TrkA abundance is increased in cutaneous nerves in bortezomib-induced neuropathy.

INTRODUCTION

Multiple myeloma (MM) is the second most common hematological malignancy, and bortezomib (BTZ), a first‐line treatment for MM, reversibly inhibits the threonine site of the 26S proteasome, leading to tumor cell cycle arrest and apoptosis [ 1 ]. Despite efforts to mitigate the BTZ side effects through combination therapies and dose adjustments [ 2 , 3 , 4 ], BTZ‐induced peripheral neuropathy (BIPN) remains a significant clinical challenge. Among patients experiencing peripheral neuropathy, 13%–36% report neuropathic pain, which typically presents in a stocking‐and‐glove distribution and severely impairs quality of life [ 5 , 6 ]. Notably, in some cases, neuropathic pain persists even after discontinuation of BTZ treatment [ 6 ]. The lack of reliable biomarkers for early prognosis and risk stratification hampers timely intervention and personalized treatment.

Neuropathic pain involves fundamental mechanisms in which the tropomyosin receptor kinase A (TrkA), a high‐affinity receptor for nerve growth factor (NGF), plays a key role [ 7 ]. In animal models of peripheral neuropathy, elevated protein levels of NGF, TrkA, and phosphorylated TrkA (p‐TrkA) have been observed in dorsal root ganglia (DRG) [ 8 ], with further activation of the p38/MAPK pathway [ 9 ]. Meanwhile, NGF/TrkA signaling enhances the channel activity of transient receptor potential vanilloid 1 (TRPV1) and promotes the release of calcitonin gene‐related peptide (CGRP) [ 8 , 10 ] in chemotherapy‐induced peripheral neuropathy (CIPN) models via the PI3K/Akt [ 11 ] and PKC/NF‐kB [ 10 ] pathways, respectively.

Beyond neuronal sensitization, the NGF/TrkA signaling may modulate vascular support mechanisms essential for nerve repair. In CIPN models, impaired vasa nervorum perfusion and density [ 12 , 13 ] correlate with neuropathy, while Vascular Endothelial Growth Factor (VEGF) therapy mitigates these deficits [ 12 ]. Notably, NGF/TrkA signaling induces VEGF release via the Akt‐NO axis [ 14 ], suggesting a dual role in both pain pathogenesis and vascular regeneration.

Based on these findings, we hypothesized that dysregulation of NGF‐TrkA signaling is associated with BIPN and investigated relevant biomarkers in skin biopsies from patients and controls to assess its potential role.

MATERIALS AND METHODS

Study population

The study included 50 patients diagnosed with BIPN, recruited from the University Hospital Würzburg between July 2021 and November 2024. All participants had a confirmed diagnosis of MM based on the International Myeloma Working Group guidelines [ 15 ] and were either currently undergoing ( n = 29) or had previously completed ( n = 21) treatment with BTZ.

Peripheral neuropathy was identified through abnormal findings in at least one of the following assessments: sensory or motor function evaluations, quantitative sensory testing (QST), or electrophysiological studies. Exclusion criteria were other serious medical conditions that could preclude participation in the study. Among the participants, 19 reported experiencing neuropathic pain as defined by a grading system [ 16 ].

A control group of 27 healthy individuals was recruited from the local community through online advertisements. These participants were screened to ensure they had no history of neurological disorders or conditions that could affect peripheral nerve function.

The study protocol received approval from the Ethics Committee of the Medical Faculty at the University of Würzburg (#98/20) and adhered to the ethical principles outlined in the Declaration of Helsinki. Written informed consent was obtained from all participants prior to their inclusion. Furthermore, the study was prospectively registered in the German Clinical Trials Register (DRKS00025196), ensuring compliance with the registry's disclosure requirements before the commencement of the research.

Function of sensory nerves

Sensory nerve function was assessed using both electrophysiological studies and QST. Neurophysiological assessments were conducted on the left sural nerve and sensory median nerve to evaluate large fiber function. Sensory nerve action potentials (SNAP) amplitude and sensory nerve conduction velocity (NCV) were recorded following standard methods [ 17 ].

QST was performed on the dorsum of the right foot of patients following the standardized protocol established by the German Research Network on Neuropathic Pain (DFNS) [ 18 ]. This protocol evaluates a comprehensive range of sensory parameters, including thermal thresholds (cold detection threshold [CDT], warm detection threshold [WDT], thermal sensory limen [TSL]), pain thresholds (cold pain threshold [CPT], heat pain threshold [HPT]), mechanical thresholds (mechanical detection threshold [MDT], mechanical pain threshold [MPT]), sensitivity to mechanical pain stimuli (mechanical pain sensitivity [MPS], wind‐up ratio [WUR]), pressure pain threshold (PPT), and vibration detection threshold (VDT).

For the analysis, data for CDT, WDT, TSL, PPT, MPT, MPS, WUR, MDT, and VDT were subjected to log10 transformation. Additionally, to account for potential confounding factors such as age and sex, all data were normalized using z‐transformation. The z‐score was calculated as follows: z‐score = (X single subject − Mean published controls)/SD published controls, referencing published control values [ 19 ]. For ease of interpretation, the algebraic sign of z‐score was adjusted so that positive z‐score indicated gain‐of‐function compared to controls, while negative z‐score reflected reduced loss‐of‐function [ 20 ].

Skin biopsy and immunofluorescence staining

Skin biopsies were obtained from the distal leg of 50 patients and 27 healthy controls using a disposable biopsy punch under local anesthesia with lidocaine. The collected tissue samples were immediately fixed in freshly prepared 4% buffered paraformaldehyde for 30 min. Following fixation, the samples were rinsed in 0.1 M phosphate buffer and subsequently transferred to a 10% sucrose solution in phosphate buffer. For preservation, the tissues were snap‐frozen in isopentane chilled with liquid nitrogen. The samples were embedded on cork plates using Tissue‐Tec® O.C.T. compound (Sakura Finetek GmbH, Freiburg, Germany).

Using a cryostat, sections of 20 μm thickness were prepared. For blocking, sections were incubated in 10% BSA‐PBS for 30 min, in a humid chamber. The primary antibodies were diluted in a solution of 1% BSA/PBS with 0.3% Triton X‐100 (Table S1 ) and were applied overnight, at 4°C, in a humid chamber. After washing three times with PBS, sections were incubated with secondary antibodies in 1% BSA/PBS for 2 h. After washing with PBS, the samples were embedded in antifade mounting medium containing DAPI (Vector Laboratories, CA, USA) and sealed with CoverGrip. Slides were stored at 4°C.

