Cerebrospinal Fluid Biomarkers of Central Nervous System Inflammation Predict Cortical Decline in Bipolar Disorder and Ventricular Enlargement in Healthy Controls

Bipolar disorder is a mental disorder involving mood episodes of (hypo)mania and depression, which are interspersed with periods of stable mood and remission of symptoms, known as euthymia [1, 2]. It is associated with impaired psychosocial functioning, increased substance use, higher risk of cardiometabolic diseases, and elevated mortality and suicide rates as compared to the general population [2, 3]. Some individuals with bipolar disorder may experience a progressively worsening illness trajectory, which includes more frequent and more severe mood episodes, incomplete remission, and insufficient treatment response [4‒6]. This pattern of symptomatic acceleration suggests a potential neuroprogressive decline in some individuals with bipolar disorder [4]. The biological processes underlying these deteriorating illness trajectories remain elusive.

Several magnetic resonance imaging (MRI) studies comparing brain structures between individuals with bipolar disorder and healthy controls have identified illness-related abnormalities in bipolar disorder [1, 7‒9]. Systematic reviews of cross-sectional studies have shown that individuals with bipolar disordered exhibit thinner gray matter in the frontal, temporal, and parietal regions, especially in the left pars opercularis (a part of the inferior frontal gyrus), fusiform gyrus, and rostral middle frontal cortex, compared with controls [7]. In studies of subcortical brain structures, enlarged ventricles were the most replicated finding, showing the largest difference from controls [10, 11].

While these cross-sectional studies highlight brain structure differences associated with bipolar disorder, they cannot determine disease trajectories or neuroprogression, nor can they distinguish abnormalities caused by the disease from those preceding its symptomatic onset. Longitudinal studies are therefore more appropriate. The most robust findings from longitudinal MRI studies include faster increases in ventricular volume and reductions in gray matter volume and cortical thickness in the frontal regions of individuals with bipolar disorder, which correlate with the occurrence of manic or hypomanic episodes [1, 12, 13]. Furthermore, faster thinning in the middle temporal cortex has been observed in cases as compared to controls [14‒16]. Some studies have also associated fluid biomarkers and brain-imaging metrics with illness staging (see [17] for an overview) but no study has ever investigated potential cerebrospinal fluid (CSF) biomarkers of neuroprogression in relation to longitudinal structural brain changes.

Similar to neurodegenerative disorders, some evidence suggests a relationship between bipolar disorder and both central nervous system (CNS) and peripheral inflammation [4, 6, 18, 19]. Inflammation has been suggested to contribute to the development of bipolar disorder and to be associated with symptomatic fluctuations, such as increased levels of pro-inflammatory cytokines like interleukin 1, 6, and 8 (IL-1, IL-6, IL-8) during both mania and hypomania [6, 18]. The latter is interesting given the link between (hypo)mania and cortical decline in frontal regions [1, 12]. IL-6 has been proposed as a biomarker of bipolar disorder, as it has been reported to be elevated across depressive, (hypo)manic, and euthymic phases [20].

Inflammation is also considered a potential driver of brain changes in bipolar disorder [19]. Peripheral inflammation markers have been associated with structural MRI changes in regions involved in affective and somatomotor processing and with lower brain volumes [15]. Higher serum concentrations of chitinase-3-like protein 1 (CHI3L1; also known as YKL-40) have been linked to lower cortical volumes in areas such as the left anterior cingulum, left frontal lobe, right superior temporal gyrus, and supramarginal gyrus in bipolar disorder [21]. Elevated serum IL-6 concentrations have been associated with thinner right middle temporal gyri [22].

However, no studies have yet linked brain abnormalities in bipolar disorder with inflammation biomarkers from CSF, which more accurately reflect CNS processes than blood serum derived biomarkers. Although CSF studies have found associations between bipolar disorder and inflammation markers like YKL-40, IL-1β IL-8, IL-6 [23‒25], these have not been investigated in relation to brain abnormalities or longitudinal structural brain changes.

