Fronto-Parietal Network Function During Cued Visual Search in the First-Episode Schizophrenia Spectrum

Abstract

Cognitive impairments account for significant morbidity in schizophrenia and are present at disease onset. Controlled processes are particularly susceptible and may contribute to pervasive selective attention deficits. The present study assessed fronto-parietal attention network (FPAN) functioning during cue presentation on a visual search task in first-episode schizophrenia spectrum patients (FE) and its relation to symptom burden and community functioning. Brain activity was recorded with magnetoencephalography from 38 FE and 38 healthy controls (HC) during blocks of pop-out and serial search target detection. Activity during cue presentation was compared between groups across bilateral FPAN regions (frontal eye fields (FEF), inferior frontal gyrus (IFG), midcingulate cortex (MCC), and intraparietal sulcus (IPS)). FE exhibited greater right hemisphere IFG activity despite worse performance relative to HC. Performance and FPAN activity were not correlated in HC. Among FE, however, stronger activity within right hemisphere FEF and IFG was associated with faster responses. Stronger right IPS and left IFG activity in patients was also associated with reduced negative symptoms and improved community functioning, respectively. Increased reliance on the FPAN for task completion suggests an inefficient cognitive control network and might reflect a compensation for impaired attentional deployment during target detection, a strategy employed by those with less severe illness. These findings represent a critical step towards identifying the neural substrates of negative symptoms and impaired neurocognition at disease onset.

Introduction

The ability to focus attention on certain percepts in our environment while ignoring others, broadly defined as selective attention, is an essential cognitive function that promotes coherent, goal-directed behaviors. Disruptions in this cognitive function have been well-documented in people with schizophrenia (Fioravanti et al 2005), are associated with negative symptoms (Cornblatt et al 1985; Sanz et al 2012) and poor functional outcomes (Miley et al 2005; Vesterager et al 2012), and remain largely impervious to current pharmacologic treatments (Millan et al 2014; Miyamoto et al 2012). Importantly, deficits of selective attention are present early in the course of the illness (Fusar-Poli et al 2012; Orellana et al 2012; Ren et al 2021), providing a meaningful treatment target during a critical period for disease intervention. However, a more mechanistic understanding of the dysfunctional neural system contributing to impairments of attention is needed to support future interventional approaches.

A frequent challenge complicating attempts to characterize disruptions in attention is the overly broad conceptualization of this cognitive construct. To address this problem, the Cognitive Neuroscience Treatment Research to Improve Cognition in Schizophrenia (CNTRICS) group has provided recommendations for paradigms with high construct validity that engage facets of the attentional system with demonstrated sensitivity to disruption in psychosis (Luck et al 2012). According to these recommendations, attentional cuing tasks emphasizing executive control of attention are particularly well suited for examining cognitive control networks with the ability to broadly modulate activity across sensory-perceptual systems. Though modulatory functions are not the exclusive source of attentional deficits in psychosis, they show greater impairment than tasks examining the implementation of attention (Lesh et al 2011; Luck et al 2008).

There is a considerable body of literature detailing a distributed fronto-parietal attention network (FPAN) responsible for the control of visual attention in healthy adults that serves as a model system to interrogate selective attention deficits (Corbetta & Shulman 2002; Vossel et al 2014). While functional distinctions between dorsal (frontal eye fields (FEF) and intraparietal sulcus (IPS)) and ventral (temporoparietal junction (TPJ) and ventral frontal cortex (VFC)) systems have been well described within this network, dynamic interactions between both are critical to the allocation of attentional resources. Interestingly, studies examining differences between location and feature (e.g., color) based cues have observed considerable overlap in the FPAN activated by both including FEF, IPS, and VFC (Egner et al 2008; Giesbrecht et al 2003). The FPAN, therefore, appears to represent a more general system reflecting internal goal representations rather than one tied to a specific type of cue.

