Background
Whole-brain irradiation (WBI) is commonly used for the treatment of primary brain tumors and brain metastases. Most patients with primary brain tumors are treated to a total dose of 55–60 Gy delivered in 25–30 fractions. A variety of conformal strategies are used to reduce dose to remote areas of the brain. This contrasts with a total dose of 18–20 Gy used for CNS treatment of children with leukemia. Overall survival is improved with modern treatment techniques [
1,
2], but patients still experience adverse late effects. Following fractionated whole-brain irradiation (fWBI), 50–90 % of long-term survivors (>6 months) have irreversible cognitive decline [
3,
4]. The underlying molecular mechanisms that result in the loss of cognitive function after radiotherapy are not completely understood, and consequently, there is no treatment to prevent these adverse effects. Improving the quality of life of the growing population of patients who have received radiation treatment is an important objective.
WBI causes a number of deleterious cellular responses including neuronal dysfunction, blood–brain-barrier damage, astrocyte and microglia activation, and infiltration of peripherally derived monocytes [
5‐
12]. WBI induces apoptosis of neural progenitor cells [
8], which would be expected to affect overall cognitive function. Transplantation of neural stem cells in rodents can ameliorate radiation-induced cognitive dysfunction [
13]. Neuronal injury and loss is not the only pathway contributing to cognitive deficits, since activation of non-neuronal cell types also affects overall brain function. We and others have shown that radiation induces infiltration of peripheral myeloid cells that depend on CCR2 signaling [
14,
15] and that loss of the cytokine receptor CCR2 prevented the development of radiation-induced long-term cognitive deficits with no influence on neurogenesis [
15]. Recently, Piao et al. demonstrated that oligodendrocyte progenitors derived from human embryonic stem cells can remyelinate the brain and rescue radiation-induced behavior deficits [
16]. However, there is no clinically available agent that targets these convergent pathways in order to ameliorate or prevent permanent loss of cognitive function induced by WBI.
WBI induces up-regulation of pro-inflammatory cytokines and chemokines, including CCL2 [
5,
17], which facilitates the recruitment of CCR2
+ monocytes into the CNS. Following a single dose of 10 Gy WBI, deletion of CCL2 ameliorates deficits in hippocampal neurogenesis [
18]. We demonstrated an increase of monocyte accumulation in the brain, as well as a decrease of microglia, 7 days following a single dose of 10 Gy WBI [
5]. The recruitment of circulating monocytes into the CNS is regulated by the production of a number of soluble chemokines that interact with their cell surface receptors. One of these, colony-stimulating factor 1 receptor (CSF-1R), is a transmembrane tyrosine kinase receptor encoded by the
c-fms proto-oncogene [
19]. CSF-1/CSF-1R signaling regulates the survival, proliferation, chemotaxis, and differentiation of monocytes and macrophages [
20‐
22]. Loss of CSF-1R results in complete elimination of microglia and severe monocyte deficits [
23‐
25], and mice lacking CSF-1 have markedly reduced numbers of microglia [
26].
Our group, and others, has used a single dose of WBI to model radiation-induced brain injury. However, in clinical treatment, virtually all patients receive fractionated brain irradiation with the goal of reducing toxicity to normal tissue. Here, we model the effects of fWBI in young adult mice by using a fractionated treatment paradigm (3 × 3.3 Gy) and explore the outcomes of CSF-1R blockade by PLX5622, analog of another CSF-1R inhibitor PLX3397 [
27]. In other preclinical studies, PLX5622 has been used to diminish peripheral monocytes/macrophages [
28,
29]. Similar to PLX3397, treatment with higher dose of PLX5622 (1200 ppm) depletes microglia in the CNS [
28‐
34]. Recently, Dagher et al. showed that PLX5622 treatment (300 ppm) ameliorated cognitive deficits in aged Alzheimer’s mice [
32]. In addition, our preliminary results (data not shown) suggest that lower (300 ppm) and higher (1200 ppm) doses of PLX5622 treatment achieved similar effect in reducing circulating monocytes in the periphery. In light of these results, we treated young adult mice with lower dose of PLX5622 (300 ppm) and evaluated cognitive outcomes at 1 month after fWBI, the earliest time point we see cognitive deficits in our hands. Our data show that fractionated brain irradiation, similar to single-dose irradiation, results in hippocampal-dependent memory deficits and loss of dendritic spine density in hippocampal granule neurons. Strikingly, CSF-1R blockade by PLX5622 can prevent memory deficits and dendritic spine density loss in mice treated with fWBI. Flow cytometry analyses of myeloid populations following treatment with PLX5622 demonstrate a strong correlation between improved cognitive performance and both decreased microglia numbers and monocyte accumulation in the brain. Using a clinically relevant model and pharmacologic approach, our data show that CSF-1R blockade by PLX5622 can prevent fWBI-induced cognitive deficits in mice by preventing loss of synaptic dendritic spines. These data implicate a new and therapeutically tractable role for infiltrating monocytes and microglia after brain irradiation in loss of synaptic function.