The sensitivity and specificity of the TrkA and GAP43 antibodies were evaluated (Figures S1 and S2 ). Detailed methods for primary antibody selection, including sensitivity and specificity assessments, were based on our previous study [ 21 ]. For testing anti‐TrkA antibodies, we also used recombinant TrkA expression. For this, Human Embryonic Kidney‐293 (HEK293) cells were transfected with a vector (pcDNA3) expressing the complete coding region of TrkA [ 22 ].

Fluorescence microscopy

Images (16‐bit) were acquired with a Leica DMi8 epifluorescence microscope, equipped with Lumencor's Bright Engine, and combined with a Thunder Imaging System (Leica Microsystems, Wetzlar, Germany). An HC PL FLUOTAR 20×/0.55 NA objective was used, and images were acquired with an x‐y resolution of 610 nm/pixel and a step size of 1800 nm. Image post‐processing included computational clearing using the Leica LAS X software and subsequent maximum intensity projection.

Confocal images (x, y, z) were acquired at 12‐bit using an Olympus FV1000 confocal laser scanning system, a FVD10 SPD spectral detector, and an Olympus UPLSAPO 60×/1.35 NA oil objective, at x‐y resolution of 98 nm/pixel and a step size of 500 nm in z.

Quantitative analysis

Quantitative analysis for fluorescence intensity

Two fluorescence images per patient were captured along the basement membrane, moving sequentially from left to right to avoid any overlap between fields. Each image included the epidermis, and a downward extension of the epidermis to ensure the inclusion of nerve fibers (Figure S5A ). For unbiased segmentation of the epidermis, dermis, nerves, and blood vessels, a deep learning‐based segmentation tool optimized for bioimage analysis was used [ 23 , 24 ], as recently described [ 21 ]. The subepidermis was then extracted by computationally expanding the epidermis by 50 μm and overlapping it with the dermis (Figure S5B ). Quantitative analysis of TrkA and GAP43 labels was restricted to nerve fiber segmentations within ROI.

Quantification of intraepidermal nerve fiber density ( IENFD )

IENFD was initially assessed using a manual counting method [ 25 ]. In this approach, all Protein Gene Product 9.5 (PGP9.5) immunoreactive fibers crossing the basement membrane in a section were counted (three sections per patient), and the counts per millimeter length of the tissue section, denoted as IENFD PGP9.5 . Similarly, nerve fibers labeled with CGRP were denoted as IENFD CGRP . The manual count served as the reference for subsequent automated quantification.

For automated quantification, nerve fibers and the epidermis were segmented using a deep learning‐based algorithm [ 23 , 24 ], as described above. The intersections between segmented nerve fibers and the epidermal outline were identified to mimic the manual counting process.

Quantification for vascular‐nerve fiber interaction

For investigating blood vessels close to nerve fibers, both structures were segmented using deepflash2. To assess potential interactions, a nerve fiber proximity region was defined as the area within a 5 μm radius surrounding the nerve fibers. Blood vessels located within this region were considered to be in close spatial proximity to the nerve fibers.

RNA isolation, cDNA synthesis and qPCR

Total RNA was extracted from skin biopsies using the Qiagen miRNeasy Microkit (Cat# 217084, Qiagen, Hilden, Germany). For cDNA synthesis, 250 ng RNA was reverse transcribed in a volume of 100 μL. For qPCR, the following primers were used: (TrkA: Hs00176787_m1; NGF: Hs00171458_m1, Thermo Fisher Scientific) and TaqMan probes (Cat# 4444557, Thermo Fisher Scientific, Vilnius, Lithuania). RPL13A primers (Hs04194366_g1) were used as the reference housekeeping gene. Reactions were run in MicroAmp Fast 96‐well reaction plates (Cat# 4346907, Applied Biosystems, Darmstadt, Germany). Subsequently, 3.5 μL cDNA samples were run in triplicates using a StepOnePlus Real‐Time Cycler (Cat# 4376598, Life Technologies, Darmstadt, Germany). The cyclin protocol was: initial incubation at 50°C for 2 min, followed by a second incubation at 95°C for 2 min, and then 40 cycles of 95°C for 3 seconds and 60°C for 30 s. Data evaluation was performed using the 2 −ΔΔCT method.

Statistical analysis

SPSS Statistics 26 software (IBM, Ehningen, Germany) was used to statistical analyze. Shapiro–Wilk test was used to test normality of each group. Descriptive statistics of the variables were analyzed and presented as mean with standard deviation (SD, normality distribution) or median with interquartile range (IQR = Q3 − Q1, non‐normality distribution).

For comparisons between controls and patients, as well as between the painful and painless subgroup of BIPN, the T‐test was applied for variables with normal distribution, while the Mann–Whitney U test (a non‐parametric test) was used for non‐normally distributed variables. Categorical variables, such as sex distribution among patients, were analyzed using the Chi‐square (χ 2 ) test. Correlations between two continuous variables were assessed using the Spearman rank correlation test. All statistical tests were two‐tailed, and significance was set at p < 0.05.

RESULTS

Demographic, neurophysiological, and QST data of patients

Demographic and clinical characteristics

The study enrolled 50 patients (mean age: 64.1 ± 9.2 years; 76% male), who were divided into two subgroups: 31 patients without pain and 19 patients with pain. As a reference group, 27 healthy volunteers were also included. Demographic and clinical characteristics are summarized in Table 1 . Body mass index (BMI) data for the control group were not available.

The numerical rating scale (NRS) based on the maximum rating in the last week without pain medicine was categorized as follows: no pain (0), n = 31, mild (1–3), n = 4, moderate (4–6), n = 4, and severe (≥7), n = 11. For patients who had taken pain medication, the NRS was adjusted to reflect the corrected pain level by incorporating the reported relief from medication (Table S4 ).

Statistical analysis revealed no significant differences in age and sex between healthy volunteers and all BIPN patients. Similarly, no significant differences were found between the painful and painless subgroups in terms of age, sex, BTZ treatment cycles, or cumulative dose. However, BMI was significantly higher in the painful subgroup compared to the painless subgroup ( p = 0.017; Table 1 ).

