Long-term lithium treatment in bipolar disorder is associated with longer leukocyte telomeres.

Telomere shortening is a hallmark of aging and has been associated with oxidative stress, inflammation and chronic somatic, as well as psychiatric disorders, including schizophrenia and depression. Additionally, antidepressants have been found to protect against telomere shortening. However, pharmacological telomere studies are lacking in bipolar disorder (BD). Therefore, the objective of this study was to explore telomere length (TL) in patients with BD in the context of lithium treatment. We determined TL by quantitative real-time PCR using peripheral blood leukocytes. Participants were outpatients diagnosed with BD type 1 or 2 (n=256) and healthy controls (n=139). Retrospective case-control and case-case study designs were applied. Lithium response (LiR) was scored using the Alda-Scale. Lithium-treated BD patients overall, as well as those on lithium monotherapy, had 35% longer telomeres compared with controls (P<0.0005, partial η(2)=0.13). TL correlated positively with lithium treatment duration of >30 months (P=0.031, R(2)=0.13) and was negatively associated with increasing number of depressive episodes (P<0.007). BD patients responding well to lithium treatment had longer telomeres than those not responding well. This is the first study to report a positive effect of long-term lithium treatment on TL. Importantly, longer TL was also associated with a better LiR in BD patients. These data suggest that lithium exerts a protective effect against telomere shortening especially when therapeutically efficacious. We hypothesize that induction of telomerase activity may be involved in LiR in BD.

Introduction

Bipolar disorder (BD) is a severe and chronic disorder affecting about 1–2% of the population causing immense suffering and excessive costs.[1, 2] BD type 1 is characterized by alternating episodes of depression and mania, whereas BD type 2 is characterized by depressions and hypomanias.[3] Rapid cycling (RC) is a severe subtype of BD defined as four or more bipolar episodes within a year, often difficult to stabilize with treatment. It has been suggested that RC should be considered a separate diagnostic entity apart from the BD, but this topic is still under discussion.[4]

Lithium is the first-line mood stabilizer in BD and has a unique role in suicide prevention.[5] However, 40–50% of bipolar patients relapse within 2 years of lithium treatment.[6] Lithium response (LiR) clusters in families to a large extent, but it is difficult to identify these responders a priori among patients because of the lack of predictive biomarkers.[7] Lithium is thought to stabilize neuronal activities and to promote neural plasticity and neuroprotection.[8] Lithium has probably several additional modes of action that are important for its therapeutic effects, but these mechanisms remain poorly understood.

Telomeres are DNA nucleoprotein complexes forming the ends of linear eukaryotic chromosomes, which protect the latter from natural cellular erosion and fusion with each other. However, telomeres erode progressively with each cell division, partly because of the end replication problem and also because of oxidative stress, finally signaling cellular senescence and apoptosis.[9] The efficiency of TL maintenance has been reported to be affected by sex,[10] genetic variation (for example, in the TERT gene encoding the telomerase reverse transcriptase) and environmental factors (for example, cigarette smoking).[11] Aging and somatic diseases, such as cancer, cardiovascular diseases, diabetes, dementia and cognitive decline, have all been associated with telomere shortening.[12] Furthermore, early life stress, depression and hypocortisolemia have also been negatively correlated with TL, suggesting that shorter telomeres might be a cumulative measure of stress in life.[13] In contrast to the above-mentioned factors that lead to TL shortening, factors like physical activity and effective pharmacological treatment have been associated with increased TL and telomerase activity, respectively.[14, 15]

In psychiatric disorders, shorter leukocyte TL (LTL) has been associated with short sleep duration, phobic anxiety, schizophrenia and depression.[12, 16, 17] Previous studies also suggest that the number of depressive episodes is associated with shorter LTL[18] and that antidepressant pharmacological treatment protects against telomere shortening.[15, 17] In addition, shorter LTL has been associated with poor response to medical treatment in schizophrenia.[19] To our knowledge, only one study with regard to BD has been conducted, which using a small group of patients (n=28) found an association of BD and number of depressive episodes with shorter LTL.[20] However, despite neuroprotective action of lithium,[8] there has been no study to examine TL in BD patients with regard to lithium treatment or LiR. Thus, in the present study, we tested whether (i) lithium treatment in BD patients may affect LTL, (ii) LTL in BD patients may associate with lithium responsiveness and (iii) number of depressive episodes, as well as RC, may be linked to LTL.

