How Does Tendon Region, Donor, and the Presence of Disease Affect Protein Composition of the Achilles Tendon?

Abstract

Background

Response to treatment for tendinopathy is variable, which may reflect variability in underlying etiology and capacity for the tendon to respond to treatment. Understanding variability in tendon protein composition may help improve our understanding of the mechanistic underpinnings of painful tendon degeneration and inform treatment targets.

Questions/purposes

(1) What factors (tendon region, individual characteristics, presence of disease) contribute to protein compositional (proteomic) and structural variation in human Achilles tendons? (2) What compositional changes characterize tendinopathy, and what protein interactions might contribute to tendon degeneration? (3) How does diabetes influence tendon composition, and what mechanisms might underlie tendon dysfunction in individuals with diabetes?

Methods

In this exploratory, cross-sectional study, human Achilles tendon specimens were obtained from individuals with (diabetes group, n = 5) or without diabetes (control group, n = 5) undergoing lower extremity amputation and from individuals undergoing tendon debridement surgeries for tendinopathy (tendinopathy group, n = 8). Specimens were collected between 2019 and 2023. Protein abundances were quantified and analyzed using mass spectrometry, hierarchical clustering, and principal component analysis. To evaluate the role of tendon region and donor on tendon protein compositional variability, we assessed proteomic differences between three regions in nontendinopathic tendons from three individuals. To identify the contribution of disease (that is, presence of tendinopathy or diabetes) on protein composition, we compared tendons from the tendinopathy (n = 8 [2 males, 6 females], mean ± SD age 48 ± 11 years), diabetes (n = 5 [3 males, 2 females], age 54 ± 9 years), and control (n = 5 [3 males, 2 females], age 42 ± 12 years) groups. Proteomic differences associated with tendinopathy and diabetes were further examined using functional enrichment and protein-protein interaction network analysis.

Results

Variability in tendon protein composition was primarily from presence of disease, followed by donor and then tendon region. Protein composition distinguished tendons with tendinopathy from controls, with 311 proteins differentially expressed (152 overexpressed and 159 underexpressed; fold change ≥ 1.5, p < 0.05) and higher Bonar scores indicating greater degeneration (mean ± SD Bonar score tendinopathy group 8.6 ± 1.2 versus control group 2.1 ± 0.7; p = 0.01). Pathway analysis identified dysregulation in extracellular matrix remodeling (TIMP1, MMP3, MMP10), inflammatory response (TNF-α, EGFR1), and metabolic reprogramming. Tendons from individuals with diabetes exhibited minimal proteomic changes compared with the control group, with 66 differentially expressed proteins (31 overexpressed and 35 underexpressed; fold change ≥ 1.5, p < 0.05) with no histopathologic differences between diabetes and control group tendons (mean ± SD Bonar score diabetes group 3.4 ± 1.0 versus control group 2.1 ± 0.7; p = 0.19). Tendons in the diabetes group showed reductions in Type I collagen, enrichment of pathways associated with fibrosis and metabolic dysfunction, and inflammatory pathways associated with α₆β₄ integrin.

Conclusion

Our findings indicate that Achilles tendon composition primarily differs based on disease etiology, with tendinopathy showing extensive extracellular matrix disruption and inflammatory activity, whereas tendons from individuals with diabetes exhibit more subtle compositional changes. This distinction suggests that tendinopathy may require targeted interventions addressing tissue remodeling and inflammation, whereas diabetes may predispose tendons to injury but not directly result in degeneration. Understanding these protein compositional variations can help refine hypotheses about disease progression, treatment response, and potential therapeutic targets.

Clinical Relevance

While proteomic analysis is not currently a part of routine clinical assessment, these findings provide a framework for identifying protein markers that may aid in early diagnosis or patient stratification to improve treatment alignment. Future studies could determine whether these proteomic changes correlate with treatment response and further inform our understanding of early-stage degeneration from chronic disease. By bridging molecular findings with clinical presentation, this study lays the groundwork for future research on precision medicine approaches for tendon disorders, with the long-term goal of tailoring treatment based on both biological and symptomatic characteristics.

Introduction

Achilles tendinopathy is characterized by chronic tendon pain and impaired function, often leading to prolonged rehabilitation and inconsistent patient outcomes [18, 58]. Tendon healing is inherently slow, and even with current best-evidence treatment, patients exhibit high variability in their responses to treatment, with some successfully returning to their prior levels of function and others not achieving this goal [58]. Clinically, this variability can lead to prolonged treatment courses in some patients with limited improvement in symptoms and function. Thus, there is an urgent need to better understand the factors driving patient heterogeneity in presentation and response to treatment, which is an important component of efficiently aligning treatment and is a goal of precision medicine. Prior studies suggest that both macroscopic (demographic, imaging, and functional performance measures) [25, 57] and microscopic (histologic and biochemical markers of degeneration) [43] factors contribute to disease variability. This heterogeneity in disease pathomechanisms has been characterized in prior studies using diagnostic imaging and histologic findings [28, 72]. For example, insertional Achilles tendinopathy may involve bony changes, bursal irritation, and intrasubstance degeneration [72]. “Tendinopathy” is increasingly recognized as an umbrella term encompassing multiple distinct pathophysiologies, further complicating the diagnosis and management. Hence, a single diagnostic label may obscure meaningful differences in disease mechanism, highlighting the need for a more nuanced approach.

Despite advancements in imaging and histologic characterization, the mechanisms underlying Achilles tendinopathy remain poorly understood. Prior research has shown that tendon degeneration involves loss of collagen organization, hypercellularity or hypocellularity, proteoglycan and glycosaminoglycan accumulation, and vascular infiltration [23, 28, 29, 37, 43], yet these changes occur heterogeneously within and between tendons. Moreover, systemic factors such as diabetes and metabolic syndrome may influence tendon health [46, 60], but their impact on tendon composition remains unclear. Understanding how tendon protein composition varies with disease state and tendon region is essential to identifying mechanisms underlying tendon degeneration and dysfunction. This study seeks to address key gaps in knowledge by characterizing the heterogeneity of the protein composition of the Achilles tendon with respect to tendon region (core versus periphery), donor, and disease status (tendinopathy versus diabetes versus control). By integrating quantitative proteomics and histopathologic analysis, we aim to provide novel insights into biological pathways associated with tendon degeneration, potentially guiding more targeted diagnostic and therapeutic strategies.

We therefore asked: (1) What factors (tendon region, individual characteristics, presence of disease) contribute to compositional (proteomic) and structural variation in human Achilles tendons? (2) What compositional changes characterize tendinopathy, and what protein interactions might contribute to tendon degeneration? (3) How does diabetes influence tendon composition, and what mechanisms might underlie tendon dysfunction in individuals with diabetes?

Materials and Methods

Overview of Study Design

The goal of this cross-sectional study was to assess the protein composition (using proteomics via mass spectroscopy) in human Achilles tendons to improve our understanding of tendon degeneration mechanisms using a data-driven approach. We were interested in understanding how tendon region, donor, and presence of disease (that is, diabetes or tendinopathy) impact tendon protein composition.

