Abstract
Previous observational studies suggested that sarcopenia is associated
with Parkinson disease (PD), but it is unclear whether this association
is causal. The objective of this study was to examine causal
associations between sarcopenia-related traits and the risk or
progression of PD using a Mendelian randomization (MR) approach.
Two-sample bidirectional MR analyses were conducted to evaluate causal
relationships. Genome-wide association study (GWAS) summary statistics
for sarcopenia-related traits, including right handgrip strength
(n = 461,089), left handgrip strength (n = 461,026), and appendicular
lean mass (n = 450,243), were retrieved from the IEU OpenGWAS database.
GWAS data for the risk of PD were derived from the FinnGen database
(4235 cases; 373,042 controls). Summary-level data for progression of
PD, including progression to Hoehn and Yahr stage 3, progression to
dementia, and development of levodopa-induced dyskinesia, were obtained
from a recent GWAS publication on progression of PD in 4093 patients
from 12 longitudinal cohorts. Significant causal associations
identified in MR analysis were verified through a polygenic score
(PGS)-based approach and pathway enrichment analysis using genotype
data from the Parkinson’s Progression Markers Initiative. MR results
supported a significant causal influence of right handgrip strength
(odds ratio [OR] = 0.152, 95% confidence interval [CI] = 0.055–0.423,
adjusted P = 0.0036) and appendicular lean mass (OR = 0.597, 95%
CI = 0.440–0.810, adjusted P = 0.0111) on development of
levodopa-induced dyskinesia. In Cox proportional hazard analysis,
higher PGSs for right handgrip strength (hazard ratio [HR] = 0.225, 95%
CI = 0.095–0.530, adjusted P = 0.0019) and left handgrip strength
(HR = 0.303, 95% CI = 0.121–0.59, adjusted P = 0.0323) were
significantly associated with a lower risk of developing
levodopa-induced dyskinesia, after adjusting for covariates. Pathway
enrichment analysis revealed that genome-wide significant
single-nucleotide polymorphisms for right handgrip strength were
substantially enriched in biological pathways involved in the control
of synaptic plasticity. This study provides genetic evidence of the
protective role of handgrip strength or appendicular lean mass on the
development of levodopa-induced dyskinesia in PD. Sarcopenia-related
traits can be promising prognostic markers for levodopa-induced
dyskinesia and potential therapeutic targets for preventing
levodopa-induced dyskinesia in patients with PD.
Subject terms: Parkinson's disease, Risk factors
Introduction
Parkinson disease (PD) is the second most common neurodegenerative
disease and is characterized by the progressive loss of nigrostriatal
dopaminergic neurons^[34]1. The pathogenesis of PD is multifactorial,
resulting from the complex interaction of polygenetic susceptibility,
the environment, and lifestyle factors^[35]2. Symptoms of PD are
complex, including classic motor features (e.g., bradykinesia,
rigidity, rest tremor, gait disturbances) and various nonmotor features
(e.g., cognitive decline, depression)^[36]1. Although levodopa is an
effective symptomatic medical treatment for PD, the chronic use of
levodopa can lead to debilitating motor complications, such as
levodopa-induced dyskinesia (LID). These complications further
complicate the management of PD^[37]3,[38]4. As attempts to develop
disease-modifying therapeutics have been unsuccessful despite extensive
research, substantial attention has shifted toward environmental and
lifestyle factors that may modulate the risk or progression of PD, such
as pesticide exposure, smoking, tea or caffeine consumption, and
diabetes^[39]1,[40]2.
Recently, an increasing body of evidence supports a potential link
between sarcopenia and PD. Sarcopenia is a progressive skeletal muscle
disorder diagnosed by the presence of low handgrip strength (HGS;
probable sarcopenia) and/or low appendicular skeletal muscle mass
(confirmed sarcopenia)^[41]5. Sarcopenia leads to adverse health
outcomes, including frailty and death, and is frequently related to
chronic degenerative diseases, including PD^[42]6. Similar to PD,
sarcopenia occurs as an age-related process influenced by genetic,
environmental, and lifestyle factors across the lifetime^[43]5. Recent
systematic reviews and meta-analyses of observational studies
demonstrated that sarcopenia is more frequent in patients with PD,
compared to healthy elderly adults, and significantly increases the
risk of falls in people with PD, indicating a possible relationship
between sarcopenia and PD^[44]7. Several pathomechanisms proposed for
sarcopenia, including chronic inflammation, autophagy, and oxidative
stress, also contribute to the development of PD^[45]1,[46]5. However,
the causal association between sarcopenia and PD remains unclear, as
the supporting evidence mainly comes from observational studies. These
studies are inherently subject to biases from confounding factors and
reverse causation.