Although few studies in psychosis have examined alterations in neural activity during anticipatory cue periods on selective attention tasks, deficits during working memory encoding reliant upon similar frontal and parietal structures have been well-described (Deserno et al 2012; Schneider et al 2007). A frequently observed phenomenon in these studies is a load-dependent dysfunction with patients exhibiting increased activity during less challenging conditions relative to healthy controls and impaired activity at higher load levels (Coffman et al 2020; Kim et al 2010; Potkin et al 2009). This effect was also observed using an implicit cueing paradigm in a Go/No-go task of behavioral inhibition with increased inferior frontal gyrus activation interpreted as an inefficient compensatory mechanism in patients to meet the demands of a fairly simple task (Arce et al 2006).

Our goal in the present study was to evaluate functioning of the FPAN during color cue processing and its relationship to the implementation of attention in first-episode psychosis. Brain activity was recorded using magnetoencephalography (MEG) during a cued visual search task with varying levels of attentional demand, and structural MRI scans were obtained to provide precise anatomic localization of sensor-level activity. The less demanding task condition (pop-out) consisted of a consistent cue-target color pair across trials and emphasized a parallel search strategy whereas the more demanding condition (serial search) emphasized a controlled serial search of stimuli with a variable cue-target color across trials. Our laboratory recently identified an impairment in the deployment of attention towards target during presentation of the search array among individuals with a schizophrenia spectrum illness at disease onset (Sklar et al 2020a). In the present investigation, we therefore examined FPAN activity during the anticipatory cue period to evaluate the integrity of this cognitive control network and its relationship to selective attention impairments and disease burden at the first episode of psychosis.

Methods and Materials

Participants

Participants included in the current report were identical to the sample in a previously published manuscript (Sklar et al 2020b). Demographic and clinical assessment information is presented in Table 1. Thirty-eight healthy controls (HC) and forty-six first-episode schizophrenia spectrum patients (FE) participated. Eight FE were excluded from analysis due to poor quality neurophysiologic data (n=2) and to obtain a matching with HC on age, gender, Wechsler Abbreviated Scale of Intelligence IQ, and parental socioeconomic status (n=6).

Table 1.

Demographics clinical assessment data (Mean ±SD) by group

	HC (n=38)	FESz (n=38)	t/χ2	p
Female/Male	13/25	14/24	0.06	1.0
Education (years)	14.1±3.2	12.4±2.4	2.58	.01
Age	22.8±5.0	21.8±4.4	0.91	.37
WASIa	108.2±10.6	107.3±14.2	0.31	.76
MATRICS-Total	48.2±7.6	38.2±14.3	3.79	.00
SESb	34.2±14.6	29.4±13.2	1.46	.15
PSESc	53.0±13.7	47.6±15.0	1.62	.11
GF: Roled	9.00±0.4	5.76±2.4	8.25	.00
GF: Sociale	9.04±0.2	5.55±1.8	12.2	.00
SANSf/SAPSg:
Inexpressivity	-	10.2±6.6
Asociality/Apathy	-	10.8±6.3
Reality Distortion	-	11.0±7.0
Disorganization	-	4.34±5.2
Percent Medicated	-	65.8
Medication (CPZh mg/day)	-	193.3±165.5

^aWechsler Abbreviated Scale of Intelligence;

^bSocioeconomic status;

^cParental socioeconomic status;

^dGlobal Functioning: Role scale;

^eGlobal Functioning: Social scale;

^fScale for the Assessment of Negative Symptoms;

^gScale for the Assessment of Positive Symptoms;

^hChlorpromazine equivalent dose

FE included individuals presenting with new-onset psychotic symptoms to any University of Pittsburgh Medical Center affiliated facility. HC were recruited from the greater Pittsburgh, PA region. All participants underwent a clinical assessment session that included a Structured Clinical Interview for DSM-IV (SCID-IV), medical screening questions to exclude individuals with a history of concussion, alcohol or substance use disorder, or neurological comorbidity, and a screening for colorblindness using pseudoisochromatic plates. All participants also completed the MATRICS Cognitive Consensus Battery (MCCB) to assess neurocognitive performance (Green et al 2004) as well as the Global Functioning: Role (GF: Role) and Global Functioning: Social (GF: Social) scales to assess community functioning (Cornblatt et al 2007). FE were assessed using Scale for the Assessment of Negative Symptoms (SANS) and Scale for the Assessment of Positive Symptoms (SAPS) to rate symptom severity prior to testing. SANS/SAPS scores were used to derive the 4-dimension symptom model (reality distortion, disorganization, inexpressivity, and apathy/asociality) validated by previous factor analyses using markers of neural dysfunction and long-term outcomes (Kotov et al 2016). Using this factor analysis, positive symptoms are comprised of reality distortion and disorganization while negative symptoms are comprised of inexpressivity and apathy/asociality.

