- Split View
-
Views
-
Cite
Cite
Guowei Wu, Lena Palaniyappan, Manqi Zhang, Jie Yang, Chang Xi, Zhening Liu, Zhimin Xue, Xuan Ouyang, Haojuan Tao, Jinqiang Zhang, Qiang Luo, Weidan Pu, Imbalance Between Prefronto-Thalamic and Sensorimotor-Thalamic Circuitries Associated with Working Memory Deficit in Schizophrenia, Schizophrenia Bulletin, Volume 48, Issue 1, January 2022, Pages 251–261, https://doi.org/10.1093/schbul/sbab086
- Share Icon Share
Abstract
Thalamocortical circuit imbalance characterized by prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity has been consistently documented at rest in schizophrenia (SCZ). However, this thalamocortical imbalance has not been studied during task engagement to date, limiting our understanding of its role in cognitive dysfunction in schizophrenia.
Both n-back working memory (WM) task-fMRI and resting-state fMRI data were collected from 172 patients with SCZ and 103 healthy control subjects (HC). A replication sample with 49 SCZ and 48 HC was independently obtained. Sixteen thalamic subdivisions were employed as seeds for the analysis.
During both task-performance and rest, SCZ showed thalamic hyperconnectivity with sensorimotor cortices, but hypoconnectivity with prefrontal-cerebellar regions relative to controls. Higher sensorimotor-thalamic connectivity and lower prefronto-thalamic connectivity both relate to poorer WM performance (lower task accuracy and longer response time) and difficulties in discriminating target from nontarget (lower d′ score) in n-back task. The prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity were anti-correlated both in SCZ and HCs; this anti-correlation was more pronounced with less cognitive demand (rest>0-back>2-back). These findings replicated well in the second sample. Finally, the hypo- and hyper-connectivity patterns during resting-state positively correlated with the hypo- and hyper-connectivity during 2-back task-state in SCZ respectively.
The thalamocortical imbalance reflected by prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity is present both at rest and during task engagement in SCZ and relates to working memory performance. The frontal reduction, sensorimotor enhancement pattern of thalamocortical imbalance is a state-invariant feature of SCZ that affects a core cognitive function.
Introduction
Schizophrenia (SCZ) is a severe and complex mental illness associated with aberrant information processing across large-scale brain networks. A thalamocortical imbalance, characterized by prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity observed in resting-state fMRI studies, has been implicated in the pathophysiology of SCZ.1–9 This abnormal thalamocortical connectivity appears consistently across the various clinical stages of SCZ, from ultra-high-risk to chronic stages.7 We have also observed this dysconnectivity pattern in healthy siblings10 and adolescents with early-onset SCZ,11 highlighting thalamocortical imbalance to be an age-unrelated and heritable feature of SCZ.3,12
The thalamus provides a central link for multiple functional and structural circuits,13 playing a vital role in the integration of neuronal signals across cortical and subcortical domains. While the critical role of the thalamocortical circuit in mental processes is undisputed, the mechanism by which prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity bring about the key features of SCZ remains hitherto unknown. An important limitation in this regard is the reliance on resting-state fMRI to uncover thalamocortical dysconnectivity in SCZ. As a result, it is unclear if thalamocortical imbalance is a fixed feature that persists even when patients are engaged cognitive activities. Furthermore, the relevance of thalamocortical dysconnectivity to symptom burden is unclear as resting-fMRI studies have shown inconsistent associations. While some report increase in psychotic symptoms with thalamocortical dysfunction,3,4,8,14–16 others report null findings,2,5–7,17 and even inverse relationship in one study.18
Several prior studies highlight the potential role of thalamocortical circuit for higher-order cognitive functions such as memory and attention. Diffusion tensor imaging (DTI) and resting-state fMRI studies in SCZ reported lower cognitive performance (outside the scanner) in patients who had reduced prefronto-thalamic structural/resting connectivity.15,17,19,20 Furthermore, working memory (WM) impairment in preclinical models of SCZ relates to disruptions in prefronto-thalamic neural pathways.21 Neuroimaging studies on cognitive remediation in SCZ highlight the role of both thalamus and prefrontal regions in cognition-enhancing therapeutic effects.22 Recently, Subramaniam and colleagues found that integrity of long-range white matter tracts connecting prefronto-thalamic-sensorimotor areas predicted the improvement in executive functions after 16-weeks of cognitive training in patients with SCZ,23 highlighting the role of thalamocortical structural connectivity in the pathophysiology of cognitive deficits in SCZ. These observations raise the possibility that the prefrontal-thalamic sensorimotor dysconnectivity observed at rest is likely to continue or be exaggerated when cognitive demands are placed, in association with the cognitive deficits of SCZ.
