Biomarkers of Inflammation – LBP and TLR– Predict Progression of Knee Osteoarthritis in the DOXY Clinical Trial

Abstract

Objective:

To evaluate systemic inflammatory biomarkers in symptomatic knee osteoarthritis (OA) and their association with radiographic and biochemical OA progression.

Methods:

Lipopolysaccharide (LPS) binding protein (LBP), soluble Toll-like receptor 4 (sTLR4) and interleukin 6 (IL-6) were measured in plasma of 431 knee OA patients from the doxycycline (DOXY) trial at baseline and 18 months. Plasma LBP was also measured at 12 months. As a biochemical indicator of disease activity and OA progression, urinary (u) C-telopeptide of Type II collagen (uCTX-II) was measured in samples collected at baseline and 18 months. Change over 16 months in radiographic tibiofemoral joint space width (JSW in mm) and joint space narrowing (JSN>0.5 mm) were used to indicate radiographic OA progression. Change over 18 months for uCTX-II was used as a secondary outcome. Both univariate and multivariable regression analyses were performed to test the association between Z-score transformed biomarkers and outcomes.

Results:

Baseline LBP and time-integrated concentration (TIC) of LBP over 12 and 18 months were associated with worsening JSW (parameter estimates: −0.1 to −0.07) and JSN (OR: 1.32 to 1.42) adjusting for treatment group, age, BMI and corresponding baseline radiographic measures. Baseline sTLR4 and TIC over 18 months were associated with change in uCTX-II over 18 months (adjusted parameter estimates: 0.0017 to 0.0020). Results were not modified by treatment with doxycycline.

Conclusion:

Plasma LBP and sTLR4 were associated with knee OA progression over 16–18 months. These results lend further support for a role of systemic low-grade inflammation in the pathogenesis of knee OA progression.

Introduction

Osteoarthritis (OA) is the most prevalent arthritic disease and a leading cause of disability, affecting more than 320 million individual globally, on the basis of age- standardized prevalence-rate estimates [1]. The past decade has marked a gradual but fundamental shift in our understanding of the mechanisms underlying OA away from the concept of it being the consequence of simple normal bodily wear and tear to an understanding of it being a complex multi-tissue pathology in which low-grade, chronic inflammation has a central role [2]. To date, studies of inflammation in OA pathogenesis and progression have focused on synovial inflammation (synovitis) [3]. However, emerging evidence suggests that synovitis is not the only form of inflammation observed in OA. Several hypotheses have been proposed in which inflammation and inflammation-related molecules from different underlying pathologies contribute to OA including: (i) previous joint trauma contributing to a local increased level of inflammatory mediators [4], (ii) metabolic syndrome and (iii) age, contributing to a systemic burden of low-grade pro-inflammatory stimuli [5–7]. These sources of local and systemic low-grade inflammation could contribute to the initiation and persistence of OA, and synergize with each other and other inflammatory mechanisms in OA [8, 9].

Biomarkers, such as toll-like receptor 4 (TLR4) [10–12], lipopolysaccharide and lipopolysaccharide binding protein (LBP) [13, 14], and interleukin 6 (IL-6) [15–17] have been associated with the burden of systemic low-grade inflammation and also implicated in inflammatory mechanisms underlying OA. Our objective was to evaluate the association of these systemic factors (LBP, soluble (s) TLR4 and IL-6), related to chronic low-grade inflammation, for associations with OA progression. Existing data and biospecimens from the completed randomized placebo-controlled doxycycline (DOXY) trial (NCT00000403) [18] provided the means of achieving this objective and evaluate, the potential modification of these inflammatory biomarkers in the context of a trial in which doxycycline modestly slowed knee OA progression [18, 19].

