Abstract

Objective To examine whether gut microbes were associated with postsurgery-sustained knee pain in patients with knee osteoarthritis (OA) by a gut microbiomics approach.

Methods Patients receiving total knee replacement (TKR) because of primary knee OA were recruited. Sustained knee pain status at ≥ 1 year after TKR was defined by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). Fasting plasma and fecal samples were collected. Metabolomic profiling was performed on fasting plasma. 16S rRNA sequencing was performed on fecal samples to determine microbial composition.

Results Twenty patients with TKR because of primary knee OA were included in the study, with 10 experiencing sustained postsurgery pain and 10 without such pain. Age, sex, and BMI (kg/m2) were matched. Linear discriminant analysis of microbiome data identified 13 bacterial taxa that were highly abundant in the pain group and 5 that were highly abundant in the nonpain group (P < 0.05 for all). Plasma metabolomic profiling measured 622 metabolites. The correlation analysis indicated the 18 taxa were significantly correlated with 231 metabolites (P < 0.05 for all). Sparse partial least squares discriminant analysis showed that 30/231 metabolites explained 29% of total variance and can be used to clearly separate patients with sustained knee pain from patients in the nonpain group. Pathway enrichment analysis showed that these significant metabolites were enriched in the arachidonic acid metabolic pathway, bile acid biosynthesis, and linoleic acid metabolism.

Conclusion Gut microbes may play a significant role in sustained knee pain in patients with knee OA after TKR, potentially through their activation of inflammatory pathways, lipid metabolism, and central sensitization.

Key Indexing Terms:

Osteoarthritis (OA) is a disabling disease with varying degrees of joint swelling and pain, affecting over 7% of the global population.1 Total joint replacement surgery is the most effective treatment for patients with end-stage OA. However, up to 20% of patients continue to experience sustained joint pain after surgery, which reduces their quality of life and causes psychological distress.2 The etiology of sustained pain in patients with OA remains elusive but is thought to be attributed to multiple factors, including sensitization, inflammation, and alterations in gut microbiome.

The gut-brain-microbiome axis is a bidirectional communication network connecting the brain, gut, and its microbiota. It has been reported to play a crucial role in several diseases, such as Parkinson disease, Alzheimer disease, and irritable bowel syndrome.3,4 Bacterial metabolism and its products can act as inflammatory activators to exacerbate pain sensitivity.5 These bacterial products can cross the intestinal blood barrier, entering the systemic circulation and potentially triggering local inflammation or causing widespread sensitization.6

Data on gut microbiomics in OA are still sparse. One study reported a positive correlation between increased abundance of Streptococcus ssp. and joint pain intensity in patients with OA.7 However, the association between gut bacterial abundance and circulating metabolite levels in the host has not yet been explored. To our knowledge, the current study was the first to identify gut microbes to be associated with sustained knee pain in patients with OA and to explore their relationship with circulating metabolite levels in the host. Understanding this relationship is the key to comprehending the microbiome’s effect on sustained knee pain and the potential corresponding mechanisms; this may provide insights into postoperative prediction as well as probiotic supplementation therapy.

METHODS

Study participants. The current study was a pilot study of the ongoing Newfoundland Osteoarthritis Study, which aims to investigate multiple aspects of OA.8 Patients with OA were recruited between November 2011 and September 2017 at St. Clare’s Mercy Hospital and Health Sciences Centre General Hospital in St. John’s, Newfoundland and Labrador, Canada. All participants underwent total knee replacement (TKR) surgery because of primary knee OA. The diagnosis was based on the American College of Rheumatology OA clinical diagnostic criteria and was confirmed by the attending orthopedic surgeon as well as the pathology reports of the removed articular cartilage following surgery.9

Sustained knee pain status was defined by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) Likert 3.0 pain subscale, which was used to evaluate the pain intensity in patients at least 1 year postsurgery.10 The questionnaire included 5 questions, each scored from 0 (none) to 4 (extreme), assessing self-reported pain when patients are walking on a flat surface, using stairs, in bed at night, sitting or lying, and standing upright. Patients with sustained knee pain were defined as those who reported pain (mild to extreme) in all 5 questions. Patients without sustained pain were defined as those who reported no pain in all 5 questions. The same definition of sustained knee pain was used in a previous publication.11 Those who reported pain in only some of the 5 questions were excluded from the current study.