HC with a current or past Axis I diagnosis were excluded. Only patients with a first-episode schizophrenia spectrum disorder were included in the present study. Diagnostic makeup for FE included 23 with schizophrenia, 6 with schizoaffective disorder (4 depressive and 2 bipolar type), and 9 with psychotic disorder not otherwise specified. FE participated within 2 months of first clinical contact and had less than 2 months of lifetime antipsychotic treatment with thirteen (~34%) being unmedicated at the time of testing.

All procedures were approved by the University of Pittsburgh IRB and participants provided informed consent prior to participation and were paid for participation.

Cued Attention Task

Participants completed a visual search task (Figure 1) described in pervious manuscripts (Sklar et al 2020a; Sklar 2020b). Each trial began with a centrally presented fixation cross (1°, 500ms). Fixation was followed by a centrally presented, feature-based cue (0.65°, 500ms) that indicated the target color for the upcoming search array. The search array (500ms) was made up of 6 eccentrically arranged colored annuli (subtending 0.65°, positioned 2° from fixation). Participants were instructed to covertly shift their attention towards the target annulus, identified based on its color, and indicate via button-press when a gap subsequently appeared to the right or left on the target.

Figure 1.

The visual search task used in the present study. In the pop-out version (A) the cue-target color remained consistent across trials and all distractors during the target array were of a uniform color. During the serial search task (B), cue-target color varied across trials and all target array distractors had unique colors. Cortical activity was examined during the Cue period.

The task consisted of 2 conditions (pop-out (PO) and serial search (SS)) with varying degrees of attentional demand. During PO, the color of the cue/target stimuli remained the same across trials and the 5 non-target annuli presented during the search array had uniform colors thereby emphasizing a parallel search strategy reliant upon bottom-up processing. In contrast, the cue/target color was variable across trials during SS and all non-target annuli in the search array had unique colors. This paradigm requires a search of individual array stimuli guided by top-down attentional systems in order to successfully locate the target annulus. Examination of the neural systems engaged by this task using lesions in non-human primates suggests that communication between cognitive control and sensory cortices is required for successful completion of the SS condition (Rossi et al 2007). Participants completed 4 task blocks, 2 of each condition with order counterbalanced across participants. Each block consisted of 144 trials.

Data Acquisition

Structural MRI scans were obtained for each participant using a Siemens Tim Trio 3T system. T1-weighted images were collected using a multi-echo 3D MP-RAGE sequence (TR/TE/TI = 2530/1.74, 3.6, 5.46, 7.32/1260 ms, flip angle = 7°, field of view (FOV) = 220 x 220 mm, 1 mm isotropic voxel size, 176 slices, GRAPPA acceleration factor = 2).

MEG data were recorded in a magnetically shielded room using an Elekta Neuromag VectorView system (Elekta Oy, Helsinki, Finland). Data was sampled from 306 sensors arranged into triplets, each with 2 orthogonal gradiometers and one magnetometer, at a rate of 1000Hz (online bandpass filter 0.1-330Hz). Head position was recorded throughout the testing session using 4 head positioning indicator coils and Neuromag MaxFilter software (http://imaging.mrc-cbu.cam.ac.Uk/meg/Maxfilter_V2.2) was used to correct for head motion during the scan. Blinks and eye movements were monitored using bipolar leads placed above and below the left eye (VEOG) and lateral to the outer canthi (HEOG), respectively, and cardiac activity was monitored using bipolar ECG leads.

MEG Data Pre-processing

Pre-processing procedures were followed to reduce the presence of artifacts from our MEG data. A temporal extension of the Signal Space Separation method (Uusitalo & Ilmoniemi 1997) was used to remove environmental magnetic artifacts from signals generated by neural activity (Taulu & Hari 2009). Data were visually inspected and the EEGLAB Toolbox (Delorme & Makeig 2004) was used to remove channels (re-interpolated following pre-processing) and segments of data corrupted by excessive noise. After applying a high-pass filter (0.5Hz; 12dB/oct), an adaptive mixture independent component analysis was preformed to isolate and remove eye-blink and cardiac activity.