Previously, doxycycline was shown to be chondroprotective for human knee OA [18-20]. The chondroprotective effect demonstrated in these human trials was consistent with a number of prior trials showing chondroprotection by doxycycline in animal models of OA, including in guinea pigs [21], rabbits [22], dogs [21, 23] and horses [24]. Tetracyclines have been reported to inhibit matrix metalloproteinases (MMPs) but the mechanism has not been completely elucidated; inhibition appears to be dependent on chelation of structural metals in MMPs rather than chelation of the active site Zn²+ [25]. The steady-state concentrations of doxycycline achieved with clinically relevant pharmacological doses of doxycycline (100 mg orally twice daily) in humans are 1.2–1.6 pg/ml; these are sufficient for an anti-microbial effect (required therapeutic range 0.5–1.0 pg/ml) [21], but generally below the range of concentrations required to inhibit the OA- associated matrix metalloproteinases: MMP-1 (collagenase) IC50 232 pg/ml; MMP-2 (gelatinase A) IC50 29 pg/ml; MMP-3 (stromelysin) IC50 16 pg/ml; and MMP-13 (collagenase) IC50 3–15 pg/ml [25, 26]. We therefore hypothesized that an MMP- independent mechanism may underlie the chondroprotective effects of doxycycline. Antibiotics are well known to alter the gut flora [27, 28]. An effect of doxycycline on gut microflora could lower endotoxemia leading to a decreased burden of systematic low- grade inflammation and thereby provide an alternative explanation for the chondroprotective effects of orally administered doxycycline.

Materials and Methods

Participants.

Data and samples corresponding to 431 participants from the doxycycline (DOXY) trial (NCT NCT00000403) [18] were available for this study (Table 1). Study participants, were obese women aged 45–64 years with unilateral radiographic knee OA according to the criteria of the American College of Rheumatology, with Kellgren/Lawrence (K/L) grade 2 or 3 in one knee (the index knee) and K/L grade 0 or 1 in the contralateral knee [29] in a conventional standing anterior-posterior (AP) knee radiograph. All participants were in the upper tertile of the age- and race-adjusted norms for body mass index (BMI) in women [30]. Considering the positive structural modifying results of the DOXY trial at 16 months [18], we chose to focus our analyses on the first 16 months of data in this study (see Figure 1 study flow diagram). Available plasma samples from the DOXY clinical trial over this timeframe included 341 from baseline, 254 from the 12-month follow-up and 282 from 18-month follow-up. Urine samples from 237 of the DOXY clinical trial were also available from baseline and 236 from the 18-month follow-up. Samples were stored at −80°C until analysis.

Table 1.

Patient demographics and plasma/urine biomarker concentrations at baseline.

	DOXY Group (n=218)	Placebo Group (n=213)	Total (n=431)
Demographics:
Age (years)	54.60 (5.47)	54.81 (5.81)	54.70 (5.64)
Height (cm)	162.84 (6.17)	162.31 (6.05)	162.59 (6.12)
Weight (kg)	97.63 (18.08)	96.06 (16.73)	96.86 (17.42)
BMI (kg/m2)	36.77 (6.29)	36.46 (5.97)	36.62 (6.13)
Outcomes:
JSW (mm)	3.84 (1.30)	3.77 (1.27)	3.81 (1.28)
uCTX-II (ng/mmol)	0.0171 (0.0277)	0.0135 (0.0194)	0.0153 (0.0240)
Biomarkers:
sTLR4 (ng/mL)	3.43 (12.10)	5.74 (30.35)	4.59 (23.17)
LBP (μg/mL)	17.07 (7.09)	17.59 (6.22)	17.32 (6.68)
IL-6 (ng/mL)	1.44 (1.22)	1.56 (3.24)	1.50 (2.45)

Values are mean (SD); BMI=body mass index; JSW=joint space width; uCTX-II=urinary collagen type II neoepitope normalized to urine creatinine; sTLR4=soluble Toll-like receptor 4; LBP=lipopolysaccharide binding protein; IL-6=interleukin-6

Figure 1.

Study flow diagram.

As described previously, exclusion criteria for the main trial included posttraumatic or any other form of secondary knee OA, the presence of inflammatory arthritis, comorbidity that would confound measurements of the outcomes, and a history of tetracycline allergy. Participants were also ineligible for enrollment if they had received an intra-articular injection of hyaluronan within the previous 6 months or corticosteroid within the previous 3 months.