Ethics. Ethics approval was received from the Health Research Ethics Authority of Newfoundland and Labrador (#2011.311). All study participants provided written informed consent for their participation.

Data and sample collection. Demographic data were obtained from general health questionnaires and medical records, age at surgery was calculated, and BMI was calculated as weight in kilograms divided by squared height in meters.

Peripheral blood samples were collected after at least 8 hours of fasting. The whole blood was centrifuged at 2000 × g for 10 minutes to separate the plasma, which was then aliquoted and stored at −80 °C until analysis.

A OMNIgene•GUT collection kit (DNA Genotek) was mailed to the home address of all participants. The kit included user instructions, a spatula, a collection tube with DNA preservation solution, a biospecimen bag, and toilet accessories. The DNA preservation solution was able to maintain DNA integrity at room temperature for up to 60 days. All participants were requested to collect their fecal sample following the manufacturer’s instructions to minimize possible contamination, record the collection date and time, and then send the tube containing the fecal sample back to our laboratory. One milliliter of the fecal sample was aliquoted and transferred into an aseptic screw cap tube and stored at room temperature until DNA extraction, with a mean (SD) storage time of 20.8 (11.8) days. Metagenomic DNA of feces was extracted using the DNeasy 96 PowerSoil Pro QIAcube HT Kit (#47021; QIAGEN) according to the manufacturer’s instruction. The extracted DNA underwent a quality check using Qubit (Invitrogen) and was stored in −20 °C freezers before microbiome analysis.

Microbiome sequencing and metabolomic profiling. 16s rRNA V4-V5 region sequencing was performed at the Integrated Microbiome Resource (Dalhousie University, Halifax, Canada). Libraries were prepared for the Illumina MiSeq platform using the Illumina DNA Prep kit (Illumina). Starting with 1 ng per sample, the process involved enzymatic shearing and adaptor tagging, PCR amplification with barcode integration, purification using columns or beads, and normalization using Illumina beads or a manual approach. Subsequently, samples were pooled and sequenced using the Illumina MiSeq platform.

The V4 to V5 variable region of 16S rRNA was amplified using Phusion High-Fidelity DNA polymerase (Thermo Fisher Scientific) and primers 515FB (GTGYCAGCMGCCGCGGTAA) and 926R (CCGYCAATTYMTTTRAGTTT) with respective barcodes. PCR products were verified using Coastal Genomics Analytical Gels (Coastal Genomics). The amplicons were cleaned and normalized using the high-throughput Charm Biotech Just-a-Plate 96-well Normalization Kit (Charm Biotech). The samples were then sequenced on the Illumina MiSeq platform using 300 + 300 base pair paired-end V3 chemistry.

Metabolomic profiling on plasma samples was performed using Biocrates MxP Quant 500 kit (Biocrates Life Sciences AG), which quantifies up to 630 different endogenous metabolites (the full list of the metabolites is provided in Supplementary Table S1, available with the online version of this article). The profiling was done on an API5500 Qtrap tandem mass spectrometry instrument (Applied Biosystems/MDS Analytical Technologies) equipped with liquid chromatography at The Metabolomic Innovation Centre (https://metabolomicscentre.ca). The Biocrates MetIDQ software was used to control the entire assay workflow from sample registration, automated calculation of metabolite concentrations, and the export of data into other data analysis programs. Metabolite concentrations were reported as μmol/L.

