Proteome-Wide Discovery of Cortical Proteins That May Provide Motor Resilience to Offset the Negative Effects of Pathologies in Older Adults

Abstract

Background

Motor resilience proteins have not been identified. This proteome-wide discovery study sought to identify proteins that may provide motor resilience.

Methods

We studied the brains of older decedents with annual motor testing, postmortem brain pathologies, and proteome-wide data. Parkinsonism was assessed using 26 items of a modified United Parkinson Disease Rating Scale. We used linear mixed-effect models to isolate motor resilience, defined as the person-specific estimate of progressive parkinsonism after controlling for age, sex, and 10 brain pathologies. A total of 8 356 high-abundance proteins were quantified from dorsal lateral prefrontal cortex using tandem mass tag and liquid chromatography–mass spectrometry.

Results

There were 391 older adults (70% female), mean age 80 years at baseline and 89 years at death. Five proteins were associated with motor resilience: A higher level of AP1B1 (Estimate −0.504, SE 0.121, p = 3.12 × 10⁻⁵) and GNG3 (Estimate −0.276, SE 0.068, p = 4.82 × 10⁻⁵) was associated with slower progressive parkinsonism. By contrast, a higher level of TTC38 (Estimate 0.140, SE 0.029, p = 1.87 × 10⁻⁶), CARKD (Estimate 0.413, SE 0.100, p = 3.50 × 10⁻⁵), and ABHD14B (Estimate 0.175, SE 0.044, p = 6.48 × 10⁻⁵) was associated with faster progressive parkinsonism. Together, these 5 proteins accounted for almost 25% of the variance of progressive parkinsonism above the 17% accounted for by 10 indices of brain pathologies.

Discussion

Cortical proteins may provide more or less motor resilience in older adults. These proteins are high-value therapeutic targets for drug discovery that may lead to interventions that maintain motor function despite the accumulation of as yet untreatable brain pathologies.

Age-related progressive loss of motor function is common, associated with diverse adverse health outcomes, and limits independent living in older adults (1). Accumulating evidence suggests that mixed-brain pathologies contribute to the heterogeneity of progressive motor decline (2). Though nearly all aging brains show the accumulation of some degree of diverse brain pathologies, the extent that these pathologies are associated with movement varies (3–5). For example, the same burden of brain pathology may be related to rapid decline in one adult and minimal motor decline in another. An adult who maintains motor function or has a slower rate of motor decline in the presence of brain pathologies can be said to manifest motor resilience (6). Modifiable risk factors, lifestyles, or proteins that provide resilience, that is, offset the negative effects of brain pathologies, may provide high-value therapeutic targets especially because most brain pathologies are currently untreatable.

Movement is a multidimensional volitional behavior, and there is substantial heterogeneity in its decline and in the extent to which its different facets are affected. Consequently, not only is the rate or severity of progressive motor loss heterogeneous, but also the specific motor abilities that decline vary widely. For example, some adults manifest loss of strength, or impaired gait and balance, whereas others display mild parkinsonian signs. These latter signs include parkinsonian gait, bradykinesia, tremor, and rigidity. Our prior work has shown that the burden of mixed-brain pathologies that accumulate in many older brains is related to the rate of progressive parkinsonism in both adults with and without a clinical diagnosis of Parkinson’s disease (PD) (5). Yet, many older adults with mixed-brain pathologies in their brain exhibit minimal signs of progressive parkinsonism, whereas others show severe progressive parkinsonism. Motor resilience proteins have not been identified in part because of the inherent challenges, of assembling the necessary resources including longitudinal motor function, adequate brain pathology measures, and proteome-wide data in large numbers of older adults.

In prior work, we found that it is possible to identify cognitive resilience proteins by isolating “cognitive resilience,” that is, cognitive decline unexplained by brain pathologies and examining its associations with brain proteome (7,8). Extending this same approach to motor phenotypes by isolating “motor resilience” might also facilitate the identification of cortical proteins that may provide resilience to motor decline in older adults (9). To discover cortical proteins that may provide resilience to progressive parkinsonism, we used data from well-characterized decedents who underwent annual assessment for parkinsonism and autopsy at the time of death which collected brain indices of 10 different brain pathologies; we also assessed the proteome from postmortem dorsal lateral prefrontal cortex (DLPFC) (10).

Method

Study Participants

Participants were older persons enrolled in one of the 2 ongoing cohort studies of chronic conditions of aging, the ROS (Religious Orders Study) and the MAP (Rush Memory and Aging Project) (10). Participants entered the studies without known dementia and agreed to annual assessments as well as brain donation after death. Both cohort studies (MAP and ROS) were approved by an Institutional Review Board of Rush University Medical Center. Written informed consent was obtained from each study participant as was an Anatomical Gift Act for organ donation.

