Ophthalmic changes in a spaceflight analog are associated with brain functional reorganization

Abstract Following long‐duration spaceflight, some astronauts exhibit ophthalmic structural changes referred to as Spaceflight Associated Neuro‐ocular Syndrome (SANS). Optic disc edema is a common sign of SANS. The origin and effects of SANS are not understood as signs of SANS have not manifested in previous spaceflight analog studies. In the current spaceflight analog study, 11 subjects underwent 30 days of strict head down‐tilt bed rest in elevated ambient carbon dioxide (HDBR+CO2). Using functional magnetic resonance imaging (fMRI), we acquired resting‐state fMRI data at 6 time points: before (2), during (2), and after (2) the HDBR+CO2 intervention. Five participants developed optic disc edema during the intervention (SANS subgroup) and 6 did not (NoSANS group). This occurrence allowed us to explore whether development of signs of SANS during the spaceflight analog impacted resting‐state functional connectivity during HDBR+CO2. In light of previous work identifying genetic and biochemical predictors of SANS, we further assessed whether the SANS and NoSANS subgroups exhibited differential patterns of resting‐state functional connectivity prior to the HDBR+CO2 intervention. We found that the SANS and NoSANS subgroups exhibited distinct patterns of resting‐state functional connectivity changes during HDBR+CO2 within visual and vestibular‐related brain networks. The SANS and NoSANS subgroups also exhibited different resting‐state functional connectivity prior to HDBR+CO2 within a visual cortical network and within a large‐scale network of brain areas involved in multisensory integration. We further present associations between functional connectivity within the identified networks and previously identified genetic and biochemical predictors of SANS. Subgroup differences in resting‐state functional connectivity changes may reflect differential patterns of visual and vestibular reweighting as optic disc edema develops during the spaceflight analog. This finding suggests that SANS impacts not only neuro‐ocular structures, but also functional brain organization. Future prospective investigations incorporating sensory assessments are required to determine the functional significance of the observed connectivity differences.

the current spaceflight analog study, 11 subjects underwent 30 days of strict head down-tilt bed rest in elevated ambient carbon dioxide (HDBR+CO 2 ). Using functional magnetic resonance imaging (fMRI), we acquired resting-state fMRI data at 6 time points: before (2), during (2), and after (2) the HDBR+CO 2 intervention. Five participants developed optic disc edema during the intervention (SANS subgroup) and 6 did not (NoSANS group). This occurrence allowed us to explore whether development of signs of SANS during the spaceflight analog impacted resting-state functional connectivity during HDBR+CO 2 . In light of previous work identifying genetic and biochemical predictors of SANS, we further assessed whether the SANS and NoSANS subgroups exhibited differential patterns of resting-state functional connectivity prior to the HDBR+CO 2 intervention. We found that the SANS and NoSANS subgroups exhibited distinct patterns of resting-state functional connectivity changes during HDBR+CO 2 within visual and vestibular-related brain networks. The SANS and NoSANS subgroups also exhibited different resting-state functional connectivity prior to HDBR+CO 2 within a visual cortical network and within a large-scale network of brain areas involved in multisensory integration. We further present associations between functional connectivity within the identified networks and previously identified genetic and biochemical predictors of SANS. Subgroup differences in resting-state functional connectivity changes may reflect differential patterns of visual and vestibular reweighting as optic disc edema develops during the spaceflight analog.
This finding suggests that SANS impacts not only neuro-ocular structures, but also functional brain organization. Future prospective investigations incorporating sensory assessments are required to determine the functional significance of the observed connectivity differences.

