Bipolar disorder (BD) and major depressive disorder (MDD) belong to different diagnostic categories; however, it is very challenging to clinically identify BD and MDD with only the onset of depression. Although there are some clinical features that could help doctors identify BD early, such as family history (Craddock and Jones, 1999; Taylor et al., 2002; McGuffin et al., 2003), atypical features (Benazzi, 2000), age at onset (Woo et al., 2015), comorbidity with attention deficit and hyperactivity disorder (Torres et al., 2015; Hunt et al., 2016; Usami, 2016), this clinical information cannot be used to reliably and specifically predict the diagnosis. It is still not sufficient for clinicians to quickly diagnose BD with depression. Early research has shown that BD is often misdiagnosed, more than 1/3 of the patients with BD were first diagnosed with depression, and the correct diagnosis took a long time (Hirschfeld et al., 2003). Several studies show that the average annual return rate of MDD diagnosis to BD was approximately 1–2% (Angst et al., 2005; Dudek et al., 2013; James et al., 2015). Among them, a 26-year follow-up study showed that the total diagnostic outcome was 39% (Angst et al., 2005). Misdiagnosis can lead to inappropriate treatment for BD patients, delay the best time to treat, increase the incidence of mania (Hirschfeld et al., 2003) and suicidal behavior, influence the prognosis of disease (Maj et al., 2002; MacKinnon et al., 2005), and increase the economic burden (Peele et al., 2003). Therefore, the early identification of BD with depression and MDD is essential.

Functional magnetic resonance imaging (fMRI) is a promising imaging method to study psychosis both in structure and function. Resting state functional connectivity (rsFC) is a frequently used method which is indicated by the temporal correlations in spatially separated brain regions (Lowe et al., 2000). It may be clinically useful to the identification of BD and MDD in depressive episode (De Almeida and Phillips, 2013; Rive et al., 2016).

Previous MRI studies indicate that BD and MDD show several disrupted brain regions, and the insula could be a key region associated with depression. A recent study on cortical structure from 20 cohorts worldwide, which contained more than two thousand MDD patients, suggest that adults with MDD had thinner cortical gray matter in the insula than controls (Schmaal et al., 2017). Insular regional homogeneity is decreased in MDD and treatment-resistant depression (Yao et al., 2009; Guo et al., 2011). There is also increased insular amplitude of low frequency fluctuation in treatment-free MDD patients (Liu et al., 2014). Cognitive task research displays that differential activation of the anterior/middle insula was significantly reduced in adolescents with MDD facing expressions of sadness and happiness compared to controls and correlated negatively with depression severity (Henje Blom et al., 2015). Meta-analyses and reviews also reported altered insular volumes in BD patient (Ellison-Wright and Bullmore, 2010; Dell’Osso et al., 2014), and found the anterior insular volume was inversely correlated with the lifetime number of depressive episodes (Takahashi et al., 2010). Depressive episodes of BD have been associated with increased serotonin transporter binding in the insula using positron emission tomography (Cannon et al., 2007). Studies also found the abnormal activation of the insular cortex in unmedicated BD patients and patients with bipolar depression (Liu et al., 2012; Yip et al., 2014). Reviews and meta-analyses for the first-degree relatives of BD patient report that the genetic liability for BD could be related to gray matter abnormalities in insular cortex (Nery et al., 2013; Lee et al., 2014).

The insula is a functionally and cytoarchitectonically diverse region of cortex located deep inside the lateral sulcus of the Sylvian fissure (Ture et al., 1999). The insula connects to several cortical regions with different neural circuits and is involved in multiple functions, including emotion, cognition, sensory perception, and somatosensation (Augustine, 1996; Craig, 2002, 2009; Nieuwenhuys, 2012). Based on the previous structural and functional studies, the insula is considered to have three different subregions, the ventral anterior insular cortex (v-AIN), dorsal anterior insular cortex (d-AIN), and posterior insula (PI) (Mesulam and Mufson, 1982a,b; Deen et al., 2011; Jakab et al., 2012). The AIN affects emotional and cognitive function more, while the PI is more associated with motor-related functions and sensory perception (Mutschler et al., 2009; Kurth et al., 2010; Kelly et al., 2012). Insular subregions exhibit different functional patterns to distinguish diseases.

In the current study, we used a seed-based fMRI method to examine the shared and distinct rsFC between participants with MDD and BD in depressive episodes (BDD). BDD patients had never had mania or hypomania episodes when they were scanned. Few studies have used insular subregion rsFC to study the difference between BDD and MDD, and the sample of BDD patients cannot exclude the effect of mania or hypomania on the brain. The v-AIN, d-AIN and PI were selected as the regions of interest (ROIs), and we explored rsFC between these seed regions to other brain regions. We hypothesized that both the MDD and BDD groups would demonstrate altered rsFC when compared to the control group; furthermore, we hypothesized that there would also be distinct changes in insular rsFC in MDD and BDD.

In this study, we compared the insular subregions rsFC with whole brain among BDD, MDD, and HC. The choice of BDD patients greatly eliminated the influence of mania or hypomania on brain function. We found that significant changes mainly exist with the frontal and parietal lobes. The PI and d-AIN aberrant rsFC were observed mainly in the parietal lobe, but the v-AIN aberrant rsFC involved a wider range of regions included both the prefrontal lobe, parietal lobes and a small part of the temporal lobe. The common change of the BDD and MDD groups was the decreased rsFC of v-AIN with the left SFG and MFG. We did not find specific changes in the MDD group compared to the other two groups. For the BDD group, specific decreased rsFC included the primary somatosensory cortex with PI and d-AIN, right postcentral gyrus with PI, motor cortex with v-AIN and left orbital frontal gyrus and left superior temporal area with v-AIN.