Statistical analyses. The Microbiome Helper pipeline based on QIIME 2 (https://github.com/LangilleLab/microbiome_helper/wiki) was used to clean the sequencing data and Deblur (https://github.com/qiime2/q2-deblur) was used to create amplicon sequence variants (ASVs). A rarefaction curve was used to determine whether a given sample had sufficient sequencing depth. α-diversity and β-diversity were used to estimate the diversity within and between samples. Chao1, abundance-based coverage estimators (ACE), Shannon, and Simpson indices were used to assess the α-diversity. Bray-Curtis distance was used to calculate β-diversity and visualized in a principal coordinates analysis (PCoA) plot. Wilcoxon signed-rank test was used to calculate the statistical significance level and analysis of similarities (ANOSIM) was used to estimate the differences between the 2 groups. Linear discriminant analysis effect size (LEfSe) was used to identify taxonomic differences between patients with and without postsurgery-sustained knee pain. Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (Picrust2; The Huttenhower Lab) was used to predict the metagenome functions based on sequences and abundance. Picrust2 predicts metagenomic functional content based on 16S rRNA gene sequences by leveraging a comprehensive reference database of known bacterial genomes to infer the abundance of gene families in microbial communities. This process involves phylogenetic reconstruction and the extrapolation of gene family presence and abundance from closely related reference genomes.

For the metabolomic data, metabolites were removed if > 25% of the samples had values below the limit of detection (LOD). Values < LOD for the remaining metabolites were imputed using a linear regression prediction model in which the concentration of a given metabolite was regressed on age, sex, and BMI. Principal component analysis was performed to assess batch effect. Spearman correlation analysis was performed to examine the relationships between each of the metabolite concentrations and the identified sustained knee pain–associated microbes. The supervised classification model, sparse partial least squares discriminant analysis (sPLS-DA), was used to reduce the number of features and seek components that best separate the samples. MetaboAnalyst 5.0 (https://www.metaboanalyst.ca) was used for the enrichment of metabolites into pathways.

All statistical analyses and visualization were performed in R version 4.3.2 (R Foundation for Statistical Computing) with vegan,12 picante,13 ggplot2,14 microbiomeMarker,15 phyloseq,16 ggpicrust2,17 caret,18 and mixOmics19 packages. Significance level was defined at α = 0.05 and correction for multiple testing was not considered because of the exploratory nature of the study.

RESULTS

Participants’ characteristics. A total of 10 patients with knee OA with postsurgery-sustained knee pain and 10 age-, sex-, and BMI-matched patients with knee OA without postsurgery-sustained knee pain (nonpain group) were included in the current study. The pain group (8 female and 2 male participants) had a mean age of 61.3 (SD 5.3) years with a mean BMI of 37.1 (SD 4.4), whereas the nonpain group (8 female and 2 male participants) had a mean age of 62.3 (SD 3.3) years and a mean BMI of 35.4 (SD 4.3). There were no significant differences in age, sex, and BMI between patients with and without sustained knee pain (P > 0.05 for all; Table). In addition, the average preoperative score for the 5 pain questions in the pain group was 3.2, whereas for those in the nonpain group, it was 2.3. Average postoperative pain score in the pain group was 2.8 compared to 0.0 in the nonpain group.

Table.

Characteristics of study participants.

Clade diversity of the gut microbiome. 16S rRNA gene sequencing produced a cumulative total of 668,272 sequence reads, averaging 33,414 paired-end reads per sample and resulting in the identification of 527 ASVs. The nearly smooth rarefaction curve, based on the observed features and sequencing depth, indicated that the sequencing data were of sufficient quality to comprehensively identify all sequences present in the samples. α-diversity variables, including richness (Chao1 and ACE) and diversity (Shannon and Simpson indices) were assessed. There were no statistically significant differences in the Chao1, ACE, Shannon, and Simpson indices between patients with and without sustained knee pain (P > 0.05 for all; Figure 1A). The separation shown in the Bray-Curtis–based PCoA plot was not apparent (Figure 1B), and ANOSIM with a permutation of 999 times revealed no differences between the 2 groups (P = 0.17).

Figure 1.