ROS and the MAP are conducted by the same team of investigators and share a large common core of harmonized clinical and postmortem data collection which allows for joint analyses. Through October 23, 2019, a total of 1 654 ROS and MAP participants had died and undergone brain autopsies. A subset of 400 decedents were part of a convenience sample, the first installment of postmortem prefrontal cortex from which tandem mass tag spectrometry (TMT-MS) proteomics were obtained. TMT-MS proteomics data collection is ongoing. As published previously, the subset of individuals with TMT-MS proteomics data and the other ROS and MAP decedents who had undergone autopsy showed similar demographics characteristics; in general, individuals with TMT-MS data had fewer neuropathologic conditions (Supplementary Table S1 in ref. (7)). A prior publication analyzed data from these same decedents to examine the associations of DLPFC proteins with cognitive resilience (7). The current analyses focus on progressive parkinsonism to discover motor resilience proteins from DLPFC.

Assessment of Parkinsonism and PD

Nurse clinicians assessed 4 parkinsonian signs including parkinsonian gait, rigidity, bradykinesia, and tremor annually using 26 items from a modified United Parkinson’s Disease Rating Scale (UPDRS). These measures have high inter-rater reliability and short-term stability both among nurses and between the nurses compared with a movement disorder specialist (11,12).

Global parkinsonian score

Using a continuous measure of parkinsonism rather than a categorical measure is useful for examining the associations of postmortem indices and proteins with the rate of change of parkinsonism. A score for each of 4 parkinsonian signs (parkinsonian gait, bradykinesia, rigidity, and tremor) was based on the sum of the scores for each of its individual items assessed with the UPDRS. The scores for the 4 parkinsonian signs were averaged to provide a continuous global parkinsonian score as previously described (11).

Clinical parkinsonism category

A categorical measure of parkinsonism is necessary to estimate the cases of incident parkinsonism. A previously validated parkinsonism category was constructed based on the number of the 4 parkinsonian signs present with UPDRS assessment. A parkinsonian sign was present if 2 or more of its items were scored as a mild or more severe abnormality. Clinical parkinsonism was present if at least 2 of the 4 parkinsonian signs were present (12).

Clinical diagnosis of PD

A clinical diagnosis of PD was based on available medical records and self-reported history as previously reported (5).

Demographic Covariates

Date of birth and sex were collected through a participant interview. Age in years was computed from self-reported date of birth and date of death.

Assessment of Brain Autopsy

Since their inception more than 2 decades ago, the parent studies (ROS and MAP) have employed a structured autopsy protocol to collect indices of diverse neuropathologies from a fixed number of predefined sites. This approach was chosen to facilitate comparison of postmortem indices and diverse aging phenotypes from large numbers of older individuals. To operationalize these novel studies, that is, to ensure high quality, reproducible data collection in large numbers of autopsies, not all brain regions could be sampled. The sites chosen spanned diverse cortical, subcortical, and brainstem sites as described below.

After death, the brain was removed and hemisected following standard procedure, as previously described (13). At the time of autopsy one hemisphere was frozen −80°C for further genomic studies. The other hemisphere was fixed in 4% paraformaldehyde for further neuropathologic and immunohistochemistry assessments conducted on fixed tissue, blinded to all clinical information. Tissue blocks were dissected from predetermined regions and are used for making the postmortem diagnosis of pathological Alzheimer’s disease (AD) or PD.

Postmortem indices of 5 neurodegenerative and 5 cerebrovascular disease pathologies were systematically assessed blinded to all clinical and cognitive data. For descriptive purposes, the 10 indices described below were dichotomized (Table 1). The primary analytic severity score for each of the 10 pathologies included in these analyses is described below.

Table 1.

Clinical and Postmortem Characteristics of the Analytic Cohort (N = 391)

Clinical Measures	Mean (SD) or n (%)
Age at baseline (y)	79.7 (6.68)
Age at death (y)	89.2 (6.51)
Female	273 (70%)
AD dementia at last visit	122 (31.2%)
Global parkinsonism score at baseline; last visit	9.8 (7.44); 16.1 (11.09)a
Parkinsonian gait score at baseline; last visit	19.1 (16.89); 22.7 (22.86)b
Bradykinesia score at baseline; last visit	12.9 (12.05); 19.0 (18.23)c
Rigidity present at baseline; last visit	n = 116 (30%); n = 151 (39%)d
Tremor present at baseline; last visit	n = 166 (42%); n = 155 (40%)e
Postmortem indices
Postmortem Interval (h)	8.1 (5.47)
Neurodegenerative pathologies
Lewy body	n = 105 (26.9%)
Nigral neuronal loss (moderate/severe)	n = 59 (15.2%)
Alzheimer’s disease (AD Reagan)	n = 233 (59.6%)
TDP-43	n = 103 (26.5%)
Hippocampal sclerosis	n = 32 (8.2%)
Cerebrovascular pathologies
Macroinfarcts present	n = 125 (31.2%)
Microinfarcts present	n = 110 (28.1%)
Atherosclerosis (moderate/severe)	n = 121 (31.0%)
Arteriolosclerosis (moderate/severe)	n = 118 (30.4%)
Cerebral amyloid angiopathy (moderate/severe)	n = 104 (26.6%)

Notes: On average, the severity of

^aglobal parkinsonism (t₃₉₀ −11.95, p < .0001);

^bparkinsonian gait (t₃₈₉ −19.92, p < .0001),

^cbradykinesia (t₃₉₀ −6.30 < 0.0001) and

^drigidity (chi-square 8.94, p = .003) were higher at the last visit compared with baseline.

