to aging three years; for memory, HGF had twice the effect of aging. CONCLUSIONS: Immune markers associate with cognitive function in a multi-ethnic cohort. Further work is needed to validate these findings and determine optimal treatment targets.

person interview and assessment were conducted by trained bilingual research assistants. A 129 subcohort of 1290 participants included 1091 subjects remaining clinically stroke-free from the 130 original cohort as well as a sample of unrelated household members (n=199). MRI and detailed 131 neuropsychological assessments were conducted, and blood samples were collected; methods of 132 recruitment and characteristics of this sample have been described previously. 14 The study was 133 approved by the Columbia University and University of Miami Institutional Review Boards, and 134 all participants provided informed consent. 135 136

Demographics and risk factors were collected through in-person interviews, as described 138

previously. 14 Race-ethnicity was determined by self-identification. Insurance status was defined 139 no insurance or Medicaid versus private insurance / Medicare 23 . Educational achievement was 140 self-reported as number of years in school and degree achieved. Physical activity was evaluated 141 by in-person questionnaire 35 is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint Neuropsychological assessments were administered by trained bilingual research staff in 151

English or Spanish, as previously described. 15 We developed cognitive domain-specific scores to 152 enable interpretability of cognitive functions (Table e-1) . Because of lack of population norms 153 for older cohorts, we used baseline neuropsychological scores as our cohort norms. 16 To achieve 154 comparability, neuropsychological tests were standardized against cohort-specific normative 155 values. For timed tests (Color Trail 1, Color Trail 2 and Grooved Pegboard), because some 156 participants did not complete tasks within the maximum time allowed, we calculated their new 157 "imputed" time by adding penalty time to the maximum time allowed. The penalty time was 158 obtained by multiplying the number of incomplete items by "expected time per correct items 159 completed" estimated using Kaplan-Meier methods. The standardized scores for individual tests 160

were then categorized into four key cognitive domains (memory, language, processing speed and 161 executive function) based on an exploratory factor analysis and prior findings. 16 Thereafter, 162 cognitive domain-specific scores were calculated by taking the mean of the individual 163 standardized test scores selected for each domain. Cognitive outcomes were domain-specific z 164 scores (memory, language, processing speed and executive function) at baseline and a composite 165 cognitive function score obtained by averaging those four domain scores. 166

We completed assays of 60 immune molecules in 1179 participants from stored fasting 168 plasma drawn at the time of a detailed cognitive assessment. 17, 18 We used a customized, highly 169 multiplexed, magnetic bead-based immunoassay. 19, 20 Our assay included molecules integral to 170 inflammatory responses, reflecting relative activation of Th1 (pro-inflammatory), Th2 (counter-171 regulatory, autoimmunity-promoting), Th17 (innate mucosal anti-pathogen defenses, including 172 in gut), T regulatory, B (antibody-producing) and NK cells, and, more generally, innate and 173 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 20, 2021. Data were processed using a quality control (QC) algorithm that calibrates performance of an 180 expanded set of serial standard curves and an in-house plasma control run on every assay plate, 181 monitoring intra-and inter-plate covariance (CV) and bead counts. Only plates with mean intra-182 assay %CV <15% were accepted. Samples failing to meet QC criteria were designated for re-run Association between cognitive outcomes and immune markers was assessed using a two-195 step hybrid Least Absolute Shrinkage and Selection Operator (LASSO) approach 22 : first, we 196 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

There were 1179 participants with immune marker data available (Table 1) . Mean age 214 was 70±8.9 years and 60% were women. There was a high proportion of Hispanics (68%); mean 215 education level was 10±5 years. There was a high burden of diabetes (28%) and other 216 cardiovascular risk factors. 217 218

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint Inclusion of LASSO-selected immune markers collectively improved model fit over age 220 alone (p-value for LRT< 0.0001 for global score and for all domains) and over age, 221 demographics including education, and traditional risk factors for each cognitive outcome (p-222 value <0.0001 for global score, memory, language, and executive function: p = 0.0081 for 223 processing speed; Table 2 ). For global cognitive function, the variance in cognition explained by 224 the immune markers was similar to that explained by age (model 0 and model 1-1), and the 225 addition of immune markers to a model that included age alone led to almost a doubling of the 226 R 2 (model 1-1 and model 1-2), implying that the effect of the immune markers was independent 227 and comparable in size to that of the effect of age. The effect of age alone was greatest for the 228 domain of processing speed (R 2 =0.17) and least for language (R 2 =0.02). The impact of adding 229 immune markers to age was greatest for executive function, for which R 2 increased from 0.04 for 230 a model including only age to 0.17 when the immune markers were added. 231

