Global Spinal Cord Peptidome Profiling in Response to Osteoarthritis in Rats

Abstract

Osteoarthritis (OA) is a complex and increasingly prevalent condition, affecting an estimated 600 million people worldwide and significantly reducing quality of life. Understanding OA as both an inflammatory and neurological disease presents challenges for diagnosis and treatment, as no curative therapy is currently available. The neuroactive effects of OA pain on neuropeptide systems, particularly within the spinal cord, remain underexplored, impeding therapeutic advances. Mass spectrometry (MS) was employed to characterize the spinal cord peptidome in a validated rat model of OA pain. The peptidome was analyzed longitudinally over 84 days using the Montreal Induction of Rat Arthritis Testing (MI-RAT^©; n = 20) model, compared to arthrotomic (Sham; n = 4) and healthy (Naive; n = 9) groups. Label-free peptidome profiling using liquid chromatography coupled with tandem MS (LC-MS/MS) revealed dynamic changes in endogenous spinal peptides during OA progression, leading to the identification of 624 peptides derived from 29 prohormone precursors. The findings reveal substantial changes in peptide levels in the spinal cord, particularly involving neuropeptide substance P and peptides derived from proenkephalin, calcitonin gene-related peptide, and somatostatin. These results provide novel insights into the molecular mechanisms underlying OA-associated pain and identify potential targets for new therapeutic interventions in neurological pain conditions.

Graphical Abstract

INTRODUCTION

Osteoarthritis (OA) is a common, disabling musculoskeletal condition that significantly impacts quality of life and poses a growing health and socioeconomic burden worldwide.^{1, 2} In 2020, OA affected an estimated 495 million people, equivalent to 7.6% of the global population.^{1, 2} With contributing factors such as aging, increasingly sedentary lifestyles, rising obesity rates, more frequent joint injuries, and longer life expectancies, the number of affected individuals is projected to increase by 75% by 2050.^{1, 2} Initially, OA was recognized as a pathology marked by joint damage, including cartilage degeneration, synovitis, subchondral bone remodeling, and soft-tissue alterations.³ However, recent studies have broadened the understanding of OA, redefining it not only as an inflammatory condition but also as a neurological disease.^{4, 5} This complexity makes OA pain particularly challenging to diagnose and treat. Currently, there are no effective pharmacological options for long-term pain management or approved disease-modifying therapies. Existing OA pain management strategies rely on symptomatic relief through analgesics, physical therapy, and lifestyle modifications, but these approaches remain insufficient for many patients.¹

Chronic OA pain arises from peripheral and central sensitization, driven by neurogenic inflammation, spinal hyperexcitability (wind-up), and an imbalance between facilitatory and inhibitory endogenous pain controls.^{4, 6–8} Furthermore, research using inflammatory animal models has demonstrated that neuronal plasticity at the spinal level contributes to mechanical tactile allodynia.^{6, 9} Gaining deeper insight into OA as a nociplastic condition requires investigating its underlying molecular mechanisms.^{10, 11}

One component that remains incompletely characterized is the OA-affected neuropeptidome, which refers to the set of translated and cell-to-cell signaling peptides in the central nervous system (CNS) modulated by chronic pain. Neuropeptides and peptide hormones, cleaved from larger precursor proteins through enzymatic processing, function as potent neuromodulators in the CNS and play vital roles in regulating various physiological functions, such as feeding, reproduction, mood, circadian rhythm.^{12, 13} It is also well documented that certain neuropeptides are essential for the onset of peripheral and central sensitization in OA.^{14, 15} Previous studies using animal models of OA have highlighted the coexistence of a painful phenotype with the temporal expression of specific pro- and anti-nociceptive spinal neuropeptides, as determined by targeted analyses using liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS).^16–23 These studies focused on quantifying selected neuropeptides known to be involved in OA pain mechanisms, rather than performing global neuropeptidome profiling. Among these models, the validated Montreal Induction of Rat Arthritis Testing (MI-RAT^©) model developed a chronic OA pain phenotype that mimics the human condition. It also exhibited biomechanical and histological alterations in the stifle joint, along with upregulated spinal expression of substance P (SP), calcitonin gene-related peptide (CGRP), bradykinin, somatostatin (SST) and enkephalins (Met- and Leu-enkephalin).^{22, 23}

While these targeted findings confirm the involvement of several key neuropeptides, the full extent of the spinal peptidome’s response to OA remains uncharacterized. This knowledge gap is significant because the effects of chronic pain on the central nervous system (CNS) neuropeptidome remain underexplored. Furthermore, most mammalian peptidome studies have focused on brain regions, with limited in-depth profiling of the endogenous peptidome in the spinal cord—another crucial component of the CNS. This is particularly relevant since the gate control theory of pain, established in 1965,²⁴ remains a cornerstone of modern pain concepts. The spinal cord, serving as the link between the brain and the peripheral nervous system, is a critical tissue for studying neural transmission, particularly in pain-related research.²⁵

Mass spectrometry (MS)-based peptidomics has emerged as a highly effective tool for comprehensive endogenous peptide profiling. Unlike conventional techniques such as radioimmunoassay, immunohistochemistry, or Edman degradation, MS-driven methods enable high-throughput detection of numerous endogenous peptides in a single analysis. Furthermore, MS-based strategies enhance the characterization of post-translational modifications, overcoming limitations of earlier approaches that often struggled to detect such biochemical alterations.²⁶

In this study, using the previously validated MI-RAT^© model of OA pain,^{22, 23} an MS-based peptidomics approach was applied to characterize the dynamic landscape of the rat spinal cord peptidome and its modulation in response to OA induction. The aim is to identify key regulatory signatures underlying spinal peptidome changes and potentially highlight novel therapeutic targets.

