Reduced association between dendritic cells and corneal sub‐basal nerve fibers in patients with fibromyalgia syndrome

In our study, we aimed at investigating corneal langerhans cells (LC) in patients with fibromyalgia syndrome (FMS) and small fiber neuropathy (SFN) as potential contributors to corneal small fiber pathology. We enrolled women with FMS (n = 134) and SFN (n = 41) who underwent neurological examination, neurophysiology, prostaglandin analysis in tear fluid, and corneal confocal microscopy (CCM). Data were compared with those of 60 age‐matched female controls. After screening for dry eye disease, corneal LC were counted and sub‐classified as dendritic (dLC) and non‐dendritic (ndLC) cells with or without nerve fiber association. We further analyzed corneal nerve fiber density (CNFD), length (CNFL), and branch density (CNBD). Neurological examination indicated deficits of small fiber function in patients with SFN. Nerve conduction studies were normal in all participants. Dry eye disease was more prevalent in FMS (17%) and SFN (28%) patients than in controls (5%). Tear fluid prostaglandin levels did not differ between FMS patients and controls. While corneal LC density in FMS and SFN patients was not different from controls, there were fewer dLC in association with nerve fibers in FMS and SFN patients than in controls (P < .01 each). Compared to controls, CNFL was lower in FMS and SFN patients (P < .05 each), CNFD was lower only in FMS patients (P < .05), and CNBD was lower only in SFN patients (P < .001). There was no difference in any CCM parameter between patients with and without dry eyes. Our data indicate changes in corneal innervation and LC distribution in FMS and SFN, potentially based on altered LC signaling.

presentation and pathophysiology. 15 FMS is distinct from small fiber neuropathy (SFN) and presents with widespread muscular pain that is regularly associated with depression, fatigue, and sleep disturbance. 16 SFN patients mostly report superficial acral pain without such additional symptoms. While nerve conduction studies are normal to marginally abnormal in FMS and SFN patients, both diseases also share pathological findings in small fiber tests. The impact of small nerve fiber impairment on FMS symptoms remains elusive. Reduction of corneal innervation was found in patients with painful (ie, herpes zoster ophthalmicus 17 ) and painless (ie, inflammatory 18 and diabetic neuropathy, 19 multiple sclerosis 20 ) neurological diseases paralleled by an increase in corneal Langerhans cell (LC) density and LC-nerve fiber contact. Integrity of the corneal sub-basal nerve plexus is regulated by interactions between immune cells and nerve fiber endings. 21,22 Their communication is crucial controlling the corneal immune status. 23,24 A disruption of these fine-tuned neuro-immune interactions may alter corneal innervation. Hence, they are promising targets for pathophysiological mechanisms in small fiber pathology.
In this prospective and controlled study, we investigated corneal immune cells of patients with FMS compared to SFN and healthy controls. We report a reduction of nerve fiber associated LC in patients with FMS compared to healthy controls and even more so in patients with SFN which may contribute to corneal small fiber pathology.

| Study participants
One hundred and thirty-four female patients with FMS and 60 age-and gender-matched healthy controls were examined between September 2014 and August 2018. We prospectively recruited study participants among patients who contacted us for enrolment. All patients were examined by a neurologist and FMS was confirmed applying current diagnostic criteria. [25][26][27] Additionally, we recruited 41 female patients with SFN treated as in-or out-patients at our department. 28 All study participants fulfilled the following inclusion criteria: diagnosis of FMS or SFN [25][26][27][28] and age ≥ 18 years. Exclusion criteria were: history of diabetes mellitus, polyneuropathy, renal insufficiency, untreated thyroid dysfunction, acute inflammatory disease, malignancy ≤5 years, drug or alcohol misuse, severe psychiatric disorder requiring treatment, usage of hard contact lenses, eye diseases or surgery, pain of other cause and undistinguishable from FMS pain, pending compensation claims. Additionally, we documented any history of autoimmune diseases. All patients and controls gave written informed consent. Our study was approved by the ethics committee of the University of Würzburg Medical Faculty (121/14).

| Laboratory and electrophysiological assessment
Serum levels of electrolytes, vitamin B12, thyroid-stimulating hormone (TSH), hepatic and renal marker proteins, and HbA1c were measured. We performed an oral glucose tolerance test (OGTT). Conduction studies of the right sural and tibial nerves were performed in all patients to exclude polyneuropathy. 29  Image size was 384 × 384 pixels with a pixel size of 1.047 μm, thus representing a 400 × 400 μm 2 field of view. Time for image acquisition did not exceed 5 minutes in any case and no participant suffered from corneal or visual complications afterwards. 30 Per patient, six different images of the sub-basal nerve plexus from the central cornea, three from each eye, were chosen by an observer blinded to group allocation, based on image quality, overall contrast, correct focal plane, and absence of artifacts as previously recommended. 31,32 Coded images were assessed offline by a second investigator unaware of the study objectives.