Materials and methods

Subjects

Patients (n=789) with clinical diagnosis of BD were consecutively recruited from the Unit of Affective Disorders at Psychiatry Southwest, Huddinge Hospital, Stockholm. Life-time manic and depressive symptoms were assessed by a psychiatrist specialized in BD or by a trained psychiatric nurse using the modules for mania and depression in the Schedules for Clinical Assessment in Neuropsychiatry as previously described.[21] On the basis of these assessments, patients were considered as fulfilling the diagnostic criteria for BD type 1, 2 or not otherwise specified. The symptoms as well as the number of manias and depressive episodes, RC and mixed episodes were assessed, including the age of onset of mania and depression. Two cohorts were selected according to predefined diagnostic and treatment response criteria. The first one (set I) consisted of patients with excellent LiR (n=76) and sex- and age-matched patients with none or partial LiR (non-LiR, n=76), all with therapeutic serum concentration (0.5–0.9 mmol l−1) during a period of at least 3 months before DNA sampling. The second set (set II) consisted of RC patients (n=98) and sex- and age-matched patients without RC (non-RC, n=98). From these four groups, DNA was available from 65, 65, 97 and 94 individuals, respectively. LiR was measured according to the yearly reduction of numbers and/or severity of episodes after the beginning of lithium treatment, controlled for the duration of lithium treatment (defined as months on treatment with serum concentration of 0.5–0.9 mmol l−1), the age of onset of BD, the number of years with BD before treatment, the duration of lithium treatment before DNA sampling and the presence of other mood stabilizers, antidepressants, neuroleptics or sleep medication according to the Alda-Scale.[7, 22] In previous studies, LiRs were those who scored ⩾7 and non-LiRs were those who scored ⩽6 (range 0–10 points). In order to achieve two well-defined groups according to LiR, patients who scored 6 points were excluded. Thus, LiR were those who scored ⩾7 and non-LiR were those who scored ⩽5. No other thresholds for LiR or non-LiR were analyzed. The BD patients on therapeutic lithium serum concentration during at least 3 months were stratified into two subgroups: lithium monotherapy and lithium plus add-on medications. Healthy non-obese controls were selected from two longitudinal population-based studies in Stockholm, Sweden: the Diabetes Prevention Program[23] and the PART study of mental health.[24]

The study was approved by the Regional Ethical Review Board in Stockholm, Sweden, in accordance with the Helsinki Declaration of 1975. For patients, both written and verbal informed consent was obtained at a visit with a specialized psychiatric nurse during euthymic state. For controls, verbal or written informed consent was obtained according to procedures approved by the ethical committee. All individuals had full capacity to consent. The procedure was documented in the research protocol, fulfilling the Swedish legal requirements.

LTL determination by quantitative real-time PCR

Peripheral blood samples were collected in EDTA tubes and DNA was extracted using an SDS-urea and phenol/chloroform/isoamylalcohol procedure, followed by purification using illustra NAP-5 columns (GE Healthcare, Buckinghamshire, UK). Relative LTL was determined according to Cawthon et al.[25] and Kananen et al.[26] In brief, DNA samples (4 ng) were processed in triplicates both for the telomere (Tel) and the single-copy gene (hemoglobin-β, Hgb) PCR, within the same 384-well plate using Platinum® SYBR® Green qPCR SuperMix-UDG (Invitrogen, by Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. The primers (10 nℳ) for the telomere PCR were Tel1 (5′-GGTTTTTGAGGGTGAGGGTGAGGGTGAGGGTGAGGGT-3′) and Tel2 (5′-TCCCGACTATCCCTATCCCTATCCCTATCCCTATCCCTA-3′) and for the Hgb PCR were Hgb1 (forward primer; 5′-GCTTCTGACACAACTGTGTTCACTAGC-3′) and Hgb2 (reverse primer; 5′-CACCAACTTCATCCACGTTCACC-3′). The PCR temperature conditions for Tel and Hgb were 95 °C for 2 min followed by 40 cycles of 95 °C for 15 s and 62 °C for 1 min, including a dissociation curve to indicate specificity. The procedures were carried out on an ABI PRISM 7900 HT qPCR instrument (Applied Biosystems, Foster City, CA, USA). DNA samples with a standard deviation of >0.4 between triplicates or a Ct value of >36 were omitted from the analyses (n=13). The relative gene expression was calculated according to the comparative 2−ΔΔCt method, where ΔΔCt=ΔCtsample—ΔCtcalibrator sample and ΔCtsample=CtTel—CtHgb. Samples from BD patients and healthy controls were assayed in the same 384-well plate. Inter-plate calibration was performed by normalizing 2−ΔΔCt of each sample to the average of three control samples. A dilution series of a synthetic oligomer for both Tel (TTAGGGx14: 107, 10.7, 1.07, 0.107, 0.0107 and 0.00107 pg) and Hgb (119-mer: 15, 1.5, 0.15, 0.015, 0.0015 and 0.00015 pg) were included in each plate to create reaction-specific standard curves. The mean of the correlation coefficient of the standard curves was >0.99 for both the Tel and the Hgb reaction.

