Morphological brain alterations in dialysis- and non-dialysis-dependent patients with chronic kidney disease

To 1) investigate the morphological brain-tissue changes in patients with dialysis- and non-dialysis-dependent chronic kidney disease (CKD); 2) analyze the effects of CKD on whole-brain cortical thickness, cortical volume, surface area, and surface curvature; and 3) analyze the correlation of these changes with clinical and biochemical indices. This study included normal controls (NCs, n = 34) and patients with CKD who were divided into dialysis (dialysis-dependent chronic kidney disease [DD-CKD], n = 26) and non-dialysis (non-dialysis patients who underwent cranial magnetic resonance imaging scans [NDD-CKD], n = 26) groups. Cortical thickness, volume, surface area, and surface curvature in each group were calculated using FreeSurfer software. Brain morphological indicators with statistical differences were correlated with clinical and biochemical indicators. Patients with CKD exhibited a significant and widespread decrease in cortical thickness and volume compared with NCs. Among the brain regions associated with higher neural activity, patients with CKD exhibited more significant morphological changes in the paracentral gyrus, transverse temporal gyrus, and lateral occipital cortex than in other brain regions. Cortical thickness and volume in patients with CKD correlated with blood pressure, lipid, hemoglobin, creatinine, and urea nitrogen levels. The extent of brain atrophy was further increased in the DD-CKD group compared with that in the NDD-CKD group. Patients with CKD potentially exhibit a certain degree of structural brain-tissue imaging changes, with morphological changes more pronounced in patients with DD-CKD, suggesting that blood urea nitrogen and dialysis may be influential factors in brain morphological changes in patients with CKD.


Introduction
Chronic kidney disease (CKD) is characterized by a decreased estimated glomerular filtration rate (eGFR) and increased urinary albumin excretion (Agarwal et al. 2021). According to the Kidney Disease Outcomes Quality Initiative (KDOQI) guidelines, patients with CKD can be classified as dialysis patients (Ammothumkandy et al. 2022) and divided into dialysis-dependent CKD (DD-CKD) and non-dialysis-dependent CKD (NDD-CKD) groups. When the disease progresses to end stage renal disease (ESRD) (Drew et al. 2019), that is, eGFR < 15 mL/min/1.73 m 2 or renal failure permanently reduced to 10% of normal function with concomitant multi-organ dysfunction, lifelong maintenance dialysis or kidney transplantation is required. Epidemiological surveys have revealed that CKD affects approximately 15% of adults in the United States and has a significantly higher prevalence in adults aged > 65 years (Dusek et al. 2021). In addition to the increased risk of Huan Yu and Chaoyang Zhang contributed equally to this work. dialysis, CKD-related multi-organ complications constitute one of the reasons for its classification as a major global health burden, especially the increased risk of central nervous system disease (Giarrocco and Averbeck 2021).
Several neuroimaging studies have demonstrated that structural changes in brain tissue are a common symptom in patients with CKD, including cerebral white matter hypersignals, asymptomatic cerebral infarction, and brain atrophy (Giovagnoli et al. 2020;Grundy et al. 2019). Since 1977, when Passer et al. first observed structural brain cortical alterations in CKD patients and noted that their brain atrophy was dominated by the frontal lobe, magnetic resonance imaging (MRI) has been widely used as a sensitive tool for the examination of brain cortical structures, and scholars have noted that the structural alterations of brain tissue in CKD patients are far greater relative to the 8-28% incidence of cerebrovascular lesions in the general population (Madre et al. 2020;Chen et al. 2021;Tsuruya et al. 2015). In 2021, Chen et al. of Zhongshan Hospital reviewed 91 studies and analyzed for the first time the correlation between eGFR, albuminuria and structural alterations of brain tissue in patients with CKD. The group pointed out that the limitation of the above studies is the homogeneity of the population, especially in Asian populations, the lack of data support, and the limited studies on morphological alterations of brain tissue in patients with different stages of hyperalgesia. We performed a cross-sectional observational study of patients with CKD (DD-CKD and NDD-CKD groups) and normal controls (NCs) using magnetic resonance imaging (MRI). FreeSurfer software was used to calculate the cortical thickness, volume, surface area, and surface curvature of the subjects in each group. Moreover, a statistical analysis of the differences among the groups, 68 brain regions, and five brain morphological indicators was performed to investigate the association between different clinical CKD subgroups and morphological changes of local brain tissue in patients. In addition, brain morphological indices with statistical differences were correlated with clinical and biochemical indices to establish the association between brain morphological changes and renal function-related indices.