Cortical Source Localization

MEG sensor data was registered to each participant’s structural MRI scan. The grey/white matter boundary was segmented using Freesurfer (http://www.surfer.nmr.mgh.harvard.edu) and tessellated into an icosahedron mesh with 5mm spacing between vertices. Subsequent processing steps were conducted using Brainstorm software (Tadel et al 2011). Sensor-level data was subjected to a low-pass (100Hz; 24dB/oct) filter and segmented into epochs constructed around the presentation of the color cue. A 300ms (−350ms to −50ms pre-cue) window was used for baseline correction and segments with data exceeding ±5pT were excluded with remaining trials used to generate participant averages. The minimum-norm estimate approach, a distributed source solution, was used to project MEG sensor data to the cortical surface. The forward solution, modeled as a series of overlapping spheres (one per sensor location), and a noise covariance matrix calculated from the baseline window were used to create a linear inverse operator using an orientation constraint of 0 (current dipoles normal to the cortex) with depth weighting applied (HÄmÄlÄinen and Hari 2002; Lin et al 2004). Cortical current density values were normalized to the baseline period in order to create a z-score for activity at each vertex.

Selection of regions of interest (ROIs) comprising the FPAN was guided by previous functional imaging work. These studies have identified a remarkably similar network for location and feature (e.g. color) based cues including the frontal eye fields (FEF), inferior frontal cortex, posterior parietal structures, and dorsal portions of the cingulate cortex (Egner et al 2008; Giesbrecht et al 2003). ROIs included in the current analysis were selected from the Destrieux parcellation available in Freesurfer (Destrieux et al 2010) based on sulci and gyri within this network exhibiting robust activity during the 500ms cue period (Figure 2). These regions included bilateral (right (RH) and left (LH) hemisphere) FEF comprised of superior and inferior pre-central sulci (Lobel et al 2001), the pars opercularis region of the inferior frontal gyrus (IFG), the intraparietal sulcus (IPS), and the middle-posterior segment of the cingulate cortex (MCC).

Figure 2.

Cortical activity averaged over the 500ms cue period for HC (A) and FE (B) as well as quantification of activity within each region displayed by group. Regions of a fronto-parietal network selected for analysis based on previous literature utilizing feature-based cues. Error bars represent SEM. *p<.01.

Data Analysis

Task accuracy and response times were analyzed using a 2 (group: HC/FE) x 2 (task: PO/SS) ANOVA. Both performance measures were subjected to log transformation due to violation of normality (Shapiro-Wilk Test p>.05) prior to analysis. Cortical activity averaged over the cue period was analyzed using a 2 (group: HC/FE) x 2 (task: PO/SS) x 8 (ROI) ANOVA. Linear regressions for each ROI were conducted including measures of task performance (accuracy and response time) for both FE and HC as well as symptom burden (reality distortion, disorganization, inexpressivity, and apathy/asociality dimensions) and community functioning (GF: Role and GF: Social scales) in FE. Finally, effects of medication on outcome measures of interest were examined by including medication status as a between-group factor in analyses of FPAN activity and performance among FE Results were considered significant at p≤0.05.

Results

Task Performance

FE responded less accurately (F_1,74=7.05, p<.01) and more slowly (F_1,74=7.19, p<.01) compared to HC. Overall, responses were less accurate during SS compared to PO (F_1,74=10.39, p<.01), though this did not differ between groups (p=.55). There was no effect of task or interaction between task and group on response times (p’s>.1).

FPAN Source Localized Activity

Cortical activity evoked during the 500ms cue period in each group is presented in Figure 2. Although overall activity across the FPAN did not differ between groups (p=.79), the pattern of activity did differ between HC and FE across ROI (Group x ROI interaction F_7,518=3.00, p<.01). Specifically, FE exhibited larger RH IFG activity compared to HC (t₇₄=−2.75, p<.01) with no differences in other ROIs (p’s>.1). No additional interactions involving group were significant (p’s>.3).