| INTRODUCTION
As NASA prepares for the first human mission to Mars, it is critical to understand how spaceflight affects the brain and body. During longduration missions onboard the International Space Station, some astronauts develop ophthalmic abnormalities including optic disc edema, choroidal folds, cotton wool spots, posterior globe flattening, and distention of the optic nerve sheath Lee et al., 2020;Mader et al., 2011). These findings have been collectively termed Spaceflight Associated Neuro-ocular Syndrome (SANS). The origin and effects of SANS are not understood, and present significant potential risk to astronaut health and performance. Here we explored whether the development of signs of SANS during a spaceflight analog is associated with altered brain functional organization.
The effects of spaceflight are often studied using ground-based analogs, with head down-tilt bed rest (HDBR) being one of the most widely used (Cassady et al., 2016;Pavy-Le Traon, Heer, Narici, Rittweger, & Vernikos, 2007). The head down-tilt posture simulates microgravity effects on the body including headward fluid shifts and axial body unloading. Headward fluid shift is hypothesized to be a contributing factor of SANS (Mader et al., 2011). Spaceflight has been shown to alter cerebrospinal fluid hydrodynamics (Kramer et al., 2020), and result in ventricular volume increases (Alperin & Bagci, 2018;Kramer et al., 2020;Van Ombergen et al., 2019), which have been associated with ocular globe deformation (Alperin & Bagci, 2018) and visual acuity visual changes (i.e., hyperopic refractive error shifts) (Van Ombergen et al., 2019). However, until recently, signs of SANS have not manifested in any spaceflight analog (Taibbi et al., 2016). Another potential contributing factor to SANS development during spaceflight is the mild hypercapnic environment of the International Space Station (Stenger et al., 2017) where CO 2 levels (0.5%) are more than 10 times higher than on Earth (Law et al., 2014).
As CO 2 is a potent arterial vasodilator, small increases in arterial CO 2 increase cerebral blood flow and intracranial blood volume (Artru, 1987;Fortune et al., 1995;Moseley et al., 1992), which can elevate intracranial pressure. As such, microgravity-induced headward fluid shifts combined with exposure to elevated CO 2 may contribute to SANS development during spaceflight.
To better replicate the International Space Station environment, NASA commissioned the current spaceflight analog study employing 30 days of strict HDBR in elevated ambient CO 2 (HDBR+CO 2 ). The ambient CO 2 was elevated to 0.5% during HDBR to match the mild hypercapnic environment of the International Space Station (Law et al., 2014). This study used "strict" HDBR such that subjects remained in the 6 head down-tilt position at all times, and were not permitted to use a standard pillow (Lawley et al., 2017). To investigate the brain's adaptation to the HDBR+CO 2 spaceflight analog, we assessed resting-state brain activity before, during, and after bed rest using functional magnetic resonance imaging (fMRI). Using this same dataset, we recently reported resting-state connectivity changes throughout the HDBR+CO 2 intervention across the entire group as well as brain-behavior associations (McGregor et al., 2021). We reported that exposure to the HDBR+CO 2 intervention resulted in gradual resting-state functional connectivity (FC) changes involving vestibular, visual, somatosensory and motor brain areas. During the HDBR+CO 2 intervention, a subset of participants developed optic disc edema-a sign of SANS Zwart et al., 2019). This occurrence afforded us the unique opportunity to explore functional brain differences between participants who developed optic disc edema during the spaceflight analog (SANS subgroup) and those who did not (NoSANS subgroup).
Here, we examined whether the SANS and NoSANS subgroups exhibited different longitudinal changes in FC from before, during, to after the HDBR+CO 2 intervention.
Genetic and biochemical risk factors for optic disc edema development have recently been identified in the SANS subgroup in the current study (Smith, Laurie, Young, & Zwart, 2019;Zwart et al., 2019) as well as in astronauts who develop SANS (Smith & Zwart, 2018;Zwart et al., 2016). Genetic risk factors include G alleles for the 5-methyltetrahydrofolate-homocysteine methyltransferase reductase (MTRR) A66G gene polymorphism and C alleles for the serine hydroxymethyltransferase 1 (SHMT1) 1420 gene polymorphism.
Low levels of B vitamins (B6, B12, and folate) have also been found to predict optic disc edema development with the current HDBR+CO 2 intervention  and with spaceflight (Smith & Zwart, 2018;Zwart et al., 2016). These findings may help explain the underlying risk for individuals who develop SANS compared to those who do not. In light of this, we further investigated FC differences between the SANS and NoSANS subgroups prior to bed rest. We present qualitative associations between identified FC differences between the SANS and NoSANS subgroups with participants' B vitamin concentration and risk alleles.
Previous studies have examined changes in FC with exposure to HDBR in standard ambient air. These studies have demonstrated FC changes among brain areas involved in somatosensory, vestibular, and motor function as well as brain areas involved in multisensory integration, and regions within the default mode network (Cassady et al., 2016;Demertzi et al., 2016;Koppelmans et al., 2017;Liao et al., 2012;Liao et al., 2015;Zhou et al., 2014). Based on the findings of these previous HDBR studies in ambient air (Cassady et al., 2016;Demertzi et al., 2016;Koppelmans et al., 2017;Liao et al., 2012;Liao et al., 2015;Zhou et al., 2014), we hypothesized that brain areas related to multisensory integration, and regions within somatosensory, motor, vestibular, visual, and default mode networks would exhibit differential FC between the SANS and NoSANS subgroups. As this was the first HDBR study to induce signs of SANS, we also used an exploratory hypothesis-free approach.
Our analysis of longitudinal FC changes showed that the SANS subgroup exhibited a distinct pattern of resting-state FC changes involving visual and vestibular brain regions during the HDBR+CO 2 intervention compared to the NoSANS subgroup. Further, the SANS and NoSANS subgroups exhibited differential resting-state brain activity prior to HDBR+CO 2 within a visual cortical network and within a large-scale network of brain areas involved in multisensory integration. While preliminary, these findings suggest that the development of optic disc edema during this spaceflight analog is associated with functional brain reorganization.