Participants and enrollment criteria
In this study, 52 patients with CKD were recruited between May 2021 and January 2022 from the ward and outpatient clinic of the Department of Nephrology, Liangxiang Hospital, Capital Medical University, and divided into nondialysis (n = 26) and dialysis (n = 26) groups according to eGFR levels. Thirty-four age-and sex-matched NCs were enrolled during the same period, and all subjects signed an informed consent form. The enrollment criteria for patients with CKD were as follows: (1) patients who met the National Kidney Foundation KDOQI diagnostic criteria for CKD, (2) age ≥ 18 years, (3) receiving maintenance hemodialysis or continuous ambulatory peritoneal dialysis for > 3 months, and (4) free of infection and other complications within the preceding 3 months. Prior to enrollment, patients provided a detailed medical history and underwent physical examination and necessary laboratory tests, none of which were contraindicated for MRI scanning. Patients with any of the following comorbidities were excluded: (1) neurological disorder; (2) any physical illness (such as hepatitis, brain tumor, trauma, or epilepsy) within the preceding 180 days; or (3) a history of major depression, bipolar disorder, schizophrenia, and substance abuse or dependence. This study was conducted in accordance with the 1964 Declaration of Helsinki and its later amendments and was approved by the Ethics Committee of Liangxiang Hospital of Capital Medical University (date of approval: May 8, 2021; approval number: 2016126).

Image processing
All MRI images were processed using FreeSurfer software (version 7.1.0, https:// surfer. nmr. mgh. harva rd. edu/), which was utilized for volumetric segmentation and cortical reconstruction (Fischl 2012). FreeSurfer was used to automatically partition the cortical surface of the bilateral cerebral hemispheres into 68 gyrus-based regions of interest (ROIs) according to the Desikan-Killiany atlas (Desikan et al. 2006). Based on each of these 68 surfaced regions, five cortical properties (i.e., cortical thickness, volume, surface area, surface area, and local gyrification index (LGI)) were extracted to represent its brain morphology.
Specifically, the original MRI data were first subjected to a series of preprocessing steps. The first step was format conversion, using MRIcron software (https:// www. nitrc. org/ proje cts/ mricr on/) to convert 3D T1 weighted images (T1WI) in DICOM format to NIFTI format. Secondly, NIFTI-format data was imported into FreeSurfer software to implement the processing pipeline in the Linux Ubuntu system. Thirdly, T1-weighted images were preprocessed with non-uniformity correction: head movement correction, skull stripping, and volume segmentation. Afterward, through encoding the shape of the corpus callosum and pons in Talairach standard space and following the intensity gradient from white matter to cerebrospinal fluid, the pial surface (the boundary between the gray matter and cerebrospinal fluid) and white surface (the boundary between the gray and white matter) were generated for each hemisphere (Dale et al. 1999;Peng et al. 2013;Fischl and Dale 2000). Then, a cortical surface-based atlas was mapped to a sphere aligning the cortical folding patterns. Desikan-Killiany atlas was used to segment the cortical surface into 68 gyrus-based ROIs, followed by the computation of five cortical metrics for each ROI, including cortical thickness, volume, surface area, surface area, and LGI. The five metrics were defined as follows: 1) Cortical thickness-the corresponding distance between pial surface and white surface at each vertex of the cortex; 2) Volume-subcortical volumes were obtained from the automatic volume segmentation procedure (Fischl et al. 2002); 3) Surface area-the area of every gyrus-based ROI was summed from the area of triangles for their vertices; 4) Surface curvature-the mean of the curvature of vertices in this ROI; and 5) LGI-measured using the method developed by Schaer (Schaer et al. 2008). The extracted cortical quantifiers were smoothed using a Gaussian smoothing kernel (full width at half maximum = 10) to improve the signalto-noise ratio and reduce individual differences. Therefore, a total of 340 cortical features for each subject were obtained.
After automated analysis of the images, FreeSurfer's QA Tools (https:// surfer. nmr. mgh. harva rd. edu/ fswiki/ QAToo ls) were then used to assess the quality of each reconstructed dataset. Additionally, standard neuroanatomical atlases were reviewed by two experts, who also consulted with each other, and all reconstructed datasets were visually inspected for the accuracy of alignment, cranial stripping, segmentation, and cortical surface reconstruction.

