Connectivity of the human insula: A cortico-cortical evoked potential (CCEP) study

Abstract

Objective:

The human insula is increasingly being implicated as a multimodal functional network hub involved in a large variety of complex functions. Due to its inconspicuous location and highly vascular anatomy, it has historically been difficult to study. Cortico-cortical evoked potentials (CCEPs), utilize low frequency stimulation to map cerebral networks. They were used to study connections of the human insula.

Methods:

CCEP data was acquired from each sub-region of the dominant and non-dominant insula in 30 patients who underwent stereo-EEG. Connectivity strength to the various cortical regions was obtained via a measure of root mean square (RMS), calculated from each gyrus of the insula and ranked into weighted means.

Results:

The results of all cumulative CCEP responses for each individual gyrus were represented by circro plots. Forty-nine individual CCEP pairs were stimulated across all the gyri from the right and left insula. In brief, the left insula contributed more greatly to language areas. Sensory function, pain, saliency processing and vestibular function were more heavily implicated from the right insula. Connections to the primary auditory cortex arose from both insula regions. Both posterior insula regions showed significant contralateral connectivity. Ipsilateral mesial temporal connections were seen from both insula regions. In visual function, we further report the novel finding of a direct connection between the right posterior insula and left visual cortex.

Significance:

The insula is a major multi-modal network hub with the cerebral cortex having major roles in language, sensation, auditory, visual, limbic and vestibular functions as well as saliency processing. In temporal lobe epilepsy surgery failure, the insula may be implicated as an extra temporal cause, due to the strong mesial temporal connectivity findings.

INTRODUCTION

Over the recent years, technological advances in noninvasive neuroimaging have further expanded a fundamental concept in neuroscience, namely, that the nervous system is an interconnected and intercommunicating network of neurons [1]. Unraveling these networks has provided the basis of the human connectome; a comprehensive description of the structural connectivity of the brain [2]. Within this structural connectome, lies a plasticity and hierarchy, designated by systems neuroscience as functional connectivity-statistically inter-dependent brain regions during which integration results in specific neurophysiological events and effective connectivity which refers to the influence of one neural system exerting its influence over another [3].

Cortico-cortical evoked potentials “CCEPs”, is a relatively novel in vivo brain mapping technique, utilizing low frequency stimulation, in patients undergoing invasive monitoring for refractory epilepsy [4]. CCEP mapping represents a form of evoked effective connectivity. Here, there is a causal influence between brain regions, where one region may influence another but not vise-versa. For example, stimulation at one site (independent measure) may result in evoked responses (dependent measure) at distant sites, but this relationship may not be reciprocal. Unlike functional connectivity, effective connectivity is a direct measure of this influence.

CCEPs have significant properties including notable spatial-temporal resolution, the ability to sample across distributed networks and the ability to accurately localize the stimulation site [5]. CCEPs have been previously utilized to define networks pertaining to language [4], limbic [6] and motor systems [7] as well as preferential excitability of seizure onset regions [8,9].

The human insula has been viewed as an enigmatic cortical region. This is partly due to its anatomical location nestled behind the frontal, temporal, and parietal opercular cortices and dense surrounding vascular network which has made it historically a challenge to study [10,11]. The adult anterior insula comprises three anterior short gyri (anterior, middle and posterior), separated from the posterior insula by a central insula sulcus. The posterior insula comprises the two long gyri (anterior and posterior) [12,13].

Despite covering only 2% of the cerebral cortex, the insula acts as a highly diverse functional highway implicated in a vast number of homeostatic, cognitive and affective processes. These include; vestibular, visceromotor and viscerosensory functions, somatomotor, pain and temperature, motor association, ocular motor, language, limbic integration, auditory function and also spans the realms of cognition including memory, body awareness, emotion and self recognition [14,15].

How the insula is able to integrate and be involved in such a wide array of functions has prompted attempts to uncover its connectivity networks. Initially such discoveries were based on tracer injection studies and dissection in both humans and primates [16–8]. With the advent of diffusion tractography and advanced fMRI algorithms, further insula connections were observed in vivo [19–21].

Given the significant efferent output from the insula, it comes as no surprise that it may potentially be able to mimic epilepsies arising from any lobe of the brain and also account for epilepsy surgery failure in temporal, frontal and parietal lobe epilepsy surgery [22–25]. Thus, the importance of sampling the insula by intracranial electrodes in lesion-negative refractory epilepsy is increasingly being recognized.

Despite this however, there are still only a handful of reports using CCEP methodology, which has attempted to describe the connections of the human insula [26–29].

In this study, CCEP data of individual gyri in the dominant and non-dominant insula was analysed, in patients undergoing epilepsy surgery with Stereo-Electroencephalography. Maps of connectivity measures, based on the averaged root mean square (RMS) values, were generated highlighting the connectivity strength of various cortical regions.

PATIENTS AND METHODS

We report how we determined our sample size, all data exclusions, all inclusion/exclusion criteria, whether inclusion/exclusion criteria were established prior to data analysis, all manipulations, and all measures in the study.

This retrospective analysis utilized 192 patients who underwent SEEG evaluation in 2014 (99 patients) or 2015 (93 patients) for epilepsy surgery. SEEG electrodes were typically 6 to 16 contacts, 2.5mm long, 0.8mm diameter, separated by 2 mm of space. Placement into the insula was either in an orthogonal or oblique manner depending on the desired coverage. Patients with electrode contacts in either of the insula regions were selected (n= 71). CCEPs were performed with the patient on full medications at the end of the evaluation.

Of the 71 patients, 42 patients had CCEPS performed in either insula region. To standardize inter-regional connectivity, patients with previous epilepsy surgery, insula abnormalities on MRI or seizure onset in the insula, were excluded (9 patients). A further 3 patients were excluded because there was ambiguity as to whether the electrode was sampling the insula or an adjacent structure (e.g. operculum). In total, 30 patients were selected for ongoing analysis.

The methods for obtaining CCEPs at our center have previously been outlined [4]. Stimulation was performed with the Grass S88 stimulator (Warwick, RI, USA) using an automated interface in Nihon-Kohden software. A paradigm using 1Hz stimulation with 0.3msec square wave pulses of alternating polarity between two adjacent electrode contacts located in the desired anatomical region was utilized. 30 trails were performed at 2mA, 4mA and 6mA intensities and 60 trials at the 8mA setting. This stimulation protocol is safe and would not cause localized tissue injury [30]. Stimulus intensity was truncated if after discharges or seizures had occurred.

Anatomical analysis and determining regions of interest

The anatomical location of each electrode contact, for every stereo-EEG case, had previously been tabulated for clinical use. Confirmation of electrode placement for all 30 patients was re-reviewed using the CT and MRI reconstructions assembled via the Curry Neuroscan software (Compumedics, USA), prior to any CCEP analysis. In particular, the insula contacts were analysed for their location with the insula itself. Any contacts that were extra-cerebral, placed in white matter or showed continued artifact, were removed from analysis. An aggregate of all the electrode contacts, across all 30 patients, superimposed onto a standardized MNI-space template, is shown in supplementary figures 11 and 12

CCEP analysis

The raw SEEG electrode CCEP averaged data was extracted from the 256-channel montage display of the Nihon Kohden digital EEG machine (NeuroWorkbench version 05–20; Nihon Kohden America, Inc., CA, USA). These raw CCEP-SEEG epoch files were then reviewed in MATLAB (MATLAB, Matworks Inc.). Synchronization pulses were captured in the NK software to define the precise time of the stimulation event. The evoked response, sampled at 1000Hz, was then bandpass filtered at 1–300Hz and notch filtered at 60Hz. The selected time window parameters included a 100msec baseline period prior to stimulation onset and 400msec time window following the onset of stimulation. Averaged evoked potentials acquired at stimulation amplitudes of 2, 4, 6 and 8mA were analyzed.