A main effect of ROI was also observed (F_7,518=81.18, p<.01). Full reporting of ROI activity differences is available in Supplemental Table 1. No main effect of task or interactions involving it were observed (p’s>.2).

Relationship Between FPAN activity, Task Performance, and Clinical Measures

Regression statistics for FE are presented in Table 2. Partial regression plots depicting significant associations are presented in Figure 3. In FE, more rapid responses were associated with greater activity in RH FEF (β=−0.52, p=.03) and RH IFG (β=−0.49, p=.04). Larger LH IFG activity was also associated with bigger Global Functioning: Role scores (β=0.52, p=.03) and larger RH IPS activity was associated with lower inexpressivity scores (β=−0.44, p=.03) in FE. No significant relationships between FPAN activity and either task performance (p’s>.05) or community functioning (p’s>. 1) were observed among HC.

Table 2.

Regression statistics for each fronto-parietal attention network region.

	LH FEF R2=.24	RH FEF R2=.27	LH IFG R2=.33	RH IFG R2=.27	LH MCC R2=.18	RH MCC R2=.20	LH IPS R2=.12	RH IPS R2=.46
Accuracy	−0.22	−0.32	−0.30	−0.17	0.03	−0.03	−0.03	−0.33
Response Time	−0.44	−0.52	−0.13	−0.49	0.07	−0.03	−0.09	−0.37
Inexpressivity	−0.33	0.1	−0.15	−0.03	−0.05	−0.12	−0.13	−0.44
Asociality	−0.34	−0.12	−0.03	−0.16	0.04	−0.07	−0.35	−0.23
Reality Distortion	−0.34	−0.19	−0.28	−0.07	−0.06	0.05	0.15	0.19
Disorganization	0.01	−0.23	0.10	−0.14	0.01	−0.17	−0.11	0.05
GFS: Role	0.06	−0.08	0.52	0.20	0.37	0.41	0.12	0.28
GFS: Social	−0.17	−0.07	0.1	−0.12	0.05	−0.26	−0.23	−0.21

FEF: Frontal eye fields; IFG: Inferior frontal gyrus; MCC: Midcingulate cortex; IPS: interparietal sulcus; GFS: Global Functioning Scale

Boldface indicates significant coefficients (p<.05)

Figure 3.

Partial regression plots depicting relationship between response time and RH FEF (A) and RH IFG (B) activity, GF: Role measure of community functioning and LH IFG activity (C), and the inexpressivity negative symptom dimension and RH IPS activity (D).

Medication Effects

There was no effect of medication status or interactions involving it on FPAN activity (p’s>.1) or behavior measures (p’s>.2).

Discussion

The present study examined functioning of the FPAN during cued visual search and its association with disease morbidity at the onset of schizophrenia spectrum illness. Using a combination of high resolution neurophysiological and structural imaging techniques, we observed increased network activity in FE, specifically within the RH IFG, despite impaired task performance relative to HC. Furthermore, associations between increased activity and improved performance among FE, a finding not present in HC, suggest a reliance on this network to meet the demands of a relatively simple selective attention task. These results suggest a cognitive control network engaged in excess of task demands to overcome deficits in the deployment of attention, a compensatory mechanism that may become overwhelmed in more complex real-world environments or during more chronic illness stages. Finally, greater disease burden was associated with an impaired ability to employ this cognitive strategy.

Group differences in cue-related network activity were restricted to the RH IFG. As a node within the ventral FPAN, this region is typically associated with bottom-up reorienting towards task-relevant stimuli (Corbetta & Shulman 2002; Vossel et al 2014). However, regions of the right inferior frontal cortex including the portion of the RH IFG examined in the current investigation represent an interface between dorsal and ventral attention networks, acting in concert with dorsal frontal and parietal regions to direct the focus of attention (Asplund et al 2010; Corradi-Dell’Acqua et al 2015). Furthermore, this region has been directly implicated in the anticipatory, top-down modulation of sensory-perceptual activity within visual cortices during stimulus color processing (Zanto et al 2010). Finally, the RH IFG is engaged in a number of cognitive control tasks involving inhibitory attentional control including gaiting of irrelevant distractors during visual search (Fockert et al 2004; Suda et al 2020). Larger RH IFG activity in FE during the current investigation can therefore be conceptualized as an enhanced utilization of cortical resources dedicated to the control of attention. While this finding appears to contradict previous reports of functional and structural abnormalities of the IFG in schizophrenia (Harms et al 2010; Iwashiro et al 2012; Jeong et al 2009), this conflict might be explained by disease chronicity with relatively preserved IFG function at illness onset.

