Similarities and Differences in the Central Nervous System between Different Types of Primary Glaucoma by Multimodal MRI

: Background. To investigate the similarities and differences in the structure and function of the central nerve system (CNS) among normal tension glaucoma (NTG), primary open-angle glaucoma with high intraocular pressure (HTG) and primary angle-closure glaucoma (PACG) patients. Methods. Ophthalmic examinations and multimodal magnetic resonance imaging (MRI) were performed. The results of brain functional connectivity (FC) were obtained based on the resting state MRI (rs-fMRI). Voxel-based morphometry (VBM) was applied to analyze structural images to obtain the volume change in the gray matter on 3D T1-weighted imaging (T1WI). Results. We found that frontal lobe FC increased in all three groups. Temporal lobe FC increased in the NTG group, while the FC of the cingulate gyrus and postcentral gyrus increased in the HTG and PACG groups by rs-fMRI. There were no gray matter volume changes in the NTG, while the gray matter of the frontal gyrus changed in the HTG and PACG. Moreover, the gray matter of the precentral gyrus changed in PACG, and the parietal lobe and supra marginal lobe changed in the HTG. The white matter of the inferior parietal lobe and postcentral gyrus increased, while that in the inferior temporal lobe decreased in the NTG group. The white matter in the HTG and PACG groups showed significant alterations in the medial occipital gyrus. compared PACG brain high-tension

3 open-angle glaucoma (POAG). POAG has been historically classified into high-tension glaucoma (HTG), defined as intraocular pressure (IOP) > 21 mmHg, and normal-tension glaucoma (NTG), defined as maximal known IOP ≤ 21 mmHg. These classifications are artificial, as IOP, the most commonly known risk factor for glaucoma, exists as a continuum. For the purposes of clinical research, this terminology has currently been maintained.
Patients with NTG, HTG, and PACG have similar visual field (VF) defects. One of the clinical differences is that both the chamber angles of NTG and HTG are open, while those of PACG is at least partially closed. Another difference is the IOP. The IOP is usually higher than normal in both HTG and PACG but remains within the defined "normal" range in NTG. There has been substantial scientific interest in the differences in optic nerve damage between NTG and HTG. [4][5][6] In the past decade, several studies have suggested that abnormalities in the CNS are present in POAG, and POAG might be accompanied by descending damage based on neuronal bidirectional trans-synaptic damage [7,8] . However, most of the previous studies did not explore differences in the structure and function of the CNS between the HTG and NTG groups.
The development of magnetic resonance imaging (MRI) makes it possible to observe changes in brain structure and function in vivo. This technique has been applied in studies of CNS changes in glaucoma [9][10][11][12][13][14][15] . The main research techniques of MRI used to evaluate the CNS in glaucoma patients include resting-state functional MRI (rs-fMRI) and voxel-based measurement (VBM) methods. Therefore, we investigated both the functional and structural differences in the CNS of different types of primary glaucoma based on multimodal MRI for the first time, and we hypothesized that there will be differences among these disease types.

Subjects
This study was a cross-sectional study. The research followed the tenets of the The ocular examination included best-corrected visual acuity (BCVA), dilated stereoscopic fundoscopy, Goldmann applanation tonometry, and standard automatic perimetry (SAP) via the Humphrey Swedish Interactive Threshold Algorithm 24-2 test strategy. All participants had undergone at least two prior VF SAP tests. Subjects with fixation loss >20% or rate of false-positive/negative response >33% were excluded from the study based on the exclusion criteria described in the supplementary materials.

Data Acquisition
All subjects were examined with a 3T magnetic resonance scanner (Achieva TX, Philips Medical Systems, Best, the Netherlands) with an 8-channel phased-array head coil. The rs-fMRI data were acquired with the following parameters: repetition time=2000 ms; echo time=30 ms; flip angle=90°; acquisition matrix=64×64; field of view=240×240 mm 2 ; slice thickness=3 mm with a 1 mm gap. Each brain volume comprised 33 axial slices, and each run contained 160 volumes. The parameters of structural MRI used by the 3D MPRAGE sequence included the following parameters: matrix=256×256 mm; TR=1830 ms; TE=4.43 ms; TI=1100 ms; flip angle=90°; slice number=128; NSA=1; Scan bandwidth=170 Hz/pixel; voxel=1×1×1 mm 3 . During the MRI data acquisition, the subjects were instructed to keep their eyes closed, relax their minds but not fall asleep, and remain as motionless as possible.