CCEP responses using subdural grid recordings have been reported to comprise two peaks-an early N1 and late N2 negative response. However in SEEG, there are often inconsistencies in the CCEP morphology, in part due to electrodes located within sulci or traversing grey and white matter [31]. Rather than relying on the latency or amplitude of the response, analysis of the root mean square (RMS) is utilized to quantify its strength. The RMS (quadratic mean) is a statistical measure of signal amplitude, representing the average absolute deviation of a signal in time, performed by taking the square root of the sum of the squared terms. We compared the ratio of the baseline RMS value at 100ms before a stimulation pulse and compared this to the RMS value at 10 to 300 milliseconds after stimulation. A kurtosis method (threshold setting at 8) was utilized to reject stimulation artifact, but not large evoked responses [32]. Rejected items were manually analyzed to confirm the presence of a stimulation artifact and not an exaggerated evoked potential.

Responses to CCEP stimulation were based on RMS values. These were calculated upon the averaged evoked potentials and baselines, consistent with prior CCEP methodology [4,6]. Electrode contacts were grouped by anatomical region of interest (ROI), with results in the same ROI averaged in order to give a connectivity summary for that ROI. It was assumed that stimulations from the same anatomical area would activate the same network, therefore if multiple stimulations were performed within the same gyri in a patient, results were averaged across stimulations. These results were ranked into Quartile Values (QV) for each ROI for the stimulated gyri, giving values of 4 (maximal) to 1. To allow for comparison between patients, weighted means (WM) were computed for RMS quartile values, which took into account the number of trials incorporated in the QV. Standard deviation of results across patients for each QV were also computed to see if the spread was large for an ROI. QVs were shown on specialized circro plots, which captured the connections of the stimulated gyri. Since this research was exploratory in nature, there were no hypotheses tested, and therefore, the standard deviation can be thought of as a reliability check for results, with larger spreads considered less reliable. Epileptic contacts were also left in the analysis for completeness.

Circle maps

Circro plots denote the projections of the stimulation site graphically and were created using a modified open source toolbox (https://github.com/bonilhamusclab/circro). The plots are organized so that right hand anatomical sites, representing the right hemisphere, are in bold, and left hemisphere anatomical sites placed on the left. Total anatomical sites differed slightly between the various gyri stimulated as there was variability in sampling across patients, due to individualized hypotheses in Stereo-EEG.

The number following the anatomical site indicates how many contacts and patients were sampled during stimulation of a particular insula gyrus. Heights of columns on the circle border illustrate the number of electrode contacts sampled graphically, and the colour highlights the standard deviation of the responses across patients, which may be thought of as a reliability measure since no statistical analysis could be performed. Lines between sites show the strength of connectivity between ROIs, with colours representing the WM of QV’s. The thickness of the line indicates the relative number of electrode contacts sampled at the response site, normalized by the number of contacts in the most sampled site on each plot. The colour bars on left and right show the scales for both standard deviation and weighted mean of quartile values. Because of the number of connections, a circro plot for each insula gyri was created.

RESULTS

Table 1 shows the patient demographics for all 30 patients. Dominant language areas were in the left hemisphere for all patients, including the 3 patients who were left handed and shown to be left dominant by WADA testing. Right and left gyri are denoted by and “R-” and “L-” prefix, respectively. Due to the volume of the data, only significant connections with a weighted mean QV greater than 3 are discussed, as these connections are considered to be the most robust. Circro plots are shown for each individual gyrus in figures 1–10 with anatomical abbreviations provided in Table 2. The complete data set is available in supplementary Table 3, with all the total number of all electrodes sampling all anatomical regions provided in supplementary Table 4.

Table 1:

Patient Demographics

Age/hand	Age Onset	SEEG Seizure focus	MRI	FDG-PET	SPECT	Insular Contact	Insula 50Hz Stimulation	Surgery
19y/RH	7y	Left Middle Frontal Gyrus (MFG)	Mild dysplasia L hippocampus	Bitemporal hypometabolism without asymmetry.	Hyperperfusion in L dorsal lateral central, adjacent frontal opercular, and posterior insula.	L MSG	Throat sensation (dry)	Left MFG resection
21y/R H	9y	Left orbitofront al and mesial temporal	Slight asymmetric enlargement of L amygdala	Reduced FDG activity in b/l mesial temporal lobes including mesial temporal regions.	N/A	L ALG	N/A	Left orbitofrontal and temporal
						L MSG	N/A
20y/R H	14y	Left Middle temporal gyrus (MTG)	Unremarkable	Hypometabolism in L lateral anterior temporal lobe.	B/l Hyperperfusion in cluster areas of L temporal, frontal and parietal region.	L ALG	N/A	Left temporal lobectomy
28y/R H	5m	Not localizable	Unremarkable	Bitemporal & L parietal hypometbolism.	Hyperperfusion in R posterior frontal, R temporal, and L anterior parietal lobe	L MSG	N/A	NA
						L PSG	N/A
						L PLG	N/A
34y/R H	24y	Left temporal	Unremarkable	B/l moderate hypometabolism in mesial temporal lobes	N/A	L ASG	N/A	Left temporal lobectomy
43y/R H	5y	Right frontal operculum	R frontal opercular cortical malformation, b/l hippocampal volume loss	Mild hypometabolism in b/l anterior medial temporal regions	Hyperpefusion in R posterior insular region or R dorsal lateral central region	R ALG	L hand burning	Frontal operculum
						L ASG	N/A
39y/R H	18y	Right frontal operculum	Unremarkable	B/l hypometabolism in anterior mesial temporal regions, more on the R side	N/A	R PLG	L hand tingling	Frontal operculum
						R ASG	Nil
						R ASG	Nil
49y/R H	22y	Bilateral orbitofront	Encephalomalacia in L frontal lobe	Left frontal hypometabolism	N/A	R MSG	N/A	NA
61y/R H	10y	Left temporal	Unremarkable	Severe hypometabolism in L mesial temporal lobe	Hyperperfusion in cluster area of L basal frontal or L lateral deep parietal region	L MSG	Difficulty reading	L lateral temporal lobectomy
						L PLG	Painful electric shock in R calf
						L ALG	Loss of verbal fluency
						L PLG	N/A
20y/R H	5y	Right pars orbitalis	Unremarkable	Mild bitemporal and focal hypometabolism in R mid to anterior dorsomedial frontal cortex.	Hyperperfusion in both hemispheres in R posterior dorsal medial frontal, R posterior basal medial temporal and L mid cingulate gyrus.	R ASG	N/A	Right pars orbitalis
12y/R H	4m	Right Superior Frontal Gyrus (SFG)	Unremarkable	Mild-moderate bitemporal hypometabolism.	No dominant foci of ictal hyperperfusion.	R MSG	N/A	Right superior frontal resection
						R MSG	N/A
30y/L H	10y	Right temporal	Unremarkable	B/l temporal hypometabolism more pronounced on the R	N/A	R ALG	N/A	Right temporal lobe resection
						R MSG	N/A
27y/R H	17y	Right Superior Frontal Gyrus (SFG)	Unremarkable	Bitemporal hypometabolism	N/A	R Circ S	N/A	Right superior frontal resection
36y/L H	1y	Right superior and Middle frontal gyrus	Unremarkable	Mild b/l temporal and frontal parasaggital hypometabolism	Hyperperfusion in L medial parietal and occipital regions	R Circ S	Whole body parasthesias	Right frontal resection
38y/R H	31y	Right neocortical temporal lobe	Unremarkable	Hypometabolism with mild asymmetry in R anterior medial basal temporal region	Hyperperfusion in R superior temporal, frontal operculum and posterior insula	R PLG	Left hemibody parasthesias	Right temporal lobe resection
20y/R H	9y	Multifocal	B/L schizencephaly and pachygyria in the parietal regions.	N/A	Hyperperfusion in bilateral temporal, opercular regions and left posterior insular region	R ASG	N/A	No surgery
						R PSG	N/A
						L PSG	R hand tingling
53y/R H	35y	Left mesial temporal region	Questionable asymmetric cortical thickening of inferior margin of L anterior temporal pole	Asymmetric hypometabolism involving L anterior temporal lobe, L orbitofrontal and L insula regions	N/A	L ASG	N/A	Left temporal lobectomy
25y/R H	4.5y	Multifocal regions	L hippocampus MTS, abnormal sulcation at rectus gyrus and anterior interior frontal lobe	Severe hypometabolism Left hemisphere	Hyperperfusion in L mid to anterior temporal regions extending to the L insular region. Also in L anterior medial frontal regions extending to the R mid cingulate gyrus.	L PSG	Nil	No surgery
						L ALG	Nil
28y/R H	19y	Right Temporal	Heterotopic gray matter in R medial temporal lobe	Bitemporal hypometabolism	Hyperperfusion in L posterior basal frontal region, b/l insula and dorsal central regions	R MSG	N/A	Right temporal lobectomy
35y/R H	unkno wn	Right mesial temporal region	Unremarkable	B/l temporal, parietal and occipital hypometabolism. R temporal asymmetry	N/A	R ASG	N/A	Right anterior temporal lobectomy
32y/R H	4y	Left Parietal Operculum	Unremarkable	Bitemporal hypometabolism without asymmetry. Mild hypometabolism in L parietal operculum region.	Hyperperfusion in L parietal operculum, L medial occipital, L lateral temporal and occipital regions	L PLG	Whole body sensation	No surgery
54y/R H	25y	Right temporal region	Unremarkable	Focal hypometabolism in R mesial R temporal	N/A	R PLG	N/A	Right temporal lobectomy
31y/R H	8m	Right Angular gyrus	Unremarkable	Bitemporal hypometabolism with asymmetry in R mesial temporal, R orbitofrontal, R anterior parasagittal frontal, R parietal operculum	Hyperperfusion in R basal medial temporal occipital junction	R PLG	N/A	Right parietal resection
52y/R H	5y	B/L temporal	Unremarkable	Diffuse cortical hypometabolism,	N/A	R ASG	N/A	Neuropace
						L ASG
24y/R H	15y	Right temporo-occipital regions	Unremarkable	Bitemporal hypometabolism with R lateral temporal asymmetry	Hyperperfusion in R lateral temporal parietal region	R PLG	Auditory- high pitch sound	Right temporo-occipital resection
69y/R H	unknown	Left Mesial temporal	Unremarkable	N/A	Hyperperfusion in L posterior lateral temporal region	L PLG	R foot painful stabbing sensation	Left temporal lobectomy
						L MSG	N/A
45y/R H	5y	Right temporal lobe	Bilateral hippocampal symmetric hyperintensity in T2/FLAIR with mild volume loss.	Diffuse hypometabolism in b/l temporal lobes more pronounced on the R.	Hyperperfusion in both hemispheres in areas of R posterior basal frontal regions and R mid deep medial frontal region	R ASG	Nil	Right temporal lobectomy
						R PLG	Warm feeling and parasthesias in both arms
17y/L H	6y	Right planum temporale and parietal operculum	Unremarkable	Hypometabolism of b/l parietal lobes more pronounced in the superior, parasagittal regions and bilateral temporal lobes more pronounced on the L	N/A	R ALG	L arm numbness	No surgery
23y/R H	14y	Left orbitofront al	Unremarkable	L frontal and temporal hypometabolism	N/A	L PLG	N/A	L orbitofrontal
27y/R H	10y	R hemisphere multifocal epilepsy	Subependymal heterotopia along atrium of R lateral ventricle. Thickened cortex in R frontal operculum	Hypermetabolism in R anterior supramarginal gyrus	N/A	R PLG	N/A	Neuropace
						R PSG