EXPERIMENTAL

Materials and reagents:

The solvents and reagents utilized, unless otherwise noted, were acquired from Thermo Fisher Scientific (Pittsburgh, PA, USA) or Millipore Sigma (St. Louis, MO, USA).

Ethics statement and animals:

The care and use of animals were in accordance with the guidelines of the Canadian Council on Animal Care and US National Institutes of Health and approved by the “Comité d’Éthique de l’Utilisation des Animaux de l’Université de Montréal” (Rech-2078). Additionally, this study followed the ARRIVE guidelines 2.0.²⁷ A total of 33 Sprague-Dawley rats (ovariectomized female; body weight 366.37 (18.67) g; age 6–8 weeks) from Charles River Laboratories (Saint-Constant, QC, Canada) were used, and housed in pairs in a stress-free environment and care with food, fruit crunchy treats, and water ad libitum. The laboratory conditions were maintained at a constant temperature of 22 °C in a 12 h light–dark cycle.

MI-RAT© osteoarthritic surgical model:

A group of 20 rats was first administered intra-muscularly 1.0 mg/kg of buprenorphine (Buprenorphine SR^®, Chiron Compounding Pharmacy Inc., Guelph, ON, Canada) as a pre-medication, 24 h before a general anesthesia using 2% isoflurane (IsoFlo^®, Abbott Animal Health, Saint-Laurent, QC, Canada) in an O₂ mixture. The MI-RAT^© was a surgical model of OA-associated chronic pain in rodents, involving right stifle instability through cranial cruciate ligament transection and destabilization of the medial meniscus procedure, along with calibrated exercise, as previously described and validated.^{22, 23} At the end of the surgery, a periarticular block of 0.25% bupivacaine solution (Marcaine^®, McKesson Canada, St.-Laurent, QC, Canada) at a dose of 0.05–0.10 mL per stifle (< 1 mg/kg) was given.

Experimental design:

After OA induction on day (D) 0, the rats in the MI-RAT^© group (n = 20) were randomly assigned to be temporally sacrificed at four different time points: (i) D21 (n = 4, OA21); (ii) D35 (n = 4, OA35); (iii) D42 (n = 6, OA42); and (iv) D84 (n = 6, OA84). Four additional rats were utilized as sham controls, maintaining all intra-articular structures intact following an arthrotomy. They were randomly sacrificed after D21 (n = 2, Sham21) or D35 (n = 2, Sham35). Finally, a naive group of healthy rats was utilized as a control with no intervention, and randomly sacrificed at D42 (n = 5, Naive42) or D84 (n = 4, Naive84) to serve as baseline values from healthy rats and to normalize values obtained from the OA and Sham groups (Figure 1A).

Fig. 1.

(A) The surgical procedure combining cranial cruciate ligament transection (CCLT) and destabilization of the medial meniscus (DMM), along with calibrated exercise, was performed to induce progressive chronic osteoarthritis (OA) pain, driven by joint instability and forced mobilization. The MI-RAT^© model (OA; n = 20) was compared with naive (n = 9) and sham (n = 4) rat groups. Spinal cord samples from each group were collected at specific time points, with biological replicates included for each condition. (B) Endogenous peptides from the rat spinal cord were extracted, processed, and analyzed via mass spectrometry for identification and quantification.

Histopathological analysis:

Histopathological analysis was conducted to confirm the induction of OA in the MI-RAT^© group (OA; n = 17; 3 ID missing used for another experiment), compared to the arthrotomic (Sham; n = 4) and healthy (Naive; n = 9) control groups. At different sacrifice time points, ipsilateral (right) and contralateral (non-altered) stifle joints were dissected and then fixed in 10% formaldehyde solution (pH 7.4), decalcified and embedded in paraffin. Frontal sections of the paraffined stifle joints, approximately 2000 μm in depth, were stained with hematoxylin and eosin, and safranin-O. Osteoarthritic damage in medial sides was assessed using the Osteoarthritis Research Society International (OARSI) scoring system, subdivided into four scores (Cartilage degeneration, Osteophyte formation, Calcified cartilage and subchondral bone damage, Synovitis), and the Proteoglycan loss score.^{28, 29} The total histological score was calculated by summing the five scores. The sections were independently evaluated by two assessors, a Ph.D. student and a pathologist, with the latter validating the final evaluation, following results comparison and reaching agreement.

Spinal cord sample collection and preparation:

Rats were decapitated following isoflurane overdose (4–5%) for the collection of the whole spinal cord achieved using a saline flush technique.^17–23 Samples were snap-frozen in liquid nitrogen, stored individually, and kept at −80 °C until further use. Rat spinal cords were individually weighed and homogenized in a bead homogenizer following the addition of 0.25% trifluoroacetic acid solution at a ratio of 1:5 (w/v) and 100 mg glass beads, as previously described.^{18, 21–23} The homogenate was centrifuged at 12,000 g for 10 min and the supernatant was then mixed with acetonitrile at a ratio of 1:1 (v/v). The samples were vortexed, precipitated large proteins on ice for 10 min, and centrifuged at 12,000 g for 10 min. The supernatant was then dried down (SpeedVac vacuum at 60°C), reconstituted in 75 μL of water containing 0.1% formic acid (FA), and mixed with 75 μL of an internal standard consisting of a 500 pg/μL heavy isotope-encoded bradykinin (RPPGFSPFR[¹³C6, ¹⁵N4]) solution. The samples were further desalted with C18 Ziptips (Millipore Sigma, St. Louis, MO, USA) following the manufacturer’s instructions. The eluent was then dried with a SpeedVac vacuum and stored at −80°C until LC-MS/MS analysis (Figure 1B).

