Beyond the pathological findings, a plethora of clinical [18,19,20], physiological [21], and imaging [22] studies support a cerebello-thalamo-cortical model of circuit dysfunction in ET. Postmortem data robustly implicate changes in PCs and surrounding neuronal populations, including basket cells [7]. Aside from ET, the SCAs and other forms of cerebellar degeneration also display varying degrees of cerebellar degeneration on postmortem examination [6]. We know, then, that the pathological features investigated in this study—reduced PC linear density, empty baskets, and basket plexus ratings—are not necessarily a distinct feature of ET. Leveraging a large body of postmortem data across a diverse range of disorders of cerebellar degeneration, as well as a large group of control brains for contextualization, we aimed to further elucidate the relationship between PC loss and basket cell morphologies. Our overarching goal was to understand more fully how PC pathology and basket cell reactive processes converge or diverge across a broad range of cerebellar degenerative diseases.

For these analyses, we focused on three primary metrics: [1] PC linear density, as reduced PC linear density relative to controls is a repeated finding in many studies of ET [2, 9, 23], percentage of empty baskets, as the remnants of basket cell synaptic contacts onto once-present PCs may provide an indirect marker of PC loss [14] and [3], basket plexus ratings, a semi-quantitative measure of the hypertrophy of basket cell axonal plexuses, which is a process hypothesized to occur when basket cell processes converge onto particular PCs, following the death or damage of neighboring PCs—providing another potential secondary marker of PC injury and death [15].

Across all clinical diagnostic groups, we observed statistically significant differences in PC linear density, empty baskets, and basket plexus ratings. Furthermore, there was a consistently significant inverse relationship between PC linear density and empty baskets within each condition (Table 2). The magnitude of this correlation for ET cases was similar to that of controls (Fig. 2b), while FA, SCA3 and SCA-m had correlations of higher magnitude. Correlations of basket plexus rating with either PC linear density or empty baskets were not significant in either ET or controls (Figs. 3b and 4b), even though ET cases differ significantly from controls in all three of these metrics (Table 1). That basket plexus rating is elevated in ET compared to controls but not associated with other markers of PC loss/damage in ET may indicate that basket cell axonal hypertrophy is not driven solely by the extent of PC loss, as will be discussed in greater detail below. Only in SCA-m and FA were basket plexus ratings significantly associated with both percentage of empty baskets and PC linear density (Figs. 3b and 4b; Table 2). Small within-group sample sizes, as demonstrated by the width of several of the CIs (Figs. 3b and 4b), and simple sampling issues (e.g., natural variation) could also have accounted for some of our null results.

In our primary analyses, we combined the majority of SCA subtypes; however, in additional analyses, we presented separable data by each subtype. These groups presented a heterogeneous picture, with differences in PC linear density and basket cell pathology, as might be expected given the known genetic, clinical and pathological heterogeneity across these diseases. In general, the correlation between PC linear density and percentage empty baskets was most robust across groups, as was observed when the groups were combined. The small n (< 10) in all but one of these groups should be noted; hence, any comparisons between groups should be approached with caution.

Understanding the degree to which primary markers of PC degeneration are associated with other markers of degeneration or the remodeling of neighboring neuronal populations has multiple clinical and mechanistic implications. Knowledge of and estimates of the magnitude of the associations add to a growing compendium of disease-specific pathological findings [7] which is useful for differentiating and diagnosing conditions with similar cerebellar involvement at post-mortem. Further, the degree to which these metrics are coupled or decoupled may provide clues to underlying mechanistic processes. For example, our results raise multiple intriguing possibilities: the relatively modest association between PC linear density and percentage of empty baskets in ET compared to other diseases such as SCAs and FA may be suggestive of heterogenous mechanisms driving basket cell remodeling. The genetic and environmental factors that underlie ET are not well established; however, they are not identical to those that underlie these other disorders. It stands to reason, then, that the unfolding downstream cascade of abnormal molecular events in ET will not be identical to that of these disorders. The precise mechanism whereby this cascade eventually results in PC loss in ET is not known. Similarly, it is not known whether the underlying genetic and environmental factors in ET produce similarly unique changes in the biology of the neuronal populations that neighbor PCs in ET. The temporal component of ET may be important as well —since ET cases live longer on average, it may be that basket cells have considerable time to undergo a more dynamic remodeling processes following PC insult. Alternatively, a more rapid rate of PC degeneration, as seen in SCA, might result in more extensive remodeling [24]. What is clear is that the pattern of basket plexus pathology is not uniform across these diseases, and these differences raise important questions about the underlying processes. What processes drive basket plexus hypertrophy and what is its time course? Are empty baskets a permanent feature of a degenerated cerebellar cortex, or do they undergo remodeling and eventual dissolution over time?