Statistical analysis of data
Based on the above methods, the cortical thickness, volume, surface area, surface curvature, and LGI of 34 subregions in each cerebral hemisphere of the three populations (NDD-CKD, DD-CKD, and NC groups) were extracted to compare the morphological indices of the brain in the different groups. The descriptions of the three groups regarding five indices in 68 brain regions of the right and left brain were expressed as either the mean ± standard deviation if the index met normality criteria, or the median (Q1, Q3) if the index did not meet normality criteria. Different methods were selected to compare the differences among the three groups of patients regarding the five indicators in 68 brain regions according to different applicable conditions: (1) if the indicator met independence, normality, and homogeneous variance, one-way analysis of variance (least significant difference method) was used, and (2) if the indicator did not meet both normality and homogeneous variance, a nonparametric test (Kruskal-Wallis H test) was used. In the comparison process, two-by-two comparisons were selected, and the P values for differences were corrected using a Bonferroni correction. Statistical significance was set at P < 0.05.
A statistical description of the clinical data and biochemical indicators of the three populations was provided as follows: continuous variables were expressed and compared referring to the above-mentioned brain morphological indicators; categorical variables were expressed as numbers (percentages) and compared by Chi-square test. Statistical significance was set at p < 0.05.
The clinical biochemical indicators that were statistically different among the three groups (12) were correlated with the brain morphological indicators that were statistically different among the three groups (58). The methods of correlation analysis were as follows: (1) two measures that conformed to a normal distribution were analyzed using Pearson's correlation; (2) two measures, at least one of which did not conform to a normal distribution, were analyzed using Spearman's correlation; and (3) categorical variables associated with numerical variables were analyzed using the Eta-squared coefficient. In multiple hypothesis testing, the false discovery rate (FDR) was used to correct the correlation analysis p-value, setting the calculated q-value threshold at 0.05. q < 0.05 was considered significant for the correlation between two variables. SPSS Statistics (version 26.0; IBM, Armonk, NY, USA) was used for all comparison analyses of variance and for the correlation analysis of categorical variables, and RStudio (version 4.0.2) was used for the correlation analysis of continuous variables.

Visualization of cortical differences
After comparing the cortical thickness, volume, surface area, and surface curvature of 68 brain regions in the NDD-CKD, DD-CKD, and NC groups, the brain regions and metrics with differences were visualized using FreeView software ( Supplementary Figures 1-4). The differences in the four cortical indicators (overall comparison P value, two-by-two comparison P value) were mapped onto the brain template, and the form of the brain template used an inflated surface (inflated surface) to exclusively exhibit the brain areas with differences (P < 0.05).