Brain networks at rest have been regarded as being in a preparatory state for task-based brain activity,24,25 evidenced by the significant association between resting-state functional connectivity (FC) and task-evoked FC.26,27 In SCZ, task-demands worsen some aspects of the resting-state dysconnectivity, while “correcting” other aspects, indicating a shift in the cooperative/competitive dynamics among task-relevant subsystems.28 On this basis, we can expect the thalamocortical imbalance to be sensitive to the degree of task demands placed on an individual with SCZ. If task engagement reduces the observed abnormality, it can provide a mechanistic basis of how repeated task-engagement during cognitive remediation programs can reverse a key pathophysiological feature.
We studied two independently recruited patient samples—discovery and replication samples with 3 aims: (1) investigate if prefronto-thalamic and sensorimotor-thalamic functional dysconnectivity in SCZ is present during WM performance; (2) examine the role of thalamocortical imbalance during WM task with that seen at rest in SCZ, (3) assess the relationship between the thalamocortical circuitry and cognitive performance and clinical symptoms in SCZ. We repeated the experiment in an independently collected rest/task dataset to assess the robustness of our observations through external validation.
Methods
Participants
One hundred and seventy-two SCZ patients were recruited from outpatient and inpatient departments of Second Xiangya Hospital of Central South University, Changsha, Hunan province, People’s Republic of China. SCZ diagnoses were confirmed by experienced psychiatrist using the Structured Clinical Interview for DSM-IV-patient version (SCID-P)29 (table 1). The inclusion/exclusion criteria and details for medication in patients can be found in the Supplementary methods 1.1 and 1.2. Patients were assessed by the Scale for the Assessment of Positive Symptoms (SAPS)30 and Scale for the Assessment of Negative Symptoms (SANS).31 One hundred and three healthy controls (HCs) from communities in Changsha city participated in the study. The inclusion and exclusion criteria were the same as those of the patient group except that the HCs and their first-degree relatives did not meet the DSM-IV criteria of any psychiatric disorders.
Characteristics . | Schizophrenia Group (n = 172) . | Control Group (n = 103) . | Significance . | . |
---|---|---|---|---|
. | Mean ± SD . | Mean ± SD . | T/χ2 . | P . |
Age (years) | 23.68 ± 5.78 | 23.50 ± 5.13 | −0.25 | 0.800 |
Education (years) | 11.60 ± 2.59 | 13.87 ± 2.31 | 7.33 | <0.001 |
Male n (%) | 111 (64.53%) | 43(41.74%) | 13.58 | <0.001 |
Illness duration (months) | 22.55 ± 31.47 | - | ||
CPZ equivalence (mg/d) | 280.47 ± 220.93 | - | ||
SAPS | 20.45 ± 15.97 | - | ||
SANS | 35.52 ± 27.13 | - | ||
WAIS-Information | 16.35 ± 4.95 | 20.75 ± 4.76 | 7.17 | <.001 |
WAIS-Digit Symbol | 62.49 ± 15.74 | 88.91 ± 13.73 | 13.85 | <.001 |
0-back performance | ||||
Target accuracy (%) | 0.74 ± 0.30 | 0.91 ± 0.16 | 5.63 | <.001 |
Target response time (ms) | 546.03 ± 156.46 | 485.35 ± 98.75 | −3.60 | <.001 |
Nontarget accuracy (%) | 0.81 ± 0.30 | 0.94 ± 0.16 | 4.22 | <.001 |
Nontarget response time (ms) | 549.82 ± 175.32 | 515.21 ± 95.81 | −1.93 | .055 |
d′ | −0.38±1.98 | 0.69 ± 1.14 | −5.17 | <.001 |
2-back performance | ||||
Target accuracy(%) | 0.50 ± 0.26 | 0.72 ± 0.20 | 7.57 | <.001 |
Target response time (ms) | 684.16 ± 196.69 | 649.53 ± 148.81 | −1.43 | .155 |
Nontarget accuracy (%) | 0.75 ± 0.26 | 0.86 ± 0.17 | 3.85 | <.001 |
Nontarget response time (ms) | 611.24 ± 203.32 | 652.08 ± 140.43 | 1.80 | .074 |
d′ | −0.53±1.62 | 0.81 ± 1.16 | −7.26 | <.001 |
