In vivo imaging of the microglial landscape after whole brain radiation therapy

Abstract

Purpose:

Whole brain radiation therapy (WBRT) is an important treatment for patients with multiple brain metastases, but can also cause cognitive deterioration. Microglia, the resident immune cells of the brain, promote a pro-inflammatory environment and likely contribute to cognitive decline after WBRT. To investigate the temporal dynamics of the microglial reaction in individual mice to WBRT, we developed a novel in vivo experimental model using cranial window implants and longitudinal imaging.

Methods:

Chronic cranial windows were surgically-implanted over the somatosensory cortex of transgenic Cx3cr1-EGFP/+ C57BL/6 mice, where microglia are fluorescently tagged with EGFP. Cx3cr1-EGFP/+ mice were also crossed with Thy1-YFP mice to fluorescently dual label microglia and subsets of neurons throughout the brain. Three weeks after window implantation and recovery, CT-image-guided WBRT was delivered (single dose 10 Gy using two 5 Gy parallel-opposed lateral beams). Radiation dosing was confirmed using radiochromic film. Then, in vivo 2-photon microscopy was used to longitudinally image the microglial landscape and microglial motility at 7 days and 16 days after irradiation in the same mouse.

Results:

Film dosimetry confirmed the average delivered dose per beam at midpoint was accurate within 2%, with no attenuation from the window frame. By 7 days after WBRT, significant changes in the microglial landscape were seen, characterized by apparent loss of microglial cells (20%) and significant re-arrangements of microglial location with time after irradiation (36% of cells not found in original location).

Conclusions:

Using longitudinal in vivo 2-photon imaging, this study demonstrated the feasibility of imaging microglia-neuron interactions and defining how microglia react to WBRT in the same mouse. Having demonstrated utility of the model, this experimental paradigm can be used to investigate the dynamic changes of many different brain cell types and their interactions after WBRT, and uncover the underlying cellular mechanisms of WBRT-induced cognitive decline.

Introduction:

Whole brain radiation therapy (WBRT) is an important treatment approach for patients with multiple brain metastases (1) or large volume multifocal glioma (2). Cranial irradiation induces a complex series of cellular and microenvironmental events that can lead to progressive cognitive decline, dementia and death, for which there are no effective treatments (3). Given the temporal dynamics of radiation injury and damage repair after WBRT in the brain, and how these changes can impact WBRT-induced cognitive dysfunction, new experimental paradigms are needed that allow for the longitudinal observation of pathology within the same animal.

Microglia, the resident immune cells of the central nervous system, are highly reactive to homeostatic perturbations and contribute to the etiology of neurodegenerative diseases and aging (4). Pharmacological depletion of microglia ameliorates WBRT-induced cognitive dysfunction in mice (5, 6), suggesting microglia also contribute to WBRT-induced pathology and cognitive decline. Precisely how WBRT affects microglia, leading to a detrimental phenotype, is not clear.

To characterize the microglial response to WBRT, we developed and validated a method to longitudinally image microglia in vivo (7), before and after irradiation in the same animal. Chronic cranial windows were implanted over somatosensory cortex, and three weeks later animals were treated with WBRT. Cortical microglia were imaged before and at four time points after irradiation. Significant alterations in the microglial landscape were evident, despite no gross changes in cellular morphology. We also demonstrate the feasibility of tracking microglial-neuronal interactions in vivo to determine the effects microglia have on neuronal dysfunction after WBRT. This approach can be extended to other paradigms of cranial irradiation, including more localized treatments.

Methods:

Animals:

All animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Rochester, NY, and conform to NIH guidelines. All mice were on a C57/BL6J background. To visualize microglia in vivo, Cx3cr1-EGFP/+ heterozygous mice were used, in which microglia are fluorescently tagged with EGFP (JAX 005582). For each group, 3 mice were used (2 male, 1 female), aged 18-20 weeks. To visualize microglia-dendrite interactions, Cx3cr1-EGFP/+ mice were crossed with Thy1-YFP mice, which label subsets of neurons throughout the brain, including somatosensory cortex Layer V neurons (JAX 003782). Images from one representative male mouse are shown.

