Decrease in Aβ42 predicts dopa-resistant gait progression in early Parkinson disease

Objective: This prospective observational study investigates the role of CSF biomarkers in predicting progression of dopa-resistant gait impairments in Parkinson disease (PD) in the first 36 months from diagnosis. Methods: Quantitative gait analysis was carried out longitudinally using an instrumented walkway (GAITRite) in 108 people with PD and 130 age-matched controls. A subgroup of 44 people with PD underwent lumbar puncture from which a battery of CSF biomarkers was measured: β-amyloid 1–42 and 1–40 (Aβ42 and Aβ40), total and phosphorylated tau protein (t-tau/p-tau181), and α-synuclein (αSyn). Linear mixed models examined the association between CSF and dopa-resistant gait characteristics (defined as substantial progression despite optimal medication). Results: Low baseline CSF Aβ42, and to a lesser extend Aβ40, predicted decline in gait characteristics in the first 3 years following diagnosis, independently explaining up to 12% of progression of step time variability (single task) and step length variability (dual-task). Interestingly, these findings were independent of age and cognition. Conclusions: These findings implicate underlying amyloid pathology in neural networks involved in locomotor control. Results suggest that disturbed Aβ metabolism may be a biomarker for dopa-resistant gait impairments in early PD. Our findings raise interesting questions regarding therapeutic interventions such as compounds or molecules aimed at reducing amyloid burden to mitigate gait disturbance in early PD and potentially falls risk. Finally, progression of discrete gait characteristics suggests they may have potential as clinical biomarkers of pathology and disease progression.

in these studies limits interpretation of findings. More importantly, longitudinal studies are lacking and urgently required in order to establish prediction.
The aims of this study were to investigate the role of CSF biomarkers to predict progression of dopa-resistant gait impairments in the first 36 months from diagnosis in PD. We were interested in the mechanisms underpinning dopa-resistant gait progression to provide an essential platform for future therapeutic interventions to mitigate gait disturbance and potential falls risk. Based on previous cross-sectional literature, we hypothesized that Ab42 and p-tau 181 would predict progression in dopa-resistant gait characteristics. METHODS

Participants. Participants were recruited into
Incidence of Cognitive Impairment in Cohorts with Longitudinal Evaluation (ICICLE)-GAIT a median of 5 months from diagnosis. This is a nested study within ICICLE-PD, an incident cohort study conducted between June 2009 and December 2011. A subset of the cohort was recruited into ICICLE-Gait at the same time (figure 1). Controls of a similar age and sex were recruited from community sources. The methods have been described in full in previous publications, [12][13][14] and are included as supplemental material at Neurology.org. Participants were tested "on" medication, which was defined as 1 hour after PD medication. Participants were evaluated at the Clinical Ageing Research Unit, Newcastle University, UK.
Standard protocol approvals, registrations, and patient consents. The study was approved by the Newcastle and North Tyneside research and ethics committee and all participants gave informed consent.