Characteristics . | Schizophrenia Group (n = 172) . | Control Group (n = 103) . | Significance . | . |
---|---|---|---|---|
. | Mean ± SD . | Mean ± SD . | T/χ2 . | P . |
Age (years) | 23.68 ± 5.78 | 23.50 ± 5.13 | −0.25 | 0.800 |
Education (years) | 11.60 ± 2.59 | 13.87 ± 2.31 | 7.33 | <0.001 |
Male n (%) | 111 (64.53%) | 43(41.74%) | 13.58 | <0.001 |
Illness duration (months) | 22.55 ± 31.47 | - | ||
CPZ equivalence (mg/d) | 280.47 ± 220.93 | - | ||
SAPS | 20.45 ± 15.97 | - | ||
SANS | 35.52 ± 27.13 | - | ||
WAIS-Information | 16.35 ± 4.95 | 20.75 ± 4.76 | 7.17 | <.001 |
WAIS-Digit Symbol | 62.49 ± 15.74 | 88.91 ± 13.73 | 13.85 | <.001 |
0-back performance | ||||
Target accuracy (%) | 0.74 ± 0.30 | 0.91 ± 0.16 | 5.63 | <.001 |
Target response time (ms) | 546.03 ± 156.46 | 485.35 ± 98.75 | −3.60 | <.001 |
Nontarget accuracy (%) | 0.81 ± 0.30 | 0.94 ± 0.16 | 4.22 | <.001 |
Nontarget response time (ms) | 549.82 ± 175.32 | 515.21 ± 95.81 | −1.93 | .055 |
d′ | −0.38±1.98 | 0.69 ± 1.14 | −5.17 | <.001 |
2-back performance | ||||
Target accuracy(%) | 0.50 ± 0.26 | 0.72 ± 0.20 | 7.57 | <.001 |
Target response time (ms) | 684.16 ± 196.69 | 649.53 ± 148.81 | −1.43 | .155 |
Nontarget accuracy (%) | 0.75 ± 0.26 | 0.86 ± 0.17 | 3.85 | <.001 |
Nontarget response time (ms) | 611.24 ± 203.32 | 652.08 ± 140.43 | 1.80 | .074 |
d′ | −0.53±1.62 | 0.81 ± 1.16 | −7.26 | <.001 |
Note: CPZ, chlorpromazine; n, number; SANS, Scale for Assessment of Negative Symptoms; SAPS, Scale for Assessment of Positive Symptoms; SD, standard deviation.
Characteristics . | Schizophrenia Group (n = 172) . | Control Group (n = 103) . | Significance . | . |
---|---|---|---|---|
. | Mean ± SD . | Mean ± SD . | T/χ2 . | P . |
Age (years) | 23.68 ± 5.78 | 23.50 ± 5.13 | −0.25 | 0.800 |
Education (years) | 11.60 ± 2.59 | 13.87 ± 2.31 | 7.33 | <0.001 |
Male n (%) | 111 (64.53%) | 43(41.74%) | 13.58 | <0.001 |
Illness duration (months) | 22.55 ± 31.47 | - | ||
CPZ equivalence (mg/d) | 280.47 ± 220.93 | - | ||
SAPS | 20.45 ± 15.97 | - | ||
SANS | 35.52 ± 27.13 | - | ||
WAIS-Information | 16.35 ± 4.95 | 20.75 ± 4.76 | 7.17 | <.001 |
WAIS-Digit Symbol | 62.49 ± 15.74 | 88.91 ± 13.73 | 13.85 | <.001 |
0-back performance | ||||
Target accuracy (%) | 0.74 ± 0.30 | 0.91 ± 0.16 | 5.63 | <.001 |
Target response time (ms) | 546.03 ± 156.46 | 485.35 ± 98.75 | −3.60 | <.001 |
Nontarget accuracy (%) | 0.81 ± 0.30 | 0.94 ± 0.16 | 4.22 | <.001 |
Nontarget response time (ms) | 549.82 ± 175.32 | 515.21 ± 95.81 | −1.93 | .055 |
d′ | −0.38±1.98 | 0.69 ± 1.14 | −5.17 | <.001 |
2-back performance | ||||
Target accuracy(%) | 0.50 ± 0.26 | 0.72 ± 0.20 | 7.57 | <.001 |
Target response time (ms) | 684.16 ± 196.69 | 649.53 ± 148.81 | −1.43 | .155 |
Nontarget accuracy (%) | 0.75 ± 0.26 | 0.86 ± 0.17 | 3.85 | <.001 |
Nontarget response time (ms) | 611.24 ± 203.32 | 652.08 ± 140.43 | 1.80 | .074 |
d′ | −0.53±1.62 | 0.81 ± 1.16 | −7.26 | <.001 |
Characteristics . | Schizophrenia Group (n = 172) . | Control Group (n = 103) . | Significance . | . |
---|---|---|---|---|
. | Mean ± SD . | Mean ± SD . | T/χ2 . | P . |
Age (years) | 23.68 ± 5.78 | 23.50 ± 5.13 | −0.25 | 0.800 |
Education (years) | 11.60 ± 2.59 | 13.87 ± 2.31 | 7.33 | <0.001 |
Male n (%) | 111 (64.53%) | 43(41.74%) | 13.58 | <0.001 |
Illness duration (months) | 22.55 ± 31.47 | - | ||
CPZ equivalence (mg/d) | 280.47 ± 220.93 | - | ||
SAPS | 20.45 ± 15.97 | - | ||
SANS | 35.52 ± 27.13 | - | ||
WAIS-Information | 16.35 ± 4.95 | 20.75 ± 4.76 | 7.17 | <.001 |
WAIS-Digit Symbol | 62.49 ± 15.74 | 88.91 ± 13.73 | 13.85 | <.001 |
0-back performance | ||||
Target accuracy (%) | 0.74 ± 0.30 | 0.91 ± 0.16 | 5.63 | <.001 |