Nerve conduction study ( NCS ) of sensory nerves

The nerve conduction parameters (see Section 2 ) were evaluated against population normative values (95% confidence intervals, CI); values falling outside this range were considered abnormal.

SNAP amplitudes of the sural nerve were abnormal in 63.3% of evaluable patients with BIPN (31/49; data missing for 1 patient). The painful BIPN subgroup showed a significantly higher rate of impairment (84.2% [16/19]) compared to the painless subgroup (50% [15/30]; χ 2 = 4.48, p = 0.034, OR = 5.33, 95% CI: 1.38–19.45) (Figure 1A ). The painful subgroup exhibited more severe reductions in sural SNAP amplitudes (median: 3.0 μV, IQR: 0–4.5 μV) compared to the painless BIPN (median: 6.65 μV, IQR: 2–12.45 μV) (Figure 1B ).

Median nerve SNAP amplitudes were abnormal in 20.48% of the BIPN patients (10/49; 1 missing), with abnormal values observed in 26.3% [5/19] of the painful subgroup and 16.7% [5/30] in the painless subgroup (not significant, n.s.) (Figure 1A ). However, the painful subgroup showed more pronounced reductions in SNAP amplitudes (median: 15.3 μV, IQR: 6.2–18.3 μV) compared to the painless subgroup (median: 21.8 μV, IQR: 13.1–27.5 μV) (Figure 1B ).

Abnormal nerve conduction velocity (NCV) was observed in 24.4% of evaluable patients for the sural nerve (11/45; 5 missing) with rates of 29.4% [5/17] in the painful subgroup and 21.4% [6/28] in the painless subgroup.

For the median nerve, abnormal NCV was found in 8.2% of patients (4/49; 1 missing) including 10.5% [2/19] in the painful subgroup and 6.7% [2/30] in the painless subgroup. No significant differences were found between subgroups for either nerve (Figure 1A ). Similarly, absolute NCV measurements did not differ significantly between subgroups for either the sural or median nerve (Figure 1B ).

Quantitative sensory testing ( QST )

QST parameters were evaluated against population normative values (95% CI) stratified by age and sex [ 19 ], with values outside this range considered abnormal (including sensitivity and insensitivity).

Among BIPN patients, abnormal rates for QST parameters were as follows: CDT (36% [18/50]), WDT (16% [8/50]), TSL (18% [9/50]), CPT (0% [0/50]), HPT (44% [22/50]), MDT (30% [15/50]), MPT (18% [9/50]), MPS (28% [14/50]), WUR (12% [6/50]), VDT (30% [15/50]), and PPT (4% [2/50]) (Figure 1C ). Patients with painful BIPN exhibited significantly higher abnormal values than BIPN patients of the painless subgroup with respect to CDT (57.9% [11/19] in comparison to 22.6% [7/31], p < 0.05) and MDT (52.6% [10/19] in comparison to 16.1% [5/31], p < 0.01) (Figure 1C ).

Z‐score profiles revealed greater loss of function in painful BIPN (vs. painless) in CDT ( p < 0.05; 0.018; −2.31 [−2.7 to −1.48] vs. −0.56 [−1.84 to 0.16]), WDT ( p < 0.05; −1.81 [−1.91 to −1.57] vs. −1.46 [−1.89 to −0.93]), TSL ( p < 0.001; −1.88 [−2.11 to −1.52] vs. −1.13 [−1.56 to −0.75]), HPT ( p < 0.01; −1.46 [−1.52 to −1.38] vs. −1.15 [−1.46 to −0.40]), and WUR ( p < 0.01; −0.99 [−1.47 to −0.63] vs. −0.39 [−0.81 to 0.13]) (Figure 1D ).

These findings suggest that BTZ predominantly affects small nerve fibers (C fibers and Aδ fibers) rather than medium or large nerve fibers (Aβ fibers) in patients with multiple myeloma.

Intraepidermal nerve fiber density in patients with BIPN

Given that electrophysiological data indicated a small nerve fiber phenotype, we investigated whether intraepidermal nerve fiber density (IENFD) is altered in patients with BIPN.

To quantify IENFD, we employed deep learning (DL)‐based nerve fiber annotation using anti‐PGP9.5 immunolabeling (IENFD PGP9.5 ) in bioimages of 20‐μm skin sections. To compare this with the standardized manual annotation method [ 25 ], which is typically conducted on 50‐μm sections, we assessed the correlation between automated and heuristic nerve fiber counts. DL‐based counts showed a significant positive correlation with manual counts ( r = 0.35, 95% CI: 0.121–0.545, p < 0.01) (Figure 2B ). As anticipated, automated counts from 20‐μm sections were lower than manual counts from 50‐μm sections, likely due to differences in section thickness. In the same set of experiments, we also performed DL‐model based annotation of CGRP+ nerve fibers (Figure 2A ).

In BIPN patients, we found a decrease in IENFD PGP9.5 compared to controls. When the IENFD was analyzed based on anti‐CGRP immunofluorescence labeling, IENFD was not different between BIPN patients and controls. It has to be noted that CGRP‐immunoreactive fibers were very rare in the epidermis (Figure 2C , Table 2 ).

CGRP immunoreactivity in human skin of patients with BIPN

In the epidermis, the total area occupied by CGRP+ nerve fibers was small and did not differ significantly among controls and patients with painless or painful BIPN (Figure 2D ). The proportion of CGRP + area relative to the total nerve fiber area (PGP9.5 + ) was similarly low in both groups (controls: 2.3%; BIPN: 3.6%) (Figure 2E ).

CGRP mean fluorescence intensity (MFI) was not significantly different between controls and BIPN patients (controls: median 1658, IQR 1314–2195; BIPN: median 2386, IQR 1372–3593; p > 0.05) (Figure 2F ).

Likewise, normalized CGRP MFI—calculated as the ratio of CGRP to PGP9.5 MFI—did not differ significantly between groups (controls: median 0.755, IQR 0–1.205; BIPN: median 0.787, IQR 0–1.715; p > 0.05) (Figure 2G ).

In the sub‐epidermis, the CGRP+ nerve fiber area was 40‐fold higher compared to the epidermis, but was not different between controls and BIPN patients (Figure 2H ). However, the proportion of the CGRP+ area to total nerve fiber area was 1.5 times higher in BIPN compared to controls (18.3% vs. 12.6%, p < 0.05) (Figure 2I ). CGRP immunofluorescence intensity values did not differ between controls and BIPN patients (median: 2583, IQR: 1722–3058 vs. median: 2685, IQR: 1954–3442, p > 0.05) (Figure 2J ). Moreover, when normalized to PGP9.5 MFI, the normalized CGRP MFI was not significantly different between BIPN patients and controls (Figure 2K ).