Mendelian randomization (MR) is a statistical framework that uses
genetic variants as instrumental variables (IVs) to serve as a proxy
for the effect of exposure on the outcome of interest^[47]8. MR can
mitigate the impact of confounding and reverse causation because
genetic variants, that are assigned randomly at conception, are not
influenced by confounding and outcomes^[48]9. MR approach has been used
to examine causality between sarcopenia and various diseases, such as
nonalcoholic fatty liver disease, coronary artery disease, type 2
diabetes mellitus, depression, and osteoporosis^[49]10–[50]13. In PD,
two prior MR studies tested the causal links between sarcopenia and the
risk of PD, despite the conflicting results^[51]14,[52]15. Moreover,
causality between sarcopenia and the progression of PD has not been
evaluated through MR approach. It is well-recognized that simple
lifestyle interventions, such as a high-protein diet and resistance
exercise, can improve sarcopenia. Accordingly, clarifying the true
nature of the association between sarcopenia and PD may provide
evidence to justify conducting high-cost, large-scale clinical trials
targeting sarcopenia and may lead to the development of effective
preventive strategies for modifying the risk or progression of PD.
Here, we evaluated the causal inference between sarcopenia-related
traits and the risk of PD using a two-sample MR analysis. We also
examined whether sarcopenia is causally associated with the progression
of PD, including motor disability, cognitive decline, and LID. We also
performed a PGS-based approach and pathway enrichment analysis to
verify significant causal associations identified by MR analyses.
Results
Overall, our MR results demonstrated a significant causal effect of
right-HGS or appendicular lean mass (ALM) on the development of LID
(Fig. [53]1 and Supplementary Fig. [54]1). In the primary
inverse-variance weighted (IVW) MR analysis, genetic predisposition to
a higher right-HGS was causally associated with a lower risk of
developing LID in patients with PD (odds ratio [OR] = 0.152, 95%
confidence interval [CI] = 0.055–0.423, adjusted P = 0.0036; Fig.
[55]1). MR-Egger (OR = 0.016, 95% CI = 0.000–0.572, nominal P = 0.0247)
and weighted median (OR = 0.182, 95% CI = 0.039–0.848, nominal
P = 0.0300) methods also showed a significant causal relationship
between right-HGS and the development of LID (Supplementary Figs. [56]1
and [57]2). Even through IVW MR analysis revealed a causal association
of nominal significance between genetically predicted left-HGS and the
development of LID (OR = 0.316, 95% CI = 0.104–0.959, nominal
P = 0.0420, adjusted P = 0.5042; Fig. [58]1), this association did not
reach the statistical significance using the MR-Egger and weighted
median methods. Genetic predisposition to a higher ALM was causally
related to a lower risk of developing LID on MR analysis (OR = 0.597,
95% CI = 0.440–0.810, adjusted P = 0.0111; Fig. [59]1), whereas a
tendency toward a causal association between ALM and LID development
was demonstrated with the MR-Egger (OR = 0.584, 95% CI = 0.293–1.166,
nominal P = 0.1278) and weighted median (OR = 0.625, 95%
CI = 0.373–1.047, nominal P = 0.0744) methods (Supplementary Figs.
[60]1 and [61]2). Bidirectional MR analysis showed no causal influence
of genetically predicted sarcopenia-related traits on the risk of PD,
progression to dementia, or progression to HY3 (Fig. [62]1 and
Supplementary Fig. [63]1). We also found no genetic evidence of reverse
causation between PD progression and sarcopenia-related traits (Table
[64]1).
Fig. 1. Results of Mendelian randomization analyses between
sarcopenia-related traits and risk or progression of Parkinson disease.
[65]Fig. 1
[66]Open in a new tab
The column labeled ‘P value’ is nominal, and an asterisk (*) indicated
significant causal associations after Bonferroni correction. HY3 Hoehn
and Yahr stage of 3, LID levodopa-induced dyskinesia, NIV number of
instrumental variables, PD Parkinson disease, CI confidence interval.
Table 1.