Differences in and diversity analysis of microbial abundances. (A) Differences in α-diversity between pain group and nonpain group assessed by Chao1, ACE, Shannon, and Simpson indices. (B) β-diversity differences between the pain group and nonpain group were estimated by PCoA based on Bray-Curtis distance. The significance level was calculated by ANOSIM. (C) The proportions of bacteria at various classification levels in each group. At the order, family, and genus levels, the subfigures only display the top 20 most abundant bacteria. ACE: Abundance-based Coverage; ANOSIM: analysis of similarities; PC: principal coordinate; PCoA: principal coordinates analysis.

Differences in microbial abundances. The relative abundance stack plot revealed that Bacteroidota, Firmicutes, Fusobacteriota, Proteobacteria, and Verrucomicrobiota were the predominant phyla in both groups, collectively accounting for ≥ 95% of the gut flora. The Wilcoxon signed-rank test revealed that the nonpain group had a significantly higher abundance of Proteobacteria (P = 0.03). Dominance at the class level was observed in Bacteroidia, Clostridia, Fusobacteriia, and Verrucomicrobiae. Similarly, at the order level, Bacteroidales, Oscillospirales, and Lachnospirales emerged as the predominant groups. At the family level, the pain group exhibited significantly higher abundance of Barnesiellaceae compared to the nonpain group. The details of each level are presented in Figure 1C. LEfSe analysis was conducted using a threshold linear discriminate analysis (LDA) score of 3.0 and P < 0.05, identifying 18 significant microbial taxa. Among them, 13 features were highly abundant in patients with sustained knee pain, whereas 5 were highly abundant in patients in the nonpain group. The feature with the highest LDA score in the pain group was Paraprevotella (LDA score of 3.9; P = 0.01), whereas it was Bacteroides in the nonpain group (LDA score of 4.8; P = 0.02; Figure 2A). Further, 9 additional microbial taxa were significantly correlated with pain-related question scores, with correlation coefficients ranging from −0.66 to −0.45 and from 0.47 to 0.57 (P < 0.05 for all). These 9 microbial taxa belong to 3 families: Erysipelotrichaceae, Lachnospiraceae, and Ruminococcaceae (details provided in Supplementary Table S2, available with the online version of this article).

Figure 2.

LEfSe results. (A) LEfSe results shown by histogram of the LDA scores; LDA score represents the effect size and rank of each differentially abundant taxon (LDA score > 3; P < 0.05). (B) sPLS-DA based on the 30 most significant metabolites that were associated with the 18 sustained knee pain–associated gut microbes. expl. var: explanatory variable; LDA: linear discriminate analysis; LEfSe: linear discriminant analysis effect size; sPLS-DA: sparse partial least squares discriminant analysis.

Association between microbiota and plasma metabolites. Plasma metabolomic profiling quantified 630 metabolites, of which 622 passed the quality control checks. No batch effect was detected. The correlation analysis indicated that 231 metabolites had significant correlations with the 18 features that were identified by LEfSe, with coefficients ranging from −0.71 to −0.44 and from 0.44 to 0.80 (P < 0.05 for all; Supplementary Table S3, available with the online version of this article). sPLS-DA analysis showed that 30 metabolites explained 29% of total variance and can be used to clearly separate patients with sustained knee pain from patients in the nonpain group (Figure 2B). These important metabolites were enriched in the arachidonic acid (AA) metabolic pathway, primary bile acid biosynthesis, and linoleic acid metabolism (Figure 3). In addition, Picrust2 identified 19 potential functional pathways that differed between the 2 groups (Figure 4; details provided in Supplementary Table S4). The results revealed that the 18 significant taxa might influence ether lipid metabolism and N-glycan biosynthesis in the pain group, whereas in the nonpain group, they appeared to affect carotenoid biosynthesis.

Figure 3.

Pathway analysis of important metabolites separating patients with sustained knee pain from patients in the nonpain group.

Figure 4.

