Recovery from the damage of cranial radiation modulated by memantine, an NMDA receptor antagonist, combined with hyperbaric oxygen therapy

Abstract

Background

Radiotherapy is an important treatment option for central nervous system malignancies. However, cranial radiation induces hippocampal dysfunction and white matter injury; this leads to cognitive dysfunction, and results in a reduced quality of life in patients. Excitatory glutamate signaling through N-methyl-d-aspartate receptors (NMDARs) plays a central role both in hippocampal neurogenesis and in the myelination of oligodendrocytes in the cerebrum.

Methods

We provide a method for quantifying neurogenesis in human subjects in live brain during cancer therapy. Neuroimaging using originally created behavioral tasks was employed to examine human hippocampal memory pathway in patients with brain disorders.

Results

Treatment with memantine, a non-competitive NMDAR antagonist, reversed impairment in hippocampal pattern separation networks as detected by functional magnetic resonance imaging. Hyperbaric preconditioning of the patients just before radiotherapy with memantine mostly reversed white matter injury as detected by whole brain analysis with Tract-Based Spatial Statics. Neuromodulation combined with the administration of hyperbaric oxygen therapy and memantine during radiotherapy facilitated the restoration of hippocampal function and white matter integrity, and improved higher cognitive function in patients receiving cranial radiation.

Conclusions

The method described herein, for diagnosis of hippocampal dysfunction, and therapeutic intervention can be utilized to restore some of the cognitive decline experienced by patients who have received cranial radiation. The underlying mechanism of restoration is the production of new neurons, which enhances functionality in pattern separation networks in the hippocampi, resulting in an increase in cognitive score, and restoration of microstructural integrity of white matter tracts revealed by Tract-Based Spatial Statics Analysis.

Key Point.

• The method for therapy and diagnosis of hippocampal function we developed can be applicable to the patients received cranial radiation so as to restore the cognitive decline.

Importance of the Study.

Using our simple, non-invasive method the patient puts on goggles that display slides in the MRI imaging room and the correct answer rate for distinguishing lure pairs, shown as images that are slightly different with completely new tasks, is assessed. In addition, blood-oxygenation-level-dependent (BOLD) reactions were also visualized as images to successfully evaluate human hippocampal function and visualize signal transmission within the hippocampus. Supplementary movies are included that show images from healthy subjects and irradiated patients. We established a method for treatment recommended as per the classification stage of neurogenesis in the human hippocampus (4 stages from normal to seriously disordered, detailed presented in Supplementary Figure S12). Importantly, while postoperative radiotherapy resulted in decreased neurogenesis, hippocampal neurogenesis of the patient recovered with treatment.

Adult hippocampal neurogenesis has been demonstrated throughout the lifespan in humans, rodents, and marmosets; ^1,2 cancer therapy-related cognitive impairment and hippocampal neurotoxicity is now a focus of attention.^3,4 Accumulating evidence reveals cognition impairment induced by cranial radiation results from dysfunctional hippocampal neurogenesis,⁵ and cerebral white matter damage.^6,7 Pathologically upregulated excitatory glutamate signaling through the N-methyl-d-aspartate receptor (NMDAR) plays a central role in the pathogenesis of neurological diseases such as brain tumors, strokes, Alzheimer’s disease, and epilepsy.^8,9 Memantine, a non-competitive antagonist restores Doublecortin (DCX) positive neural precursor cells in the hippocampus and oligodendrocytes (myelin sheath-forming cells) in the cerebrum. In preclinical models and human clinical trials, memantine exhibited neuroprotective effects and prevented pathological NMDAR-mediated signaling.^10,11 Oxygen also regulates neural stem cells and oligodendrocytes in development and disease.¹² Preconditioning hyperbaric oxygen (HBO) attenuates postoperative cognitive impairment in aged rats ¹³ and stimulates peripheral nerve sheath regeneration by attenuating pro-inflammatory cytokine production.¹⁴ In humans, HBO induces a relative regional cerebral blood flow (rCBF) increase in areas encompassed in the dorsal attention pathway of the superior frontal, ventral premotor, parietal, and middle temporal cortex in the left hemisphere, and in default mode networks including left cuneus, calcarine cortex, and posterior cingulated cortex; this suggests a possible beneficial effect for improving externally directed cognitive performance and internally directed cognitive state in the resting-state.¹⁵

We show here that HBO preconditioning (2.5 atm, 100% oxygen, 40 min) ¹⁶ and NMDAR antagonist use concomitant with radiotherapy effectively ameliorates hippocampal dysfunction and preserves white matter integrity.

Methods

Details of the methods are provided in the Supplemental Methods.

Patient and Healthy Volunteer Recruitment

A total number of 173 patients including 61 patients with malignant tumor receiving radiotherapy, and 36 healthy volunteers were participated in this study (clinical and demographic information were shown in Supplementary Figure S2 and Table S1). We examined the safety, feasibility and restorative effect on cognitive function by HBO and memantine combined with RT. The trial were conducted in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki. This study was approved by the ethical committee of the University of the Ryukyus (Code No. 99 and 510); written informed consent was obtained from all participants.