Results of reverse Mendelian randomization analyses between
sarcopenia-related traits and progression of Parkinson disease
Exposures Outcomes NIVs Method Odds ratio (95% CI) P value
Risk of PD
Right handgrip strength 1 Wald ratio 0.996 (0.965, 1.028) 0.787
Left handgrip strength 1 Wald ratio 0.988 (0.958, 1.020) 0.470
Appendicular lean mass 2 IVW 0.989 (0.973, 1.004) 0.157
Progression to dementia
Right handgrip strength 11 IVW 0.999 (0.997–1.001) 0.1988
Right handgrip strength 11 MR-Egger 0.998 (0.989–1.007) 0.6413
Right handgrip strength 11 Weighted median 0.999 (0.997–1.002) 0.5105
Left handgrip strength 11 IVW 0.998 (0.996–1.000) 0.0640
Left handgrip strength 11 MR-Egger 0.997 (0.988–1.006) 0.4961
Left handgrip strength 11 Weighted median 0.998 (0.996–1.000) 0.1065
Appendicular lean mass 13 IVW 1.001 (0.998–1.003) 0.5792
Appendicular lean mass 13 MR-Egger 1.001 (0.995–1.006) 0.7973
Appendicular lean mass 13 Weighted median 1.000 (0.997–1.003) 0.9684
Progression to HY3
Right handgrip strength 5 IVW 1.000 (0.996–1.005) 0.8224
Right handgrip strength 5 MR-Egger 1.003 (0.997–1.009) 0.3630
Right handgrip strength 5 Weighted median 1.004 (1.000–1.007) 0.0453
Left handgrip strength 5 IVW 1.002 (0.999–1.005) 0.2272
Left handgrip strength 5 MR-Egger 1.004 (1.000–1.008) 0.1682
Left handgrip strength 5 Weighted median 1.004 (1.001–1.008) 0.0082
Appendicular lean mass 6 IVW 1.000 (0.997–1.002) 0.8397
Appendicular lean mass 6 MR-Egger 1.001 (0.996–1.005) 0.8094
Appendicular lean mass 6 Weighted median 1.000 (0.996–1.003) 0.8103
Development of LID
Right handgrip strength 6 IVW 1.000 (0.997–1.003) 0.9793
Right handgrip strength 6 MR-Egger 1.005 (0.995–1.016) 0.3868
Right handgrip strength 6 Weighted median 0.999 (0.995–1.003) 0.6917
Left handgrip strength 6 IVW 0.999 (0.997–1.002) 0.6993
Left handgrip strength 6 MR-Egger 1.005 (0.995–1.014) 0.3841
Left handgrip strength 6 Weighted median 0.998 (0.995–1.002) 0.3854
Appendicular lean mass 7 IVW 1.001 (0.997–1.004) 0.7841
Appendicular lean mass 7 MR-Egger 1.002 (0.990–1.014) 0.7166
Appendicular lean mass 7 Weighted median 1.002 (0.997–1.006) 0.4127
[67]Open in a new tab
CI confidence interval, HY3 Hoehn and Yahr stage of 3, IVW
inverse-variance weighted, LID levodopa-induced dyskinesia, NIVs number
of instrumental variables.
In sensitivity analyses, the zero-intercept test of MR-Egger regression
showed no evidence of horizontal pleiotropy across single-nucleotide
polymorphisms (SNPs) for the three associations (Supplementary Table
[68]5). In addition, there was no heterogeneity in causal estimates
according to MR-Egger or IVW methods using Cochran’s Q statistics
(Supplementary Table [69]5). Leave-one-out sensitivity analysis
confirmed that observed associations were not influenced by single SNPs
(Supplementary Figs. [70]3–[71]5), and funnel plots exhibited symmetry,
suggesting no horizontal pleiotropy (Supplementary Fig. [72]6). We
could not perform complete sensitivity analyses between PD risk and
sarcopenia-related traits because of the insufficient number of IVs
after the stringent IV selection process.
Based on our MR results, we performed PGS-based analysis to test
whether PGS for sarcopenia-related traits was associated with the risk
of developing LID among patients with PD in the Parkinson’s Progression
Markers Initiative (PPMI) database. The occurrence of LID was defined
as the first time the patient reported a score ≥1 in the “Time spent
with dyskinesias” item of the Movement Disorder Society-Unified
Parkinson’s Disease Rating Scale part IV. LID-free survival duration
was defined as the time from treatment initiation to LID occurrence for
patients with an LID event or to the last visit for those without LID.
We used age at PD onset, sex, disease duration, body mass index (BMI),
levodopa equivalent daily dosage (LEDD) at the time of LID or the last
visit, and the top three principal components as covariates.
Multivariable Cox proportional hazard models showed that a higher PGS
for right-HGS (hazard ratio [HR] = 0.225, 95% CI = 0.095–0.530,
adjusted P = 0.0019) and a higher PGS for left-HGS (HR = 0.303, 95%
CI = 0.121–0.759, adjusted P = 0.0323) were significantly associated
with a lower risk of developing LID. An elevated PGS for ALM was
nominally associated with a reduced risk of developing LID (HR = 0.840,
95% CI = 0.708–0.998, nominal P = 0.0472, adjusted P = 0.1416).
Patients with PD were also grouped into quintiles based on the PGS for
each sarcopenia-related trait, and the risk of developing LID was
compared between patients in the highest and lowest quintiles for each
trait. The Kaplan–Meier (KM) analysis revealed that the risk of
developing LID was significantly lower in patients in the highest PGS
quintile for right-HGS, compared to those in the lowest quintile
(adjusted P = 0.0014; Fig. [73]2). The risk of developing LID did not
differ significantly between patients in the highest PGS quintile for
ALM (adjusted P = 1.000) or left-HGS (adjusted P = 0.2693), compared to
those in the lowest quintile for each trait.