To understand the effect of tendon region and donor on variability of tendon protein composition, Achilles tendon samples were collected from three individuals without known tendon injury. Each tendon was divided into three regions: the left periphery, the core, and the right periphery. To understand the effect of disease (tendinopathy or diabetes) on variability of tendon protein composition, Achilles tendon samples were collected from three groups of individuals: individuals with tendinopathy (tendinopathy group, n = 8), individuals with diabetes (diabetes group, n = 5), and tendons from individuals without tendinopathy and without diabetes (control group, n = 5).

We present clinical data gathered from participants and their medical records along with histologic analysis [23, 37] to characterize the severity of patient presentation and degeneration of the tendon specimens. Proteomics was used to quantify protein abundances in tendon specimens (Fig. 1). Bioinformatics and statistical analysis were used to assess contributors to variability in tendon protein composition (region, donor, presence of disease). Functional enrichment analysis was used to suggest mechanisms of tendon degeneration.

Fig. 1.

Overview of proteomics workflow showing specimen collection, tissue preparation, mass spectrometry (MS) of tagged peptides, and data analysis. (Step 1) Human Achilles tendon samples are collected and (Step 2) processed using a Covaris cryopulverizer for tissue lysis. (Step 3) Proteins are extracted with a urea buffer and (Step 4) digested to generate peptides. (Step 5) Peptides are labeled with TMT-10 isobaric mass tags and (Step 6) combined for simultaneous MS analysis. (Step 7) Protein quantification is performed by LC-MS using a Thermo Fisher Scientific Q-Exactive Plus™ mass spectrometer. (Step 8) Protein identification and data analysis are conducted using Proteome Discoverer and Mascot, followed by PCA and cluster analysis to interpret the results. A color image accompanies the online version of this article.

Proteomics data associated with this publication is freely available for the broad scientific community at Digital Commons Data@Becker Repository (DOI: 10.17632/cg3zcvrdy7.1). A Data Transfer and Use Agreement may be needed for qualified researchers to access the data set. The data set meets PRIDE database’s definition of a complete proteomics data set.

Ethical Approval

We obtained ethical approval from the Washington University School of Medicine in St. Louis Institutional Review Board (protocol ID 202111069 and 201505110). All participants gave their written informed consent. All procedures were performed in accordance with the ethical standards of the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Study Population and Selection Process

Participants were enrolled between April 2019 and August 2023. All participants needed to be at least 18 years of age. Participants in the diabetes group needed to have a diagnosis of Type 2 diabetes. Exclusion criteria included pregnancy, previous amputations that had removed the Achilles tendon tissue, active infection extending into the tendon, and certain infectious diseases (history of human immunodeficiency virus or hepatitis C, current methicillin-resistant Staphylococcus aureus, vancomycin-resistant enterococcus, or COVID-19 infection). Additional exclusion criteria for individuals in the diabetes and control groups included presence of macroscopic tendon injury or tendon thickening on visual inspection at the time of tissue collection. Additional exclusion criteria for the tendinopathy group included history of collagen disorders (such as Ehlers-Danlos syndrome), rheumatoid arthritis, or lupus.

For the diabetes and control groups, tissue used for this analysis was collected as part of a larger tissue procurement study. For the diabetes group, 5 of 10 total tissue specimens were available for proteomic analysis. For the control group, 5 of 7 total tissue specimens were available for proteomic analysis. For the tendinopathy group, participants scheduled for Achilles tendon debridement with repair were approached about the study. Only patients having surgeries with study-affiliated surgeons, with surgeries scheduled at a time when the study team was available to collect debrided tissues, were informed about the study. Two individuals deferred participation in the study, leaving 8 of 10 eligible participants enrolled. Therefore, this study population represents a convenience sample of a subset of eligible individuals where tissue specimens were available rather than a complete cohort of all patients treated at our institution during this period. Because of the limited availability and variability between surgical specimens, matching between the groups was not feasible.

Participant Demographics and Clinical Characteristics

Among individuals in the control group (Table 1), reasons for amputation included congenital foot deformity (1 of 5) and chronic lower extremity pain with or without history of trauma (4 of 5). Among individuals in the diabetes group, reasons for amputation included Charcot deformity (5 of 5), with associated complications such as nonhealing ulcers (2 of 5) and infection (1 of 5). The most recent hemoglobin A1c results (taken within 1 year prior to surgery) were extracted from the medical record where available. The median hemoglobin A1c level was 7.1% (range 6.3% to 8.8%), reflecting glycemic control around the time of surgery.

Table 1.

Participants’ characteristics

	Diabetes (n = 5)	Control (n = 5)	Tendinopathy (n = 8)
Age in years	54 ± 9	42 ± 12	48 ± 11
Sex
Male	3	3	2
Female	2	2	6
Height in cm	185.9 ± 9.2	174.2 ± 7.8	164.1 ± 8.7
Weight in kg	129.2 ± 30.6	79.7 ± 14.8	73.6 ± 24.1
BMI in kg/m2	36.8 ± 6.2	29.6 ± 5.0	26.9 ± 7.5
Hemoglobin A1c in %	7.28 ± 0.83	-	-
VISA-A score (n = 7 participants)	-	-	38 ± 20
Severity of symptoms (NPRS) (n = 7 participants)	-	-	7.0 ± 2.6
Symptom duration in months (n = 6 participants)	-	-	68.7 ± 114.1

Data presented as mean ± SD or the number. VISA-A scores range from 0 to 100, with lower scores indicating greater symptom severity and functional impairment. NPRS scores range from 0 to 10, where 0 represents “no pain” and 10 represents “worst imaginable pain.”

All participants in the tendinopathy group received radiographs and/or MRI for confirmation of diagnosis and presurgical planning as part of standard of care provided by the orthopaedic surgeon. Three reported being avid runners prior to symptom onset; however, running was currently pain limited. Comorbidities included: contralateral insertional Achilles tendon debridement (2 of 8), ipsilateral Haglunds deformity resection without tendon debridement or repair (1 of 8), Stage II chronic kidney disease (1 of 8), and history of spinal surgery (1 of 8). Participants reported using statins (2 of 8) and oral steroids in the last 6 months (1 of 8). No participant had used fluoroquinolones in the last 6 months. Participants reported prior treatments including physical therapy, walking boot immobilization, platelet-rich plasma, steroids, and antiinflammatory agents.

Participants in the tendinopathy group were administered the Victorian Institute of Sport Assessment – Achilles (VISA-A) questionnaire [53] before surgery as a self-reported measure of their pain and symptoms. This is a diagnosis-specific questionnaire with scores ranging from 0 to 100, with lower scores indicating more severe pain and functional impairment. The median (range) VISA-A score was 41 (10 to 61). The Numeric Pain Rating Scale (NPRS) was used to assess the participant’s worst pain in the last week. The scale ranged from 0 to 10, where 0 represents “no pain” and 10 represents “worst pain imaginable.” The median (range) NPRS score was 7 (3 to 10) (n = 6 patients who reported NPRS ratings). Participants in the tendinopathy group reported a median (range) symptom duration of 36 months (6 to 300) (n = 6 patients who reported symptom duration).