Patient data collectionAbbreviation - y: year, m: month, R: R, L: L, N/A: not applicable, FCD: focal cortical dysplasia, RH: R-handed, LH: L-handed, b/l: bilateral, ASG: Anterior short gyrus, MSG: Middle short gyrus, PSG: Posterior short gyrus, ALG: Anterior long gyrus, PLG: Posterior long gyrus

Fig. 1.

Right ASG

Fig. 10.

Left PLG

Table 2:

Abbreviations for figures 1–10:

AC- Anterior Cingulate	Hip t- Hippocampal tail	MTG- Middle temporal gyrus	Precu post- Precuneus posterior
AG- Angular Gyrus	IFG PO- Inferior frontal gyrus pars opercularis	OcG lat- Occipital gyrus lateral	Precu sup- Precuneus superior
Amy- Amygdala	IFG POrb- Inferior frontal gyrus pars orbitalis	OcG mes- Occipital gyrus mesial	Put- Putamen
Cin s- Cingulate Sulcus	IFG PT- Inferior frontal gyrus pars triangularis	OrF lat- Orbitofrontal lateral	SFG lat- Superior frontal gyrus lateral
Clau- claustrum	IFS- Inferior frontal sulcus	OrF mes- Orbitofrontal mesial	SFG mes- Superior frontal gyrus mesial
Col s- Collateral Sulcus	Ins ALG- Insula anterior long gyrus	Para lob- Paracentral lobule	SFS- Superior frontal sulcus
CS- Central Sulcus	Ins ASG- Insula Anterior short gyrus	PC- Post Cingulate	SMG- Supramarginal gyrus
Cun- Cuneus	Ins Cir s- Insula circular sulcus	PHG- Parahippocampal gyrus	SPL- Superior parietal lobule
Ent c- Entorinhal cortex	Ins MSG- Insula middle short gyrus	Pl. pol- Planum polare	STG- Superior temporal gyrus
Fim- Fimbriae	Ins MSG- Insula middle short gyrus	Pl. tem- Planum Temporale	STS- Superior temporal sulcus
FO- frontal operculum	Ins PSG- Insula posterior short gyrus	PO-Parietal Operculum	SubC s- Subcentral sulcus
FP lat- Frontal pole (lateral)	IPL- Inferior parietal lobule	PO s- Parieto-occipital sulcus	SubC g- Subcentral gyrus
FP mes- Frontal pole (mesial)	IPS- Inferior parietal sulcus	PostC g- Post central gyrus	SyFi- Sylvian Fissure
Fus- Fusiform	ITG- Inferior temporal gyrus	PostC s- Post central sulcus	T.Pol lat- Temporal pole lateral
GP- Globus pallidus	ITS- Inferior temporal sulcus	PreC g- Precentral gyrus	T.Pol mes- Temporal pole mesial
GyR- Gyrus Rectus	Lin- Lingula	PreC s- Precentral sulcus	Unc- Uncus
Heschl- Heschl’s gyrus	MC- Mid cingulate	Precu ant- Precuneus anterior
Hip h- Hippocampal head	MFG- Middle frontal gyrus	Precu inf- Precuneus inferior

The Anterior Short Gyrus (ASG)

Six patients had CCEPs in the R-ASG, with one patient having two electrode pairs located within the R-ASG, totaling 7 separate pairs of CCEP stimulations. A total of 639 electrode contacts in 76 bilateral grey matter regions were analyzed. 457 contacts were located in the right hemisphere across 46 anatomical areas, while 182 contacts in the left hemisphere accounted for 30 anatomical regions.

Five pairs of electrodes from the L-ASG had CCEP responses obtained from 5 patients. A total of 430 electrodes were analyzed across 67 regions of the brain, with 125 contacts in the right hemisphere, across 28 anatomical areas and 305 contacts in the left hemisphere derived from 39 regions.

Figures 1 and 2 represent all the various connections of the insula ASG across these contacts.

Fig. 2.

Left ASG

Ipsilateral connections:

Frontal connections:

Both ASG had strong connections to the frontal opercular regions (right-24 contacts, WM 3.5 and left 26 electrodes, WM 3). Both areas were also strongly connected to the inferior frontal gyrus-pars opercularis (right-22 electrodes, WM 4 and left-15 electrodes, WM 3.6).