Target response time (ms) | 546.03 ± 156.46 | 485.35 ± 98.75 | −3.60 | <.001 |
Nontarget accuracy (%) | 0.81 ± 0.30 | 0.94 ± 0.16 | 4.22 | <.001 |
Nontarget response time (ms) | 549.82 ± 175.32 | 515.21 ± 95.81 | −1.93 | .055 |
d′ | −0.38±1.98 | 0.69 ± 1.14 | −5.17 | <.001 |
2-back performance | ||||
Target accuracy(%) | 0.50 ± 0.26 | 0.72 ± 0.20 | 7.57 | <.001 |
Target response time (ms) | 684.16 ± 196.69 | 649.53 ± 148.81 | −1.43 | .155 |
Nontarget accuracy (%) | 0.75 ± 0.26 | 0.86 ± 0.17 | 3.85 | <.001 |
Nontarget response time (ms) | 611.24 ± 203.32 | 652.08 ± 140.43 | 1.80 | .074 |
d′ | −0.53±1.62 | 0.81 ± 1.16 | −7.26 | <.001 |
Note: CPZ, chlorpromazine; n, number; SANS, Scale for Assessment of Negative Symptoms; SAPS, Scale for Assessment of Positive Symptoms; SD, standard deviation.
Additionally, an independent replication sample, including 49 SCZ patients and 48 HCs (Supplementary table S1) who met the identical inclusion/exclusion criteria were recruited at the same hospital and Changsha communities, respectively. No significant differences as for the socio-demographical, clinical, and cognitive characteristics between the discovery and replication patient groups were found except for the age, education, SAPS score (Supplementary table S2). All participants gave written informed consent in the study approved by the ethics committee of the Second Xiangya Hospital of Central South University.
Cognitive Function Assessment
All participants completed the 0-back and 2-back WM task during their MRI scan (Supplementary methods 1.3). Moreover, all participants were evaluated with Information and Digital symbol subscales of Wechsler Adult Intelligence Scale Chinese Revised (WAIS-CR). Particularly for the n-back WM task, besides the traditional measurement including task accuracy and response time, an additional index, d prime (d′), was calculated for n-back tasks.32d′ is calculated using the formula d′ = ZHit – ZFA, where Hit represents the proportion of hits when a signal is present (hits/(hits + misses)), also known as the hit rate, and FA represents the proportion of false alarms when a signal is absent (false alarms/(false alarms + correct negative)), the false-alarm rate. Z represents a transformation of the two distributions allowing for comparison of measures with different ranges of absolute values. The better the participant maximizes hits and minimizes false alarms (equals to maximizing correct rejections), the better the participant is able to discriminate target from nontarget when performing a task. It is also demonstrated that d′ has the advantage of capturing executive skills needed to perform cognitive tasks in patients with SCZ, without being influenced by demographic variables or IQ.32
Experimental Design, fMRI Data Acquisition, and Preprocessing
Details of participants, task design, imaging acquisition protocols, and preprocessing steps for 2 datasets are provided in the Supplementary methods 1.3 and 1.4, Supplementary table S3, and also described elsewhere.33–35 Specifically, in the study stimulation blocks and resting intervals alternated within the experiment with a total of four 2-back and four 0-back blocks. Each stimulation block and resting interval last for 40 s and 20 s, respectively. After ruling out participants with overt head movements during acquisition (translation ≥ 3mm, rotation ≥ 3°), the n-back task data of 190 participants (112 SCZ) and resting-state data of 241 participants (146 SCZ) in the discovery sample, and n-back task data of 72 participants (33 SCZ) and resting-state data of 90 participants (44 SCZ) in the replication sample were selected for the following analysis in the present study.