TrkA gene expression and immunoreactivity in human skin

Next, we asked whether BIPN is associated with altered expression and abundance of NGF and its receptor TrkA. On the RNA level, qPCR revealed a trend toward lower TrkA expression in the skin of BIPN patients compared to controls (median: 1.349 vs. 2.046; Mann–Whitney U = 372, p > 0.05), with a small effect size (rank‐biserial r = −0.21) (Figure 3A ). Moreover, NGF RNA expression was significantly decreased in BIPN patients compared to healthy controls (median: 0.92, IQR: 0.69–1.14 vs. median: 1.22, IQR: 1.03–1.43, p < 0.01) (Figure 3A ). Within the BIPN cohort, patients with painful neuropathy exhibited higher NGF mRNA levels than those with painless BIPN (median: 1.12, IQR: 0.90–1.14 vs. median: 0.794, IQR: 0.64–1.03, p < 0.05) (Figure 3A ).

At the protein level, anti‐TrkA labels showed substantial colocalization with anti‐PGP9.5 labels (Figure 3B ). In this dataset, quantification of PGP9.5 immunofluorescence revealed a two‐fold reduction in total intraepidermal nerve fiber area in BIPN patients compared to controls (Figure 3C ). Similarly, PGP9.5 MFI was decreased in BIPN patients (Figure 3D ). In contrast, TrkA MFI was more than five times higher in BIPN patients than in healthy controls (Figure 3E ). When normalized to PGP9.5, TrkA MFI was elevated by more than six‐fold in BIPN patients (Figure 3F ). Both absolute and normalized TrkA MFI showed moderate positive correlations with the number of bortezomib (BTZ) treatment cycles ( r = 0.33, p < 0.05; r = 0.36, p < 0.05), but neither correlated with cumulative BTZ dose.

The same analysis was applied to nerve fibers in the sub‐epidermal region. Here, the PGP9.5 area did not differ significantly between controls and BIPN patients (Figure 3G ) although PGP9.5 MFI was decreased in BIPN patients (Figure 3H ). In contrast, both TrkA MFI and normalized TrkA MFI were significantly increased—by approximately six‐fold—in BIPN patients compared to controls (Table S2 ; Figure 3I,J ). However, in contrast to the epidermal findings, neither TrkA MFI nor normalized TrkA MFI in the subepidermal region correlated with the number of BTZ treatment cycles or cumulative BTZ dose ( p > 0.05 for all comparisons).

Vascular‐nerve fiber interaction

Given reported effects of NGF‐TrkA signaling on blood vessel formation [ 28 ], we assessed dermal blood vessel area and vessel innervation in the skin samples. The vascular area was increased in BIPN patients compared to controls (median: 926.0, IQR: 714.8–1710 vs. median: 705.1, IQR: 496.5–1190, p < 0.05), with no differences observed between the BIPN subgroups (Figure 4A ).

We also quantified the area of blood vessels located close to the nerve fiber region, defined as the area within a 5 μm radius around the nerve fiber segmentation (see workflow in Figure 4B ). In BIPN patients, the vessel‐nerve area (blood vessel area within the extended nerve fiber region) revealed two phenotypes: In the painless BIPN subgroup, the area was three times larger than in controls (median: 164.0, IQR: 81.86–308.4 vs. median: 54.2, IQR: 0.31–105.9, p < 0.001) (Figure 4C ). The painful BIPN subgroup showed a vessel‐nerve area similar to that of controls (median: 48.96, IQR: 0.62–167.5) (Figure 4C ).

GAP43 expression and nerve regeneration

NGF‐TrkA activation has been shown to increase GAP43 immunoreactivity, associated with painful diabetic peripheral neuropathy [ 29 ]. Given the increased TrkA levels, we investigated whether GAP43 immunoreactivity was also increased in the case of BIPN.

Immunolabeling showed that GAP43+ nerve fibers largely overlapped with PGP9.5+ nerve fibers in all samples (Figure S3A ). Quantitative analysis revealed no significant differences in GAP43 MFI or normalized GAP43 MFI (relative to PGP9.5) between controls and BIPN patients in either the epidermis or sub‐epidermis (Figure S3B ).

To explore whether peptidergic fibers might be specifically affected, we also analyzed GAP43 immunoreactivity in CGRP+ nerve fibers. Similarly, GAP43 MFI and normalized GAP43 MFI (relative to CGRP) showed no significant differences between controls and BIPN patients in either the epidermis or sub‐epidermis (Figure S3B ; Table S3 ). Notably, a decrease in GAP43 MFI was observed on all nerve fibers in the painful subgroup of the BIPN cohort when compared with the painless subgroup ( p < 0.05) (Figure S3B ), although no significant differences in normalized GAP43 intensities were observed (Figure S3B ).

DISCUSSION

In this study, we found decreased intraepidermal nerve fiber density, but increased TrkA abundance on remaining nerve fibers in lower leg skin from patients with BIPN, accompanied by enhanced dermal vascularization. Elevated NGF mRNA expression and heightened TrkA immunoreactivity in the painful subgroup indicate a potential link between NGF–TrkA signaling and symptom manifestation.

Consistent with the reduced IENFD, BIPN patients in our cohort exhibited decreased sensory nerve conduction velocities and reduced sensory nerve action potential amplitudes in the sural nerve, along with thermal sensory deficits – findings that typically accompany a reduction in IENFD PGP9.5 [ 30 , 31 ]. These results indicate that both large and small sensory nerve fibers are impaired in patients with BIPN.

Notably, axonal degeneration, as assessed by reductions in IENFD and nerve fiber area, did not show a preferential effect on CGRP+ nerve fibers. However, the relative percentage of CGRP+ nerve fiber area (to PGP9.5+ nerve fiber area) and normalized CGRP MFI were increased in the sub‐epidermal region of BIPN patients. This finding aligns with earlier studies demonstrating an increase in CGRP+ neurons in the DRG of BIPN mouse models [ 32 ] and higher CGRP+ fiber density in painful diabetic polyneuropathy patients [ 33 ], suggesting a potential compensatory or maladaptive upregulation of CGRP expression in response to nerve injury.