Statistical analyses

Differences in clinical characteristics between study groups were assessed using Pearson's chi-square test or Mann–Whitney U-test. LTL dependence on sex and age was determined using analysis of covariance (ANCOVA). Correlation between LTL and age was determined using nonparametric Spearman's correlation test. Likewise, partial correlation between LTL and duration of lithium treatment, controlled for age and gender, was also determined using nonparametric Spearman's correlation test. Difference in duration of lithium treatment between LiR and non-LiR groups was determined using nonparametric Mann–Whitney U-test. Dependence of LTL on duration of lithium treatment and on number of depressive episodes was assessed using linear regression while adjusting for age and sex; partial Cohens eta square (η2) indicates effect size (proportion variance explained) similar to R2 in ANCOVA; η2=0.01 small; η2=0.06 moderate; η2=0.14 large effect. Residuals from linear regression analyses were controlled for normality. Dependence of LTL on the dichotomous variables RC and response to lithium treatment was tested using ANCOVA adjusted for age and sex. Similarly, difference in LTL between BD patients on lithium and healthy controls was assessed in the sex- and age-matched sample (age range: 33–77 years) using ANCOVA, while adjusting for sex and age, and was then verified using nonparametric Mann–Whitney U-tests. A P-value of <0.05 was regarded as statistically significant. The analyses were performed using IBM SPSS Statistics v. 20 (IBM Corporation, Armonk, NY, USA).

Results

The characteristics of the study groups are shown in Table 1. LTL correlated negatively with age (F=42.6, P<0.0005, Spearman's ρ=−0.28), but was independent of sex (P>0.05) in patients and controls. There was no age × sex interaction on LTL (P>0.15).

Table 1

Clinical characteristics of the study groups

Group	na	Males (%)	Age (years)a	Disease duration (years)a^	Mixed episode (%patients)	Mania or hypomania (nepisodes) ab	Depression (nepisodes) ab	RC (%patients)	BD2 (%patients)
Set I
Lithium responder	65	44.6	50 (36,63) 21–89	28 (15–40) 4–72	48.4*	3 (2,6)*	6 (5,6)***	21.5*	8.6*
Non-lithium  responder	65	45.3	52 (44,62) 22–84	30 (22,38) 1–58	32.3	6 (3,6)	6 (6,6)	43.8	17.9
Set II
Rapid cycling	97	40.4	49 (35,61) 23–80	22 (13,39) 1–56	46.4	6 (3,6)**	6 (6,6)***	100	0*
No rapid cycling	94	46.4	52 (40,62) 21–78	26 (15,36) 6–62	31.9	4 (2,6)	6 (3,6)	0	23.7
Combined
Bipolar disorderc	256	43.0	50 (38,62) 21–89**	27 (15,37) 1–72	39.6	5 (2,6)	6 (4,6)	43.8	12.8
Bipolar disorderd	207	45.5	53 (43,63) 33–77	29 (20,38) 1–58	39.9	5 (3,6)	6 (5,6)	42.7	12.6

Abbreviations: BD, bipolar disorder; RC, rapid cycling.

Median (25th, 75th percentile) range.

Six means more than five episodes.

Set I plus Set II. Some patients were included in both Set I and Set II.