fMRI Data Analysis
According to the experimental paradigm of the study and the prior researches, we separated the 0-back condition, 2-back condition, and rest interval by choosing their one-to-one time points.25,36 We extracted 80 scans of four 2-back blocks, 80 scans of four 0-back blocks, and 240 scans of resting-state (from an independent resting-state scanning, not included in the n-back experiment, for more details please see the Supplementary methods 1.3), respectively, from each subject. Next, the FC analysis was applied separately to the scans collected during either the 2-back task, the 0-back task, or the resting-state by a seed-based approach.37 We examined the FC between each of the seed regions, namely the 16 thalamus subdivisions, and the rest of the brain in a voxel-wise manner.38 Details about the 16 thalamic subdivisions extracted from The Human Brain Atlases39 can be found in the Supplementary methods 1.5 and our prior studies.10,11 Finally, we transformed the correlation coefficient in each voxel to z-value by the Fisher r-to-z transformation to improve normality.
Statistical Analysis
For three conditions including resting-state, 0-back task and 2-back task, analysis of covariance (ANCOVA) was employed, respectively, to perform voxel-based comparison of z-value maps between HC and SCZ groups by including gender, education, and head-movements as covariates with a significance threshold set at P < .05 after the False Discovery Rate (FDR) correction at the voxel level. After identifying brain regions with altered FC showing significant differences in the patients vs. HCs contrasts, the coefficients were extracted from these regions. Then all the links with hyperconnectivity or hypoconnectivity were averaged, respectively, to obtain the mean value of the regions with hyperconnectivity or hypoconnectivity for each group at each condition (Supplementary table S4). The Spearman’s correlation analysis examined the associations of the mean hyperconnectivity or hypoconnectivity of each group at each condition with cognitive measures and clinical symptoms, respectively (PFDR < .05). Group differences on demographic, mean value of FC, and clinical variables were analyzed with t-test or chi-square test where applicable using the SPSS (SPSS, Chicago, IL, USA).
Results
The demographic, clinical characteristics and task performance are shown in table 1. The sex (χ2 = 13.58, P < .001) and education level (T = 7.33, P < .001), as well as the head-movements measured as framewise displacement (FD) (Supplementary method 1.4 and table S4), were not fully matched and were included as covariates during between-group comparisons of thalamocortical connectivity. Relative to HCs, patients exhibited lower target and nontarget accuracy both in 0-back and 2-back tasks (Ps < .001), longer target response time in 0-back task (P < .001), and marginally significant longer nontarget response time both in 0-back (P = .055) and 2-back tasks (P = .074). Particularly, patients showed impairments in discriminating target from nontarget both in 0-back and 2-back tasks (d′SCZ < d′HC,Ps < .001). Moreover, patients also showed lower scores on digit symbol and information compared with HCs (P < .05).
Thalamocortical Functional Dysconnectivity: Discovery Sample
Compared to HCs, patients showed significantly decreased prefronto-thalamo-cerebellar connectivity and increased sensorimotor-thalamic connectivity across 0-back, 2-back, and resting-state conditions. To better illustrate the similarities of thalamocortical dyconnectivity between the task and rest conditions in patients, we performed a conjunction analysis (Supplementary methods 1.6). A notable overlap was seen in regions with increased or decreased FC between 0-back and rest conditions (figures 1 and S2), as well as those between 2-back and rest conditions (figures 2 and S3). Regions with increased FC (red) overlapped at the sensorimotor cortex, and those with decreased FC (blue) overlapped at the prefrontal, parietal, cerebellar, and striatum regions (Supplementary results 2.1.1). Detailed description on the locations of thalamic subregions and cortical regions with abnormal FC can be found in the Supplementary results 2.1.2 (figures S4–S6; tables S5–S7).
To confirm the results were not affected by antipsychotics and illness duration, we also examined the characteristics of the thalamocortical FC in 66 first-episode drug-free (Supplementary table S8–S11) and 95 medicated patients (Supplementary table S12–S15) in our discovery sample separately. Both drug-free and medicated patient groups demonstrated a similar pattern of thalamocortical disruptions during 0-back, 2-back tasks, and resting-state. In addition, we carefully matched the gender, age, and education and obtained a sub-sample consisted of 76 SCZ patients and 67 HCs. In this subsample, we repeated our analysis and found very similar results in line with the initial findings in the larger sample (Supplementary tables S16–S20; figures S7–S8).