Mass spectrometry data acquisition:

Samples were reconstituted in 40 μL of 0.1% FA in water. Prior to MS injection, sample concentrations were measured using the Thermo Scientific NanoDrop One Microvolume Spectrophotometer to ensure around 1 μg of peptide sample was loaded onto a self-packed 18 cm length, 75 μm i.d. Bridged Ethylene Hybrid C18 (1.7 μm, 130 Å, Waters) microcapillary column. The Orbitrap Fusion Lumos Tribrid Mass Spectrometer (Thermo Fisher Scientific, Pittsburgh, PA, USA) coupled to a Dionex UltiMate 3000 UPLC system was employed to analyze samples. The LC separation was conducted with 0.1% FA in water as mobile phase A, 0.1% FA in acetonitrile as mobile phase B, and kept at a flow rate of 0.3 μL/min. Peptides were separated by a 120 min gradient elution with 0–2 min 3–7% B, 2–17 min 7–12% B, 17–67 min 12–22% B, 67–82 min 22–32% B, 82–97 min 32–55% B, 97–97.5 min 55–95% B, 97.5–105 min 95% B, 105–105.5 min 95–3% B, and 105.5–120 min 3% B. The MS data were collected in a m/z range of 300–1500 at a resolving power of 60k, an AGC target of 2 × 10⁵, and a maximum injection time of 100 ms. The top 20 precursors were then selected for MS2 fragmentation with a normalized collision energy of 30%, a resolving power of 15k, an AGC target of 1 × 10⁴, and a maximum injection time of 100 ms. Precursors were subjected to dynamic exclusion for 45 s with a 10-ppm tolerance.

Database search and data analysis

The raw MS data were analyzed using PEAKS Studio Xpro software (Bioinformatics Solutions Inc, Waterloo, ON, Canada) against a custom database. This database contains 1,497 precursor protein sequences with signal peptides, which were filtered from an initial set of 8,147 rat proteome entries obtained from UniProt on September 25, 2022, using SignalP 6.0.³⁰ The filtering process was based on the premise that neuropeptide precursors generally feature an N-terminal signal peptide required for secretion. However, while SignalP reliably detects proteins with signal peptides, it does not exclusively identify neuropeptide precursors but also includes enzymes and other secretory proteins that possess signal peptides. The parameters for database searching include 10 ppm parent mass error tolerance, 0.02 Da fragment mass error tolerance, unspecific digest mode. The variable modifications include oxidation on the methionine, pyro-glutamination formation on the N-termini glutamic acid and glutamine, and amidation on the C-termini. Peptides were searched with a maximum of two post-translational modifications per peptide. Label-free quantitation was carried out using the PEAKS Q module, with normalization based on the integrated peak area of isotope-encoded bradykinin. Missing values in the label-free quantitation were excluded from the statistical analysis.

The normalized peptide peak area ratios used for clustering were calculated using an in-house Python script to highlight the relative expression pattern of each peptide across the experimental groups. Specifically, for each peptide, the mean of the internal standard-normalized peak areas was first determined for each of the six groups. Then, the mean value for each group was divided by the sum of the mean peak areas across all six groups. The resulting ratios, which sum to 1.0 for each peptide across all conditions, were used as input for the K-means clustering algorithm.

Statistical comparisons between experimental groups were performed for each peptide using an independent two-sample Welch’s t-test, which does not assume equal variance. To account for the multiple comparisons inherent in large-scale peptidomics analysis, the resulting p-values from each comparison series were adjusted using the Benjamini-Hochberg procedure to control the False Discovery Rate (FDR). A peptide was considered to be significantly regulated if its corresponding adjusted p-value (q-value) was less than 0.2, a threshold chosen to balance sensitive discovery with statistical control in this exploratory study. The MS data were analyzed using Microsoft Excel and in-house-built Python scripts. The histological scores were analyzed using the generalized linear model in SPSS software (IBM^® SPSS^® Statistics Server version 29.0, New York, NY, USA).

RESULTS AND DISCUSSION

Using a well-established rat model of OA, the objective of this study was to investigate the spinal cord peptidome in response to OA pain progression. To achieve this, three groups were studied, each containing biological replicates: Naive, Sham, and OA (utilizing the MI-RAT^© model). All animals successfully completed the study, and there were no complications following the surgical procedure. The OA group was further subdivided based on the time post-surgery into D21, D35, D42, and D84 subgroups. At these designated time points, the rats were euthanized, and their stifle joints and spinal cords were harvested. Endogenous peptides were extracted, followed by sample cleanup, and subsequently analyzed by MS (Figure 1).

Histological analysis results:

The MI-RAT^© ipsilateral stifle total histological score was significantly higher than those of the Sham (p < 0.01) and Naive (p < 0.001) groups over time (Table 1, Data File S1).

Table 1.

Histological analysis parameters are presented as least squares means (LSM) ± standard error of the mean (SEM) in the MI-RAT^© model (OA; n = 17; three IDs excluded for use in another experiment), compared to the Sham (n = 4) and Naive (n = 9) groups.