| Image evaluation
We used purpose-written, proprietary image analysis software ACCMetrics and CCMetrics (M.A. Dabbah, Imaging Science, Manchester, UK) to determine the following parameters: corneal nerve fiber density (CNFD, that is, number of main nerve fibers [no./mm 2 ]), nerve fiber length (CNFL, that is, total length of nerve fibers [mm/mm 2 ]), nerve fiber width (CNFW, that is, the average axial diameter of all nerve fibers analyzed [mm]), nerve branch density (CNBD, that is, number of branches arising from the main nerves [no./mm 2 ]), and nerve fiber fractal dimension (CFracDim, that is, measure of spatial distribution and structure complexity of corneal nerve fibers). CNFD, CNFL, CNFW, and CFracDim were automatically determined by ACCMetrics software and CNBD was manually measured using CCMetrics software.
LC were counted manually on the same images used for nerve fiber quantification. We defined all hyperreflective structures showing a cell body as LC. The number of LC per mm 2 is referred to as the total number of cells (LC total ). Based on morphological appearance, 33 LC were subclassified as dendritic cells (dLC, that is, cells showing dendrite-like elongations in addition to their cell body, Figure 1A) and non-dendritic cells (ndLC, that is, cells only consisting of a cell body, In some SFN patients, more than one etiology was assumed, therefore percentages add up to >100%. b Reduced corneal sensitivity was defined as Cochet-Bonnet esthesiometry <5 mm in at least one eye. Figure 1B). For every LC, we determined whether it was in association with nerve fibers (LC fiber assoc. /LC no assoc. ) as described previously. 18 Nerve fiber association was assumed when either the cell body or a dendrite touched a nerve fiber. To assess whether changes in LC fiber assoc. were due to an increased/ decreased chance of LC nerve fiber interaction resulting from changes in CNFL, we also calculated the LC fiber assoc. /CNFL and dLC fiber assoc. /CNFL ratio.

| RESULTS
3.1 | Study population, laboratory, and electrophysiological findings 3.2 | Lower density of dendritic LC in association with nerve fibers in FMS and SFN patients compared to controls Table 2 summarizes findings for LC and the cell subclasses. There were no differences in LC total between groups ( Figure 2A). The number of LC no assoc. did not differ between groups ( Figure 2B).
However, the LC fiber assoc. count was lower in SFN patients compared to controls (P < .01) but did not differ between FMS patients and controls (P > .05; Figure 2C). dLC total density was similar in FMS patients, SFN patients, and controls ( Figure 2D). dLC fiber assoc. were lower in both FMS and SFN patients compared to healthy controls (P < .01 each; Figure 2E). In FMS patients, the LC fiber assoc. / CNFL ratio was as high as in controls. SFN patients presented with lower values than controls (P < .01, data not shown) while the dLC fiber assoc. /CNFL ratio was lower in both patient groups compared to controls (P < .05 each, Figure 2F). There were no intergroup differences for ndLC no assoc. , ndLC fiber assoc. or dLC no assoc.
(data not shown). Figure 3 illustrates CCM fiber findings. CNFL was lower in FMS and SFN patients compared to controls (P < .05 each; Figure 3A). CNFD was lower in FMS patients than in controls (P < .05) but did not differ between SFN patients and controls ( Figure 3B). In contrast, CNBD did not differ between FMS patients and controls but was lower in SFN patients than in controls (P < .001) or FMS patients (P < .001; Figure 3C). There were no intergroup differences for CNFW and CFracDim ( Figure 3D,E).

| LC density does not correlate with age or CNFD
CCM parameters did not correlate with age or disease duration in patients and controls. In addition, dLC fiber assoc. did not correlate with CNFL, CNFD, or CNBD (  Figure 4 shows the distribution of OSDI scores. . Each patient and control group was tested separately for differences between participants with and without DED. DED did not influence CNFL, CNFD, CNBD, CNFW, CFracDim, dLC total , dLC fiber assoc. , or LC fiber assoc. ( Figure S1).