Thalamocortical Functional Dysconnectivity: Replication Sample
The present findings were replicated in another sample of 49 SCZ patients and 48 HCs, recruited and assessed independently (Supplementary results 2.2), highlighting the robustness of the observed pattern. We observed very similar thalamocortical disturbances during 2-back task (Supplementary table S21; figure S9), and at resting-state (Supplementary table S22; figure S9) though the 0-back results were not significant in this smaller sample. A conjunction analysis revealed that the regions with increased FC (red) overlapped at the sensorimotor cortex, and those with decreased FC (blue) overlapped at the prefrontal regions between 2-back and rest conditions (figure S10).
Cognition, Symptoms, and Thalamic Hypo\Hyper-Connectivity in Schizophrenia
We performed the Spearman’s correlation analysis between mean FC of voxels with hypo/hyper-connectivity with 0-back and 2-back task performances (including task accuracy, response time, and d′), WAIS Information and Digit Symbol scores, as well as SANS and SAPS total scores. In all participants, the mean hyper-connectivity was negatively correlated with target/nontarget accuracy at 0-back and 2-back conditions, while the mean hypo-connectivity was positively correlated with target accuracy in the 2-back condition. Further, the mean hyper-connectivity was positively correlated with target response time in the 0-back condition (figure 3A). In patients only, the mean hypo-connectivity was negatively (r = −0.30, P = .029) and hyper-connectivity was positively (r = 0.24, P = .038) correlated with nontarget response time, respectively, in the 2-back condition. Furthermore, the mean hypo-connectivity and hyper-connectivity were positively and negatively correlated with the d′ score, respectively, at both 0-back and 2-back tasks in all participants (figure 3B). These relationships were not statistically significant when the correlations were restricted to either SCZ or HC groups only. In the replication sample, the significant correlations of the mean hypo/hyper-connectivity with WM performance and the d′ score were also identified (table S23). Moreover, no significant correlations with WAIS Information and Digit Symbol subscales or clinical symptoms were found.
Relationship Between Mean Hypo- and Hyper-Connectivity
The present results reveal two distinct patterns in SCZ: sensorimotor-thalamic hyperconnectivity and prefronto-thalamic-cerebellar hypoconnectivity. To verify the proposal24 that they reflect a coordinated disturbance in this thalamocortical circuit, the correlation between the mean strength of connectivity was performed between regions with hyper- and hypo-connectivity at each state (0-back, 2-back, or resting state). As shown in the figure 4, during both task and resting-state conditions, patients and controls exhibited a negative relationship between the prefronto-thalamic connectivity and the sensorimotor-thalamic connectivity (figure 4). Using a mixed linear model, we further tested if diagnosis and load have independent as well as interacting effects on this anticorrelation, and observed significant main effects of diagnosis (anticorrelation was stronger in controls than patients, PFDR = .024, effect size: Partial eta squared = 0.009) and load (higher inverse correlation at rest, intermediate at 0-back, lower at 2-back, PFDR = .0006, effect size: Partial eta squared = 0.199).There was no statistically significant interaction between diagnosis and load (P = .13). We report the inverse correlation across the loads in both groups in Supplementary table S24. Finally, a similar but attenuated inverse correlation pattern between mean hypo- and hyper-connectivity was detected in our replication sample (Supplementary figure S11).
Relationship Between Thalamic Dysconnectivity at Rest and Task Conditions
Finally, the relationships of the mean hypoconnectivity and mean hyperconnectivity under different conditions (0-back, 2-back, and resting conditions) were examined in SCZ patients. Positive correlations between the mean hypoconnectivity under resting-state and that under 2-back task (r = 0.20, PFDR = .045), as well as between the mean hyperconnectivity under resting-state and that under 2-back task (r = 0.22, PFDR = .030) were noted. We could not replicate this pattern in the smaller replication patient sample.
Discussion
The thalamocortical dysconnectivity pattern has been identified in previous studies and extensively investigated in SCZ.40 However, patterns of thalamic dysconnectivity when patients are performing cognitive tasks remain unclear. To address this knowledge gap, we combined n-back task- and resting-fMRI in a large sample, characterized the thalamocortical dysconnectivity in association with task performance and clinical measurements in SCZ, and replicated this in a smaller independent sample, for the first time. We identified complex patterns in agreement with recent discoveries in SCZ:1–9 hyperconnectivity between thalamus and the sensorimotor cortices but hypoconnectivity between thalamus and the prefrontal cortex and cerebellum during both n-back WM task and rest conditions. Both patterns were observed in first-episode drug-free and drug-treated patients in our discovery sample, supporting that the identified hypo- and hyperconnectivity in the thalamocortical circuit may not be influenced by the use of antipsychotics and illness duration in SCZ. Furthermore, both patterns were observed robustly in our replication sample, lending further support to our findings. Both patterns were significantly correlated with n-back WM task performance but not with symptoms assessed by the SAPS and SANS: the more abnormal the connectivity, the poorer the task performance, providing the first direct evidence linking thalamocortical functional dysconnectivity with cognitive deficits in patients with SCZ. Across samples, we showed that the two patterns of dysconnectivity were functionally related with each other, which agrees with previous studies1 and again strengthens the pathophysiological model of overlapped hypo- and hyper-connectivity in the thalamocortical circuit in SCZ. Notably, the correlation strength between the hypo- and hyper-connectivity was modulated by cognitive task load: higher in the resting-state, lower in the 2-back task, and intermediate in the 0-back task. Finally, we found that the mean connectivity during 2-back task correlated with that during resting-state positively, indicating a continuum of functional dysconnectivity across cognitive states.