NGF binding to its high‐affinity receptor TrkA triggers receptor internalization, followed by either recycling to the plasma membrane or degradation via the lysosomal pathway—a process partially regulated by proteasome‐mediated deubiquitination [ 34 ]. In the context of bortezomib‐induced peripheral neuropathy (BIPN), the observed increase in TrkA immunoreactivity within nerve fibers, despite reduced NGF mRNA expression in whole‐skin biopsies, may reflect the impact of bortezomib (BTZ) treatment itself. As a proteasome inhibitor, BTZ disrupts the ubiquitin–proteasome system, which governs the degradation of approximately 80% of cellular proteins [ 35 ]. Inhibition of this system is likely to impair TrkA deubiquitination and turnover, leading to its intracellular accumulation in peripheral nerve fibers. The hypothesis is further supported by the moderate positive correlation between the number of BTZ treatment cycles and normalized TrkA MFI.

Interestingly, while TrkA abundance is elevated in both painless and painful BIPN cohorts, higher NGF mRNA levels are detected in the painful subgroup. Although the underlying mechanism remains unclear, it is conceivable that increased NGF availability in this group enhances both canonical [ 36 ] and non‐canonical [ 22 , 37 , 38 ] TrkA signaling. Elevated NGF may promote receptor internalization and delayed degradation, sustaining signaling activity. Moreover, TrkA can be transactivated by local G protein–coupled receptors, a process that facilitates its translocation to the plasma membrane and may further amplify NGF‐mediated signaling [ 39 , 40 ]. While speculative, such a mechanism could explain how local NGF production, together with TrkA accumulation, contributes to proalgesic signaling in BIPN. If this interpretation holds true, clinically approved Trk kinase inhibitors, such as larotrectinib, could be evaluated in preclinical BIPN models to explore their therapeutic potential in modulating aberrant NGF/TrkA signaling associated with painful neuropathy.

Experimental studies have shown that NGF injection induces long‐lasting thermal hyperalgesia in both animals [ 41 ] and humans [ 42 ]. Conversely, neutralizing antibodies against NGF or TrkA as well as TrkA mutations attenuate behavioral responses to noxious stimuli in animal models [ 43 , 44 ]. In line with our findings in BIPN, NGF mRNA levels have been reported to be decreased in the skin of patients with peripheral neuropathy (PNP) [ 45 ]. Moreover, clinical phase II trials investigating anti‐NGF antibodies for the treatment of diabetic peripheral neuropathic pain and postherpetic neuralgia (PHN) have demonstrated significant pain reduction [ 46 , 47 ].

In CIPN models, reduced nerve blood perfusion and diminished density of vasa nervorum have been reported [ 12 , 13 ]. Consistent with these findings, we observed a reduction in blood vessel area within the nerve fiber region in patients with BIPN, despite an overall increase in total dermal blood vessel area. This suggests that the enhanced dermal vascularization in painful BIPN may not confer functional benefit, as these patients continue to experience significant neuropathic symptoms.

In contrast, painless BIPN patients exhibited a different vascular pattern, implying that not just the quantity, but the spatial distribution and functional interaction between blood vessels and nerve fibers may be critical for symptom modulation.

Effective oxygen, nutrient, and hormone exchange likely depends on the close anatomical and functional relationship between nerve fibers and blood vessels [ 48 ], as observed in the painless BIPN group (Figure S4 ).

The usability of GAP43 as a biomarker for peripheral neuropathy remains controversial. While an increased IENFD ratio of GAP43/PGP9.5 was proposed as a biomarker to distinguish painful diabetic neuropathic (DNP) from painless DNP [ 29 , 49 ], other studies reported conflicting results. For instance, IENFD of GAP43 was decreased in painful DNP [ 50 ], and the GAP43/PGP9.5 ratio did not significantly change before and after chemotherapy in patients [ 51 ]. These findings are similar to observations in nerve compression cases before and after surgery [ 52 ] as well as to our results, further complicating the interpretation of GAP43.

Our research has limitations: First, the method of calculating non‐peptidergic nerve fibers by subtracting peptidergic nerve fibers from the total nerve fiber count is overly simplistic. This approach does not account for the substantial overlap in the expression of non‐peptidergic and peptidergic markers in human DRG neurons [ 53 , 54 ]. Second, skin sections were quantified at a single time point, which limits the ability to capture dynamic changes in nerve fibers. Third, the quantification of bioimages was conducted using a local view (region of interest), which may overlook broader spatial patterns and lead to incomplete or biased data interpretation. Finally, the TrkA antibody used in this study targets total TrkA rather than activated (phosphorylated) TrkA. To further validate TrkA activation, future studies should employ antibodies specific to phosphorylated TrkA or assess downstream signaling pathways, such as PI3K/Akt and MAPK cascades, which are critical mediators of TrkA signaling.

In conclusion, this study suggests that BTZ, a first‐in‐class proteasome inhibitor used in the treatment of multiple myeloma, damages small nerve fibers, and the resulting accumulation of functional TrkA receptors in remaining nerve fibers is associated with the development of pain symptoms in BIPN. These findings establish a testable mechanistic hypothesis for symptom development that requires confirmation in subsequent preclinical models.

AUTHOR CONTRIBUTIONS

Yuying Jin and Julian Brennecke contributed to the methodology of the study; Yuying Jin and Annemarie Sodmann developed the software; Yuying Jin, Robert Blum, and Claudia Sommer performed data curation; Yuying Jin conducted investigation; Annemarie Sodmann carried out validation; Yuying Jin and Annemarie Sodmann performed formal analysis; RB and CS provided supervision and funding acquisition; Yuying Jin and Robert Blum contributed to visualization; Martin Kortüm, Yuying Jin, Nadine Cebulla, Daniel Schirmer, Eva Runau, Leon Flamm, Calvin Terhorst, Laura Jähnel, Johanna Güse, Nicola Giordani, Annett Wieser, Julian Brennecke, Annemarie Sodmann, Robert Blum, and Claudia Sommer participated in project administration; Robert Blum supplied resources. Yuying Jin wrote the original draft, and Yuying Jin, Robert Blum, and Claudia Sommer contributed to review and editing.

FUNDING INFORMATION

This study was conducted as part of the ResolvePAIN project (Project ID: 426503586), which is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation). Yuying Jin received a scholarship from the China Scholarship Council (CSC, No. 202106230110) to support her participation in this research.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Data S1: Supporting Information.

