Six-month-old C57BL/6J male mice (n = 84 mice in total) were obtained from Jackson Laboratories, Bar Harbor Maine. The biological age of the mice was selected to be relevant to the biological age of astronauts during space missions. The mice were shipped from Jackson Laboratories to Brookhaven National Laboratory (BNL), Upton, New York, and allowed to accommodate to the housing facility there for 1 week. Subsequently, the mice were irradiated with 0.3, 0.6, or 0.9 Gy of 600 MeV ²⁸Si ion or sham-irradiated (n = 42 mice/radiation dose). The duration of the radiation exposures was 0.99 ± 0.08 min, 1.53 ± 0.08 min, and 2.10 ± 0.07 min, for the 0.3, 0.6, and 0.9 Gy dose respectively. For irradiation, mice were individually loaded into 8 × 3 × 3 cm plastic square enclosures with air holes and placed in a foam fixture in the beam line of the NASA Space Radiation Laboratory (NSRL). They were exposed to a rectangular beam of approximately 20 × 20 cm. The focused beam of high-energy was generated by the Booster accelerator at BNL and transferred to the experimental beam line at the NSRL facility. Dose calibration was performed so that the desired dose could be delivered. Sham-irradiated mice were placed into the plastic enclosures for the same time as the irradiated mice. Mice were randomly assigned to the experimental groups. The week after the irradiation or sham-irradiation, the mice were shipped to Oregon Health & Science University (OHSU) and cognitive testing started 2 weeks following irradiation. Mice were tested in the open field and for novel object recognition in week 1. In week 3, mice were tested for hippocampal network stability and euthanized immediately after by cervical dislocation. The hippocampus of one hemibrain was dissected for DNA methylation and RNAseq analyses. The other hemibrain was frozen in −70°C isopentane (2-methyl butane, Sigma) and processed for Arc mRNA and TET2 immunohistochemical analyses. The hemibrains were blocked together such that each slide contained sections from different experimental groups. All slides were cryosectioned and stored at −70°C until processed for immunocytochemistry or fluorescence in situ hybridization (FISH). All protocols were reviewed and approved by the Institutional Animal Care and Use Committees (IACUC) of OHSU and BNL and were in compliance with all Federal regulations.

The novel object recognition test was performed as described (Raber et al., 2002). The mice were habituated to an open field (16 × 16 inches, Kinder Scientific, Poway, CA) for 3 times for 10 min each over three subsequent days. On day 4, the mice were placed in the open field containing two identical objects and they were allowed to freely explore for 15 min. On day 5, the mice were placed again in the open field, but one familiar object was replaced with a novel object. The mice were allowed to explore for 15 min. Movement and time spent exploring each object was recorded using Ethovision XT video tracking system (Noldus Information Technology, Sterling, VA) and hand scored by a researcher blinded to the treatment of the mice. The percent time exploring the novel object, out of the total time exploring the novel and familiar objects on day 5, was used to assess novel object recognition. For each group, the preference for the novel versus the familiar object was assessed. The open field arena and objects were cleaned with 5% acetic acid between mice and trials.

Exploration of identical or different environments was used to study the stability of hippocampal networks (Guzowski et al., 1999), using a method called catFISH (cellular compartment analysis of temporal activity using fluorescence in situ hybridization) that relies on the precise temporal kinetics of the IEG and neuronal gene Arc, which is involved in synaptic plasticity and memory, as described previously (Impey et al., 2017). When neurons are engaged in information processing, Arc is rapidly transcribed and can be visualized and quantified after ∼5 min. Subsequently, the mRNA is translocated to the cytoplasm where it remains detectable for ∼20–30 min after the initial transcription. The mRNA is translocated to tagged synapses for protein synthesis. Two different cellular compartments (nuclear and cytoplasmic) can be clearly distinguished, allowing identification of neurons active during distinct behavioral experiences (Guzowski et al., 1999).

Eight mice from each experimental radiation condition were placed individually into a novel environment (A; a square open field; 61 × 61 cm box with 20-cm high walls) and allowed to explore for 5 minutes. Mice were returned to their cage for 25 min, returned to the same environment for an additional 5 min (AA Paradigm). Another 8 mice per radiation dose were allowed to explore environment A for 5 min, and 25 min later they were placed in a different environment (B; a circular arena 45 cm in diameter, and allowed to explore for 5 min; AB Paradigm). Following the last environmental exposure, the mice were killed by cervical dislocation and the brains quickly removed, as described above.

Four to six slides from each treatment group, each containing sections from all the mice in that group, were prepared for fluorescence in situ hybridization. Arc mRNA were detected as previously reported in detail (Rosi et al., 2009; Ramirez-Amaya et al., 2013). Briefly, hapten-labeled antisense riboprobes were hybridized together with the tissues overnight. The digoxigenin-labeled Arc full-probe riboprobe was detected with anti–digoxigenin-HRP conjugate (Roche) and revealed with a cyanine-3 (CY3) substrate kit. Nuclei were counterstained with sytox-green (Molecular Probes).

