Altered Functional Connectivity Density in Young Survivors of Acute Lymphoblastic Leukemia Using Resting-State fMRI

Introduction

Acute lymphoblastic leukemia (ALL) is the most common malignancy in childhood. The application of chemotherapy has greatly improved the event-free survival rate and outcomes over the last few decades. However, chemotherapy drugs such as methotrexate have been shown to be neurotoxic, with disruption of astrocytes, inflammatory responses and microvascular damage.¹ Accumulating evidence has shown that chemotherapy-related cognitive impairment (CRCI) is a non-negligible issue for long-term survival states, as there are studies concerned with the deterioration of intelligence, executive function and processing speeds of adult survivors.^2,3 The childhood brain is at an important stage of growth and development for myelination, synaptogenesis and the establishment of complex neuronal connections. Resting-state magnetic resonance imaging, whether functional (fMRI) or structural, has been widely used as an objective method to understand the possible neuroimaging basis of cognitive deficits.^4–7 Survivors of ALL who have undergone chemotherapy show a reduction in fractional anisotropy,⁸ smaller cortical surface area,⁵ smaller white matter volumes^9,10 and altered spontaneous activity in the hippocampus,¹¹ accompanied by poor neurocognitive performance.

However, to date, there have been few studies aimed at measuring global functional changes in juvenile survivors of ALL. We used functional connectivity density (FCD) mapping, as a known method in relation to the spontaneous release of synaptic neurotransmitters,^12,13 aiming to determine the widespread alterations in brain function when cognitive deficits occur. Therefore, any abnormality in neural activity, such as apoptosis, and developmental, plastic, and degenerative mechanisms, would result in a change in FCD. Further, we sought to explore the relationship between the disorder of FCD and cognitive impairment in young survivors of ALL and the probable functional manifestations.

Materials and Methods

Subjects

In total, 24 young survivors of ALL and 22 age-matched healthy controls (HCs) were recruited into the study. All of the patients, who had completed standard-risk treatment based on Chinese Children Leukemia Group (CCLG)-ALL 2008, were included according to the following criteria: 1) were off-therapy for at least 12 months at the time of enrollment; 2) had received a standard chemotherapy regimen; and 3) had no history of cranial radiation, CNS involvement or gross neuropathologies. All of the HCs were recruited through community postings, and had no major sensory impairments, MRI contraindications or any significant cognitive disorders.

The results of neurocognitive scales for ALL patients and HCs, such as the intelligence quotient (IQ) (full scale of the Wechsler Adult Intelligence Scale, Fourth Edition [WAIS-IV]), Trail Making Test part A (TMT A), and Wechsler Memory Scale Digit Span Test (Forward) (DST), were collected by two experienced psychiatrists.

This study was conducted in accordance with the Declaration of Helsinki, and the Human Research Ethics Committee of the First Affiliated Hospital of Nanchang University approved the study protocol., Because all of our participants were under the age of 18 years, written informed consent was obtained from a parent or legal guardian of each participant prior to the study.

MRI Data Acquisition

The fMRI data were acquired on the same Siemens Trio 3.0 T scanner, implementing an eight-channel phased-array head coil, in the First Affiliated Hospital of Nanchang University, China. Each subject was fixed comfortably by a belt and foam pads during resting-state fMRI (rs-fMRI) scanning, keeping the supine position and with the head in a neutral position, eyes closed and awake. The rs-fMRI data were acquired with two sequences: a gradient-echo echo-planar imaging (EPI) sequence of 240 volumes, and high-resolution brain structural images using a T1-weighted 3D magnetization-prepared rapid gradient-echo (MP-RAGE) sequence. The parameters of the EPI sequence were: repetition time (TR)=2,000 ms, echo time (TE)=40 ms, flip angle=90°, slice thickness/gap=4.0/1 mm, field of view (FOV)=240×240 mm, in-plane resolution=64×64 and 30 axial slices to cover the whole brain; scanning time was 8 minutes. The parameters of the 3D MP-RAGE sequence were: TR=1,900 ms, TE=2.26 ms, flip angle=9°, matrix=256×256, FOV=240×240 mm, thickness=1.0 mm and 176 sagittal slices; scanning time was 3 minutes.

fMRI Data Preprocessing

Preprocessing of fMRI data was carried out using Data Processing Assistant for rsfMRI (DPARSF, http://rfmri.org/dpabi) in the MATLAB (MathWorks Inc., Natick, MA, USA) working platform. The first ten volumes of each data set were discarded for signal equilibrium. Then, slice-timing correction of temporal differences was applied to the remaining 230 volumes. The resulting images were realigned according to the first volume following a criterion of head motion within 3 mm translations and 3° rotations in any axis. Subjects were excluded if their framewise displacement (FD) exceeded 0.3, where FD was calculated as an index of head motion. Four ALL patients and four HCs were eventually excluded because of head motion parameters. Afterward, the realigned functional data underwent co-registration based on their respective high-resolution T1-weighted structural images, then were spatially normalized to standard Montreal Neurological Institute (MNI) space with a resampled voxel size of 3×3×3 mm³, having regressed out the nuisance variables. Finally, band-pass filtering (0.01–0.08 Hz) was performed to reduce low-frequency drifts and high-frequency physiological noise.