Sample Collection

For the tendinopathy group, tendon specimens were obtained from tissue sections debrided by the surgeon (JB, JM, AT) as part of routine surgical procedure during Achilles tendon debridement with repair surgery and passed to a member of the study team (RSV, JAZ). For the diabetes and control groups, tendon samples were obtained during lower extremity amputation procedures. A member of the study team (JAZ) immediately dissected the Achilles tendon after surgical removal of the affected limb. All specimens were wrapped in phosphate-buffered saline-soaked gauze, transported on ice, and stored at -80°C until analysis.

To assess regional variability, a 1-mm transection was taken from the distal aspect of each Achilles tendon and divided into three regions: two periphery (noted as 1 and 3) and one central “core” specimen (noted as 2). The core specimen was used for all primary analyses, while the peripheral and core specimens were compared in the regional analysis. An additional 1-mm transection from the distal aspect of the remaining surgical tendon specimen was taken and fixed in neutral buffered formalin for 24 to 48 hours for histologic processing.

Histology

To characterize severity of tendon tissue degeneration, all Achilles tendon specimens underwent histologic processing and scoring with the Bonar score [23, 37], where higher scores indicate greater degeneration severity. Tendon samples were fixed in 4% paraformaldehyde for 48 hours, underwent dehydration with ethanol, were embedded in paraffin wax, and 5-μm longitudinal sections were prepared and stained with hematoxylin and eosin (Leica Biosystems) for general tissue morphology and alcian blue (Sigma-Aldrich) for glycosaminoglycan accumulation [23, 37]. Sections were independently evaluated by four raters (including RSV, NH). Reliability of the Bonar score has been previously established [23]. The intraclass correlation coefficient (2-way mixed, absolute agreement) between raters on a subset of histologic specimens (n = 7) was 0.71 (95% confidence interval [CI] 0.37 to 0.93). Each rater reviewed and scored the sections twice, and inconsistencies in scoring of individual items for each section were resolved by discussion between raters until consensus was achieved. Raters were blinded to diabetes status but not tendinopathy status at time of evaluation. Because of inadequate fixation, sections from two tendons in the diabetes group and two tendons from the control group were excluded from statistical analysis.

Protein Extraction and Preparation for Proteomic Analysis

To identify differentially expressed proteins between the tendinopathy, diabetes, and control groups, we performed proteomic analysis using tandem mass tag (TMT)–based quantitative mass spectrometry (Fig. 1). This allowed us to quantify protein abundances across the three groups. Proteomic validation analysis was performed to ensure data quality (Supplemental Fig. 1 and Appendix 1; http://links.lww.com/CORR/B430).

Dissected human tendon samples were frozen in liquid nitrogen in 1.7-mL tubes, transferred to cryopulverization tubes (tissueTUBE TT05 [Covaris]) and cryopulverized using the CP02 CryoPrep system (Covaris) at an impact level of 5 to ensure uniform sample homogenization [41]; 100 µL of lysis buffer (50 mM of tris HCl [pH 8.0], containing 8 M of urea, 75 mM of NaCl, 1 mM of ethylenediaminetetraacetic acid (EDTA), 10 mM of NaF, phosphatase inhibitor cocktail 2 [1:100] and cocktail 3 [1:100], 2 μg/mL of aprotinin, 10 μg/mL of leupeptin, and 1 mM of PMSF [pH 8.0]) was added to each sample. Each sample was then transferred to sonication vials (Covaris) and sonicated in the Covaris S220X for two cycles of 8 minutes at 8°C with the following settings: PIP 100 watts, cycles per burst 500, and duty factor 10%. Sonicated samples were transferred to 1.7-mL Eppendorf tubes (Axygen) and centrifuged at 14,000 RCF for 10 minutes to pellet debris. The resultant supernatant was collected for analysis. Protein concentration was determined using a bicinchoninic assay according to kit instructions (Pierce), and the remaining lysate was stored at -80 °C.

To prepare samples for digestion, 100 µg of total protein per sample was reduced with 10 mM of dithiothreitol (Pierce) at room temperature for 60 minutes and then alkylated with 4 µL of 250 mM of iodoacetamide (Pierce) in the dark at room temperature for 60 minutes. The urea concentration was diluted to 1.5 M using 50 mM of tris HCl buffer (pH 8.0) before adding 10 µg of hyaluronidase (Millipore Sigma) per sample to breakdown glycosaminoglycans [8]. Samples were incubated at room temperature for 1 hour, followed by digestion with 2.0 µL of LysC (1 µAU) in 100 mM of tris HCl buffer (pH 8) for 2 hours to initiate protein cleavage while maintaining enzyme activity in the presence of residual urea. Trypsin (1:20) was then added, and samples were incubated overnight at room temperature to ensure complete digestion into peptides suitable for mass spectrometry analysis. To remove high–molecular weight contaminants (MWCO), samples were filtered through a 10,000 MWCO filter (Millipore) and then acidified to 1% (vol/vol) trifluoroacetic acid (Sigma) and desalted using stage tips [40]. The peptides were eluted with 60 µL of 50% (vol/vol) acetonitrile (J. T. Baker) in 0.1% (vol/vol) trifluoroacetic acid and dried in a Savant DNA 120 concentrator Speed-Vac (Thermo Scientific). Dried peptides were dissolved in 20 µL of 1% (vol/vol) acetonitrile in water, with 10% of each sample removed for quantification using the Pierce Quantitative Fluorometric Peptide Assay kit (Thermo Scientific). An aliquot of peptides (1 µg per sample) was transferred to 1.7-mL tubes (Axygen), dried, and stored at -80ºC for TMT labeling.

Preparation of TMT-labeled Peptides and Nanoliquid Chromatography Tandem Mass Spectrometry

To compare protein expression across the tendinopathy, diabetes, and control groups, peptides were labeled using TMT-based mass spectrometry. This process allowed for simultaneous quantification of proteins from different samples in a single analysis while preserving individual sample identities.

