We established and applied a protocol for FGFR3 testing from the reference laboratory (center 1) at center 2 according to previously published information only (4), and in a second step adapted the protocol to the details provided by center 1, as shown in Table 1 . For center 1, cut-off values to define an anti-FGFR3-positive result had been determined as previously described from 65 healthy control sera without reported autoimmune disease, neurological disease including neuropathic pain, or malignancy (4). For center 2, these cut-off values were determined using 50 healthy controls whose sera had been recruited during previous studies (10, 11) or routine diagnostic testing.

A schematic illustration of the study design including patient and control cohort sizes is shown in Figure 1 . To screen for anti-FGFR3 antibodies at center 2, we included sera of 60 patients with SFN fulfilling the diagnostic criteria of “definite SFN” proposed bei Devigili et al. (12, 13), namely clinical symptoms of small fiber damage, exclusion of large fiber pathology by sural nerve conduction studies, and either abnormal results in quantitative sensory testing, or reduced intraepidermal nerve fiber density in skin biopsy at the ankle, or both. Further, we included sera of 17 patients with SNN who fulfilled the diagnostic criteria proposed by Camdessanché et al. (14) (n = 11 possible, n = 5 probable). As a control group, we assessed 68 patients with fibromyalgia syndrome (FMS) diagnosed according to the American College of Rheumatology (ACR) 1990 or ACR 2010 classification without further autoimmune disease, polyneuropathy, or pain of other origin as previously reported (15). Of these FMS patients, intraepidermal nerve fiber density was reduced in 33/68 (proximal) or 30/68 (distal) individuals (15).

For the inter-laboratory comparison study, we included all seropositive patient samples from center 2 detected by the screening assay (n = 6), and matched them with equal numbers of seronegative sera, healthy controls (n = 6), and samples from center 1 recruited in previous studies (4, 9). This resulted in a cohort of 36 patients with sensory neuropathies (18 with SFN and 18 with SNN), each fulfilling the respective diagnostic criteria (12, 14), of whom 18 had been considered seropositive and 18 had been considered seronegative. Demographic data of patients included in the inter-laboratory testing are shown in Table 2 . The study was approved by the Ethics Committees of the Universities of Würzburg and St.-Étienne (reference number 220/20 and IRBN742021/CHUSTE, respectively). Patients and controls either gave informed consent or in case they were lost to follow-up, the Ethics Committee of the University of Würzburg or St.-Étienne approved the use of diagnostic serum samples and retrospective analysis of documented clinical data (reference 2021031501).

Patient and control sera were sent to the other center on dry ice and stored at -20°C or -80°C until testing. ELISA using the protocol described in Table 1 was performed three times in independent assays with serum samples from the other center, and evaluated by blinded researchers. Sera which showed three consecutive values above 3 standard deviations (SD) of the means of controls were considered positive. The degree of positivity was assessed semi-quantitatively according to the mean optical density (OD) level (weakly positive if 3–5 SD and strongly positive if > 5 SD above the mean of controls).

To confirm ELISA results for anti-FGFR3 antibodies, 41/42 sera included in the inter-laboratory validation study (n = 1 one serum missing due to insufficient volumes) as well as sera tested positive at center 2 during screening (n = 3 FMS and n = 2 healthy control, see Figure 1 ) were stained on 5-µm cryosections of rat DRG in center 2. After fixation for 10 minutes in acetone at -20°C and blocking with 4% normal goat serum, 4% fetal calf serum, and 0.3% Triton-X-100 in Phosphate-Buffered Saline for one hour at room temperature (RT), serum diluted 1:500 in blocking solution, rabbit Anti-FGFR3 (1:1000, # GTX31549, Genetex, Irvine, CA) and Alexa-647-labeled mouse anti-peripherin (1:1000, # sc-377093, Santa Cruz Biotechnology, Dallas, TX) were incubated overnight at 4°C. After washing, Cy3-labeled anti-human and Alexa-488-labeled anti-rabbit IgG secondary antibodies (1:300, Jackson Immuno Research, Westgrove, PA) were incubated for 1h at RT. Slides were mounted with Vecatshield Mounting Medium with DAPI 1.5 µg/ml (Vector Laboratories, Newark, CA) and assessed using a fluorescence microscope (Zeiss Axiovert 200m, Oberkochen, Germany). Cross-reaction was ruled out by performing single stainings with the respective serum/commercial antibody and by using healthy sera as controls. Staining clusters and colocalization of serum reactivity with commercial anti-FGFR3 and peripherin reactivity were assessed from photomicrographs semiquantitatively by two independent and blinded researchers. For display, identical image modification protocols were applied to photomicrographs of both patient and control sera using ImageJ (16). Further, we aimed to establish cell-based assay (CBA) using plasmids for full-length human FGFR3 and the extracellular domain of FGFR3 as previously described (4, 11), including double-staining with patient sera and commercial Anti-FGFR3 (Genetex), but results were not reliably reproducible for positive sera. Therefore, CBA was not applied as further confirmation assay in this study.

The frequency of positive/negative results between testing via a self-established protocol and the detailed protocol ( Table 1 ) was assessed using Fisher’s exact test. Inter-laboratory concordance of positive/negative results was calculated using Cohen’s Kappa coefficient. OD values between centers were compared using the Mann-Whitney test and Spearman correlation, or t-test depending on normal distribution. DRG binding rates were compared by Chi-square test, and ELISA sensitivity and specificity was calculated by Fisher’s exact test. Testing and display were performed using Prism V9.3.0 (GraphPad Software, San Diego, CA).

