Small ﬁ ber neuropathy in Parkinson's disease: A clinical, pathological and corneal confocal microscopy study Parkinsonism and Related Disorders

Autonomic and somatic denervation is well established in Parkinson's disease (PD). Objectives: (1) To determine whether corneal confocal microscopy (CCM) can non-invasively demon-strate small nerve ﬁ ber damage in PD. (2) To identify relationships between corneal nerve parameters, intraepidermal nerve ﬁ ber density (IENFD) and clinical features of PD. Methods: Twenty-six PD patients and 26 controls underwent CCM of both eyes. 24/26 PD patients and 10/26 controls underwent skin biopsies from the dorsa of both feet. PD patients underwent assessment of parasympathetic function [deep breathing heart rate variability (DB-HRV)], autonomic symptoms [scale for outcomes in Parkinson's disease e autonomic symptoms (SCOPA-AUT)], motor symptoms [UPDRS-III “ ON ” ] and cumulative Levodopa dose. Results: PD patients had signi ﬁ cantly reduced corneal nerve ﬁ ber density (CNFD) with increased corneal nerve branch density (CNBD) and corneal nerve ﬁ ber length (CNFL) compared to controls. CNBD and CNFL but not CNFD correlated inversely with UPDRS-III and SCOPA-AUT. All CCM parameters correlated strongly with DB-HRV. There was no correlation between CCM parameters and disease duration, cu- mulative Levodopa dose or pain. IENFD was signi ﬁ cantly reduced in PD compared to controls and correlated with CNFD and UPDRS-III. However, unlike CCM measures, IENFD correlated with disease duration and cumulative Levodopa dose but not with autonomic dysfunction. Conclusion: CCM identi ﬁ es corneal nerve ﬁ ber pathology, which correlates with autonomic symptoms, parasympathetic de ﬁ cits and motor scores in patients with PD. IENFD is also reduced and correlates with CNFD and motor symptoms but not parasympathetic de ﬁ cits, indicating it detects different aspects of peripheral nerve pathology in PD.


Introduction
Parkinson's disease (PD) is a neurodegenerative disease recognized clinically by its motor symptoms of rest tremor, bradykinesia, rigidity and postural instability. The motor syndrome results from degeneration of dopaminergic neurons in the substantia nigra. Abnormal aggregates of alpha-synuclein in neuronal cells form the histopathological hallmark of PD.
A substantial body of evidence demonstrating peripheral nerve pathology has challenged the traditional view of PD as a disorder of the central nervous system. Phosphorylated alpha-synuclein deposits have been demonstrated in the autonomic nerves of the colon, cardiac plexus and more recently cutaneous C fibers [1e5].
Several studies have shown clinical and histopathological evidence of small and large fiber peripheral neuropathy in PD [6e10]. To date there is no consensus on the underlying pathophysiology of peripheral neuropathy in PD. Cumulative Levodopa exposure has been proposed as a potential cause for large fiber neuropathy via homocysteine accumulation and reduced levels of folate as well as vitamin B12 [6,9,10]. However several studies have found no relation between peripheral denervation and Levodopa exposure suggesting that peripheral nerve involvement may be an intrinsic feature of the disease, especially in relation to small fiber neuropathy [1,7,11]. Furthermore, the identification of alpha-synuclein in cutaneous nerves has led to the consideration of skin tissue as a potential biomarker in PD [2,5].
The cornea receives its innervation from the ophthalmic branch of the trigeminal nerve. Corneal nerves have an important neurotrophic role in maintaining corneal integrity. Small unmyelinated nerve fibers enter the cornea from the periphery and course through the corneal stroma before they penetrate the subepithelial Bowman's membrane, giving rise to the sub-basal nerve plexus [12]. This nerve plexus is of particular relevance for defining neuropathic changes since it supplies the overlying corneal epithelium. We have developed corneal confocal microscopy (CCM) as a rapid, non-invasive technique for in vivo visualization of corneal nerves and shown it to be a valid surrogate measure of diabetic somatic and autonomic neuropathy [13,14]. CCM can also detect early neuropathy in patients with recently diagnosed type 2 diabetes and subjects with impaired glucose tolerance [15,16]. Furthermore, CCM has a high sensitivity in detecting early nerve regeneration following simultaneous pancreas and kidney transplantation in diabetic neuropathy [17]. We have further extended the role of CCM to detect small fiber neuropathy in Fabry's disease, idiopathic small fiber neuropathy and CharcoteMarieeTooth 1A [18e20].
One study that included 4 patients with PD reported no significant reduction in corneal nerve density compared to controls [21]. However, assessment of corneal nerve morphology was not the primary aim of the study and the small number of PD participants precludes meaningful conclusions regarding the utility of CCM in PD.
Given the strong evidence for small fiber neuropathy in PD, we hypothesized that CCM would allow rapid, non-invasive detection of corneal nerve pathology and that this would be related to intraepidermal nerve fiber density and clinical symptoms in PD, in particular those related to pain and autonomic dysfunction.