Demographic and clinical characteristics
The results revealed no significant differences in age, sex, blood potassium, or blood sodium among the three groups. Patients in the NDD-CKD and DD-CKD groups had significantly higher hypertension, diabetes mellitus, incidence of dyslipidemia, urea nitrogen, creatinine, and phosphate levels as well as significantly lower hemoglobin levels, erythrocyte pressure volume, and albumin levels than those in the NC group. Only hypertension, diabetes mellitus, incidence of dyslipidemia, urea nitrogen, creatinine, and phosphate levels were significantly different between the NDD-CKD and DD-CKD groups compared with those in the NC group. The incidence, hemoglobin, protein, and chloride levels were statistically different (Table 1).
A total of 696 correlation analyses were performed between 12 clinical variables and 58 brain morphological indicators. Twelve clinical variables, which are significantly different among three groups, were listed in Table 1: nine continuous variables (hemoglobin, erythrocyte pressure product, protein, albumin, urea nitrogen, creatinine, chloride, calcium, and phosphorus) and three categorical variables (hypertension, hyperglycemia, and dyslipidemia). The 58 brain morphological variables were shown in Figs. 1, 2, 3, 4, including 25 variables for cortical thickness, 15 variables for volume, 5 variables for surface area, and 13 variables for surface curvature.

Analysis of differences in the cortical thickness of the whole brain area between the CKD-patient and healthy-control groups
Cortical thickness in the left and right hemispheres was statistically different between the CKD-patient group and healthy NCs (Fig. 1). Cortical thickness in nine regions of the left cerebral cortex was significantly lower in patients with CKD than in NCs, including the fusiform gyrus syrinx, inferior temporal gyrus, lateral occipital cortex, lingual gyrus, paracentral gyrus, pars triangularis, postcentral gyrus, superior temporal gyrus, and transverse temporal gyrus. The thickness of the right cerebral cortex was also significantly lower than that of the NC group, including the fusiform gyrus, inferior parietal, inferior temporal gyrus, isthmus cingulate cortex, lateral occipital cortex, lateral orbital frontal cortex, lingual gyrus, middle temporal gyrus, paracentral gyrus central paracentral gyrus, pars opercularis insularis, parstriangularis trigeminalis, rostral middle frontal cephalic, superior temporal gyrus, supramarginal gyrus, frontal pole, and temporal pole; the difference was statistically significant (P < 0.05).

Analysis of differences in the cortical volume of the whole brain area between patients with CKD and healthy controls
Cortical volumes were also reduced in the CKD group compared with those in the NC group, with statistically different regions: (1) the left hemisphere including the fusiform gyrus syrinx, lateral occipital cortex, lateral orbital frontal cortex, middle temporal gyrus, pars orbitalis, and transverse temporal gyrus and (2) the right hemisphere including the bank of the superior temporal sulcus, inferior parietal cortex, inferior temporal gyrus, lateral orbital frontal cortex, lingual gyrus, middle temporal gyrus, superior temporal gyrus, supramarginal gyrus, and temporal pole (Fig. 2).

Analysis of differences in the cortical surface area of the whole brain area between patients with CKD and healthy controls
Overall, brain regions with significant differences in cortical surface area included the following: (1) the left hemisphere including the lateral orbital frontal cortex and medial orbital Fig. 1 Comparison of cortical thickness of the whole brain area between patients with CKD and healthy controls. The figure displayed the cortical thickness of 9 regions in the left hemisphere (a) and 16 regions in the right hemisphere (b). Statistical analyses were performed using two-way ANOVA (LSD) or Kruskal-Wallis H test, with *** indicating P < 0.001, ** indicating P < 0.01, and * indicating P < 0.05.
All P values were corrected by Bonferroni correction frontal cortex and (2) the right hemisphere including the inferior temporal gyrus, lingual gyrus, and middle temporal gyrus (Fig. 3). Among them, there was no significant difference between the NDD-CKD and NC groups. The surface area of the DD-CKD group was significantly lower than that of the NC group in the three brain regions on the right side; the surface area of the DD-CKD group was significantly lower than that of the NDD-CKD group in the lateral orbital frontal cortex on the left side and middle temporal gyrus on the right side.