Fig. 2. Kaplan–Meier estimates of survival without levodopa-induced
dyskinesia.
Fig. 2
[74]Open in a new tab
The probability of being levodopa-induced dyskinesia (LID)-free was
compared between patients with Parkinson disease (PD) in the highest
polygenic score (PGS) quintile (blue, n = 63) for right handgrip
strength (HGS) and those in the lowest quintile (red, n = 63). The P
value was calculated using the log-rank test and adjusted with the
Bonferroni correction. The displayed rows of numbers indicated the
count of PD patients at risk of LID and the censored observations
during the follow-up period.
Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment
analysis for genome-wide significant SNPs of right-HGS identified
several significantly enriched pathways (Fig. [75]3 and Supplementary
Table [76]6). Among these, the mitogen-activated protein kinase (MAPK)
signaling pathway (adjusted P = 0.0036)^[77]16, cyclic GMP-dependent
protein kinase G (cGMP-PKG) signaling pathway (adjusted
P = 0.0087)^[78]17, long-term depression (LTD, adjusted
P = 0.0157)^[79]18, and long-term potentiation (LTP, adjusted
P = 0.0259)^[80]19,[81]20 have been suggested as contributors to the
pathogenesis of LID via control of synaptic plasticity.
Fig. 3. Pathway enrichment analysis for single-nucleotide polymorphisms
associated with right handgrip strength.
Fig. 3
[82]Open in a new tab
The nodes represent the pathways, with their sizes denoting the number
of genes enriched in the pathway, while their colors corresponding to
adjusted P value using Bonferroni correction. The enrichment represents
the proportion of enriched genes in the given pathway to annotated
genes.
Discussion
The present MR analysis showed that increased right-HGS or ALM
significantly reduced the risk of LID. Furthermore, patients with a
higher PGS for right-HGS had a lower risk of developing LID, possibly
through the control of synaptic plasticity. In contrast, we found no
causal link between genetically determined sarcopenia-related traits
and the risk, motor progression, or cognitive decline of PD.
To our knowledge, this is the first study to identify a significant
causal link between sarcopenia-related traits and development of LID
using MR analysis and a PGS-based approach.
LID is a major motor complication that develops in patients with PD on
long-term levodopa treatment^[83]1. Approximately 40% of patients with
PD experience LID after 4–6 years of levodopa treatment, and its
frequency increases overtime, up to 90% of patients receiving levodopa
for more than 9–15 years^[84]3,[85]4. Because LID limits the
administration of levodopa sufficient to provide maximal
antiparkinsonian effects and nullifies the benefit of levodopa on motor
disability, LID is a devastating complication that deteriorates quality
of life^[86]3,[87]4. Accordingly, it is crucial to identify detrimental
or protective factors for the occurrence of LID and initiate
appropriate strategies to prevent its development. Clinical factors,
such as longer disease duration, younger age at onset, higher levodopa
dose, female sex, and lower BMI, are known to be associated with an
increased risk of developing LID^[88]21,[89]22. The relationship
between sarcopenia and LID has been explored in a few observational
studies. One cross-sectional study demonstrated that HGS, but not
skeletal muscle mass, was lower in PD patients with LID, compared to
those without LID^[90]23. However, other cross-sectional studies
reported that the frequency of LID did not differ between PD patients
with sarcopenia and those without^[91]24,[92]25. A recent longitudinal
study reported that the presence of sarcopenia at baseline, assessed by
temporalis muscle thickness, was unrelated to the risk of developing
LID^[93]26. Nevertheless, these observational findings may not be
sufficient to draw a definitive conclusion due to a small sample size
as well as possible confounding and reverse causation that are
difficult to control in observational studies. In this regard, this MR
analysis provided genetic evidence to support a causal influence of
sarcopenia-related traits on the risk of LID in PD.
