INCIDENT HAND OA IS STRONGLY ASSOCIATED WITH REDUCED PERIPHERAL BLOOD LEUKOCYTE TELOMERE LENGTH

Abstract

Objective:

To evaluate the relationship of telomere length to the prevalence and incidence of hand osteoart hritis in a longitudinal cohort.

Design:

We conducted a cross-sectional and longitudinal analysis of data from a subset of participants in the Osteoarthritis Initiative recruited between February 2004 and May 2006. 274 individuals were eligible for the study based on availability of both baseline and 48-month hand radiographs and peripheral blood leucocyte telomere length data. Mean telomere length of PBLs from the DNA samples was determined using a validated quantitative PCR-based assay, and hand radiographs were analyzed and graded using the Kellgren-Lawrence scale.

Results:

In joint –level analyses, prevalent IPJOA was significantly associated with PBL telomere length in the baseline sample in unadjusted analyses (RR=2.84; 95% CI:0.87–9.29) or in models adjusted for age, sex, and body mass index (aRR=1.10; 95% CI: 0.96–1.27). The association in crude and adjusted analyses appeared slightly stronger with incident IPJOA, especially in the subset with normal hands at baseline (aRR=1.62; 95% CI: 1.02–2.57). PBL telomere length was also associated with prevalent HOA at baseline (significant in unadjusted analysis: RR=1.22; 95% CI 1.06–1.42), but not after adjusting for covariates: aRR=1.12; 95% CI: 0.96–1.30). The magnitude of association was stronger for incident HOA, especially incident symptomatic HOA (aRR=1.53; 95% CI: 1.09–2.15).

Conclusions:

In summary, the results of this exploratory analysis are confirmatory of previous work showing a cross-sectional relationship between telomere length and HOA and add to the field by demonstrating an even stronger association with incident IPJOA, both radiographic and symptomatic.

INTRODUCTION

The ends of eukaryotic chromosomes are capped by a particular chromatin structure which constitutes the telomere. Telomeres are composed by specific tandemly repeated G-rich DNA sequences organized with arrays of protein complexes and a long non-coding RNA (TERRA: Telomere-Repeat-Containing-RNA), constituting a tightly packed higher order structure. This special organization is essential for the maintenance of the integrity of the genome, protecting the chromosomes from terminal degradation and end-to-end fusion^1,2. As a normal consequence of the process of DNA replication, telomeres lose ~50–150 base pairs in length during each cycle of division^3,4. Telomerase is a reverse transcriptase that replaces the telomeric sequences maintaining telomere length. This enzyme is mainly active in germ cells. Nevertheless, most of somatic cells do not show telomerase activity or it is not effective enough, so telomeres progressively shorten with ageing. Furthermore, telomere shortening may be increased by various stressors and by DNA damage induced by oxidative stress⁵. Once telomere sequence shortening reaches a critical threshold, the telomeric chromatin loses their capping structure and chromosomes become vulnerable to damage, leading to replicative senescence, cell arrest, apoptosis, and pathophysiological changes in tissues.

Since telomere DNA sequences decrease with ageing in somatic cells, their assessment is being considered a marker of biological age. In fact, several reports, suggest that the telomere length determined in peripheral blood leukocytes (PBL) predicts longevity^6–8 and susceptibility to age-related disease such as cardiovascular disease, diabetes, some cancers and dementia^9–13. Regarding osteoarthritis (OA), chondrocyte senescence associated with increased telomere shortening has been demonstrated in osteoarthritic cartilage¹⁴. Nevertheless, the relationship between OA and telomere size in PBL seems controversial. Whereas some epidemiologic studies have found reduced PBL telomere length in samples of people with OA relative to controls, other reports showed no evidence of statistically significant differences^15–17. These studies were vulnerable to imprecision consequent on small sample sizes and cross-sectional design. The objective of this study therefore was to test the relationship of PBL telomere length to the actual prevalence of hand OA in a longitudinal cohort. Furthermore, the longitudinal approach allowed us, for the first time, to prospectively assess the usefulness of PBL telomere length as a risk factor for future hand OA incidence.