Immunohistochemistry

Mice were deeply anesthetized with isoflurane, and perfused transcardially with 4% paraformaldehyde. After deparaffinization and blocking, tissues were incubated with primary antibodies at 4°C overnight. Immunohistochemical staining was performed using the high polymer (HISTOFINE simple stain, NICHIREI, Japan) method.

Golgi Stain

Golgi-staining was performed using FD Rapid Golgistain Kits (FD NeuroTechnologies, INC. MD21041, USA).

Behavioral Experiment

The detailed experimental protocol is presented in Supplementary Figure S1

Radiation Treatment in Human Patient

Radiotherapy plans were generated using the RayStation treatment planning software (RaySearch Laboratories, Stockholm, Sweden) and performed by Radixact tomotherapy delivery systems (Accuray, Incorporated, Sunnyvale, CA, USA). The irradiation dose to the human hippocampus was also calculated utilizing irradiation dose distribution images created by Raystation for the treatment (Supplementary Figure S8D).

Event-Related fMRI Study

We used an fMRI behavioral task to assess the hippocampal pattern separation function (Supplementary Figure S3). The experimental paradigm previously we reported was adapted ¹⁷ with permission.

Analysis of Neurocognitive Function

Eight tests: (I) mini-mental state examination (MMSE), modified MMSE (3MS), and Hasegawa Dementia Scale-Revised (HDS-R) for global cognitive screening, (II) Trail Making Test (TMT) and Stroop test (ST) for executive function, (III) Wechsler Adult Intelligence Scale-Revised (WAIS-R) digit span subtest (DS) for working memory, (IV) WAIS-R digit symbol test (DST) for psychomotor speed, (V) partial WAIS-R block design subtest (fifth and ninth items) and the cube-copying test for visuospatial ability, were evaluated for 119 preoperative patients (Supplementary Table S12).

Human Tract-Based Spatial Statics (TBSS) Analysis

We performed whole brain analysis with TBSS ^18,19 to examine eight association and projection fibers of the unaffected side including anterior thalamic radiation (ATR), cingulum cingulate gyrus part (CGC), corticospinal tract (CST), cingulum hippocampal part (CGH), inferior front-occipital fasciculus (IFOF), inferior longitudinal fasciculus (ILF), superior longitudinal fasciculus (SLF), and uncinate fasciculus (UNC) in the contra-lesional hemisphere.

Results

Rodent Model of Cranial Radiotoxicity

In primary brain tumors, extended local irradiation with intensity-modulated radiotherapy ²⁰ is commonly applied. In this study, we used whole brain irradiation fields instead of the hippocampal target fields used in rodent models.⁵ We developed a rodent fractionated whole cranial radiation model to reflect cranial radiotherapy-induced toxicity in humans, instead of localized irradiation of the hippocampus for the ablation of hippocampal neurogenesis (Figure 1A).

Fig. 1.

Irradiated neurons affected by HBO and memantine treatment. (A) Apparatus for hyperbaric oxygen (HBO) and radiation therapy (RT) for humans and rodents, and time line of the experiment. (B) Immunohistochemical analysis of Doublecortin (DCX) in mice dentate gyrus (DG) in the control, RT, HBO, memantine (M), RT + M, RT + HBO and RT + M + HBO groups. Scale bar: 20 (left), 100 (right) µm. (C) Bromodeoxyuridine staining, and DCX-positive cells/HPF ×20 in hippocampi of mice (representing mean values of each four independent experiment). (D) Golgi-staining cells in the subgranular layers of the hippocampal DG (n = 4) in the control, RT, RT + memantine (M), and RT + HBO groups. Scale bar indicates 2, 50, 200 µm, in higher, intermediate, lower magnification images. Number of thin-, stubby-, and mushroom-type spine/10 μm in each groups (right lower column). “total”: “Number of thin + stubby + mushroom-type spine/10 μm”. *P < .05, **P < .01, ***P < .001.

Memantine Reverses Radiation-Induced Hippocampal Neurotoxicity; HBO Stimulates Myelin Remodeling

To detect the early events of hippocampal neurogenesis after exposure to ionizing radiation, we performed a histological examination in bromodeoxyuridine (BrdU)-labeled specimens; DNA in the S-phase was analyzed in animals treated with fractionated whole brain irradiation (2 Gy/day/fraction for five consecutive days, total dose: 10 Gy/5 fractions) (four independent experiments, three mice were used per one experiment) (n = 12) or concomitant HBO preconditioning prior to radiation (n = 12), and/or memantine (5 mg/kg for 5 consecutive days) (n = 12); these data were compared with data from non-irradiated mice (n = 12). Specimens were prepared 2 days after the final treatment. Memantine also induced migratory or dividing morphology in DCX-positive neuronal progenitor cells (Figure 1B). These cells migrated horizontally in the subgranular layer, and vertically throughout the granular layers. Oxygenation influences differentiation and proliferation in neural stem cells (NSC). In vitro, hypo-oxygenation (<5%) induces self-renewal in type-B NSCs, while hyper- oxygenation (>20%) induces neuronal maturation.¹² There were significantly fewer BrdU-positive cells in the irradiated group compared to the control group (P < .001); in the irradiated + memantine group, DNA synthesis increased and was comparable to that in controls (baseline values) (Figure 1C). As evidenced by BrdU-labeling experiments, HBO alone did not stimulate DNA synthesis compared to controls (Figure 1C). DCX-positive cells existed in the subgranular layers (Figure 1B, C). Moreover, in the HBO + memantine group, bipolar migratory cells frequently induced by memantine-alone treatment were absent, as were migratory and dividing cells that were observed in irradiated mice treated with memantine (Figure 1B). The BrdU-labeling indices in the Radiation Therapy (RT) + HBO + memantine group were significantly increased compared to those in the RT alone group (P < .001), but not changed compared with the RT + memantine group. Memantine treatment prevented the loss of dendritic spines in the hippocampal neurons of irradiated mice. While RT lowered the number of Golgi-staining cells in the subgranular layer, this number was restored to the non-irradiated control level with memantine treatment. Moreover, spine retraction and shrinkage caused by RT were reversed by memantine treatment (P < .01) (Figure 1D).