^eTremor was uncommon and its severity was similar at baseline and the last visit (chi-square 0.924, p = .337).

Neurodegenerative Pathologies

Nigral neuronal loss

Dissection of diagnostic blocks included a hemisection of midbrain including substantia nigra. Nigral neuronal loss was assessed in the substantia nigra in the mid to rostral midbrain near or at the exit of the third nerve using hematoxylin and eosin (H&E) stain and 6 micron sections using a semiquantitative scale (0–3) (14).

Lewy Bodies

Seven regions (15) (substantia nigra, anterior cingulate cortex, entorhinal cortex, amygdala, midfrontal cortex, superior or middle temporal cortex, inferior parietal cortex) were assessed for Lewy bodies using α-synuclein immunostaining (LB509; 1:150 or 1:100, Zymed Labs, Invitrogen, Carlsbad, CA, USA; and pSyn#64; 1:20 000; Wako Chemicals, Richmond, VA, USA). The presence or absence of Lewy bodies in any of the 7 regions was recorded.

AD pathology

A modified Bielschowsky silver stain was used to visualize neuritic plaques, diffuse plaques, and neurofibrillary tangles in 5 cortical areas (hippocampus, entorhinal, midfrontal, middle temporal, and inferior parietal). A board-certified neuropathologist, blinded to clinical data, determined the pathological diagnosis of AD based on an intermediate to high likelihood of AD according to NIA Reagan criteria (16). In addition, a summary global AD pathology score was made by averaging regional scores. Amyloid-β was labeled with an N-terminus–directed monoclonal antibody (10D5; Elan, Dublin, Ireland; 1:1 000). PHFtau was labeled with an antibody specific for phosphorylated tau (AT8; Innogenetics, San Ramon, CA; 1:1 000) (17).

Transactive DNA response binding protein 43

Transactive DNA response binding protein (TDP) staging (0: none; 1: localized to amygdala only, 2: extended to other limbic regions, and 3: extended to neocortical regions) was determined using a phosphorylated monoclonal TAR5P-1D3 antibody (pS409/410; 1:100, Ascenion, Munich, Germany) (18).

Hippocampal sclerosis

The presence of hippocampal sclerosis was identified by severe neuronal loss and gliosis on H&E-stained sections in CA1 or subiculum (19).

Cerebrovascular Disease Pathologies

Infarcts

Chronic macroinfarcts identified during gross examination were confirmed histologically. The presence of microinfarcts was determined on sections in a minimum of 9 regions stained with H&E (20,21).

Cerebral amyloid angiopathy

Sections in 4 neocortical regions (ie, midfrontal, midtemporal, angular, and calcarine) were immunostained for β-amyloid (4G8; 1:9 000, Covance Labs, Madison, WI; 6F/3DDako; 1:50, North America Inc., Carpinteria, CA; and 10D5; 1:600, Elan Pharmaceuticals, San Francisco, CA). Meningeal and parenchymal vessels were assessed for amyloid deposition. Individuals with moderate or severe amyloid angiopathy were identified (22).

Atherosclerosis

A semiquantitative scale (0–3) was used to assess the severity of atherosclerosis was determined by visually examining the cerebral arteries and visible branches of the circle of Willis.

Arteriolosclerosis

The severity of arteriolosclerosis in the small vessels of the anterior basal ganglia was assessed using a semiquantitative scale (0–3) (23).

Mass Spectrometry–Based Proteomics Using Isobaric TMTs

TMT proteomics analysis was conducted on frozen tissue samples of the DLPFC. Details on the mass spectrometry–based proteomics, database searches, and quality control have been previously described (7). In brief, the samples were homogenized, and the protein concentration was determined. After protein digestion, isobaric TMT peptide labeling and high pH fractionation were performed. Fractions were then analyzed by liquid chromatography–mass spectrometry. The resulting mass spectrometry spectra were searched against the UniProthuman protein data base, with individual protein abundance checked against the global internal standard. Technical confounders including sequencing batch, MS mode, and postmortem interval were regressed out (7). A total of 8 356 proteins in 391 persons passed the final quality control.

Statistical Analysis

The frequencies of each of the age-related neuropathologies are reported. As we have done in prior publications, we used the square root transformation for global parkinsonism score. We first described the characteristics of the participants and used pairwise t-tests to compare the severity of parkinsonian signs at baseline and last visit.

Motor resilience has been understudied in part because it is rare to obtain (i) repeated measures of motor function in large numbers of older adults together with (ii) indices of diverse brain pathologies and (iii) proteome data. The approach used in the current study to isolate motor resilience has been successfully employed in our prior studies to discover cognitive resilience proteins (7,8). We used all available postmortem brain indices to isolate progressive parkinsonism not related to brain pathologies. Then we identified cortical proteins that may provide resilience to offset the negative effects of brain pathologies accumulating in aging brains.