For TMT labeling, the lyophilized peptides were dissolved in 40 µL of HEPES buffer (Alfa Aesar) (100 mM, pH 8.5) and labeled according to the vendor protocol using the TMT-10 reagent kit (Thermo Fisher Scientific) This step enabled multiplexed quantification of proteins across all samples. To remove salts and contaminants that could interfere with downstream mass spectrometry analysis, the labeled peptides were combined, dried, and dissolved in 120 µL of 1% (vol/vol) formic acid (Sigma-Aldrich) before desalting using stage tips. The labeled peptides were then analyzed using high-resolution nanoliquid chromatography tandem mass spectrometry. Chromatographic separation was performed with an Acclaim PepMap 1000 C18 RSLC column (75 µm inner diameter × 50 cm) (Thermo Fisher Scientific) on an EASY nanoLC (Thermo Fisher Scientific). The column was equilibrated with 11 µL of solvent A: 1% (vol/vol) formic acid at a pressure of 700 bar. The samples in 1% (vol/vol) formic acid were loaded (2.5 µL) onto the column with 1% (vol/vol) formic acid at 700 bars. Peptide chromatography was initiated with mobile phase A (1% formic acid) containing 5% solvent B (100% acetonitrile, 1% formic acid) for 5 minutes, and then increased to 25% B over 195 minutes, to 35% B over 40 minutes, to 70% B over 6 minutes, to 95% B over 2 minutes, and held at 95% B for 18 minutes, with a flow rate of 250 nL/minutes.

Spectral acquisition was performed in data-dependent acquisition mode. The full-scan mass spectra were acquired using the Orbitrap mass analyzer (Thermo Scientific) with a scan range of m/z = 350 to 1800 and a resolving power set to 70,000. Twelve data-dependent high-energy collision dissociations were performed with a mass resolving power set to 35,000, a fixed lower value of m/z 100, an isolation width of 1.2 Da, and a normalized collision energy setting of 32. Maximum injection time was 60 milliseconds for parent-ion analysis and 120 milliseconds for product-ion analysis. The target ions that were selected for MS2 were dynamically excluded for 40 seconds. Automatic gain control was set at a target value of 3×10⁶ ions for MS1 scans and 1×10⁵ ions for MS2 analysis. Peptide ions with charge states of +1 or +7 or greater were excluded for high-energy collision dissociation acquisition.

The machine data from the analysis of isobarically labeled peptides, using the Q-Exactive mass spectrometer (Thermo Scientific), were converted to peak lists using Proteome Discoverer, version 2.1.0.81 (Thermo Scientific). Mascot software, version 2.8.0.1 (Matrix Science) [47] was used to search MS2 spectra against a UniProt (downloaded January 2023) database of human proteins (20,539 entries) and common contaminant proteins (common Repository of Adventitious Proteins [cRAP], version 1.0, January 1, 2012; 116 entries). Data were searched with a fragment ion mass tolerance of 20 ppm and a parent ion tolerance of 20 ppm. Enzyme was specified as trypsin/P with allowance of a maximum of four missed cleavages. Carbamidomethylation of cysteine was specified as a fixed modification. Deamidation of asparagine, formation of pyroglutamic acid from N-terminal glutamine, acetylation of protein N-terminus, and oxidation of methionine were specified as variable modifications. Peptide spectral matches were filtered at 1% false-discovery rate by searching against a reversed database, and the ascribed peptide and protein identities were accepted. A minimum of two peptides with unique sequences was required to accept a protein identification. To reduce the interference of blood subproteomes, 154 common blood proteins were removed from further analysis.

Analytic Approaches and Statistical Analyses

To identify proteomic differences between the tendinopathy, diabetes, and control groups, we performed statistical normalization, quality assessment, and bioinformatics analyses. These steps ensured that the data were reliable, comparable across groups, and provided insights into tendon degeneration mechanisms of meaningful magnitude.

Normalization and Quality Assessment

To ensure accurate and unbiased protein quantification, we processed and normalized multiplex TMT data using proteoQ, version 1.7.5.1 (https://github.com/qzhang503/proteoQ), an open-source tool developed with the tidyverse approach [69] in R [49] and RStudio [55]. For data transformation and outlier removal, the raw precursor intensities were converted to logarithmic ratios (base 2) relative to the average precursor intensity across all samples. Within each sample, outliers were identified and removed using the Dixon test, applied recursively to peptides with greater than two identifying peptide spectral matches [33]. The median of the ratios of peptide spectral matches assigned to the same peptide was used to derive peptide-level values, which were then aggregated at the protein level. To align protein ratios across samples, likelihood functions were first estimated for the log ratios of proteins using finite mixture modeling, assuming two-component Gaussian mixtures [10]. The ratio distributions were then aligned so that the maximum likelihood of log ratios was centered at 0 for each sample. Scaling normalization was performed to standardize the log ratios of proteins across all samples. To reduce the influence of extreme values from either log ratios or reporter-ion intensities, only the values between the 5th and 95th percentile of log ratios and 5th and 95th percentile of intensity were used in the calculations of SDs.

Bioinformatic and Statistical Analysis

To determine how protein composition varied between groups, we conducted dimensionality reduction and statistical testing. Metric multidimensional scaling and principal component analysis (PCA) of protein log₂ ratios were performed with the base R function stats:cmdscale and stats:prcomp, respectively. Dendrograms and heatmaps were generated with pheatmap [32] to visualize clustering of samples by protein log₂ ratios. To assess the statistical significance in protein abundance differences between indicated groups of contrasts, we applied linear modeling using the contrast fit approach in limma [52]. Because this study emphasizes discovery and includes a small number of specimens, we initially did not correct the p value for multiple comparisons. This allowed us to identify differentially expressed proteins with a smaller effect size in this small number of samples with a tolerance for false discovery. All p values below the α = 0.05 threshold were considered statistically significant. Volcano plots, displaying log₂ fold change (FC) versus unadjusted p values, were used to highlight differentially expressed proteins between groups. We then repeated the analysis adjusting p values for multiple comparisons with Benjamini-Hochberg correction to control for false discovery rate and reduce false positives [11] (Supplemental Table 1; http://links.lww.com/CORR/B430). A priori data on which to base a power analysis were not available at the time of study initiation; thus, sample size was determined by resource constraints.

Functional Annotation and Enrichment Analysis

Identifying patterns in proteins that are overexpressed or underexpressed can uncover biological pathways and protein interactions associated with tendinopathy and diabetes. To identify these patterns, differentially expressed proteins were analyzed using functional enrichment and protein-protein interaction network analysis. Differentially expressed proteins were analyzed using the STRING database, version 11.0 (https://www.string-db.org/) for functional interaction and pathway enrichment [63]. A default threshold of confidence scores > 0.5, which corresponds to “medium confidence interactions,” was chosen to capture meaningful interactions while minimizing false positives [62]. The functional enrichment analysis was performed to identify significantly enriched biological processes, pathways, or molecular functions. Fisher exact test is used for enrichment by STRING, with a significance determined at an adjusted p value of < 0.05. The top-5 enrichment measures for each category are displayed in a bar plot, created using GraphPad Prism 10, with associated proteins list.

To further explore the specific inflammatory pathways enriched in differentially expressed proteins, pathway enrichment analysis was also performed using InnateDB (www.innatedb.ca) with Network Pathway (NETPATH) selected as the primary database, given its focus on inflammation-related signaling. The hypergeometric algorithm was used for pathway overrepresentation analysis, using Benjamini-Hochberg correction applied for multiple testing. Pathways with an adjusted p value of < 0.05 were considered significantly enriched.