Those in Set I and Set II with therapeutic serum lithium concentration (0.5–0.9 mmol l−1) in at least a period of 3 months before DNA sampling, with same age range as controls. Some patients were included in both Set I and Set II.

*P<0.05; **P<0.01; ***P<0.001 compared with the ‘No' category (Set I and Set II), compared with controls (combined sample) using Pearson's χ2 test or Mann–Whitney U-test.

LTL and lithium treatment duration in BD

As lithium has a neuroprotective effect,[8] we tested whether lithium exposure influenced LTL. This was assessed in Set I comprising of age- and sex-matched LiR and non-LiR BD patients with a duration of lithium treatment that was for LiR: 52 (21,172), 3–468 months, and for non-LiR: 46 (21,113), 3–360 months (numbers indicate: median (25th,75th percentile) range). Lithium duration did not differ between the LiR and the non-LiR groups (P=0.36, Mann–Whitney U-tests). LTL correlated positively with lithium treatment duration given exposure for more than 30 months (P=0.044, Spearman's ρ=0.24) and was also positively dependent on lithium treatment duration as determined by linear regression (n=72; lithium treatment duration: standardized (std) β=0.29, t=2.2, P=0.031; age: std β=−0.41, t=3.1, P=0.003, R2=0.13; Figure 1). Sex had no effect on this association. In the whole of Set I, with a lithium duration range of 3–468 months, LTL was not significantly dependent on lithium treatment duration (n=121; lithium treatment duration: std β=0.13, t=1.2, P=0.22; age: std β=−0.48, t=4.5, P<0.0005).

Figure 1

Lymphocyte telomere length (LTL) was positively dependent on duration of lithium treatment among those exposed for more than 30 months (n=71, P=0.032, Spearman's ρ=0.24). Axes represent unstandardized residuals of regressing LTL on age and sex, and regressing duration of lithium treatment on age and sex.

LTL and number of episodes

There was a tendency for shorter LTL in the RC cases compared with the age- and sex-matched non-RC cases (Set II), which was, however, not statistically significant (Δ=6%, F=2.16, P=0.14, ANCOVA adjusted for age (P<0.0005) and sex (P=0.57); P=0.093, Mann–Whitney U-tests; Table 1 and Figure 2a). When assessing the combined group of RC and non-RC cases, there was a significant effect of the number of depressive episodes on LTL, with an effect size similar to that of age (depressive episodes: std β=−0.20, P=0.007; age: std β=−0.26, P<0.0005, R2=0.14; Figure 2b). The LTL (relative quantity) marginal mean was reduced 0.075 units per depressive event after adjusting for age and sex assuming linear model. This dependence on number of depressive episodes was stronger in males (std β=−0.32, P=0.006) than in females (std β=−0.11, P=0.29). As there was an effect of the number of depressive episodes on LTL, the LTL dependence on lithium treatment duration was corrected for the number of depressive episodes. However, the number of depressive episodes was not found to influence the association between LTL and lithium treatment duration (P=0.62). For manic/hypomanic episodes, no dependence between LTL and number of episodes was found.

Figure 2

Lymphocyte telomere length (LTL) point estimate was reduced in those with rapid cycling (RC; P=0.14) (a), and LTL was significantly reduced with increasing number of depressive episodes (P=0.007) (b), with larger effect in males (P=0.006). The scatter plot data shown are from males and females. In (a), the y axis represents LTL relative quantity (RQ) values and patients in RC and non-RC groups are matched for age and sex. In (b), the y axis represents unstandardized residuals of regressing LTL (RQ) on age and sex. Bars and dots indicate the mean and error bars indicate standard error of the mean.

LTL and response to lithium

LTL was also assessed for difference between LiR and non-LiR groups in Set I. LTL was 10% longer in the LiR patients compared with the non-LiR patients after adjusting for all the parameters here suggested to influence LTL (LiR vs non-LiR: F=4.31, P=0.047, partial η2=0.13; age: F=10.8, P=0.003; number of depressive episodes: F=2.29, P=0.014; lithium treatment duration: F=1.41, P=0.15; sex: F=0.61, P=0.44: ANCOVA; Figure 3). As shown here, increasing number of depressive episodes had a shortening effect on LTL also in Set I. The absence of a significant effect of lithium treatment duration on LTL in the above-mentioned model was due to the inclusion in the analysis of those individuals with a short lithium treatment duration.