Isolating motor resilience

An adult who maintains movement or has a slower rate of progressive parkinsonism in the presence of diverse brain pathologies can be said to manifest motor resilience to parkinsonism. Therefore, to identify proteins that might be associated with motor resilience, we needed to isolate progressive parkinsonism over many years prior to death that was not explained by indices of brain pathologies. We added terms for 10 brain pathologies and their interaction with time to the previous linear mixed-effect models. Lewy bodies (Estimate = 0.035, SE 0.018, p = .047 and nigral neuronal loss (Estimate = 0.020, SE 0.009, p = .030) were associated with progressive parkinsonism. TDP-43 (Estimate = 0.012, SE 0.006, p = .066), macroinfarcts (Estimate = 0.029, SE 0.015, p = .055), arteriolosclerosis (Estimate = 0.014, SE 0.008, p = .070) showed marginal associations with progressive parkinsonism. The summary global AD pathology score, hippocampal sclerosis, microinfarcts, atherosclerosis, and CAA were not related to progressive parkinsonism.

To visualize the heterogeneity of resilience to progressive parkinsonism, Figure 1A and B show the different trajectories of progressive parkinsonism we chose 100 randomly selected individuals examined in these analyses with and without adjustment for indices of 10 brain pathologies. The heterogeneity of progressive parkinsonism can be observed in both figures, that is, the heterogeneity of the trajectories as each line represents the trajectory for a different individual during the study follow-up. The variance of person-specific slopes of motor decline (light lines) after controlling for AD and related pathologies in Figure 1B is 16% less when compared with the person-specific variance of slopes for motor decline in Figure 1A. The reduction in the variance of person-specific slopes of decline after addition of brain pathologies indicates the part of motor decline that was due to the burden of brain pathologies, yet additional factors or proteins that are independent of brain pathologies and drive motor resilience remain to be identified.

Figure 1.

Heterogeneity of motor decline with and without controlling for Alzheimer’s disease and other brain pathologies. We randomly selected a subset of 101 individuals included in the analyses of the full cohort to illustrate the heterogeneity of progressive parkinsonism without adjustment for Alzheimer’s disease and other brain pathologies (A) and for residual progressive parkinsonism after including terms for indices of 10 brain pathologies (B). (A) Trajectories of progressive parkinsonism based on repeated annual measures of UPDRS testing prior to death. Each individual light line represents the estimated person-specific progressive parkinsonism for a single adult with the length of the line based on the number of years of follow-up. Bold black line represents average progressive parkinsonism. (B) Trajectories of residual motor decline in the same individuals shown in (A) to illustrate the heterogeneity of residual progressive parkinsonism after adding terms to the models in (A) for 10 indices of brain pathologies. Light lines show person-specific residual progressive parkinsonism, and bold black line represents average residual progressive parkinsonism.

Next, we performed a proteome-wide association analysis to examine the association of individual proteins with motor resilience by conducting 8 356 parallel linear mixed-effects models. To minimize false-positive results due to multiple testing, statistical significance was determined a priori at an α level of 5 × 10⁻⁶, corresponding to a Bonferroni correction for 10 000 tests. Implementing a strict Bonferroni correction may be too conservative and can exclude important proteins for further mechanistic and drug studies. Therefore, given our modest sample size and as we have done in prior studies, we also included proteins with a suggestive threshold of p < 1 × 10⁻⁴ (24).

For proteins with significant associations, we extended and “validated” the results by replacing their protein abundance with their corresponding gene expression from RNA-Seq. Regression models examined the association of proteins with individual neuropathologic indices. For targeted analyses, statistical significance was determined at a nominal level of α = .05 unless otherwise specified. The analyses were conducted using SAS/STAT software, version 9.4 (SAS Institute Inc, Cary, NC). All the models were adjusted for age and sex. Gene Ontology (GO) analysis was conducted using the Bioconductor package topGO (version 2.34.0) using the classic Fisher test. UniProt IDs were mapped to GO terms using the packages org. Hs.eg.db (version 3.7.0) and GO.db (version 3.7.0).

Standard Protocol Approvals, Registrations, and Patient Consents

Both studies were approved by an Institutional Review Board of Rush University Medical Center. Written informed consent was obtained from all study participants as was an Anatomical Gift Act for organ donation.

Results

Clinical and Postmortem Characteristics

There were 391 older persons included in these analyses and their clinical and postmortem characteristics are detailed in Table 1. During a mean (SD) of 8.7 (4.5) years of annual follow-up, the global parkinsonism score increased from 9.8 at baseline (median 8.1) to 16.1 proximate to death (median 14.5). Global parkinsonism is constructed from scores of 4 parkinsonian signs. On average, the severity of global parkinsonism, parkinsonian gait, bradykinesia, and rigidity was higher at the last visit compared with baseline. Tremor was uncommon and its severity was similar at baseline and at the last visit (Table 1). More than a third of the participants (37%) showed parkinsonism at study entry and more than half (53%) showed parkinsonism at the last visit proximate to death.