Differences between the right and left ASG were seen in connections to the following: Lateral pre-frontal areas. The R-ASG connections were strongly represented in the inferior frontal gyrus pars triangularis (PT) and orbitalis (PO) (13 and 17 electrodes respectively) and also in the lateral frontal pole (15 contacts). In contrast the L-ASG had only a very weak connection to pars orbitalis (WM 1.4, 5 electrodes) and to the entire frontal pole contacts (WM 1 in a total of 14 electrodes from mesial to lateral). No IFG pars triangularis representation was available on the left.

The L-ASG also strongly linked to the primary motor area (2 contacts). This was not seen with the right.

Temporal connections:

Shared strongly connected common efferents from the right and left ASG, were seen in the following temporal lobe locations:

1. Mesial temporal areas (R-ASG: uncus (2 contacts), hippocampal head (11 contacts), amygdala and entorhinal cortex (18 and 1 contacts respectively), and L-ASG: hippocampal head (8 contacts), uncus (1 contact,), entorhinal cortex (5 contacts), amygdala and collateral sulcus (6 and 3 contacts respectively))

2. temporal pole (R-ASG; mesial to lateral 7 contacts), L-ASG; mesial temporal pole (12 contacts)

3. temporal neocortical areas (R-ASG; superior temporal sulcus (STS) 2 contacts and superior temporal gyrus (STG) 16 contacts). L-ASG; STS (6 electrodes).

4. temporal operculum (anterior) (R-ASG; planum polare (5 contacts), L-ASG planum polare (8 contacts)

Differences in the strength of connectivity was found between the L-ASG which was seen to have a strong connections in the basal temporal region (fusiform gyrus (4 contacts, WM 3), ITG (8 contacts, WM 3) while the R-ASG was poorly connected to these areas (fusiform, 14 contacts, WM 1.4 and ITG 6 contacts WM 1).

Parietal connections:

Both ASG were highly connected to the angular gyrus

Differences included the R-ASG being strongly connected with the lateral parietal connections (R-ASG; supramarginal gyrus (SMG), 14 contacts and intraparietal sulcus (IPS), 2 contacts). The L-ASG however had a poor connection to the left SMG (WM 1.7, 6 contacts) and no electrodes sampling the IPS.

Other regions

The L-ASG strongly connected to the ipsilateral claustrum (1 contact), This was not sampled on the right.

Contralateral connections:

Contralateral connections from the R-ASG were to the left IFG PT (5 contacts), claustrum (1 contact), lateral orbitofrontal region (5 contacts) and the left anterior cingulate (8 contacts). Of note, the right anterior cingulate also had contralateral connectivity via the L-ASG but less so (WM 2, 6 contacts)

The L-ASG had a very strong contralateral connection to the right anterior precuneus (1 contact). Contralateral strong connectivity was also seen in the angular gyrus (2 contacts), frontal operculum (5 contacts), hippocampal head (3 contacts), inferior frontal sulcus (IFS) and the STS (2 electrodes each).

In comparison, the R-ASG also showed a moderately strong connection to the opposite frontal operculum, however, not clearly as strong (WM 2.4 but with a SD of 0.8 in 14 electrodes). The contralateral connection from R-ASG to left hippocampal head was also seen to be present but modest (8 contacts, WM 2). The R-ASG lacked sampling in the contralateral (left) precuneus and angular gyrus.

Intra and inter-insula connectivity:

R-ASG was strongly connected to the ipsilateral PSG (8 contacts). Contralateral connections were modest.

The L-ASG is most strongly connected to the ipsilateral ALG (WM 4, 3 electrodes), MSG (WM 3.2, 5 contacts), PLG (WM 3, 2 contacts) and shows strong contralateral connectivity to the R-ASG (WM 3.2, 5 contacts) and R-PSG (WM 3, 3 contacts).

The Middle Short Gyrus (MSG)

Four patients had CCEPs performed in the R-MSG, with one patient having 2 pairs of electrodes placed orthogonally, giving a total of 5 stimulated CCEP pairs. In total, 218 grey matter electrodes were analyzed in 40 anatomical regions. This was divided into 149 electrode contacts in the right hemisphere across 21 anatomical sites and 69 contacts in the left hemisphere representing 19 individual anatomical areas.

Five pairs of L-MSG CCEPs were tested across 5 patients. In total, 538 electrode contacts were analyzed from 78 regions of the brain. 92 contacts were placed in the right hemisphere across 21 anatomical regions and 446 contacts in the left hemisphere across 57 areas.

Figures 3 and 4 represent the right and left MSG.

Fig. 3.

Right MSG

Fig. 4.

Left MSG

Ipsilateral connections:

Frontal connections:

Both MSG were strongly connected to their respective frontal operculae (R-MSG: 8 contacts and L-MSG 22 contacts).

Differences in connectivity were seen in the following connections:

1. Anterior Cingulate (AC)-The L-MSG was strongly connected to the AC (6 contacts).

2. Orbitofrontal regions-the R-MSG was strongly connected to the right mesial to lateral orbitofrontal regions (8 contacts in total). The L-MSG had a descending ipsilateral connectivity strength from mesial (WM 2.7, SD 1, 11 electrodes) to lateral (WM 2, 3 contacts) and unknown contralateral orbitofrontal connection.

3. Gyrus rectus and IFG- the L-MSG was strongly connected to the following areas; Gyrus rectus (6 contacts), IFG- PO & PT (3 and 6 contacts respectively). On contrast, the R-MSG was not as robustly connected to the ipsilateral IFG PO (WM 2.7, SD 1.3, 7 contacts) and gyrus rectus (WM 2.6, SD 1.3, 4 contacts).

4. Other- The R-MSG also was strongly connected to the ipsilateral IFS (only 1 contact) and SFS (4 contacts). The status of these connections from the left MSG is unknown. The L-MSG strongly connected to the left motor cortex (13 contacts), which was not sampled on the right.

Temporal connections

Both MSG connected strongly to the temporal poles (R-MSG 8 contacts, L-MSG 4 contacts), and middle temporal gyrus (MTG) (R-MSG 6 contacts and L-MSG 27 contacts).

The R-MSG had strong connections to the hippocampal head (6 contacts, SD 0), whereas the left was weaker (WM 2.6, SD 0.5, 7 contacts) although the connection to the hippocampal tail and fimbriae were stronger (WM 4 and 3, SD 0 respectively).

Furthermore, L-MSG stimulation showed strong responses in the ipsilateral temporal operculum (planum polare and temporale 12 and 15 contacts respectively), the STG (10 contacts), ITS (5 contacts), uncus (3 contacts) superior temporal sulcus (9 contacts).

These areas were not sampled on the right.

Parietal connections

The L-MSG also had a strong bilateral connection to the superior parietal lobules, whereas the R-MSG was only modestly connected to its own side (WM 2, SD1, 6 electrodes).

The L-MSG strongly connected the ipsilateral primary sensory cortex (5 contacts). This was not represented on the right.

Other regions:

L-MSG showed strong connections to the subcortical structures (globus pallidus and putamen) and to both (ipsilateral and contralateral) claustra. The connectivity of the R-MSG to these areas is unknown.

Contralateral connections:

Differences in frontal lobe connectivity were seen in both MSG as follows: The L-MSG also showed a strong connection to the right frontal operculum (WM 3, SD 0, 3 contacts) and to the right IFG PO (WM 2.8, SD 0.98, 5 contacts) although this was not mirrored from the right (R-MSG to left FO WM 1, SD 0, 3 contacts and IFG PO WM 2,SD 0, 3 contacts).

The R-MSG was seen to strongly connect the left AC (2 contacts), but a poor connection ipsilaterally (WM 2.2, SD 0.6, 9 contacts). In contrast, the L-MSG had a very weak connection to the right AC (WM 1, SD 0, 3 contacts). The R-MSG also projected to the whole left orbitofrontal region (6 contacts mesial to lateral) and to the gyrus rectus (WM 3, 4 electrodes). A stronger connection to the left MFG (WM 3, SD0, 5 contacts) was seen from the R-MSG, than ipsilaterally (WM 2.9, SD 0.8, 16 electrodes).

Contralateral connections to the left MTG from the R-MSG were also significant (3 electrodes).