This study aimed to investigate the association between CSF inflammation biomarkers and longitudinal changes brain structure, as measured with MRI. Based on prior research, we expected higher concentrations of IL-6, IL-8, and YKL-40 to be associated with cortical decline in the inferior frontal gyrus and middle temporal cortex, as well as volumetric enlargements of the lateral ventricles in bipolar disorder.

Study participants included individuals diagnosed with bipolar disorder type 1 or type 2, and healthy controls from the St. Göran Bipolar project, a naturalistic, longitudinal study conducted at the Northern Stockholm psychiatric outpatient clinic for bipolar disorders. All participants underwent a comprehensive diagnostic evaluation performed by board certified specialists in psychiatry or psychiatric residents. Diagnoses were made using the DSM-IV repetition time criteria and the SCID-I diagnostic interview, following a modified Swedish version of the Affective Disorder Evaluation [26]. The DSM-IV criteria have consistently been utilized since the study’s inception in 2005 to ensure reliable assessments and comparisons over time. The M.I.N.I. Neuropsychiatric interview was used for differential diagnostics and to screen for comorbid psychiatric conditions [27]. Final diagnoses were determined by consensus among a panel of psychiatrists specialized in bipolar disorder. Additional details on recruitment, inclusion criteria, and diagnostic procedures are provided in [27].

Healthy controls were randomly selected by Statistics Sweden, a government agency for official statistics, and matched by age and sex to the bipolar disorder participants. They were evaluated by a psychiatrist using the M.I.N.I. interview and parts of the Affective Disorder Evaluation [27]. Exclusion criteria for controls included any previous or ongoing psychiatric illness, any ongoing somatic illness, first-degree relatives with schizophrenia or bipolar disorder, substance or alcohol abuse, or neurological conditions, see [28] for further details.

Written and informed consent was obtained from all participants before inclusion. The study adhered to the Declaration of Helsinki and was approved by the Regional Ethics Committee in Stockholm, Sweden.

For this study, we only included participants who completed MRI scans at two timepoints and provided CSF and blood samples at baseline assessment. This resulted in a total of 63 participants: 29 with bipolar disorder and 34 healthy controls.

Manic symptoms were assessed using the Young Mania Rating Scale (YMRS), and depressive symptoms with Montgomery-Åsberg Depression Rating Scale (MADRS). Global functioning was evaluated using the function and severity ratings of the Global Assessment of Functioning (GAF) scale. Participants with bipolar disorder were euthymic at all study timepoints as assessed by the treating psychiatrist during clinical interviews.

MRI data were collected at the MR Research Centre, Karolinska University Hospital, Stockholm using a General Electric 1.5 T Signa Excite MRI scanner with an 8-channel head coil. Coronal 3D T1-weighted images were acquired in spoiled gradient echo recall sequences (3D-SPGR) with the following parameters: repetition time = 21.0 ms, echo time = 6 ms, field of view = 18 cm, flip angle = 30°, acquisition matrix = 256 × 256, 128 slices, and voxel size = 0.7 × 0.7 × 1.8 mm³. Scans were conducted with identical protocols and on the same scanner at two timepoints (mean interval between scans 5.7 ± 0.7 years) to provide baseline and follow-up measurements.

Structural T1-weighted images were processed using FreeSurfer’s cortical surface reconstruction methods [29‒32]. Cortical thickness was calculated as the shortest distance between brain matter and the pial surface at each vertex, using Desikan-Killiany parcellation in FreeSurfer. All reconstructions were visually inspected and manually corrected as needed (for details, see [12, 16]). Given previous findings, the regions of interest (ROIs) were cortical thickness of the middle temporal cortex and the inferior frontal gyrus (consisting of pars opercularis, pars triangularis, and pars orbitalis), as well as the volume of the lateral ventricles.

