Functional connectivity of sensorimotor network before and after surgery in the supplementary motor area

After resective glioma surgery in the Supplementary Motor Area (SMA), patients often experience a transient disturbance of the ability to initiate speech and voluntary motor actions, known as the SMA syndrome (SMAS). It has been proposed that enhanced interhemispheric functional connectivity (FC) within the sensorimotor system may serve as a potential mechanism for recovery, enabling the non-resected SMA to assume the function of the resected region. The purpose of the present study was to investigate the extent to which changes in FC can be observed in patients after resolution of the SMAS. Eight patients underwent resection of left SMA due to suspected gliomas, resulting in various levels of the SMA syndrome. Resting-state functional MR images were acquired prior to the surgery and after resolution of the syndrome. At the group level we found an increased connectivity between the unaffected (right) SMA and the primary motor cortex on the same side following surgery. However, no significant increase in interhemispheric connectivity was observed. These findings challenge the prevailing notion that increased interhemispheric FC serves as the only mechanism underlying recovery from SMA syndrome and suggest the presence of one or more alternative mechanisms.


Introduction
The supplementary motor area (SMA) is located in the superior frontal gyrus anterior to the primary motor area and dorsal to the cingulate gyrus (Penfield and Welch, 1951;Ibe et al., 2016).It appears to be important for the initiation and planning of voluntary, non-stimulus driven, movement and speech in humans.Evidence for this comes from three different lines of research.First, neurophysiological data, such as the readiness potential initially described by Kornhuber and Decke (1965), (Kornhuber andDeecke, 1965, 2016) indicated an increased activity above the dorsal medial frontal cortex where the SMA is located, prior to the initiation of non-stimulus driven motor movements and speech (Libet et al., 1983;Orgogozo and Larsen, 1979;Eccles, 1982).Second, direct electrical stimulation of the SMA has been shown to promote urges to move limbs in the contralateral part of the body and at increased stimulation intensities, initiation of movement that is subjectively perceived as voluntary (Fried et al., 1991).Finally, neurosurgical resections of the SMA during epilepsy and/or glioma surgery may lead to a transient, dramatic disturbance of voluntary action and speech known as the SMA syndrome (SMAS).(Laplane et al., 1977;Kasasbeh et al., 2012;Rajshekhar, 2000;Sjöberg, 2024).
Resolution of the SMAS is apparent approximately two weeks on average but can take up to three months (Zentner et al., 1996).According to the prevailing theory recovery occurs through alterations in functional connectivity, (Pinson et al., 2022;Otten et al., 2012) such that the sensorimotor regions of the affected hemisphere exhibits elevated functional crosstalk with the contralateral intact SMA through the corpus callosum (Oda et al., 2018).Resting-state functional connectivity analyses has provided valuable insight about functional alterations during manifestation of the syndrome, but also after recovery (van Meer et al., 2010;Fox and King, 2018;Cargnelutti et al., 2020;Hart et al., 2017).Results from one such study revealed a large-scale reorganization, such that decreased inter-hemispheric connectivity of lesion side motor area and contralateral SMA postoperatively was followed by elevated functional connectivity during the recovery process, consistent with the classical theory (Vassal et al., 2017).Similarly, results from task-based fMRI have revealed elevated activation in the contralateral SMA during both motor (Orgogozo and Larsen, 1979;Krainik et al., 2004;Sailor et al., 2003) and speech tasks (Quirarte et al., 2020;Chivukula et al., 2018).
However, functional re-organization within the sensorimotor network, which encompasses increased inter-hemispheric connectivity, may not be the sole pathway toward recovery from the SMAS.An alternative perspective on the role of the SMA suggests that it is not solely responsible for initiations and sustaining of motor activity (Laplane et al., 1977).Instead, it might be better understood as part of a broader network of brain regions that collectively generate and control voluntary actions (Haggard, 2017;Sjoberg, 2021;Lau et al., 2004).One observation in support of this view is that recovery from the syndrome has been observed in a patient with a complete agenesia of the Corpus Callosum, which may be indicative of mechanisms less dependent on interhemispheric connectivity (Obaid et al., 2022).Furthermore, injury to the frontal aslant tract, a fiber bundle connecting the SMA with the ipsilateral inferior frontal gyrus, (Dick et al., 2019;Catani et al., 2012) has been found to be associated with an increased likelihood of developing the SMAS following injury to the dorsal medial frontal cortex (dMFC), which is the anatomical location of the SMA, (Young et al., 2021;Briggs et al., 2021) and might reflect aberrant connectivity between the sensorimotor network and frontal cognitive control networks (Dick et al., 2019).This finding suggests that connectivity of the sensorimotor network to cognitive control networks may serve as a foundation for recovery from the SMAS.The possibility aligns with observations highlighting impaired cognitive control and working memory concomitant with the occurrence and resolution of the SMAS (Nakajima et al., 2014;Sjöberg et al., 2019;Canas et al., 2018).
In the present study we investigated resting state functional connectivity pre-and post-glioma surgery in the left SMA in a clinical sample of eight patients with different levels of the SMAS.The primary aim was to investigate the extent to which the resolution of the SMAS can be attributed to the intact SMA assuming the role of the resected SMA in interaction with the primary motor cortex (M1) within the affected hemisphere.More specifically we tested whether increased interhemispheric connectivity between left M1 and the intact right SMA can be manifested after surgery.Second, we investigated changes in between-network connectivity of the sensorimotor network with a cognitive-control network.