	Naive (n = 9)		Sham (n = 4)		MI-RAT (OA; n = 17)
Histological OARSI score (LSM ± SEM)	D42 (n = 5)	D84 (n = 4)	D21 (n = 2)	D35 (n = 2)	D21 (n = 4)	D35 (n = 4)	D42 (n = 6)	D84 (n = 3)
Total histological score (0–55)	0.30 ± 1.08***	1.00 ± 1.21***	4.25 ± 1.71**	2.00 ± 1.71***	9.75 ± 1.21a	11.63 ± 1.21a	10.42 ± 0.99a	17.83 ± 1.40b
Cartilage degeneration (0–30)	0.20 ± 0.53***	0.75 ± 0.59***	1.50 ± 0.83	0.50 ± 0.83**	3.25 ± 0.59a	3.50 ± 0.59a	4.17 ± 0.48a	6.33 ± 0.68b
Osteophyte formation (0–4)	0.00 ± 0.27	0.00 ± 0.30**	0.00 ± 0.43	0.00 ± 0.43	0.50 ± 0.30	1.00 ± 0.30	0.67 ± 0.25	1.33 ± 0.35
Calcified cartilage and subchondral bone damage (0–5)	0.00 ± 0.33	0.00 ± 0.37***	0.00 ± 0.53	0.00 ± 0.53	0.50 ± 0.37a	0.25 ± 0.37a	0.83 ± 0.30a	2.00 ± 0.43b
Synovitis (0–4)	0.00 ± 0.32***	0.00 ± 0.36***	2.00 ± 0.51*	1.00 ± 0.51**	3.25 ± 0.36a	3.00 ± 0.36a,c	1.83 ± 0.29b	2.00 ± 0.42b,c
Proteoglycan loss (0–12)	0.10 ± 0.51***	0.25 ± 0.57***	0.75 ± 0.80	0.50 ± 0.80***	2.25 ± 0.57a	3.88 ± 0.57c	2.92 ± 0.46a,c	6.17 ± 0.65b

A generalized linear model was used to evaluate the significance of differences between groups.

^*p < 0.05

^**p < 0.01

^***p < 0.001 (differences between Naive and Sham groups and the MI-RAT group at the same time point).

^a,b,cp≤ 0.043 (indicate significant differences between time points within the same group).

Data sourced from Data File S1.

There were no between-group differences for the contralateral stifle joint (Data File S1). The OA right stifle showed progressive cartilage degeneration (a vs. b, p < 0.01), calcified cartilage and subchondral bone damage (a vs. b, p < 0.027), and proteoglycan loss (a vs. b vs. c, p < 0.043). Synovial inflammation declined by approximately 39% over time (without fully resolving, a vs. b vs. c, p < 0.024), while proteoglycan loss mirrored the evolution of the total histological score in the MI-RAT^© group. These results confirmed the efficient induction and expression of OA in the MI-RAT^© model and justified further analysis of the spinal peptidome.

Peptide identification:

Overall, the MS-based spinal cord peptidome approach uncovered 992 endogenous peptides across all groups (Figure 2A, Data File S2), with 624 peptides derived from 29 prohormone precursors (Table S1).

Fig. 2.

(A) Total number of endogenous spinal cord peptides identified and quantified in this study. (B) Distribution of identified peptides across various neuropeptide precursors. VGF: Neurosecretory protein VGF. (C) Box plots showing the distribution of m/z, peptide length, and GRAVY (grand average of hydropathy) values for the identified peptides, highlighting their physicochemical properties. (D) Sequence logo plot illustrating amino acid residue preferences at the N- and C-terminal regions flanking the identified peptides. (E) Dynamic range of peak areas for quantifiable peptides in the OA42 group (n = 6). Each peptide is represented by a blue dot, with gray error bars indicating standard deviation. Selected bioactive peptides are highlighted.

These identified peptides include not only classical neuropeptides (endogenous ligands that bind to G protein-coupled receptors) but also many proteolytic fragments of propeptides and peptide variants of known peptides (i.e., peptides containing parts of or overlapping with known neuropeptides). Some truncated mature neuropeptides or proteolytic fragments of propeptides may have yet undiscovered biological functions. The sample preparation process and the high sensitivity of the MS platform enabled the discovery of these typically low-abundance, diverse forms of truncated mature neuropeptides and proteolytic peptide fragments, broadening our understanding of endogenous peptide cleavage and degradation processes. These findings highlight the complexity of the peptidome and suggest that many potentially bioactive peptides remain to be characterized.

Among the identified prohormone precursors, several generated a particularly large number of peptides (Figure 2B). For instance, precursors such as secretogranin-2, proenkephalin-A, proSAAS, SST, and calcitonin each produced over 50 distinct peptides. The high number of peptides derived from these precursors may reflect their significant roles in spinal cord function and pain modulation.

The identified peptide distribution in the rat spinal cord shows distinct differences when compared to a previous peptidomic analysis of the rat brain from our lab³¹, which utilized a similar sample processing and mass spectrometry workflow. While some prohormone families, such as secretogranin-2, proenkephalin-A, and proSAAS, represented a similarly high percentage of the total identified peptides in both tissues, other precursors demonstrated clear tissue-specific patterns (Figure S1). Notably, calcitonin-derived peptides were frequently identified in the spinal cord but were undetected in the brain. Conversely, peptides from cholecystokinin were far more prevalent in the brain than in the spinal cord. These findings strongly suggest variations in prohormone localization and processing within the two major components of the rat CNS.