Approximately 48% of the variance in global cognition was explained by a model 232 containing age, other demographics, and risk factors; the model fit improved by approximately 233 4.0% (from 48% to 50%) after inclusion of the immune markers. 234 235

A subset of immune markers and their two-way interactions that predicted each cognitive 237 function were selected using LASSO (Table e-2) . C-C Motif Chemokine Ligand 11 (CCL11, or 238 eotaxin), C-X-C Motif Chemokine Ligand 9 (CXCL9), hepatocyte growth factor (HGF) and 239 serpin E1 (or plasminogen activator inhibitor-1) were consistently selected as being associated 240 with overall cognitive function (Table 3 ) and with each of the neuropsychological domains (table  241 e-3-6). In addition, sFasL, SCF, leptin, CCL3, CSF3 and CCL2 contributed to multiple cognitive 242 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint domains. Interactions between CCL5 and serpin E1, between CCL3 and TNFβ, between CCL11 243 and IL1RA, between HGF and leptin, between TRAIL and IL12p40, between CCL11 and CCL5, 244 and between CXCL10 and CXCL9 were also found in 2 or more domains. 245 246

The magnitude of immune marker effects was comparable to that of conventional risk 248 We found several immune molecules are cross-sectionally associated with cognitive 255 function in a predominantly Hispanic, but ethnically diverse, population. Many immune markers 256 were correlated, prompting us to use the LASSO technique to perform data reduction, thereby 257 extracting those molecules that were independently and most informatively associated with 258 cognition. As a group, those markers that were selected using LASSO had an explanatory impact 259 comparable and independent to that of age. After accounting for sociodemographic and vascular 260 risk factors, as well, the immune panel significantly improved model fit for the association with 261 cognition by approximately 4%. 262

These results suggest that serum immune markers have an independent association with 263 cognition beyond that due to traditional risk factors for cognition and vascular disease. While the 264 collective magnitude of these markers on cognition was modest, the magnitude of association for 265 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint several markers was comparable to that of other well-accepted risk markers of cognitive 266 impairment, including age. Cognitive decline is a complex, multifactorial phenomenon, and it is 267 unlikely that any single risk factor, immune pathway or marker accounts for a substantial 268

proportion of the variance in cognition. The total variance explained by models not including the 269 immune panel did not exceed 50%. Our results suggest that several immune pathways may be 270 involved and that approaches that affect multiple pathways simultaneously will be most likely to 271

have an impact, although further research is needed. effect in humans appears to be reversed. 26 

CXCL9 is a member of the subfamily of CXC chemokines that lack a glutamic acid-281 leucine-arginine motif. They all share the CXCR3 receptor, and participate in Th1-type immune 282 responses, involving tissue infiltration by T cells as part of the innate immune response. CXCL9 283 is constitutively expressed on human brain-derived microvascular endothelial cells and 284 astrocytes. 27 CXCL9 is upregulated by IFN-γ, induces activation of extracellular signal-regulated 285 kinases in cortical neurons in mice, and is involved in neuronal-glial interactions. 28 The CXC 286 chemokines increase with inflammation and infection of the CNS, 29 and are also up-regulated in 287 human brains with Alzheimer pathology. 30 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint HGF is a neurotrophic factor with anti-apoptotic and angiogenesis effects that facilitates 289 dendritic arborization and has been associated with small artery disease among patients with 290 cognitive impairment and AD. 31 Consistent with our findings, these associations were 291 independent of other vascular risk factors, suggesting that the mechanisms by which HGF 292 contributes to cognitive impairment may be independent of traditional vascular risk factors. It is 293 increasingly evident that vascular and neurodegenerative contributions to cognitive aging and 294 dementia overlap, and that vascular mechanisms contribute to neurodegeneration. Immune 295 mechanisms that lead to small vessel disease may thus enhance the degenerative pathways that 296 lead to AD, either by impairing blood flow, enhancing inflammation, reducing regenerative 297 capacity, or inhibiting amyloid clearance. 298