Subjects
Thirty-three patients (22 males, 11 females) fulfilling the UK Brain Bank criteria for the diagnosis of Parkinson's disease were recruited from neurology clinics. Patients with a known history of cancer, chemotherapy, diabetes, alcoholism, vitamin deficiencies, celiac disease and autoimmune conditions were excluded. All patients underwent an oral glucose tolerance test and fasting blood screening to include: vitamin B12, folate, methylmalonate, and homocysteine to exclude other causes of neuropathy. Six patients (5 males, 1 female) were excluded due to impaired glucose tolerance, leaving 27 patients who fulfilled the inclusion criteria. CCM was not undertaken in 1 patient due to previous refractive eye surgery. Twenty-six age-matched healthy volunteers served as controls. All participants gave their written informed consent.

Disease duration, severity and levodopa exposure
PD duration was calculated from reported symptoms onset until the date of assessment. UPDRS-III was used to assess motor severity. Patients were examined in the "ON" state in an outpatient setting. All patients except one were on Levodopa therapy. Cumulative Levodopa dose was calculated from the date of first prescription until the date of assessment.

Non-motor symptoms
Non-motor symptoms were quantified using the non-motor symptoms scale NMSS [22]. Pain was quantified separately using the recently devised King's PD pain scale [23] as well as the short form McGill Pain Questionnaire (SFMPQ).

Autonomic symptoms and function
Autonomic symptoms were assessed using the scale for outcomes in Parkinson's disease e autonomic symptoms (SCOPA-AUT) [24]. Deep breathing heart rate variability (DB-HRV) provided an estimation of cardiac parasympathetic dysfunction and was assessed with an ANX 3.0 autonomic nervous system devise (ANSAR Medical Technologies Inc., Philadelphia, PA). Cardiovascular sympathetic function was assessed by measuring blood pressure and heart rates in the supine position and after standing up for 5 min.

Evaluation of neuropathy
Nerve conduction studies were performed on all PD patients. Patients with evidence of a large fiber neuropathy underwent assessment of immunoglobulins and protein electrophoresis to rule out the presence of a paraproteinemic neuropathy. The neuropathy disability score (NDS) was obtained in all participants. NDS is a bedside measure to stratify diabetic neuropathy [25]. A clinician assesses vibration perception using a tuning fork, temperature perception using a cold and warm metal rods, pinprick perception using Neurotips™ and ankle reflexes using a tendon hammer producing a score from 0 to 10. Quantitative Sensory Testing was assessed using the MEDOC TSA II (Medoc Ltd., Ramat-Yishai 20095, Israel) device on the dorsum of the left foot. Quantitative sensory testing included heat sensation threshold, heat-induced pain threshold, cold threshold and cold-induced pain threshold. Vibration perception threshold was measured using a Neurothesiometer (Horwell, Scientific Laboratory Supplies, Nottingham, UK) on the hallux of both feet.

Skin biopsies
Two 3-mm punch skin biopsies were taken from the dorsa of both feet. The biopsies were immediately fixed in 4% paraformaldehyde, cryoprotected in graded solutions of sucrose, frozen and cut on a cryomicrotome (HM450, Microm International, Germany). Six 50 mm sections per biopsy were immuno-stained using anti-human PGP 9.5 antibody and nerve fibers were demonstrated using SG chromogen (Vector Laboratories, Peterborough, U.K.). A pathologist blinded to participants' details performed tissue analysis. Intraepidermal nerve fiber density (IENFD), i.e., the number of nerve fibers crossing basement membrane, was quantified according to established criteria and expressed as number per millimetre of epidermal length [26]. The mean between right and left IENFD was calculated for each patient and used for analysis.

Corneal confocal microscopy
Corneal microscopy was performed on both eyes using a Heidelberg Retina Tomograph III with a Rostock Cornea Module (HRT III RCM; Heidelberg Engineering GmbH, Heidelberg, Germany), as previously described [17]. Four to six high-resolution (1e2 mm) images of the sub-basal plexus of each eye were obtained for all participants. A trained investigator who was blinded to participants' details analysed corneal images separately. Corneal Nerve Fiber Density (CNFD): The number of main nerves per square millimetre, Corneal Nerve Branch Density (CNBD): The number of branches emanating from each main nerve per square millimetre and Corneal Nerve Fiber Length (CNFL): The length of all nerve fibers and branches (mm per square millimetre) were quantified and the mean derived from the right and left eye for each parameter. Quantification was undertaken using semi-automated, purposewritten, proprietary software (CCMetrics; M.A. Dabbah, Imaging Science and Biomedical Engineering, Manchester, UK). To estimate the error in measuring CNFD, CNBD and CNFL, we acquired images and determined each of these parameters in 15 subjects on two occasions separated by at least 48 h. The coefficient of variation of these parameters was 12% for CNFD, 24% for CNBD and 9% for CNFL.