In line with many prior studies, thalamic hyperconnectivity was detected across motor and sensory cortices in the present study, especially the bilateral postcentral gyrus, precentral gyrus, temporal cortex, occipital cortex, and lingual gyrus which are important for auditory, visual and motor processing. Of hyperconnectivity with target/nontarget accuracy and positive correlation with target response time of n-back task indicate that a pronounced sensorimotor-thalamic hyperconnectivity relates to poorer n-back task performance. Chen and colleagues recently reported a similar relationship between WM performance outside the scanner and resting-state sensorimotor-thalamic hyperconnectivity in SCZ.15 The central cognitive role of thalamus appears to be shaping mental representations (held in WM), either by maintaining relevant mental constructs online or by updating those no longer relevant.41 Consistent with this idea, in our study the thalamocortical dysconnectivity in SCZ relates to difficulties in both mental representation of the target and discarding the irrelevant (nontarget) stimuli. Importantly, using the d′ to assess the ability of discriminating target from nontarget in n-back task, we found a significantly negative correlation with the sensorimotor-thalamic connectivity and the d′ score irrespective of diagnosis (higher the sensorimotor-thalamic connectivity, poorer the performance to discriminate target from nontarget), further lending support to the role of this circuit in the impairments of mental representation of the target and discarding the irrelevant stimuli.
We report that reduced prefronto-thalamic connectivity also relates to impoverished n-back task performance, particularly to the inability of discriminating target from nontarget during cognitive behaviors (d’ score, across diagnosis). This is consistent with previous DTI and resting-state fMRI findings relating prefronto-thalamic disconnectivity in SCZ to WM.17,19 A large body of work has consistently documented the critical roles of prefrontal and cerebellar regions in cognitive function.42–44 Our findings may specify the involvement of the reduced prefronto-thalamo-cerebellar connectivity and increased sensorimotor-thalamic connectivity in the impairments of discriminating target from nontarget in SCZ, which is of important relevance to clinical manifestation of psychotic symptoms such as delusion and hallucinations.
In our study, no significant correlation was noted between thalamocortical connectivity and clinical symptoms assessed by the SAPS and SANS, again highlighting the complexity of interpreting prior findings in this regard.2,6,7,16–18 Apart from the clinical heterogeneity of SCZ with regard to symptom profiles, illness stage, and medication use, it is also likely that the thalamocortical imbalance primarily affects cognitive function, with any share variance with symptom severity being a secondary effect. In line with thalamocortical imbalance, the cognitive dysfunction in SCZ is relatively stable over time and independent of psychotic symptoms.45 It is interestingly to note that Northoff and colleagues46 have demonstrated a close relationship between the sensorimotor-thalamic circuit and psychomotor behaviors, relating increased sensorimotor-thalamic FC with hyperactivity in mania and a decreased sensorimotor-thalamic FC with motor retardation in depression.47 However, we did not specifically assess psychomotor symptoms in SCZ in this study. Future studies exploring more fine-grained symptom measures are required to parse these complex relationships.
Prior studies have observed that the prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity were anti-correlated at rest.1,7 Our findings extend the evidence of this anti-correlation across cognitive task and resting-state in both discovery and replication samples. This concurrent hypo- and hyper-connectivity may reflect an “imbalance” between prefronto-sensorimotor information flowing through the thalamus.1 The biochemical mechanism involved in this “imbalance” of the neural circuit is unclear. While the thalamus is known as an important relay for glutamatergic and GABAergic signals between cortical and subcortical areas bidirectionally,48,49 a proposal that excitatory-inhibitory (E/I) imbalance between glutamatergic and GABAergic neurons give rise to the dysfunction of thalamocortical neurocircuits has been raised in this field.50 Additionally, it is possible to attribute to deficits of other neurotransmitter systems. Northoff et al. suggest that an increased dopaminergic and decreased serotonin signaling may result in the increased sensorimotor-thalamic FC in psychotic disorders including SCZ.47
We also found that lower the task load, higher the inverse correlation between hypoconnectivity and hyperconnectivity in both patients and HCs, suggesting that the physiological anti-correlation seen at rest in HC is exaggerated in SCZ, but weakens when task demands arise in both groups. However, as the task-dataset was smaller in size in our study than the rest-dataset (only 78 HCs and 112 patients has n-back data), this finding should be replicated in other larger samples.