METHOD

Study Sample

To study the association between radiographic hand OA and telomere length we conducted a cross-sectional and longitudinal analysis of data from a subset of participants in the Osteoarthritis Initiative (OAI) who had measurement of PBL telomere length from biospecimens obtained at baseline.

The OAI is a multicenter cohort study of 4,796 adults with or at risk for symptomatic knee OA. Four clinical sites (Memorial Hospital of Rhode Island, The Ohio State University, University of Maryland and Johns Hopkins University, and the University of Pittsburgh) recruited participants between February 2004 and May 2006. The eligibility criteria for the OAI were designed to enrich the cohort with individuals at risk for knee osteoarthritis and to exclude individuals with end-stage knee OA or inflammatory rheumatic disease. Ultimately three subcohorts were identified; Progression subcohort (n=1390), Incidence subcohort (n=3284) and Non-exposed control subcohort (n=122). OAI data and protocols are available for free public access¹⁸. Institutional review boards at each OAI clinical site and the OAI coordinating center (University of California, San Francisco) approved the OAI study. All participants provided informed consent prior to participation.

Biospecimens from a subsample of OAI participants had been analyzed for PBL telomere length for a previous investigation of mtDNA haplogroups and telomere length and knee OA among participants of Caucasian ancestry^18,19. Samples of Caucasian participants were selected as follows: all of the non-exposed subcohort (n=96; no specific distribution of haplogroups); 120 random samples from the progression subcohort with a similar distribution of the mtDNA haplogroups H, J and T (33% for each haplogroup); 96 random samples from the incidence subcohort with the mtDNA haplogroups H, J and T following the frequencies of these haplogroups in the general population (no specific distribution of haplogroups). The total number of DNA samples analyzed was 312.

Assessments

OAI participants received an inventory of assessments at baseline and at the 48-month visit that included postero-anterior radiographs of one or both hands, biospecimen collection, and questions about hand pain referencing a homunculus “During the past 30 days, which of these joints have had pain, aching, or stiffness on most days? By most days, we mean more than half the days of a month”. Self-reported physician-diagnosed hand OA was based on the response to “Has a doctor ever told you that you have osteoarthritis or degenerative arthritis in your hand or fingers?”

Telomere analysis

Biospecimens from a small subsample of OAI participants at baseline had been analyzed for PBL telomere length as a component of a research program evaluating the influence of mtDNA haplogroups on telomere length in the OAI as a comparison group to a white Caucasian Spanish cohort^11,20. Because of population genetic considerations this OAI sample was confined to Caucasian participants and was sampled across the progression (n=129), incidence (n=87) and non-exposed control subcohorts (n=96).

Mean telomere length of PBLs from the DNA samples was determined using a validated quantitative PCR-based assay²⁰. This measures the average ratio of telomere repeat copy number to a single gene (36B4) copy number (T/S ratio) in each sample. The quantitative PCR technique utilized a LightCycler II thermocycler (LightCycler 480, Roche Diagnostics, Laval, Quebec, Canada) in a 96-well format. Duplicate DNA samples were amplified in parallel 20uL PCR reactions that included 30 ng of sample DNA, the DNA master SYBR Green I kit (LightCycler® 480 Sybr Green I Master, Roche Diagnostics) and 500nM of specific primers. The technique was optimized by developing standard curves using serial dilutions from a reference DNA.

Hand radiographic scoring

We evaluated the severity of OA in each of the 2nd-5th distal interphalangeal (DIP) joints, 2nd-5th proximal interphalangeal (PIP) joints, 1st-5th metacarpophalangeal (MCP) joints, thumb interphalangeal (IP) joint, and thumb-base joints (i.e., first carpometacarpal (CMC-1) joint and the scaphotrapezial (ST) joint) of the dominant hand.

These joints were graded using a modified Kellgren-Lawrence (KL) scale that is based on presence of individual radiographic OA features as follows: 0 = no OA (no osteophyte or joint space narrowing); 1 = minimal OA, i.e. questionable osteophyte or joint space narrowing (JSN); 2 = mild OA, i.e. small osteophyte(s) and/or or mild JSN, sclerosis may be present; 3 = moderate OA, i.e. moderate osteophyte(s) and/or JSN, sclerosis and erosions may be present; 4 = severe OA, i.e. large osteophyte(s) and/or severe JSN, sclerosis and erosions may be present²¹.