Imaging

For this study, we used the previously described radiographic joint space width (JSW) data derived by automated image analysis of fluoroscopically standardized semiflexed AP views of knees obtained at baseline and 16 months [18]. The change in minimum JSW of the medial tibiofemoral compartment was used as the indicator of structural change from baseline to 16 months. Change in JSW over 16 months (ΔJSW₁₆ = JSW at 16 months - JSW at baseline) was analyzed as a continuous variable. We also analyzed, the change in JSN over 16 months (JSN₁₆) as a dichotomized variable using reduction of 0.5 mm in JSW over 16 months as the threshold to distinguish knee OA JSN progressors (-ΔJSW₁₆ > 0.5 mm) from non-progressors (-ΔJSW₁₆ <0.5 mm) based on a variable determined for analyses in the original study and contemporaneously recommended by the Group for the Respect of Ethics and Excellence in Science (GREES) [31].

Biomarker quantification

Only EDTA plasma was archived and available from the DOXY study. Plasma biomarker analyses were performed blinded to the clinical data and outcomes in the Kraus laboratory. Soluble TLR4 (sTLR4) was measured at baseline and 18 months using the Human TLR-4 high-sensitivity sandwich-based ELISA (RayBiotech®, Norcross, GA, USA). Lipopolysaccharide binding protein (LBP) was measured at baseline, 12 and 18 months using the Human LBP ELISA kit (Hycult®Biotech, Plymouth Meeting, PA, USA). The manufacturer has verified that plasma and sera yield similar LBP concentrations with this ELISA kit (mean LBP based on 6 donors of 11.2 μg/mL (range 5.3–28.1 μg/mL) for EDTA plasma and 11.9 μg/mL (range 4.3–27.3 μg/mL) for sera. IL-6 was measured at baseline and 18 months using the V-PLEX Human IL-6 kit from Meso Scale Discovery (MSD, Rockville, MD, USA). CTX-II (IDS, Bolton, UK) in the urine was quantified at baseline and 18 months using an ELISA based on a sequence found exclusively in human type II collagen. Urinary CTX-II (uCTX-II) was corrected for creatinine excretion levels quantified by ELISA (Quidel, San Diego, USA). The coefficients of variation (CVs) for the assays conducted in the Kraus laboratory were as follows: Intra-assay CVs were sTLR4 5.24%, LBP 6.28%, IL-6 2.19%, uCTX-II 3.5%; Inter-assay CVs were sTLR4 6.46%, LBP 8.80%, IL-6 6.26%, uCTX-II 6.42%. For the purpose of statistical analyses, any value that was below the lowest limit of detection, (LLOD) for the biomarker assay was imputed as ½ LLOD of the assay. Although one of many, this imputation method is well-established and known for its simplicity and low bias [32]. Imputations for values below LLOD were performed as follows: sTLR4: 98 baseline and 77 at 18 months; LBP: only 2 at baseline, 6, 12 and 18 months; CTX-II: only one sample at baseline; IL-6: no sample values imputed.

Statistical analysis

Outcomes of interest included continuous outcomes: ΔJSW₁₆ and ΔuCTX-II₁₈ (uCTX- II/creatinine at 18 months - baseline), and a binary outcome of radiographic progression (JSN). The outcome variable, uCTX-II, was log₁₀ transformed to achieve a normal distribution for purposes of statistical analysis for the measures at single time points. For change in uCTX-II over 18 months (ΔuCTX-II₁₈), to avoid negative values, we used the following transformation prior to log₁₀ transformation: ΔuCTX-II₁₈ + 1 − min(ΔuCTX-II₁₈ for the cohort). We also considered change in total Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) score over 18 months although a threshold level of knee pain was not among the inclusion criteria for this study and not unexpectedly, baseline clinical manifestations of OA in these subjects was mild as reported in the original clinical trial report [18].

