Who to Enroll in Parkinson Disease Prevention Trials?

LRRK2

Pathogenic LRRK2 variants account for 5%–6% of familial and 1%–2% of sporadic PD in North American and European populations.¹⁰ Missense variants vary in frequency between populations, with the G2019S variant by far the most frequent pathogenic variant in Western populations. Penetrance of the G2019S variant is incomplete and population dependent.⁹ It is estimated that 26% of Ashkenazi Jewish (AJ) carriers and 43% of non-Ashkenazi mixed Europeans develop PD by 80 years of age.¹⁰ The overall carrier frequency of G2019S is estimated at 0.48%^7,11 and approximately 2% in AJ groups where there is a population founder.¹⁰ However, there is still limited information on the frequency and types of pathogenic variants in populations of Latino and African ancestry.^11,12 G2019S carriers interested in research have been identified through family members of affected patients, direct-to-consumer genetic testing, and more than 300 have participated in initiatives thus far.¹³ Because of reduced penetrance, many carriers will not be prime candidates for primary prevention trials that are designed to assess phenoconversion to PD.

Recruitment of at-risk individuals with more penetrant variants such as I2020T is limited by the low frequency of carriers.⁶ Conversely, the G2385R variant, which is present in up to 10% of East Asian populations, but of lower penetrance, represents one of the largest genetically defined at-risk PD cohorts in the world.¹⁴ PD in G2385R carriers more closely approximates idiopathic PD (iPD) than G2019S-related disease. This presents an interesting conundrum for trial consideration. While the higher frequency of carriers has the advantage of easier enrollment, the potential lower specificity and greater variability of etiology associated with lower penetrance require much larger sample size to demonstrate risk reduction. In addition, there may be a greater contribution of other factors to disease etiology, thereby reducing the benefits of genetic homogeneity.¹⁵ Thus, we will focus further discussion on the G2019S variant.

GBA1

GBA1 variants are present in 5%–10% of PD worldwide and approximately 20% in the AJ^4,16 While homozygous or compound heterozygous variants lead to Gaucher disease (GD), a multisystem lysosomal storage disorder, both biallelic and monoallelic carriers have an increased PD risk.^4,17 GBA1 variants are classified into 3 groups based on the severity of their phenotypic effects: severe, mild, and risk variant. Severe variants are also termed neuronopathic (i.e., L444P), and mild are considered non-neuronopathic (i.e., N370S). Risk variants (i.e., E236K and T369M) do not cause GD in the biallelic state but have been associated with decreased glucocerebrosidase activity and are found at an increased frequency in both PD and dementia with Lewy bodies (DLB).^17,18 However, they are associated with a much lower risk of developing PD. The distribution of the variant type varies depending on the population, with the N370S variant most prevalent in AJ groups and L444P and N370S most frequent of the mild and severe variants in White mixed European samples.¹⁸ While the overall gene frequency is greater than LRRK2, the penetrance of GBA is lower and depends on mutation severity, with an overall estimated 10%–19% of mild and severe variant carriers developing PD by 80 years of age.⁸ More than 400 unaffected carriers have participated in observational studies thus far.¹³ As with LRRK2, selection of at-risk groups will need to consider the breadth of variants included, and most GBA1 risk variant carriers will never develop PD.¹¹

LRRK2

The LRRK2 gene encodes LRRK2, a multidomain protein containing a kinase, GTPase, and protein-protein interaction domains.^2,7 Pathogenic variants are believed to lead to a toxic gain-of-function mechanism resulting in an increased kinase activity generated either directly, by variants in the kinase MAPKKK-like kinase domain (G2019S and I2020T), or indirectly, by variants in the Rab-like ROC (R1441G/C/H) or the adjoining C-terminal of Ras COR domain (Y1699C/G).^2,7 LRRK2 plays an important role in vesicular trafficking and degradation of SNCA and the inflammatory response. Other pathways implicated in LRRK2-mediated neurodegeneration include those related to lysosomal, endosomal, mitochondrial, and autophagosomal functions.