CSF samples were collected at baseline by lumbar puncture at the L3/L4 or the L4/L5 interspace between 9 and 10 am after overnight fasting. A total of 12 mL of CSF was collected, gently inverted to prevent gradient effects, aliquoted and stored at −80°C pending analysis. Samples were thawed and refrozen once before analysis. All bipolar disorder participants were euthymic at the time of collection, as previously defined.

Blood samples were collected between 8 and 9 am after overnight fasting, were allowed to clot at room temperature for 30–60 min, and were then centrifuged for 10 minutes at 1700 × g. The supernatant was kept at temperatures below 5°C until transported to the biobank within 4 hours and stored long-term at −70°C.

Based on previous research, the biomarkers of interest in this study were IL-6, IL-8, and YKL-40. Protein concentrations in blood serum and CSF were measured using a proximity extension assay from the Proseek Multiplex CVD-I panel (Olink Bioscience, Uppsala, Sweden) [33], as described in previously published research [34, 35]. For complementary analysis, a secondary validation assay of the CSF data was used [24, 36]. Additionally, we conducted an analysis including blood serum C-reactive protein (CRP), which was analyzed using a high-sensitivity immunoturbidimetric assay (Siemens Healthcare Diagnostics Inc. and Beckman Coulter Inc.) at Unilabs, Stockholm, Sweden [36].

For each brain ROI, we calculated the percentage change between baseline and follow-up using the formula: (brain measure at TP2 − brain measure at TP1)/brain measure at TP1. Negative values indicate a decrease, while positive values indicate an increase in cortical thickness or subcortical volume over time, see [16] for details.

We first compared the bipolar disorder and healthy control groups regarding the percentage change in brain ROIs and CSF biomarkers concentrations using a t test. Linear regression models were used to examine the effect of CSF biomarkers at baseline on longitudinal brain change in the whole sample, controlling for group status, age at baseline, and sex. Separate regression models were then conducted for each group, with baseline CSF biomarker concentrations as the independent variable, and age at baseline and sex as covariates, using longitudinal percentage change in brain ROIs as the dependent variable. The same analyses were repeated for blood serum biomarkers.

To validate associations, we repeated the analyses using the secondary CSF biomarker assay (see online suppl. Tables S5–10; for all online suppl. material, see https://doi.org/10.1159/000542888). To control for non-CNS inflammation, a partial correlation was conducted between brain ROI percentage change and blood serum CRP concentrations, adjusting for age and sex as covariates. In a secondary analysis, we conducted a regression between the biomarkers and brain change adding medication as a covariate within the bipolar group. Analyses for both the bipolar disorder and control groups were independently corrected for multiple comparisons using the false discovery rate (FDR). Only results with FDR-adjusted p values <0.05 were considered statistically significant [37].

This longitudinal case-control study included 63 participants: 29 with bipolar disorder (type 1 or 2) and 34 healthy controls. GAF function scores at both baseline and follow-up were significantly lower in the bipolar disorder group compared with controls. MADRS scores were significantly higher in the bipolar disorder group at both timepoints, with both groups’ mean values below the clinical cut-off of 7. There were no statistically significant differences between groups in sex, age, BMI, or YMRS scores (see online suppl. Table S1 for demographic and clinical details).

In this sample, there were no statistically significant differences in baseline concentrations of the selected biomarkers in either CSF or blood serum between cases and controls (see online suppl. Table S2).

The ventricular volume increased in both cases and controls, with a significantly higher percentage change in bipolar disorder (t_(34.95) = 2.429, p = 0.02 for left and t_(37.432) = 2.781, p = 0.008 for right lateral ventricle). For participants with bipolar disorder, the left lateral ventricle had a mean change of 0.209 (SD = 0.253) and the right lateral had a mean change of 0.225 (SD = 0.233). In healthy controls, the left ventricle had a mean change of 0.088 (SD = 0.097) and the right lateral ventricle had a mean change of 0.095 (SD = 0.104). There were no significant differences between cases and controls in the percentage change of the middle temporal cortex and the inferior frontal gyrus (see online suppl. Table S3).