Descriptive statistics are presented as means±SD for continuous variables and frequency (percentage) for categorical variables. To be able to directly compare strengths of associations across the biomarkers, sTLR4, LBP and IL-6 were Z-score transformed. We considered the following four types of biomarker measurements: baseline, 12 months (only for LBP), 18 months, and the time-integrated concentration (TIC) of the biomarker from baseline to 18 months as previously described [33]. Pearson, correlations were calculated between Z-score transformed biomarkers (LBP, sTLR4 and IL-6), and the continuous outcomes ΔJSW_16, ΔuCTX-II₁₈). To assess the association between biomarkers and outcomes, univariate linear regression was used for continuous outcomes ΔJSW₁₆, ΔuCTX-II₁₈), and a logistic regression model was used for the binary outcome (JSN₁₆). In addition to univariate analysis, multivariable regression analyses were performed to test the association between biomarkers and outcomes with covariate adjustment of treatment group and baseline outcome measure (e.g. baseline JSW or baseline uCTX-II/creatinine) in the initial model. Thereby, we analyzed change in the outcomes over time with adjustment for the baseline measure, and including age and BMI in the final model. Network plots were created to visualize the strength of correlations (Pearson using transformed variables for the analyses for the plot) among demographics, biomarkers, and outcomes of interests. All analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) with supplemental use of R software (version 3.3.2).

Results

Patient characteristics and outcomes

A total of 431 participants from the DOXY clinical trial were included in this study; 218 subjects were from DOXY treatment group and 213 from the Placebo group. All participants had unilateral knee OA at baseline, were females of similar age and in the upper tertile of the age-, race-, and sex-appropriate norms for body mass index (BMI) established by the Second National Health and Nutrition Examination Survey [34] (Table 1). The baseline biomarker (LBP, IL-6, and sTLR4) concentrations in the placebo and treated groups were comparable in magnitude and variability (Table 1).

Association of inflammatory biomarkers with OA structural changes

Adjusting for treatment group and baseline outcome variable, higher concentrations of LBP at baseline and TICs over 12 and 18 months (TIC(LBP)₁₂, TIC(LBP)₁₈) were associated with OA progression based on (decreasing) ΔJSW₁₆ (parameter estimates (β coefficient): −0.096 to −0.142) and (increasing) JSN₁₆ (OR: 1.36 to 1.50) (Table 2). Because the biomarker measures were Z-score transformed for all these analyses, these parameter estimates and OR represent the effect of one standard deviation increase of the biomarker predictor. With additional adjustment for age and BMI, the associations of LBP variables with OA progression were slightly attenuated but essentially unchanged (Table 3). The strongest of these associations was between TIC(LBP)₁₂ and ΔJSW₁₆ (parameter estimate −0.111), and between TIC(LBP)₁₈ and JSN₁₆ (OR 1.42) (Table 3). There were no statistically significant associations of sTLR4 or IL-6 with OA radiographic progression. The association of LBP with ΔJSW₁₆ and JSN₁₆ remained significant upon further adjustment for baseline sTLR4 (data not shown).

Table 2.

Association of inflammatory biomarkers with knee osteoarthritis structural progression (adjusted for treatment group and baseline outcome variable).