Protein-Protein Interaction Network Analysis and Identification of Hub Proteins

Protein-protein interaction networks provide information about how strongly different proteins are expressed together, which can be helpful in identifying proteins that are most important in driving a tissue’s characteristic protein signature. Protein-protein interaction networks were constructed using STRING and visualized in Cytoscape (v3.10.2), an open-source software platform for visualizing complex networks and integrating these with any type of attribute data. Interaction data were imported from STRING into Cytoscape, where nodes represent proteins, and edges represent interactions between them. To identify hub proteins within the network, the CytoHubba plugin in Cytoscape was utilized. The degree metric, which counts the number of edges connected to a node, was selected to identify the top-10 hub proteins. This metric is based on the assumption that proteins with more interactions are likely to be more important in cellular processes. The resulting network, highlighting the top-10 hub proteins, was visualized using the yfiles circular layout tool in Cytoscape. Each hub protein’s role was interpreted in the context of tendinopathy, providing insights into potential pathophysiologic mechanisms and therapeutic targets.

Results

Factors Contributing to Protein Compositional Variation in Human Achilles Tendon

Achilles tendon composition primarily varied by disease status (tendinopathy and diabetes), followed by individual donor and tendon region (Fig. 2A). This was evident from the hierarchical clustering dendrogram (Fig. 2), which shows that samples first group by individual donor before separating by tendon region (1, 2, 3) within each individual. The dendrogram at the top illustrates this clear separation, with distinct branches forming for each donor, indicating that protein abundance profiles are more similar within an individual than between individuals (Fig. 2B). The heatmap further supports this pattern by displaying protein abundance variations across samples. The color gradient (red for high expression, blue for low expression) shows that samples from the same individual share more similar expression patterns, reinforcing the dominant effect of individual-specific variability.

Fig. 2.

Proteomic analysis of different regions within human Achilles tendons. (A) Diagram showing sampling locations from the tendon including (1) left periphery, (2) core, and (3) right periphery. (B) Heatmap displaying protein expression patterns across the three regions (1, 2, 3) within three tendon samples (C1, C2, C3), with clusters indicating groups of proteins that show similar expression profiles. Colors represent differences in protein abundance, with red indicating higher expression and blue indicating lower expression. (C) PCA highlights both regional (core versus periphery) and individual (sample-specific) proteomic differences. Separation between each dot shows how the core region (2) is distinct from the peripheral regions (1, 3) within each tendon sample. The ellipses indicate the clustering of regions within each sample, emphasizing that each tendon sample forms a distinct cluster where the color represents tendon samples (C1, C2, C3). (D) Volcano plot showing differentially expressed proteins that are relatively overexpressed (red) or underexpressed (blue) in the peripheral (P) compared with core (C) regions. The Venn diagram shows the total number of proteins (center) that are overexpressed (pink) and underexpressed (blue).

The secondary clustering by tendon region within each donor suggests that regional differences exist but are less pronounced. PCA further supports these results, where samples colored by individual form distinct, well-separated clusters along the PC axes (Fig. 2C). Additionally, within each individual’s cluster, the periphery regions (1, 3) of each tendon are positioned more closely compared with the core region (2). Comparisons of protein expression between tendon region periphery and core revealed that only 38 proteins were differentially expressed in the periphery region compared with the core, with log₂ FCs of < 1, as seen on the volcano plot (Fig. 2D). Of 38 differentially expressed proteins, 9 proteins were found to be overexpressed in the periphery regions compared with the core, 29 proteins were underexpressed in the periphery compared with the core; the log₂ ratio and p values for differentially expressed proteins have been provided (Supplemental Table 1; http://links.lww.com/CORR/B430).

Hierarchical clustering analysis of tendon proteomes across all individuals from the three groups (tendinopathy, diabetes, and control) demonstrated that disease status is the primary factor driving clustering (Fig. 3). The dendrogram at the top of the heatmap shows that most samples in the tendinopathy group formed a distinct and separate cluster from the diabetes and control tendon groups (Fig. 3A), indicating unique expression patterns associated with tendinopathy. In contrast, the diabetes and control tendon groups cluster more closely together, indicating a higher degree of similarity in their proteomic profiles compared with the tendinopathy group. PCA further supports the proteomic differences driven by disease status, with tendinopathy samples (green) separated from diabetes (orange) and control (yellow) samples primarily along the PC1 axis (Fig. 3B). Most tendinopathy samples are positioned on the right side of the plot, whereas samples in the diabetes (pluses) and control (triangles) groups are clustered on the left. Some tendinopathy samples are positioned closer to the control and diabetes group along both principal components, indicating a degree of similarity between these samples.

Fig. 3.

Proteomic analysis of human Achilles tendons in the tendinopathy (T), diabetes (D), and control (C) groups. (A) Heatmap displaying hierarchical clustering of samples based on protein expression levels. The color gradient from blue to red represents the range from low to high expression levels (log₂ ratio of FC). (B) PCA showing the first two principal components (PC1 and PC2), which account for a total of 37.3% of the total variance.

On histologic analysis, tendon degeneration was most severe in the tendinopathy group and did not differ between the diabetes and control groups (Fig. 4). Control samples exhibited low mean ± SD Bonar scores (5.0 ± 4.6), with well-organized collagen fibers, low cellularity, and minimal vascularity. Tendons in the diabetes group showed no differences in Bonar scores compared with the control group (8.3 ± 5.1, mean difference 3.8 [95% CI -3.7 to 21.4]; p = 0.19). Tendons in the tendinopathy group had higher Bonar scores compared with controls (12.4 ± 4.0, mean difference 7.4 [95% CI 9.1 to 15.8]; p = 0.01), indicating greater severity of degenerative changes. These samples exhibited disrupted collagen arrangement, increased cellularity and vascularity, and altered ground substance. Additional histologic findings included regions of neovascularization and focal areas of hypercellularity.

Fig. 4.

These images demonstrate the histopathologic findings of control, diabetes, and tendinopathy specimens. (A) The Bonar degeneration score for Achilles tendon samples in the control (C), diabetes (D), and tendinopathy (T) groups. The scores are categorized based on six histologic features: collagen arrangement (black), cell morphology (tan), cellularity (blue), vascularity (red), ground substance (gray), and additional findings (light brown). (B) Representative histologic images of tendon samples stained with hematoxylin and eosin (HE) and alcian blue (AB) from the control group. (C) Representative images of tendon specimens stained with HE and AB from the diabetes group. (D) Representative images of tendon specimens stained with HE and AB from the tendinopathy groups. (E) Representative images showing examples of increased lipid deposits and osteoid metaplasia with increased vascularity from diabetic and tendinopathic tendons on HE-stained sections. Scalebar represented as black bar at the bottom left on all images is 100 µm.