Next, we investigated whether treatment with HBO or memantine ameliorates radiation-induced white matter damage. Because excessive glutamate signaling through oligodendrocyte NMDARs mediates axon-to-myelin signaling,¹² we speculated that NMDAR activation by excessive glutamate generated due to cranial irradiation can cause acute excitotoxicity in myelin-forming oligodendrocytes. Indeed, cranial radiation exerted neurotoxicity towards white matter tracts (Figure 2A). HBO renewed myelin remodeling more effectively than memantine, as estimated by myelin basic protein expression (myelin sheathing marker in oligodendrocytes) (Figure 2B). We also found that radiation caused severe morphological alterations to pyramidal neurons as evidenced by immunoreactivity towards high molecular neurofilament protein (NFP-200kDa) antibody (anti-neurofilament H 200kD antibodies; AHP2259GA; polyclonal, Bio-Rad, CA, USA). We also observed marked perikaryal degeneration and axonal degeneration and deformities; both memantine and HBO ameliorated axonal degeneration (HBO more effective than memantine) (Figure 2C, D). Therefore, HBO efficaciously repairs white matter damage and restores cerebral neurons, while memantine effectively combats radiation toxicity towards hippocampal neurogenesis (Figures 1D and 2E). Dendritic morphology was examined by Golgi-staining. In memantine-treated irradiated mice, degenerative and retracted spine morphologies and sparse dendritic spine densities showed significant recovery and were comparable to those in non-irradiated animals (Figure 2E). This indicates that memantine preserves neuronal morphology and dendritic spine density and promotes neurogenesis, and thus shows neurotrophic effects (Figures 1D and 2E).

Fig. 2.

Effect of HBO and memantine on axons, the myelin sheath, and dendritic spines. (A) Immunohistochemical expression of NF-200kDa in the dentate gyrus (DG) of the control, radiation therapy (RT), RT + memantine (M), and RT + hyperbaric oxygen (HBO) groups. (B) Immunohistochemical expression of MBP in mice brain sections indicated by a white square in each groups. (C) Higher magnification of the area demarcated by a white box in panel a. Scale bar: 200 µm in (A), 50 µm in (B) and (C). (D) Semiquantitative analysis of MBP expression (right column), and morphometry of axonal length (center column) in each treatment group. “axonal length (b/a)” was defined as “length of axon (b)/maximal radius of cell body (a)”. (E) Golgi-staining in the retrosplenial dysgranular cortex of mice in each treatment group. Red-boxed area was magnified from left to right in each treatment group. Scale bar: 200 µm, 50 µm and 10 µm from left, mid and right images, respectively. Number of thin-, stubby-, mushroom-type spine/10 μm in each groups (lower column). “total”: “Number of thin + stubby + mushroom-type spine/10 μm”. *P < .05, **P < .01, ***P < .001.

Increased Neurogenesis Produces a Behavioral Response in Mice with Radiation Toxicity

Ionizing radiation affects associative memory and anxiety.²¹ The hippocampus plays an important role in the integration of emotion and cognition.²² Using elevated plus maze experiments, we assessed whether ionizing radiation influences anxiety-like behavior (Figure 3A). The time and the frequency of visits to and time spent in the open arm were higher in irradiated mice. In context fear conditioning test, freezing behavior of HBO treated mice group were ameliorated most (Figure 3B). Increased adult hippocampal neurogenesis leads to improvements in pattern separation in rodent models.²³ We next examined whether the modulation of hippocampal neurogenesis by memantine, HBO treatment, or a combination of the two can improve behavioral responses in mice with radiation toxicity. To this end, pattern separation ability was evaluated using a delayed non-matching to place protocol in the radial arm maze while varying the angles between the sample and reward arms (Figure 3C). The sample and reward arms were separated by 45°, 90° and 135° in separation 1–3 cases, respectively (Figure 3D). We found that memantine treatment led to increased correct choice selection in only the separation 3 experiment, though without statistical significance (n = 6) (Figure 3E). The 2-object novel object recognition task was used to evaluate recognition memory (Figure 3F); a 10-min exposure phase and a 5-min test phase were used. Mice receiving cranial radiation showed reduced exploratory behavior. A significant increase in the novel object exploration time compared to the familiar object exploration time was observed in the control and RT + memantine groups (Figure 3G). Memantine was reported to induce the neurogenesis in the hippocampal dentate gyrus (DG) region.²⁴ Our data would indicate neurogenesis in the DG region play a crucial role for the protection to the RT-induced NMDAR inactivation (Supplementary Figures S10 and S11).