Microscopic imaging for Arc mRNA, was performed using a Zeiss AXIO IMAGER Z1 microscope with motorized Z-drive for transmitted light and epifluorescence (Rosi et al., 2009). For each end point, the four to six coronal sections per mouse were used to image CA1 and CA3.

Manual cell counts of cells expressing Arc mRNA were performed by an experimenter blind to the relationship between the experimental conditions they represented. Cytoplasmic Arc mRNA–positive neurons were identified when the staining constituted at least 60% of the cell body (Rosi et al., 2009) and was detectable throughout three planes across the Z-stack. To avoid classification errors, we carefully verified that the staining belonged to the cell of interest by checking the nuclear counterstaining (Rosi et al., 2009). The neuronal nuclei were classified as follows: negative (no staining), Arc mRNA foci positive (containing only Arc intranuclear foci staining detected with the Arc intron probe), Arc mRNA cytoplasmic positive (containing only cytoplasmic staining detected with the full-length probe) (Arc-cyto), or as double-labeled for Arc pre-mRNA foci and Arc mRNA cytoplasmic (containing both intranuclear Arc pre-mRNA foci and cytoplasmic Arc mRNA transcripts in two different colors). When a count included all classifications we refer to it as an Arc-positive neuron (this includes the total number of the three above mentioned classifications). The percentages of Arc-foci, Arc-cyto, or Arc-double were calculated relative to the total number of pyramidal neurons included in the analysis as previously reported (Rosi et al., 2009).

Hemibrains were processed for Tet2 immunoreactivity using a specific primary antibody from Santa Cruz Biotechnology (Tet2 S-13, 1:250, catalog number sc-136926) and donkey anti-rabbit Alexa 488 (1:200) as secondary antibody and n = 3 sections per hemibrain, approximately 200 µm apart, per mouse, as described (Impey et al., 2017). Background threshold levels were set and applied to all images. Pixel intensities above this threshold were used for quantification measures (area occupied by pixels and intensities of pixels). The total intensity was also quantified as a measure of overall pixel intensity within a specific brain region.

DNA was isolated from the hippocampus. Antibodies against 5 mC and 5 hmC were used to immunoprecipitate sonicated DNA preparations for methyl-DNA immunoprecipitation (meDIP--anti-5mC mouse mAb, EMD-Millipore NA8; catalog number 162 33 D3) and hydroxymethyl-DNA immunoprecipitation (hmeDIP; anti-5hmC rabbit polyclonal; catalog number 39769; 2 ul, Active Motif)), respectively, and Dynal anti-mouse IgG beads from pools of tissues. Beads were rinsed 7 times with IP buffer, eluted with 1% SDS at room temperature and the eluted DNA purified and subjected to limited amplification (∼18 cycles). Libraries were sequenced on the HiSeq2000 platform at the OHSU Massively Parallel Sequencing Shared Resource or the Oregon State University Center for Genome ReseArch. DIP-Seq regions methylated above “background” were identified using a sliding window method and enriched regions selected via a Monte Carlo-permutation test (Fejes et al., 2008).

To facilitate direct comparison of DIP-Seq data with gene expression data, RNA-Seq was used to profile transcription from the same animals used for the DIP-Seq experiments. RNA was isolated using the NEBnext poly A selection kit (New England Biolabs). Illumina high-throughput sequencing technology was used to profile RNA levels in an unbiased manner. For Illumina RNA-Seq library preparation, NEBnext Ultra kit was used according to the manufacturers specifications (New England Biolabs). Libraries were sequenced on the HiSeq2000 platform at the OHSU Massively Parallel Sequencing Shared Resource. Illumina data were mapped to the UC Santa Cruz assembly using Bowtie (Langmead et al., 2009). For RNA-Seq analyses, tags that overlap with known RefSeq gene models (UCSC RefSeq annotation) were counted using R scripts (Gentleman et al., 2004). Significance was assessed using the DESeq2 package (Love et al., 2014). The Storey Q-test was used to adjust for multiple comparisons (Storey and Tibshirani, 2003).

The cognitive and Arc data are shown as mean ± SEM. The statistical analyses of the data were performed using SPSS™ (Chicago, IL) and GraphPad Prism™ (San Diego, CA) software packages. To analyze locomotor activity over 3 days, ability to locate a visible platform over 2 days, and ability to locate a hidden platform over 3 days, repeated measures ANOVA was used. To compare exploration of the objects and the percentage of Arc-positive cells and total number of Arc cells following exposure to the two different environmental conditions, 2-tailed t-tests were used. The cognitive and Arc figures were generated using GraphPad Prism software. We considered p < 0.05 as statistically significant.