Contrary to the intended design of the visual search task, there was no significant difference in FPAN recruitment between PO and SS. In addition, the lack of any relationship between FPAN activity and performance among HC emphasizes the relative simplicity of the task. In contrast, FE exhibited significant correlations between two frontal regions within this network and response time, indicating an increased reliance on an executive control network to complete an otherwise simple task. Interestingly, an inefficient use of cognitive resources was also observed in our recent analysis of attentional deployment following presentation of the stimulus array in which FE exhibited an increased reliance on parietal structures during the SS condition (Sklar et al 2020a). In contrast, HC relied more heavily on sensory-perceptual cortices to carry out this process. Together, these results are consistent with findings of a load-dependent difference in executive network activity between HC and people with schizophrenia in the working memory literature used to support the “inverted U” model of cognitive efficiency (Coffman et al 2020, Kim et al 2010; Potkin et al 2009). An alternative interpretation of the larger RH IFG activity is that it reflects an increased focus of attention on singleton items such as the centrally presented color cue in people with schizophrenia, a phenomenon referred to as “hyperfocusing” (Luck et al 2019).

Beyond revealing a difference in cognitive strategy relative to HC, FPAN activity in FE and its associations with their clinical assessments underscore the impact of cognitive dysfunction on disease burden during this early illness stage. Despite a lack of any significant correlation with positive symptom dimensions, weaker cue-related FPAN activity within RH IPS and LH IFG was associated with greater inexpressivity dimension scores and lower GF: Role scores, respectively. This pattern is consistent with previous reports documenting robust associations between attention impairments and both negative symptoms and impaired global functioning in schizophrenia with relatively weak or absent associations with core positive symptoms of the disorder (Chen & Faraone 2020; Curtin et al 2019). While the causal relationships underlying these correlations remain speculative, previous regression analyses suggest that negative symptoms act as mediators of the relationship between neurocognitive deficit and impaired community functioning (Ventura et al 2009; Meyer et al 2014). Regardless of the directional nature of these interactions, identifying the neural substrates of negative symptoms and impaired neurocognition at disease onset for future targeted interventions would be of considerable clinical value.

Limitations

There are limitations to be considered when interpreting results from the present investigation and to be addressed by future work. First, as previously noted, our manipulation of attentional load did not have as substantial an effect as desired, limiting our ability to assess efficiency of the FPAN across a meaningful range of task complexity. A task employing a more complex cue-target design would be better suited for this purpose. Second, our participants engaged several structures outside the previously defined color-cue FPAN, most notably in temporal regions. This temporal lobe recruitment might have resulted from participants verbalizing the color of the cue at the moment of its presentation, though this possibility cannot be addressed with the data at hand. Finally, the cue delay period used in the current paradigm was fairly short (500ms). A longer epoch would have provided a better opportunity to capture dynamic spectral changes, particularly within lower frequency bands with longer period lengths, to assess effective connectivity and provide a direct measure for top-down sensory-perceptual modulation. However, while cue periods of 1000ms or longer are more common, previous work has documented the top-down modulation of visual perception by fronto-parietal structures within 200ms of cue stimulus onset (Zanto et al 2010).

Conclusion

In conclusion, examination of the cognitive control FPAN during a low-complexity visual search task revealed enhanced cue-related IFG activity within a first-episode schizophrenia sample. Combined with the correlations between FPAN activity and task performance, these findings suggest an inefficient use of cognitive resources and may reflect a compensation for impairments in the downstream deployment of attention. Furthermore, implementation of this strategy was associated with reduced negative symptom burden and better community functioning, providing meaningful targets for disease modifying treatments during early illness stages.