Data Preprocessing and Analysis
Data on rs-fMRI processing were obtained via DAPRSF software (Release 2.1; State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China; available in the public domain at http://restfmri.net/forum/DPARSF). For each subject, the preprocessing steps were as follows: a. The first 10 volumes were discarded to allow for magnetization equilibration.
b. The remaining volumes were corrected for the acquisition time and realigned to correct for head motion. Subjects with maximum displacement in any direction 5 of >3.0 mm or head rotation of >3.0 mm were excluded from the study. Subjects with incomplete data were also excluded. As a result, the data for 3 NTG patients, 5 PACG patients, and 4 control subjects were excluded, and 88 subjects (25 HTG patients, 22 NTG patients, 20 PACG patients, and 21 controls) were analyzed.
c. The realigned images were then spatially normalized to the Montreal Neurological Institute space and resampled to 3-mm isotropic voxels. d. For seed-based FC, linear detrending, bandpass filtering, and smoothing were performed before the computation.

Computation of the rs-fMRI parameter and VBM parameters
The FC strength of the rs-fMRI was defined as the Pearson correlation between the time series within other brain regions and the averaged time series within the seed regions. In this work, BA 18 and BA 19 were selected as the seed regions, and Pearson's correlation coefficient was calculated at the voxel level.
VBM has been used to investigate structural changes in the brain [16] . The

Statistical Analysis
We compared the anatomic and functional data in the control and glaucoma groups.
The baseline characteristics of the groups are shown in Table 1.
For each parameter, an analysis of variance (ANOVA) was first performed to 6 determine the differences between the groups, including the 3 subtypes and control groups. Then, multiple comparison analyses (P<0.05 with FDR correction and a cluster size of >20 voxels) were performed to find the difference between each group of the 3 subtypes and the normal control group. In all statistical analyses, age, gender, and right/left eye were used as covariates.
All data acquisitions were masked. MRI data statistical analyses were performed by two individuals blinded to the diagnostic groups.

Rest-fMRI
Compared with the normal controls, the results revealed that temporal lobe FC

Discussion
The blood oxygen level-dependent functional MRI (BOLD-fMRI) technique is an imaging method that uses deoxyhemoglobin as an endogenous contrast agent that relies on changes in regional cerebral blood oxygen saturation, which reflects changes in blood volume, blood flow during nerve activity, and hemoglobin oxygenation. This principle can be used to investigate the functions and activities of the CNS [17] , including the 7 changes in FC in cerebral areas. To date, BOLD-fMRI has been applied to investigate changes in the CNS in patients suffering from various diseases [18][19][20][21] .
Characteristic clinical manifestations of glaucoma include an enlarged optic cup, narrowed rim, retinal nerve fiber layer loss, and VF defects. Previous neuroimaging studies have demonstrated that glaucoma might be a neurodegenerative disease, not only referring to an eye disease [1,2] , but may also induce functional changes in brain areas [13,[22][23][24] . To explore the changes in the CNS of different types of adult primary glaucoma, we investigated the similarities and differences in CNS structure and function in NTG, HTG and PACG patients.