We dichotomized the 10 brain pathologies to count how many pathologies were observed in each individual. AD pathology was the most common neuropathology observed but one or more cerebrovascular disease pathologies were also very common (Table 1). At autopsy, 233 participants (59.6%) had a pathological diagnosis of AD according to modified National Institute on Aging Reagan criteria. About 97% of all autopsied individuals showed 1 or more pathologies: 1 pathology (41, 10%), 2 pathologies (n = 81, 21%), 3 pathologies (n = 95, 24%), 4 pathologies (n = 63, 16%), 5 pathologies (n = 54, 14%), 6 pathologies (n = 30, 8%), and 7–9 pathologies (n = 14, 4%). The average individual had 3 pathologies (median, 3; interquartile range, 2–5).

Identifying Motor Resilience Proteins

We used annual global parkinsonism scores as the continuous longitudinal outcome with linear mixed models which controlled for age and sex. On average parkinsonism increased by 0.12/points/y (Estimate = 0.115; SE, 0.008, p < .001). We then isolated parkinsonism resilience by after accounting for the effects of 10 brain pathologies, we then performed 8 356 parallel analyses examining the association of individual cortical proteins with parkinsonism resilience as illustrated in Figure 2.

Figure 2.

Proteome-wide association of cortical proteins with motor resilience. Each point on the plot represents the association of an individual protein with progressive parkinsonism after controlling for age, sex, and 10 neuropathologic indices. The horizontal coordinate is the corresponding gene location within the chromosome (defined as the midpoint of the start and end positions). The vertical coordinate is −log10 of the p-value if the protein is associated with a slower rate of progressive parkinsonism (ie, more resilience) and log10 of the p-value if it is associated with a faster rate of decline (ie, less resilience). The dashed lines correspond to the reference significance level representing α = 5 × 10⁻⁶. Tetratricopeptide Repeat Domain 38 (TTC38) was the lead protein associated with motor resilience after correction for multiple testing. Four additional proteins: carbohydrate kinase domain containing protein (CARKD); adaptor-related protein complex 1 subunit beta 1 (AP1B1), abhydrolase domain containing 14B (ABHD14B), and G protein subunit gamma 3 (GNG3) were also associated with motor resilience.

Tetratricopeptide repeat domain 38 (TTC38) was the lead protein with a higher level of protein associated with less motor resilience. Four additional proteins: carbohydrate kinase domain containing protein (CARKD); adaptor-related protein complex 1 subunit beta 1 (AP1B1), abhydrolase domain containing 14B (ABHD14B), and G protein subunit gamma 3 (GNG3) were also associated with motor resilience (Figure 3A–E). A higher level of AP1B1 and GNG3 was associated with more motor resilience and slower progressive parkinsonism. In contrast, a higher level of CARKD and ABHD14B was associated with less motor resilience and a faster rate of progressive parkinsonism (Table 2, left). These results were unchanged in sensitivity analyses when we excluded 18 of 391 (4.6%) individuals with a clinical diagnosis of PD for which they received levodopa or dopamine agonist (Table 2, right).

Figure 3.

Association of target protein with motor resilience and RNA expression. (A–E) Associations of individual proteins with motor resilience. In each panel, the circles indicate motor resilience; y-axis (ie, person-specific rates of progressive parkinsonism estimated from linear mixed models adjusted for demographic characteristics and neuropathologic conditions) plotted against the protein level (x-axis), with a corresponding regression line with 95% CI (shaded area). Motor resilience (y-axis) is unitless as it represents residual motor decline after regressing out age, sex, and the effects of 10 indices of brain pathologies. Tetratricopeptide repeat domain 38 (TTC38) was the lead protein associated with motor resilience after correction for multiple testing. Four additional proteins: carbohydrate kinase domain containing protein (CARKD); adaptor-related protein complex 1 subunit beta 1 (AP1B1), abhydrolase domain containing 14B (ABHD14B), and G protein subunit gamma 3 (GNG3) were also associated with motor resilience. (F) Associations of RNA expression of corresponding genes (x-axis) with motor resilience protein (y-axis). Red tile represents a positive association, and purple tile represents a negative association.

Table 2.

Motor Resilience Proteins and Progressive Parkinsonism

Series	Protein × Time	Model A All Decedents		Model B No PD
		Estimate (SE)	p-Value	Estimate (SE)	p-Value
1	TTC38 × Time before death	0.140 (0.029)	1.87 × 10−6	0.141 (0.030)	1. 80 × 10−6
2	CARKD × Time before death	0.413 (0.100)	3.50 × 10−5	0.417 (0.101)	4.02 × 10−5
3	ABHD14B × Time before death	0.175 (0.044)	6.48 × 10−5	0.166 (0.043)	1.17 × 10−4
4	AP1B1 × Time before death	-0.504 (0.121)	3.12 × 10−5	-0.488 (0.122)	6.78 × 10−5
5	GNG3 × Time before death	-0.276 (0.068)	4.82 × 10−5	-0.280 (0.072)	9.40 × 10−5

Notes: PD = Parkinson disease. Each row shows the results for the interaction of the protein (left column) with Time in the study before death (rate of change of progressive parkinsonism prior to death) from 2 linear mixed-effect models. Each model included an additional 26 terms (not shown) including: Time in the study before death; Age, Sex, 10 indices of brain pathologies and the protein in column 1 as well as the interaction of Time with each of the 13 cross-sectional terms. Model 1 includes all decedents (n = 391). Model 2 repeated Model 1 after excluding decedents (n = 18) with a clinical diagnosis of PD prior to death. Tetratricopeptide Repeat Domain 38 (TTC38) was the lead protein associated with motor resilience after correction for multiple testing. Four additional proteins: carbohydrate kinase domain containing protein (CARKD); adaptor-related protein complex 1 subunit beta 1 (AP1B1), abhydrolase domain containing 14B (ABHD14B), and G protein subunit gamma 3 (GNG3) were also associated with motor resilience.