A radiologist in training (LFS) performed the OA scoring after receiving training from a Board Certified Musculoskeletal radiologist (SES) and confirming good intra-reader agreement. The reader scored 100 randomly selected pairs of hand radiographs twice. Intra-reader agreement, based on weighted kappas, was good (weighted kappa > 0.84). The radiograph scoring process was facilitated by customized software, which displayed baseline and follow-up images side by side but blinded the reader to time. The completed readings then received a secondary review by a rheumatologist with an expertise in hand OA epidemiology and radiographic scoring (IDH). Scores for which there were disagreements of more than 1 KL grade (0.8%; out of 115,224 joints with readable KL grade) were adjudicated in consensus with the musculoskeletal radiologist (SES).

Definitions of interphalangeal and hand OA

We used joint-level and person-level definitions of hand OA for this analysis.

Joint-level:

• Prevalent interphalangeal OA (IPJOA) was defined on the basis of a KL score ≥2 in a finger or thumb interphalangeal joint

• Incident interphalangeal OA (IPJOA) was defined on the basis of an increase in KL score from ≤1 at baseline to ≥2 at 48 months

Person-level:

• Prevalent radiographic hand OA was defined on the basis of having prevalent IPJOA in one or more joints (excluding thumb base joints) on at least 2 different fingers

• Incident radiographic hand OA was defined on the basis of meeting the definition for prevalent at 48 months but not at baseline

• Incident symptomatic hand OA was defined when a participant did not meet criteria for SxHOA at baseline but did at 48 months

Statistical analysis

Descriptive statistics were generated to characterize the analytic sample. Telomere length was slightly skewed with a long upper tail, so we performed a log transformation to make the variable more normally distributed (prior to transformation, skewness=2.06 and kurtosis=6.53; after log transformation, skewness=0.33 and kurtosis=1.72). Since there are no clinical cut-points for telomere length, the variable was standardized as a z-score in order to provide easier interpretation. Similarly, all continuous variables were standardized to facilitate comparison in the multivariable models. In addition, we examined the relationship of telomere length with age using correlations and explored the linearity of the association using polynomial regression analysis with log(telomere) as the dependent variable and age as the independent variable. After determining that the correlation with age was modest [r(95% CI): −0.165 (−0.278, −0.047)] and linear within the scope of our sample (in the regression analysis, log (telomere) was modeled with age, (age)², (age)³ and (age)⁴; all higher order terms above linear (age) were not statistically significant (p=0.861)), we analyzed the standardized log of telomere length as a risk for OA prevalence and incidence.

Analyses were divided into a joint-level analysis (binary variables of prevalent IPJOA and incident IPJOA) and a person-level analysis (binary variables of prevalent hand OA, incident HOA, and incident symptomatic HOA). Additionally, we examined the total number of prevalent IPJOA and the total number of incident IPJOA per person. The statistical approach used to estimate the relative risk associated with telomere length and each of the previously mentioned outcomes was accomplished through the use of a modified Poisson regression with a robust variance estimator²². Using generalized estimating equations (GEE) with an underlying Poisson distribution for the dependent variable, robust variance estimator, unstructured correlation matrix, and a log link function, the relative risk of each standard deviation decrease in log of telomere length was estimated for each outcome in both unadjusted and adjusted GEE models. Model covariates included age, gender, and body mass index. In the joint-level analysis, the GEE model as parameterized not only allowed for the estimation of risk but also accounted for the within-person correlation among joints. Total number of OA joints and change in the number of OA joints were also modeled with the modified Poisson regression described above. Associations between model covariates and telomere length were examined in order to reduce the influence of potential multicollinearity. All analyses were conducted with SAS version 9.4 (Cary, NC).

RESULTS

Sample characteristics

274 individuals (156 women) were eligible for the study based on availability of both baseline and 48-month radiographs and PBL telomere length data. As a consequence of selection criteria for the parent genetic study, the sample compared to those of the overall OAI hand cohort was all Caucasian, younger, leaner, and had indices indicative of higher socioeconomic status as well as lower frequency of comorbidities including radiographic hand osteoarthritis (Table 1).