ACKNOWLEDGMENTS

The authors would like to express their gratitude to Kathleen Stahl for processing skin biopsies and all patients as well as controls for participating in this study.

DATA AVAILABILITY STATEMENT

Anonymized clinical data are available from the corresponding author upon request, pending approval and a signed data‐sharing agreement with Würzburg University. Data from patients with BIPN within the KFO5001 project have been partially published previously [ 26 , 27 ]. Deep‐learning segmentation models, original bioimages, and corresponding predicted segmentations generated by our models are publicly available at Zenodo (DOI: 10.5281/zenodo.15773443 ).

REFERENCES

1 Fricker LD . Proteasome inhibitor drugs . Annu Rev Pharmacol Toxicol . 2020 ; 60 : 457 – 476 . 31479618 10.1146/annurev-pharmtox-010919-023603

2 Attal M , Lauwers‐Cances V , Hulin C , Leleu X , Caillot D , Escoffre M , et al. Lenalidomide, bortezomib, and dexamethasone with transplantation for myeloma . N Engl J Med . 2017 ; 376 ( 14 ): 1311 – 1320 . 28379796 10.1056/NEJMoa1611750 PMC6201242

3 Cavo M , Gay F , Beksac M , Pantani L , Petrucci MT , Dimopoulos MA , et al. Autologous haematopoietic stem‐cell transplantation versus bortezomib‐melphalan‐prednisone, with or without bortezomib‐lenalidomide‐dexamethasone consolidation therapy, and lenalidomide maintenance for newly diagnosed multiple myeloma (EMN02/HO95): a multicentre, randomised, open‐label, phase 3 study . Lancet Haematol . 2020 ; 7 ( 6 ): E456 – E468 . 32359506 10.1016/S2352-3026(20)30099-5

4 Grosicki S , Simonova M , Spicka I , Pour L , Kriachok I , Gavriatopoulou M , et al. Once‐per‐week selinexor, bortezomib, and dexamethasone versus twice‐per‐week bortezomib and dexamethasone in patients with multiple myeloma (BOSTON): a randomised, open‐label, phase 3 trial . Lancet . 2020 ; 396 ( 10262 ): 1563 – 1573 . 33189178 10.1016/S0140-6736(20)32292-3

5 Cavaletti G , Nobile‐Orazio E . Bortezomib‐induced peripheral neurotoxicity: still far from a painless gain . Haematologica . 2007 ; 92 ( 10 ): 1308 – 1310 . 18024368 10.3324/haematol.11752

6 Argyriou AA , Iconomou G , Kalofonos HP . Bortezomib‐induced peripheral neuropathy in multiple myeloma: a comprehensive review of the literature . Blood . 2008 ; 112 ( 5 ): 1593 – 1599 . 18574024 10.1182/blood-2008-04-149385

7 Mantyh PW , Koltzenburg M , Mendell LM , Tive L , Shelton DL . Antagonism of nerve growth factor‐TrkA signaling and the relief of pain . Anesthesiology . 2011 ; 115 ( 1 ): 189 – 204 . 21602663 10.1097/ALN.0b013e31821b1ac5 PMC3121917

8 Nakahashi Y , Kamiya Y , Funakoshi K , Miyazaki T , Uchimoto K , Tojo K , et al. Role of nerve growth factor‐tyrosine kinase receptor a signaling in paclitaxel‐induced peripheral neuropathy in rats . Biochem Biophys Res Commun . 2014 ; 444 ( 3 ): 415 – 419 . 24480438 10.1016/j.bbrc.2014.01.082

9 Ji RR , Samad TA , Jin SX , Schmoll R , Woolf CJ . p38 MAPK activation by NGF in primary sensory neurons after inflammation increases TRPV1 levels and maintains heat hyperalgesia . Neuron . 2002 ; 36 ( 1 ): 57 – 68 . 12367506 10.1016/s0896-6273(02)00908-x

10 Kim Y , Je MA , Jeong M , Kwon H , Jang A , Kim J , et al. Upregulation of NGF/TrkA‐related proteins in dorsal root ganglion of paclitaxel‐induced peripheral neuropathy animal model . J Pain Res . 2024 ; 17 : 3919 – 3932 . 39588524 10.2147/JPR.S470671 PMC11586490

11 Zhang X , Huang J , McNaughton PA . NGF rapidly increases membrane expression of TRPV1 heat‐gated ion channels . EMBO J . 2005 ; 24 ( 24 ): 4211 – 4223 . 16319926 10.1038/sj.emboj.7600893 PMC1356334

12 Kirchmair R , Tietz AB , Panagiotou E , Walter DH , Silver M , Yoon YS , et al. Therapeutic angiogenesis inhibits or rescues chemotherapy‐induced peripheral neuropathy: taxol‐ and thalidomide‐induced injury of vasa nervorum is ameliorated by VEGF . Mol Ther . 2007 ; 15 ( 1 ): 69 – 75 . 17164777 10.1038/sj.mt.6300019

13 Taib S , Durand J , Dehais V , Boulay AC , Martin S , Blugeon C , et al. Vascular dysfunction is at the onset of oxaliplatin‐induced peripheral neuropathy symptoms in mice . Life Sci Alliance . 2025 ; 8 ( 2 ): e202402791 . 39578077 10.26508/lsa.202402791 PMC11584327

14 Emanueli C , Salis MB , Pinna A , Graiani G , Manni L , Madeddu P . Nerve growth factor promotes angiogenesis and arteriogenesis in ischemic hindlimbs . Circulation . 2002 ; 106 ( 17 ): 2257 – 2262 . 12390957 10.1161/01.cir.0000033971.56802.c5

15 Rajkumar SV , Dimopoulos MA , Palumbo A , Blade J , Merlini G , Mateos MV , et al. International myeloma working group updated criteria for the diagnosis of multiple myeloma . Lancet Oncol . 2014 ; 15 ( 12 ): e538 – e548 . 25439696 10.1016/S1470-2045(14)70442-5

16 Finnerup NB , Haroutounian S , Kamerman P , Baron R , Bennett DLH , Bouhassira D , et al. Neuropathic pain: an updated grading system for research and clinical practice . Pain . 2016 ; 157 ( 8 ): 1599 – 1606 . 27115670 10.1097/j.pain.0000000000000492 PMC4949003