Figure 3

Lymphocyte telomere length (LTL) was increased in those responding well to lithium treatment compared with those not responding well to lithium treatment. LTL was increased in bipolar disorder (BD) patients, also in those on lithium monotherapy (BDLi-mono), compared with healthy controls with similar sex and age distribution. Not only the lithium responders, but also the non-lithium responders had longer LTL compared with healthy controls (P<0.0005). The y axis represents LTL relative quantity (RQ) values and patients in LiR and non-LiR groups are matched for age and sex. Bars indicate the mean and error bars indicate standard error of the mean.

LTL in BD compared with healthy controls

The mean value of LTL was 34.5% higher in BD cases (n=202) compared with healthy controls (n=135; F=49.9, P<0.0005, partial η2=0.13, ANCOVA adjusted for sex (P=0.10) and age (P<0.0005, partial η2=0.043); P<0.0005, Mann–Whitney U-tests; Figure 3). Not only LiR, but also non-LiR individuals (n=60) had longer LTL compared with controls (F=37.8, P<0.0005, ANCOVA adjusted for sex (P=0.096) and age (P=0.035); P<0.0005 Mann–Whitney U-tests). Furthermore, the patients on lithium monotherapy (n=39; 53.8% males; age 55 (39, 63), 33–73 years) had a mean LTL value that was 34.7% higher compared with controls (F=26.7, P<0.0005, partial η2=0.14, ANCOVA adjusted for sex (P=0.080) and age (P=0.008); P<0.0005 Mann–Whitney U-tests; Figure 3).

Discussion

This is the first study to report an association between lithium treatment in BD patients and LTL. BD patients on lithium had markedly longer LTL telomeres compared with healthy controls. This latter result remained significant, and equally strong, when LTL was assessed in BD patients on lithium monotherapy. The findings also suggest that duration of lithium treatment (more than 30 months) in BD patients correlate positively with LTL. Further, there was a tendency for BD patients responding well to lithium to have longer LTL than that poor responders had. Conversely, the number of depressive episodes in BD patients was associated with shorter LTL, with a stronger effect in males. This is in line with the only previous study of TL in BD, which found that BD type 2 and number of depressive episodes were associated with shorter LTL.[20] However, the latter study did not investigate drug response on LTL, as only 2 of the 28 patients were on mood stabilizers (lithium or valproate). The effect sizes of the findings were moderate to large (partial η2 and R2⩾0.13).

With regard to the lithium findings, one should bear in mind that (i) there are no lithium receptors, and (ii) the small Li+ ions enter all cells, replace sodium ions, compete with magnesium binding, and have a variety of inhibitory and/or facilitatory effects on enzymes and signaling systems. A wealth of findings has documented effects of lithium on monoaminergic systems, neurotrophic factors, neuropeptides and the arachidonic acid cascade. Lithium has also been found to enhance neuroprotective proteins and pathways (such as Bcl-2 and the Wnt signaling pathway), and to inhibit the key enzymes glycogen synthase kinase-3 (GSK-3, isoforms α and β) and phosphatidylinositol phosphatases, either directly or indirectly by affecting protein complexes that can modify the Akt/GSK-3 signaling.[27, 28, 29, 30] Besides the wide usage of lithium in both phases of BD and in major depressive disorder, investigations from the past decade have also focused on the drug's effects on a number of central nervous system disorders, including Alzheimer's disease, multiple sclerosis, amyotrophic lateral sclerosis, Huntington's disease and stroke (glutamate neurotoxicity).[8] The widening of lithium's therapeutic perspective has emerged as a consequence of experiments showing that lithium is neuroprotective and anti-inflammatory, and that it also affects cell proliferation/neurogenesis, neuronal plasticity and hippocampal volume.[31, 32, 33, 34]