Table 1.

Characteristics of the Osteoarthritis Initiative (OAI) Hand Osteoarthritis (HOA) Sample

		OAI (N=3312)		SAMPLE (N=274)		p-value
Age, years	Mean (sd)	61.2	(9.1)	59.1	(8.7)	<0.01
Sex (female)	N (%)	1891	(57.1)	156	(56.9)	1.00
Race (white)	N (%)	2658	(80.3)	274	(100.0)	<0.01
BMI	Mean (sd)	28.6	(4.8)	27.4	(4.4)	<0.01
Smoking (never)	N (%)	1788	(54.0)	158	(57.7)	0.50
Income (>50k)	N (%)	1154	(36.2)	63	(24.0)	<0.01
Degree	N (%)	1316	(40.0)	139	(50.9)	<0.01
Diabetes	N (%)	229	(7.1)	8	(3.0)	0.01
Hypertension	N (%)	1694	(51.1)	109	(39.8)	<0.01
Physician diagnosed HOA	N (%)	531	(16.6)	39	(14.6)	0.39
Hand Pain	N (%)	737	(22.3)	54	(19.7)	0.40
Baseline RHOA	N (%)	1383	(41.8)	101	(36.9)	0.01
Baseline knee ROA	N (%)	1899	(57.3)	113	(41.2)	<0.01
Pain in joint(s) most days past 30 days
Either shoulder	N (%)	732	(22.1)	44	(16.1)	0.02
Back	N (%)	527	(15.9)	38	(13.9)	0.37
Either elbow	N (%)	298	(9.0)	28	(10.2)	0.50
Either ankle	N (%)	333	(10.0)	36	(13.1)	0.11
Either foot	N (%)	363	(11.0)	27	(9.9)	0.57
Either hip	N (%)	776	(23.4)	56	(20.4)	0.26
Jaw	N (%)	308	(9.3)	17	(6.2)	0.09
Neck	N (%)	559	(16.9)	34	(12.4)	0.06
Either wrist	N (%)	364	(11.0)	27	(9.9)	0.56
Either knee	N (%)	1702	(51.4)	77	(28.1)	<0.01

sd = standard deviation; ROA = radiographic osteoarthritis; RHOA = radiographic hand osteoarthritis

Telomere length was inversely associated with age, but the magnitude of the correlation was small [Spearman r(95% CI): −0.165 (−0.278, −0.047)]. Telomere length was also inversely associated with body mass index [Spearman r(95% CI): −0.234 (−0.343, −0.119)] and waist circumference [Spearman r(95% CI): −0.140 (−0.254, −0.022)].

At baseline among the 274 hands available for analysis there were 3832 readable joints (14 joints in 274 hands, 4 of which were unreadable), of which 481 had prevalent IPJOA (Table 2). Thus, there were 3351 joints eligible for incident IPJOA among which 47 (1.4%) developed this outcome. Among the subset with no hand OA at baseline (2370 joints eligible for incident IPJOA after excluding hands with any IPJOA at baseline), 18 joints developed this outcome.

Table 2.

Interphalangeal Joint Osteoarthritis (IPJOA) and Telomere Length

Prevalent IPJOA
N eligible	N cases	RR (95% CI)	aRR (95% CI)
3832	481	2.84 (0.87, 9.29)	1.10 (0.96, 1.27)
Incident IPJOA
3351	47	1.59 (1.20, 2.12)	1.55 (1.12, 2.16)
2370	18	1.52 (0.97, 2.34)	1.62 (1.02, 2.57)*

^*sensitivity analysis excluding individuals with any HOA at baseline

aRR = Relative risk adjusted for age, sex and body mass index; CI = confidence interval; HOA = hand osteoarthritis

Among the 274 hands at baseline, 101 met the definition for prevalent HOA and 21 for symptomatic HOA. There were 9 cases of incident radiographic HOA and 32 cases of incident symptomatic HOA. Of the 32 cases of incident symptomatic hand OA, 2 (6.3%) had symptoms at baseline but no radiographic hand OA, 3 (9.4%) had neither symptoms nor radiographic hand OA, and 27 (84.4%) had radiographic hand OA at baseline but no symptoms.