Although not currently being tested in nonmanifesting LRRK2 carriers, major LRRK2-specific therapeutic approaches include blocking LRRK2 kinase activity with selective small molecule inhibitors and brain-specific modulation through intrathecal injection of antisense oligonucleotides (Table).^1,2,4,19 Potential concerns have been raised regarding the safety of systemic kinase inhibition because LRRK2 is not only expressed in the brain but also in immune cells and several other peripheral organs, and kinase inhibitors can vary in selectively.^19,20 Other future considerations include molecules that specifically inhibit variant LRRK2, targeting the LRRK2 Rab-like ROC domain, and more generalized inhibition of reactive oxygen species.¹

While both the effect size and risk profiles are likely greater for agents that directly affect kinase activity or expression, there are fewer ethical and adherence concerns with lower risk interventions for LRRK2-PD primary prevention. A form of vitamin B₁₂, 5′deoxyadenosylcobalamin, has been identified as a novel potential modulator of LRRK2 activity—although it is unclear whether physiologically tenable levels will be achieved.²¹ Epidemiologic studies also suggest promise for other strategies such as regular nonsteroidal anti-inflammatory drug (NSAID) usage,²² higher daily caffeine intake,²³ elevation of antioxidants such as urate,²⁴ and anti-tumor necrosis factor therapy.²⁵ Finally, while not specific to LRRK2, observational data suggest that certain dietary and lifestyle interventions, such as stress management, vitamin D supplementation, regular high-intensity aerobic exercise, and the Mediterranean diet, may be neuroprotective.¹⁵ Commensurate with the lower effect sizes of lower-risk interventions, such trials would require larger sample sizes, but may be amendable to innovative trial designs such as remote/virtual trials, which can enroll larger cohorts.²⁶

GBA1

Building on years of GD research, the GBA pathway is well-studied, with potential pharmacologics for intervention and related biomarkers to demonstrate both pharmacoengagement and disease effect. The GBA1 gene encodes glucocerebrosidase (GCase), a lysosomal enzyme that facilitates the processing of glycosphingolipids.¹⁶ GCase dysfunction leads to the accumulation of the downstream substrates: sphingolipid, glucosylceramide, and glucosylsphingosine.²⁷ Peripheral replacement of GCase and substrate reduction therapies form the cornerstone of GD treatment, and it has been postulated that glucosylsphingosine accumulation in the brain may also mediate PD.¹⁶ Although the pathophysiology of GBA-PD is still being elucidated, at its core is the dysregulation of brain cellular GCase. GCase reduction or dysfunction is postulated to impair lysosomal protein degradation, which has been postulated to lead to toxic accumulation and aggregation of SNCA in a bidirectional feedback loop that promotes further SNCA aggregation. It also reduces mitochondrial function, leads to stress of the endoplasmic reticulum, and activates microglia causing neuroinflammation.

Accordingly, there are a plethora of GBA-specific pharmacologic approaches (though none being tested in nonmanifesting carriers). These include, but are not limited to, increasing gene expression, gene editing, activating mutant GCase, chaperoning misfolded proteins, targeting SNCA aggregation, and inhibiting substrate accumulation (Table).^1,2,4

LRRK2

Some of the highest LRRK2 levels can be found in the blood (Gtexportal.org). LRRK2 and phospho-LRRK2 rapidly reduce in the blood in response to LRRK2 kinase inhibitors, providing a strong target engagement opportunity for LRRK2 inhibitors.^29,30 LRRK2 protein and autophosphorylated LRRK2 protein have been found to be upregulated in both LRRK2 variant carriers and iPD, and urinary autophosphorylated LRRK2 and phospho-Rab decrease with kinase inhibitor treatment in nonhuman primates. Urinary basic metabolic panel levels depend on LRRK2 kinase activity, and isoforms were found to be higher in G2019S carriers with and without PD when compared with nonmutant PD, with the highest levels in disease-manifesting carriers. Finally, LRRK2 protein has been detected in CSF, and autophosphorylated LRRK2 protein may be elevated in CSF. Both total LRRK2 protein and autophosphorylated LRRK2 protein diminish with LRRK2 inhibitor treatment in CSF from nonhuman primates.³⁰ Thus, between blood, urine, and CSF, a comprehensive vantage of target engagement is possible.