Furthermore, the higher PGS for HGS was significantly correlated with a
lower risk of LID after adjusting for clinical factors associated with
LID, including age, sex, disease duration, BMI, and LEDD. HGS is known
to be associated with various disease conditions and prognoses,
irrespective of muscle mass^[94]27. Pathway enrichment analysis may
give an insight into the shared biological mechanisms linking between
HGS and the development of LID. KEGG pathway analysis revealed that
genome-wide significant SNPs for right-HGS were substantially enriched
in several biologic pathways associated with the control of synaptic
plasticity, including LTP, LTD, the MAPK signaling pathway, and the
cGMP-PKG signaling pathway. Although the pathogenesis of LID is not
fully understood, a substantial body of research suggests that aberrant
control of synaptic plasticity plays a central role. A link between LID
and lack of depotentiation, the ability to reverse previously induced
LTP, at corticostriatal synapses is well-described in in vivo and in
vitro models of LID and PD patients with LID^[95]19,[96]20. Similar
changes in synaptic plasticity also occur in basal ganglia output
neurons, globus pallidus internus and substantia nigra pars reticulata,
of PD patients with LID^[97]28. An electrophysiologic study reported
that LID was associated with loss of bidirectional corticostriatal
synaptic plasticity, with only LTD being induced in the indirect
striatal pathway while only LTP being induced in the direct striatal
pathway^[98]18. It is well-established that the MAPK signaling cascade,
that leads to the activation of extracellular signal-regulated
kinases-1 and -2 (ERK1 and ERK2), plays a vital role in the control of
synaptic plasticity. In dyskinetic hemiparkinsonian murine models, the
MAPK signaling pathway was abnormally activated in direct pathway
striatal neurons^[99]16. It also found that Ras-ERK signaling is
required to elicit the LPT induction and depotentiation at the
corticostriatal synapses^[100]29. The nitric oxide/cGMP pathway, that
activates PKG, is involved in the induction of corticostriatal LTD and
was reduced in the dopamine-depleted striatum of dyskinetic
rats^[101]17. An increase of cGMP level by phosphodiesterase inhibitors
restores the reduced LTD and decreases the severity of dyskinesia in
hemiparkinsonian rats treated with levodopa^[102]30. Taken together,
higher HGS may protect against pathophysiologic processes of LID
through the control of synaptic plasticity in PD.
HGS is widely used in clinical and research settings because it is
measured with non-invasive, simple-to-test, and inexpensive tools and
reliably reflects general muscle strength^[103]31. Accordingly,
measuring HGS can facilitate detection and regular monitoring of
patients with PD who are more susceptible to the development of LID.
This may lead to the implementation of preventive strategies for LID
from the early disease stage, including consideration of dopamine
agonist as an initial treatment option, limiting the levodopa dosage,
administration of rotigotine transdermal patch, and adjunctive
treatment with amantadine^[104]32–[105]36. In addition to the
predictive utility, our findings suggest sarcopenia as a modifiable
treatment target for preventing LID, adding the lifestyle
recommendations for management of PD. It is well-recognized that
sarcopenia can be improved by simple lifestyle interventions. These
include oral nutritional therapy, such as a high-protein diet and
supplementation of leucine and vitamin D, in combination with
resistance exercise^[106]37–[107]40. Our data propose that patient with
PD may benefit from the lifestyle interventions for sarcopenia to
reduce the risk of LID.
In this MR study, there was no genetic evidence supporting a causal
effect of sarcopenia-related traits on the risk of PD. Our MR results
are consistent, in part, with the findings of a recent MR study that
examined causal relationships between 401 exposures, including
sarcopenia-related traits, and PD risk^[108]14. On the contrary, a
recent MR study by She et al. reported that low HGS was causally
associated with the risk of PD, while ALM was not^[109]15. These
conflicting findings might be mainly derived from the differences in
GWAS data used in each study. Especially, She et al. used GWAS summary
statistics of HGS that was dichotomized as low or high HGS according to
the cut-off values defined by the European Working Group on Sarcopenia
in Older People or Foundation for the National Institutes of Health. In
both the study by Noyce et al. and the current study, MR analyses were
conducted using GWAS data on continuous values of HGS. Previous
observational studies have suggested a link between sarcopenia and the
risk of PD, mostly based on the presence or absence of sarcopenia that
was determined by diagnostic criteria according to cut-off values of
HGS or ALM^[110]7,[111]23,[112]25. Assumably, MR analysis using
summary-level GWAS data of clinically confirmed sarcopenia, which is
currently unavailable, may provide more definitive results on the
causality with the risk of PD.
A growing body of evidence has suggested that sarcopenia is
significantly associated with cognitive impairment or Alzheimer’s
disease. In addition to the epidemiological data, recent MR studies
demonstrated that genetically determined sarcopenia-related traits were
casually related to AD or lower cognitive performance, indicating that
this association is causal^[113]41,[114]42. Several hypotheses, such as
imbalance of myokine secretion and poor vascular homeostasis, have been
proposed as shared pathomechanisms underlying these
associations^[115]43. In PD, however, a few observational studies
demonstrated inconsistent findings on the relationship between
sarcopenia and cognitive impairment. In our MR analysis, there was no
causal inference between sarcopenia and the progression to dementia in
PD patients. However, current studies only provide limited evidence to
determine the association between sarcopenia and cognitive impairment
in PD, due to insufficient statistical power and differences in study
design and methodologies. Future studies in large-scale datasets are
needed to verify the association between sarcopenia and cognitive
impairment in PD through various statistical methods, including MR.