At baseline, each hand had on average 1.76 joints (Standard Deviation [SD]=2.43, range 0–13) with prevalent IPJOA. By 48 months, this average per hand increased by 0.16 OA joints (SD=0.43, range 0–2).

Associations

In joint –level analyses, prevalent IPJOA was statistically associated with PBL telomere length in the baseline sample in unadjusted analyses and in models adjusted for age, sex, and body mass index (Table 2). The PBL telomere length association in crude and adjusted analyses was statistically significant with incident IPJOA, especially in the subset with no HOA hands at baseline (62% increase in the risk of incidence of IPJOA for every standard deviation decrease in log of telomere length).

At the person-level, PBL telomere length was associated with prevalent HOA at baseline. This association was statistically significant in the unadjusted analysis, but the sample size was smaller, and the confidence limits were imprecise after adjusting for covariates (Table 3). The magnitude of association was stronger for incident HOA and incident symptomatic HOA (68% and 53% increase risk for incident HOA and incident symptomatic HOA, respectively, for each standard deviation decrease in log of telomere length).

Table 3.

Hand Osteoarthritis (HOA) and Telomere Length

Prevalent HOA
N eligible	N cases	RR (95% CI)	aRR (95% CI)
274	101	1.22 (1.06, 1.42)	1.12 (0.96, 1.30)
Incident HOA
173	9	1.64 (1.01, 2.68)	1.68 (1.02, 2.15)
Incident Symptomatic HOA
253	32	1.53 (1.15, 2.03)	1.53 (1.09, 2.15)

aRR = Relative risk adjusted for age, sex and body mass index.

Telomere length was also associated with the average number of joints per hand with prevalent IPJOA at baseline and with the increase in the number of OA joints (Table 4). For every standard deviation decrease in log of telomere length, we would expect a 21% increase in the prevalent number of joints affected with hand OA. Similarly, for each standard deviation decrease in log of telomere length, we would expect a 46% increase in the number of new joints affect by OA/hand.

Table 4.

Hand Osteoarthritis (HOA) and Telomere Length

Prevalent number of affected joints/individual
N eligible	N cases	RR (95% CI)	aRR (95% CI)
274	481 joints in 148 subjects	1.30 (1.11, 1.52)	1.21 (1.02, 1.43)
Increase in number of affected joints/individual
N eligible	N cases	RR (95% CI)	aRR (95% CI)
274	46 incident OA joints in 40 subjects	1.48 (1.09, 1.99)	1.46 (1.05, 2.05)

aRR = Relative risk adjusted for age, sex and body mass index.

DISCUSSION

In this exploratory analysis of a sample of participants in the OAI, we found statistically significant associations of PBL telomere length with incident radiographic OA in the interphalangeal joints of the hands. We also found a strong association of telomere length with incidence of both hand OA and symptomatic hand OA. The association with symptomatic hand OA appears to be predominantly driven by the report of new hand symptoms.

The telomere results that made this analysis possible were available serendipitously because of a separate genetic investigation. As such we regard this as a convenience sample. The sample was small, included only Caucasian individuals and was not representative of the overall OAI cohort. Individuals in this analysis were younger, leaner, had less comorbidity, and were of higher socioeconomic status. The main implication is that these results are not generalizable to non-Caucasian populations and so such studies need to be replicated in racially diverse samples. In addition, biases resulting from a convenience sample from a cohort that was selected on the basis of having risk factors for knee OA could influence the strength (but not validity) of observed relationships.