Methods
Compound
Control and PLX5622 (300 ppm formulated in AIN-76A standard chow, Research Diets, Inc.) chows were provided by Plexxikon Inc (Berkeley, CA). Approximately 1.2 mg of PLX5622 was ingested by each mouse per day (calculation based on 4 g/mouse chow daily).
Animal procedures
All animal experiments were conducted in compliance with animal protocols approved by the Institutional Animal Care and Use Committee at the University of California, San Francisco (UCSF), following the National Institutes of Health Guidelines for animal care. C57BL/6J male mice were purchased from the Jackson Laboratory. CX3CR1
+/GFP/CCR2
+/RFP animals were generated by crossing CX3CR1
GFP/GFP/CCR2
RFP/RFP with C57BL/6J mice as previously described [
5]. Starting at 8 weeks old, C57BL/6J mice were treated with PLX5622 or control chow for 21 days. Cranial irradiation started 7 days after drug treatment was initiated. Mice used for Golgi staining were euthanized at the end of the first NOR test (33 days after fWBI) were euthanized and perfused with ice-cold PBS, and the right hemispheres were used for Golgi staining. Mice used for flow cytometry analyses were euthanized at indicated time relative to the last day of fWBI.
Radiation treatment
All the mice were anesthetized with intraperitoneal injection of a ketamine (100 mg/kg)/xylazine (10 mg/kg) mix and placed 16.3 cm from a cesium-137 source (JL Shepherd & Associates). The eyes and body were shielded by a lead collimator that limited the beam to a width of 1 cm. An extra lead plate was used to block exposure of the trachea. Irradiated groups received 1.65 Gy of irradiation on both side of the head to accumulate 3.3 Gy for each fractionated irradiation. Three fractions were delivered every other day over 5 days to accumulate a total dose of 10 Gy. Sham animals underwent the same procedures without radiation.
Novel object recognition test
All the mice used for the NOR test were housed in a room with reversed light cycle (12 light/12 dark) for at least 2 weeks before tests. Tests were conducted during the dark cycle. The mice were handled 5 min each day for 5 days before habituation. An open arena (30 cm × 30 cm × 30 cm; L × W × H) was placed in a dimly lit behavior test room with an overhead camera. The mice were allowed to explore the open arena for 10 min for two consecutive days. On day 3, two identical objects were placed in the arena and mice were allowed to explore for 5 min. On day 4, one of the objects was replaced by a novel object and mice were allowed to explore for 5 min (Additional file
1: Figure S1A). Trials were recorded by the overhead camera and analyzed by an automatic video tracking system (EthoVision, Noldus) for movement tracking or by manual scoring for exploratory behavior. Exploratory behavior was defined as the animal directing its nose toward an object at a distance less than 2 cm. Objects were secured in the arena with magnets. Arena and objects were wiped with 70 % ethanol between trials to eliminate odor cues.
Delayed matching-to-place dry maze test
Delayed matching-to-place (DMP) dry maze test was used to measure special working memory as described by Faizi et al. [
35]. Briefly, we used a modified Barnes maze with 40 escape holes (
D = 5 cm, 16 holes on the outer ring with 50-cm distance to the center, 16 holes on the middle ring with 35-cm distance to the center, and 8 holes on the inner ring with 20-cm distance to the center). All holes were uncovered with the exception of the escape hole, which is covered with a dark escape tube (a black PVC tube). The light was set to approximately 1200 lux, and a noise (2 kHz, 85 dB) was used during the test. Visual cues were placed on three sides of the maze. Mice were giving four trials each day with interval of 10 min. Mice were placed at the center of the maze under a dark box for 30 s. The trial started when the box was removed and ended when the mice found the escape hole within 90 s. Mice were guided to the escape hole by the experimenter if they could not find it within 90 s. Noise was turned off, and the escape hole was covered immediately after the mice entered. The mice were returned to their home cage after a 10-s delay. The maze surface and the escape tube were cleaned with 70 % ethanol (
v/
v) after each trial to minimize odor cues. The escape tube was kept at the same location and changed on each test days. Trials were recorded by an overhead camera and analyzed by Ethovision (Ethovision, Noldus).