Cranial Window Implantation:

Mice were anesthetized with a fentanyl-based cocktail (fentanyl 0.05 mg/kg; midazolam 5.0 mg/kg; dexmedetomidine 0.5 mg/kg). A circular glass window (3 mm coverslip glued to a 5 mm coverslip) was placed over a 3mm-craniotomy above somatosensory cortex. Dental cement was used to seal the coverslip and attach a metal headplate, as previously described (8). After implantation, mice were allowed to recover for at least 3 weeks, giving time for any inflammation induced by the procedure to subside (9). Implanted windows are stable for several months, allowing for repeated imaging over a long period of time.

Whole brain irradiation:

A Small Animal Radiation Research Platform (SARRP) irradiator (XStrahl) was used to perform CT image-guided WBRT. Mice with cranial window implants were anesthetized (2.8% isoflurane, balance air) and given a single 10 Gy dose (two parallel-opposed beams) using the 10x10 mm² aperture, encompassing the forebrain and thalamus. Irradiated mice were under isoflurane anesthesia for less than 15 minutes and recovered within several minutes. Non-irradiated control mice with cranial window implants were not placed in the SARRP irradiator or exposed to isoflurane.

Dosimetry calibration:

EBT3 radiochromic film was calibrated from 3.0-7.0 Gy in 0.5 Gy intervals using 3 shots at each calibration dose, using the same SARRP 10x10 mm² aperture. These films were then scanned on an Epson 11000XL flatbed scanner at least 24 hours post-irradiation with no software corrections. The three red channel measurements were averaged for each calibration point, and the resulting curve was fit to a 3^rd order polynomial.

2-photon in vivo imaging:

A custom-built 2-photon laser scanning microscope (2PLSM; 20x 0.96 NA water-immersion objective, Olympus) was used to image cortical EGFP+ microglia and YFP+ dendrites as described previously (10). For imaging, were anesthetized with a fentanyl-based cocktail (fentanyl 0.05 mg/kg; midazolam 5.0 mg/kg; dexmedetomidine 0.5 mg/kg). Two-photon excitation wavelength of 920 nm was used for both fluorophores. Control animals were imaged at −2d, 2d, 7d and 16d; while WBRT-treated animals were imaged at −2d, 6hr, 24hr, 7d and 16d, relative to treatment. To image the same field of view across imaging sessions, the cortical blood vessel pattern was used to position the mouse under the microscope. Blood vessels appear on 2-photon imaging as a relative absence of signal and are outlined in Fig 2.

Figure 2: WBRT alters the microglial landscape.

(a-b) Maximum Z-projection images of EGFP-expressing microglia in the same imaged over time in a non-irradiated mouse (No RT control; a) and a mouse that received 10 Gy of WBRT (WBRT; b). Overlays show pre-treatment (−2 days) in magenta and post-treatment (7 d, or 16 d) in green. Dotted red lines indicate locations of blood vessels used to identify the same area. Red arrows point to representative cells whose locations were unchanged. Scale bar 50 μm. (c) Total number of cells in the same region of interest, normalized to pre-treatment time point. Black circles represent No RT controls, imaged at −2d, +2d, +7d, and +16d relative to treatment (n=3 mice). Magenta squares represent WBRT-treated mice imaged at −2d, +6hrs, +24hrs, +7d, and +16d (n=3) (d) Comparison of the normalized cell number 7d and 16d after WBRT vs. No RT controls (2-way ANOVA with Sidak’s multiple comparisons test; p-values as written). (e) Illustration of cell classification scheme comparing 7d post- and 2d pre-treatment from a representative WBRT-treated animal . (f) Change in persistent cell fraction over time between treatment groups, with time points and samples as described in (c). (g) Comparison of persistent cell fraction 7d and 16d post-WBRT vs. no RT controls (2-way ANOVA with Sidak’s multiple comparisons test; p-values as written). (h) Change in new cell fraction over time between treatment groups, with time points and samples as described in (c). No significant effects of treatment or time were observed comparing 7d and 16d post-WBRT vs. no RT by 2-way ANOVA. Data are presented as mean ± SEM. Details of statistical analyses presented in Supplementary Table 1.