Next, we examined the variance of progressive parkinsonism accounted for by a model that included age, sex, and terms for 10 brain pathologies (Model 1). The terms in Model 1 accounted for 17% of the variance of progressive parkinsonism. Then in a second model, we sequentially added each of the 5 proteins to the terms in the first model, that is, age, sex, and terms for 10 brain pathologies (Model 2). We compared the variance of Model 2 for each of the 5 proteins to the variance accounted for by Model 1. The difference between these models represents the variance accounted for by the protein(s) over and above the variance due to age, sex, and the 10 brain pathologies (Supplementary Table S1). These proteins accounted for additional unexplained variance of progressive parkinsonism (TTC38 [10.0%], CARXD [6.2%], ABHD14B [5.5%], AP1B1 [8.3%], and GNG3 [7.6%]). In a single model, the 5 proteins together accounted for an additional 25% of the variance of progressive parkinsonism compared with the 17% accounted for by the 10 indices of brain pathologies together (Supplementary Table S1).

Global parkinsonism score is constructed from 4 parkinsonian signs. In further analyses, we examined the associations of motor resilience proteins with the progression of the 4 parkinsonian signs. The 5 motor resilience proteins were associated with the progression of parkinsonian gait and bradykinesia, but were not associated with the progression of tremor or rigidity (Supplementary Table S2). The lack of association with tremor and rigidity may be due to reduced power as both were uncommon in this cohort.

To identify candidate pathways which may underlie parkinsonism resilience, we conducted an exploratory Gene Ontology (GO) analysis. For this analysis, proteins with p ≤ .05 were considered as potentially associated with parkinsonism resilience (623 of 7 616 proteins with GO annotation). A total of 1 896 biological pathways with at least 50 annotated proteins were assessed for an enrichment with the parkinsonism resilience-associated proteins (Supplementary Figure 1). Five pathways closely related in the GO hierarchy achieved an enrichment p-value < .001 and suggest the ubiquitin-dependent proteasome pathway as a molecular mechanism providing parkinsonism resilience (table in Supplementary Figure 1).

Motor Resilience Proteins Associations With RNA Expression and Genetic Risk of PD

To further examine these motor resilience proteins and their associations with other streams of genomic data, we examined their associations with RNA gene expression levels in DLPFC, which were available in all decedents. As illustrated in Figure 3F, the correlations between the identified resilience proteins and corresponding gene expression levels vary. TTC38 (r = .21, p < .0001), CARKD (r = .24, p < .0001), and ABHD14B (r = .14, p = .005) were related to corresponding gene expression levels but AP1B1(r = .06, 0.267) or GNG3 (r = .08, p = .100) were not. Although TTC38, CARKD, and ABHD14B gene expression levels were related to resilience proteins, they were not related to parkinsonism resilience (all p’s > .155).

We then examined the association of these proteins with a previously published polygenic risk score developed for PD based on a recent meta-analysis (25,26). The polygenic risk score was associated with TTC38 (R = .14, p = .013) and ABHD14B (R = .13, p = .013), but not with AP1B1, CARKD, or GNG3.

In further analyses, we examined whether proteins from 62 candidate genes for PD were related to progressive parkinsonism in the current study (27,28). Thirty-two of 62 were detected in the current data set. Three proteins KPNA1 (Karyopherin Alpha 1), KCNIP3 (Calsenilin), and FYN (protein-tyrosine kinase oncogene) were nominally associated with the rate of progressive parkinsonism in the current study (Supplementary Table S3).

Parkinsonism Resilience Proteins and Postmortem Indices of Brain Pathologies

Although our models for parkinsonism resilience controlled for indices of 10 brain pathologies, proteins that provide resilience may be pleotropic and have multiple functions including independent associations with the postmortem neuropathologies measured in this study. We consider a significance level of <.005 to correct for multiple comparisons because we examined the association of each protein with multiple neuropathologies (Supplementary Table S4). Using this cutoff, 3 of the 5 motor resilience proteins were associated with tangle pathology. A higher level of AP1B1 (Estimate −2.312 [SE 0.763, p < 3.0 × 10⁻³]) or GNG3 (Estimate −1.456 [SE 0.434, p < 8.7 × 10⁻⁴]) and a lower level of ABHD14B (Estimate 1.010 [SE 0.305, p < 1.0 × 10⁻³]) were associated with a lower burden of tangle pathology. CARKD, ABHD14B, and AP1B1 showed marginal associations with β-amyloid (CARKD and ABHD14B), nigral neuronal loss (CARKD and ABHD14B), and macroscopic infarcts (AP1B1) (Supplementary Table S4), but these proteins were not related to Lewy body pathology, nigral neuronal loss, or the other brain pathologies. Marginal associations were noted for ABHD14B and CARKD with β-amyloid (p’s < .03) and ABHD14B and CARKD were weakly associated with nigral neuronal loss (p ≤ .05), but not with Lewy body pathology or the other neuropathologies. TTC38 and AP1B1 were not associated with any of the neuropathologies (Supplementary Table S4).