This study has several limitations. First, we could not determine the
causal effects of sarcopenia defined according to specific diagnostic
cut-off values, despite HGS and ALM being included as major diagnostic
criteria for sarcopenia. This is because the current public database
does not contain GWAS data for sarcopenia confirmed by the diagnostic
algorithm. MR analysis using summary-level GWAS data of clinically
confirmed sarcopenia may yield more discrete results on causal
relationships between sarcopenia and the risk or progression of PD.
Second, potential nonlinear effects could not be explored because of
the use of aggregate data. Third, sex-stratified GWAS data were not
available for sarcopenia-related traits or PD, prohibiting us from
examining potential sex-related differences. Fourth, we could not
exclude the potential influence of survival bias, as sarcopenia may
affect patient mortality before PD is diagnosed. Fifth, the GWAS
databases for sarcopenia-related traits and PD were mostly derived from
individuals of European ancestry. Causal associations between
sarcopenia-related traits and PD risk or outcomes may differ according
to ethnicity. Sixth, given that the MR approach estimates the lifetime
effects of constant exposures on outcomes, the results of MR analysis
using time-varying exposures, such as sarcopenia-related traits, should
be interpreted cautiously^[116]44. Seventh, we could not exclude the
potential reverse causation between PD risk and sarcopenia-related
traits due to the insufficient number of IVs, even though we
additionally performed MR analysis using the Wald ratio method as an
ancillary test. Eighth, as we used GWAS data of PD risk from the
Finnish population, a potential founder effect could not be excluded.
However, the genetic variants associated with PD are largely consistent
between the Finnish population and the general European cohort,
indicating that the founder effect on the genetic architecture of PD is
limited (Supplementary Table [117]7). Finally, our MR study showed that
causal association between left-HGS and the development of LID was
nominally significant, while genetically predicted right-HGS was
significantly associated with LID after multiple testing correction.
This may be due to the moderately limited sample size of GWAS of LID,
which might decrease the statistical power to detect the small effects.
Alternatively, this asymmetric effect of HGS, biased to right-HGS, can
be interpreted that HGS of dominant hand may have a more influential
role in predicting the development of LID in PD. This possibility can
be further explored using GWAS data of HGS of the dominant/nondominant
hand that are currently unavailable.
Our data support the causal influence of sarcopenia-related traits,
specifically increased HGS or ALM, on reducing the risk of developing
LID from a genetic perspective. These findings require further
validation through large-scale longitudinal cohort studies or
randomized controlled trials. Our approach suggests that
sarcopenia-related traits may be easily used, reliable biomarkers for
monitoring the development of LID and may lead to simple lifestyle
modifications to prevent LID in patients with PD.
Methods
Study design
We conducted a two-sample MR study to assess the causal relationship
between sarcopenia-related traits and risk or progression of PD. The MR
analysis was based on three assumptions, as shown in Fig. [118]4. We
used right-HGS, left-HGS, and ALM for sarcopenia-related traits.
Markers of PD progression included progression to Hoehn and Yahr stage
3 (HY3), progression to dementia, and development of LID. We then
calculated PGS of the sarcopenia-related traits using individual
genotype data from the PPMI cohort^[119]45 and conducted survival
analysis for PD outcomes with significant causal associations
identified in the MR study.
Fig. 4. Overview of the two-sample Mendelian randomization analysis and major
assumptions.
[120]Fig. 4
[121]Open in a new tab
The solid lines represent the direct causal associations estimated
using genetic variants as instrumental variables for the factor of
interest. The dashed lines represent potential causal associations
between variables that may violate the MR assumptions. The red crosses
mark the pathways that should be avoided in MR analysis. GWAS
genome-wide association study, HY3 Hoehn and Yahr stage 3, IVW
inverse-variance weighted, LID levodopa-induced dyskinesia, MR
Mendelian randomization, PD Parkinson disease.
Genome-wide association study summary statistics and instrumental variables
selection
GWAS data of sarcopenia-related traits, including right-HGS, left-HGS,
and ALM, were retrieved from the IEU OpenGWAS database (Supplementary
Table [122]1)^[123]46. GWAS for right-HGS and left-HGS was conducted in
461,089 and 461,026 UK Biobank participants (data fields 47 and 46),
respectively, using 9,851,867 SNPs^[124]47. Handgrip strength was
assessed using a calibrated Jamar [125]J00105 hydraulic hand
dynamometer and adjusted for the patient’s hand size^[126]48. The total
observed scale h^2 for right-HGS and left-HGS were 9.95% and 9.96%,
respectively. GWAS for ALM was conducted in 450,243 individuals from
the UK Biobank using 18,071,518 SNPs^[127]49. ALM is the most commonly
used index of muscle mass in the sarcopenia literature. It is
considered more appropriate for quantification of muscle mass than
whole-body lean mass because ALM mitigates the effects of nonadipose
tissue, cardiac/vascular smooth muscle, and the systemic water
component. ALM was estimated using bioelectrical impedance analysis. To
quantify ALM, fat-free masses in the arms and legs were summed and
adjusted for appendicular fat mass, age, square of age, and other
covariates^[128]49. The phenotypic variance explained by the 1059 lead
variants in ALM was 17.8%^[129]49.