GBA

As with LRRK2, there are several promising biochemical outcome measures that demonstrate pharmacodynamic engagement with the GCase-sphingolipid pathway. Of primary interest is GCase and substrate and product glycosphingolipid levels.^27,31 Measures related to SNCA are also relevant as supported by the inverse correlation between serum GCase activity and SNCA in GBA-related PD.³ CSF GCase and SNCA levels have been used as biomarkers for drug engagement with the GCase/SNCA pathway in the investigation of Ambroxol, a proposed pharmacologic chaperone of variant GCase.^1,2,4 Severe variants exhibit the lowest levels of CSF SNCA, which has been correlated with impaired cognitive function and REM behavior disorder (RBD).²⁷ While synuclein imaging is still under development, SNCA oligomers in CSF, skin,³² and other tissues,²⁸ including a novel method using real-time quaking-induced conversion, demonstrate promise in identifying individuals with early SNCA deposition in GBA trials.¹⁹

LRRK2

Unlike RBD cohorts, which may develop to a range of outcomes from PD to DLB to multiple systems atrophy, G2019S variants are predominantly associated with PD.^34,35 While similar to iPD, G2019S-associated PD typically has an earlier AAO, usually 50s–60s, and generally exhibits a less severe phenotype with mildly slower motor progression.^10,34 Early gait involvement is a predominant feature, offering an opportunity for an early biomarker and suggesting gait instability and falls as relevant outcome measures.^3,10 Nonmotor features, especially RBD and cognitive decline, are less frequent in this group.^10,35 Generally, olfactory impairment is less severe, although there are better and worse performing olfactory subgroups, which may guide in the selection of at-risk individuals for post hoc stratification.³⁶

While mutation type helps reduce disease heterogeneity, even among G2019S carriers, phenotypic and pathologic diversity remains.^5,6 Most LRRK2-PD is associated with nigral degeneration and Lewy body (LB) formation, but up to 30% of cases have isolated nigral degeneration without LBs.³ This is particularly relevant because the presence of LBs has been correlated with nonmotor symptoms, including cognitive impairment, anxiety, and dysautonomia, whereas their absence with a predominantly motor phenotype.⁵ Unfortunately, there are currently limited means of separating this group in vivo without biopsy, but potential biomarker proxies include SNCA, as discussed. While these promising methods are currently limited by substantial interindividual and technical variability,²⁸ they have the future potential to stratify subtypes of G2019S disease based on pathophysiology. Regardless, if the therapeutic intervention is upstream of the LB pathology and there is a shared mechanism for the onset of disease, then agents for LRRK2 could work in both pathologic outcomes.

Subtle preclinical findings such as parkinsonian motor signs, gait abnormalities, arm swing asymmetry, and spiral drawing variability have been demonstrated in nonmanifesting LRRK2 carriers (LRRK2-NMC) and may imply a distinct LRRK2-PD prodrome that can be used to identify at-risk patients.^37,38 Various imaging modalities in LRRK2-NMC suggest viable compensatory network changes and imply early synaptic plasticity possibly amendable to disease modification.^35,39

GBA

Compared with LRRK2 and iPD, GBA-related PD tends to have a more rapid progression with more prevalent motor and nonmotor features, especially hyposmia, cognitive impairment, hallucinations, depression, RBD, and dysautonomia.^3,16 Severe GBA variants are associated with a shorter survival and an earlier, more extreme phenotypic expression of motor and nonmotor features.¹⁷ However, severe phenotypic features can occur with mild variants, typically later in disease course.⁴⁰