17 Kane NM , Oware A . Nerve conduction and electromyography studies . J Neurol . 2012 ; 259 ( 7 ): 1502 – 1508 . 22614870 10.1007/s00415-012-6497-3

18 Rolke R , Baron R , Maier C , Tolle TR , Treede DR , Beyer A , et al. Quantitative sensory testing in the German research network on neuropathic pain (DFNS): standardized protocol and reference values . Pain . 2006 ; 123 ( 3 ): 231 – 243 . 16697110 10.1016/j.pain.2006.01.041

19 Magerl W , Krumova EK , Baron R , Tolle T , Treede RD , Maier C . Reference data for quantitative sensory testing (QST): refined stratification for age and a novel method for statistical comparison of group data . Pain . 2010 ; 151 ( 3 ): 598 – 605 . 20965658 10.1016/j.pain.2010.07.026

20 Rolke R , Magerl W , Campbell KA , Schalber C , Caspari S , Birklein F , et al. Quantitative sensory testing: a comprehensive protocol for clinical trials . Eur J Pain . 2006 ; 10 ( 1 ): 77 – 88 . 16291301 10.1016/j.ejpain.2005.02.003

21 Jin Y , Brennecke J , Sodmann A , Blum R , Sommer C . Antibody selection and automated quantification of TRPV1 immunofluorescence on human skin . Sci Rep . 2024 ; 14 ( 1 ): 28496 . 39557902 10.1038/s41598-024-79271-9 PMC11574049

22 Gupta R , Dittmeier M , Wohlleben G , Nickl V , Bischler T , Luzak V , et al. Atypical cellular responses mediated by intracellular constitutive active TrkB (NTRK2) kinase domains and a solely intracellular NTRK2‐fusion oncogene . Cancer Gene Ther . 2024 ; 31 ( 9 ): 1357 – 1379 . 39039193 10.1038/s41417-024-00809-0 PMC11405271

23 Segebarth D , Griebel M , Stein N , von Collenberg CR , Martin C , Fiedler D , et al. On the objectivity, reliability, and validity of deep learning enabled bioimage analyses . Elife . 2020 ; 9 : 9 .

24 Griebel M , Segebarth D , Stein N , Schukraft N , Tovote P , Blum R , et al. Deep learning‐enabled segmentation of ambiguous bioimages with deepflash2 . Nat Commun . 2023 ; 14 ( 1 ): 1679 . 36973256 10.1038/s41467-023-36960-9 PMC10043282

25 Lauria G , Hsieh ST , Johansson O , Kennedy WR , Leger JM , Mellgren SI , et al. European Federation of Neurological Societies/Peripheral Nerve Society Guideline on the use of skin biopsy in the diagnosis of small fiber neuropathy. Report of a joint task force of the European Federation of Neurological Societies and the Peripheral Nerve Society . Eur J Neurol . 2010 ; 17 ( 7 ): 903 – 912 . e44‐9. 20642627 10.1111/j.1468-1331.2010.03023.x

26 Cebulla N , Schirmer D , Runau E , Flamm L , Gommersbach S , Stengel H , et al. Neurofilament light chain levels indicate acute axonal damage under bortezomib treatment . J Neurol . 2023 ; 270 ( 6 ): 2997 – 3007 . 36802032 10.1007/s00415-023-11624-2 PMC10188420

27 Zhou X , Han S , Cebulla N , Haertle L , Steinhardt MJ , Schirmer D , et al. Bortezomib induced peripheral neuropathy and single nucleotide polymorphisms in PKNOX1 . Biomark Res . 2023 ; 11 ( 1 ): 52 . 37194045 10.1186/s40364-023-00490-9 PMC10189922

28 Calza L , Giardino L , Giuliani A , Aloe L , Levi‐Montalcini R . Nerve growth factor control of neuronal expression of angiogenetic and vasoactive factors . Proc Natl Acad Sci U S A . 2001 ; 98 ( 7 ): 4160 – 4165 . 11259645 10.1073/pnas.051626998 PMC31196

29 Cheng HT , Dauch JR , Porzio MT , Yanik BM , Hsieh W , Smith AG , et al. Increased axonal regeneration and swellings in intraepidermal nerve fibers characterize painful phenotypes of diabetic neuropathy . J Pain . 2013 ; 14 ( 9 ): 941 – 947 . 23685187 10.1016/j.jpain.2013.03.005 PMC3994562

30 Bechakra M , Nieuwenhoff MD , Rosmalen JV , Groeneveld GJ , Huygen FJPM , et al. Pain‐related changes in cutaneous innervation of patients suffering from bortezomib‐induced, diabetic or chronic idiopathic axonal polyneuropathy . Brain Res . 2020 ; 1730 : 146621 . 31926911 10.1016/j.brainres.2019.146621

31 Park SB , Cetinkaya‐Fisgin A , Argyriou AA , Hoke A , Cavaletti G , Alberti P . Axonal degeneration in chemotherapy‐induced peripheral neurotoxicity: clinical and experimental evidence . J Neurol Neurosurg Psychiatry . 2023 ; 94 ( 11 ): 962 – 972 . 37015772 10.1136/jnnp-2021-328323 PMC10579520

32 Bruna J , Udina E , Ale A , Vilches JJ , Vynckier A , Monbaliu J , et al. Neurophysiological, histological and immunohistochemical characterization of bortezomib‐induced neuropathy in mice . Exp Neurol . 2010 ; 223 ( 2 ): 599 – 608 . 20188093 10.1016/j.expneurol.2010.02.006

33 Gylfadottir SS , Itani M , Kristensen AG , Tankisi H , Jensen TS , Sindrup SH , et al. Analysis of macrophages and Peptidergic fibers in the skin of patients with painful diabetic polyneuropathy . Neurol Neuroimmunol Neuroinflamm . 2022 ; 9 ( 1 ):e1111.