As mention, the L-MSG also had strong connectivity to the contralateral superior parietal lobule and claustrum.

Intra and inter-insula connectivity:

R-MSG was strong connected with the ipsilateral PSG (WM= 4, 4 contacts).

L-MSG showed strong connections (WM 3–4) in the ipsilateral ALG (4 contacts), MSG (11 contacts), PSG (2 contacts), PLG (12 contacts). Strong contralateral connections were seen in the right MSG (1 contact), and PSG (1 contact).

The Posterior Short Gyrus (PSG)

3 patients had CCEPs stimulated from the R-PSG. In total 253 grey matter electrodes were tested from 46 locations in the brain. 202 electrodes were found in the right hemisphere, representing 36 distinct areas. Stimulations of the R-PSG to 51 electrodes regions in the left hemisphere were present, representing 10 distinct anatomical sites. In the L-PSG, 3 pairs of CCEPs across 3 patients were recorded, spanning 252 contacts in 48 brain regions. 57 electrode contacts had been placed in the right hemisphere spanning 13 anatomical regions, while 195 electrodes where place in the left hemisphere across 35 anatomical areas.

Figures 5 and 6 represent connections from both PSG.

Fig. 5.

Right PSG

Fig. 6.

Left PSG

Ipsilateral connections:

Frontal connections:

Both PSG had strong connectivity to the frontal operculum (R-PSG WM 3.75, 20 electrodes, L-PSG WM 4, 17 electrodes) and the mesial SFG (R-PSG WM 3, 2 contacts; L-PSG WM 3, 4 contacts).

Differences between the PSG included the following areas:

1. Anterior Cingulate (AC)-The R-PSG was strongly connected to the right AC (2 contacts), whereas the left AC was only moderately well connected by the left PSG (WM 2, SD 0, 2 contacts).

2. Orbitofrontal gyrus-the R-PSG strongly connected to the mesial orbitofrontal gyrus (2 contacts), which was weaker on the left (WM 2, SD 0, 8 electrodes)

3. L-PSG to paracentral lobule (5 contacts) and left subcentral gyrus (2 contacts) was strong. The connection of R-PSG to right subcentral gyrus was present but weaker (WM 2, 2 contacts) and unknown for the paracentral lobule.

Temporal connections:

The temporal operculum is very highly connected by the PSG on both sides (Heschl’s gyrus: R-PSG WM 4, 4 contacts; L-PSG WM 3, 6 contacts. Planum temporale: R-PSG WM 4, 16 contacts; L-PSG 3.25, 16 contacts and planum polare: R-PSG WM 3, 5 contacts and L-PSG WM 4, 5 contacts).

The mesial temporal structures comprising the amygdala, hippocampus and mesial temporal pole showed strong efferents from the R-PSG (all WM 3, 3 contacts each). The L-PSG also had strong links to the hippocampus (WM 4, 2 contacts) and uncus (WM 3, 2 contacts) but only a modest connection to the amygdala (WM 2, 2 contacts). The mesial temporal pole region on the left was not represented.

The temporal neocortical areas of the STG and STS were strongly connected by the R-PSG (WM 3.3 and 4 and contacts 15 and 1 respectively) and the STG and MTG by the L-PSG (WM 3.1 and 3 and 7contacts each respectively). In contrast, the R-PSG to ipsilateral MTG was however modestly connected (WM 1.8, SD 0.8, 11 electrodes)

Parietal connections:

The R-PSG connected strongly to the ipsilateral SMG (2 contacts). The connection for L-PSG was not sampled. The L-PSG had a very strong ipsilateral connection to the ipsilateral parietal operculum (WM 4, 16 contacts), whereas the right PSG connection was not as strong (WM 2.6, 19 electrodes)

Other regions:

The R-PSG was strongly linked to the lateral occipital gyrus (7 contacts). The left was unknown.

Contralateral connections:

The R-PSG was highly connected to the contralateral (left) parietal operculum (PO) (WM 4, 2 contacts), whereas the L-PSG had a modest connection only to the right PO (WM 2, 8 contacts, SD 0). Conversely, the L-PSG had a strong connection to the right (contralateral) FO (WM 3, 5 contacts), whereas the right PSG to the left FO was weaker (WM 2, 9 contacts).

Both the PSG’s were well connected to the contralateral STG, although the right was stronger (R-PSG to left STG-WM 3, 8 contacts and L-PSG to right STG- WM 2.7, 7 contacts).

Intra and inter-insula connectivity:

Stimulation of the R-PSG resulted in the strongest connectivity (WM= 4) in the ipsilateral insula contacts comprising the ASG, ALG and PLG. A strong connection was also seen to the left PLG (5 contacts)

The L-PSG connectivity was mainly to ipsilateral ALG. Only moderate connections (WM 2) were seen to the right ASG, PLG and PSG.

The Anterior Long Gyrus (ALG)

Four pairs of electrodes, from 4 patients had CCEP stimulation from the R-ALG. 304 electrode contacts were studied from 47 anatomical regions. 292 electrode contacts highlighted 42 separate anatomical areas in the right hemisphere, while 12 electrode contacts were placed in the left hemisphere across 5 anatomical regions.

Five pairs of L-ALG CCEP responses were acquired from 5 patients, spanning 559 electrodes from 68 regions of the brain. 70 electrodes were implanted in the right hemisphere across 17 anatomical areas and 487 in the left hemisphere across 51 anatomical regions.

Figures 7 and 8 show the connectivity of both ALG

Fig. 7.

Right ALG

Fig. 8.

Left ALG

Ipsilateral connections:

Frontal connections:

Both ALG were strongly linked to the frontal opercular regions (R-ALG WM 4, 10 contacts; L-ALG WM 3.6, 15 electrodes).

The L-ALG was highly linked to the ipsilateral gyrus rectus (WM 3, 16 contacts), the precentral gyrus (WM 3.6, 13 contacts) and sulcus (WM 3 but only 1 contact). The R-ALG however, only showed a weaker connectivity to these regions (right Gyrus Rectus WM 1.7, SD 0.9, 6 contacts and right pre-central gyrus WM 2, SD 0, 6 electrodes) R-ALG strongly connected to the ipsilateral IFG pars orbitalis (POrb) (5 contacts), the subcentral gyrus (3 contacts) and mesial frontal pole (3 contacts). On the left, the connection to mesial frontal pole was much weaker (WM 1.8, SD 0.4, 13 electrodes) and those to the subcentral gyrus and IFG POrb were unknown.

Temporal connections:

Both ALG had strong ipsilateral connections to mesial temporal areas including the uncus (R-ALG WM 4, 1 contact; L-ALG WM 3.2 6 contacts), amygdala (R-ALG WM 3, 7 contacts; L-ALG WM 3.6 10 contacts) and parahippocampal gyrus (R-ALG WM 3, 3 contacts; L-ALG WM 3.4 and 8 contacts). L-ALG additionally revealed major efferents to the collateral sulcus (WM 4, 2 contacts) and fimbria (WM 3, 2 contacts) although these were not sampled from the right.

Lateral temporal contacts showed a difference between the two regions. Firstly, the L-ALG showed significant efferents to the lateral temporal pole (WM 3, 6 contacts) and the STG (WM 3.3, 11 contacts). Conversely, the R-ALG displayed a slightly less strong connection to STG (WM 2.9, 16 contacts) and very weak connections to the lateral temporal pole (WM 1, 4 contacts). The R-ALG showed strong connections to the STS (WM 4, 2 contacts), which was weaker from the L-ALG (WM 2, 2 contacts).

Both ALG were also highly connected to their respective temporal opercular regions (right ALG- planum polare and temporale (WM 3.2 and 4 and 2 and 4 electrodes respectively. L-ALG- planum temporale and polare (WM 3.2 and 3, 10 and 12 electrodes respectively). It is worth noting that the L-ALG strongly connected to Heschl’s gyrus (WM 3, 14 contacts) whereas the R-ALG connection to ipsilateral Heschl’s was weaker (WM 1.8, SD 0.4, 5 contacts).