Cognitive impairment in GBA-related PD demonstrates the greatest deficits in working memory, executive function, and visuospatial processing^16,40 and may be seen before the onset of motor symptoms, thus representing another prospective genotype-specific clinical endpoint. GBA-related PD could also represent a cohort for secondary prevention trials because agents to prevent dementia in PD are sorely needed. For such studies, the ability to predict the course of cognitive decline is useful, and mutation status may serve to stratify patients because severe mutations have the most rapid cognitive decline. A hypothetical power estimate for a 3-year trial on a therapy to halt cognitive decline in PD showed a 25-fold reduction of sample size when restricting enrollment to severe GBA1 variants rather than all-comers with PD.⁴¹

Although some GBA1 risk variant carriers develop PD and others DLB, both are disorders of similar SNCA aggregation, akin to iPD, though with even more widespread cortical LB involvement.³ Thus, an advantage of GBA1 risk variant cohorts is improved neuropathologic homogeneity. Although the mechanism by which carriers develop different phenotypes on the spectrum of PD to PD-dementia to DLB needs further elucidation, it is assumed that putative agents would treat the entire GBA phenotypic spectrum. Because biological sex may affect metabolism, glycosphingolipid expression, and possibly phenotypic expression, trial randomization schema incorporating sex might be considered (Steffany Bennett, PhD, oral communication 5/13/22).

As in LRRK2-PD, evidence suggests there may be a distinct prodrome of GBA-related PD. Retrospectively, patients with GBA-related PD report a shorter and more severe prodromal phase than those with iPD.³⁵ Studies in nonmanifesting GBA carriers (GBA-NMC) report early depression, RBD, subtle motor signs, autonomic dysfunction, olfactory loss, and varying degrees of cognitive impairment, primarily in verbal and executive domains, although some studies indicate that cognitive features may be very mild.^35,40,42,43 It is unclear whether the difference in studies may be attributable to the severity of mutation or other factors.

Solution 1: Incorporate Biomarkers Into Trial Design, Using a Multipronged Approach With Genetic Information Available in the Development of Trial and for Post Hoc Analysis

Other yet to be discovered genetic and epigenetic modifiers such as age, sex, ethnicity, and environmental factors are likely at play, affecting disease onset and development.^9,15,44 Parsing these may aid in identifying enriched cohorts of at-risk individuals and understanding and reducing heterogeneity. Therefore, additional DNA samples should be considered in all trials in genetic cohorts regardless of predominant variant² Information can be used during study design to limit studies to include those with certain modifiers that predict a greater risk and, thus, oversample risk groups in the recruitment schema. It can also be used to allow balanced allocation schema such that individuals are randomized to study arms based on genetic factors. Known genetic information can also be used post hoc for stratification, and/or as new genetic and epigenetic modifiers emerge, having genetic data will facilitate post hoc analyses. In addition, the application of PRSs, which aggregate the effects of multiple protective and risk genetic factors, has the potential to facilitate enrollment decisions.

Solution 2: Overenrollment of Family Members Could Also Increase the Likelihood of Additional Genetic Modifiers and Improve Enrollment

As anticipated, because most cases with PD do not have a known attributable genetic variant, family members of affected patients with PD could not be included in prevention trials in most populations. However, enrolling carriers of known risk variants who additionally have a family history of PD may help address the challenge of reduced penetrance because a positive family history is a known risk factor for the development of PD and specifically is associated with an increased penetrance of GBA1 mutations.⁸ In addition, positive prodromal PD markers may be more frequent, even in variant family members, as seen with abnormal transcranial sonography in nonaffected family members of patients with LRRK2-PD.⁴⁶ Thus, a family history of PD may be used to identify the highest risk individuals to be prioritized, and oversampling could increase the likelihood of including additional modifiers of disease. Family members may also be highly motivated to participate in trials both for their own health and for the potential benefit of others because they have personally experienced the effect of disease and may be compelling advocates for research.