Analysis of differences in the cortical-surface curvature of the whole brain area between patients with CKD and healthy controls
The left and right hemispheres comprise two areas of common reduction: the superior temporal gyrus and transverse temporal gyrus; areas of reduced surface curvature belonging to their respective parts also exist. In the left hemisphere, the bank of the superior temporal sulcus, entorhinal cortex, superior temporal gyrus, inferior parietal cortex, and two-way ANOVA (LSD) or Kruskal-Wallis H test, with *** indicating P < 0.001, ** indicating P < 0.01, and * indicating P < 0.05. All P values were corrected by Bonferroni correction Fig. 3 Comparison of the surface area of the whole brain regions between patients with CKD and healthy controls. The figure displayed the surface area of 2 regions in the left hemisphere (the left) and 3 regions in the right hemisphere (the right). Statistical analyses were performed using two-way ANOVA (LSD) or Kruskal-Wallis H test, with *** indicating P < 0.001, ** indicating P < 0.01, and * indicating P < 0.05. All P values were corrected by Bonferroni correction transverse temporal gyrus were statistically different among three groups. In the right hemisphere, the fusiform gyrus, lingual gyrus, pars opercularis insularis, posterior cingulate gyrus, precuneus gyrus precuneus, superior parietal gyrus, superior temporal gyrus, and transverse temporal gyrus were statistically different among three groups (P < 0.05) (Fig. 4).

Correlation analysis of morphological changes in brain and renal function
As shown in Table 1, we identified the following statistically significant variables: (1) nine continuous variables including hemoglobin, erythrocyte pressure product, protein, albumin, urea nitrogen, creatinine, chloride, calcium, and phosphorus and (2) three categorical variables such as hypertension, hyperglycemia, and dyslipidemia. We performed a correlation analysis of the above 12 variables and 58 brain morphological indices. The results of the correlation analysis between the nine continuous clinical indicators and the 58 continuous morphological brain indicators demonstrated that only six groups owned significant correlation with corrected p-values (q values) lower than 0.05, as detailed in (Table 2, Fig. 5). The absolute values of their correlation coefficients (r) ranged from 0.2 to 0.4, indicating a weak correlation between two variables. The results revealed that, in terms of blood urea nitrogen, for example, the higher the value of urea nitrogen, the lower the cortical thickness of the transverse temporal gyrus, lateral occipital cortex, and paracentral gyrus; moreover, it was lower in the CKD group than in the NC group, with the DD-CKD group exhibiting a lower value than that in the NDD-CKD group.
Correlations among the three clinical categorical variables and continuous variables of brain morphology were assessed using the Eta-squared coefficient ( Table 3). The higher the Etasquared value, the stronger the correlation. Only the groups with an Eta squared value > 0.1 were listed here, and the results demonstrated that blood glucose abnormalities correlated more significantly with cortical thickness and curvature.