Discussion

This proteome-wide discovery study identified 5 cortical proteins that may provide resilience to progressive parkinsonism and offset the negative effects of brain pathologies that accumulate in aging brains. An adult with progressive parkinsonism may have more or less resilience relative to the average progression in parkinsonism associated with brain pathologies. While controlling for indices of 10 brain pathologies, we found that higher levels of TTC38, CARKD, and ABHD14B were associated with less resilience and a faster rate of progressive parkinsonism. By contrast, higher levels of AP1B1 or GNG3 were associated with more motor resilience and a slower rate of progressive parkinsonism in older adults. These 5 proteins together accounted for an additional 25% of the variance of progressive parkinsonism compared with 17% accounted for by the 10 brain pathologies together. Some but not all of these resilience proteins were related to RNA gene expression levels as well as to a polygenic risk score for PD and candidate genes for PD. Although these associations were modest, the convergence of some gene and protein expression levels lend confidence to our protein findings. On the other hand, some gene and protein expression were not correlated. This may indicate that there are other pathways leading to expression of these proteins associated with progressive parkinsonism (24). Targeting these proteins in further drug discovery studies may result in approaches to slow progressive parkinsonism offsetting the negative effects of brain pathologies that accumulate in aging brains and are generally untreatable.

Parkinsonism affects 50% or more of adults by age 85 years and its presence is associated with diverse adverse health outcomes (29). Although parkinsonism can be an early sign of PD, the prevalence of the clinical diagnosis of PD in the general population is only 2%–5% (30). Thus, even the occurrence of PD in about 10% of older adults without a clinical diagnosis of PD does not account for the estimated prevalence of 50% with clinical parkinsonism that has been reported (12,31). Furthermore, work has shown that a higher burden of mixed-brain pathologies especially cerebrovascular disease pathologies that commonly accumulate in older brains are associated with more rapid progressive parkinsonism. Given the magnitude and consequence of parkinsonism in our aging population it is crucial to identify factors such as lifestyle or molecular mechanisms that can provide resilience to progressive parkinsonism and offset the negative effects of brain pathologies that are generally untreatable.

Resilience refers to a person’s ability to tolerate life-course-related brain changes such as injury or the negative effects of brain pathologies while maintaining function (6). In addition, to structural redundancies, the brain is plastic, actively responding to damage, behavior, and past experiences. Resilience includes both functional compensation through engagement of redundant neuronal populations and dynamic molecular resilience to maintain cellular homeostasis to counteract senescence. Most prior work about brain reserve or resilience has focused almost exclusively on identifying risk factors or modifiable behaviors that promote resilience for cognitive decline (6).

The current study focusing on progressive parkinsonism extends our recent work which identified cortical proteins that provide resilience for cognitive and motor decline (7,9). The current study identified 5 novel proteins associated with progressive parkinsonism not explained by 10 diverse brain pathologies. As we reported in our prior publications, brain pathologies in the current study accounted for only a minority of the variance of progressive parkinsonism (17%) (3). By contrast, the individual parkinsonism resilience proteins accounted for 5%–10% of progressive parkinsonism and when considered together in a single model they accounted for an additional 25% of progressive parkinsonism compared with a model with only brain pathologies and demographics. These proteins have important clinical implications as they may serve as high-value therapeutic targets for drug discovery that may lead to interventions that maintain motor function despite the accumulation of brain pathologies that are generally untreatable.

Of the proteins identified, TTC38 that is associated with more motor resilience may play a role in folic acid metabolism that is crucial for CNS function and is associated with carboxypeptidase N deficiency that is implicated as a regulator for inflammation (32). CARKD, also associated with more resilience, catalyzes the dehydration of the S-form of NAD(P)HX, metabolism of water-soluble vitamins and is associated with a novel neurodegenerative disorder exacerbated by febrile illnesses (33). ABHD14B is involved in metabolism by its hydrolase activity and transferring an acetyl group from an acetylated lysine residue of a protein to coenzyme A making acetyl-coenzyme A (34). It also interacts with other molecules in formation of complexes required to activate transcription processes and the expression level of its gene has been found associated with cancers (35,36). GNG3, associated with less motor resilience, is abundantly and widely expressed in the brain and contributes to the specificity of many receptor signaling pathways. Its deficiency manifests as seizures in mice and may be a biomarker for glioblastoma in humans (37,38). It was identified as a hub gene and downregulated in aging human brain. Among its related pathways are GABAergic synapse and p75 NTR receptor-mediated signaling (39). AP1B1, also associated with less resilience, is an adaptin that recruits membrane proteins for vesicles. AP1B1 mutations are associated with several neurological diseases including meningiomas, sensorineural hearing loss, peripheral neuropathy, and mental retardation. Trem2 variants with increased expression upregulate AP1B1 in a microglial anti-inflammatory phenotype (40).