| Participants
Twelve healthy adults enrolled in this study (6 males, 6 females). Participants were nonsmokers between 24 and 55 years of age with a body mass index ranging from 19 to 30 kg/m 2 . Exclusion criteria included any chronic illness and pain, elevated risk of thrombosis, chronic hypertension, diabetes, obesity, arthritis, hyperlipidemia, hepatitis, or metabolic bone diseases. Participants were also screened for any eye conditions that significantly impair vision. All participants passed psychological screening and an Air Force Class III equivalent physical examination. One female participant withdrew from the study on the first day of HDBR+CO 2 . Eleven participants in total completed the study; their demographics are summarized in Table 1. All participants provided written informed consent to study protocols approved by the ethics committee of the local medical association (Ärztekammer Nordrhein) and the institutional review boards at The University of Florida and NASA Johnson Space Center.
During the HDBR+CO 2 intervention, 5 participants developed optic disc edema ("SANS subgroup"), and 6 participants did not ("NoSANS subgroup"). Development of optic disc edema did not result in measurable visual changes in the SANS subgroup. As reported previously , there was no statistically reliable age difference between participants in the SANS and NoSANS subgroups.

| Study design
This study was carried out at: envihab, the German Aerospace Center's medical research facility in Cologne, Germany. This experiment was conducted as part of the VaPER (Vision Impairment and Intracranial Pressure and Psychological: envihab Research) bed rest campaign.
Participants resided at the: envihab facility for the duration of this 58-day study. Figure 1 shows the study design and each phase is summarized below.

| Baseline data collection (BDC)
Baseline data collection (BDC) occurred during a 14-day phase (BDC-14 to BDC-1). Participants were ambulatory and the ambient air was standard (i.e., not hypercapnic). Participants remained inside the bed rest module of the: envihab facility, and their activity level was restricted to walking. Participants slept in a horizontal (i.e., 0 tilt) position. Resting-state fMRI scans were acquired twice during the baseline phase: 13 and 7 days before the bed rest phase (BDC-13 and BDC-7, respectively).