Study design
This study is a single center clinical observational study of eight patients affected by the SMAS after resection of the left SMA due to glioma.Resting State functional Magnetic Resonance Imaging (rsfMRI) data were obtained on two occasions and was done in conjunction with clinically relevant structural MRI sequences and task fMRI (mapping of motor function, speech function and speech lateralization).The first MRI session took place before surgery and the second after remission of the SMAS (M = 9.62 month, SD = 9.80 month).

Patients
Eight patients, consecutively operated in the left SMA between 2015 and 2018 by the last author (RLS) at the University Hospital of Northern Sweden and who developed a full or partial postoperative SMAS were included in the present study.In seven cases the indication for surgery was resection of a presumed low-grade glioma (WHO grade II) and in one case a suspected high grade (WHO IV) glioma.These glioma grades were all postoperatively confirmed by pathological anatomical diagnosis.Two of the surgeries (P1 and P8) were reoperations due to recurring glioma growth.Seven of the surgeries were performed with mapping of Motor Evoked Potential from the primary motor area in full anesthesia in accordance with previously established procedure (Zentner et al., 1996).One of the surgeries (which was a reoperation) was performed without such monitoring.Clinical characteristics, neuropsychological function, and quality of life for this cohort of patients has previously been reported elsewhere (Sjöberg et al., 2019;Stalnacke et al., 2020).
Clinical characteristics of the patients are presented in Table 1.The patients had a mean age of 49.56 years (Range = 39.47-69.65).Five of the patients were female and three were male.
The SMAS was assessed with a neurological examination preoperatively and daily screenings of speech and motor function during the patient's recovery at the neurosurgical ward postoperatively.Before the rsfMRI was acquired at follow-up the SMAS was determined to be in remission by the surgeon during a routine neurological examination that occurred at a 3-month follow-up appointment.
Hand motor function was assessed with a writing task (WAIS-IV symbol-digit) (Wechsler and Assessment, 2008) using the dominant (right) hand, contralateral to the affected hemisphere.Here coding is a measurement of the speed needed to copy symbols in designated locations.Speech was assessed with letter fluency (D-KEFS, phonemic fluency).This test assesses a form of verbal fluency defined as the number of words starting with a particular letter that can be verbally listed by the subject in a set amount of time (Delis et al., 2001).Motor and speech testing were administrated pre-and postoperative follow-up, adjoining time with rsfMRI.Speech function was additionally assessed during SMAS, 1-2 days after surgery (Table 1).

Informed consent
Surgeries were performed on clinical indication.Patients gave informed oral and written consent to participate in this study.The study was approved by the regional ethical committee at Umeå, Sweden (Dnr: 2016/479-31 and Dnr: 2018-402-32M).
Before image acquisition ten dummy scans were collected and discarded.The patients were instructed to stay awake and look at a cross presented on a display (white cross on black background).