The properties of the identified peptides were examined (Figure 2C). The most prevalent forms were peptides with m/z values between 500 and 800 and lengths of 10 to 18 amino acids. This predominance suggests selective processing in the spinal cord, as well as MS’s inherent preferences for peptide selection and identification. The grand average of hydropathy (GRAVY) scores, which indicate peptide hydrophobicity, ranged from approximately –3 to 1, implying that the majority of peptides identified are relatively hydrophilic. This hydrophilicity may influence peptide solubility and could also reflect biases introduced during sample processing and selection.

Linear sequence motif analysis of the identified peptides revealed that the most prevalent amino acid residues flanking the N-termini or C-termini were dibasic residues (lysine and arginine) and leucine (Figure 2D). Basic residues can act as proteolytic cleavage sites for prohormone convertases, leading to their enrichment.³² The enrichment of leucine residues at cleavage sites in spinal cord peptides may indicate tissue-specific processing mechanisms or protease preferences active in the spinal cord, which remain to be further investigated. Together, these data demonstrate that spinal cord peptides exhibit specific cleavage patterns.

Peptide quantitation:

Furthermore, label-free quantitation based on LC-MS/MS-detected peptide peak areas was perfromed using PEAKS Studio Xpro software. This approach enabled the quantification of 411 peptides in the spinal cord across different experimental conditions (Figure 2A, Data File S3). These peptides exhibited a high dynamic range, spanning over two orders of magnitude, a hallmark of MS-based peptidomic analyses. This wide dynamic range underscores the sensitivity of the MS approach, allowing for the detection of both highly abundant and low-abundance peptides. Due to differences in endogenous abundance, ionization efficiencies, and mass spectrometer parameters, the peak areas of different endogenous peptides varied significantly (Figure 2E).

For subsequent analysis, the temporal subgroups of the control groups were consolidated. The Naive42 (n=5) and Naive84 (n=4) subgroups were pooled into a unified Naive cohort. This decision was supported by minimal differences in their histological scores (Table 1) and the observation that the vast majority of peptidome remained stable, with no widespread or systematic changes between the two time points (Figure S2A). Similarly, the Sham21 (n=2) and Sham35 (n=2) subgroups were combined into a unified Sham group. This approach was justified by two key factors: first, the minimal differences in their histological scores (Table 1) and the overall stability of the peptidome (Figure S2B); and second, the statistical necessity of increasing the sample size. Given the low number of replicates at each individual time point (n=2), pooling was essential to create a more statistically robust control group (n=4) for comparison against the OA groups. In contrast, the OA progression time points (OA21, OA35, OA42, and OA84) were retained as distinct subgroups to allow for the temporal analysis of OA-induced changes.

To investigate spinal cord peptidome responses to chronic OA pain, the K-means clustering of normalized peptide area ratios was conducted across six experimental groups (Naive, Sham, OA21, OA35, OA42, OA84). This analysis assessed changes in peptide levels across groups and over the course of OA progression. Peptides were grouped into three clusters (Cluster 1, Cluster 2, Cluster 3), indicating distinct expression patterns over time (Figure 3A).

Fig. 3.

(A) Heatmap displaying three clustered peptide expression changes across the Naive, Sham, and OA model groups (OA21, OA35, OA42, OA84). Color intensity represents the normalized peptide peak area ratio. This ratio was calculated for each peptide by dividing its mean peak area in a given group by the sum of its mean peak areas across all six groups. (B) Volcano plot comparing changes in prohormone-related peptide levels between the conditional groups (Sham and OA) and the Naive group. Thresholds for significant changes were set at a log₂ fold change of < –1 or > 1 and a false discovery rate adjusted q-value of < 0.2. (C) Box plot illustrating the quantified peak area of Substance P across experimental groups. The statistical significance for the OA groups was assessed relative to both the Naive group and the Sham group. p-values were calculated using Welch’s t-test and corrected for multiple comparisons using the Benjamini-Hochberg procedure to obtain q-values. Significant changes relative to both Naive and Sham groups are marked in red (q < 0.2).

Most quantified peptides (312 out of 411) were classified into Cluster 2, characterized by a gradual increase in peptide expression in the OA group, peaking at D42 before slightly decreasing at D84, indicating a marked response during OA progression. It is important to note that this temporal pattern was identified by the K-means clustering algorithm, which groups peptides with similar expression profiles across all conditions. Due to the high biological variability often observed in chronic pain models, the primary statistical tests in this study were designed to identify significant differences between the OA groups and the control groups, rather than between individual OA time points. The other two clusters showed different patterns: Cluster 1 peaked at D21, and Cluster 3 peaked at D84 in the MI-RAT^© model (Figure S3).

Significance testing (threshold: ≥ 2-fold change, a false discovery rate adjusted q-value of < 0.2 ) revealed widespread changes in peptide level across experimental groups compared to the Naive group (Figure 3B). Volcano plots showed that significantly upregulated peptides predominantly originated from the OA group over time. Focusing on the D42 peak observed in Cluster 2, 278 of the 312 peptides in this cluster exhibited significantly higher (>2-fold) peak areas in the OA42 group compared to the Naive group (Data File S3). For example, SP, a well-known pain facilitating neuropeptide, showed a significant increase in peak area from the Naive and Sham groups to OA42, remaining elevated at OA84 (Figure 3C), demonstrating the feasibility of this approach for characterizing pain-related peptides. Overall, 196 prohormone-derived peptides in Cluster 2 were significantly upregulated in the OA42 group compared to the Naive group (Data File S3). These included known neuropeptides like SP (Figure 3C) and cerebellin (CBLN) (Figure S4A); named peptides such as antrin (Figure S4B), BigLEN (Figure S4C), and little SAAS (Figure S4D); truncated forms of bioactive peptides like neuropeptide Y and Met-enkephalin-Arg-Gly-Leu; and many other peptides cleaved from regions flanking the prohormone (Table S2).