Outcome Variable	Biomarker†	n	Parameter Estimate (SE) or Odds Ratio* (95% CI)	p Value
	Z-LBP baseline	341	−9.63×10−2 (4.06×10−2)	0.018
	Z-TIC (LBP)12	253	−1.42×10−1 (4.90×10−2)	0.004
	Z-TIC (LBP)18	282	−1.20×10−1 (4.59×10−2)	0.009
ΔJSW16¶	Z-sTLR4 baseline	345	4.67×10−2 (5.03×10−2)	0.354
	Z-TIC (sTLR4)18	288	6.08×10−2 (4.64×10−2)	0.191
	Z-IL-6 baseline	343	−4.66×10−2 (3.95×10−2)	0.239
	Z-TIC (IL-6)18	285	−4.88×10−2 (4.63×10−2)	0.292
	Z-LBP baseline	341	1.357 (1.068, 1.724)	0.013
	Z-TIC (LBP)12	253	1.452 (1.097, 1.921)	0.009
	Z-TIC (LBP)18	282	1.497 (1.133, 1.977)	0.005
JSN16§	Z-sTLR4 baseline	345	0.751 (0.392, 1.439)	0.388
	Z-TIC (sTLR4)18	288	0.502 (0.161, 1.564)	0.235
	Z-IL-6 baseline	343	1.493 (0.878, 2.55)	0.142
	Z-TIC (IL-6)18	285	1.362 (0.884, 2.10)	0.161
	Z-LBP baseline	235	−2.87×10−4 (5.93×10−4)	0.6290
	Z-TIC (LBP)12	206	−1.34×10−4 (6.49×10−4)	0.836
	Z-TIC (LBP)18	233	−7.55×10−4 (5.81×10−4)	0.897
Log10(ΔuCTX-Il18)¶	Z-sTLR4 baseline	234	1.96×10−3 (6.19×10−4)	0.002
	Z-TIC (sTLR4)18	233	1.68×10−3 (5.42×10−4)	0.002
	Z-IL-6 baseline	233	1.61×10−3 (1.26×10−3)	0.203
	Z-TIC (IL-6)18	230	2.09×10−3 (1.18×10−3)	0.079

^¶Models were adjusted for treatment group and corresponding baseline radiographic measures (JSW baseline for ΔJSW₁₆; uCTX-II at baseline for ΔuCTX-II18);

^§Models were adjusted for treatment group;

^†Data were Z score transformed;

^*Parameter estimates (beta coefficient estimates and standard error) from linear regression for continuous outcomes. Odds ratios only given for logistic regression models. ΔJSW₁₆=JSW at 16 months– JSW at baseline; JSN₁₆ dichotomized1 vs 0 based on joint space narrowing >0.5mm vs <0.05mm over 16 months; ΔCTX-II18=urinary CTX-II at 18 months – uCTX-II at baseline; TIC=time integrated over 12 or 18 months as indicated by subscript; remainder of abbreviations as in Table 1; p<0.05 in bold

Table 3.

Association of inflammatory biomarkers (LBP and sTLR4) with knee osteoarthritis progression (adjusted for treatment group, BMI and baseline outcome variable^*).

Outcome Variable	Biomarker†	n	Parameter Estimate (SE) or Odds Ratio*** (95% CI)	p Value
	Z-LBP baseline	341	−7.08×10−2 (4.18×10−2)	0.090
ΔJSW16	Z-TIC (LBP)12	253	−1.11×10−1 (5.21×10−2)	0.035
	Z-TIC (LBP)18	282	−7.90×10−2 (4.79×10−2)	0.100
	Z-LBP baseline	341	1.319 (1.029, 1.692)	0.029
JSN16	Z-TIC (LBP)12	253	1.406 (1.042, 1.898)	0.026
	Z-TIC (LBP)18	282	1.418 (1.06, 1.897)	0.019
Log10 (ΔuCTX-Il18)**	Z-sTLR4 baseline Z-TIC (sTLR4)18	234 233	1.98×10−3 (6.25×10−4) 1.71×10−3 (5.48×10−4)	0.0017 0.0021

^†Biomarker Data were Z score transformed.

^*Covariates include treatment group, age, BMI, and corresponding baseline radiographic measures (JSW baseline for ΔJSW₁₆; uCTX-II at baseline for ΔuCTX-II₁₈)

^**log ((Δ uCTX-II₁₈) was based on log₁₀₍(Δ uCTX-II₁₈+1-min((Δ uCTX-II₁₈ for the cohort)).