Resting-state data analysis
Structural and functional images were preprocessed using Statistical Parametric Mapping software (SPM12; Wellcome Department of Imaging Science, Functional Imaging Laboratory, University College London) using an inhouse program, DataZ.Structural images were segmented into grey matter (GM), white matter (WM), and cerebrospinal fluid (CSF) likelihood maps in native and DARTEL (Ashburner, 2007) space.
The rsfMRI were corrected for slice-time acquisition, movements by realigning to the first volume in series followed by unwarping to account for non-linear movement effects.The rsfMRI were coregistered to the corresponding structural image for each subject and time point.Then, a subject-specific template followed by a group template were created using DARTEL on GM/WM/CSF.(Ashburner, 2007) rsfMRI series were normalized into DARTEL space in a two-step-dartel manner, using flow-field maps from the templates.Finally, all rsfMRI were affine-aligned into the Montreal Neurological Institute (MNI) space, smoothed by convolving with an isotropic gaussian curve with 8 mm full width at half maximum.Post filtering of the files included global-WM, global-CSF, movement parameters, and its squares and derivatives, (Friston et al., 1996;Yan et al., 2013) and a bandpass filter of 0.008Hz-0.1Hzwas applied.A group image was created in MNI space by normalizing structural images in the same way as the functional images and then calculating a mean image.
Independent Component Analysis (ICA) was applied to the preprocessed rsfMRI using the group ICA fMRI toolbox (GIFT v.3.0b)(Allen et al., 2011;Calhoun et al., 2001).ICA identifies spatially independent but temporally coherent components of the brain and has been frequently used in previous rsfMRI studies.We applied ICA on both the preoperative and postoperative rsfMRI data (as different sessions; for details see Salami et al., 2016(Salami et al., 2016)).Using minimum description length (MDL), ICA identified 47 components, 17 of which were deemed to be resting-state components (for criteria selecting components, please see Salami et al., 2014, 2016(Salami et al., 2016;Salami et al., 2014)).Given the a priori research question, we identified a single sensorimotor functional network (Fig. 2A) with nodes in right SMA (blue) and left and right M1 (red, green) and two cognitive-control networks: the left and right frontoparietal networks with nodes in left and right dorsolateral prefrontal cortex (DLPFC) (blue, yellow) and parietal lobes (red, green) (Fig. 2B).
To identify components overlapping with the area that was resected during surgery, spatial correlation was conducted between all resting state components and a binary mask delineating the resected tissue (Fig. 1).The mask was manually traced to the outline of the resection cavity on the post-operative T1 image in native space for all eight patients using itk-SNAP 3 (Yushkevich et al., 2006).The mask for all patients were then normalized to the group template and compiled to an aggregate mask with gradation from 1 to 8, depending on how many patients had resection in each voxel.Spatial correlation between resection area and ICA-components were very low (with the largest r = 0.1), suggesting that no component manifested spatial overlap with the resection area.
We restricted our search space in a two-step fashion to identify regions in the unaffected right SMA.A right SMA area was defined by selecting an area in the SMA region (from the sensorimotor network) which showed spatial overlap between pre-and post-operative.We also identified areas in the left M1 (LM1) by the overlap between the sensorimotor network and the left precentral mask from an AAL mapping, (Tzourio-Mazoyer et al., 2002) right M1 (RM1) by the overlap between the sensorimotor network and the right precentral mask from AAL.
Clusters in the dorsolateral prefrontal cortex (DLPFC) from left and right frontoparietal networks, identified by ICA, were selected after thresholding the two networks.Parietal areas in left and right parietal lobes were selected using the same method (Fig. 2B).
Using these ROIs, further analyses of correlation within the sensorimotor network and the frontoparietal network as well as between RSMA and nodes in the frontoparietal network were performed to be able to investigate possible changes in crosstalk between nodes.
Since the present study was concerned with surgery induced changes between pre-and post-operative scans we also wanted to establish nonsurgery related changes in relation to unspecific factors in non-operated controls for comparison and to evaluate test-retest stability.For this purpose, the two-step method described above was applied in a sample of 221 healthy patients to identify right and left M1.The rsfMRI images from the healthy sample were obtained from the Betula-project (Nyberg et al., 2020).The data were gathered at Betula timepoint 5 and 6, separated by four years, from the same site and scanner, using the same Symptoms at follow up for P6 was characterized by a severe right-arm paresis and a minor paresis of the right leg.Speech lateralization was determined by fMRI.+1 indicates a strong left-side dominance and − 1 a strong right-side dominance.≥ 0.2 = left-side dominance, ≤-0.2 = right-side dominance.The method is described in Sjöberg et al., 2019(Sjöberg et al., 2019).Speech and motor function results are standardized, M = 10, SD = 3. Scores of speech and motor function are standardized; M = 10, SD = 3.
The two-step method for identifying ROI:s in the sensorimotor network was compared to MNI-space coordinates for S1 (− 56 -8 32; 52 -6 30) in both the SMA-and Betula group.
To assess the test-retest reliability for the two-step method used to identify the sensorimotor and frontoparietal network, the pre-and postoperative correlation between functional network nodes in SMA-  patients, as well as the healthy Betula group, was analyzed by means of intra class correlation (ICC), utilizying ICC-package (Ver 1.3.1.0)made by Arash Salarian in Matlab with ICC of type "C-1".Outliers (>+3.29sd) were removed from the Betula sample (two participant were removed from the M1 ICC and one was removed from the S1 ICC).General signal reduction was applied.
The results of these analyses showed correlations of 0.332 for controls and 0.245 for the study group for left and right S1.That is, how strong the correlation was between the pre-operative correlation between left and right S1 and the post operative correlation between the same nodes.For M1 the correlation for controls were 0.34 whereas it was only 0.002 for the study group.This suggests results for our sample that were on par with the larger sample for areas not affected by the surgery (S1) but, as could be expected, clear changes in the study sample as compared to controls for an area involved in a network immediately affected by the surgery (M1).