With regard to a specific effect of lithium on TL, mechanistic support for the association between lithium treatment and LTL was recently provided.[35] hTERT, a catalytic subunit bearing the enzymatic activity of telomerase, is the rate-limiting determinant of human telomerase activity, whereas the other subunits are constitutively expressed.[36] In general, hTERT is considered to be regulated primarily on the transcriptional level. Specifically, transcription of hTERT has been found to be activated by β-catenin, which binds to and removes repressor T-cell factor 4 (TCF4) from the hTERT promoter, resulting in increased telomerase activity and longer TL in human cancer cell lines and human embryonic kidney cells.[35] β-Catenin is indirectly regulated by lithium with the main mediator between lithium and β-catenin being GSK-3β. β-Catenin is phosphorylated by GSK-3β, leading to β-catenin degradation. This elimination of β-catenin prevents it from reaching the nucleus for removing TCF4. Lithium inhibits GSK-3β by several mechanisms[33] leading to retention of β-catenin that can activate hTERT transcription.[33] GSK-3β, β-catenin and hTERT are expressed in lymphocytes as well as in the brain.[35, 37, 38] GSK-3β is a kinase abundantly expressed in the brain regulated by numerous signaling pathways, thus activate hTERT transcription. Lithium inhibits GSK-3β directly and reversibly, by regulating its transcription and by competing with Mg2+ and thereby disrupting the catalytic function of GSK-3β.[39] Lithium also indirectly regulates GSK-3β activity by inducing various kinases and phosphatases, such as PI3K, PP2A, PKC and Akt, controlling phosphorylation at GSK-3β serine-9, which keeps GSK-3β inactive. The downstream molecular targets of GSK-3β are many, and the resulting functional consequences of GSK-3β inhibition depend on the target. Inhibition of GSK-3β is proposed to be one of the two core mediating events of lithium's neurotrophic and neuroprotective effects including increased neurogenesis, axonal elongation, dendrite growth, as well as reduced apoptosis and neurodegeneration.[40]

The second core mediating event of lithium's neuroprotective effect is the inactivation of phosphatidylinositol phosphatases.[40] This inactivation blocks the recycling of inositol, which is essential for IP3-mediated signaling from G-coupled receptors and tyrosine receptors. Furthermore, this lithium-induced inositol depletion increases intracellular degradation of dysfunctional proteins or cellular components (autophagy) and stabilizes the neuronal membrane through downregulation of Myristoylated Alanine-Rich Protein Kinase C Substrate.[41] Thus, in the form of downstream events, lithium influences a variety of biochemical systems that are involved in neuronal survival and differentiation.[40, 42, 43] Because of the numerous downstream molecular targets of GSK-3β, the abundance of hTERT regulators and the plethora of enzymes dependent on Mg2+ with which Li+ may compete, there are probably several links between lithium and TL yet to be discovered.

In this study, the association of lithium with LTL was seen only after long-term treatment at a therapeutic dose (serum concentration of 0.5–0.9 mmol l−1). Clinical empirical data have shown that it usually takes 6–12 months of therapeutic lithium dose to reach maximum prophylactic effect in BD.[3] The temporal association of lithium with LTL might be related to the clinical response to treatment.

An attempt to score the clinical response to lithium treatment has been provided through the Alda-Scale, where high score indicates LiR, that is, marked reduction in yearly number of episodes.[7] Those with good response to lithium treatment had longer LTL than those with poor response. The LiR-LTL association did not appear to be confounded by the longer duration of lithium treatment in the LiR group compared with the non-LiR patients. Instead it may be due to the higher number of depressive episodes among non-LiR than among LiR (P<0.0005, Mann–Whitney U-tests). Also, we cannot rule out the possibility of different lithium concentrations in the serum of LiR and non-LiR groups, although such a difference would be unexpected due to the treatment routines that are followed.

Previous studies have linked stress and psychiatric disorders such as schizophrenia and depression to shorter telomeres.[12, 16, 17, 19] Accordingly, in this sample an increased number of depressive episodes was associated with shorter LTL in the BD patients. Thus, we hypothesize that LTL in BD patients will reflect a balance between the number of depressive episodes and the duration of treatment with therapeutic lithium concentrations. Remarkably, the LTL was markedly longer in BD patients compared with healthy controls. The LTL was longer than the controls in both the LiR group and the non-LiR group. Previously, shorter LTL in schizophrenia was associated with poor response to antipsychotic treatment,[19] and higher telomerase activity was associated with antidepressant pharmacological treatment in unipolar depression.[19] Similarly, lithium might influence the telomerase activity and contribute to longer LTL. Therefore, we studied BD patients on lithium monotherapy as a separate subgroup and the results showed that the lithium treatment association with LTL was not confounded by other drugs. There is, to our knowledge, no previous report on LTL with regard to manic/hypomanic episodes. In this study, no association was found between LTL and number of manic or hypomanic episodes. On the other hand, TL has repeatedly been found to be inversely correlated with increased age. Accordingly, we found that LTL decreased with increased age both in patients and controls. Consequently, age was controlled for in the statistical analyses. Similarly, we also controlled for sex and the association between LTL and number of depressive episodes was found to be stronger in males.