| Tear fluid prostaglandin concentrations do not differ between FMS patients and controls
The obtained volume of tear fluid was sufficient for prostaglandin

| DISCUSSION
We examined corneal LC and sub-basal nerve fibers in FMS and SFN patients compared to healthy controls and report on lower numbers of dLC in association with nerve fibers in FMS and SFN patients. We further show lower CNBD in patients with SFN compared to FMS independent of DED.
We assume that the immune cells observed via CCM are Langerhans cells since corneal epithelial dendritic cells exclusively resemble LC in terms of antigen expression and ultra-structural morphology in healthy human eyes. 40 There was no sign of corneal inflammation in our patients that might have caused migration of other immune cells into the cornea.
Lower CNFD and CNFL counts were reported in small cohorts of patients with FMS when investigated with CCM compared to healthy controls. 9,13,14 Also, reduced nerve fiber branching 9 and thinning of F I G U R E 4 Distribution of OSDI severity among FMS patients, SFN patients, and healthy controls. Distribution of OSDI among healthy controls, A; FMS patients, B; and SFN patients, C. Dry eye disease symptoms assessed by OSDI score were either classified as normal (<13), mildly (13)(14)(15)(16)(17)(18)(19)(20)(21)(22), moderately (23)(24)(25)(26)(27)(28)(29)(30)(31)(32), or severely pathological (>32). Abbreviations: FMS, fibromyalgia syndrome; OSDI, ocular surface disease index; SFN, small fiber neuropathy corneal stromal nerve fibers was found in FMS patients compared to healthy controls. 13 We reproduced these findings for CNFD and CNFL, though did not observe a difference in nerve branching and higher corneal nerve tortuosity 43 than in healthy controls. LC counts did not differ between SFN patients and controls. 43 We showed lower CNFL and corneal nerve fiber branching in SFN patients than in controls, whereas CNFD did not differ. Furthermore, we observed a lower density of LC in association with nerve fibers in SFN patients than in healthy controls but did not detect any difference in total LC counts which matches previous findings. 43 These results may indicate that LC do not change in total number but reduce their interactions with nerve fibers in SFN.
Corneal innervation and immune cell density appear to be altered similarly in FMS and SFN patients as we found lower CNFL and lower counts of dLC with nerve fiber association also in FMS patients. A reduction of LC-nerve associations may indicate a lack of neurotrophic signaling and nerve growth. However, there was no direct correlation between any nerve fiber parameter and dLC fiber assoc. density.
We cannot exclude an influence of low CNFL on the observed LC fiber assoc. and dLC fiber assoc. counts. If there are fewer nerve fibers per mm 2 there may simply be fewer chances for LC to contact a fiber. However, CNFL did not correlate with LC fiber assoc. and dLC fiber assoc. in our study. Also, the ratio of dLCfiber assoc. /CNFL was lower in FMS and SFN patients than in controls. This means that low dLCfiber assoc.
values are not generally found in patients with low CNFL and that there is fewer dLC-nerve fiber co-localizations per mm of nerve fiber found in patients. We conclude from this that dLC fiber assoc. is reduced independently of CNFL in FMS and SFN patients.
The co-occurrence of changes in corneal innervation and corneal immune cell density has previously been shown in patients with chronic inflammatory demyelinating polyneuropathy (CIDP) who had lower CNFD, CNFL, and CNBD compared to healthy controls. The number of immune cell infiltrates was also higher in CIDP patients compared to controls and decreased with disease duration; further, cell counts positively correlated with the degree of motor impairment. 18 In patients with multiple sclerosis, CNFD, CNFL, and CNBD were lower than in healthy controls, whereas the density of dendritic cells was higher in the patient group. 20 Examining patients with diabetes mellitus with and without neuropathy in comparison to healthy controls, lower CNFD, CNFL, and CNBD, and higher nerve fiber tortuosity and LC density were found. 19 These findings were in line with data from mouse models of type 1 diabetes in which an increase of corneal LC density was found to correlate with rising blood glucose levels after induction of diabetes and a gradual decrease in CNFD. 44 Patients with DED of any etiology have a higher LC density in their central cornea than healthy controls, with the highest densities found in autoimmune-associated DED. 45 Tear film concentration of PGE 2 were higher and of PGD 2 lower in patients with dry eyes correlating with symptom severity. 37 We did not find any differences in tear fluid prostaglandin concentrations between those patients with and without DED. This may be due to a selection bias since analysis was only possible in patients with a sufficient amount of tear fluid collected.
There are some limitations to our study: Our assessment included the examination of the corneal sensitivity using Chochet-Bonnet esthesiometry prior to CCM. Although we believe that this should be of minor influence, we cannot exclude a potential effect on corneal LC activation. Next, CCM can indicate contact between LC and nerves, but to prove an ultrastructural interaction, highresolution ex vivo techniques are required. We can only assume that dLC fiber assoc. are interacting with associated nerve fibers. Finally, we enhanced tear flow for tear fluid collection by facial application of mint oil to the study participants, which may have influenced prostaglandin concentrations.
We demonstrate that CCM is a powerful tool to identify even subtle differences in small fiber pathology. Our finding of differences between FMS and SFN patients in CNBD indicates that precise examination of the corneal sub-basal nerve plexus may be instrumental to differentiate between both conditions. We also identify a reduced interaction between dendritic LC and corneal nerve fibers and speculate that reduced neurotrophic support from LC may contribute to nerve fiber loss in patients with FMS and SFN.