Serpin E1 is the major inhibitor of the serine proteases tissue plasminogen activator and 299 urokinase, and thus acts as an inhibitor of fibrinolysis. Elevated levels potentially contribute to 300 increased atherosclerosis, and risk of coronary artery events and stroke. 32 In addition, serpin E1 301 inhibits activity of matrix metalloproteinases, which play a role in blood-brain barrier breakdown 302 after ischemia. 33 Serpin E1 also plays a role in regulation of brain-derived neurotrophic factor 303 (BDNF), which must be converted from a pro-protein by plasmin. BDNF plays an important role 304 in neuronal activity and memory. 34 A small study showed that serpin E1 is elevated in peripheral 305 blood with mild cognitive impairment; 35 however, our findings of higher levels of serpin E1 in 306 association with cognitive performance, and without interaction with BDNF, suggest that serpin 307 E1 production may exert a small protective effect. 308

Associations of immune markers with cognition were present for each of the individual 309 domains of cognitive function investigated, although several markers, as well as a global 310 cognitive measure, were present across all areas. Other markers selected by our data reduction 311 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint technique as contributing to two or more individual cognitive domains included sFasL (memory, 312 language function, processing speed, and global cognition), SCF (memory, processing speed, and 313 global cognition), CSF2 (memory and processing speed), leptin (language, executive function, 314 and global cognition), and the chemokines CCL2 (processing speed, executive function) and 315 CCL3 (language, executive function, and global cognition). 316

Evidence that levels of the chemokines CCL2, CCL3, CCL11 and CXCL9 are associated 317 with cognitive decline in our cohort support the hypothesis that the innate immune system 318 contributes to cognitive impairment and dementia. 36, 37 Recent studies provide evidence that 319 increased amyloid-beta peptide activates the innate immune system via pattern recognition 320 receptors on microglia and astrocytes, contributing to hyperphosphorylation of tau. Sterile 321 inflammation can also be provoked by recognition of intracellular contents released from 322 necrotic cells through damage-associated molecular patterns. Elevations of these chemokines 323 could reflect chronic activation of innate immunity in brains of patients with cognitive 324 impairment. There is also evidence that infections, such as herpes simplex virus-1, provoke 325 innate immunity, cause tau phosphorylation, and contribute to neurodegeneration. 38 Previous 326 studies in our cohort demonstrate that serologies against herpesviruses are associated with lower 327 cognitive test scores. 39,40 328 Our results could have clinical implications since inflammation is potentially modifiable. 329

IL1β inhibition diminishes residual inflammatory risk for cardiovascular events, 41, 42 and some 330 antiviral agents are being tested for efficacy against AD, with variable results. 43,44,45 331 Acetylcholinesterase inhibitors also decreased IFNγ, TNFα, IL1β and IL6 secretion by immune 332 cells and enhanced innate resistance to viral infection in vitro. 46 Furthermore, the timing of 333 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint intervention also appears to be critical, underscoring the significance of long-term prospective 334 investigations such as NOMAS. 335

Our study has limitations. First, measures of immune molecules and cognition were 336 obtained at a single time point, limiting our ability to make causal determinations. We continue 337 to follow our cohort over time, however, to provide longitudinal data. Second, immune markers 338

were not assessed in the cerebrospinal fluid or brain tissue, the site where injury occurs. 339

Nonetheless, blood biomarkers are increasingly recognized as indicators of neurodegenerative 340 activity and predictors of AD. Third, though our panel included 60 different immune-related 341 markers, many others were not included. Fourth, we did not include MRI-derived biomarkers in 342 this study, nor did we have positron emission tomography to assess amyloid, tau, or microglial 343 activations. Our outcomes, however, were cognitive test scores, which may reflect multiple 344 pathologies and may ultimately have greater clinical relevance for patients. Our assay does not 345 provide a clinically useful panel of markers for the prediction of dementia currently. Our 346 findings require confirmation in other cohorts, and in longitudinal analyses, before they can be 347 used for clinical purposes. 348

Our study also has strengths. Our cohort represents a racially and ethnically diverse, 349 elderly cohort at highest risk for cognitive decline and dementia. Many prior studies have not 350 included large numbers of Black and Hispanic/Latino people. Our population was also well-351 characterized for demographic and medical risk factors. Our multiplex immune assay included a 352 large number of molecules that may contribute to cognitive decline and dementia. 353

In conclusion, peripheral blood levels of several immune-related molecules, representing 354 pathways related to innate immunity, neural repair, and thrombosis, may play a role in cognitive 355 decline and dementia. As a group these factors modestly contribute to cognitive impairment, 356 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint independently of vascular risk factors associated with cognitive decline. These results speak to 357 the ability of a multiplex immune assay to add information regarding pathways that may be 358 involved in neurodegeneration in a diverse population. These findings also indicate the need for 359 further research on immune mechanisms in cognitive aging. 360 361 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21250068 doi: medRxiv preprint

A 20-Year Journey from Axonal Injury to Neurodegenerative Diseases and the Prospect of Immunotherapy for Combating Alzheimer's Disease

Inflammatory risk factors, biomarkers and associated therapy in ischaemic stroke

Innate Immunity: A Common Denominator between Neurodegenerative and Neuropsychiatric Diseases

Do infections have a role in the pathogenesis of Alzheimer disease?