GWAS summary statistics for the risk of PD were obtained from FinnGen
release R9, which was derived from a nationwide network of Finnish
biobanks (Supplementary Table [130]1)^[131]50. This PD GWAS consisted
of a total of 20,170,236 SNPs in 4235 PD cases and 373,042 controls
([132]https://r9.finngen.fi/pheno/G6_PARKINSON). Summary statistics for
the progression of PD were obtained from a recently published GWAS
identifying genetic variants associated with 25 cross-sectional and
longitudinal phenotypes of PD. The study included 4,093 patients with
PD recruited from 12 longitudinal cohorts across North America, Europe,
and Australia and followed for a median of 3.81 years (Supplementary
Table [133]1)^[134]51. In the current study, we evaluated three
specific traits for PD progression: progression to HY3, progression to
dementia, and development of LID. The HY scale rates the severity of
Parkinsonian motor disability from stage 1 (HY 1) to stage 5 (HY 5),
with higher stages representing more severe motor symptoms. HY3 is
clinically important because it is the stage when postural instability
begins to appear.
IVs were selected stringently according to the following steps: (1) we
first extracted exposure-related SNPs that reached genome-wide
significance, with a threshold of P value < 5E-8; (2) we then excluded
SNPs with a minor allele frequency <0.01 from consideration; (3) to
obtain independent SNPs associated with the exposure of interest, we
applied linkage disequilibrium clumping (setting R^2 threshold as 0.001
and kilobase pairs as 10,000) and retained SNPs with the most
significant P value; and (4) we harmonized the exposure and outcome
data using the same set of SNPs and removed palindromic SNPs
(Supplementary Tables [135]2–[136]4). We also performed MR analysis
using the Wald ratio method because the reverse causal association
between PD risk and sarcopenia-related traits could not be verified
through IVW, MR-Egger or weighted median methods due to the
insufficient number of IVs following the stringent selection process
for IVs. This method can provide reliable estimates with even a single
IV, enabling investigation of potential causal associations.
Mendelian randomization and sensitivity analyses
The primary MR analysis was performed using the random effect IVW
method^[137]52. The Bonferroni method was applied to correct for
multiple testing. MR-Egger^[138]53 and weighted median^[139]54 methods
were used to complement and enhance the reliability of the MR findings.
The MR-Egger method offers less biased effect estimates in scenarios of
directional pleiotropy and considerable heterogeneity, assuming there
is no measurement error. The weighted median method can provide
consistent effect estimates even if up to 50% of the genetic variants
do not meet the instrument strength independent of direct effect
assumption. We utilized the MR-Egger and weighted median methods as
auxiliary approaches to achieve unbiased results, despite the presence
of some invalid instruments, at the cost of reduced statistical power.
To check the robustness of the MR estimates, we conducted various
sensitivity analyses. Egger intercept analysis was used to test the
horizontal pleiotropic effects, and P value < 0.05 were considered
indicative of possible horizontal pleiotropy^[140]53. MR-Egger and IVW
methods using Cochran’s Q statistics were employed to identify
potential heterogeneity among the SNPs, with P value < 0.05 considered
suggestive of potential heterogeneity^[141]53,[142]55. Leave-one-out
sensitivity analysis was used to assess whether the causal effect was
driven by single SNPs. Significant changes in causal effects after
excluding any specific SNP indicated the presence of
heterogeneity^[143]56. Finally, a funnel plot was used to detect
directional pleiotropy. Asymmetry of the plot indicated strong effects
of certain SNPs on the outcome, despite their low precision, suggesting
the potential presence of pleiotropy^[144]53. These analyses were
conducted using the TwoSampleMR package (version 0.5.6) within the R
software environment (version 4.0.5).