Resting-State FCD Calculation

Long- and short-range FCD calculations were carried out on Graph Theoretical Network Analysis (GRETNA, http://www.nitrc.org/projects/gretna/). The correlation coefficient threshold (Tc) was set to 0.25 to reduce the false-positive rate and lower the sensitivity. The rate (r) of the correlation coefficient was calculated through Pearson correlation between time-varying signals of a certain voxel and every other voxel in the gray matter (GM); these two voxels could be considered to be relative if r>0.25. The threshold of the signal-to-noise ratio (SNR) was set as 0.5 within the GM regions to minimize unwanted susceptibility-related signal loss. The calculation result of each voxel made up the global FCD, which means the functional connection density distribution of the whole brain (Tomasi et al, 2010).¹⁴ According to He,¹⁵ a connection within a given anatomical distance of 75 mm can be considered an intraregional effective functional connection, which makes up the short-range FCD. Then, the long-range FCD exists outside the region, equaling the difference between the global FCD and long-range FCD. In order for the result to be in a normal distribution, the long- and short-range FCD maps of each dataitem were rescaled by the individual average FCD value and converted to Z scores. Finally, the normalized FCD data were spatially smoothed with a Gaussian kernel of 6×6×6 mm³ depending on full width at half maximum (FWHM), in order to increase the SNR and eliminate artifacts.

Statistical Analysis for FCD

First, the spatial distribution of mean short- and long-range FCD was estimated in the ALL group and HC group, respectively. Then, general linear model (GLM) analysis of between-group FCD comparisons was performed in DPARSF, while nuisance covariates such as age, gender and average FD were included. Two-tailed Gaussian random field (GRF) correction was used to reduce the false-positive rate, with a corrected threshold of voxel p-value <0.01, cluster p-value <0.05 and a GM mask. The mean short- and long-range FCD in clusters with significant differences between groups were further estimated by Cohen’s d to determine the effect size (ES) (Mégevand et al, 2014).¹⁶ Moreover, a partial correlation analysis, with age and gender as covariates of no interest, was performed to investigate whether the FCD in the regions showing significant group differences correlated with the outcomes of neurocognitive scales.

ROI-Based FC Calculation and Statistical Analysis

Region of interest (ROI)-based FC analysis was performed on the ROIs that showed significant differences between the ALL and HC groups in the long-range FCD. The mean time course for each ROI was extracted using DPABI (http://rfmri.org/DPARSF) and Pearson correlation coefficients of the mean time courses were calculated between pairwise ROIs for each subject. After Fisher’s r-to-z transformation, calculations were performed in the following procedure. First, we used the FC values for between-group comparison only if the FC values extracted from the ROIs between ALL patients and HCs were statistically significant within each group after a random effects one-sample t-test in SPM8. Next, the z-values for all ROIs of the ALL and HC groups were evaluated by a GLM analysis in SPM8 after controlling for age and gender with the GRF correction (voxel p<0.01, cluster p<0.05). The ES was also estimated by Cohen’s d to determine the significant differences between groups. Finally, we assessed the correlations between values of altered long- and short-range FCD and ROI-based rsFC in ALL patients with clinical data by partial correlation analysis.

Results

Demographic and Clinical Data

Table 1shows that the mean age of young ALL survivors was 10.15±3.1 years, and the mean duration off therapy ranged from 12 to 18 months. There were no significant differences between the two groups in age, gender or handedness.

Table 1 General Clinical Information for Young Survivors of Acute Lymphoblastic Leukemia (ALL) and Healthy Controls (HCs)

Spatial Distribution of FCD

The group-averaged FCD maps showed similar spatial patterns in the ALL patients and HCs. There was a wide spatial extent of both long-range FCD and short-range FCD in the temporal, parietal and limbic lobes, and the cuneus. Besides, short-range FCD was higher in the precuneus and cingulate, and long-range FCD was observed to be higher in the postcentral, precentral and medial frontal gyrus (Figure 1).