First, we established an anti-FGFR3 ELISA at center 2 using previously published information only (4). The ELISA cut-off value from n = 50 healthy controls for a positive result was at an OD₄₄₀ of 0.102 (0.020 + 3x standard deviation of 0.027). Using this self-established protocol, 1/17 (5.9%) patients with SNN and 1/60 (1.7%) patient with SFN were considered seropositive for anti-FGFR3 antibodies, but also 1/50 healthy controls (2%). We then repeated the assay using a very detailed version of the protocol provided by reference center 1, as shown in Table 1 , and including three reproductions of the ELISA. The cut-off value was at an OD₄₄₀ of 0.122 (0.039 + 3x standard deviation of 0.028). We still identified 1/17 seropositive patients with SNN, but now 5/60 patients with SFN (8.3%, p = 0.08 compared to first protocol), and also 2/50 healthy controls (4%, p = 0.04 compared to first protocol). The two positive controls were re-assessed, and one of them had developed malignancy (pulmonary adenocarcinoma) four years after recruitment, the other was tested positive for anti-cortactin-antibodies (17). Moreover, 3/68 (4.4%) of patients with FMS were tested seropositive (15). These three individuals had normal results in quantitative sensory testing, partially reduced intraepidermal nerve fiber density, but normal results in clinical examination, thus not fulfilling the SFN diagnostic criteria (12, 13). Thus, controls with autoimmune/malignant context and patients with FMS can also show positive FGFR3 ELISA results. Furthermore, this direct comparison of two protocol versions shows that small modifications can impact cut-off values and results.

To validate the anti-FGFR3 ELISA assay, we compared the results of 42 sera between two independent laboratories at center 1 and 2, using identical sera and protocols. Cut-off values for positive results were higher in center 1 compared to those in center 2 (OD₄₄₀ 0.371 vs 0.122), as well as mean ODs in general (median = 0.293 vs 0.076, p < 0.0001, Figure 2A ). OD values correlated strongly between center 1 and 2 (r = 0.68, p < 0.0001, Figure 2B ). In total, 34/42 samples (81%) were evaluated in concordance (13 positive, 21 negative, see Figure 2B ). A Cohen’s kappa coefficient of κ = 0.61 indicated a substantial agreement (18) between both tests. Discordance occurred in one healthy control, four SFN and three SNN sera. Seropositive sera which were evaluated in concordance tended to exhibit higher ODs than sera that were rated as positive only in one center (median of 0.655 vs. 0.537 for those rated positive center 1 only, p = 0.350; median of 0.208 vs. 0.137 for those rated positive in center 2 only, p = 0.08), and 7/8 discordant sera were rated only as weakly positive. In conclusion, we could successfully validate the anti-FGFR3 assay in two independent laboratories with high levels of concordance. Still, cut-off values have to be determined separately for every lab, as mean ODs can vary among different settings.

To further validate ELISA results, we performed immunofluorescence binding assays on rat DRG sections with the sera included in the inter-laboratory validation study and FGFR3 seropositive sera from center 2 (including n = 2 positive healthy controls and n = 3 positive FMS sera). Inter-rater comparison revealed almost perfect agreement on semiquantitative assessment and classification of staining cluster regarding the stained cell type (43/46 sera, 93.5%, Cohen’s κ = 0.849). The staining pattern of DRG positive sera included staining of small, peripherin-positive DRG neurons (see Figure 2E ). Triple immunofluorescence with a commercial anti-FGFR3 antibody revealed colocalized binding of serum, anti-FGFR3, and anti-peripherin on small DRG neurons, but no colocalization in anti-FGFR3 negative sera (see Figures 2D, E ). While anti-peripherin and commercial anti-FGFR3 stained the cytosol, staining with patient serum also exhibited membrane localization (see Figure 2E, arrow ).

11/16 (68.8%) ELISA anti-FGFR3-positive sera (confirmed by both centers) and in total 17/26 (65.4%) of single- or double-center positive sera showed binding to DRG small neurons, whereas only 1/20 (5%) sera rated as FGFR3-negative in both centers showed binding to small DRG neurons (p < 0.0001). Out of the 11 anti-FGFR3-positive sera binding small neurons, seven were rated as strongly positive in ELISA in both centers, two as strongly positive in one center (n = 1 center 1 and n = 1 center 2), and two as weakly positive in both centers. Patients with positive DRG neuron staining had higher OD values in the FGFR3 ELISA than patients without specific DRG binding (ELISA at center 1: mean 0.65 vs. 0.23, p < 0.0001; ELISA at center 2: median 0.208 vs. 0.06, p = 0.0011). In general, sera rated as anti-FGFR3-positive in only one center showed higher DRG binding rates than negative sera (3/8, 37.5% vs. 5%, p = 0.026), but a trend towards lower DRG binding-rates compared to double-confirmed positive sera (37.5% vs. 68.8%, p = 0.14). If only double positive (ELISA and DRG staining) were considered as a gold standard for anti-FGFR3 diagnostics, ELISA testing in general (including single-center positive and double-center positive results) showed a sensitivity of 93.3% and a specificity of 61.3%, with a false positive rate of 38.7% (p = 0.0004, Fisher’s exact test). Using DRG staining as a confirmation method, anti-FGFR3 antibodies were not confirmed in FMS sera (0/3 positive), and only in 1/2 healthy control sera tested positive for anti-FGFR3. All other healthy control sera (n = 7) were confirmed negative.

Thus, FGFR3 antibodies can be visualized on small DRG in most patients with high ODs in the anti-FGFR3 ELISA test. The combination of ELISA and binding assays to rat DRG could increase diagnostic specificity for anti-FGFR3 antibodies.