Another neuropeptide family previously reported to be closely related to pain modulation is the enkephalin-related peptides. In the fold-change and significance-filtered analysis, many endogenous peptides derived from pro-enkephalin A and pro-enkephalin B were clearly identified (Data File S2, S3). Many of which were truncated form of enkephalin-derived peptides. Two short neuropeptides Met-enkephalin (YGGFM) and Leu-enkephalin (YGGFL) may have failed to be detected due to their short sequences, limited fragmentation ion information, and the sequence length limitations of database search software. However, numerous C-terminally extended peptides containing the core Met-enkephalin sequence were identified and quantified. For example, the proenkephalin-A-derived peptide YGGFMRRVG was found to be significantly upregulated in the OA group at days 21, 42, and 84 when compared to both Naive and Sham controls (Figure S5). This further supports the close relationship between the enkephalin peptide family, opioid receptors, and OA progression in the MI-RAT^© model.

In MS-based peptidomics studies, the identification of disulfide bond-containing neuropeptides is particularly challenging, primarily because disulfide bonds hinder adequate dissociation of these neuropeptides. This results in less informative fragmentation ions, complicating peptide sequence matching. However, intramolecular disulfide bonds are widespread in neuropeptides, such as SST, CGRP, oxytocin, and vasopressin. Although these intact neuropeptides may be difficult to identify via MS, their MS-detected truncated forms can serve as useful indicators of expression levels across groups.

Moreover, two active SST neuropeptides -- somatostatin-14 (SST-14) and somatostatin-28 (SST-28) -- have been reported as products of SST precursors (Figure 4A).

Fig. 4.

(A) Schematic of somatostatin-28 (SST-28), highlighting the SST-28 (1–12) fragment and its cleavage site, marked with a red arrow. (C) Schematic of the calcitonin gene-related peptide 1 (CGRP 1) highlighting the CGRP 1 (18–37) fragment and its cleavage site, marked with a red arrow. (B, D) Box plots illustrating the quantified peak areas of the SST-28 (1–12) fragment and CGRP 1 (18–37) fragments across experimental groups. Statistical significance for the OA groups was assessed relative to both the Naive group and the Sham group. p-values were calculated using Welch’s t-test and corrected for multiple comparisons using the Benjamini-Hochberg procedure to obtain q-values. Significant changes relative to both Naive and Sham groups are marked in red (q < 0.2).

The SST-28 peptide includes a pair of basic residues adjacent to the SST-14 sequence, recognized as a cleavage site involved in prohormone processing. Cleavage at this dibasic residue generates SST-28 (1–12). Previous literature has shown that SST-28 (1–12) can serve as an excellent marker for the proSST system in mammalian tissue.³³ In this OA study, SST-28 (1–12) exhibited a typical Cluster 2 expression pattern (Figure 4B), suggesting that SST signaling pathways may be actively involved in pain modulation during OA progression. Similarly, another bioactive peptide cleaved from proSST, antrin, showed comparable expression changes to SST-28 (1–12) across experimental groups (Figure S4B).

CGRP, which contains 37 amino acid residue and one intracellular disulfide bond, also exists in two major forms known as CGRP 1 and CGRP 2 (Figure 4C).³⁴ Previous peptidome workflow identified Ser17-Arg18 as an in vivo cleavage site, possibly processed by insulin-degrading enzyme.³⁵ In this dataset, CGRP 1 (18–37), a proteolytic fragment of CGRP 1, was clearly identified. Its level peaked in the OA42 group (Figure 4D). Similarly, for CGRP 2, the corresponding proteolytic fragment showed a similar trend (Figure S6). The label-free quantitation results further support consistent expression changes in CGRP neuropeptides. The reliable detection of linear proteolytic fragments from cyclic neuropeptides may offer new insights into the degradation and regulation of neuropeptides.

Summary of results:

Translating findings from experimental pain models into effective clinical treatments for chronic pain remains challenging, partly due to limitations in both clinical trials and animal models. Previously, monosodium iodoacetate injection into the articular space of the rat stifle was a common OA pain model.³⁶ However, this model has been criticized for its acute and transient nature, and for being more representative of inflammatory and nociceptive rather than nociplastic pain characteristic of OA.^{17, 20, 37, 38} In contrast, the MI-RAT^© model demonstrates a progressive and persistent chronic OA pain phenotype, associated with structural damage to the stifle joints as well as neuroepigenetic and neuropeptidomic alterations. The phenotypic changes include behavioral and biomechanical alterations, peripheral sensitization (evidenced by mechanical hypersensitivity), central sensitization (indicated by a diminished response to mechanical temporal summation) and activation of endogenous pain inhibitory control (reflected by increased conditioned pain modulation).^{22, 23}

The current investigation into OA-induced alterations in the neuropeptidome has provided valuable insights. Notably, this study validated the use of LC-MS/MS for the identification and quantification of a broader range of spinal peptides regulated by chronic pain in a rat OA model. To our knowledge, the spinal cord peptidome in this context has not been previously characterized. In this study, 992 peptides with defined sequences and post-translational modifications were identified, and 411 of them were quantified, marking a significant advancement. Furthermore, several OA-regulated peptides and their associated families were identified, establishing a strong foundation for future research aimed at elucidating the mechanisms through which these peptides mediate OA-related effects on the CNS. The significant increase of pain-related spinal cord peptides in Cluster 2 aligns with their known roles in pain transmission and modulation, suggesting that these peptides may contribute to the development and maintenance of chronic pain in OA. The potential functional relevance of these peptides is further discussed in the following section (§ Functional considerations).