Analysis of microglial re-arrangement:

Z-stacks were taken in the same location for each imaging session. Image volumes from subsequent imaging sessions were aligned using the BigWarp plugin in FIJI (imageJ.net/fiji). A maximum Z projection of 80 slices (1 μm step size) was made. The 2D coordinates of microglial cell bodies at each time point were marked using the Cell Counter plugin. A nearest neighbor analysis was then used to quantify microglial rearrangements between pre-radiation (−2 days) and post-radiation timepoints in MATLAB (Mathworks). Cells at the pre-radiation time point were classified as ‘persisten’ if there was a cell at the subsequent time point <10 μm away; otherwise they were labeled as ‘los’. Cells at a given post-radiation time point were labeled as ‘new’ if there were no cells at the pre-radiation timepoint <10 μm away; otherwise they were labeled as ‘original’. Thus ‘original’ and ‘persisten’ cells are overlapping populations. MATLAB scripts available at (Anonymized for Review).

Statistics:

Statistical analyses were performed in Prism 8.0 (GraphPad). Comparisons between non-irradiated control mice and WBRT-treated mice at 7 days and 16 days post-treatment were done using repeated-measures 2-way ANOVA with Sidak’s multiple comparisons test. Results from the multiple comparisons tests are presented in the figures. Full results from the statistical analyses are presented in Supplementary Table 1.

Results:

To test the feasibility of longitudinal in vivo imaging of the brain following WBRT, cranial windows were implanted over somatosensory cortex in adult mice, and microglial cells were imaged using 2-photon in vivo microscopy before and after CT image-guided WBRT using the SARRP irradiator (Fig 1a-c). The computed dose-volume histogram shows that the majority of the targeted brain region received the desired dose of 10 Gy (Fig 1d, red), while a control region outside of the beam path received little radiation (Fig 1d, green). To empirically confirm the calculated dosimetry, EBT radiochromic film was placed either along the midline of the brain or directly underneath the cranial window and metal headplate implant and one 5 Gy dose was delivered for each measurement. We found that both regions received within 95% of the 5 Gy target dose over 3 trials (Fig 1e), with a midpoint dose average of 4.97 (SD ±0.10) Gy, confirming our ability to effectively deliver WBRT to mice with chronic cranial window implants.

Figure 1: WBRT dosimetry in the brain of a mouse with a cranial window implant.

(a) Timeline of experimental design. Green arrows indicate imaging timepoints. (b) Picture of mouse with cranial window implant in the SARRP irradiator. (c) CT-images in the (i) coronal and (ii) sagittal planes. Blue box shows the extent of the 10mm x 10mm beam used for WBRT. Regions highlighted in red (irradiated) and green (non-target) in (c) show the volumes used to calculate the Dose Volume Histogram (d). (e) Plot of empirically measured delivered dose in indicated areas using radiochromic film. The film was placed in the midline of the brain and underneath the cranial window and a single 5 Gy dose was delivered. Dotted line indicates the target dose of 5 Gy.

The effects of WBRT on cortical microglia were then determined using this experimental paradigm. Imaging the same location before and at several timepoints after WBRT revealed substantial changes in the microglial landscape as early as 7 days after radiation (Fig 2; Supplementary Table 1). While the microglial landscape in non-irradiated control mice was relatively stable over time (Fig 2a), that of WBRT-treated mice was substantially altered after treatment (Fig 2b). WBRT led to a significant decrease in overall cell numbers by 16 days after radiation compared to non-irradiated controls (Fig 2c-d).