Exploratory gene ontology analysis suggested that the ubiquitin-dependent proteasome pathway is a molecular mechanism providing motor resilience. The ubiquitin–proteasome pathway targets and degrades unneeded or misfolded proteins and has been implicated in neurodegenerative diseases (41–43). Normal proteostasis protects neurons from pathological protein aggregates and may therefore be a shared mechanism providing resilience against various neuropathologies common in our cohort (44). In addition to their association with progressive parkinsonism, some but not all of the motor resilience proteins showed associations with transcriptome and a polygenic risk score for PD. The approach employed in the current study could be extended and applied to explicate the neurobiology underlying other important aging phenotypes as well as provide novel drug discovery targets to mitigate the negative effects of neuropathologies accumulating in aging brains.

Some limit the concept of resilience to factors that maintain function in the face of pathologies or other insults (6,45,46). Although the term vulnerability is often used to refer to what is here considered less resilience, we find this conceptual approach problematic. Most importantly, considerable empirical evidence unequivocally demonstrates that the vast majority of risk factors, be they psychosocial exposures, diet, or biologic factors, cover a range for which high (or low) is beneficial and low (or high) is detrimental. Thus, the approach presented in the current study based on proteins that can either hasten or slow the rate of motor decline conforms with the way in which human biologic processes work. Furthermore, treating resilience and vulnerability as different constructs rather than 2 sides of the same “coin” markedly limits power by effectively cutting the distribution of the exposure in half. Finally, words constrain thinking and using 2 terms with opposite meanings can result in the misconception that there are 2 processes. Hence, considering all of the proteins identified as subsumed under a single biologic process suggests that all adults may show some degree of resilience some having higher and some lower-than-average resilience (47,48).

Elucidating the basis for motor resilience may have important public health consequences in directing efforts to prevent late-life motor impairment. Our recent work that has shown that multiple brain pathologies commonly accumulate in aging brains and may contribute to progressive parkinsonism (5). Thus, it is likely to be impractical to significantly affect the negative consequences of these brain pathologies in older adults even if new treatment were discovered for AD and other pathologies. Alternatively, the current report suggests that targeting genes and proteins that provide motor resilience may offer a practical and complementary approach for potential therapies that can maintain motor function in aging adults even in the absence of treatments for accumulating brain pathologies.

Our study has several strengths. The data come from 2 prospective cohort studies with high follow-up and autopsy rates. Proteome-wide association analysis links thousands of high-abundance cortical proteins to progressive parkinsonism over many years prior to death while controlling for diverse common brain pathologies. The approach provides a comprehensive analysis of protein signals that are associated with motor resilience. Bonferroni correction for multiple testing reduces the chance of spurious findings. This threshold may be overly conservative for a discovery study and if strictly employed may exclude identification of important proteins (24).

This study also has some limitations, both of these select cohorts may not represent the general populations as most are non-Latino Whites with higher-than-average educational levels. Findings in this study will need to be replicated in other more diverse cohorts. The paucity of proteins associated with rigidity and tremor in this study may be due to reduced power since both were uncommon in this cohort of older adults without PD. Neuropathologic conditions are assessed in multiple brain regions, whereas the tissue collection site for TMT-MS proteomics was restricted to a single site in the prefrontal cortex. Brain regions and other pathologies such as white matter abnormalities not measured may account for additional unexplained progressive parkinsonism. Neuropathologies outside of the brain that contribute to motor decline were not measured in the current study neural control systems underlying motor decline extend beyond the brain to spinal networks that via peripheral nerve regulate muscle firing in the periphery outside the CNS. So further studies examining proteome of other motor-related tissues outside the brain are needed to complement this current study. This would provide a more complete description of the genes and proteins driving motor resilience as well as signaling and cross-talk between spatially distinct motor networks that underlie movement and parkinsonism in older adults.

Funding

This research was supported by National Institutes of Health grants: P30AG10161, P30AG072975, U01AG61356, U01 AG061357, R01AG15819, R01AG17917, R01AG59732 and K01AG054700. The funding organizations had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.

Conflict of Interest

None declared.

Authors Contributions

A.S.B.: drafting initial/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. L.Y.: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data; H.-U.K.: drafting/ revision of the manuscript for content, including medical writing for content; analysis or interpretation of data; A.R.Z.: drafting/revision of the manuscript for content, including medical writing for content; S.O.: drafting/revision of the manuscript for content, including medical writing for content; F.G.: drafting/revision of the manuscript for content, including medical writing for content; S.T.: drafting/revision of the manuscript for content, including medical writing for content; N.T.S.: drafting/revision of the manuscript for content, including medical writing for content, major role in acquisition of data; A.I.L.: drafting/revision of the manuscript for content, including medical writing for content, major role in acquisition of data; D.A.B.: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data and funding its collection.