Parietal connections:

Both ALG were highly connected to their corresponding parietal operculum (R-ALG-WM 3.6, 11 contacts; L-ALG WM 4, 27 contacts).

The R-ALG connected strongly to the ipsilateral primary sensory cortex (4 contacts). Interestingly, there was almost no connection seen from the L-ALG to this area (WM 1, SD 0, 2 contacts)

The L-ALG showed strong connectivity to the ipsilateral posterior precuneus (4 contacts). The R-ALG was also connected to precuneus (anterior) but moderately so (WM 2.4, SD 0.5, 5 electrodes)

Other areas:

The L-ALG was strongly connected to the ipsilateral claustrum and globus pallidus. These connections were unknown from the right.

Contralateral connections:

R-ALG connected to the left mesial orbitofrontal region and L-ALG to the right inferior frontal sulcus (both areas single contacts). The left ALG was also strongly connected to the right ASG (WM 3, 2 contacts).

Intra and inter-insula connectivity

CCEP analysis in the R-ALG revealed highly connected areas in the right (10 contacts). L-ALG connections were best in ipsilateral ASG (WM 4, 3 contacts), MSG (WM 3.8, 10 contacts) and PSG (WM 3, 3 contacts).

The Posterior Long Gyrus (PLG)

Six pairs of CCEPs from 6 patients were recorded from the R-PLG. In total, 556 contacts from 64 anatomical regions were analyzed. 515 electrode contacts were located in the right hemisphere in 52 areas and 41 electrodes in the left hemisphere in 12 anatomical regions.

Six pairs of L-PLG CCEPs were performed in 5 patients, spanning 648 electrodes over 77 regions of the brain. 70 electrodes in the right hemisphere in 14 anatomical areas and 578 electrodes were located in the left hemisphere across 63 anatomical areas. Figures 9 and 10 outline the full connectivity profiles from both these regions.

Fig. 9.

Right PLG

Ipsilateral connections:

Frontal connections:

Both PLG demonstrated strong ipsilateral connections to the primary motor region (R-PLG WM 3.6, 7 contacts to the right precentral gyrus. L-PLG WM 3, 4 contacts to the precentral sulcus. Of note, no precentral sulcus contact was available on the right, and the precentral gyrus contact on the left showed a weaker connection (WM 2.2, SD 0.6, 18 contacts).

The R-PLG also had a strong connection to the frontal operculum (WM 4, 13 contacts), although the left PLG wasn’t as strongly connected (WM 2, SD 1.0, 25 contacts).

The L-PLG was well connected to the IFG-P Operc and IFG-PT (WM 3.4 and 3.2 and 8 and 10 contacts respectively) with poor connectivity to IFG POrb (WM 1, 3 contacts). In contrast, the R-PLG showed strong connectivity with IFG-POrb (WM 3.8, 14 electrodes), moderate connectivity with IFG-PO (WM 2.3, SD 0.7, 25 electrodes) and very poor connectivity with IFG PT (WM 1, SD 0, 6 electrodes).

The R-PLG was well connected to the ipsilateral IFS (WM 3, 2 contacts), but the L-PLG was a weak connection in only 1 representative contact (WM 2)

Temporal connections:

Both PLG shared similar qualities being highly connected to the temporal operculum (R-ALG- WM 3.8–4, Heschl’s gyrus (9 contacts), planum polare and temporale (10 and 20 contacts respectively)); L-ALG- WM 3.2–4, planum temporale (21 contacts), Heschl’s gyrus (12 contacts) and planum polare (17 contacts)

Both regions also demonstrated links to certain mesial temporal areas but to varying degrees. Strong links were seen from the R-PLG to the ipsilateral amygdala (WM 4, 10 contacts), uncus (WM 4, 1 contact), parahippocampal gyrus (WM 3.8, 4 contacts), entorhinal cortex (WM 3, 1 contact) and the hippocampus (WM 3.3, head and tail 23 contacts total). However, the L-PLG only showed a strong connection to the tail of hippocampus (WM 3.5, 8 contacts) and the more anterior mesial temporal connections were weaker (hippocampal head (WM 2.5, 4 contacts, SD 0.8), amygdala (WM 1.8, 13 contacts), uncus (WM 1.5, 8 contacts), parahippocampal gyrus (WM 1, 1 contact)). Both PLG also strongly connected to the ipsilateral STG (R-PLG WM 3., L-PLG WM 3,15 contacts each). The R-PLG also highlighted strong efferents to the STS (WM 3, 15 contacts), while the left side was less well connected to the this region (WM 2, 9 contacts) The L-PLG strongly connected with the temporal pole (WM 4, 16 contacts in total, stronger connections lateral to mesial), ITG and ITS (WM 3 and 4, 9 and 5 contacts respectively). In comparison, connection to the ITG from the R-PLG was modest (WM 2.1, 21 contacts) and a poor to the temporal pole (WM 2 for mesial part and WM 1 for the lateral part in total of 4 contacts)

Parietal connections:

Both PLG were also highly connected to the parietal opercular regions (R-PLG WM 3.2, 9 contacts and L-PLG WM 3.2, 27 contacts).

The R-PLG was strongly connected to the ipsilateral sensory cortex (WM 3, 7 contacts). The L-PLG showed strong connectivity to the left postcentral sulcus (WM 3, 3 contacts) but a weaker connection to the postcentral gyrus (WM 2.2, 22 contacts,)

The L-PLG was strongly connected to the supramarginal gyrus (WM 2.6, 25 contacts), whereas the right PLG was moderately well connected (WM 2.6, SD 1.1, 19 electrodes). Connectivity of the L-PLG to the inferior precuneus was strong (WM 3, SD 0, 4 electrodes) but weaker from the R-PLG (WM 2, 2 contacts)

Other regions

Both PLG were also highly connected to the claustrum (WM 4, 2 contacts from right and 4 from the left)

The L-PLG had very strong bonds (WM4) with the subcortical areas (globus pallidus and putamen), representation of these areas on the right was absent.

The left PLG connected strongly to the paracentral lobule (WM 3.1, 11 contacts), whilst the right PLG had a weaker connection to the right paracentral lobule (WM 2, 6 contacts).

Contralateral connections:

The R-PLG was the highly connected to the contralateral lingula (WM 4, 2 contacts, SD 0), the temporal regions (collateral sulcus (WM3, 4 contacts), ITG (WM 3, 3 contacts), temporal pole (WM 3, 2 contacts)), including the temporal operculum (planum temporale-WM 3, 5 contacts). Contralateral connections from the L-PLG to the right planum temporale, STG and STS were weak.

Interestingly, the left PLG had a bilateral projection to the frontal opercular regions with an equally connected moderate value (WM 2).

Intra and inter-insula connectivity:

Strong intra-insula contacts from the L-PLG were seen in the ipsilateral ALG, MSG and PSG. Ipsilateral intra-insula connections from the R-PLG, were weaker and included the right ASG (WM 2.6, 9 contacts) and the ALG (WM 2, 1 contact). Contralateral L-PLG to the right MSG was present but weak (WM 1).

Contralateral insula connections from the right were not sampled.

DISCUSSION

CCEPs of the individual gyri in the human insula in both hemispheres, highlight the extensive and complex networks arising from each sub-region of this cortical area. These efferent connections explain why the insula is a major integrative hub and so highly involved in a large number of important functions.

Language function and the Insula

The insula is increasingly being recognized as having a major role in the language network [33–36]. Cortical regions involved in language function are complex and involve two major modalities (i) regions for speech articulation and motor control-including Broca’s region (IFG pars opercularis and triangularis) and the primary motor cortex and (ii) language association areas, comprising widespread cortical regions including the temporal pole, primary auditory cortex, fusiform gyrus, supramarginal gyrus, angular gyrus and the superior, middle and basal temporal gyri [37] and the SMA [38].