^***Odds ratios only given for logistic regression models. Others are beta coefficient parameter estimates and standard error (SE) from linear regression for continuous outcomes. p<0.05 in bold

However, sTLR4 was significantly associated with worsening OA disease activity based on change (increase) over 18 months in the biochemical marker uCTX-II (ΔuCTX-II₁₈). Adjusting for treatment group and baseline outcome variable, both sTLR4 at baseline and TIC over 18 months (TIC(sTLR4)₁₈) were positively associated with ΔuCTX-N₁₈ (Table 2). With additional adjustment for age and BMI, all sTLR4 associations with ΔuCTX-N₁₈ were unchanged (Table 3).

Of the biomarkers, only IL-6 (both at baseline and TIC over 18 months) was statistically significantly associated with change in total WOMAC over 18 months. With adjustment (for treatment, age, BMI and baseline WOMAC), baseline IL-6 was marginally associated with change in WOMAC over 18 months (parameter estimate −1.26, standard, error (SE) 0.76, p=0.09) but IL-6 TIC-18 was not associated (parameter estimate −1.03, SE 0.87, p=0.24).

The network plot for measures across time points up to 18 months showed significant correlations among BMI, baseline LBP, TIC(LBP)₁₂ and TIC(LBP)₁₈ (Figure 2). Also, a significant correlation was observed between baseline uCTX-II and ΔuCTX-II₁₈ (Figure 2).

Figure 2.

Network plot including demographics, biomarkers and outcomes of interest. The network plot represents one-to-one Pearson correlations performed with z-score normalized biomarkers and log₁₀ transformed uCTX-II to satisfy assumptions of normality. Positive correlation (>0.1)-red, negative correlation (<−0.1)-green. For the measurements derived from the same biomarkers (e.g. LBP baseline, TIC(LBP)₁₂, TIC(LBP)₁₂), we do not show their correlation since we are mainly interested in correlation among different biomarkers and other measures.

Discussion

In this substudy of the DOXY clinical trial, plasma LBP at baseline and TICs of LBP over 12 and 18 months were associated with radiographic structural worsening of knee OA over 16 months. The fact that adjusted analyses, accounting for treatment, did not alter the association of inflammatory mediators (LBP and sTLR4) with OA progression outcomes suggests that the chondroprotective effects of doxycycline in this study were likely not mediated by antibiotic induced changes in LBP. LBP is a well-known acute- phase reactant produced primarily by hepatocytes [35]. LBP is activated by inflammatory mediators, such as IL-1 and IL-6, and directly or indirectly by LPS itself [36–38]. Systemic concentrations of LBP have been reported to range from 5–10 pg/mL in a healthy population measured using a sandwich ELISA [39], whereas the range in our OA cohort was 2.5–94.92 pg/mL similarly using an ELISA based assay. In humans, LBP triggers a dynamic endotoxin (such as LPS) transfer cascade by binding and transferring LPS to CD14, which transfer LPS to TLR4-MD-2 receptors on immune cells; LBP thereby concentrates LPS at the cell membrane of immune cells to stimulate an inflammatory response [40]. LBP also recognizes pathogens such as spirochaeta, mycobacteria, gram-positive bacteria and mycoplasma [41]. LBP binds pro-inflammatory constituents of both Gram positive and Gram negative bacteria [42], making it a more general marker of, bacterial exposures than LPS that stems only from Gram negative bacteria [38]. Measurement of LBP as an inflammatory mediator is generally preferred over direct measures of LPS due to the short half-life of LPS [43–45]. Moreover, although LPS measurements in blood are possible, the traditional Limulus amebocyte lysate (LAL) assay for LPS is not recommended for blood due to a host of inhibitors that interfere with the assay. Consequently, our study and most prior studies have utilized measures of LBP alone [46, 47]. Recently, Ge [48] and Nakarai et al. [49] suggested that LBP could also be considered an adipokine. Under conditions of positive energy balance, insulin resistance and inflammation, LBP is produced in excess through a TLR4 receptor- mediated mechanism by human adipocytes in response to inflammatory LPS or palmitate stimulation. Other studies indicate that LBP is a necessary factor for triggering an inflammatory cascade by saturated fatty acid and metabolic endotoxemia [50, 51]. In our previous study that evaluated the association of LPS with the severity of inflammation, symptoms and radiographic abnormalities in a relatively small knee OA cohort (n=25; 31 knees) [52], we observed higher serum LBP concentrations (16.4±5.3 pg/mL, range 8.2 to 26.1 pg/mL) than those reported for a healthy population [39]. This finding was further confirmed by the current study with a larger sample size. These results support a link of OA progression with low-grade inflammation that could be related, as we previously proposed [7], to an altered gut microbiome or metabolic stress.