Figure 1 Spatial distribution of short- and long-range FCD in ALL and HCs, visualized with the BrainNet Viewer (http://www.nitrc.org/projects/bnv/). Abbreviations: FCD, functional connectivity density; L, left; R, right; ALL, acute lymphoblastic leukemia; HC, healthy control; k/k0, normalized FCD.

Group Differences in FCD

A between-group comparison was conducted to reveal differences in long-range FCD and short-range FCD (Table 2; Figure 2). Compared with the HCs, decreased regions in the long-range FCD map were the bilateral lingual gyrus, cingulate cortex, parahippocampal gyrus and right calcarine fissure. Simultaneously, decreased regions in the short-range FCD map were the bilateral cuneus, lingual gyrus, cingulate cortex, hippocampal gyrus and precuneus (GRF correlation, cluster p<0.01, voxel p<0.05).

Table 2 Differences in Short-Range FCD and Long-Range FCD Between the ALL and HC Groups (GRF Correction, Voxel-Wise p<0.01, Cluster-Wise p<0.05)

Figure 2 Brain regions with significant changes in short- or long-range FCD in ALL (voxel p<0.01, cluster p<0.05, GRF corrected), visualized with the DPABI slice viewer (http://rfmri.org/dpabi). Notes: A negative T value means that the resting-state FCD in ALL is smaller than in HCs. Significant group differences in the mean strength of the short- (above) and long-range (below) FCD in the above regions (p<0.05). The error bars are standard errors of the mean.Abbreviations: FCD, functional connectivity density; ALL, acute lymphoblastic leukemia; HC, healthy control; GRF, Gaussian random field; LING, lingual gyrus; CC, cingulate cortex; HIP, hippocampal gyrus; R, right; CAL, calcarine fissure and surrounding cortex; CUN, cuneus; PHG, parahippocampal gyrus; preCUN, precuneus; ES, effect size.

Alteration in FC Based on Seed Region

The region of the bilateral lingual gyrus, cingulate cortex, parahippocampal gyrus and right calcarine fissure, which showed a decreased long-range FCD, was defined as an ROI. Table 3 and Figure 3 show the between-group results of FC: there was significantly increased FC between ROI and the posterior cerebellar lobe, and decreased FC between ROI and the middle occipital gyrus, cuneus and lingual gyrus in ALL patients.

Table 3 Differences in Functional Connectivity Between the ALL and HC Groups (GRF Correction, Voxel-Wise p<0.01, Cluster-Wise p<0.05)

Figure 3 Brain regions with significant changes in rsFC with the bilateral CUN/LING/CC/PHG/preCUN in ALL (voxel p<0.01, cluster p<0.05, corrected), visualized with the DPABI slice viewer (http://rfmri.org/dpabi). Note: Significant group differences in the mean strength of rsFC in the above regions (p<0.05). The error bars are standard errors of the mean.Abbreviations: rsFC, resting-state functional connectivity; ALL, acute lymphoblastic leukemia; HC, healthy control; MOG, middle occipital gyrus; CUN, cuneus; LING, lingual gyrus; CbPL, posterior cerebellar lobe.

Correlations

The decreased short-range FCD value of ALL patients was positively correlated with the score on the DST (r=0.5, p<0.05) and the increased rsFC value was negatively correlated with the score on the TMT A (r=0.7, p<0.05), as shown in Figure 4.

Figure 4 Relationships between clinical indices and FCD and rsFC value in ALL. Note: The score on the DST was positively correlated with the increased rsFC value, and the score on TMT A was negatively correlated with the decreased FCD value of ALL patients. Abreviations: FCD, functional connectivity density; rsFC, resting-state functional connectivity; ALL, acute lymphoblastic leukemia; DST, Digit Span Test; TMT A, Trail Making Test A; LING, lingual gyrus; CC, cingulate cortex; PHG, parahippocampal gyrus; R, right; CAL, calcarine fissure and surrounding cortex; CbPL, posterior cerebellar lobe.

Discussion

As far as we know, this is the first study to apply FCD mapping to investigate the altered spontaneous neuronal activity of young survivors of ALL. Our results indicate that there is some overlap between long- and short-range FCD in the lingual gyrus and cingulate cortex, which reflects the considerable degree of functional disruption in these regions. Besides, altered FCD occurred in the cingulate cortex, parahippocampal and hippocampal gyrus, right calcarine fissure, cuneus and precuneus, indicating that chemosensitivity or toxicity may be more pronounced in the posterior region of the brain. Previous studies were more concerned with structural changes in ALL showing widespread alterations.^5,6 Genschaft et al found smaller volumes in the left calcarine gyrus, both lingual gyri and the left precuneus, smaller hippocampal volumes and lower figural memory.¹⁷ Therefore, structural abnormalities probably give rise to local and remote FCD changes.