Metric distance test
Metric distance test was used to measure hippocampal function as previously described by Goodrich-Hunsaker et al. [
36] with alterations to suit test in mice (Additional file
1: Figure S1E). Briefly, the test was performed on two consecutive days during the dark cycle with one habituation phase and 3 trials each day. On day 1, the mice were put into an open arena (30 cm × 30 cm × 30 cm, L × W × H) for 5 min. On trial 1, the mice were put into the arena with two identical objects placed at a distance of 28 cm to each other for 5 min. On trial 2, the mice were put into the arena with the same setting as trial 1 for 5 min. On trial 3, the distance between the objects was shortened to 14 cm and the mice were allowed to explore for 5 min. There was a 3-min interval between each trial, and the mice were put back into their home cage after each trial. One day 2, all trials were performed with the same setting as day 1 except that the distance between objects on trial 3 was changed to 21 cm. The objects and the arena were cleaned with 70 % (
v/
V) ethanol after each trial to minimize odor cues. Trials were recorded, and total time exploring the object pair was scored. Total time exploring during trial 1 on each test day was used as baseline of exploration.
Flow cytometry
The mice used for flow cytometry analysis were euthanized and perfused with ice-cold PBS. The brains were removed and immediately placed into ice-cold HBSS. Brain samples were then dissociated using a Neural Tissue Dissociation Kit (P) (Miltenyi Biotec). Dissociated cells were resuspended in 10 ml of 30 % Percoll solution (Sigma) in an RPMI medium and laid over a 1 ml 70 % Percoll solution layer. After centrifugation at 800g for 30 min at 4 °C, interphase cells were transferred to a new 15-ml Falcon tube and washed with RPMI. Cell pellets were resuspended with FACS buffer (DPBS with 0.5 % BSA fraction V) and blocked with one volume of blocking solution (5 % normal mouse serum, 5 % normal rat serum, 5 % normal rabbit serum, 2 % FBS, and 1 % BSA fraction V in ×1 DPBS) for 30 min and stained for 30 min with fluorophore-conjugated antibodies on ice (CD45-BV711, CD11b-AF700, Ly6C-Pacific Blue, and Ly6G-PE were purchased from BD Pharmingen); 7AAD was used to exclude dead cells. Data were collected on an Aria III sorter (BD) and analyzed with FlowJo v10 software (Tree Star Inc.). At least 20,000 and 200,000 viable events were collected from each brain and blood sample, respectively.
Golgi staining
Brain hemispheres stayed in Golgi staining solution (A Modified Golgi-Cox Stain for Neural Cells, Docket No. D4433, Cornell University) for 14 days and were transferred into 30 % sucrose in ×1 PBS overnight at 4 °C. The next day, tissues were transferred into fresh 30 % sucrose solution, protected from light and stored at 4 °C for at least 2 days. Brains were cut into 100 μm sections with a vibratome (VT1000 S, Leica, Wetzlar, Germany), mounted on gelatin pre-treated slides and dried for 2 days. Samples were then developed with the developing solution (A Modified Golgi-Cox Stain for Neural Cells, Docket No. D4433, Cornell University), covered, and dried. Images were taken on a Keyence 7000 system under a ×100 objective lens with immersion oil for hippocampal granule neurons. For each sample, 18–30 images were taken (2–3 images per section, 8–12 sections per mouse, N = 5–6 per treatment group) and used for dendritic spine density analysis. All protrusions from the dendrites were manually counted as spines regardless of morphology. A total length of at least 3000 μm of dendrites was analyzed from each animal using ImageJ (National Institutes of Health).
Statistical analysis
NOR test results are shown as mean percentage of time spent on exploring each object (time exploring familiar or novel object/total exploring time) or mean discrimination index ((time exploring novel object − time exploring familiar object)/total exploring time) ± SEM. Metric distance test results are shown as percentage of time spent on exploring both objects in trial 1 on each test day. Results for DMP and metric distance tests were analyzed with ordinary two-way ANOVA with Bonferroni’s test for post hoc comparisons using day and experimental group as independent factors. Results for NOR test, dendritic spine density analysis, and flow cytometry with PLX5622 and fWBI treatments were analyzed with ordinary two-way ANOVA with Tukey’s test for post hoc comparisons using PLX5622 and fWBI as independent factors. Results for temporal analysis of monocyte accumulation were analyzed with one-way ANOVA with Bonferroni’s test for post hoc comparisons. All other comparisons between two sets of data were determined using t test. Error bars are shown as mean ± SEM. Details of each statistical analysis were described in figure legends. Graphs were analyzed and plotted with GraphPad Prism 6 software (GraphPad Software, Inc).