Polygenic score calculation
Quality control procedures were conducted on 619 genotype samples from
the PPMI^[145]45 dataset, which included 267,607 SNPs, at both the
subject and SNP levels. During this process, 13 samples were excluded
due to genotype missingness exceeding 0.05 and 5 samples were removed
due to mismatched sex. In addition, 12,476 SNPs with an overall
missingness >0.05, 190 SNPs not meeting Hardy–Weinberg equilibrium with
a P value < 1E-6, 9 SNPs with a test-mishap P value < 1E-9 and 206,479
SNPs with a minor allele frequency <0.01 were excluded from subsequent
analyses. To assess relatedness among samples, we selected 28,378
linkage disequilibrium–independent SNPs out of the remaining 48,453
SNPs and estimated pairwise identity by descent using the PLINK routine
“--indep 50 5 2”. Five samples exhibiting a heterozygosity rate
deviating more than 4 standard deviations from the mean were excluded.
If the pi-hat value, representing the proportion of shared alleles
between two related samples, was 0.1875 to 0.9, we retained the sample
with a higher genotyping call rate and excluded the other. If the
pi-hat value was >0.9, both related cases were excluded to ensure
integrity of the data. One sample was excluded in this step. Principal
component analysis was employed to estimate genetic structure and
population stratification within PD samples, using GCTA (version
1.94.1)^[146]57. Genotype quality control was conducted using PLINK
(version 1.90b6.21)^[147]58 and in-house scripts. After filtering the
low-quality samples and SNPs, we utilized 595 samples and 46,898
autosomal SNPs for imputation using Minimac4 (version 1.7.3) at the
Michigan online imputation server
([148]https://imputationserver.sph.umich.edu/)^[149]59. Haplotype
Reference Consortium (HRC version r1.1)^[150]60 was used as the
reference panel. Haplotype phasing was estimated using Eagle2 (version
2.3)^[151]61, with a chunk size of 20 Mb. After removing duplicate
SNPs, we obtained 5,580,714 imputed SNPs.
PGS for sarcopenia-related traits was calculated as follows:
[MATH: PGS=∑iNβi
*Gi :MATH]
where
[MATH: βi
:MATH]
was the effect size of SNP
[MATH: i :MATH]
on the trait,
[MATH: Gi
:MATH]
was the number of effect alleles in the sample, and N was the total
number of SNPs used to calculate the PGS. We excluded SNPs from GWAS
that exhibited a minor allele frequency <0.01, as well as those
identified as palindromic or duplicate. A genome-wide significance
threshold of 5E-8 was utilized to filter SNPs, and these were then
overlapped with imputed genotype data from the PPMI. To ensure the SNPs
associated with each exposure were independent, linkage disequilibrium
clumping was applied with an R^2 threshold of 0.1 and a distance of 250
kilobase pairs. Finally, PGS was calculated for 358 patients with PD
from the PPMI dataset with available clinical data, using 2490 SNPs
associated with ALM, 250 SNPs associated with left-HGS and 285 SNPs
associated with right-HGS. PLINK2 (version v2.00a3.7LM) was used to
calculate PGS^[152]58.
Survival analysis
To examine whether PGS for sarcopenia-related traits can predict PD
outcomes that exhibited significant casual associations in the MR
study, we conducted Cox proportional hazards analysis, while
controlling for relevant covariates. To handle observations with tied
survival times, we employed the Breslow method. The Bonferroni method
was employed for multiple testing correction. We further stratified
patients with PD based on each quintile of PGS for sarcopenia-related
traits and conducted KM analysis of the probability of PD outcomes. The
log-rank test was used to compare KM plots between patients in the
highest quintile of PGS and those in the lowest PGS quintile. Cox
proportional hazards and KM survival analyses were performed using the
coxph function in the survival R package (version 3.5.3) and ggsurvplot
function in the survminer R package (version 0.4.9), respectively.
Pathway enrichment analysis
Pathway enrichment analysis was conducted using g:Profiler (version
e109_eg56_p17_1d3191d)^[153]62 which supports employing SNPs as query
to perform enrichment analysis. We utilized genome-wide significant
SNPs, which were previously used in MR and survival analysis
(Supplementary Tables [154]2–[155]4), as inputs for the enrichment
analysis. All known genes within the human genome were considered as
the statistical domain scope. The reference pathways and gene sets were
obtained from the KEGG database^[156]63. To increase the reliability of
our results, reference pathways and gene sets containing either a
single gene or >1000 genes were discarded from the analysis. The
Bonferroni method was applied for multiple testing correction, and
adjusted P value < 0.05 were considered statistically significant.
Pathways that met these criteria were included in our final analysis.
Standard protocol approvals, registrations, and patient consent
For the Mendelian randomization analysis, we used summary statistics of
GWAS from individual studies that were approved by the relevant
institutional review boards. For polygenic score analysis,
individual-level clinical and genotype data were obtained from PPMI,
which was approved by the ethical standards committee on human research
at each participating site. Written informed consent was obtained from
all participants prior to inclusion in the studies. No new ethical
approval was required from the ethics committees for the current
report.
Supplementary information
[157]Supplementary Information^ (2.8MB, pdf)
Acknowledgements