Previous studies suggested that receiving chemotherapy at a younger age may be predictive of poor neurocognitive outcomes.^18,19 A decline in the cognitive performance of cancer survivors was confirmed in intelligence, attention, working memory, executive function and other aspects.¹⁸ For patients in our study, the poor performance on the DST was negatively correlated with the short-range FCD value of the bilateral lingual gyrus, cingulate cortex, hippocampal gyrus and right calcarine fissure. The DST is a measure of working memory, especially short-term memory and attention. Impaired neurocognitive function of ALL patients can be measured using this neurocognitive scale.

The pathophysiology of methotrexate-induced neurotoxicity is associated with oxidative stress injury of vascular endothelial cells and neurons, diffuse demyelination and myelin formation disorders of white matter.^2,3 CRCI caused by other commonly used drugs is considered to be related to glial progenitor cell death, destruction of the microvasculature and so on.^20–22 A previous study clarified the hippocampal toxicity of chemotherapy and hippocampal-dependent brain plasticity.²³ Considering the important role of the hippocampus in memory formation, memory maintenance, learning and anxiety, Monje and Dietrich reported reduced hippocampal atrophy and increased BOLD signal in the hippocampus during an unsuccessful encoding task.¹¹ In the resting state, Kesler et al found altered clustered connectivity in hippocampal and occipital regions, and lower IQ was related to decreased FC between the left hippocampus and left lingual gyrus.²⁴ In addition to decreased hippocampal volumes with lower figural memory in Genschaft’s study, another group of survivors of ALL showed decreased hippocampal GM with reduced verbal learning/memory.²⁵ Combined with our result of suppressed spontaneous hippocampal activity, in proportion to working memory, we can hypothesize that maladaptive hippocampal functioning and disordered functional networks may be the pathophysiological cause of memory impairment.

In a previous study, Tomasi and Volkow indicated that the precuneus was the region with the highest short-range FCDs and significant long-range FCDs in healthy people.²⁶ The precuneus, the key node (hub) in the default brain functional network, which is associated with conscious recall of short-term memory, is vulnerable to oxidative stress injury. Young survivors of ALL in our study showed decreased long-range FCD in the precuneus. Research using voxel-based morphometry revealed decreased GM volume in the precuneus.²⁸ On the other hand, functional and structural research in brain cancer patients could be included for reference, since CRCI has also been reported in breast cancer patients after chemotherapy hwhen neurotoxic drugs such as methotrexate and fluorouracil were included in the regimen. For instance, Conroy et al revealed lower working memory-related activation in the right precuneus in breast cancer patients.²⁸ Wang et al demonstrated that there was compensatory increased FC between the precuneus and dorsolateral prefrontal cortex.²⁹ Contradictory findings may be related to altered connectivity and confusion of physiological functions in the precuneus after chemotherapy.

With the ROI of altered long-range FCD regions, we also documented significantly increased FC in the posterior lobe of the cerebellum (lobule crus I/VIIIa/VIIIb/IX). Lobule VIIIa of the cerebellum is a region associated with visuospatial attention and working memory, and patients with lesions to VIIIa (and to crus I and crus II) showed attention disorder.³⁰ Similar enhanced activation was also reported in breast cancer survivors, who showed increased cerebral blood flow of the posterior cerebellum during a short-term word-pair association task using positron emission tomography (PET)³¹ and increased BOLD signal in the resting state using fMRI.³² In early imaging work by Kingma et al, children with ALL after chemotherapy showed a volume reduction in the cerebellum.³³ Our results indicated a correlation between increased FC in the posterior lobe of the cerebellum and patients’ performance on the TMT A, so we supposed that the increased activation in the cerebellum was a kind of compensation in ALL patients.

Thus, we used the method of FCD mapping to characterize the effects of chemotherapy on brain functional networks in young survivors of ALL. The objective description of neuroimaging and brain activity could help us to understand better the unique pattern of cognitive impairment associated with chemotherapy and the potential mechanisms for compensation. However, there are two limitations to our study. First, the sample size is relatively small, which may increase false-positive and false-negative findings. Secondly, the cognitive function scale adopted did not cover all aspects of neurocognitive function.

Conclusion

Our results suggest the altered functional connectivity of young survivors of ALL in the posterior region of the cerebrum and posterior lobe of the cerebellum. Alterations in spontaneous neuronal activity seem to parallel the neurocognitive testing, which indicates that rs-fMRI could be used as a neuroimaging marker for neurological impairment in ALL patients.