Picrust2 identified potential functional pathways that differed between the 2 groups (pain and nonpain) based on sequencing and abundance. ECM: extracellular matrix; mTOR: mammalian target of rapamycin; Picrust2: Phylogenetic Investigation of Communities by Reconstruction of Unobserved States.

DISCUSSION

Our data suggest that sustained knee pain in patients with OA after TKR may be partially attributed to some highly abundant gut microbes that are involved in lipid metabolism and central sensitization through various mechanisms. Bacteria that were more prominent in the nonpain group appeared to play a role in antiinflammatory processes and in reducing neuropathic pain.20,21

The highly abundant taxa in patients with sustained knee pain might be involved in pain and inflammation through multiple mechanisms. Paraprevotella, which had the highest LDA score in the pain group, has been reported to increase in mice with colitis and showed a positive correlation with lysophosphatidylcholine.22 Previous work from our laboratory revealed that phosphatidylcholine, the precursor of lysophosphatidylcholine, was negatively associated with sustained knee pain.11 During the conversion of phosphatidylcholine to lysophosphatidylcholine, long-chain fatty acids such as AA were released,23 which is the precursor of eicosanoids such as prostaglandins and leukotrienes, which play a vital role in inflammatory response. Thus, our results suggested that Paraprevotella in the gut might be involved in promoting this process. Additionally, Alloprevotella, which belongs to the same family as Paraprevotella, has been reported to be positively associated with erythrocyte sedimentation rate and C-reactive protein in patients with rheumatoid arthritis (RA).24 Another family member, Prevotella, has been linked to inflammatory diseases mediated by Th17-related immune responses.25 The second and third (Barnesiellacea and Barnesiella) differentially abundant taxa both belong to the Barnesiellaceae family and have been reported to be positively associated with tumor necrosis factor (TNF).26 It has been demonstrated that arthritic mice overexpressing human TNF showed more intense pain responses when exposed to noxious stimuli.27Coprobacter is positively associated with sarcopenia and plasma trimethylamine-N-oxide levels, which can also promote inflammation.28,29 Sarcopenia leads to weakened muscle mass and strength, resulting in pain or fatigue in patients during activity or at rest.30Christensenellaceae has been reported to have an increased abundance in patients with obesity with low back pain.31 Further, among the 5 highly abundant taxa in the nonpain group, Anaerostipes caccae can produce acetate, butyrate, and lactate as the end products of glucose metabolism.20 Butyrate is a type of short-chain fatty acid, which has been widely reported to be involved in antiinflammatory processes and in relieving neuropathic pain.32 Particularly, butyrate has been shown to alleviate pain in rats with OA by decreasing inflammation levels and downregulating the transcription of genes related to pain signaling.33

The Ruminococcaceae family was negatively correlated with WOMAC pain score in our study, which is consistent with a previous study.7 However, the Subdoligranulum strain showed a positive correlation, which causes joint swelling in colonized mice and has been identified in individuals at high risk or in the early stages of RA.34 The role of Subdoligranulum in postsurgery-sustained pain requires further investigation. Moreover, Blautia also showed a negative correlation with WOMAC pain score. Blautia has been reported at a higher abundance in a nonpain group compared to a sustained-pain group following breast cancer surgery.35

Consistently, the following 3 pathways were identified as being associated with significant microbial taxa for sustained knee pain: the AA metabolic pathway, linoleic acid metabolism, and bile acid biosynthesis. Among them, AA and linoleic acid both belong to the category of polyunsaturated omega-6 fatty acids. AA was recognized as a crucial signaling molecule that can initiate and activate various pathways involved in inflammation and pain processes.36 Linoleic acid and its derivatives were associated with pain and inflammation.37 Further, as a precursor of AA, linoleic acid can be converted into AA and promote inflammation.38 In addition, the correlation between bile acid metabolism and lipid metabolism is also important, particularly in the digestion and absorption of lipids.39 AA can be converted to prostaglandins by cyclooxygenases. The role of prostaglandins in increasing pain sensitization has been widely reported.40 The transient receptor potential vanilloid 1 (TRPV1) is well-recognized for its important role in modulating pain and inflammatory processes.41 Numerous lipids, including lipoxygenase metabolites of AA, have been identified as activators of TRPV1 receptors.42 This results in an increased inflammatory response and increased pain intensity. Overall, the 3 identified pathways were all related to lipid metabolism, which might be a key factor in causing postsurgery-sustained knee pain.