Our work shows strong connectivity to both language motor and association areas, from the dominant insula. What is striking is the finding of maximal language efferents from each opposing pole of the insula, meaning that the ASG sends long efferents to the posterior brains regions and the PLG conversely, connects more anterior brain regions. The ASG and PLG both relay connections to the basal temporal language areas (ITG and fusiform) and parietal language areas (angular gyrus and supramarginal gyrus).

Speech articulation and motor control have been shown to involve the precentral gyrus of the insula [33,39–43] as well as the recognized inferior frontal gyrus (Broca’s) and primary motor area. Cortical stimulation studies in the insula, resulting in speech disturbances, have also helped confirm this [44–45]. The obtained CCEP results provide insight into this mechanism, showing that the left ASG and MSG are strongly connected with IFG-pars opercularis and with the primary motor cortex. However, we further report involvement of the posterior insula. The ALG also connects to the primary motor cortex and along with the PLG, both are strongly linked to IFG pars opercularis, and pars triangularis. These regions also show strong connectivity to the primary auditory cortex, a key language association area that modulates functional connectivity during motor speech [46]. This posterior insula-auditory-speech motor network that we describe, is implicated in studies of deaf individuals, utilizing lip-reading and articulatory-based cues, which show grey matter hypertrophy in the posterior insula [47].

The insula PSG also shows a similar connectivity pattern. Although it is well connected to the auditory cortex and Broca’s areas, it is conversely not strongly connected to the primary motor cortex. This may explain the recent study, arguing against the specific role of the precentral gyrus of the insula in articulation [48]. Rather it suggests that the role is in fact a modulation of articulation in combination with the IFG pars operculum, a role that is also shared by the SMA [38]. Our own work shows that the PSG is strongly connected to these regions too.

Language association regions are also highly connected by the dominant insula, with the lateral temporal language areas receiving strong projections from the importation of the insula namely MSG, PSG and ALG. The temporal pole is also highly connected from both anterior and posterior insula regions (ASG, MSG, ALG and PLG), which correlates well with primate studies, where these strong connections are attributed to a limbic integrating system for complex, preprocessed visual and auditory information with the insula [16].

Stimulation findings have previously reported speech disturbances arising from both insula, particularly during stimulation of the MSG [45]. Our study may explain these findings to some degree given that the non-dominant insula projects to contralateral language areas (namely R-ASG to left pars triangularis; R-MSG to left MTG, R-PSG to left STG and R-PLG to the left temporal pole). We observe that while the right MSG has moderate connections to left language networks, it is also highly connected from the left MSG. The right anterior insula has shown to be an important mediator in speech and language function [36] but also strongly activated during non-lyrical singing (non-lyrical tune) [49]. While we did not find strong intra-insular connectivity from the right to left MSG, the right MSG was still fairly well connected to the ipsilateral Heschel’s gyrus and primary motor cortex. We hypothesis that in situations such as non-lyrical singing, the left MSG may enhance the right, or even suppress the left, in order to engage non-verbal language function.

We report that the intra-insular connectivity of the dominant insula appears to have a “closed-loop” like connectivity. The anterior insula structures are highly connected in a “downstream-upstream” gradient (ASG to MSG, ALG, PLG; MSG to PSG, ALG, PLG; PSG to ALG, PLG; ALG to all and PLG to ALG, PSG, MSG). It remains to be elucidated how and if such connectivity has a role to play in language function.

Sensation: Pain, Interceptive awareness and somatosensory afferents

Multimodal information processing networks across sensorimotor, emotional and homeostatic domains are increasingly being attributed to the insula [50–51]. Although some studies suggest an asymmetry in function, with the right/non-dominant insula being attributed to negative affect, hunger survival, avoidance behaviors [52], cardiac control [53] and also in pain [54], such lateralization of valence is not so clearly defined at present [55–56]. Our own study, however, reveals an asymmetry in connectivity. The right insula, at least for somatosensory function, has more ipsilateral and contralateral connections than the left. This may be due to earlier development of the right hemisphere, making developmental connections more complete [50], especially in the context of patients with epilepsy, where pathological activity may impair plasticity, thus our results may be bias with respect to lateralization.

Sensory integration appears to follow a hierarchy within the insula in a “bottom-up”/posterior-to-anterior axis, whereby sensory inputs to the posterior insula are propagated to the mid-insula region for homeostatic processing and then onto the anterior regions for saliency processing [54,57–60]. Our research supports the existence of this axis to a degree; especially as both posterior insula regions (ALG and PLG) are highly connected to the ipsilateral anterior gyri. We also demonstrate that the left posterior intra-insula contacts are more widely connected to their ipsilateral short gyri than on the right. The left ALG is also strongly connected to both the ASG. These additional connections from the left may provide an explanation as to why the left hemisphere shows dominance for positive emotional stimuli in the anterior and mid insula regions and during the perception of emotion in others, despite both posterior insula regions being activated in both cases [54]. Having developed later in embryological terms, the left insula may be responsible for higher cognitive situational assessment, and thus support the notion that the right insula is a default network concerned mainly with “negative” emotional processing (hunger, avoidance, pain) to ensure species survival [53]. Further evidence comes from the findings that the right anterior insula becomes activated by emotionally aversive stimuli and harm avoidance behavior [61] and that dysfunction of this system may be the key in everyday anxiety disorders [62].

Pain processing is also a role ascribed to the insula along with the anterior cingulate region (AC), the parietal operculum (SII), the primary sensory cortex (S1) and the limbic structures including the amygdala, hippocampus and orbitofrontal cortex and precuneus [63–65]. Of these regions, the AC is thought to be more specific to pain processing [66–68] compared to SII, where historical stimulation failed to produce nociceptive experiences [69].

Specifically, the MSG has been implicated in pain processing with stimulation studies [70]. The CCEP studies support this finding given that both MSG are highly connected to the left AC with additional cross connectivity from the right ASG. The right AC appears to be less well connected by the insula receiving inputs mainly from the right PSG. Connectivity of the right anterior insula to the AC has been shown to be important in bladder control and urgency [71]. Another CCEP study did not find evidence of connections from the insula to anterior cingulate [29].

Insula to cingulate connections involve Von Economo neuronal (VENs) networks, specialized neurons present in the anterior cingulate and frontoinsular cortices [72], which are more abundant on the right side [73]. If we extend our findings, of the dual efferent innervation of the left AC, further afield, in neurology, it may provide addition insight into the neurological illness of fronto-temporal dementia (FTD). In this disorder, decreased VEN densities are a pathological marker [74] but only patients with right-sided FTD present with behavioural disturbances [75]. This is presumably since the right AC only receives a paucity of ipsilateral insula connections, resulting in a frontal disconnection syndrome, whereas the left is relatively protected by the benefit of bilateral innervation.

The right and left MSG and PSG also show strong connections to mesial temporal and frontal limbic areas as well as SI and SII as well as the novel finding of the right PSG connecting to the contralateral SII, which supports the hypothesis that the mid-posterior insula is in fact a multimodal homeostatic or interoceptive area [76].

We highlight the strong connectivity of both posterior insula regions to S1 and SII, confirming their role in the sensory network. This was also demonstrated through fMRI [77]. Significant connectivity of the posterior insula regions to their respective amygdala and hippocampi were seen, but with additional right posterior insula cross-connections to the contralateral left orbitofrontal region. Studies in anger responses have shown increased activity in the right posterior insula and STG and to a lesser degree the left posterior insula and hippocampus [78].

We hypothesize that the posterior insula may itself have independent emotional sensory processing capabilities rather just a relay to the anterior insula for processing.

The precuneus has also shown to be part of the sensory network, especially in visuo-tactile texture matching in fMRI studies in healthy volunteers [79]. The precuneus and mesial limbic regions are connected via the insula, although only one study has demonstrated this in humans [80]. Our research demonstrates this connection of the left posterior insula (ALG and PLG) and left ASG being strongly connected to the ipsilateral precuneus and also the mesial limbic networks.