Fig. 3.

Increased neurogenesis produces a behavioral response in mice subjected to radiation toxicity. (A) Anxiety-like behavior was assessed using the elevated plus maze test. (B) Context fear conditioning test. (C–E) Pattern separation ability was tested using a delayed non-matching to place task protocol in a radial arm maze in which the angle between the sample and reward arms was varied during the experiment (S: start-arm; R: reward-arm) separated by 45°, 90°, 135° in separations 1–3, respectively (D). Proportion of correct choice responses for each angle (E). (F, G) Schematic presentations of the two-object novel object recognition task (F) and results (G). Control (n = 6), radiation therapy (RT) (n = 6), RT + memantine (M) (n = 6), the RT + hyperbaric oxygen (HBO) (n = 6), RT + HBO + M groups (n = 6). *P < .05, **P < .01, ***P < .001.

Non-invasive Neuroimaging of Human Hippocampal Neurogenesis

The investigation of pattern separation by fMRI is a unique non-invasive approach for evaluating the function of the DG in the hippocampus, and can be applied in human health and disease.^17,25 In rodents, young granule cell (GC) neurons in the DG mediate pattern separation processes to help discriminate between similar, overlapping, and closely related images, as well as episodes or spatial configurations. Old GC neurons aid in pattern completion and confer retrieval ability to create explicit memory from partial cues from stored memory or preexisting knowledge via the EC (entorhinal cortex)/DG/CA3 (cornu ammonis) pathway.¹⁶ However, this aspect is yet unclear in humans.

To fill the gaps in our understanding of adult neurogenesis in rodents and humans, we developed a novel strategy based on the assessment of the generation and ablation dynamics of young neurons. This strategy involves the elucidation and monitoring of hippocampal function during cranial radiation administration in patients with brain tumors; these data are compared with data from young healthy volunteers [approved by the ethical committee of the University of the Ryukyus (No.99, 510)]. We used a previously described fMRI behavioral paradigm (Supplementary Figure S3) ¹⁷ in our study. A significant decline in the correct response rates was observed during lure tasks in patients receiving radiotherapy. Cranial irradiation therapy (cIR) was performed in 61 patients (Supplementary Figure S2), among them 13 people were irradiated on both sides of the hippocampus (Supplementary Table S5), with a mean irradiation dose to hippocampus of 22.7 ± 18.3 Gy (0.5–60 Gy). The correct response percentage for the lure task in patients before cIR was 38.0 ± 19.0% (n = 13, age: 28.7 ± 11.3 years; 13–41 years), which was reduced to 25.0 ± 27 % (n = 13, age 29.0 ± 10.5 years; 13–41 years) after cIR. In comparison, the correct response percentage (lure task) in healthy volunteers was 47.0 ± 19.0% (n = 36, age 25.2 ± 2.9 years; 23–35 years) (Supplementary Table S2), and that in the non-irradiated group was 43.0 ± 21.0% (n = 31, age 24.6 ± 8.1 years; 7–38 years) (Supplementary Table S3). Importantly, analysis of the correct response rate for the new and repeated tasks (but not for the lure task) revealed no significant differences across groups (Figure 4A). Furthermore a significant difference was found in the lure correct response rate of the healthy group and the irradiated patient group (P < .01, Mann-Whitney U-test), the benign tumor group and the irradiated patient group (P < .05, Mann-Whitney U-test) and before and after the irradiation group (P < .05, Mann-Whitney U-test) (Figure 4A). Figure 4B shows the relationship between the correct answer rate and radiation doses to the right or left hippocampus in individual patients. Decreased incorrect response rates in the lure task were observed when the mean radiation dose to right hippocampus reached 8.5 ± 5.0 Gy (0.5–6.2 Gy; n = 10), and that to the left hippocampus reached 7.9 ± 5.4 Gy (0.5–16.0 Gy; n = 10) (Supplementary Tables S1 and S6). Four among 10 patients with left hippocampal exposure to ionizing radiation, and four among 10 patients with right hippocampal exposure to radiation exhibited a 0 % correct response score in the lure task, indicating complete ablation of neurogenesis (Figure 4B). Five patients received ionizing radiation to both the right and left hippocampi, and were included in the right and left hippocampal exposed groups as well. In this overlapping group, we found two patients with an increased correct response score of lure task after radiation therapy, with one patient increasing from 56 % before treatment to 81 % after treatment and another from 6 to 13%. Two other patients in the group with exposure to both the right and left hippocampi showed a decrease in the lure task score with radiation within 2 Gy, indicating high sensitivity to radiotoxicity (Figure 4B). Although sensitivity towards hippocampal toxicity was not uniform among individuals, we found no differences in laterality and dominance regarding radiation exposure site between the right and left hippocampi. We also found that irradiation induced a reduction in pattern separation ability but not in correct responses to new or same stimuli (Figure 4A). The analysis of the reaction times for new, lure, and repeated tasks revealed no significant differences across groups (Supplementary Figure S4). Next, we analyzed BOLD response patterns in the right and left during lure tasks. BOLD is an established functional brain imaging signal,²⁶ and reflects a neural response that leads to changes in oxyhemoglobin versus deoxyhemoglobin ratios to support energy demands related to functional responses.²⁷ Neurophysiologically, this hemodynamic response correlates with local field potential, incoming input, and the local processing of a given area rather than with action potentials (MUV: multi-unit activity).²⁸ In healthy volunteers (n = 36, 25 ± 15.0 years-old), the initial dip in the BOLD response peak (first negative-peak; N1) occurred at 1.7 ± 1.5 s, followed by a fractional increase in the initial positive-peak (P1) within 4.1 ± 1.1 s. The subsequent signal decrease was delayed by 8.4 ± 1.6 s (second negative-peak; N2), and the %BOLD signal change from the resting-state was -0.19 ± 0.27; this was followed by a slope that reached a plateau or peak value for long pulses (>20 s) (Figure 4C, G and Supplementary Table S2). On the other hand, a low-amplitude P1 followed by low-amplitude N1 and P2 (second positive-peak) which were then followed by a slope that reached a low-amplitude plateau value were characteristic of non-irradiated patients with brain tumors (n = 31, 25 ± 15.0 years-old) (Figure 4D, G and Supplementary Table S3). Furthermore, a marked reduction in the %BOLD signal change was found in patients with cIR (n = 13, 25 ± 15.0 years; Figure 4F). A delayed latency of N1 (4.7 ± 0.59 s) with a small amplitude of %BOLD signal change (-0.08 ± 0.24) subsequently followed by a delayed N2 (10.4 ± 0.87 s) was found in patients with cIR (Figure 4G and Supplementary Table S5). In the RT group before receiving RT (n = 10, 25 ± 15.0 years-old), we found a small P1 (0.02 ± 0.13 s) without an initial dip, followed by a N1, and a P2 with a large SD value (amplitudes ‐0.11 ± 0.29, and 0.08 ± 0.31, respectively), indicating signal fluctuations in this group (Figure 4E, G and Supplementary Table S4). Both the non-IR group and healthy volunteers exhibited a positive correlation between the BOLD response and a correct response rate in the lure task (Figure 4C, D). However, there was no such correlation in the cIR-group (Figure 4E, F).