Additionally, the evolution of peptide clustering (Clusters 1, 2, and 3), as revealed by LC-MS/MS analysis in the MI-RAT^© model, may correspond to the behavioral and physiological chronicization of pain – an area that warrants further investigations. Indeed, as reported in previous studies, the pain phenotype of the MI-RAT^© model has shown three distinct phases of somatosensory sensitivity: i) an early phase (up to D21) characterized by increased sensitivity, possibly corresponding to acute postoperative pain; ii) an intermediate phase (D35 to D49) with reduced tactile sensitivity, potentially reflecting CNS adaptation to pain; and iii) a late phase (around D56) with sustained tactile sensitivity, which may indicate the development of chronic pain.^{22, 23} Moreover, these earlier studies demonstrated that the MI-RAT^© model exhibited activation of endogenous inhibitory pain control at D21 and D35, which disappeared at D49 and D56, possibly due to fatigue of the control system, as has been observed in human OA pain phenotype.^{23, 39}

Comparison with prior MS-based literature on spinal cord peptidome:

Research utilizing MS to profile molecular alterations in the rat spinal cord has been extensively documented. In the realm of proteomics, spinal cord proteome studies under various disease and injury conditions have been reported.^40–43 These investigations predominantly focused on the global proteome, encompassing the full complement of proteins within the tissue. Notably, the proteins analyzed in these studies were not restricted to prohormones and typically exhibit large molecular masses, often requiring enzymatic digestion (e.g., trypsin) to generate smaller fragments compatible with MS analysis.

In contrast, the spinal cord peptidome, defined as the comprehensive suite of short, endogenous peptides derived from the proteolytic cleavage of prohormone precursors by endogenous processing enzymes within the spinal cord, remains underexplored. While targeted analyses of neuropeptides in the spinal cord have been conducted by Beaudry and colleagues,^{16–23, 44, 45} and Sweedler and coworkers have investigated quantitative peptidomic changes in mouse dorsal root ganglia and dorsal horn tissues in chronic itch models,^{46, 47} broader profiling of the spinal cord peptidome is still lacking. This gap underscores the need for systematic investigations into the spinal cord peptidome, which are addressed in this study through a workflow and experimental design specifically optimized for comprehensive peptidome characterization.

Notably, the current study diverges fundamentally from prior investigations in both experimental design and biological insights. This work provides the first critical perspective on time-dependent peptidome reorganization in the spinal cord during chronic OA pain, a complex neurological and pathological state. Unlike conventional cross-sectional approaches that contrast single disease or stress conditions with controls, this longitudinal analysis delineates dynamic peptide fluctuations across multiple stages of OA progression, capturing evolving molecular signatures that align with disease chronicity.

Methodologically, the integration of advanced MS instrumentation with optimized sample preparation workflows enabled the high-confidence identification of nearly 1,000 distinct peptides within the spinal cord—a marked increase in coverage compared to earlier studies.⁴⁴ Strikingly, a significant proportion of these peptides has not been previously reported in the context of OA (e.g. antrin, a peptide derived from proSST), underscoring the novelty of the current findings. Consequently, this study delivers unprecedented depth and breadth in peptidomic profiling, revealing nuanced alterations across diverse neuropeptide families and their potential roles in OA-driven neurobiological adaptations.

Functional considerations:

The analysis of the spinal peptidome highlighted the significant impact of several targeted neuropeptides in the MI-RAT^© model, such as SP, CGRP, SST and enkephalins-- previously reported as upregulated,^{22, 23} – as well as others newly identified, such as antrin, CBLN. The SP, a cleavage product from the tachykinin family, plays a prominent role in the neurotransmission of nociceptive (pain) stimuli.^{48, 49} SP is ubiquitously distributed across mammalian tissues and body fluids, reflecting its diverse physiological functions. Within the nervous system, the highest concentrations of SP are found in the dorsal horn of the spinal cord, the substantia nigra, and the amygdala.⁵⁰ In the spinal cord, the selective expression of SP in the dorsal root ganglia and dorsal horn (but not in the ventral horn) supports its critical involvement in primary afferent sensory neurotransmission, particularly in pain pathways.^{51, 52}

Tachykinin family members bind to neurokinin receptors, with SP exerting its effects primarily through the neurokinin-1 (NK1) receptor, leading to modulation of pain perception and transmission.^{49, 53} Experimental studies have demonstrated that inhibiting SP release or blocking its interaction with NK1 receptors produces analgesia following spinal injections.⁵⁴ These findings have spurred the investigation of compounds that inhibit SP’s action as potential analgesic drugs. Although early clinical trials with NK1 receptor antagonists have shown mixed results regarding their efficacy in human pain management, ongoing research continues to explore and refine these therapeutic strategies to develop effective treatments for pain and other SP-related disorders.^{55, 56}

Another important neuropeptide implicated in the development and maintenance of neuropathic pain is CGRP, which is produced through alternative RNA processing of the calcitonin gene.³⁴ The highest levels of CGRP are found in the outer layers of the spinal cord dorsal horn and the trigeminal nucleus caudalis, corresponding to the central terminals of primary afferent neurons originating from the dorsal root and trigeminal ganglia, respectively.⁵⁷ CGRP is commonly co-localized with SP.⁵⁸ Measured CGRP levels have been closely associated with various types of pain, including somatic, visceral, neuropathic, and inflammatory pain, and OA.^{59, 60} In particular, CGRP receptor antagonists and monoclonal antibodies targeting CGRP or its receptor have become mainstream treatments for migraine, while CGRP’s potential in treating other types of pain remains under investigation.^{56, 61} Moreover, a recent study showed that a CGRP receptor antagonist (rimegepant) could reduce joint damage and phenotypic alterations due to OA in a mouse model.⁶²

Although SP and CGRP are considered crucial to pain sensation, a recent study found that Tac1 and Calca double-knockout mice still displayed intact responses to pain stimuli.⁶³ Additionally, the loss of these two peptides did not affect chronic pain or neurogenic inflammation.⁶³ These findings suggest that other signaling molecules also contribute to the pain response process.