Interoceptive awareness (conscious awareness of inner bodily signals) [51] is mediated by the right anterior insula to higher-order areas of visceral representation such as the orbitofrontal cortex [81–82]. The CCEP data highlight the right MSG and PSG efferents to the ipsilateral orbitofrontal cortex and the novel finding that the right anterior (ASG) and posterior insula (ALG) are also highly integrated with the left (contralateral) orbitofrontal region. The left insula is not as well connected to either orbitofrontal area and this may explain why previous CCEP studies may have failed to show insula-orbitofrontal connectivity [29] or that only the left ASG has some connection to the orbitofrontal regions [83].

The insula-orbitofrontal networks are attributed to processing the final evaluation of reward or punishment implicated by a certain stimuli [84]. An example of this network at play comes from the sensation of taste. A recent fMRI study implicated the left MSG and ALG as the determinants of pleasantness of taste, and the right ALG and PLG being involved in taste concentration processing, in a male-only cohort [85]. Our connectivity study confirmed the findings of insula-orbitofrontal connectivity to be present, but weak, from the left and absent from the right. However, a partial explanation may be due to our cases being of mixed sex.

Vision

Few studies in the human insula have reported direct links between the visual cortex and insula. Recently, high-resolution tractography studies have reported, for the first time in humans, ipsilateral connections between the insula lobe and the visual cortex as well as parietal and temporal visual association areas (supramarginal and angular gyri, fusiform gyrus, cuneus and precuneus) [80]. The insula, with respect to visual function, is thought to act as a cross modal cortical integrator for stimulus perception [86–87] as well as visual consciousness [88]. The latter also requires parietal integration (precuneus and lateral parietal cortex) as part of the default mode network [89].

Our study supports such findings and adds a novel dimension to the literature. Firstly, it appears that connectivity from the left insula, to visual association areas is more ubiquitous, although this may be due to limited sampling between insula and occipital regions within our database. The ASG on both sides were found to be more highly connected to structures implicated in the ventral visual stream [90], including the fusiform gyrus (left ASG), the inferior temporal gyrus (left ASG), the angular gyrus (both) and the intra-parietal sulcus and supra-marginal gyrus (right ASG respectively). This may explain why dysfunction of the anterior insula has recently been implicated in migraine with aura [91]. The right posterior short gyrus also shows strong connectivity between the lateral occipital cortex and supra-marginal gyrus with respect to the visual integration from the right insula.

One finding, unreported to date, is the strong contralateral connectivity from the right posterior long gyrus to the left lingula and inferior temporal gyrus, with weak ipsilateral connections to these same areas. These findings may provide clues in the specific role of the right PLG in visuospatial navigation, namely that activation of visual cortex and motion-sensitive areas (such as the inferior temporal gyrus) are associated with a deactivation of the right posterior insula [92], whereas in the blind, the opposite is true and imagined locomotion activates the posterior insula [93], which may act as an alternative sensorimotor region when vision is impaired or limited (e.g. navigating at night)

Hearing:

The insula is an integral part of the auditory system, along with the primary auditory cortex (Heschl’s Gyrus), the planum temporale and lateral superior temporal gyrus [94]. Connectivity studies in monkeys support the involvement of the posterior insula as an important auditory region [95]. Our CCEP findings confirm strong connections to Heschl’s gyrus from the right (PSG and PLG) and left insula (PSG, ALG, PLG). Interestingly, the right ALG was not well connected to the ipsilateral Heschl’s and left sided insula connections appeared more extensive when compared with the right. Planum temporale was strongly connected ipsilaterally from both mid-posterior insula regions (PSG, ALG and PLG on the right and MSG, PSG, ALG and PLG) and the lateral superior temporal gyrus as similarly connected by the posterior long gyri on both sides and additionally by the posterior short gyrus on the left. Impairment of these mid-posterior insula connections has previously been implicated in music aversion in dementia [96].

A novel finding in our CCEP research is the finding of connectivity from both PLG to the contralateral Heschl’s, in both hemispheres, along with MSG from the right and PSG from the left. This may explain the reported findings of contralateral ear hypoacusis during posterior insula stimulation [64].

Furthermore, the anterior insula showed weaker connections to Heschl’s but stronger connections to secondary auditory areas, suggesting they may have an integrative role in auditory processing, as previous suggested in stroke studies of the insula [97].

Intra and inter-insula connectivity:

Connectivity between the two insula lobes has been reported to occur from both anterior and posterior regions in animal studies [98–99] but is sparsely reported in humans. Using CCEPs one study in humans, could not find proof of connectivity between the posterior insula regions [29]. In another single case report, utilizing latency as a connectivity marker, the anterior insula regions on both sides were found to be linked (right PSG to left PSG and left MSG and left PSG to right PSG) [28]. The same study could not however confirm contralateral connectivity from the posterior insula.

Our own results highlight strong cross-connectivity mainly from the left ASG, MSG and ALG and only moderate connections from the right ASG and PSG. The PSG appear to be the major hubs between anterior and posterior insula regions within each respective insula. Additionally, the polar opposites of each insula do show connectivity (ASG to PLG and vice versa) although this is only a moderately strong connection within the right insula but a stronger connection from the left.

Limbic connectivity and epilepsy

Surgical failure in patients diagnosed with temporal lobe epilepsy is increasingly being attributed to a failure of recognition of the insula as an epileptogenic zone due to the significant connections to the mesial temporal regions [25,69,100]. Our own findings confirm the significant connections to the mesial temporal structures from all gyri of the right insula. The left insula similarly showed full connectivity to the hippocampus but the amygdala received connections from the ASG, PSG and ALG only. The left ASG was also strongly connected to the right hippocampus. This may explain why patients with left sided temporal lobe epilepsy exhibit greater white matter damage [101], especially as both hippocampi are highly connected to their ipsilateral insula [6].

Structural imaging studies have shown strong connectivity of the amygdala and temporal pole by the anterior insula in the human and macaque monkey [14,19,20,101]. Conversely, prior CCEP studies in humans showed the reverse situation with connectivity of the amygdala with posterior but not anterior insula connections [29]. Similarly, this study showed hippocampal connections from all the insula except PSG. In our results, the right temporal pole was connected by the right ASG and MSG while the left temporal pole received connections from the ASG, MSG, ALG and PLG. Cross connectivity from the right PLG was also seen to the left temporal pole. The strong connectivity to this region from both insula may explain why it is also implicated in language and naming [102].

Vestibular

The insula has been implicated in vestibular function, particularly the right insula. Some studies have implicated both anterior and posterior insula regions [103], whereas others have implicated the poster-superior part in conjunction with the adjacent parietal opercular region, termed the parieto-insula vestibular cortex [104–105]. Cortical stimulation studies in epilepsy patients undergoing SEEG have confirmed vestibular sensations in the posterior insula [106]. While the extended vestibular network is still not well established [105], cortical stimulation studies in humans have also implicated the superior and middle temporal gyri, inferior parietal lobule, intraparietal sulcus and angular gyrus [107–110]. This multi-lobar involvement has also been confirmed by animal studies [111–112]. Neurons comprising the vestibular cortex are multisensory incorporating optokinetic and somatosensory stimuli [113].

Our results show connectivity to the vestibular network mainly from the right insula with bilateral connectivity from the right anterior insula (ASG, MSG and PSG). What is particularly pertinent is the finding that both posterior long gyri are well connected to the STG and parietal opercular regions and that it is the right PSG, which connects the STG and parietal opercular regions on both sides. Vestibular stimulation, during fMRI, highlighted increased neural activity in the posterior insula and STG [114], but conversely during visual stimulation of a rotating stimulus bilateral insula blood flow decrease was observed on PET [115]. The right PSG also has contralateral connections to the left PLG, which in turn has strong connections to most of the left insula. As the PSG has been shown to have afferents from all types of vestibular receptors [116], we propose that it may be a major hub for sensory switching reciprocal inhibition, allowing for a change in the dominant sensory system, depending on the prevailing sensory mode.

CONCLUSION

CCEPs of both human insulae efferents confirm this structure to be a major multi-modal network hub within the cerebral cortex. The sheer amount of connections explains the complex neurophysiological function and that it is a key region involved in both functional and effective connectivity.