The intriguing possibility that a decreased length of telomeres could be a marker of BD disease, based on that a wide variety of insults and pathologies reduce TL, cannot be answered by our data as almost all BD patients were treated with lithium, well established as mood stabilizer in Sweden.[44] In this study, BD patients had markedly longer telomeres than the healthy controls. Our heuristic hypothesis is that the effects of lithium treatment were overriding the impact of negative events. In order to clarify this issue, samples from a cohort of not previously treated BD patients are needed, which is part of our current project planning. Other weaknesses in this study include: (i) The study design was cross-sectional/retrospective yielding data that were only associative in nature. Investigations of the putative causality between lithium and LTL should be pursued in the future. (ii) Shorter LTL has been reported to be associated with cigarette smoking, obesity, inflammation and several somatic diseases.[11, 34] These factors were not controlled for in the present study as these data were not available. Nonetheless, there is no clinical indication of less nicotine use or lower body weight in BD patients compared with the general population in Stockholm, Sweden, where the samples derived from; (iii) Almost all Swedish BD patients have received other pharmacological treatments before lithium because of the well-known delay in correct BD diagnosis. Therefore, it was not possible to study the effect of each drug separately before the start of lithium treatment using the current study design. Instead, BD patients on lithium monotherapy were studied as a subgroup, which supported that the data showing longer LTL after lithium treatment were not a consequence of therapy with other drugs. (iv) The number of depressive episodes and the number of manic or hypomanic episodes larger than six was indicated as ‘6' implying loss of information. However, the LTL residuals, after correction for age and sex for those with more than five depressive episodes, displayed a narrow 95% confidence interval at LTL values higher than the point estimate of the five episodes group. (v) Similar to previous TL studies of psychiatric disorders and stress, TL was here measured in a peripheral tissue and not in the brain where lithium is considered to exert its therapeutic action. (vi) For the test of association for LTL to RC and number of depressive episodes in sample Set II, a significance threshold of P<0.025 may be appropriate as LTL was studied for dependence also on number of manic/hypomanic episodes in this set. The finding that LTL was negatively associated with number of depressive episodes (P=0.007) survived the more stringent significance threshold, and was also confirmed using Set I.

Clinical implications

Our findings add significantly to knowledge regarding the putative mechanisms of lithium action. Specifically, the results showed markedly longer LTL in BD patients treated with lithium, suggesting a lithium-induced increase of telomerase activity during lymphopoiesis. A similar lithium effect during neurogenesis would be in line with previously proposed neuroprotective effects of lithium, as telomerase activity and TL influence neural stem cell proliferation and neurogenesis in neurogenic regions of the adult brain.[45] Importantly, the LTL was increased in BD patients responding well to lithium compared with non-responders. Lithium treatment is the first line of maintenance treatment in BD but a significant number of patients are only partial responders and a significant proportion of patients require add-on pharmacological treatment with antiepileptics and antipsychotics in order to prevent relapses. Although a distinct family history of bipolarity, typical clinical picture and the mania–depression–remission sequence of disease are good clinical LiR predictors, no biomarkers of response have been identified so far. Identification of a response trait marker would alleviate a significant unmet medical need as it would enable a priori choice of optimal individual maintenance treatment. Furthermore, identification of such a marker could guide additional investigations of lithium's therapeutic role in neurodegenerative and other disorders.[8] Finally, it remains to be tested if agents influencing TL might be useful as augmentation agents in treatment of intermediate lithium responders.

Conclusions

These data suggest that long-term lithium treatment in BD patients results in longer LTL with an effect size that is stronger than the telomere shortening effect of depressive episodes. The association between lithium and LTL is supported by a previously reported mechanistic pathway linking lithium and telomerase in vitro. We propose prospective studies of telomerase activity and TL in lithium-treated BD patients in order to elucidate LTL's role in BD and in the therapeutic effects of lithium.