Despite these issues, our results are remarkably consistent with previous epidemiologic studies in this field. Zhai et al measured relative leukocyte telomere length in relation to presence of radiographic hand OA among 1086 Caucasian participants in the TwinsUK Adult Twin Registry²³. The sample was predominantly female with a mean age of 55 years, and 160 of these had prevalent hand OA. They found that after adjustment for age, sex, body mass index and smoking, PBL telomere length was shorter on average by 178 base pairs in subjects with hand osteoarthritis (p=0.04) and had a dose-response association with semi-continuous measures of osteoarthritis. Poonpet et al also found reduced leukocyte telomere length in a case control study comparing 80 patients with knee osteoarthritis receiving knee surgery compared with 60 age-matched controls, and an association of telomere length with certain angiogenic cytokines¹⁵. Tamayo et al measured mean PBL telomere length in patients with various rheumatologic diseases²⁴. However, the sample with osteoarthritis was small (N=34) and did not exhibit a difference from the control group.

A prior concern was that a strong relationship of telomere length with age could confound or threaten interpretation of the analytic models through excessive collinearity or residual confounding. Because of this we examined both the strength of the relationship and its linearity, using polynomial regression techniques. We found that although the relationship was statistically significant, the magnitude of the correlation was not sufficient to indicate vulnerability to excessive collinearity in the models. Also, using numerous polynomial models, we found that the relationship was sufficiently linear to permit simple adjustment for age by including it as a covariate in the multivariable models.

Another issue to consider is the likelihood of correlated outcomes among joint within an individual, which we addressed by using generalized linear mixed models. Collider bias can also be a problem in analyses nested in cohort studies²⁵; however, the selection of this sample was not conditioned on any outcome and both prevalent and incident OA were represented among the participants.

We found the distribution of telomere length to be highly skewed at the longer end of the range, so it was necessary to perform a log transformation in order to achieve sufficient normality for analysis. Since we also standardized this and all covariates, the interpretation of the relative risks generated by the multivariable models needs to take these transformations into account. Thus, for example, the relative risk of 1.68 for the relationship of baseline telomere length with incident symptomatic hand OA indicates that there was a 68% increase in the risk of incident symptomatic hand OA for each standard deviation decrease in log of telomere length.

Discerning the direction of cause and effect can be a problem in epidemiologic studies, especially in cross-sectional analyses. Availability of follow-up hand radiographs and longitudinal data was, therefore, a strength of our study. The cross-sectional relationships that we observed at baseline between telomere length and IPJOA were also present (and stronger) in the longitudinal analyses of incident OA and incident symptomatic OA. These findings strongly suggest an influence of telomere length of the subsequent development of OA. Moreover, the associations with both radiographic and symptomatic outcomes indicate that telomere length is broadly predictive for the clinical construct of OA in the hands and not just the structural component.

The precise nature of the relationship between telomere length from blood leukocytes and the presence or risk of future degenerative articular disease deserves further research. PBL telomere size decline and OA development are both age related phenotypes. PBL telomeres would reflect the biological age of the individual, which may not be coincidental to chronological age. Telomere length is determined by both genetic and non-genetic factors interacting throughout life. More precisely, heritability of PBL telomere size is estimated to be moderate to high, between 36% to 84%²⁶. Otherwise, environmental stressors and lifestyle factors like smoking or chronic psychological stress influence PBL telomere size^27,28. This complex and dynamic interaction of genetic and non-genetic factors make the PBL telomere length a proxy for assessing the arrangement of harmful determinants of disease²². Numerous population-based studies indicate that peripheral blood leukocytes telomere length may predict longevity^6–8 and susceptibility to age-related disease such as cardiovascular disease, diabetes, some cancers and dementia9–13. For example, studies of telomere length and cardiovascular disease in aggregate show relative risks in the range of 1.4 – 1.8 (shortest vs. longest third)¹¹. In the Bruneck cohort short telomere length at baseline predicted incident cancer independently of standard cancer risk factors (multivariable hazard ratio 1.60) and cancer mortality (HR 2.1)²⁹.

Regarding osteoarthritis, leukocytes and chondrocytes belong to different tissues with different cellular replacement kinetics, so their telomere sizes are different¹⁷. Nevertheless, the telomere size from leukocytes and from other tissues appears highly correlated in the same individual^30,31. One possibility is that PBL telomeres would be a passive indirect marker, secondarily affected in hand OA. It is being recognized that PBL telomere length points to systemic influences on the telomere maintenance in other tissues²² and this should also be true for articular chondrocytes. The genetic and non-genetic factors that intensify telomere shortening in OA chondrocytes, including inflammation and oxidative stress^32,33 could also exert their influence in hematopoietic stem cells, so short telomeres in blood leukocytes would be correlated to relative short telomeres in articular tissue in comparison to healthy cartilage¹⁷.