Solution: Consider Restricting Studies to Classes or Specific GBA1 Variants

The tremendous variant heterogeneity in GBA1 may be reduced by focusing on a specific variant class (severe, mild, and variant) and type. Furthermore, clinical trial endpoints may be selected based on variant class because severe variants are associated with a more rapid disease course and earlier AAO by 5 years.¹⁷

However, restricting enrollment to severe variant carriers may improve trial efficiency at the cost of decreasing recruitment because there are fewer eligible participants. Similarly, not all agents will work on all GBA1 variants and the specific variant; not only the class must be considered.² Therapies may differ in their efficacy depending on residual enzyme activity and underlying mechanism. GCase activity can be helpful for stratification and generally correlates to variant severity.³ In addition to class, variant type also affects GCase function differently, with some reducing GCase activity, affecting endoplasmic reticulum quality control or trafficking to the lysosome through Saposin C and LIMP2.^16,27 Therefore, depending on the particular therapeutic and desired outcome, GBA1 trials will need to account for variant type and class.

Solution 1: Develop Risk Profiles by Combining Markers

One strategy is combining individual biomarkers with genetic and epigenetic information to develop risk profiles that can be used to stratify risk and determine proximity to phenoconversion. While epigenetic modifiers have not been definitively demonstrated, a family history in a first-degree relative may increase the likelihood of identifying at-risk carriers. In addition, gene-environment interactors discussed as candidate interventions for LRRK2, such as urate levels, and NSAID and caffeine use^22,-,24 might be included in risk profiles. Furthermore, high urate levels and abnormal DaT imaging may have compounding effects in LRRK2-NMC,²⁴ suggesting a higher risk and possibly closer proximity to phenoconversion. In addition, RBD and hyposmia are uncommon among G2019S carriers, but due to their high predictive value for phenoconversion their presence may indicate a higher risk.^19,35

In addition, the MDS Research Criteria for Prodromal PD could be applied to LRRK2-NMC to address the hurdle of incomplete penetrance to exclude carriers from prevention trials who are unlikely to phenoconvert in a certain time window because it demonstrated a high specificity, sensitivity, and negative predictive value for the diagnosis of PD within 5 years in healthy G2019S carriers.⁴⁸ However, the criterion does not determine proximity to phenoconversion. One solution to address this challenge is to apply PRS and the MDS Research Criteria for Prodromal PD to perform screening for the highest risk carriers and begin enrollment a few years before the median AAO of LRRK2-PD.

Clustering clinical and historical data in GBA-NMC with biofluid and imaging profiles has the potential to discern preclinical/prodromal PD. For example, nonmotor symptoms such as olfaction and RBD may cluster with serum SNCA to denote potential patients on the trajectory for decline.⁴⁹ RBD, which is common in GBA-PD and in itself has a high positive predictive value for the development of disease, also has been associated with cognitive impairment.³⁵ Therefore, in secondary prevention trials aiming to prevent cognitive decline in GBA-PD, RBD may be an important potential eligibility criterion.

Solution 2: Consider Alternative Trial Designs

Adaptive trial designs may also help offset the limitation of a smaller sample size from genetic cohort studies. Trials that allow for inclusion or exclusion based on drug engagement or response might help to improve efficiency. Basket trials may reduce sample sizes by reusing controls. Platform trials that permit the testing of multiple therapies simultaneously may be advantageous because successful neuroprotection may require simultaneous application of multiple agents acting on different targets. Platform trials have demonstrated success in cystic fibrosis when testing was limited to genetic cohorts and have been implemented by the DIAN-TU Next Generation Alzheimer's disease prevention study (ClinicalTrials.gov NCT01760005). There are additional ethical concerns that arise with requirement for knowledge of genetic status from NMCs, and designs to allow participants to participate without requiring knowledge of their own genetic status would be ideal. Modeling after the Trial of Parkinson's And Zoledronic acid, piloting virtual, home-based trial designs is another innovative approach of attaining sufficient sample size for prevention trials and has the additional advantages of convenience, efficiency, and cost-effectiveness.²⁶ Remote trials might be feasible to demonstrate even small effects for lower-risk interventions (i.e., exercise, caffeine, and NSAIDs) when applied to a very large cohort.