Demographic and clinical measures. Clinical assessments included a standardized neurologic examination and the Movement Disorder Society-revised Unified Parkinson's Disease Rating Scale, 15 from which Hoehn & Yahr stage 16 and motor phenotype were calculated. 17 Levodopa equivalent dose (LEDD) scores were calculated according to established methods. 18 Global cognition was assessed using Montreal Cognitive Assessment (MoCA). 19 Quantitative gait analysis and gait characteristics. Gait was assessed using a 7 meters long 3 0.6 meters wide instrumented walkway (Platinum model GAITRite, software version 4.5, CIR Systems, Franklin, NJ). Participants were instructed to walk at their comfortable walking pace for 2 minutes around a 25-meter oval circuit under single and dual-task conditions. The dual-task protocol involved walking and memorizing digits, based on the Wechsler Forward Digit Span, which was used as the concurrent cognitive task. 20,21 Gait was repeatedly sampled as participants walked over the GAITRite mat (included in the 25-meter circuit) for a minimum of 5 passes (.40 steps per participant). 22 Gait was quantified according to an a priori model developed for older adults 23 and validated in PD 24 that describes 16 discrete gait characteristics. 25 We examined change in each gait characteristic over 36 months and characteristics that exhibited substantial change (despite optimal medication) were defined as doparesistant. Figure e-1 provides further details of the acquisition and processing of the gait data.
Quantification of CSF biomarkers. CSF biomarkers were measured using a robust protocol. 13 Lumbar puncture (LP) was performed on a subset of consenting participants using a standardized method as detailed previously. 13 All LPs were done between 8 and 10 AM after an overnight fast and while withholding PD medications. Samples were centrifuged within 15 minutes of collection at 2,000 g at 48C for 10 minutes. The supernatant was divided into aliquots and frozen at 2808C, then analyzed for Ab42 and Ab40 using commercially available assays: Ab42: Statistical analysis. CSF biomarkers were selected for analysis on the basis of previous reports. 10,11 Dopa-resistant gait impairments were identified as follows: first, change per year for all (16) gait characteristics derived from single and dual task testing was assessed with a linear mixed-effects model (lme4 package, R statistical software version 3.2.2, Vienna, Austria). 28,29 Participants and time (from baseline assessment to subsequent testing sessions) were included as random effects and age at baseline and sex as fixed effects. We then examined between-group change in gait with group (control, PD) as a fixed effect. Rate of progression was determined in the total cohort (n 5 108 PD and 130 controls) over 36 months (repeat assessments every 18 months from diagnosis) and then extracted for each individual for further analysis. Finally, bivariate correlations between change in LEDD and change in gait over 36 months for all CSF markers were conducted (data not shown), with significant relationships revealing dopa-resistant gait characteristics. These 3 steps validated the dopa-resistant classification; namely, substantial progression of gait impairment despite optimal medication; progression greater than controls; and no association between change in LEDD and gait over 36 months. The second stage of analysis established whether CSF markers could predict progression in dopa-resistant gait characteristics using general linear modeling and controlling for age, global cognition (MoCA), and baseline gait. Preliminary data analysis suggested a potential interaction between baseline gait and CSF markers in predicting gait progression, and models were examined with and without this interaction. Baseline gait was dichotomized around the group median (to ensure balanced groups) for each variable.
Linearity of progression, normality, and homoscedasticity were inspected to ensure robustness of the fitted mixed and general linear models. There was moderate positive skewness in gait decline for step length and step time variability. Rerunning the models with log transformations did not affect findings; therefore the final models are presented using untransformed data to aid interpretation. A threshold of p , 0.05 was used to inform interpretation and 2-tailed 95% confidence intervals were calculated as interval estimates for rate of progression of gait impairment and its predictors in the general linear model.