Statistical analysis
IBM SPSS version 22 was used to compute the results. All parameters were normally distributed except for cumulative Levodopa dose, vibration and cold perception thresholds as well as heat and cold pain thresholds. Independent samples t-test or Man-neWhitney U test were used to compare means as appropriate. Cohen d was calculated to measure effect size. Two-tailed Pearson's correlation or Spearman's correlations were used as appropriate to determine relationships between continuous variables. Categorical data were compared with Chi square test. P value of <0.05 is considered statistically significant.

Study population
A total of 27 PD patients and 26 controls were enrolled in the study. There was no significant difference in demographics between PD patients and control subjects (Table 1).
Both IENFD and CCM parameters displayed asymmetry between sides in PD patients. However, when CCM measures and IENFD were separated according to the clinically more affected side (defined by UPDRS-III scores) and the clinically less affected side there was no significant difference in CCM measures or IENFD.

Clinical, neurophysiological, quantitative sensory and autonomic measures of neuropathy
NDS, DB-HRV and sensory thresholds all showed impairment in PD compared to controls (Table 1). Overall, there was no significant postural drop in blood pressure in PD (defined as mean systolic drop of >20 mm Hg and/or diastolic drop of >10 mm Hg) ( Table 1). Three out of 27 PD patients (11.1%) had large fiber axonal neuropathy on nerve conduction studies and 2 of these patients had raised

Relation between corneal nerve parameters, intraepidermal nerve fibers and clinical features
CNFD correlated positively with IENFD in PD patients (Pearson's r ¼ 0.464, P ¼ 0.026). Correlations between CCM parameters, IENFD and clinical data are summarized in Table 2. CNBD and CNFL but not CNFD correlated inversely with UPDRS-III. There was no correlation between any of the corneal nerves parameters and disease duration, cumulative Levodopa dose, non-motor symptoms, pain scales or sensory thresholds. However, there was an inverse correlation between heat-induced pain threshold and SFMPQ (Spearman's rho À0.450, P ¼ 0.036). A significant positive correlation was found between DB-HRV and all corneal nerve parameters, but not IENFD (Table 2). However, IENFD correlated with disease duration, disease severity and cumulative Levodopa dose. Both CNFD and IENFD correlated with NDS. CNFD, IENFD and NDS were independent of vitamin B12, folate, methylmalonate and homocysteine levels.