| Head down-tilt bed rest with elevated carbon dioxide (HDBR+CO 2 )
Following the BDC phase, participants underwent 30 days of strict 6 HDBR in elevated CO 2 (HDBR+CO 2 ). The ambient CO 2 level was maintained at 0.5% during HDBR, corresponding to the average CO 2 concentration onboard the ISS (3.8 mmHG ambient partial pressure of carbon dioxide) (Law et al., 2014). Participants adhered to bed rest 24 hours per day and were continuously monitored to ensure compliance. Participants were instructed to always have at least one of their shoulders in contact with the mattress. They were not permitted to raise or prop up their head at any time. We employed strict HDBR to keep the head at a À6 angle at all times and maintain elevated fluid pressure in the head (Lawley et al., 2017 Resting-state fMRI scans were acquired twice during bed rest on days 7 and 29 of HDBR+CO 2 (HDT7 and HDT29, respectively).

| Recovery (R)
Upon the completion of the HDBR+CO 2 intervention, participants underwent a 14-day recovery phase (R + 0 to R + 13). As in the BDC phase, participants were ambulatory and the ambient air was standard (i.e., not hypercapnic). Participants remained inside the bed rest module of the: envihab facility, and their activity level was restricted to walking. Participants slept in a horizontal (i.e., 0 tilt) position. Onehour rehabilitation sessions took place 3, 5, 7, 8, 9, and 11 days following bed rest during which participants performed active stretching and exercises to improve strength and coordination using a coordination ladder or Bosu ® ball. Resting-state fMRI scans were acquired twice during the recovery phase: 5 and 12 days after the end of bed rest (R5 and R12, respectively).

| MRI acquisition
For all six scan sessions, the participant's body was in the 6 HDT position, but the head was horizontal in the MRI coil. A mask and tank system were used to maintain CO 2 levels at 0.5% throughout MRI scans during the HDBR+CO 2 phase. Neuroimaging data were acquired on Whole-brain resting-state fMRI data were acquired using a T2*weighted EPI sequence (TR = 2,500 ms, TE = 32 ms, flip angle = 90 , 37 axial slices, slice thickness = 3.5 mm, duration = 7 minutes). During resting-state scans, participants were instructed to remain awake and to fixate a red on-screen circle without thinking about anything in particular.

| Image preprocessing
Image preprocessing has been detailed elsewhere ( To improve cerebellar normalization to MNI space, we extracted the cerebellum from each functional and anatomical image then normalized the cerebellums to the SUIT cerebellar template. To accomplish this, we used CERES to segment the cerebellum from each subject's native space anatomical scan (Romero et al., 2017). CERES uses an automatic pipeline to perform cerebellar segmentation, and has been shown to yield more accurate segmentations than other toolboxes including SUIT (Carass et al., 2018). CERES-generated cerebellar masks were applied to coregistered T1 and functional images, allowing us to extract the cerebellum. The extracted cerebellums were then normalized to the SUIT cerebellar template. We chose to use the SUIT template because it shows greater anatomical detail within the cerebellum relative to whole brain MNI templates, yielding better normalization for fMRI data (Diedrichsen, 2006).
Whole-brain MRI-normalized functional images were smoothed using a 5 mm FWHM Gaussian kernel. MNI-normalized cerebellums extracted from functional scans were smoothed using a 2 mm FWHM Gaussian kernel. A smaller smoothing kernel was used to reduce smoothing across tissue types within the cerebellum's smaller internal structures.
F I G U R E 1 Baseline data collection (BDC) was carried out during a 14-day ambulatory phase in standard ambient air. Participants then underwent 30 days of strict 6 head down-tilt bed rest with 0.5% ambient carbon dioxide (HDBR+CO 2 ). The bed rest intervention was followed by a 14-day ambulatory recovery (R) phase in standard ambient air. Resting-state fMRI scans were acquired on sessions BDC-13, BDC-7, HDT7, HDT29, R5, R12