Proteomic Dysregulation in Tendinopathy and Associated Pathways and Interaction Network

In tendon samples from the tendinopathy group, a total of 311 differentially expressed proteins were identified. Hierarchical clustering between the tendinopathy and control groups showed that tendinopathy samples cluster separately from control samples based on the dendrogram on the top (Fig. 5A). The volcano plot represents the differentially expressed proteins in the tendinopathy group compared with controls (Fig. 5B). Of the 311 differentially expressed proteins, 152 were overexpressed and 159 were underexpressed (Supplemental Table 2; http://links.lww.com/CORR/B430). Among these, the upregulated proteins with the highest FC included TNC (FC = 1.42; p = 0.002), CLEC3A (FC = 1.35; p = 0.04), PAPSS2 (FC = 1.32; p = 0.001), H3C1 (FC = 1.07; p = 0.02), and S100A1 (FC = 1.05; p = 0.03), while the downregulated proteins with the highest FC included proteins such as MMP3 (FC = -1.86; p < 0.001), MMP10 (FC = -1.64; p < 0.001), ADH1B (FC = -1.38; p = 0.001), ANGPTL7 (FC = -1.35; p = 0.005), and MYOC (FC = -1.31; p = 0.002). Data on exploratory relationships between participant self-reported symptoms and protein compositional and histologic findings have been provided (Supplemental Fig. 2; http://links.lww.com/CORR/B430).

Fig. 5.

Proteomic analysis of samples in the tendinopathy (T) and control (C) groups. (A) Heatmap representing the hierarchical clustering of differentially expressed proteins between tendinopathy (T) and control (C) Achilles tendon samples. The color gradient indicates protein expression levels (red: upregulated; blue: downregulated), demonstrating distinct clustering of tendinopathic versus control samples. (B) The volcano plot illustrates the FC (log₂ ratio) versus statistical significance (-log₁₀ p value) of differentially expressed proteins between the tendinopathy and control groups. Red dots indicate upregulated proteins, while blue dots indicate downregulated proteins in tendinopathic tendons. Proteins showing no change are shown in gray. The top-20 most upregulated and downregulated targets are labeled (1 to 20) based on log₂ ratio. The inset Venn diagram shows the total number of differentially expressed proteins (center), upregulated (pink), and downregulated (blue). (C) This bar plot represents the enriched functional categories identified through STRING analysis, comparing protein expression between tendinopathy and control tendon samples. Each bar represents the strength ratio of enrichment in specific categories. The bar length indicates the degree of enrichment in each category, with specific proteins involved listed at the end of each bar, where red = upregulated and blue = downregulated proteins. (D) The protein-protein interaction network highlighting the top-10 hub proteins identified through network analysis, ranked by degree of centrality using the CytoHubba plugin in Cytoscape. Proteins with higher connectivity are highlighted in red, indicating their central role in the protein-protein interaction network and potential importance in the pathophysiology of tendinopathy.

Functional enrichment analysis was mapped out of 308 of the 311 differentially expressed proteins to known biological pathways, cellular components, and molecular functions (Fig. 5C). Most enriched pathways include purine ribonucleoside salvage, negative regulation of metallopeptidase activity, regulation of collagen fibril organization, and protein folding in the endoplasmic reticulum. Most enriched components included proteasome storage granule, collagen Type I trimer, endoplasmic reticulum chaperone complex, and interstitial matrix. Most enriched molecular functions included glyceraldehyde-3-phosphate dehydrogenase activity, protein disulfide isomerase activity, metallopeptidase inhibitor activity, and aldehyde dehydrogenase (NAD+) activity. The most enriched KEGG pathways included glycolysis/gluconeogenesis, pentose phosphate pathway, ECM-receptor interaction, fatty acid degradation, and the citrate cycle (TCA cycle). The strength ratio of each pathway is shown on the x-axis, indicating the degree of enrichment among the identified differentially expressed proteins. Inflammatome analysis identified that the tumor necrosis factor (TNF-α) signaling (adjusted p value = 0.02) was enriched in the tendinopathic tendons compared with the control tendons (Supplemental Fig. 3; http://links.lww.com/CORR/B430).

Protein-protein interaction network analysis of tendon samples in the tendinopathy group revealed that the 10 hub proteins with the largest degree values were HSPA5, HSPA8, EEF1A1, EEF2, RPS3, RPLP0, RPS2, RPS9, PKM, and EEF1G (Fig. 5D). All identified hub proteins were found to be upregulated in tendinopathy samples compared with control samples. The most highly connected proteins in the network were HSPA8, HSPA5, EEF1A1, and EEF2.

Proteomic Dysregulation in Diabetes and Associated Pathways and Interaction Network

In tendon samples from the diabetes group, a total of 66 differentially expressed proteins were identified (Supplemental Table 3; http://links.lww.com/CORR/B430). The hierarchical clustering between the diabetes and control groups showed protein expression patterns that are different between the two groups (Fig. 6A). The volcano plot represents the differentially expressed proteins in the diabetes group compared with the controls (Fig. 6B). In the diabetes group, 31 proteins were upregulated, with the highest FC observed in PLA2G2A (FC = 1.41; p = 0.001), OAF (FC = 1.22; p = 0.03), TIMP3 (FC = 1.13; p = 0.009), APCS (FC = 1.10; p = 0.02), and LBP (FC = 1.10; p = 0.003). Conversely, 35 proteins were downregulated, with the highest FC in proteins COL25A1 (FC = -1.09; p = 0.003), COL1A1 (FC = -0.96; p = 0.001), PPA2 (FC = -0.90; p = 0.03), SCUBE1 (FC = -0.9; p = 0.009), and COL1A2 (FC = -0.85; p = 0.006). Note that no differentially expressed proteins continued to meet the threshold of statistical significance with p value adjustment.

Fig. 6.

Proteomic analysis of Achilles tendons obtained from individuals in the diabetes (D) and control (C) groups. (A) Heatmap illustrating the hierarchical clustering of differentially expressed proteins between diabetes and control Achilles tendon samples. Each row represents a protein, and each column represents a sample. The color gradient from blue to red indicates low to high expression levels, respectively. Distinct clustering patterns reveal differences in protein expression profiles between diabetes and control tendons. (B) Volcano plot showing the FC (log₂ ratio) of protein expression (x-axis) against statistical significance (-log₁₀ p value, y-axis). Proteins upregulated in diabetes tendons are shown in red, while those downregulated are shown in blue. Gray dots indicate no changes. Dashed lines represent thresholds for significance. The top-20 most upregulated and downregulated targets are labeled (1 to 20) based on log₂ ratio. The inset Venn diagram shows the total number of differentially expressed proteins (center), upregulated (pink), and downregulated (blue). (C) The bar plot depicting enriched functional categories identified through STRING analysis, comparing protein expression between tendons in the diabetes and control groups. Each bar represents the strength ratio of enrichment in specific categories. The bar length indicates the degree of enrichment in each category, with specific proteins involved listed at the end of each bar, where red = upregulated and blue = downregulated proteins. (D) The protein-protein interaction network highlighting the top-10 hub proteins identified through the network analysis, ranked by degree centrality using the CytoHubba plugin in Cytoscape. Proteins with higher connectivity are highlighted in red, indicating their central role in the protein-protein interaction network and potential importance in the pathophysiology of tendinopathy.