Discussion
CKD is considered an independent risk factor for the development and progression of cognitive impairment, with a prevalence ranging from 16 to 38%. Although cognitive impairment can develop at any stage of CKD, patients at greatest risk are those receiving renal replacement therapy, also known as hemodialysis (Liu et al. 2021;Kelly et al. 2021). However, to our knowledge, there are relatively few studies on patients undergoing hemodialysis, mostly of whom are from European populations (Miglinas et al. 2020;Murphy et al. 2021), and on cognitive impairment and structural changes in brain tissue in dialysis patients in the East Asian population; scholars believe that patients at all stages of CKD may have cognitive deficits or more severe organic impairment, which subsequently seriously affects the prognosis of ESRD patients and exacerbates the risk of death in dialysis patients (Narasimhan et al. 2022;Querfeld et al. 2020). Whether this analyses were performed using two-way ANOVA (LSD) or Kruskal-Wallis H test, with *** indicating P < 0.001, ** indicating P < 0.01, and * indicating P < 0.05. All P values were corrected by Bonferroni correction is also true in Asian populations remains to be elucidated. This study included changes in cortical anatomy, mainly in cortical thickness, volume, surface area and surface curvature, in patients with NDD-CKD and DD-CKD in Beijing, which were compared with the NC group. The present study also examined their cognitive function and degree of renal function decompensation for correlation analysis. When compared with the NC group, NDD-CKD and DD-CKD patients showed extensive reduction in cortical thickness and volume in the paracentral gyrus, transverse temporal gyrus, and lateral occipital cortex. The latter two were more sensitive to brain alterations in CKD patients and analysis of the results showed that the reduction in cortical thickness was mainly concentrated in the fusiform gyrus, inferior temporal gyrus, lateral occipital cortex, lateral occipital cortex, lingual gyrus, and lingual gyrus, paracentral gyrus central paracentral gyrus, and superior temporal gyrus. The decrease in cortical volume was mainly concentrated in lateral orbital frontal cortex and middle temporal gyrus. Thus, it is hypothesized that cortical atrophy in patients with CKD begins in the frontal, temporal, and occipital lobes, and gradually expands in brain tissue atrophy as renal function decreases and dialysis becomes dependent. The structural alterations in their brain tissue may become the structural basis of cognitive decline in patients (Sanchez-Meza et al. 2021;Towle et al. 2021).
Uremic toxins can cause chronic inflammation, endothelial dysfunction, and vascular calcification, increasing the cerebrovascular burden associated with CKD patients. Although dialysis facilitates the removal of uremic toxins, it also triggers rapid changes in body fluid osmolality, which can exacerbate cognitive decline (Vinti et al. 2021;Yang 2021). It can also affect hematopoiesis due to its insufficient erythropoietin production, and the patient's organs have impaired iron utilization, aggravating anemia (Zhang and Parikh 2019). The latter correlates with the same brain alterations in patients. The combined results of the laboratory tests and other indicators of the patients enrolled in this study suggest that there is a significant correlation between the patients' blood creatinine, uremic toxin, and anemia and the cerebral cortical thickness and curvature of CKD patients, considering that the above biochemical indicators may be the main influencing factors of the morphological changes of the brain in CKD patients.
The present study also found that brain tissue atrophy in NDD-CKD subjects was slightly less spatially significant than in DD-CKD patients, but occurred in similar locations; that is, cortical atrophy in CKD patients also began in the frontal, temporal, and occipital lobes and gradually expanded with disease progression and dialysis dependence. This is consistent with the cross-sectional findings of Drew et al., but the former lacked data from normal controls. The mechanism of this trend is considered related to microvascular injury caused by systemic hypertension and endothelial dysfunction, which are mostly associated with focal and diffuse structural abnormalities of brain tissue, which in turn cause different degrees of cognitive impairment in patients with renal insufficiency. It is thus hypothesized that cortical thickness, volume, surface area and surface curvature may be sensitive indicators of early cortical atrophy in patients with CKD, which would contribute to the early detection and early intervention of cognitive dysfunction in patients with CKD.

Limitations
Our study had a small sample size and degree of data serendipity. In a follow-up study, we intend to continue increasing the number of patients included in the cohort for cranial MRI analysis to confirm the differences between the various groups. The current study was based on the results of only one MRI study per patient at a specific stage of the disease, thus constituting a crosssectional study and having a persistently large deficit in the observation of the dynamic course of disease progression. Further studies should take a longitudinal cohort approach to gain more insight into how structural changes in the brain tissue of patients with CKD develop dynamically over time and coincide with the decline in renal reserve function. In each plot, two histograms represented the distribution of two variables, and the scatters represented the subjects. Also, the estimated linear regression was demonstrated by the blue line with 95% confidence interval marked in grey

Conclusion
In summary, we systematically studied the cortical thickness, volume, surface area, and surface curvature of the cerebral hemispheres in the NDD-CKD, DD-CKD, and NC groups using neuroimaging. Patients with CKD with or without dialysis dependence exhibited a certain degree of structural brain-tissue imaging changes, with more significant decreases in cortical thickness and curvature, as well as decreased cortical thickness in some brain regions related to cognitive function.