Accelerated telomere shortening implies that some telomeres from chondrocytes would achieve excessive attrition prematurely leading to accelerated articular senescence³⁴. In fact, telomere erosion and indicators of cellular senescence increase with age in articular cartilage¹⁴ and these changes appear to be amplified proximal to osteoarthritic lesions in the cartilage^16,35,36. Moreover, senescent cells due to short telomeres secrete a characteristic pattern of cytokines which induce a local inflammatory microenvironment, named SASP (Senescent-Associated Secretory Phenotype)³⁷. In senescent chondrocytes, the proinflammatory cytokine profile is accompanied by increased metalloproteinases, destroying the cartilage matrix and aggravating the OA evolution. Cell free telomeric RNA fragments (cfTERRA) have been recently identified as an extracellular mediator of the senescent secretory phenotype³⁸. Telomeric stress induces the over transcription of TERRA which is secreted in exosomes, inducing the cytokine expression in adjacent cells. Senescent secreted factors could reach the blood circulation and possibly have systemic effects. Under this hypothesis, chondrocyte telomeres would have a direct causal role in the development of OA.

Another possibility is that PBL telomeres reflect a systemic background of telomere shortness, i.e. of global accelerated ageing, which actively synergize with other local hazardous etiologies enhancing their effect and increasing the risk of development of the disease²². Depending on the target tissue affected by the harmful factors, different age-related diseases would possibly be boosted to develop. This hypothesis is attractive and may explain the appearance of different age-related diseases in different individuals, all of them demonstrated to be associated to shorter leukocyte telomeres, like cardiovascular disease, diabetes, dementia or certain types of cancer^9–13. Accordingly, a primary causal role of genetics of PBL telomere length in disease was disentangled from other environmental and life-style factors in a meta-analysis of genome wide association studies including 37,684 individuals⁹. It was seen that certain common alleles in seven genes related to telomere biology, which associate to shorter PBL telomere length, increased the risk of coronary artery disease.

In our case, predetermined short telomeres in chondrocytes may heighten the effect of present OA risk factors. One known high-risk factor in OA is mechanical stress. The application of shear stress to chondrocytes in culture causes release of reactive oxygen species (ROS), oxidative stress, and development of chondrocyte senescence, suggesting that cartilage degeneration in post-traumatic osteoarthritis may be mediated by accelerated chondrocyte senescence³⁹. Several studies have shown that the presence of oxidative stress induces telomere genomic instability, replicative senescence and dysfunction of chondrocytes in OA cartilage, suggesting that oxidative stress, leading to chondrocyte senescence and cartilage ageing, might be responsible for the development of OA^5,40,41. Oxidative stress may be more harmful for chondrocytes with previously shortened telomeres since they contain guanine-rich sequences, which are especially prone to oxidative damage. As a consequence, telomeres would accumulate single-strand DNA breaks and shorten under conditions of oxidative stress^5,32,33. ROS attack would accelerate the loss of telomeric sequences to the critical size mainly when telomeres are initially short, resulting in premature senescence of chondrocytes.

Oxidative stress and chronic sub-clinic inflammation associated with an altered metabolic profile are characteristic of obesity, a classical OA risk factor. PBL telomere length is shorter in obese individuals^42,43 and our study showed a clear influence of BMI in OA development and PBL telomere shortening.

In summary, the results of this exploratory analysis are confirmatory of previous work showing a cross-sectional relationship between telomere length and OA and add to the field by demonstrating an even stronger association with incident IPJOA, both radiographic and symptomatic. Even though these findings were generated from analysis of a convenience sample, they show remarkable internal and external consistency. These results strengthen the hypothesis of the presence of an accelerated ageing phenotype which associates or promotes premature chondrocyte senescence, compromising the capacity of cartilage and joint structures to maintain health and respond to biomechanical and inflammatory or metabolic stressors, leading to articular degeneration.