| Image denoising
Head movement is known to influence measures of resting-state FC, introducing spurious correlations between distinct structures throughout the brain (Power, Barnes, Snyder, Schlaggar, & Petersen, 2012;Van Dijk, Sabuncu, & Buckner, 2012). We used the ARTifact detection tool (ART; http://www.nitrc.org/projects/artifact_detect), which is integrated into CONN, to detect and reject motion artifacts. Using Preprocessed resting-state runs were denoised by regressing out confounding effects of head motion (6 head motion parameters and their 6 first-order temporal derivatives), 5 principal components modeling estimated noise from the white matter, 5 principal components modeling estimated noise from the CSF, and a "scrubbing" regressor modeling ART-detected outlier volumes within the run. Resting-state fMRI images were then band-pass filtered between 0.008 and 0.09 Hz (Biswal, Yetkin, Haughton, & Hyde, 1995;Damoiseaux et al., 2006).

| Subject-level analyses
2.6.1 | Hypothesis-driven seed-to-voxel analyses We used a hypothesis-driven approach to examine resting-state functional connectivity (FC) between 14 regions of interest (ROIs) and the rest of the brain. Drawing upon the findings of previous HDBR studies in standard ambient air, we hypothesized that the two subgroups would exhibit differential patterns of FC involving brain areas related to multisensory integration, somatosensation, motor function, vestibular processing, and regions within the default mode network (Cassady et al., 2016;Demertzi et al., 2016;Liao et al., 2014;Liao et al., 2015;Zhou et al., 2014). The coordinates for the ROIs were the same as those used in our previous resting-state fMRI study using HDBR intervention in standard ambient air (Cassady et al., 2016). We also included ROIs located within left frontal eye field and bilateral primary visual cortices, the coordinates of which were determined using the automatic anatomic labeling (AAL) atlas included in the CONN toolbox.
ROIs were 8-mm diameter spheres centered on coordinates presented in Table 2.
We computed the average time series of unsmoothed functional data within each ROI. Time series for ROIs located in the cerebrum were extracted from the whole brain unsmoothed func- analog studies, we also employed a hypothesis-free approach to explore connectivity differences between the SANS and NoSANS subgroups. This voxel-to-voxel analysis does not require a priori ROI selection and as such allowed us to identify subgroup differences that may otherwise be missed by our seed-to-voxel approach using hypotheses based on the findings of HDBR studies in standard ambient air. For this approach, we estimated the intrinsic connectivity contrast, a measure reflecting the strength of connectivity between a given voxel and every other voxel in the brain (Martuzzi et al., 2011).
First, we computed Pearson's correlation coefficients between the time course of each voxel with that of every other voxel in the brain (or within the cerebellum). We calculated the root mean square of each correlation coefficient and then performed a Fisher Z-transformation. This procedure was repeated for each participant's six restingstate runs, yielding per-session Z score maps reflecting the strength (i.e., magnitude) of connectivity between each voxel and the rest of the brain.

| Subgroup differences in longitudinal changes in functional connectivity
We first examined if the SANS and NoSANS subgroups exhibited differential patterns of FC changes from before, during, to after the HDBR +CO 2 intervention. We performed this longitudinal analysis using both hypothesis-based and hypothesis-free approaches. In our general linear model (GLM), subgroup was used as a between-subjects factor (i.e., SANS, NoSANS) and session was used as a within-subjects factor (i.e., 6 time points). The longitudinal within-subject contrast model used for this analysis was informed by the ophthalmological findings of another research group that participated in this HDBR+CO 2 study . Laurie and colleagues measured total retinal thickness in these same participants using optical coherence tomography on days 1, 15, and 30 of the HDBR+CO 2 intervention, and on the 6th and 13th day of the recovery phase. The SANS subgroup showed a gradual increase in total retinal thickness during HDBR+CO 2 which was followed by a gradual post-bed rest decrease towards baseline values . Our longitudinal analysis aimed to identify brain networks in which FC changes mirrored the time course of optic disc edema onset and recovery. We therefore used a within-subject contrast model (shown in Figure 2) to identify brain areas in which FC was stable across the baseline phase (BDC-13, BDC-7) and then showed a linear FC change throughout HDBR+CO 2 followed by a gradual return towards baseline in the subsequent recovery phase. Participant age and sex were included as covariates. The significance level was set at an uncorrected voxel threshold of p < .001 with a cluster-size threshold of p < .05 (twotailed) corrected for multiple comparisons according to the false discovery rate (FDR) method.