Discussion
Radiotherapy is routinely delivered in fractions to treat cancers in order to reduce toxicity to normal tissues. However, most patients who receive WBI, or radiation to both temporal lobes, will develop cognitive deficits that can be profoundly disabling. Although conformal techniques and lowered CNS doses have reduced some adverse effects, there has been little research in the area of modifying the cellular response in the CNS following radiation, with the goal of reducing long-term cognitive deficits in patients. In this study, we modeled a clinical treatment paradigm by dividing 10 Gy WBI into three equal fractions delivered every other day and examined the cellular and behavioral consequences of that treatment. Our data show that fractionated brain irradiation results in hippocampal-dependent memory deficits and loss of dendritic spine density. CSF-1R blockade appears to rescue memory deficits and dendritic spine density loss in hippocampal granule neurons in the mouse model we have studied. Cumulatively, our findings offer novel insight into the mechanism of radiation-induced injury and demonstrate that the CSF-1R is a relevant and rational therapeutic target that could be used clinically to prevent irradiation-induced sequelae.
We previously reported that a single dose of 10 Gy WBI selectively disrupts hippocampal-dependent memory functions by affecting the cellular infrastructure responsible for plasticity and memory formation [
13,
15]. Notably the hippocampus is exquisitely sensitive to WBI as non-hippocampal functions were intact [
15]. Our current data demonstrate that fractionated radiation causes persistent cognitive deficits similar to those induced by 10 Gy WBI was delivered as a single dose. The NOR test examines hippocampal-dependent recognition memory for objects in rodents, and 1 month after irradiation, the mice had markedly impaired memory function that persisted up to 3 months. Further assessment of hippocampal dentate gyrus function with metric distance test showed significant impairment in animals that received fWBI. These cognitive changes are almost certainly the consequence of the activation of a diverse set of cellular responses that lead to synaptic alterations. Interestingly, higher cortical functions involved in episodic-like and working memory were not significantly affected by this irradiation paradigm. These data suggest while fractionation may reduce other types of radiation-induced normal brain injury, the hippocampus remains a selectively vulnerable structure.
Dendritic spines are postsynaptic components of excitatory synapses in the CNS. Their structural and density changes play fundamental role in synaptic functions, which are crucial for learning and memory [
42,
43]. Decreased density and malformation of dendritic spines have been observed in neurodegenerative diseases, and loss of synapses is strongly correlated with cognitive decline [
44,
45]. In the current study, we observed a decrease of dendritic spine density after fWBI comparable to what we and others previously reported using a single dose of 10 Gy WBI [
46,
47]. These results suggest that fractionated irradiation routinely used to treat patients with primary brain tumors and metastases has substantial deleterious effects, similar to those observed following single-dose irradiation.
The exact mechanism of how ionizing radiation leads to impaired neuronal function, as demonstrated by the reduction of dendritic spines, is unclear. It is possible that some of this effect is directly related to DNA and cellular damage occurring in neurons following irradiation. However, there is also evidence to suggest that other cellular and molecular pathways activated after brain irradiation can contribute to neuronal dysfunction. Relevant to our study, analyses of myeloid cell populations reveal an accumulation of monocytes in the CNS after fWBI. CCR2 is a chemokine receptor expressed on cells of the myeloid cell lineage, and we have previously noted that CCR2 deficiency prevents radiation-induced hippocampal neuronal dysfunction from a cellular and behavioral perspective [
15]. Based on these results, we postulated that inhibition of monocyte accumulation could prevent cognitive deficits induced by cranial irradiation. Because inhibition of CCR2 results in a more selective inhibition of myeloid cells, we chose to broadly block monocyte accumulation by treatment with PLX5622, a small-molecule selective CSF-1R inhibitor, which was tested in phase I clinical trials (ClinicalTrials.gov Identifier: NCT01329991 and NCT01282684). CSF-1R is essential for the survival and differentiation of macrophages and cells of the monocyte lineage [
26,
41]. The association of increased macrophage infiltration with poor diagnosis in many types of cancers has led to some interest in targeting CSF-1R for cancer therapy [
48], and studies have shown that CSF-1R inhibitors have anti-tumor effects [
33,
49]. CSF-1R blockade has also showed efficacy in ameliorating other neuroinflammatory diseases. Gomez-Nicola et al. reported that blockade of CSF-1R with another tyrosine kinase antagonist inhibits microglia proliferation and slows neuronal damage in prion disease models [
50].