Statistical analysis
Pearson correlation was used to quantify connectivity between the areas identified by the ICA analysis.A one-tailed Paired Wilcoxon Signed ranked test was utilized to compare pre and postoperative rvalues.P-value consideration: p < 0.05 was considered significant for Paired Wilcoxon Signed ranked test.
At the individual level, the functional connectivity between RSMA-LM1 of four patients (P2, P3, P4 and P6) decreased, whereas four patients revealed elevated connectivity (Fig. 3A and Table 3).Three of the four patients with a decrease in RSMA-LM1 connectivity also were found to have a decrease in LM1-RM1 connectivity (P6 had unchanged connectivity LM1-RM1) (Fig. 3B and Table 3), suggesting the possibility that remission of SMAS in these patients was not dependent on an increase in interhemispheric connectivity between either M1 to M1 or from LM1 to RSMA.All four patients with an increased RSMA-LM1 connectivity after surgery also had increase in LM1-RM1 connectivity.
We next, in an exploratory attempt, assessed whether there were any changes in functional connectivity of the parietal cortex to the M1 and found no changes between sessions.

Discussion
The main purpose of the current study was to utilize rsfMRI to investigate potential network reorganizations that may account for recovery from the SMAS.Our findings provide only partial support for the hypothesis that the contralateral SMA, via its increased connectivity with M1 from the lesion hemisphere, plays a role in the recovery from the SMAS.Although we did observe a modest increase in connectivity between right M1 and left M1, it did not reach statistical significance (Table 3), potentially due to the limited sample size.When looking at correlations between right SMA, left M1 and right M1 at an individual level an interesting pattern appeared.Four patients exhibited postsurgery increases in connectivity between right SMA and left M1 and between left M1 and right M1 (Fig. 3A).The other four patients, however, showed the opposite patterna decrease in connectivity between right SMA and left M1 as well as between left M1 and right M1 (except for P6 whose LM1-RM1 connectivity was unchanged) (Fig. 3B) We were unable to identify any clear distinguishing characteristics between these groups of patients with regard to other variables (i.e.follow up time, symptomatology, severity of SMA syndrome etc).
This finding implies that increased interhemispheric sensorimotor connectivity, with the healthy SMA compensating for the resected SMA, may not be essential for recovery from the SMAS.This observation concords well with previous studies reporting full recovery from SMAS in a patient with complete corpus callosum agenesia (Obaid et al., 2022).
In addition to the observation of connectivity changes within the sensorimotor network, alterations in functional connectivity between the sensorimotor and the frontoparietal network were investigated.Building on our previous behavioral findings in a partially overlapping dataset we also examined alterations in connectivity between cognitive control and sensorimotor networks (Sjöberg et al., 2019).However, in the present study, no such changes were identified.
Methodologically all studies focusing on functional connectivity, including longitudinal lesional studies such as this one, are based on the often implicitly made assumption of a reasonable test-retest reliability of the analyses made over time(NOBLE).This reliability is supposed to Correlation between the nodes of, LM1 and RM1 when comparing pre-and postoperative rsfMRI sequences.Z-values are from the Signed Ranks Test for the difference between pre and post* = p < 0.05.Correlation between RSMA-LM1 and LM1-RM1 pre-and postoperatively.
mirror the temporal stability of network function in normal subjects (Damoiseaux et al., 2006).A further strength of the present study is that we chose to provide data from a large control group to explicitly shed light on this assumption in the context of interhemispheric connectivity between nodes in the right and left sensorimotor networks respectively.We did so by analyzing test-retest reliability between left and right M1 and S1 respectively in a sample of 221 healthy volunteers using the exact same protocol as the one used to define networks in the study sample.This approach gave a the-test-retest reliability of 0.34 surpassing mean results for similar correlations published over the latest decade (Noble et al., 2019).Still, in the functional connectivity literature test-retest correlations, including the ones reported by us typically fall into the category described as "weak" by statistical