Fig. 4.

Irradiation reduces pattern separation ability in human subjects. (A) Mean percentage of correct responses for new, lure, and same hippocampal memory tasks. Black dots represent individual data. **P < .01. (B) Relation between the correct response rate in the lure task and radiation doses to the right, and left hippocampus (n = 10). (C–F) Each panel represents from left to right, blood-oxygenation-level-dependent (BOLD) response curves, activation maps, %BOLD signal change, and correlation between %BOLD signal in healthy volunteers (C), patients with brain tumors in the non-irradiated group (D), patients with brain tumors before cranial irradiation (E), and patients with brain tumors after cranial irradiation (F). (G) Showing the mean of latency and amplitude of BOLD response for a lure task associated with dentate gyrus in healthy volunteers, patients with brain tumors in the non-irradiated group, patients with brain tumors before and after cranial irradiation. The first negative-peak is defined as N1, the second peak as N2, the first positive-peak as P1, that the second positive-peak as P2. “None”: absence of a peak.

Similar to rodent hippocampal X-ray ablation models, the analysis of human subjects using a lure task-based fMRI paradigm (hippocampal functional test for the ability to distinguish between discrete similar representations, Supplementary Figure S3) revealed that cranial irradiation affected hippocampal function and induced partial DG dysfunction. Further, in support of the above contention, we observed inactivation of the DG in the metabolic map (Figure 4F), and high error rates in the lure task, but not in the new tasks (Figure 4A). Taken together, these findings represent the first successful imaging of cranial radiation-induced toxicity with reference to human adult neurogenesis. Next, we analyzed the intra-hippocampal pathway using serial metabolic maps obtained during fMRI tasks. In healthy volunteers, bilateral hippocampi were similarly activated in DG, CA3, CA1, and subiculum (Supplementary Table S2; n = 36; Figure 5A, Supplementary Movies S1–S6). In the cIR-group, signals were restricted to the right hippocampus, and the activation spot was limited to the right perirhinal cortex; inactivated spots were observed in the DG, CA3, CA1 and subiculum (n = 13) (Figure 5B; Supplementary Movies S7–S12). In the new task, activation was observed in bilateral, perirhinal cortex, DG, CA3, CA1, and subiculum in healthy volunteers; however, in the cIR- group, activation was detected in the bilateral perirhinal cortex and subiculum, and inactivation was found in bilateral CA1 (Figure 5C, D; Supplementary Movies S13–24). Lastly, in the same tasks, activation signals in healthy volunteers were detected bilaterally like those for the lure tasks; the DG, CA3, CA1 and subiculum pathway activation was also observed in addition to direct CA1 activation in both hippocampi (Figure 5E; Supplementary Movies S25–S30). In the irradiated patients group, hot spots were found in the right EC, bilateral CA1, and subiculum, but not in the DG or CA3 (Figure 5F; Supplementary Movies S31–S33).

Fig. 5.