Mook-Jung I. Peripheral inflammatory biomarkers in Alzheimer's disease: a brief review

An update on the interactions between Alzheimer's disease, autophagy and inflammation

A Framework of Minority Stress: From Physiological Manifestations to Cognitive Outcomes

Role of stress, depression, and aging in cognitive decline and Alzheimer's

Examining Cognitive Decline Across Black and White Participants in the Harvard Aging Brain Study

Extracellular matrix inflammation in vascular cognitive impairment and dementia

Emerging Role of Immunity in Cerebral Small Vessel Disease

Trained Immunity Characteristics Are Associated With Progressive Cerebral Small Vessel Disease

Circulating biologic markers of endothelial dysfunction in cerebral small vessel disease: A review

Inflammatory Biomarkers of Vascular Risk as Correlates of Leukoariosis

Infectious Burden and Cognitive Decline in the Northern Manhattan Study

The Statistical Value of Raw Fluorescence Signal in Luminex xMAP

Bead-based multiplex immuno-assays for cytokines, chemokines, growth factors and other analytes: median fluorescence intensities versus their derived absolute concentration values for statistical analysis

Distinct plasma immune signatures in ME/CFS are present early in the course of illness

Cytokine network analysis of cerebrospinal fluid in myalgic encephalomyelitis/chronic fatigue syndrome

Multiplexed particle-based flow cytometric assays

Least angle regression

Revisiting the Role of Eotaxin-1/CCL11 in Psychiatric Disorders. Front Psychiatry

The ageing systemic milieu negatively regulates neurogenesis and cognitive function

CCL11 enhances excitotoxic neuronal death by producing reactive oxygen species in microglia

CCL11 (Eotaxin-1) Levels Predict Long-Term Functional Outcomes in Patients Following Ischemic Stroke

Review: The chemokine receptor CXCR3 and its ligands CXCL9, CXCL10, and CXCL11 in neuroimmunity -a tale of conflict and conundrum

Expression of the chemokine receptor CXCR3 on neurons and the elevated expression of its ligand IP-10 in reactive astrocytes: In vitro ERK1/2 activation and role in Alzheimer's disease

Regulation of CNS precursor function by neuronal chemokines

Scheltens P; Dutch-Italian Alzheimer Research Group. Chemokines in serum and cerebrospinal fluid of Alzheimer's disease patients

Serum Hepatocyte Growth Factor Is Associated with Small Vessel Disease in Alzheimer's Dementia

The roles of PAI-1 gene polymorphisms in atherosclerotic diseases: A systematic review and meta-analysis involving 149,908 subjects

Tissue plasminogen activator induced delayed edema in experimental porcine intracranial hemorrhage: reduction with plasminogen activator inhibitor-1 administration

Amyloid beta soluble forms and plasminogen activation system in Alzheimer's disease: Consequences on extracellular maturation of brainderived neurotrophic factor and therapeutic implications

Identification of inflammatory and vascular markers associated with mild cognitive impairment

Neuroinflammation in Alzheimer's disease

Inflammation in Alzheimer's disease: amyloid-beta oligomers trigger innate immunity defence via pattern recognition receptors

Alzheimer's disease-specific tau phosphorylation is induced by herpes simplex virus type 1

APOE ϵ4 modifies the relationship between infectious burden and poor cognition

Antiinflammatory Therapy with Canakinumab for Atherosclerotic Disease

Anticytokine Agents: Targeting Interleukin Signaling Pathways for the Treatment of Atherothrombosis

Recommendations for Anti-inflammatory Treatments in Alzheimer's Disease: A Comprehensive Review of the Literature

Anti-herpetic Medications and Reduced Risk of Dementia in Patients with Herpes Simplex Virus Infections-a Nationwide, Population-Based Cohort Study in Taiwan

Porphyromonas gingivalis in Alzheimer's disease brains: Evidence for disease causation and treatment with small-molecule inhibitors

Acetyl Cholinesterase Inhibitors and Cell-Derived Peripheral Inflammatory Cytokines in Early Stages of Alzheimer's Disease