Brain region of functional connectivity
In the present study, compared with normal controls, frontal lobe FC increased in all three groups. The frontal lobe is involved in optic localization, determining the fixation point and coordinating multiple retinal fields. The frontal lobe also transmits visual information to the oculomotor complex [16] . Some prior studies had similar results, although none of these studies analyzed different types of primary glaucoma simultaneously [25,26] . Moreover, in our study, when we used BA18 and BA19 (related to superior visual cortex) as the seeds, we found that the connection between the frontal lobe and the occipital lobe significantly increased in all three groups of visual field-defected glaucoma patients, which indicated that the frontal lobe may have a special relationship with the visual pathway. It was recently reported that optic nerve damage leads to disturbed occipitofrontal functional connectivity, and vision recovery seemed to enhance these connections [27,28] . These findings demonstrated that frontal lobe activity might be involved in vision impairment in glaucoma. However, how the frontal lobe contributes to vision loss is still weakly defined and needs to be explored.
Nevertheless, these results might indicate that glaucoma damage involved not only the visual center but also other parts of the central nervous system.
Notably, there were also some differences between the three groups. Compared with the normal controls, the anterior cingulate cortex and postcentral gyrus FC increased in both the HTG and PACG groups, whereas the temporal lobe FC increased in the NTG group. The functions of the limbic system include emotion, motivation, 8 olfaction and long-term memory [29] . The cingulate cortex, as part of the limbic lobe, played a crucial role in depression and schizophrenia [30][31][32], and its connection to the prefrontal cortex is also associated with pleasure [33] . The postcentral gyrus is a significant structure in the parietal lobe, which is located in the primary somatosensory cortex. Disorders of this area may cause reduced nociception, and patients may become sensitive. Several studies have reported that glaucoma patients have a high prevalence of anxiety and depression, and the authors considered that the oculopathy, vision loss and side effects of treatment might be the reason for this low quality of life [29,[34][35][36][37] .
Therefore, we deduced that both HTG and PACG patients might have CNS damage related to emotional changes. Unfortunately, the reason could not be explained by available studies.
Conversely, temporal lobe abnormalities usually occur in patients with cerebral degenerative disorders, such as Alzheimer's disease [38,39] and frontotemporal dementia.
The patients in our NTG group showed a temporal lobe abnormality in the FC parameter of BOLD-fMRI, which was widely reported to be a typical brain disorder manifestation of BOLD-fMRI. Moreover, patients with Alzheimer's disease were reported to have a higher probability of having open-angle glaucoma [40] . The present study indicates that our NTG patients might have a neurodegenerative change in the central nervous system, which is similar to the neurodegeneration in senile dementia.

Brain region in voxel-based morphometry analysis
VBM is an objective method used to evaluate brain volume changes. This method identifies structural changes throughout the brain and permits the recognition of unsuspected potential regions in the brain but is not biased to any particular area [41] .
VBM has been reported to be used in studies of patients with different kinds of visual field defects and has demonstrated changes in cortical gray and white matter volume [16] .
In our study, compared with normal controls, there was no volume change in the gray matter in the NTG group. However, the gray matter of the middle frontal gyrus was significantly changed in the HTG and PACG groups. The middle frontal gyrus is strongly related to the parietal lobule, temporal gyrus, cingulate gyrus, parahippocampal gyrus, angular gyrus, fusiform gyrus and lingual gyrus. As described above, these areas are 9 primarily responsible for emotion, behavior, and long-term memory. Therefore, this result is consistent with our previous FC results showing that both HTG and PACG patients might have CNS damage related to negative emotion performance, such as depression.
In this study, no change in white matter volume in the occipital lobe was observed in the NTG group, but the white matter volume of the medial occipital lobe was significantly increased in the PACG group and decreased in the HTG group. Boucard et al.
also reported changes in cortical matter density in POAG in the medial occipital cortex [42] .
There were also several other previous studies that showed visual cortex atrophy and functional impairment in HTG patients, although some of the brain areas were different from each other, possibly because of the different methodology of the MRI scanning and data analysis. Moreover, the different inclusion criteria of the subjects could be another reason for this difference [14,43,44] .
However, the changes in the white matter of the occipital lobe were also different between HTG and PACG patients. As mentioned above, the white matter volume of the medial occipital lobe was significantly increased in the PACG group and decreased in the HTG group. Thus far, we do not know the exact reason for this difference, but we speculate that this difference might be due to the different courses of these two diseases. In the case of similar visual field damage, the course of HTG is usually longer than that of PACG. A recent study on early glaucoma showed that before the brain structures start to atrophy, they first become larger and therefore the author reported that the brain might have a temporary compensated stage [45] . Therefore, we deduced that at the early stage of PACG, the gray matter volume of the occipital lobe became larger due to brain compensation because of the relatively shorter duration of high IOP attack. However, compared to PACG, HTG patients usually have a long history of moderately elevated IOP and the occipital lobe is decompensated to atrophy. However, this speculation needs to be proven by further study.