The current study found a decreased rsFC of the left dorsolateral SFG and MFG with v-AIN in both BDD and MDD patients. These two aberrant regions of dorsolateral prefrontal cortex (DLPFC) were consistent with previous studies. Previous studies reported that first-episode and medication-naive MDD patients had altered gray matter volume in SFG and MFG, which was negatively correlated with the duration of illness and the severity of clinical symptoms (Lai and Wu, 2014; Peng et al., 2016). Functional studies reported abnormal activation in the SFG, which was associated with rumination in MDD (Guo et al., 2013; Schiller et al., 2013). Task-related and resting state functional studies reported the abnormal activation in the MFG (Naismith et al., 2010; Liu et al., 2013), which was associated with anticipatory emotional processes and depressive symptoms such as apathy, anergia and loss of motivation in MDD (Feeser et al., 2013). Resting state functional studies also revealed abnormal regional homogeneity in the MFG on BD in depressive episodes (Liang et al., 2013; Gao et al., 2014). From these studies and our results, we considered that altered rsFC of DLPFC with v-AIN was the common pathophysiological basis of MDD and BDD. In addition, we found that there was a lower trend of decreased rsFC in BDD. One structural MRI study also found that depressed BD patients showed reduced thickness in the DLPFC compared to MDD (Lan et al., 2014). We surmised that there was more serious dysfunction in BDD.

BDD patients showed specific decreased v-AIN rsFC in left orbital frontal gyrus and left STG (included the left superior temporal pole), which were involved in fronto-limbic neural circuit. The orbital frontal cortex (OFC) participates in the process of learning, predicting and making emotional decisions, as well as the process of rewarding related behaviors (Kringelbach, 2005). Altered function in OFC might be associated with symptoms such as emotional instability and indecisiveness in BDD (Liu et al., 2012). Studies have reported that OFC volume decreases in both adults and children with BD (Nugent et al., 2006; Najt et al., 2007). Recent twin and family fMRI studies of BD also found that OFC rendered the most significant heritability estimates and was significantly correlated with the BD phenotype (Sugihara et al., 2017). These studies indicate that OFC could play an important role in BD. Decreased OFC-insular connectivity was correlated with the processing of emotional information, which leaves the patients with the inability to properly treat their emotions, leading to increased negative self-focus (Critchley et al., 2004; Guo et al., 2015). The superior temporal lobe participates in the processing of emotional experiences. Studies reported that the STG and the adjacent cerebral cortex play important roles in the processing of information related to individual communication (Narumoto et al., 2001). Functional and DTI studies reported the abnormal degree centrality and mean diffusivity of the temporal pole in BD (Zhou et al., 2017; Spuhler et al., 2018). Reduced rsFC between insula and the left STG related areas might result in social withdrawal and be related to the time and duration of the episode (Guo et al., 2015). Together with these results, we considered that the decreased rsFC of vAIN-OFC and vAIN-superior temporal gyrus were specific in the pathophysiology of BDD.

Other significant regions of BDD were in the motor cortex and included the primary somatosensory cortex and supplemental motor area (SMA), which had decreased rsFC with all three insular subregions. The primary somatosensory cortex mainly participates in the processing of sensory information and face emotion (Keysers et al., 2010). Previous studies found that abnormal overactivation by BD depressed patients in response to happy faces and fearful faces (Chen et al., 2006). A study found increased rsFC between the somatosensory cortex and insular cortex in euthymic BD subjects compared to HC (Minuzzi et al., 2017). This was mutually verified by our results. The SMA participants in movement control and is involved in identifying and describing emotional feelings (Han et al., 2018). A study found different activation patterns of SMA in manic and depressed BD during reaction time tasks (Caligiuri et al., 2004). Overall, we speculated that the rsFC of the insula with the primary somatosensory cortex and SMA could be a mood state-related indicator in BD that increases in manic/hypomanic episodes and decreases in depression. The decreased rsFC could distinguish BDD from MDD.

In addition, we found that the rsFC between PI and the right lateral postcentral gyrus was correlated with anxiety in patients. However, the correlation patterns were different. Depression often has anxiety symptoms. The mechanism is still unclear. In this study, we considered that the insula regulates part of the somatosensory cortex and plays an important role. Further research is needed in the future.

The present study has several limitations. First, the sample size of BDD patients was small because the BDD patients were drawn from the initial group of MDD patients after follow-up. Second, we are not sure whether the drug will affect the results. Finally, we did not conduct cognitive tests and we did not carry out classified studies to further verify our results.

BDD and MDD patients showed similar and distinct decreased insular rsFC changes. BDD patients have specific decreased insular connectivity, especially in the somatosensory and motor cortex, which may be used as imaging evidence for clinical identification.

All proceduresperformed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the Medical Science Research Ethics Committee of the First Affiliated Hospital of China Medical University (approval reference number [2012]25-1). All participants provided written informed consent by themselves or via their parents/guardians if they were under 18 years old after a complete description of the study.

ZY designed the study, collected and analyzed the data, and drafted the manuscript. SW, XJ, YZ, and JL collected the data. MC and LC performed the MRI scan. FW and YT designed the study and revised the manuscript. All authors approved the final version to be submitted.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.