Activation maps for behavioral tasks (lure, new, and same) showing hippocampal subregions in healthy volunteers, and patients with brain tumors after cranial irradiation (A–F), Axial, coronal, and sagittal slices showing activation or inactivation spots (a voxel-level threshold of P < .001 uncorrected, and a cluster-level threshold of FWE-corrected P < .05). Color-bars: T-value. Also see full data given in movies (Supplementary Movies S1–S33). CA1, cornu Ammonis 1; CA3, cornu Ammonis 3; DG, dentate gyrus; L, left; R, right.

Neuromodulation of the Human Brain by Memantine and HBO

Adult human neurogenesis plays a crucial role in learning and memory, so restoring neurogenesis is important for improving cognition in patients. Based on the clinical trial previously reported ¹⁰ and our own results from rodent models, we selected memantine, a non-competitive NMDAR inhibitor to be used with HBO to improve neurogenesis. We asked whether a correct response rate obtained for an established task can reflect hippocampal neurotoxicity (indicating neuromodulation-induced neurogenesis). To resolve this question, we analyzed hippocampal function using tasked-fMRI (behavioral paradigm) in patients receiving radiation. The patients divided information was summarized in Supplementary Table S16. Nineteen patients were administered 5 mg memantine at 9.0 ± 10.0 days after starting radiotherapy (1.0 ± 5.0 Gy). At pre-radiotherapy, the correct response rate for a lure task was 17.9 ± 16.9% (n = 11). Patients received memantine during radiotherapy (45.0 ± 15.0 Gy). After completion of radiotherapy, the correct response rate of the lure task increased to 29.5 ± 14.1% (n = 11). In non-memantine group, a significant deterioration was found from 19.6 ± 20.3% at pre-radiotherapy to 11.5 ± 12.7% (n = 8) post-radiotherapy, indicating that the lure task score reflects the disruption and regeneration of hippocampal neurogenesis, and that neuromodulation using an NMDA antagonist concomitantly with HBO can be effective in this regard. Amplitude and latency of BOLD response between before (n = 10) and after (n = 13) radiation was compared by Man-Whitney U-test (Figure 4G, Supplementary Tables S4 and S5). But there was no significant difference (Supplementary Table S17). We next examined the correct response rate of the lure task was compared with each three treatment groups, and found that the group with HBO + memantine significantly improved the pattern separation ability, reflection of adult hippocampal neurogenesis, after radiation treatment (P < .01). Moreover, the signal change rate (0.48 ± 0.66) of the BOLD reaction of the right DG of the memantine combined with HBO group was significantly higher than that of the memantine non-combined group (‐0.16 ± 0.45) (P < .05) (Supplementary Figure S13), indicating hippocampal neuroimaging estimation can be detected human adult neurogenesis instead of battery of neuropsychological examinations.

Promoting Neurogenesis Improves Higher Cognitive Brain Function

In rodents, an increase in adult hippocampal neurogenesis induced enhanced normal object recognition, spatial learning, contextual fear conditioning, and extinction learning; a higher efficiency in differentiating between overlapping contextual representations was also observed, which was indicative of enhanced pattern separation.²⁹ Furthermore, stimulation of adult hippocampal neurogenesis, when combined with an intervention such as voluntary exercise produced a robust increase in exploratory behavior and improved cognition and mood.²² To clarify whether increased neurogenesis improves higher cognitive function, we performed neuropsychological assessments, analyzed pattern separation ability, and examined alterations in white matter integrity. In preoperative analysis, we found that higher cognitive function was influenced at least by hippocampal neurogenesis (Figure 4A). Neuropsychological assessment was performed in 119 preoperative patients (Supplementary Table S12). The same participants who completed the neuropsychological tests were included in the fMRI pattern separation ability test. The group with a correct response rate higher than the median score (25%) in a lure task (high score group, n = 60, 47.0 ± 16.8%, range: 25–88%; low score group, n = 59, 6.5 ± 6.6%, range: 0–23.1%) also achieved a higher score for 3MS, HDS-R, DS, DST, ST-naming, and ST-interference (Supplementary Figure S14 and Tables S12, S18); this indicates the importance of hippocampal memory function on general cognitive function (3MS and HDS-R), psychomotor speed (DST) and executive function (TMT and ST) (Figure 6A, Supplementary Table S7). Strikingly, patients receiving cranial radiation showed significant reduction of neurogenesis as shown in Figure 4A. Moreover, impaired hippocampal neurogenesis exacerbated neurocognitive dysfunction in patients receiving cranial radiation (Supplementary Figure S5 and Table S13). An interesting representative case is shown in Supplementary Figures S6 and S7; the relationship between hippocampus-associated pattern separation ability and neurocognitive domains is depicted, showing coordinated hippocampal DG deterioration at the early stages of cranial radiation and restoration of hippocampal function and higher cognitive function at the end of radiochemotherapy. Among the four neuropsychological test domains (MMSE, TMT, DS, and DST), irradiated patients treated with memantine + HBO showed significant elevation of DST scores, though we found no significant results in other domains (Figure 6A, Supplementary Table S7). We performed regression analysis between neuropsychological score and signal change rate of BOLD response and latency of the right DG, but no significant relationship was found (Supplementary Table S14). Next, we tried to analyze χ² test and residual analysis for the amount of effect of each neuropsychological domain, but χ² test was not applied by Cochran’s rule because of small sample size. There was no significant difference between each three treatment groups and their psychological score of each domain by Kruskal–Wallis test (Supplementary Table S15). DST can estimate psychomotor speed, and we speculated that white matter integrity is an important factor for signal transduction in neural networks. Therefore, we next estimated cranial radiation-induced toxicity in the contra-lesional hemisphere by applying diffusion tensor imaging data and examined the microstructural integrity [fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD)^30] of eight white matter functional fibers including ATR, CGC, CST, CGH, IFOF, ILF, SLF, and UNC (Figure 6B, Supplementary Tables S8–S11). The FA values significantly increased at CGH, ATR, and CST after the therapy, suggesting an improvement in white matter integrity. The MD value significantly decreased in CST, CGH, and ILF, at the same time suggesting that myelination may be promoted.