Functional enrichment analysis of the differentially expressed proteins returned no biological processes predominantly enriched in diabetes. However, there was enrichment in proteins involved in specific molecular function, cellular component, and KEGG pathways (Fig. 6C). The most enrichment was observed in cellular components including collagen Type I trimer, collagen-containing extracellular matrix, and the basement membrane. Additionally, molecular functions such as platelet-derived growth factor binding, cell adhesion molecule binding, and structural molecule activity were enriched. The KEGG pathway for ECM-receptor interaction was enriched. Results of the inflammatome analysis did not identify any inflammatory pathways that were enriched with diabetes (Supplemental Fig. 4; http://links.lww.com/CORR/B430).

Protein-protein interaction network analysis revealed that the top-10 hub proteins with the largest degree values were COL1A1, HNRNPC, RPL3, RUVBL1, RPS13, COL1A2, ILF3, COL6A1, ITGA1, and RPL27 (Fig. 6D). Of these proteins, HNRNPC, RPS13, and RPL27 were found to be upregulated in diabetic tendons compared with control samples, while key structural and signaling proteins such as COL1A1, COL1A2, COL6A1, ITGA1, and others were found to be downregulated.

Discussion

The mechanisms underlying tendinopathy remain incompletely understood, but prior research suggests a role for protein compositional dysregulation, inflammation, and impaired tissue remodeling. Similarly, diabetes increases the risk for tendon injury [60], yet the extent to which it alters tendon composition predisposing individuals to injury is unclear. The purpose of this study was to take a data-driven approach to understand contributors to variability in tendon protein composition and identify key protein compositional changes accompanying tendinopathy and diabetes. We found that tendon samples from individuals with tendinopathy exhibited extensive proteomic dysregulation, with changes in proteins associated with extracellular matrix turnover, stress response pathways, and metabolic regulators. In contrast, tendons from individuals with diabetes exhibited minimal proteomic and structural abnormalities, with mild reductions in Type I collagen. These findings suggest that tendinopathy is driven by an active, insufficient, or dysregulated remodeling process, whereas diabetes may contribute to subtle compositional changes in tendons without inducing substantial structural degeneration. Clinically, these findings can inform future targeted therapies for tendinopathy to restore tissue homeostasis and improve tendon repair. Furthermore, diabetes was associated with reduced Type I collagen, the primary component of tendon extracellular matrix. This finding suggests that diabetes-related changes in tendon composition may represent one mechanism contributing to the increased risk of tendon injury observed in this diagnosis group [60].

Limitations

This study had several limitations that should be considered when interpreting the findings. Differences in tissue acquisition because of different types of surgical procedures (tendon debridement for tendinopathy samples versus lower extremity amputation for tendons in the diabetes and control groups) could have influenced proteomic differences. However, our data align well with existing reports of proteins known to be dysregulated in tendinopathic tendons [7, 29, 31, 38, 39, 51]. Additionally, technical replicates performed as quality control measures show high consistency across process repeats in all clinical conditions (Appendix 1 and Supplemental Fig. 1; http://links.lww.com/CORR/B430); thus, between-group differences likely reflect underlying biological differences rather than procedural artifacts.

The small sample size, particularly in the diabetes group, limited statistical power and our ability to generalize findings across the spectrum of diabetes severity. Given limitations of statistical power, we were only able to detect large-magnitude differences in differentially expressed proteins, and it may be that additional differentially expressed proteins will be able to be detected in larger cohorts. For this reason, we decided to interpret data a priori based on nonadjusted p values. While this may introduce false discovery in the diabetes-control comparisons, there is good alignment of our findings with those of prior reports, which increases our confidence in our findings [30, 36, 44, 66, 70]. While we report hemoglobin A1c levels to provide context on glycemic control, other clinical factors such as diabetes duration and medication use were not systematically available. Future studies with larger cohorts, including individuals at different stages of diabetes progression, will be essential to determine whether different severities of systemic metabolic alterations yield different impacts on tendon health. Additionally, the proteomics data reflect tendon composition at a single time point, capturing a snapshot of disease rather than its progression. While we acknowledge that prior treatments (for example, physical therapy, injections) in the participants with tendinopathy may have influenced proteomic profiles, all participants were undergoing surgery because of failure of nonoperative management, ensuring that their tendons represent a chronic, recalcitrant injury state.

The absence of cell-specific analyses limited conclusions on the cellular origins of proteomic changes. While histologic assessment confirmed changes in cell morphology and extracellular matrix composition, we did not perform transcriptional or immunohistochemical profiling to distinguish tenocytes from chondrocyte-like cells based on function (transcriptional profile), which could provide additional insight into the observed shifts in ground substance and proteoglycan expression in the future.

This study was designed as a hypothesis-generating, data-driven proteomics analysis rather than a targeted mechanistic investigation. While our findings highlight potential molecular pathways involved in tendon degeneration and diabetes-related changes, future studies will be needed to validate these findings through functional analyses and larger-scale clinical studies. Despite these limitations, our study provides a proteomic characterization of Achilles tendons in tendinopathy and diabetes, identifying key molecular signatures that may guide future investigations. These findings underscore the utility of proteomics in tendon research and provide a foundation for future mechanistic and translational studies exploring targeted interventions for tendon healing.

Key Findings and Clinical Implications

We found that the presence of disease (that is, tendinopathy, diabetes) was the primary contributor to variability in tendon protein composition, followed by donor and then tendon region. While previous studies have reported faster turnover in tendon protein content in the interfascicular matrix compared with the collagen-rich fascicular bundles [19, 64], we observed only minor regional variation in protein composition. Clinically, our findings support prior work that has indicated differences between the core and periphery of tendons [26], suggesting that regional differences in tendon composition may be attributed to differences in rates of tissue turnover and could contribute to the development of tendon injury. From a methodological standpoint, by demonstrating that proteomic composition is largely determined by donor factors rather than regional variations, we established that sampling from a single core region minimizes regional differences and ensures consistency in comparative analyses of disease conditions such as tendinopathy and diabetes.

Pathophysiology of Symptomatic Tendinopathy

Achilles tendinopathy is characterized by a complex interplay of degenerative, inflammatory, and metabolic processes, contributing to progressive tendon dysfunction [4, 28, 42]. Individuals in the tendinopathy group reported moderate to severe pain and functional impairment, as reflected by low VISA-A scores and high NPRS scores. While exploratory, the negative correlation between Bonar scores and VISA-A scores suggests that higher degeneration may correspond to more severe patient-reported symptoms.