We present a model whereby basket plexus hypertrophy may arise from recruitment and remodeling of basket cell axonal processes following degeneration or loss of neighboring PCs. However, as noted above, some of our data do not support this model; e.g., in this sample of ET cases, and in SCA3, basket plexus ratings were not significantly correlated with either PC linear density or empty basket percentage. These findings raise the possibility that basket cell remodeling may be regulated by mechanisms beyond simple PC loss. Alternative biological mechanisms for formation of empty baskets could include alterations in GABAergic neurotransmission, neuroinflammatory processes, dysregulation of the SEMA3A/NRP1 signaling pathway and/or activity-dependent structural plasticity within cerebellar microcircuits.

Much is known about the molecular basis of basket plexus formation from a developmental perspective. Studies of mouse mutants with targeted deletions of the axon guidance molecule Semaphorin 3 A (SEMA3A, secreted by PCs) [25] or its receptor Neuropilin-1(NRP-1) [26] demonstrate that PC–derived SEMA3A initially attracts and orients NRP-1–expressing basket cell axons towards their targets. NRP-1 then mediates subcellular target recognition at the PC axon initial segment by trans-synaptic interaction with neurofascin-186, a L1 immunoglobulin family cell adhesion molecule that is required for the formation of the pinceau and is expressed by both PCs and basket cells. Notably, inactivation of neurofascin-186 in adult PCs destabilizes the axon initial segment and pinceau structure, potentially implicating this molecule in neurodegenerative processes within the cerebellum [27,28,29]. Basket cell pinceau size also varies with cerebellar zonal organization, being larger around PCs with higher firing rates (e.g., aldolase C-negative PCs); however, this size differential depended on PC neurotransmission rather than basket cell GABAergic input onto PCs, pointing to a target cell–driven mechanism of structural plasticity [30]. Nonetheless, early hyperactivity of molecular layer interneurons potentiates PC degeneration in SCA1 mice [31], indicating a complex interplay between PC and interneuron activity in both normal physiology and disease progression. Whether these developmental and organizational processes persist across the lifespan — and whether their disruption may relate to the pathological findings observed in this study — remains an important question for future investigation [25, 30].

Some of the metrics we study in detail here have been examined in prior work, including assessments of their associations. In an earlier study of 32 ET cases and 21 controls, there was a marginal inverse association between PC loss and basket plexus rating (p = 0.06) [15], and in another study of 127 brains (controls, ET, SCA3, SCA-m), PC linear density, % empty baskets and basket plexus rating were all correlated with one another, although the analyses did not stratify by diagnosis as we have done here [14].

We recognize certain limitations to this study. First, although the number of brains was large, compared to ET and controls, there were relatively fewer cases of SCA-m, SCA3, and FA. This may have made it more difficult to test associations in our correlation analyses; however, despite this, robust correlations were detected for SCA-m and FA (Figs. 2B, 3B and 4B) and even for SCA3 (Fig. 2B). Second, the basket cell rating is a semi-quantitative ordinal scale; thus, compared to the other two continuous measures utilized in the study, presents certain limitations. Ordinal measures provide limited resolution compared to continuous measures, given the lack of equivalence between unit measurements. In addition, our studies have demonstrated that the basket plexus rating varied considerably even within a single slide, meaning that the assigned rating is an overall rating of a heterogenous and variable process [32]. This creates statistical noise. Third, while a prior study in ET demonstrated an increased number of pinceau processes that abnormally targeted PC axon segments more distal to the axon initial segment [33], we have not examined this morphologic change in the large cohort of brains in this study. Fourth, we restricted our analysis to a particular region of the hemispheric cerebellar cortex but encourage further exploration of other regions. Fifth, we did not exclude cases based on extra-cerebellar pathologies such as Lewy pathology or Alzheimer’s type changes. Previously, it has been shown that there is not an association between these metrics of PC loss in ET cases and the presence and/or severity of any accompanying Alzheimer’s type neuropathological changes or Lewy body pathology [5, 32]. The study had numerous strengths including: [1] a large number (> 300) of brains [2], inclusion of multiple cerebellar degenerative diseases [3], inclusion of controls of varied ages, and [4] the fact that the research question has not been studied to any great degree in ET, in controls or in the setting of other cerebellar degenerative diseases.