Discussion
This is the first study to systematically characterize corneal nerve changes in PD using the non-invasive technique of corneal confocal microscopy. Compared to controls, PD patients demonstrated a reduction in CNFD, a marked increase in CNBD and an increase in CNFL. This pattern of corneal nerve pathology is notably different from diabetes, Fabry's disease, chronic inflammatory demyelinating polyneuropathy and CharcoteMarie Tooth 1A, where a reduction of all three corneal nerve parameters has been documented [13,18,20,27]. Thus the reduced CNFD with a markedly increased CNBD appears to be unique in PD. Pathologically this represents small fiber neuropathy, characterized by the reduced CNFD. However, the increased CNBD is likely to reflect attempted nerve regeneration. Increased branching of nerve fibers in PD has been previously reported on skin biopsies signifying attempts at nerve regeneration [8]. In our PD patients CNBD correlated inversely with motor severity suggesting that nerve regeneration occurs in the early stages of the disease and declines as the disease becomes more advanced. Indeed, increased corneal nerve branch density was the first measure to increase after simultaneous pancreas and kidney transplantation or continuous subcutaneous insulin infusion in type 1 diabetes [17,28].
We confirm the results of a number of studies, which have shown a reduction of distal IENFD in PD patients [5,7,8]. However, we are the first to report a reduction of CNFD, which correlates with IENFD indicating widespread involvement of small nerve fibers in PD. This is further supported by the significant impairment of modalities relating to both parasympathetic and sensory functions (DB-HRV, thermal perception and pain thresholds). We did not document significant postural drop in blood pressure or heart rate (sympathetic cardiovascular function) in our PD population. Given that postural hypotension is a manifestation of severe sympathetic dysfunction this may reflect the disease stage of most of our PD cohort with relatively few patients in very advanced stages ( Table 1).
The change in corneal nerve parameters was independent of age, cumulative Levodopa dose, methylmalonate, homocysteine, B12 and folate levels, suggesting that damage to corneal nerves is a result of an intrinsic disease process rather than external factors. Furthermore, unlike diabetes where corneal nerve damage was symmetrical [29] we found asymmetry in corneal nerve parameters between sides in PD patients suggesting that corneal neuropathy in diabetes is due to a global metabolic process, whereas in PD may be due to a more localized degenerative process consistent with PD pathology.
Corneal nerve changes strongly correlated with autonomic symptoms (SCOPA-AUT) and parasympathetic dysfunction (DB-HRV), but did not correlate with disease duration or cumulative Levodopa dose. However, IENFD correlated with disease duration, and cumulative Levodopa dose but not with autonomic symptoms or parasympathetic deficits (Table 2). These findings suggest that CCM and skin biopsies may be reflecting different aspects of neuronal pathology in PD and indeed corneal nerves changes rather than IENFD are more closely associated with autonomic pathology. In support of this we have recently shown that CCM has a very high sensitivity and specificity for the diagnosis of diabetic autonomic neuropathy and correlates with autonomic symptoms and deficits [14]. The moderate correlation between IENFD and CNFD emphasizes that skin biopsy and CCM are visualizing different aspect of small fiber pathology. Indeed in a recent study of 134 patients with early diabetic neuropathy the association between IENFD and CCM pathology was even lower [16]. It would be interesting to investigate whether corneal denervation is present during the pre-motor phase when autonomic symptoms are commonly reported which can potentially add to the sensitivity and specificity of pre-motor features such as anosmia and REM sleep behaviour disorder in diagnosing pre-motor PD.
Previous studies of skin biopsies in PD have revealed preferential deposition of phosphorylated alpha-synuclein in autonomic cutaneous nerves, with a higher yield for alpha-synuclein from proximal as opposed to distal sites [2,5]. IENFD reduction, on the other hand, follows a length-dependent pattern [7]. This has led to the suggestion that alpha-synuclein deposition and a lengthdependent reduction in IENFD are different pathological processes [7]. Distal skin denervation in our PD patients was more pronounced than corneal denervation (Fig. 2). Taken together one may conclude that the difference of alpha-synuclein deposition between proximal and distal sites may reflect neuronal availability with distal axons being more vulnerable to degeneration compared to proximal sites.
Similar to previous reports we found otherwise unexplained clinical and neurophysiological evidence of large fiber neuropathy in a proportion of our PD patients but we could not establish a link between levodopa exposure or vitamin deficiencies and large fiber neuropathy. The question of Levodopa induced neuropathy in PD remains a subject of debate. While there have been reports linking large fiber neuropathy with prolonged exposure to Levodopa [6,9], evidence for such a link with small fiber neuropathy is lacking [1,2,7,11]. Although we have found a correlation between IENFD and  cumulative Levodopa dose, establishing causality is difficult as reduced IENFD also correlated with disease duration and severity. Correcting for these factors using multiple regression analysis may address this question, but our study was not powered for this. However, given the lack of relationship between changes in IENFD and methylmalonate, homocysteine or B12 as well as the asymmetry between skin biopsies, our results suggest that cutaneous denervation may be due to an intrinsic disease process rather than an extrinsic toxicity due to Levodopa. Pain is very common in PD, affecting up to 70% of PD patients [30]. We found no correlation between self-reported pain scores and small fiber nerve density suggesting that small fiber pathology may make a lower contribution to pain in PD and that more central mechanisms are involved. This is further supported by the inverse correlation between heat-induced pain threshold and SFMPQ indicating a reduced pain threshold in painful PD phenotypes. Thus pain in PD is more likely to be due to an impaired central processing of nociceptive inputs rather than damaged peripheral nerves.
Both CNBD and CNFL as well as IENFD correlated negatively with UPDRS-III. These findings strengthen the argument that skin biopsy and CCM may be useful biomarkers in PD. Of course our study was not designed to assess the utility of these measures as biomarkers and is limited by the cross sectional design and the assessment of UPDRS-III in the "ON" state, potentially affecting the ability to interpret the motor severity data. Furthermore, skin tissue coimmunostaining with vasoactive intestinal peptide (VIP) or tyrosine hydroxylase would have been required to detect sympathetic adrenergic fibers and differentiate autonomic from sensory nerve involvement. These limitations, however, did not affect our primary objective, which was to assess corneal nerve pathology in PD compared to healthy controls and this was demonstrated with significant effect size.
Corneal confocal microscopy, a rapid non-invasive imaging technique, allows in vivo visualization and quantification of corneal nerves. This enables longitudinal assessment of small nerve fiber pathology, which is not readily achieved using skin biopsies in PD. Further longitudinal studies using UPDRS-III "OFF" scores are needed to establish its relationship to motor deficits. Whether CCM can indeed fulfil the criteria of a non-invasive biomarker to detect sub-clinical disease, predict progression and the effect of treatment in PD, remains to be answered.