Some pain-relieving neuropeptides were also clearly identified and found to be elevated at D42 in the MI-RAT^© model, closely associated with OA progression. Among these, CBLN emerged as a notable candidate. In the literature, CBLN is known to regulate the formation, differentiation, and maintenance of neuronal synapses.⁶⁴ Additionally, peptides derived from CBLN1 and CBLN2 have been shown to facilitate pain signaling by binding to glutamate receptors.⁶⁵

In rats, the endogenous opioid system comprises several precursor proteins that give rise to active opioid peptides. The two primary precursors are proenkephalin-A and proenkephalin-B (also known as prodynorphin). The active peptides include Met-enkephalin, Leu-enkephalin and several extended enkephalin forms, such as synenkephalin and alpha-neoendorphin. In the MI-RAT^© model, elevated levels of extended enkephalin peptides were also observed, consistent with previous reports.²² The binding of proenkephalin-derived peptides to opioid receptors can reduce pain transmission.⁶⁶

Similar to opioid peptides, SST is also highly expressed in the CNS. Beyond its well-known inhibitory effects on cell proliferation and hormone secretion, SST also plays a role in regulating inflammation and pain.⁶⁷ The SST analogue octreotide has been shown to produce analgesic effects in post-operative and cancer-related pain.⁶⁸ SST modulates pain by binding to somatostatin receptors, particularly somatostatin receptor 4 (SSTR4).⁶⁹ Targeting SST receptors can yield potent anti-inflammatory and analgesic effects. The SST and opioid systems are the two primary inhibitory systems in mammals. Developing SSTR4 agonists represents a promising non-opioid strategy for managing chronic neuropathic, inflammatory, and mixed pain. However, SST appears to play a paradoxical role: while it inhibits pain signals, this inhibition may also lead to spinal disinhibition and facilitation of pain transmission.⁷⁰ Interestingly, a rare peptide derived from proSST(1–10), named antrin, emerged from the analysis in the MI-RAT^© model. Antrin was initially discovered in the secretory granules of gastric delta cells, and no studies have reported on this neuropeptide since 1993.^{71, 72} The study is the first to report the presence of the antrin neuropeptide in the rat spinal cord in a model of OA pain.

Furthermore, peptides derived from proSAAS were significantly upregulated in the MI-RAT^© model. These peptides are generally associated with food intake and body weight regulation.⁷³ Specifically, BigLEN has been reported to bind to GPR171 to mediate its feeding regulatory effects.⁷⁴ Another proSAAS-derived peptide, little SAAS, lacks an identified receptor but is known to play a significant role within the neuropeptidergic system of suprachiasmatic nucleus.^{73, 75} The observed changes in BigLEN and little SAAS levels mirrored those of SP, CBLN and antrin, suggesting that proSAAS-derived peptides may also play a significant role in pain modulation in the MI-RAT^© model.

CONCLUSIONS

This study provides novel insights into the reorganization of the spinal cord peptidome in response to OA. The dynamic changes observed in neuropeptide levels underscore the critical role of the spinal cord peptidome in OA-related chronic pain. These findings not only highlight the sensitivity and efficacy of the label-free tandem MS platform in detecting and quantifying these peptides but also open avenues for developing targeted therapeutic interventions through neuropeptide modulation. Moreover, these insights extend beyond OA, potentially offering valuable strategies for modulating pain-related pathways in other chronic conditions

DATA REPORTING

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE⁷⁶ partner repository with the dataset identifier PXD060558.

(Reviewer access details: Project accession: PXD060558, Token: RGtd3btGTMDb)

Supplementary Material

Supplemental Information

Table S1. List of protein precursors identified in this study.

Table S2. Prohormone-derived peptides significantly altered in the OA42 group compared to the Naive group in Cluster 2.

Fig. S1. Bar chart of the proportion of identified peptides from specific protein families in the brain and spinal cord.

Fig. S2. Volcano plot of prohormone-related peptide level changes between Naive84 vs. Naive42 and Sham35 vs. Sham21

Fig. S3. Normalized peptide peak area ratio trends for peptides in three clusters.

Fig. S4. Quantitative analysis of antrin, cerebellin, BigLEN, and little SAAS levels across experimental groups.

Fig. S5. Quantitative analysis of peptide YGGFMRRVG across experimental groups.

Fig. S6. Cleavage pattern and quantitative analysis of CGRP 2 (18–37) fragment levels across experimental groups.

Supplemental Data File S1

Supplemental Data file S1. Excel spreadsheet detailing the histological analysis results of experimental groups.

Supplemental Data File S2

Supplemental Data file S2. Excel spreadsheet detailing the identified peptides from rat spinal cords.

Supplemental Data File S3

Supplemental Data file S3. Excel spreadsheet detailing the label-free quantitation results of peptides from rat spinal cords