Visual pathway in high tension glaucoma and normal tension glaucoma
In addition to the different brain regions that changed in high tension glaucoma and normal tension glaucoma by using rs-fMRI and VBM analysis, we speculated that there 10 might also be some difference in visual pathway injury between these disease types.
The visual pathway can be divided into 2 pathways [46] , the magnocellular pathways and parvocellular pathways. After visual information integration and processing in the primary visual cortex (V1), the visual signal is transmitted to the secondary visual cortex (extrastriate cortex) by dorsal and ventral pathways.
The dorsal pathways, which receive signals from the magnocellular pathways, are responsible for spatial information and motional orientation of an object and are mainly located in the parietal lobe, occipital lobe and frontal lobe [47] . In our study, we found that frontal lobe FC increased, the gray matter of the frontal gyrus, parietal lobe changed, and the white matter of the medial occipital lobe changed in the HTG and PACG patients, but only frontal lobe FC increased in the NTG patients, which indicated that early and middle stage HTG and PACG patients might show significant changes in the magnocellular pathways, which is consistent with the study by Jiang [48] . A previous study also reported that an experimental animal model with high IOP induced early stage glaucoma showed magnocellular pathway disorder, which is also consistent with our results [49] .
However, we did not observe occipital lobe changes in our NTG patients via fMRI or VBM analysis in this study. However, we found that temporal FC increased and the white matter of the temporal lobe changed in the NTG group, whereas there were no changes in temporal lobe in the HTG and PACG groups. The ventral pathways are mainly located in the temporal cortex and hippocampus. These results indicated that NTG and HTG had different visual pathway disorders in the early and middle stages. Although all three groups of glaucoma patients showed frontal lobe changes, we deduced that glaucoma and neuropathy patients all had emotion problems, as we mentioned before, which might be the reason for this similarity.
Our findings suggested that although similar visual field deficits are observed in NTG and HTG, these diseases seem to express differential pathophysiology [50] . Our speculation is that HTG and PACG might have a different pathological mechanism compared with NTG, and the damage probably results from retinal ganglion cells due to high IOP. In this study, we recruited NTG patients based on the criteria of POAG 11 with IOP≤21 mmHg, which might include a few NTG patients whose initial IOP was in the low teen values and had the same optic nerve damage mechanism as HTG, although their diagnostic IOP was within the normal range. However, our results showed that our NTG group might also include some patients with other neurodegenerative diseases, such as senile dementia or ischemia.
One limitation of our study is that we had limited cases of all three types of glaucoma patients, which might lead to bias in our results. To verify our observations or to correct some inconsistencies in our data, an investigation with a larger sample size will be necessary to account for glaucoma-related CNS changes [45] . Moreover, the consistency and accuracy in the MRI methodology could be the other limitation of our study, which might be influenced by the attention, head movement, or excessive nervousness of the patient. Nevertheless, the main point of our study was to explore the differences in the CNS in primary glaucoma subtypes.

Conclusion
This study indicated that different types of primary glaucoma with similar visual field damage showed some differences in the structure and function of the brain obtained via multimodal MRI. The NTG group showed some unique manifestations compared to the HTG and PACG groups. Future studies are needed to further explore the structural and functional changes in different types of primary glaucoma.

Authors' contributions
XY, JL, XW, YF, and YP were responsible for the project design. All authors collected the data. XY, JL, and ZZ contributed to the data analysis. XY made major contributions to writing the manuscript.
JL, RR and YP revised and edited the manuscript. All authors critically reviewed and approved the final manuscript.

Funding
This study was supported by grants from Capital health development and scientific research projects (number: 2011-4021-05).

Availability of data and materials
The data used to support the findings of this study are available from the corresponding author upon request.

Ethics approval and consent to participate
The study complied with the principles of the Declaration of Helsinki, and was approved by the National Unit of Clinical Trial Ethics Committee, Peking University First Hospital. Written informed consent was obtained from all participants.

Consent for publication
Not applicable.