RESULTS
Participant characteristics. Demographic and clinical data are shown in table 1. Of the 158 participants recruited to ICICLE-PD, 56 consented to LP and of these, 47 also consented to ICICLE-Gait. Participants without longitudinal assessment at each time point were excluded, leaving 44 participants for analysis. Participants were a median of 5 months postdiagnosis at baseline and 41 months at the 36-month follow-up assessment. They had mild disease with little difference between the subgroup who underwent CSF examination and the total PD cohort. There were no differences in global cognitive scores between the subgroup and total group of participants with PD. The UPDRS-III scores, PIGD phenotype, and mean LEDD dose were lower than the entire group, indicating less severe disease.
Gait progression in early PD. Gait outcomes were stable over 36 months for controls. For participants with PD, dopa-resistant gait characteristics (and % yearly progression rate) under single task conditions included variability of step time (4.5% per year), step length (5.7% per year), and step width (3.4% per year), and under dual-task conditions, variability of step length (5.7% per year) (table 2 for descriptive data and change depicted in absolute values). Progression of gait impairment was defined by an increase in variability (expressed as the SD). Increased dosage of levodopa (LEDD) over the 36 months was not related to progression of variability of step time (single-task r 5 0.015, p 5 0.877), or step length (single-task r 5 20.039, p 5 0.687; dual-task r 5 0.155, p 5 0.108), corroborating that these variables represented doparesistant progression. However, an increase in levodopa dosage was related to an increase in single-task step width variability (r 5 0.218, p 5 0.023), and was therefore not included in subsequent analysis. Table 3 and figure 2 describe the relationship between baseline CSF and change in gait. Low baseline Ab42 was an independent predictor of progression of step time variability (single task) in people with high baseline step time variability (above median: .16.7 ms). The overall model accounted for 48.5% of variance in progression of step time variability with Ab42 independently accounting for 12.0%, independent of age and global cognition, which were not significant predictors. Ab42 predicted progression of step length variability (dualtask) independent of baseline gait values, explaining 26.6% of the overall model variance and Ab42 independently explaining 9.8%. Age and global cognition were not significant predictors in the model. In addition, we found that Ab40 was a significant predictor of progression of step length variability (dual-task), with the full model explaining 22% of variance and Ab40 alone explaining 11.1%. A significant interaction indicated a low baseline Ab40 predicted progression in step length variability in participants with a high baseline step length variability (above median: .2.36 3 10 22 meters). There were no associations with total and p-tau 181 , Ab40/42, or aSyn. NFoG, n (%) who report FoG - Ab40, pg/mL ---10,532 (4,035) a-Synuclein, pg/mL  Because previous reports highlight a relationship with motor phenotype (PIGD), we carried out additional analysis to see if PIGD phenotype predicted gait progression. While we saw a relationship at baseline, additional analysis of variance showed that gait progression did not differ between PIGD and TD motor phenotypes (p 5 0.407). DISCUSSION We report a study exploring the role of CSF biomarkers in the progression of gait impairment in early PD. We prioritized dopa-resistant gait characteristics because of the need to understand mechanisms explaining progression and identify potential therapeutic targets. Low baseline CSF Ab42, and to a lesser extent Ab40, predicted decline in gait characteristics in the first 3 years following diagnosis. The nature of the findings suggests a role for amyloid pathology. Our study included longitudinal evaluation in very early PD, the use of a robust model to inform (a priori) the selection of gait characteristics, and a broad battery of CSF proteins. The study was carried out in a community-based prospective incident cohort with robust case ascertainment and thus generalizes to idiopathic PD. Our findings are important for 2 key reasons. First, they point to potential mechanisms and novel targets for intervention to mitigate dopa-resistant gait disturbance (and by implication fall risk) at a very early stage of PD, and second, they identify discrete gait characteristics as potential clinical biomarkers for disease progression and therapeutic response.
Two previous studies have explored the relationship between CSF proteins and motor features, classified by motor phenotype. Lower values of Ab42 and to a lesser extent Ab40 were predictors independently explaining up to 12% of the variance in progression of dopa-resistant gait characteristics compared with CSF total and p-tau 181, Ab40/42, and aSyn, which were not associated with gait progression. Our findings broadly concur with previous cross-sectional reports showing a relationship with amyloid (from CSF and PET imaging) and motor disturbance especially in the PIGD phenotype in early and advanced PD. 10,11,30 However, importantly, our findings highlight limitations of previous work and argue for a more selective investigation of gait impairment.
We found that motor phenotype is not a proxy for dopa-resistant gait impairments as it is not specific to dopa-resistant features or their progression, highlighting the limitations of cross-sectional study design and the use of motor phenotype. This is most likely explained by recent work that shows motor phenotype is unstable in early disease and individuals transition between phenotypes, questioning its longitudinal utility. 31 This study provides