To better quantify cellular re-arrangements microglia in the pre-radiation session were classified as either ‘persisten’ or ‘los’ in subsequent imaging, and microglia in post-radiation sessions were classified as ‘new’ or ‘original’ (when compared to the pre-radiation session), where ‘original’ and ‘persisten’ cells are overlapping populations (Fig 2e). Compared to non-irradiated controls, the fraction of persistent cells from 2 day pre- to 7 days post-WBRT was significantly decreased, but was not significantly different at 16 days (Fig 2f-g). The fraction of new cells was not statistically significant between treatment groups (Fig 2h). The decrease in cell numbers could be secondary to radiation-induced microglial cell death (11, 12). Subsequent re-arrangements may be a result of remaining microglia re-distributing throughout the cortex, but this remains to be tested.

Finally, the feasibility of this method to track microglial interactions with dendrites was demonstrated before and after WBRT. By crossing the Cx3cr1-EGFP and Thy-YFP mouse lines, we were able to image the dendritic arbors in somatosensory cortical layers 2/3 emerging from a subset of layer 5 neurons (Fig 3a-b). This method has sufficient resolution to detect interactions between microglia and the same dendrites or spines following WBRT (Fig 3c), providing an avenue to determine how WBRT affects microglial-neuronal interactions in real time.

Figure 3: Longitudinal imaging of microglia and dendrites after WBRT.

(a) Low magnification maximum Z projection of EGFP-expressing microglia (green) and YFP-expressing neuronal dendrites (magenta), demonstrating labeling of a subset of dendrites in layers 1-2/3 of somatosensory cortex. (b) Side view (maximum Y-projection) showing neuronal cell bodies in layer V extending dendrites into the upper cortical layers. (c) High magnification imaging of the same dendrites longitudinally, before and after WBRT. Representative of 3 individual mice. Scale bars 50 μm in (a,b) and 10 μm in (c).

Discussion:

Combining CT image-guided WBRT with longitudinal in vivo imaging after brain irradiation can be a fruitful approach to studying microglial physiology in the context of therapeutic irradiation. In this study, we found significant alterations in the microglial landscape in somatosensory cortex that occur within one week of WBRT, possibly through a combination of cell death and the re-arrangements of remaining microglia (Fig 2).

Because of its role in learning and memory and the presence of neurogenesis, the hippocampus has been the main focus of several studies examining WBRT-induced pathology (13). In the hippocampus, radiation induces a long-lasting aging-like transcriptional profile in microglia (14) and leads to prolonged loss of microglial cells (11, 12, 15). Our observations of decreasing microglial numbers in cortex as well as the recent finding that injury-induced microglial proliferation is impaired after radiation (16) suggest a brain-wide impairment of microglial viability. Post-radiation, microglia recruit peripheral monocytes through secretion of CCL2, exacerbating inflammation (17–20). The Cx3cr1-EGFP/+ mouse line cannot distinguish between resident microglia and monocyte-derived macrophages. However, in the cortex 3 weeks after cranial irradiation, infiltrating monocytes are about 50-fold less abundant than resident microglia (20). Thus, while infiltrating monocytes may contribute to the general inflammatory milieu, the changes in microglial landscape observed in this study are likely due to alterations of resident microglia.

Here, we show the feasibility of imaging microglia-dendrite interactions in vivo before and after cranial irradiation (Fig 3). While little is known about the effects of radiation on synapses in the cortex (13), the microglial CR3 receptor mediates dendritic spine loss in the hippocampus (21). Future studies using the methodology described here will be able to determine dendritic spine dynamics following radiation and whether they are affected by microglial contacts.

This study is a methodological investigation to demonstrate the feasibility of examining real-time microglial responses after brain irradiation. A key advantage to longitudinal in vivo imaging is the ability to track pathology in the same animal over time. We envisage this approach being used to investigate in detail changes in microglial behaviors, such as surveillance, morphology, and phagocytic function, as well as their interactions with neurons. Given the central role of microglia in shaping the inflammatory environment of the irradiated brain, addressing changes to microglial physiology could be a promising avenue for treatment.