| Subgroup differences in functional connectivity at baseline
In light of recent work identifying genetic and biochemical predictors of optic edema development in these same participants , we also tested if there were FC differences between the SANS and NoSANS subgroups before bed rest. This analysis aimed to uncover functional brain networks in which resting-state activity was predictive of which participants would develop signs of SANS in the subsequent HDBR+CO 2 intervention. For this analysis, our GLM included subgroup (i.e., SANS, NoSANS) as a between-subjects factor and baseline session (i.e., BDC-13, BDC-7) as a within-subjects factor.
The within-subjects contrast modeled the average FC across the 2 BDC sessions. Participant age and sex were included as covariates.

| Associations between functional connectivity and genetic and B vitamin status
Biochemical and genetic data were collected from participants as part of NASA's international standard measures assessments and other studies.
A fasting blood sample was collected on BDC-11, 11 days before the HDBR+CO 2 intervention. Samples were analyzed for MTRR 66 G alleles, SHMT1 1,420 C alleles, B6 and B12 vitamin levels, and homocysteine as previously described . We quantified genetic risk for SANS by summing the number of MTRR 66 G alleles and SHMT1 1420 C alleles, as has been done in other studies .
For functional brain networks identified by our longitudinal analysis, we qualitatively examined associations between each risk factor and changes in FC from pre-to post-HDBR+CO 2 (i.e., BDC-7 to HDT29). For functional brain networks identified by our baseline analysis, we qualitatively examined associations between each risk factor and average baseline FC values (i.e., across BDC-13 and BDC-7). As these genetic and biochemistry measures predict optic disc edema, which we used as a regressor (SANS vs. NoSANS) in our resting-state fMRI, it follows that FC differences in our identified networks should be related to risk factors. Due to the nonindependence of this analysis, we present associations for illustrative purposes only, not as the basis for inference (Poldrack & Mumford, 2009).

| Subgroup differences in longitudinal changes in functional connectivity
Our hypothesis-based seed-to-voxel analyses revealed that the ROI in right posterior parietal cortex (PPC) and clusters within bilateral insular cortices exhibited differential longitudinal changes between the SANS and NoSANS subgroups (Figure 3a). The NoSANS subgroup showed decreased FC between these regions during HDBR+CO 2 followed by a post-bed rest FC increase. In contrast, the SANS subgroup showed the opposite pattern, increasing FC during HDBR+CO 2 followed by a post-bed rest decrease.
Cluster coordinates and statistics are presented in Table 3.
Furthermore, the ROI in left PCC exhibited differential longitudinal changes with right V1 between the two subgroups ( Figure 3b).
The NoSANS subgroup showed a FC increase between these regions during HDBR+CO 2 which decreased after bed rest whereas the SANS subgroup showed decreased FC between these regions during HDBR +CO 2 followed by a post-bed rest reversal. Cluster coordinates and statistics are presented in Table 4.
Our hypothesis-free voxel-to-voxel analysis did not yield any statistically significant clusters.

| Subgroup differences in functional connectivity at baseline
We also identified two brain networks in which average FC across the baseline scan sessions significantly differed between the SANS and NoSANS subgroups. The first network, shown in Figure 4a, consisted of right primary visual cortex (V1) and higher order visual areas V3 and V4 within the left hemisphere. The NoSANS subgroup exhibited positive connectivity between these visual brain areas at baseline. In contrast, the SANS subgroup exhibited lower FC or an anti-correlation (i.e., negative FC) between right V1 and left V3 and V4 relative to the NoSANS subgroup. Cluster coordinates and statistics are presented in Table 5.
The second network, identified using left insular cortex as a seed  Table 6.
Our hypothesis-free voxel-to-voxel analysis did not yield any statistically significant clusters.
To assess the reliability of FC values between BDC-13 and BDC-7, we computed the intraclass correlation coefficient for each unique ROIcluster connection for the networks shown in Figure 4. Table 7 shows the intraclass correlation coefficients and 95% confidence intervals for each unique ROI-cluster connection. For the network shown in