34 Makkerh JP , Ceni C , Auld DS , Vaillancourt F , Dorval G , Barker PA . p75 neurotrophin receptor reduces ligand‐induced Trk receptor ubiquitination and delays Trk receptor internalization and degradation . EMBO Rep . 2005 ; 6 ( 10 ): 936 – 941 . 16113645 10.1038/sj.embor.7400503 PMC1369184

35 Collins GA , Goldberg AL . The logic of the 26S proteasome . Cell . 2017 ; 169 ( 5 ): 792 – 806 . 28525752 10.1016/j.cell.2017.04.023 PMC5609836

36 Huang EJ , Reichardt LF . Trk receptors: roles in neuronal signal transduction . Annu Rev Biochem . 2003 ; 72 : 609 – 642 . 12676795 10.1146/annurev.biochem.72.121801.161629

37 Watson FL , Porcionatto MA , Bhattacharyya A , Stiles CD , Segal RA . TrkA glycosylation regulates receptor localization and activity . J Neurobiol . 1999 ; 39 ( 2 ): 323 – 336 . 10235685 10.1002/(sici)1097-4695(199905)39:2<323::aid-neu15>3.0.co;2-4

38 Lee FS , Chao MV . Activation of Trk neurotrophin receptors in the absence of neurotrophins . Proc Natl Acad Sci U S A . 2001 ; 98 ( 6 ): 3555 – 3560 . 11248116 10.1073/pnas.061020198 PMC30691

39 Rajagopal R , Chen ZY , Lee FS , Chao MV . Transactivation of Trk neurotrophin receptors by G‐protein‐coupled receptor ligands occurs on intracellular membranes . J Neurosci . 2004 ; 24 ( 30 ): 6650 – 6658 . 15282267 10.1523/JNEUROSCI.0010-04.2004 PMC6729703

40 Vaegter CB , Jansen P , Fjorback AW , Glerup S , Skeldal S , Kjolby M , et al. Sortilin associates with Trk receptors to enhance anterograde transport and neurotrophin signaling . Nat Neurosci . 2011 ; 14 ( 1 ): 54 – 61 . 21102451 10.1038/nn.2689 PMC3808973

41 Khodorova A , Nicol GD , Strichartz G . The TrkA receptor mediates experimental thermal hyperalgesia produced by nerve growth factor: modulation by the p75 neurotrophin receptor . Neuroscience . 2017 ; 340 : 384 – 397 . 27826102 10.1016/j.neuroscience.2016.10.064 PMC5154773

42 Dyck PJ , Peroutka S , Rask C , Burton E , Baker MK , Lehman KA , et al. Intradermal recombinant human nerve growth factor induces pressure allodynia and lowered heat‐pain threshold in humans . Neurology . 1997 ; 48 ( 2 ): 501 – 505 . 9040746 10.1212/wnl.48.2.501

43 Ugolini G , Marinelli S , Covaceuszach S , Cattaneo A , Pavone F . The function neutralizing anti‐TrkA antibody MNAC13 reduces inflammatory and neuropathic pain . Proc Natl Acad Sci U S A . 2007 ; 104 ( 8 ): 2985 – 2990 . 17301229 10.1073/pnas.0611253104 PMC1815293

44 Pacifico P , Testa G , Amodeo R , Mainardi M , Tiberi A , Convertino D , et al. Human TrkAR649W mutation impairs nociception, sweating and cognitive abilities: a mouse model of HSAN IV . Hum Mol Genet . 2023 ; 32 ( 8 ): 1380 – 1400 . 36537577 10.1093/hmg/ddac295 PMC10077510

45 Uceyler N , Riediger N , Kafke W , Sommer C . Differential gene expression of cytokines and neurotrophic factors in nerve and skin of patients with peripheral neuropathies . J Neurol . 2015 ; 262 ( 1 ): 203 – 212 . 25371017 10.1007/s00415-014-7556-8

46 Wang H , Romano G , Frustaci ME , Bohidar N , Ma H , Sanga P , et al. Fulranumab for treatment of diabetic peripheral neuropathic pain: a randomized controlled trial . Neurology . 2014 ; 83 ( 7 ): 628 – 637 . 25008392 10.1212/WNL.0000000000000686 PMC4141990

47 Bramson C , Herrmann DN , Carey W , Keller D , Brown MT , West CR , et al. Exploring the role of tanezumab as a novel treatment for the relief of neuropathic pain . Pain Med . 2015 ; 16 ( 6 ): 1163 – 1176 . 25594611 10.1111/pme.12677

48 Eichmann A , Thomas JL . Molecular parallels between neural and vascular development . Cold Spring Harb Perspect Med . 2013 ; 3 ( 1 ): a006551 . 23024177 10.1101/cshperspect.a006551 PMC3530036

49 Galosi E , La Cesa S , Di Stefano G , Karlsson P , Fasolino A , Leone C , et al. A pain in the skin. Regenerating nerve sprouts are distinctly associated with ongoing burning pain in patients with diabetes . Eur J Pain . 2018 ; 22 ( 10 ): 1727 – 1734 . 29885017 10.1002/ejp.1259

50 Shillo P , Yiangou Y , Donatien P , Greig M , Selvarajah D , Wilkinson ID , et al. Nerve and vascular biomarkers in skin biopsies differentiate painful from painless peripheral neuropathy in type 2 diabetes . Front Pain Res (Lausanne) . 2021 ; 2 : 731658 . 35295465 10.3389/fpain.2021.731658 PMC8915761

51 Albayrak M , Figueras C , Segui E , Campolo M , Gabarron E , Moreno R , et al. Prognostic value of cutaneous reinnervation with GAP‐43 in oxaliplatin‐induced neuropathy . J Neurol . 2022 ; 269 ( 8 ): 4174 – 4184 . 35258850 10.1007/s00415-022-11035-9 PMC9293807

52 Carroll L , Sandy‐Hindmarch O , Baskozos G , Zhu GC , McCarthy J , Schmid A . Cutaneous expression of growth‐associated protein 43 is not a compelling marker for human nerve regeneration in carpal tunnel syndrome . PLoS One . 2022 ; 17 ( 11 ): e0277133 . 36383568 10.1371/journal.pone.0277133 PMC9668135

53 Tavares‐Ferreira D , Shiers S , Ray PR , Wangzhou A , Jeevakumar V , Sankaranarayanan I , et al. Spatial transcriptomics of dorsal root ganglia identifies molecular signatures of human nociceptors . Sci Transl Med . 2022 ; 14 ( 632 ): eabj8186 . 35171654 10.1126/scitranslmed.abj8186 PMC9272153

54 Shiers SI , Sankaranarayanan I , Jeevakumar V , Cervantes A , Reese JC , Price TJ . Convergence of peptidergic and non‐peptidergic protein markers in the human dorsal root ganglion and spinal dorsal horn . J Comp Neurol . 2021 ; 529 ( 10 ): 2771 – 2788 . 33550628 10.1002/cne.25122 PMC8113129