Fig. 6.

The effect of the HBO and memantine therapy on the neuropsychological domain and white matter in brain tumor patients receiving radio therapy. (A) Results of neuropsychological test scores for mini-mental state examination (MMSE), Trail Making Test (TMT) B–A, digit span subtest (DS), and digit symbol test (DST). Patients were divided into three groups according to the presence or absence of hyperbaric oxygen (HBO) with/or memantine therapy. *P < .05. (B) Figure showing the values of four parameters [fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD)] of diffusion tensor imaging-derived measures in irradiated patients without HBO and memantine (n = 3), with HBO (n = 11), and with HBO and memantine (n = 10) respectively. ATR, anterior thalamic radiation; a.u., arbitrary unit; CGC, cingulum cingulate gyrus part; CGH, cingulum hippocampal part; CST, corticospinal tract; FA, fractional anisotropy; IFOF, inferior fronto-occipital fasciculus; ILF, inferior longitudinal fasciculus; SLF, superior longitudinal fasciculus; UNC, uncinate fasciculus. *P < .05, **P < .01, ***P < .001.

The AD and RD level significantly decreased after radiation therapy at ILF and CGH. These results are somewhat complex, and may suggest that hyperbaric oxygen therapy has a multifaceted effect on the neural network, though it is not inconsistent with the results of the finding by our animal model experiments as shown in Figure 2A–C. We found recovery of the microstructural integrity of CGH, ATR, and CST fibers in the HBO + memantine group (n = 10) (Figure 6B), and two fibers including ATR and IFOF in a representative case treated by HBO + memantine (Supplementary Figures S8 and S9).

Discussion

Recent studies report that cancer treatments including radiation and chemotherapy cause cognitive impairment that correlates with hippocampal dysfunction and adult neurogenesis. Enhancing neurogenesis is thus an attractive option to improve brain health care. Importantly, we show here that not all patients are affected by cranial toxicity, and significant heterogeneity exists in this regard. Chemotherapy-related toxicity is also reported to show similar heterogeneity.³ Evaluating neurogenesis by non-invasively measuring the correct response rate in pattern separation ^17–19 is useful in medicine. In clinical practice, a large part of the brain is exposed to radiation; furthermore, human cognition involves multiple large-scale interactions among various regions. Thus, in addition to the hippocampus, other cerebral regions contribute to radiation- and chemotherapy-induced cranial toxicity. Recently, the importance of white matter function in accuracy, speed, and efficiency of information processing has been highlighted.³¹ Therefore, we investigated the effect of radiation on brain networks by diverse experimental modalities including structural and functional analysis both in human and animal models. Our study provides a basis for formulating therapies for restoring cognitive function in various brain disorders, and for the non-invasive estimation of hippocampal neurogenesis.

Segmentation of Hippocampal Subfields

For the elucidation of hippocampal function, it is important to accurately and reproducibly identify the structure and function of hippocampal subfields such as the DG and CA1–CA4, the subiculum, and subregions of the parahippocampal gyrus such as entorhinal, perirhinal, and parahippocampal cortices. Efforts have been underway to harmonize protocols among researchers.³² We performed behavioral analysis using fMRI, and successfully identified hippocampal information processing circuits in human healthy volunteers and patients with brain disorders. We found that the EC/DG/CA3 pathway was preferred during a lure task, consistent with a previous study³³ However, for a new task, we found direct relays from the EC to CA1 in human subjects. Memory processing for same tasks seemed to be activated both in the EC/DG/CA3 circuits as well as via direct CA1 pathways. Intra-hippocampal information processing is deranged by radiotherapy in patients receiving RT compared to that in healthy volunteers. The DG exhibited a cold spot in fMRI lure tasks in irradiated patients, and a marked reduction in %BOLD signal change. A delayed latency of the N1 with a small amplitude of %BOLD change subsequently followed by a delayed N2 was found in patients with cIR, indicating metabolic and synaptic failure in the region in irradiated patients. In healthy volunteers, bilateral hippocampal circuits were activated in every new, lure, and same tasked-fMRI scans, although the activation was not completely symmetrical. Memory processing detected by BOLD signals in bilateral hippocampi indicated that memory engrams were stored in bilateral hippocampi as reported in bilateral information on turn direction in rodents.³⁴ These bilateral BOLD activations were impaired in new or same-task fMRI data in patients with RT, but were preserved in lure task fMRI data; however, unlike in healthy volunteers, patients with RT showed negative BOLD responses in bilateral hippocampi. This bilateral hippocampal activity was not considered as simple coordinated functioning, but as a possible safeguarding mechanism for the memory store, because patients receiving cranial radiation retained correct responses in new or same tasks, even though the activation spots were detected only in the left hippocampus (Supplementary Movies S1–33); in patients with incorrect responses to the same task, we could not detect any signals in the bilateral hippocampi.