convention.Part of the reason for this may be "noise" in the data and part of the explanation may be that data mirrors a true amount of inherent instability of networks and connections.As discussed above, the main finding of the present study was that surgery in the right SMA does not cause changes that are uniformly predictable across subjects even in subjects that have recovered from the SMAS.Since our reliability analyses suggest that we have meaningfully identified and defined relevant dimensions of functional connectivity in the sensorimotor network and in light of previous literature, our finding does seem interesting and potentially valid.However, the fact that test-retest correlations in our study (and other similar studies) are, (as noted above and described in our methods Fig. 3. Changes in connectivity after remission of the SMAS.Correlations between RSMA-LM1 and LM1-RM1 are presented for each patient.A:.Four patients had an increased connectivity between RSMA-LM1.All four of whom also showed a tendency for increased connectivity between LM1-RM1.B: Of the four patients who had a decreased connectivity between RSMA-LM1, three also had a decreased connectivity between LM1-RM1.One patient (P6) had an unchanged connectivity between LM1-RM1.Y-axis is r-value.Correlations between regions in the dorsolateral prefrontal cortex, RSMA and parietal regions.Z-values are from the Signed Ranks Test for the difference between pre and post* = p < 0.05.section), typically "weak" highlights the fact that the possibility of chance findings also needs to be considered and that replications of our results in larger samples would be important.The present study is to our knowledge the second one published that has studied patients with rsfMRI before surgery induced SMA syndrome as well as after resolution of the syndrome (>3 months post op).The previous one, by Vassall et al., included six patients (Vassal et al., 2017).In addition to this we are aware of one study with presenting 33 preoperative scans of patients with SMA syndrome, (Fang et al., 2022) and one study including six patients with postoperative deficits after resections (Yamao et al., 2023).
Seen in the context of previous literature our sample size of eight patients can thus be considered large.However, compared to the average rsfMRI study it is not and it is important to keep in mind that this means that our sample (as well as previously published ones) is vastly underpowered to detect medium-sized and small effects at the group level.It may furthermore be noted that our 6 min scanning protocol is relatively short for a resting state protocol (though short scanning time might be important in a clinical context such as this one).Improvement on both these issues in future replication studies may further decrease the noise component in collected data thereby improving reliability (Birn et al., 2013).
Furthermore, with regard to the possibility of a link between the executive regulation of motor function, beyond the nodes in the sensorimotor system it should be noted that this study, which primarily focused on functional connectivity within the sensorimotor network, left several possibilities unexplored.This means that the fact that no alteration in functional connectivity between the frontoparietal network and the sensorimotor network was found after surgery in these patients does not disprove the possibility that effects on executive function may be an important part of the SMAS.
In sum our main results suggest that the functional reorganization of sensorimotor connectivity and function that occurs after the SMAS can, but does not need to, involve transfer of function from the resected SMA to the contralateral SMA.This result is broadly consistent with previous observations of increased activity in the contralateral SMA in some patients after recovery from the SMAS, (Orgogozo and Larsen, 1979;Vassal et al., 2017) but the present findings indicate that recovery from the syndrome can happen without greater involvement of the contralateral SMA.

Declaration of interest
The authours reports no conflict of interest.

Fig. 1 .
Fig. 1.The colored area represents the resection cavities.White represents areas where all eight patients had tissue removed during surgery.Additional colors represent increasingly fewer patient with resected tissue from 8 to 1 as displayed above.The T1-weighted structural image is a normalized mean of all eight patients.

Table 1
Patient characteristics; Age, Sex, WHO-grade of glioma, speech lateralization, SMAS symptoms and speech and motor function.

Table 2
Correlations between the nodes of RSMA, LM1 and RM1.