The aim of this study was to predict cortical thickness change by the selected inflammation biomarkers. We first conducted analyses with case-control status as an interaction term to evaluate if such predictions would differ between cases and controls. Significant interactions were found for all CSF biomarkers with the percentage change of ventricular volumes (see online suppl. Table S4). There was no interaction effects related to associations with other brain ROIs. Based on these interactions, we ran separate regression models for cases and controls.

In the bipolar group, baseline CSF IL-8 concentrations predicted a decline of cortical thickness in the right middle temporal cortex (β = −0.56, p = 0.003, FDR-q = 0.009, CI 95% = −0.026 to −0.11), the left middle temporal cortex (β = −0.48, p = 0.011, FDR-q = 0.022, CI 95% = −0.026 to −0.004) and the right inferior frontal gyrus (β = −0.58, p = 0.002, FDR-q = 0.009, CI 95% = −0.052 to −0.014). These results are shown in Figure 1. No other biomarker was significantly associated with changes in brain metrics.

In healthy controls, baseline CSF concentration of YKL-40 predicted ventricular enlargement in both hemispheres at follow-up: right lateral ventricle (β = 0.67, p = 0.007, FDR-q = 0.021, CI 95% = 0.018–0.108) and left lateral ventricle (β = 0.62, p = 0.006, FDR-q = 0.021, CI 95% = 0.017–0.092). Figure 2 shows the association of YKL-40 with the lateral ventricles. No other associations were significant (see online suppl. Table S5, S6).

All reported results include both p values and the FDR-corrected q values. The results were replicated using an independent CSF assay as described above (see online suppl. for details). In this validation CSF assay, the association between IL-8 and cortical decline in the inferior frontal gyrus was significant for both hemispheres (see online suppl. Table S9). No significant associations were found between baseline biomarker concentrations in blood serum and brain change over time (see online suppl. Tables S5–10 for details).

To verify that our findings were not related to general inflammation, we correlated blood serum CRP concentrations with brain metrics and found no significant associations. Furthermore, a secondary analysis of potential medication effects (lithium, antipsychotic, antiepileptic medication intake) within the bipolar group showed no impact on the association between IL-8 and cortical decline (data not shown).

This study investigated the impact of selected CSF inflammation biomarkers on long-term brain structural decline in individuals with bipolar disorder and controls. The main findings were: (1) baseline CSF IL-8 concentration predicted cortical thickness decline in individuals with bipolar disorder, and (2) baseline CSF YKL-40 concentration was associated with greater ventricular enlargement in healthy controls.

We found that higher CSF concentrations of IL-8 predicted a decrease of cortical thickness in the right inferior frontal gyrus and both the left and right middle temporal cortex in individuals with bipolar disorder, consistent with our hypotheses. Previous studies have shown that individuals with bipolar disorder experience a greater decline in cortical thickness over time in these brain regions [12, 14]. The inferior frontal gyrus plays a key role in executive functions such as cue detection for initiation of cognitive tasks and inhibitory control [38]. In bipolar disorder, this region has been associated with reduced functional connectivity to brain areas involved in emotion regulation [39]. The middle temporal cortex is involved in memory processing and retrieval, as well as semantic and perceptual processing [40, 41]. In bipolar disorder, lower functional connectivity within the middle temporal cortex has been shown [42].

Biomarker studies have previously linked higher levels of CSF IL-8 to bipolar disorder [24] although findings are conflicting [34]. As a pro-inflammatory cytokine, the association of IL-8 with cortical decline may suggest increased immunological activation in individuals experiencing more significant brain decline over time [24]. Some studies propose that an interaction of immunological dysregulation, accelerated brain decline, impaired self-regulation, and increased frequency of mood episodes could contribute to disease progression in bipolar disorder [4, 15]. The results of this study could be viewed within this framework, with baseline CSF IL-8 concentration potentially predicting long-term brain deterioration. This interpretation is further supported by the fact that all measurements were collected during a euthymic state, making it unlikely that the association is due to merely temporarily elevated IL-8 concentrations caused by acute mood episodes.