| Subgroup differences in longitudinal changes in functional connectivity
In Figure 6, we show relationships between previously identified genetic and biochemical predictors of SANS and participants' FC changes from pre-to post-HDBR+CO 2 within the networks identified F I G U R E 3 Differential longitudinal FC changes between the SANS and NoSANS subgroups. (a) The NoSANS subgroup (gray squares) showed a FC decrease between the ROI in right posterior parietal cortex (PPC) and clusters in bilateral insular cortices during HDBR+CO 2 followed by a post-bed rest reversal. The SANS subgroup (triangles) exhibited FC increases between these regions during HDBR+CO 2 followed by a post-bed rest decrease. (b) The NoSANS subgroup (gray squares) showed a FC increase between the ROI in left posterior cingulate cortex (PCC) and right primary visual cortex (V1) during HDBR+CO 2 followed by a post-bed rest reversal. The SANS subgroup (triangles) exhibited a FC decrease between these regions during HDBR+CO 2 followed by a post-bed rest increase.     in which multisensory integration occurs, combining inputs from the visual, somatosensory and auditory systems (Lewis & Van Essen, 2000). The insula is a region within the vestibular system (Zu Eulenburg, Caspers, Roski, & Eickhoff, 2012), which uses multisensory integration for self-motion perception, spatial orientation, gaze stabilization, and postural control (Balaban, 2016 and HDBR, the vestibular system receives altered or reduced sensory inputs and it is thought that the brain adaptively reweights sensory inputs as a result (Mulavara et al., 2018;Young et al., 1986;Young, Oman, Watt, Money, & Lichtenberg, 1984). The brain increases weighting of reliable sensory inputs or use in guiding performance while decreasing weighting of altered or unreliable inputs (Assländer & Peterka, 2014). The current results suggest differential patterns of vestibular reweighting during the HDBR+CO 2 between the SANS and NoSANS subgroups. The SANS subgroup's coupling between PPC and the insula may reflect progressively increased weighting of vestibular inputs during the spaceflight analog as optic disc edema develops. However, we did not find subgroup differences in vestibular processing in response to pneumatic skull taps from preto post-HDBR+CO 2 (Hupfeld et al., 2020). Future studies incorporating behavioral metrics of vestibular input reliance are required to investigate the functional significance of this finding.
The SANS subgroup exhibited reduced FC between posterior cingulate cortex (PCC) and primary visual cortex (V1) during HDBR+CO 2 and then recovered post-bed rest whereas the NoSANS subgroup exhibited FC increases between these regions during HDBR+CO 2 .
PCC is a key region within the default mode network. Although its function remains a topic of debate, recent theories suggest that the PCC is involved in task-independent introspection and self-referential processes (Leech & Sharp, 2014). V1 is the first cortical stage of the visual system wherein low-level visual feature processing occurs (Tootell et al., 1998 (Pechenkova et al., 2019). These studies have also provided evidence for vestibular cortex reorganization and multisensory reweighting following spaceflight. The authors did not report if the cosmonauts who participated in these studies developed SANS during spaceflight, though it has been reported that cosmonauts do not develop SANS perhaps due to differences in in-flight countermeasures used by astronauts and cosmonauts (Jillings et al., 2020).
As CO 2 is a potent arterial vasodilator, it is not possible to distinguish CO 2 -induced changes in vascular function from changes in neural activity on the basis of BOLD fMRI alone (Drew, 2019).
Throughout this study, arterial partial pressure of carbon dioxide, assessed as part of NASA's standard measures, showed a small but significant increase during the HDBR+CO 2 phase (McGregor et al., 2021), though see also (Laurie et al., 2020b;. Cerebral perfusion was also assessed during our MRI sessions using arterial spin labeling (Roberts et al., 2021). As reported by Roberts et al, both subgroups showed a reduction in average perfusion across the whole brain during the HDBR+CO 2 phase relative to baseline measures (Roberts et al., 2021 These subgroup differences in pre-bed rest FC are likely markers of other differences that make subjects in the SANS subgroup more prone to developing optic disc edema. For example, recent work has identified genetic and biochemical predictors of optic disc edema development in the SANS subgroup of this study Zwart et al., 2019) as well as in astronauts who develop SANS (Zwart et al., 2016). Low levels of B vitamins (B6, B12, riboflavin, and folate) and 1-carbon pathway risk alleles (all of which are linked to circulating homocysteine concentration) predict optic disc edema during HDBR +CO 2 Zwart et al., 2019) and spaceflight (Smith & Zwart, 2018;Zwart et al., 2017). There is currently little evidence that individual differences in genetics and/or B vitamin status directly relate to functional connectivity (Downey et al., 2019). Genetics, B vitamin status, and functional connectivity have all been linked to individual differences in anatomical connectivity. Lower vitamin B12 status and higher homocysteine concentration are associated with lower white matter volume in healthy older adults (Feng et al., 2013;Vogiatzoglou et al., 2008). Functional connectivity also shows a close correspondence to anatomical connectivity (Fox et al., 2005). It is thus feasible that subgroup differences in B vitamin levels and risk alleles in the current study Zwart et al., 2019) contribute to anatomical connectivity differences within the identified networks at baseline, which in turn result in individual differences in functional connectivity. However, if this were the case, one would expect the SANS subgroup to exhibit lower FC compared to the NoSANS group for both networks. Inclusion of anatomical brain connectivity measures are required to directly test this hypothesis.
Analyses of average whole brain cerebral perfusion in these same subjects have shown no significant differences between the SANS and NoSANS subgroups prior to bed rest (Roberts et al., 2021). However, since perfusion analyses were performed at the whole-brain level, they may not have captured regional differences in perfusion between the subgroups.
Future research is required to further investigate the cause(s) of SANS, and whether or how these pre-bed rest FC differences relate to the cause(s). Direct regional comparisons of structural, perfusion, and resting-state fMRI data between the SANS and NoSANS groups during the baseline phase are required to disentangle the origin of the subgroup FC differences within the FC networks reported here.