Postural control
Step length asymmetry, evidence implicating amyloid in the pathogenesis of dopa-resistant gait characteristics in early PD, although direct causality is unknown.
Our findings highlight a possible role for disturbed amyloid metabolism in the neural networks involved in gait control. A number of explanations are possible to explain the mechanisms by which disturbed amyloid metabolism may exert an influence on gait. Lower values of Ab42 and Ab40 are generally considered to correlate with plaque formation in the brain. 7,32 Ab plaque formation may exert a direct effect on synaptic function in the absence of tau 33 influencing the neural circuitry subserving dopa-resistant gait. Alternatively, it is possible that amyloid has a synergistic effect on other proteins such as tau or aSyn catalyzing protein misfolding, 34-36 leading to a more aggressive rate of progression in gait disturbance. However, Table 3 General linear models to predict change in gait using CSF (n 5 44) of people with Parkinson disease

Models including Ab42
Change in single-task step time SD, m/s (R 2 adj 5 48.5%) contrary to our hypothesis, we did not see a relationship with total tau or p-tau 181 . These findings, however, concur with other reports that tau per se is not a primary driver of pathogenesis in PD. 8 A combination of plaque formation, aSyn aggregation, and disturbance of neurotransmitter function (e.g., through synaptic rarefication via protein aggregation) is most likely to contribute to mechanisms underpinning gait progression. Age-related white matter burden may also contribute, although our results argue that for this cohort age was not a significant feature. Postmortem follow-up is required to confirm the precise mechanism. It appears that a combination of different pathologic influences mediate gait progression in a discrete manner in PD, emphasizing the need for a multimodal therapeutic approach. We determined whether gait characteristics were dopa-resistant by modeling gait progression over 3 years from baseline, despite optimal medication. We also identified only those variables that showed significantly greater change compared to controls over the same time period to avoid confounding due to age-associated decline. Controls showed very little change in gait.
Participants with PD exhibited significant decline in 3 gait characteristics over and above controls under single task (variability in step time, step length, and step width) and one under dual task (variability of step length). Variability of step time and step length were prominent, in accordance with earlier work.
Step time variability in particular is a sensitive marker of incipient pathology. 37 Establishing the functional and structural correlates using multimodal brain imaging will help inform the neural basis of these findings. We were also able to discern those gait characteristics that were controlled by levodopa (not reported) (e.g., step length and asymmetry), highlighting specific positive benefits of levodopa medication in early stages.
The effect of single and dual task protocols on results is also worthy of comment. During singletask gait, participants are able to recruit additional attentional resources to maintain gait performance, in contrast to dual-task conditions, which preclude cognitive compensation by distracting attention. Our findings imply that dual task step length variability is mediated through motor circuits, whereas single task step time variability is more likely to be subserved by motor and cognitive networks, as Correlation between baseline CSF and change in gait Relationship between baseline CSF markers and change in gait in people with Parkinson disease.
indicated by a (tentative) significant association between the MoCA and step time variability. Brain imaging with amyloid and tau in early PD will help clarify the topographic distribution of pathology driving these changes. Our findings have interesting clinical implications. CSF Ab42 is a validated in vivo marker of Ab accumulation and is implicated in progression of dopa-resistant gait impairments in early PD. Early targeting of Ab pathology has been proposed as a potential strategy to halt or prevent cognitive decline and dementia in PD. 8,9 Therapies to reduce accumulation of amyloid or target its production are under way. We argue that this may also offer the potential to target early gait impairments, and moreover, may provide a global strategy to concurrently target gait and cognitive impairment. Recently, amyloid deposition in the brain and decreased CSF amyloid have also been shown to predict falls in older adults at risk of dementia, 37 in support of this. Longer follow-up will determine the relationship with fall risk in our cohort.
Our data also suggest that selective gait characteristics may have a role to play in identifying early PD pathology and specific gait characteristics may be useful markers of progression in disease modification trials. Evidence that motor changes precede cognitive changes is common in older adults 38 and has fueled an interest in the role of gait as a clinical biomarker of cognitive decline. 6 The shared pathologic mechanisms of gait and cognitive decline support the potential of gait characteristics as clinical biomarkers of cognitive decline in PD 14,39 and work is under way investigating this.
Some study limitations should be acknowledged. The population was drawn from an incident PD cohort followed from diagnosis with repeat assessments every 18 months. While misdiagnosis may have contributed, this is unlikely to have had a major effect. Diagnosis followed a stringent process and the flowchart highlights revised diagnosis over the time course of the study, showing that the numbers are low. Dropout may also have confounded our analysis; however, this was low, suggesting it would not exert any undue influence. It is difficult to compare the absolute values of our CSF data with those of others due to the use of different assays across studies. Moreover, we did not have a control cohort with CSF. However, we were able to control for the effects of aging in our gait characteristics, only considering those that changed with respect to controls, and are therefore confident that our findings are PD-specific. Given the difficulty of obtaining CSF, this emphasizes the importance of clinically available biomarkers (such as high step time variability at diagnosis).
Our findings suggest that disturbed Ab metabolism is a biomarker for dopa-resistant gait impairments in early PD. They also raise interesting questions with respect to therapeutic interventions such as compounds or molecules aimed at reducing amyloid burden to mitigate gait disturbance in early PD and by implication fall risk. Comprehensive reporting of gait characteristics was critical to this study and helped explain the heterogeneity of gait disturbance and therapeutic response. Moreover, selective progression of gait characteristics suggests they may have potential as clinical biomarkers of pathology and disease progression. Finally, the overlap in the pathologic substrates of gait and cognitive decline raises an interesting question as to whether a monotherapy may be able to concurrently target gait and cognitive dysfunction in the future.

AUTHOR CONTRIBUTIONS
Prof. Rochester: PI ICICLE-Gait; responsible for study concept and design, analysis and interpretation, drafted first manuscript. Dr. Galna: data analysis and interpretation, manuscript revision. Dr. Lord and Dr. Yarnall: interpretation and manuscript revision. R. Morris, Dr. Duncan, and Dr. Khoo: data acquisition and manuscript revision. Dr. Mollenhauer: manuscript revision and interpretation. Prof. Burn: PI on ICICLE-PD; manuscript revision.