Reorganization of Mossy Fiber (MF) Axons, Negative BOLD Signals, and Pattern Separation

Recent imaging studies in humans indicate that pattern separation is mediated by circuitry consisting of EC, DG and CA3.³⁵ Reorganization of the MF axons is needed for pattern separation. The morphological maturation of hippocampal MF synapses impact cognitive ability, and its deficit leads to intellectual disability.³⁶ Moreover, dendritic spine changes are associated with long-term potentiation.^37,38 In our study, the immunohistochemical analysis of DCX was applied to detect neural progenitors, and the numbers of DCX-positive cells were significantly observed in the HBO + memantine groups. CA3 pyramidal cells and MF cells orthogonally correspond to pattern separation.³⁹ A more optimal three dimensional whole brain visualization method such as the clarity method ⁴⁰ is required to further elucidate MF reorganization and alterations in global brain network connectivity.

In this study, patients receiving cranial radiation showed marked reduction in BOLD signal changes, which indicated neurogenesis failure in the DG. In the serial analysis of two representative brain tumor cases with cranial RT (gliomatosis cerebri and atypical meningioma), we observed the following therapy-associated negative BOLD response: “a delayed latency of the N1 (4.7 ± 0.59 s) with a small amplitude of %BOLD change (‐0.08 ± 0.24 s) subsequently followed by a delayed N2 without significant P1”. Negative BOLD phenomenon is indicative of functional inhibition of regional synapses,⁴¹ and four mechanistic explanations have been proposed: 1) a large decrease in rCBF compared to the cerebral metabolic rate of oxygen (CMRO₂),⁴² 2) a larger fractional increase in CMRO₂ compared to rCBF,⁴³ 3) comparable decreases in rCBF and CMRO₂,⁴⁴ and 4) a slow rise in rCBF and rapid increments in neural activity.⁴⁵ Interestingly, we have shown that this negative BOLD response can be restored by HBO + memantine (Supplementary Figure S6A) or HBO (Supplementary Figure S6B) treatment with improved pattern separation ability.

White Matter Reorganization in Diffusion MRI-Based Tractography

Diffusion MRI is useful for unraveling the structural organization and connectivity of the human brain. Diffusion MRI-based tractography can non-invasively trace the three principal macroscopic white matter tracts including long association tracts, commissural fibers (connecting cortical areas to subcortical structures), and short-association tracts (u-fibers).⁴⁶ We found that combined HBO and memantine therapy affected more effectively in ameliorated on white matter integrity than that of HBO mono-therapy.

Chemotherapy-Induced Neurotoxicity

We must take into consideration that in addition to radiation, chemotherapeutic agents also exhibited neurotoxicity in patients receiving chemoradiotherapy. Temozolomide (TMZ) and alkylating DNA cross-linking cancer drugs are commonly used concurrently with radiation. In rodent experiments, several weeks of TMZ treatment attenuated spontaneous hippocampal theta (4–10 Hz) activity; this activity has been reported reduce hippocampal neurogenesis.⁴⁷.Patients recovered from the chemoradiotherapy-induced cranial toxicity when treated with HBO and memantine at post-chemoradiotherapy. However, patients deteriorated when the chemotherapy was restarted. TMZ alone is reported to be detrimental to neurogenesis.⁴⁷ Addition of radiotherapy to chemotherapy regimens led to hippocampal dysfunction and impeded neurogenesis (Supplementary Figure S6). Moreover, cognitive decline was observed in 40% of newly diagnosed glioblastoma patients after standard chemoradiotherapy,⁴⁸ and was enhanced in elderly patients. It is important to monitor the status of adult hippocampal neurogenesis for the long term follow-up of patients in the maintenance chemotherapeutic phase (Supplementary Figure S12). Based on promising results of this study, a randomized, double-blind, placebo-controlled trial to test recovery of higher cognitive function and effect on survival of malignant glioma patients receiving radiotherapy combined with hyperbaric oxygenation and memantine (Trial ID: jRCTs071190010) is now underway.

Funding

Supported by Grants-in-Aid for Scientific Research (B)(23390352) &(A)(17H01403); Challenging Exploratory Research (24650168),Special Account Budget for Education and Research (2011-2013, 2011-2015, 2014-2018, 2015-2019) by Ministry of Education, Culture, Sports, Science and Technology; the Industrial Disease Clinical Research Grants by Ministry of Health, Labour and Welfare (14050101-01); Grants for the Princess Takamatsu Cancer Fund& for Takeda Science Foundation, all these to S.I..

Conflict of interest statement. None declared.

Authorship Statement. Y.H. and M.N. were equally contributed. Y.H.: Human neuroimaging, writing and editing the manuscript. M.N.: Human fMRI imaging, writing and editing the manuscript. R.U.: Human neuroimaging. C.K.: Rodent behavioral analysis. K.F.: Histological analysis. H.T.: Ca² imaging, editing of the manuscript. S.I.: Project administration, funding acquisition, conceptualization, writing, review and editing of the manuscript.