| Limitations
The small sample size is a significant limitation of our work. This study was the first long-duration bed rest study to implement strict HDBR and use a mild hypercapnic environment. As such, we are unable to determine whether the development of optic disc edema was due to strict HDBR, elevated CO 2 levels, or a combination thereof. Moreover, optic disc edema in the SANS subgroup was more pronounced than typically seen in astronauts post-flight .
In addition, cases of SANS in astronauts typically include other ophthalmic changes including globe flattening, choroidal folds, cotton wool spots, etc. (Mader et al., 2011). Thus, it remains unclear how closely the signs of SANS observed in the current dataset mimic those seen in astronauts.
Visual deterioration during long-duration spaceflight represents a significant potential risk to human health and performance on current and future missions. Therefore, despite the limitations noted above, these data have provided important insights into the underlying cause of SANS (Smith & Zwart, 2018;Zwart et al., 2017) and understanding the breadth of its effects. This is particularly the case given that prior work has only related SANS to central nervous system structures, and not to connectivity or activity.

| Conclusion
Here we have shown that the SANS and NoSANS subgroups exhibited differential resting-state brain activity prior to the spaceflight analog within a visual cortical network and within a largescale network of brain areas involved in multisensory integration.
We also showed that the development of signs of SANS during the spaceflight analog is associated with differential patterns of resting-state connectivity changes within visual and vestibularrelated brain networks. This finding suggests that SANS impacts not only neuro-ocular structures, but also brain function. Forthcoming investigations by our group will combine functional and structural brain analyses to assess whether astronauts who develop and decision to submit the manuscript for publication.