Interestingly, another inflammatory mediator related to innate immunity, sTLR4, both at baseline and over 18 months, was associated with change in uCTX-II over 18 months, although we did not find a correlation between sTLR4 and radiographic progression based on ΔJSW₁₆ or JSN. To our knowledge, the current study is the first to quantify plasma sTLR4 in knee OA patients. Endotoxin tolerance is induced by reducing cell surface TLR4 through CD14-dependent endocytosis or protease induced shedding of the ectodomain of TLR4 [53]. This phenomenon has been reported for many tissues, including articular cartilage [54]. Thus, based on other studies, the concentration of soluble TLR4 (sTLR4) in the current study was taken to reflect a global burden of low- grade inflammation. These findings are consistent with an important role of innate immunity and TLR4 inflammatory pathways in OA. These systemic biomarker measures (sTLR4 and uCTX-II) are both dynamic and likely reflect immediate OA disease activity rather than subsequent anatomic severity represented by radiographic measures of OA. Many studies have shown that multiple cell types within the joint (including macrophages, fibroblasts, and chondrocytes) increase expression of TLR4 throughout the process of joint space narrowing [55–59]. The release of sTLR4 has been related to the degree of systemic inflammation in body fluids/organs such as amniotic fluid, saliva, blood, cerebrospinal fluid, kidney and liver [60–64]. In the context of renal injury, TLR4 is shed from the cell surface to become soluble (sTLR4), which is believed to prevent further activation of TLR4 signaling [61]. Ten Oever et al. [65] proposed sTLR4 as a potential systemic biomarker for inflammatory conditions, including rheumatoid arthritis. Recently, Barreto et al. [54] detected elevated sTLR4 in synovial fluid from OA compared with healthy knees. Our results are consistent with sTLR4 as a biomarker of OA disease activity.

In the current study, we also measured plasma IL-6 in all available samples from the DOXY clinical trial. Although baseline IL-6, after covariate adjustment, was marginally associated with worsening OA symptoms over 18 months, it was not associated with radiographic OA progression over 16 months or change in uCTX-II over 18 months. The interval of radiographic OA progression assessed in this substudy (16 months) may have been too brief to reflect the resulting structural abnormalities resulting from the concurrent molecular level disease activity indicated by the IL-6 and CTX-II results.

The current study was novel for its evaluation of LBP as a mediator of DOXY treatment effects on OA structural progression. Although our results did not support, microbiome modification (based on LBP) as the means by which DOXY slowed knee OA progression, it supports the involvement of low-grade inflammation as a pathoetiological mechanism in OA progression.

The current study had several strengths. First, this study was based on a multicenter, randomized, double-blind, placebo-controlled trial with a relatively large sample size. Second, we used features from carefully acquired radiographs and uCTX-II as indicators of OA progression. Third, all participants in the Doxycycline clinical trial were female. This would potentially enhance the homogeneity of the study population but potentially limit its generalizability to the population in general. One minor limitation related to the measurement of analytes in samples that had been stored frozen (albeit −80°C) for up to 18 years. Although, there is no available information on the stability of LBP in frozen samples over time, our results were in a similar range to those reported in studies in different populations using fresh samples [66, 67].

Conclusion

Our findings in the DOXY clinical trial indicate that LBP and sTLR4 demonstrate capabilities as prognostic biomarkers in knee OA patients. Our results support a role for low-grade inflammation, possibly related to metabolic stress and/or gut microbiome alteration in the pathogenesis of knee OA.