IL-6 is the most replicated inflammation biomarker associated with bipolar disorder, according to a meta-analysis of cross-sectional studies on peripheral biomarkers in blood serum [20]. It has been associated with abnormally thin cortical regions, such as the right middle temporal cortex and superior frontal gyrus, suggesting a role in cortical decline [15, 22]. However, we found no association between longitudinal brain changes and IL-6 concentrations in either CSF or blood serum. A previous study in our cohorts found no significant differences in CSF IL-6 concentrations between cases and controls [24]. Furthermore, most studies associating IL-6 with cortical metrics have relied on blood serum biomarkers, potentially indicating different processes involved in peripheral versus CNS inflammation and the impact on brain structure in bipolar disorder [15].

The lack of significant associations between biomarkers and ventricular volume changes in bipolar disorder is intriguing, given that this group displayed significantly greater increases of ventricular volumes compared with controls in this study and in previous studies of ours [12]. This suggests that other pathological factors, such as cardiometabolic conditions or oxidative stress, may contribute to the observed ventricular enlargement in bipolar disorder [4, 43, 44]. Furthermore, studies on other psychiatric and neurodegenerative disorders indicate that abnormal ventricular enlargements may be a common feature across various conditions [45, 46].

Nevertheless, we found that greater ventricular volume changes in healthy controls were associated with baseline CSF concentrations of YKL-40. This observation, indicating that markers of CNS inflammation relate to greater volumetric changes even in a nonpsychiatric population, is not surprising. CNS inflammation is strongly implicated in conditions like neurodegenerative disorders and multiple sclerosis, suggesting that inflammatory processes in the brain can be generally detrimental to gray matter integrity [47, 48]. Our control cohort included only individuals with no history of psychiatric, neurological, or somatic illnesses. Therefore, the observed effect may suggest that even within a presumably healthy range, elevated CSF YKL-40 concentration could partially predict the extent of ventricular deterioration, indicating a potential susceptibility to faster brain aging in these individuals.

We observed no effect of YKL-40 on longitudinal brain changes in individuals with bipolar disorder. This is intriguing, given previous studies showing that CSF concentrations of YKL-40 are associated with the frequency of (hypo)manic episodes, which in turn have been linked brain decline in the frontal cortex in bipolar disorder [14, 23]. We expected CSF YKL-40 to predict cortical decline in the inferior frontal gyrus, but this was not the case. We speculate that the relationship between YKL-40, (hypo)mania, and brain change may be more complex or indirect. For instance, some other factor might influence both YKL-40 and brain changes, with an indirect link to (hypo)mania, rather than YKL-40 directly varying the intensity or frequency of (hypo)manic episodes. Alternatively, larger sample size might be needed to demonstrate an association between YKL-40 and brain change.

Our findings were evident only in CSF-derived inflammatory biomarkers, not in blood serum. The lack of associations with blood serum biomarkers suggests that peripheral inflammation is unlikely to influence the observed brain changes and reduces the likelihood of blood contamination during lumbar puncture affecting CSF biomarker concentrations. The notion that CNS rather than peripheral inflammatory processes contribute to brain decline is further supported by the lack of associations between peripheral CRP levels and brain changes. Our results highlight the distinction between peripheral and CNS inflammatory processes in bipolar disorder, where only the latter potentially play a role in brain decline [12].

The longitudinal study design, involving two timepoints of MRI brain scans and one timepoint of CSF sampling, is a strength of this study though it also limits the sample size. Future replication in independent cohorts is necessary to assess the generalizability of our findings. While the present study focused on three inflammation biomarkers, we acknowledge that other proteins could also be promising candidates for similar analyses in future studies. Of note, previously reported case-control differences in CSF concentrations of IL-8 and YKL-40 in this cohort could not be replicated in a second, independent cohort in an earlier study of our group [34]. In the present study, despite no group differences in CSF biomarker concentrations, we observed associations between CSF biomarkers and brain change based on case-control status.