Further, ether lipid metabolism and N-glycan biosynthesis were predicted to play roles in the pain group using Picrust2. Ether lipids are involved in signal transduction, neurotransmission, and inflammatory processes.43 Alterations in N-glycans have been associated with increased sensitization to mechanical force in a rat model and chronic low back pain in humans.44,45 Although carotenoid biosynthesis was predicted to have a role in the nonpain group, the antiinflammatory effects of β-carotene have been demonstrated in many in vivo models and can be used as an adjuvant treatment for neuropathic pain.21 Additionally, research involving diabetic neuropathic pain zebrafish demonstrated that carotene can reduce neuropathic pain in this model by scavenging free radicals and reducing lipid oxidation.46

There are some limitations in the study. First, the study was an exploratory effort with a small sample size, and we did not consider correction for multiple testing; it is essential to validate these results in further studies with large sample sizes. Additionally, synovitis was associated with higher scores for postsurgery-sustained pain, and cystine can be a biomarker for knee OA with synovitis.47,48 However, no significant difference was shown in cystine levels between the pain group and the nonpain group in the current study, which might also be attributed to the small sample size. Second, our results suggest that microbiome-related lipid metabolism may be responsible for sustained knee pain, but the concentrations of these measurable lipids did not differ significantly between the 2 groups. The targeted approach applied in the current study might have been unable to detect all the bacteria-related metabolites. Conducting a metabolomic analysis on fecal samples using an untargeted approach might provide more valuable and comprehensive insights.49 Third, gut microbiota can be influenced by age and sex.50 We matched age, sex, and BMI between the 2 groups, but the majority of our participants were elderly female individuals (mean age 62.2 years). Last, antibiotic drug use and dietary intake could have effects on the gut microbial composition. Unfortunately, we did not have data on these factors.

To the best of our knowledge, this was the first study to integrate gut microbiota analysis with blood metabolites in research on patients with sustained knee pain in OA postsurgery. Our findings highlight the significance of gut microbiome composition in patients with OA experiencing postsurgery-sustained knee pain. Changes in these highly abundant microbes could partially explain the persistence of pain in these patients. Our findings may assist in understanding the role of the microbiome in the development of postoperative pain in OA and could offer valuable insights for postoperative pain prediction and potential probiotic supplementation strategies.

In conclusion, some highly abundant gut bacteria may contribute to the postsurgery-sustained knee pain in patients with OA, potentially because of their involvement in lipid metabolism, activation of inflammatory pathways, and contribution to central sensitization.

ACKNOWLEDGMENT

We thank all the study participants who made this study possible. The microbiomic profiling was done at the Integrated Microbiome Resource (Dalhousie University, Halifax, Canada) and the metabolic profiling was done at The Metabolomics Innovation Centre (Edmonton, Alberta, Canada).

Footnotes

This study was funded by the Canadian Institutes of Health Research (FRN# 175015, 153298, 143058, 132178, and 191966), Arthritis Society (ID# 22-121), Research and Development Corporation of Newfoundland and Labrador (5404.1423.102), and Memorial University of Newfoundland Medical Research Fund.

The authors declare no conflicts of interest relevant to this article.

Accepted for publication August 14, 2024.Copyright © 2024 by the Journal of Rheumatology

This is an Open Access article, which permits use, distribution, and reproduction, without modification, provided the original article is correctly cited and is not used for commercial purposes.

DATA AVAILABILITY

Data are available upon reasonable request from the corresponding author.