We noted that the prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity during resting-state related to the same patterns during 2-back task-state in SCZ. Growing evidence suggests that the inherent connectivity measured by resting-state and task fMRI separately may have large substantial overlapping, sharing the same variance more than 80%.26 This also adds to the notion that a large portion of the resting-state functional dysconnectivity seen in SCZ represents a genetic diathesis effect, and thus continuous across task states. Further, mental manipulations occurring during task conditions have been linked to resting-state functional circuitries in healthy individuals.51 The present study extends prior evidence that disturbance of large-scale brain circuit at rest also links to that while responding to cognitive demands. While our findings highlight the continuity of aberrant FC between rest and task states, interestingly, recent studies have focused on the degree of evoked changes from pre-stimulus baseline to task (namely “reduced rest/prestimulus-task modulation”) on the cognitive impairments of SCZ.52 Such an adaptive rest-to-task modulation in the thalamocortical connectivity is a promising area of study to further delineate the role of this circuit in the pathophysiology of SCZ.
Limitations
Our study has some limitations of note. First, we cannot fully exclude the possibility that the effects of medications and duration of illness contributed to our findings. However, we observed the similar patterns of thalamocortical connectivity both in first-episode drug-free patients and medicated ones in our discovery sample. Consistent with this notion, previous studies of high-risk patients with SCZ and patients’ healthy siblings have suggested that thalamocortical dysconnectivity occurs without the use of antipsychotics.2 Second, FDR was performed for group comparisons correction at the voxel level to control for type I error in the fMRI data. Our independent sample replicated the similar pattern. However, for the secondary analysis involving 16 thalamic subdivisions, we cannot exclude the possibility of a type I error. We urge caution when interpreting the sub-regional findings. Third, the correlations of the thalamocortical connectivity with the task accuracy and the d′ score were found in all the participants, but not in the independent patient sample, possibly due to relatively smaller sample, which needs to be replicated in a larger patient cohort in the future. Finally, although global mean signal regression (GSR) is controversial,53,54 in line with prior studies examining thalamocortical dysconnectivity in SCZ,2,24,25 we performed the GSR on our data to ensure comparability. However, after repeating the data analysis without GSR, we did not find the typical pattern of decreased prefrontal-thalamo-cerebellar connectivity and increased sensorimotor-thalamic connectivity in our data. Given its extensive connections with cortical and sub-cortical areas and its role in vigilance and arousal, thalamic connectivity may be highly sensitive to the global signal effect; we suggest a cautious interpretation of our findings in the context of global signal regression.
Conclusion
In summary, the present study documents that the combination of prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity is present at both the task- and resting-state in SCZ patients. Both prefronto-thalamic hypoconnectivity and sensorimotor-thalamic hyperconnectivity are related to WM deficits in SCZ. Moreover, these two features are anticorrelated, suggesting an “imbalance” between prefronto-thalamic circuitry and sensorimotor-thalamic circuitry in SCZ from a circuit perspective. The imbalance dissipates to some extent when facing higher task load requiring more cognitive demands, but the extant dysconnectivity impairs WM performance in patients.
Funding
This work was supported by the China Precision Medicine Initiative (grant number 2016YFC 0906300 to Dr Liu), the National Natural Science Foundation of China (grant number 81701325 to Dr Wu, 81873909 to Dr Luo, 82071506 to Dr Liu, 81671335 to Dr Xue, and 81801353 to Dr Ouyang), Natural Science Foundation of Shanghai (grant number 20ZR1404900 to Dr Luo), Hunan Key Research and Development Program (grant number 2020SK2090 to Dr Zhang) and Shanghai Municipal Science and Technology Major Project (grant number 2018SHZDZX01 to Dr Luo). Dr. Palaniyappan acknowledges salary support from the Tanna Schulich Chair of Neuroscience and Mental Health.
Conflict of Interest Statement
Dr. Palaniyappan reports personal fees from Otsuka Canada, SPMM Course Limited, UK, Canadian Psychiatric Association; book royalties from Oxford University Press; investigator-initiated educational grants from Janssen Canada, Sunovion and Otsuka Canada outside the submitted work. All other authors report no financial relationships with commercial interests. No other authors reported conflicts of interest.