The naturalistic design of the study is both a strength and a limitation, as all participants with bipolar disorder were on medication. Although our results were controlled for medication status, the impact of brain decline in an unmedicated population remains unknown. Additionally, we pooled bipolar disorder types 1 and 2, preventing analysis of potential differences based by subtype. Future research could explore whether mania is more strongly associated with inflammation biomarker concentrations than hypomania.

Another area for future investigation is the inclusion of longitudinal CSF data alongside MRI data over multiple timepoints, which would allow for a better understanding of associations between brain metrics and biomarker concentrations over time. Studies should also investigate whether CSF derived IL-8 and other inflammation biomarkers predict longitudinal brain changes in brain regions other than those investigated in this study. In general, it is important to note that biomarkers distribute evenly in CSF, and we have no knowledge about regional specificity of CSF biomarker secretion.

Higher concentrations of IL-8 in the CSF, but not in blood serum, predict cortical thickness decreases in the left and right middle temporal cortex and the right inferior frontal gyrus in bipolar disorder over time. These findings suggest that pro-inflammatory molecules may contribute to long-term changes in brain structure in bipolar disorder. Additionally, the association between CSF-derived YKL-40 and ventricular enlargement in healthy controls may suggest that inflammatory proteins play a role in normal brain aging.

Our findings emphasize the value of utilizing CSF biomarkers when investigating psychiatric disorders as they reflect brain chemistry more closely than peripheral biomarkers, which are influenced by non-CNS processes and confounders. This study enhances our understanding of neurobiological processes underlying bipolar disorders by highlighting the role of inflammation biomarkers in relation to longitudinal structural brain changes. Our findings suggest the potential inclusion of CSF inflammation biomarkers into future models of individualized diagnosis and treatment and highlight the need for future studies on anti-inflammatory interventions in bipolar disorder.

We are very grateful to all our study participants and the staff at the St. Göran bipolar affective disorder unit. We especially acknowledge our research nurses Lena Lundberg, Stina Stadler, Agneta Carlswärd-Kjellin, and Martina Wennberg, as well as data manager Mathias Kardell. We thank Rouslan Sitnikov for help with the MRI protocol. Yngve Hallström is acknowledged for performing lumbar punctures.

The study was conducted in concordance with the Declaration of Helsinki and was approved by the Regional Ethics Committee (Regionala etikprövningsnämnden) in Stockholm, Sweden (Approval No. 2012/559-32). Written informed consent was obtained from all study participants.

M.L. declares that he has received honoraria from Lundbeck pharmaceuticals outside the present work. All other authors declare no conflict of interest.

This research was supported by grants from the Swedish Research Council (Vetenskapsrådet, 2022-01643), the Swedish Brain foundation (Hjärnfonden, FO2022-0217), the Swedish Foundation for Strategic Research (KF10-0039), the Swedish Federal Government under the LUA/ALF agreement (ALFGBG-716801), and the Märta-Lundqvists’s Foundation. The funders had no role in the design, data collection, data analysis, and reporting of this study.

T.B.J.: methodology, formal analysis, writing – original draft, review and editing, and visualization; A.L.K.: conceptualization, methodology, formal analysis, writing – original draft, review and editing, visualization, and supervision; A.G.: conceptualization, methodology, and writing – review and editing; C.A.: methodology, formal analysis, investigation, and writing – review and editing; C.M.S.: investigation, writing – review and editing; M.L.: conceptualization, investigation, resources, writing – review and editing, supervision, project administration, and funding acquisition.

Tobias Bellaagh Johansson and Anna Luisa Klahn contributed equally to this work.

The data that support the findings of this study are not publicly available due to data protection regulation and decisions by the Ethical Review Board of Sweden but are available from the corresponding author A.L.K. upon reasonable request.