Key proteins underlying tendinopathy as a degenerative process include dysregulation of extracellular matrix remodeling proteins. Key matrix regulators including TIMP1, MMP3, and MMP10 were downregulated, suggesting their involvement in tendinopathy. TIMP1 is an inhibitor of matrix metalloproteinases preventing breakdown of the extracellular matrix, while MMP3 and MM10 are involved in collagen turnover and tissue repair [29, 31, 39]. Lack of or poorly timed TIMPs and MMP expression may contribute to tendon degeneration, fibrosis, and impaired healing [22, 39]. Our findings align with existing reports of downregulation of MMP3 and MMP10 in tendinopathy, even though other matrix metalloproteinases are typically overexpressed with tendon repair and remodeling [7, 29, 31, 39, 51]. Additionally, we observed decreased levels of OLFML3 and ITGA1 in tendinopathic tendons, proteins involved in extracellular matrix repair [22] and tissue mechanosensitivity [24], respectively, which suggests compromised tendon function and an impaired response to mechanical stress.

Inflammation also plays a central role in tendon degeneration, as evidenced by the dysregulation of inflammatory mediators in tendinopathic tendons. Inflammatome analysis revealed that the TNF-α pathway was highly enriched in tendinopathy samples, supporting the involvement of chronic inflammatory signaling in tendon degeneration. Additionally, S100A10, a calcium-binding protein involved in proinflammatory signaling [59], was upregulated, while stress-response proteins (HSPA5, HSPA8) were among the most highly connected hub proteins in the protein-protein interaction network. Of note, HSPA5 has been linked to acute Achilles tendon rupture [48], while EEF2 has been identified as a novel biomarker and strongly prognostic of 1-year clinical outcomes in ruptured Achilles tendons [17]. These proteins are associated with cellular stress, protein misfolding, and chronic inflammation, suggesting that tendinopathic tendons experience persistent cellular stress and an unresolved inflammatory response.

Beyond extracellular matrix proteins, tendinopathic specimens exhibited metabolic reprogramming, with enrichment in purine ribonucleoside salvage and negative regulation of metallopeptidase activity, suggesting increased nucleotide salvage demand and a protective response against extracellular matrix degradation. Downregulation in proteins associated with enriched metabolic pathways such as glycolysis/gluconeogenesis and the TCA cycle indicates a shift toward less efficient energy production. Additionally, the protein-protein interaction network identified PKM and ribosomal proteins (RPLP0, RPS2, RPS9, RPS3) as highly connected, highlighting altered protein translation processes in tendinopathic tissues. It is important to note that the enrichment of these proteins is not overwhelmingly strong, highlighted by a moderate strength ratio, suggesting that while these proteins are likely relevant in tendinopathy, they are not the only key players or are part of a broader, more distributed network of interactions.

Our study identified several proteins not previously linked to tendinopathy that were differentially expressed in tendinopathic tendons. This included proteins involved in metabolism (AKR1B1, ALDH6-A1, DPP4, SCP2, GRHPR), cell communication (THBS4), vascular biology (SCUBE-1), osteoclast/osteoblast regulation (MYDGF) [71], and the immune response (DDT). These novel findings provide a foundation for further investigations into the molecular mechanisms of tendinopathy and potential therapeutic targets.

Pathophysiology of Tendons in Diabetes

Tendons from individuals with diabetes in our study demonstrated minimal histopathologic degenerative changes compared with tendons in the tendinopathy group, as reflected in Bonar scores, which showed no difference from controls. However, changes in tendon architecture, collagen organization, abnormal lipid deposits, abnormal mineralization, and enthesopathy were observed in tendons from this group, which aligns with prior studies [65, 67, 73]. The clinical presentation of patients in the diabetes cohort had relatively well-controlled diabetes at the time of surgery and underwent lower extremity amputation primarily because of complications of Charcot arthropathy and in the absence of symptomatic tendon injury. This suggests that systemic metabolic changes may predispose tendons to injury [60]; however, additional systemic or behavioral triggers are likely needed for progression to tendinopathy.

Proteomic analysis revealed subtle differences in the extracellular matrix composition of tendons in the diabetes group, particularly with downregulation of key structural proteins, including COL1A1, COL1A2, and COL6A1, suggesting potential alterations in tendon integrity. Enrichment in extracellular matrix–receptor interaction pathways further supports the idea that diabetes may affect tendon mechanotransduction, potentially impairing tendon resilience over time. In contrast to tendinopathy, these changes were not accompanied by major histologic disruptions or severe matrix remodeling, suggesting a distinct disease trajectory. Given prior reports of increased tendon stiffness and altered mechanical properties in diabetes [1, 2, 9], our findings suggest that while fibrosis may be present at the molecular level, its clinical relevance remains unclear.

While diabetes is often associated with systemic inflammation, tendons in the diabetes group showed no strong enrichment of inflammatory signaling cascades, suggesting that inflammation may play a less central role in tendon alteration in diabetes. Furthermore, while tendinopathy showed marked alterations in energy metabolism and stress-response pathways, tendons in the diabetes group displayed proteomic changes that align with systemic metabolic dysfunction. Enrichment in platelet-derived growth factor binding and cell adhesion molecule binding indicates potential alterations in cellular repair mechanisms and signaling. Protein-protein interaction network analysis highlighted upregulation of RNA binding proteins (HNRNPC, RSP13, and RPL27), suggesting increased protein synthesis activity, which may reflect an adaptive or compensatory response to chronic metabolic stress. Interestingly, proteomics was able to capture many of the differentially expressed proteins that are known markers of diabetes in other tissues, such as PLA2G2A [34], DIABLO [61], VTN [5, 6, 12, 14, 45, 68], ATP1A1 [3, 15, 27, 35], SFRP2 [13, 20, 54], CLU [21], and GPX4 [16, 56, 74], indicating systemic rather than tendon-specific changes. Taken together, both proteomic and histologic changes in diabetes tendons were less pronounced than those observed in tendinopathic tendons, suggesting that diabetes alone may not be the primary driver of tendon degeneration. Instead, additional systemic or mechanical stressors may be necessary to trigger symptomatic tendon injury [50, 60].

Conclusion

Our findings provide a comprehensive proteomic and histologic characterization of Achilles tendons in tendinopathy and diabetes. We demonstrated that tendinopathy is associated with substantive extracellular matrix dysregulation, inflammatory signaling, and metabolic alterations, suggesting an active degenerative process that may drive tendon dysfunction and pain. In contrast, diabetes appears to induce more subtle protein compositional changes, primarily affecting collagen composition and mechanotransduction pathways without the widespread degeneration observed in tendinopathy. Future studies should expand on these findings by incorporating larger, more diverse cohorts, including individuals with varying degrees of diabetes severity and different tendinopathy stages, to better understand disease progression and patient-specific risk factors. By integrating clinical, histologic, and proteomic insights, our study lays the foundation for future mechanistic investigations into tendon degeneration and metabolic dysfunction. These findings emphasize the utility of proteomics in musculoskeletal research and highlight potential molecular pathways that may be leveraged for therapeutic development in tendinopathy and diabetes-related tendon dysfunction.