Here, we irradiated young adult mice at 2 months of age (equivalent of 20 years of age in humans [
51]) to reflect a population in human patients with longer survival and high risk to develop cognitive deficits after radiotherapy. We performed cognitive tests when mice were 3–5 months of age (equivalent of 24–30 years of age in humans [
51]) to represent delayed time points when cognitive deficits are seen in humans. We demonstrate with flow cytometry analyses that PLX5622 treatment inhibits Ly6C
high monocyte accumulation in the brain after fWBI, possibly due to the reduced numbers of circulating Ly6C
high monocytes in the peripheral vasculature. We observed a 35–50 % decrease of microglia number in the brain during PLX5622 treatment, similar to the data reported by Dagher et al. with an Alzheimer’s model [
32]. The numbers of brain microglia and peripheral blood monocyte both recovered after PLX5622 treatment was stopped, and PLX5622 treatment alone did not affect dendritic spine density or cognitive performance, suggesting that the effect of PLX5622 is transient and non-toxic. Ten days after fWBI, we observed a trend toward a decrease in dendritic spine density in both control and PLX5622-treated groups, but this did not reach statistical significance. However, at 33 days after fWBI, there was a statistically significant 20 % reduction of spine density in control groups, and CSF-1R blockade by PLX5622 treatment completely reversed the spine density loss. Given the limitations of Golgi staining and the manual scoring method used for dendritic spine counts [
52], it is possible that we underestimated early-delayed structural and functional changes at 10 days post fWBI. Dye loading and electrophysiology techniques can be used in future studies to more accurately assess neuronal functions with CSF-1R blockade after fWBI. Nonetheless, our results are consistent with previous findings that suggest radiation-induced cognitive dysfunction is an ongoing process [
3,
4]. We cannot, however, definitively conclude from our data whether the preserved dendritic spine density and cognition after fWBI are due to temporarily decreased microglia number or impaired monocyte accumulation during and immediately after fractionated brain irradiation. We observed a trend of better performance in the NOR test in PLX5622-treated animals compared to animals on control diet (Figs.
2b, c, PLX5622-sham vs control sham). However, given that PLX5622 specifically acts on the CSF-1R, which is expressed in myeloid cells, it is unlikely that PLX5622 have direct neurotrophic effects. It is possible that PLX5622 has secondary anti-inflammatory effects due to reduced microglia and monocyte numbers in the CNS during and after radiation. Further studies specifically targeting either cell population might help answer this question.
The duration of treatment required to modulate cognitive function is unclear. Several PPAR agonists and RAS blockers have been shown to be effective in ameliorating radiation-induced cognitive dysfunctions [
53‐
55]. These studies all utilized continuous treatments starting from 3 to 7 days before radiation until the end of cognition assessment. In this study, we observe that transiently inhibiting the CSF-1/CSF-1R signaling reduces microglia number during radiation and blocks monocyte infiltration after radiation and is sufficient to ameliorate fWBI-induced neuronal and cognitive dysfunction. However, our previous report demonstrated that radiation-induced inflammation in the brain persists for at least 3 months after a single dose of 10 Gy cranial irradiation [
15]. It is possible that dendritic spine loss in hippocampal granule neurons is a secondary effect caused by radiation-induced inflammation and is an ongoing process which lasts a substantial period of time. One limitation of our current study is that we only assessed dendritic spine density and cognitive performance up to 1 month after fWBI with PLX5622 treatment. Further studies are needed to determine if dendritic spine loss occurs at longer time points after brain irradiation, and whether temporary blockade of CSF-1R can permanently rescue this effect and ameliorate cognitive deficits. In addition, given the fact that PLX5622 treatment reduces a substantial portion of microglia, the safety, especially long-term effect of CSF-1R blockade, remains to be tested.
Acknowledgements
We thank Jane Gordon from the FACS core at HDFCC, UCSF, for assistance in flow cytometry. We thank Phillip Yang and Dr. Tingting Huang at Stanford University for the help with the Golgi staining and Dr. Jennifer Punk for the assistance with imaging.