Single read sequence data was mapped to the mouse reference genome (UCSC mm9) using the Bowtie algorithm using standard flags and allowing 2 mismatches (Langmead et al., 2009). Sequences that map to a single location were selected and domains enriched for 5 mC or 5 hmC were selected using a parameter-optimized Monte-Carlo-based segmentation algorithm (Fejes et al., 2008). A 1000 bp sliding-window was used based on iterative analyses that maximized the number of enriched regions. A comparison of different high-throughput sequencing based methods to study DNA methylation concluded that MeDIP-Seq covers ∼67% of genomic CpGs (Harris et al., 2010).

For statistical comparisons of biological samples, regions of methylation enrichment were merged and differences in methylation interrogated with FDR-adjusted negative binomial statistics (Anders and Huber, 2010). Statistical and visualization studies involved the R programming language and Bioconductor packages (Gentleman et al., 2004). Gene ontology analyses utilized the Bioconductor Goseq package, which adjusts for sequence-length bias artifacts in genomic data (Young et al., 2010). KEGG analyses were conducted using the DAVID-EASE site (Dennis et al., 2003). For gene ontology analyses the top DMRs (differentially methylation regions) or DHRs (differentially hydroxymethylated regions) (p < 0.01) within a 50 kb window centered on the RefSeq transcriptional start site or within the RefSeq gene body were non-redundantly annotated. Unless otherwise stated, overlap between DMRs and RNA-Seq data was analyzed using a similar windowing approach. For statistical comparisons of gene-annotated DHRs with DHRs or RNA-Seq gene data we used the Fisher Exact test (“stats” R core library) or for multi-set comparisons (>2 sets) the SuperExactTest R package (Wang et al., 2015). For Fisher Exact tests FDR-adjusted p < 1 × 10⁻³ was considered significant.

DIP sequence-tag heatmaps were generated in R by plotting median-normalized DIP-Seq tag density in gene bodies and indicated flanking regions with color-maps scaled to the 80% quantile. Statistical analyses of pathway data were conducted using FDR-adjusted Fisher exact test. Statistical analyses of DHR density in genomic regions were conducted using a Monte-Carlo-based permutation statistic (“coin” R package). Unless otherwise stated, FDR-adjusted p < 0.01 was considered statistically significant.

The mice were first habituated to an open field for three subsequent days. ²⁸Si ion irradiation did not affect activity levels in the open field over these 3 days (Figure 1A). On day 4, the mice were exposed to the open field containing two objects for 15 min. On day 5, the mice were exposed to the open field with two objects for 15 min, but one familiar object (present on day 4) was replaced with a novel one. There was no effects of radiation on the time the mice spent exploring the objects (Figure 1B). The percent time exploring the novel object was used to assess novel object recognition. Following ²⁸Si (600 MeV) ion irradiation, mice irradiated with 0.3 Gy were impaired, as revealed by no discrimination between the familiar and novel object, while those irradiated with higher doses were not (Figure 1C).

We next assessed whether the mechanisms mediating the cognitive effects of ²⁸Si ion irradiation might be associated with reduced ability of hippocampal neurons to recognize similar environments and to discriminate distinct environments using the temporal kinetics of mRNA encoding Arc. Exploratory behavior activated a comparable total number of neurons in sham-irradiated and irradiated mice in the CA1 (Figures 2, 4A) and CA3 (Figures 3, 4B) regions of the hippocampus. However, there was an effect of ²⁸Si ion irradiation on the percent of neurons activated by both experiences. The percentage of Arc-positive neurons expressing Arc mRNA in the nucleus and cytoplasm in the CA3 region of the hippocampus of sham-irradiated mice was significantly higher following exposure twice to the same environment, as opposed to exposure to two different environments (t = 3.191, p = 0.0110, Figure 4D) and there was a trend towards a difference in the CA1 region of the hippocampus of sham-irradiated mice (t = 2.108, p = 0.0567, Figure 4C). This was not seen in mice irradiated with 0.3 Gy that showed impaired cognitive performance (CA1: t = 1.286, p = 0.2273, Figure 3C; CA3: t = 1.051, p = 0.3280; Figure 2D). In mice irradiated with 0.9 Gy, who showed no impaired cognitive performance (Figure 1B), there was a trend towards a difference in the CA3 region (t = 2.237, p = 0.00557, Figure 4D), but not the CA1 region (t = 0.5842, p = 0.5752, Figure 4C).

We next determined whether ²⁸Si ion irradiation-induced cognitive injury is associated with pathway changes in the hippocampus by comparing hippocampal pathway changes in mice irradiated at a lower dose that caused cognitive injury with hippocampal pathway changes in mice irradiated at a higher dose that did not cause cognitive injury. We did not see effects of ²⁸Si ion irradiation on TET2 immunoreactivity in the dentate gyrus, CA1 or CA3 region of the hippocampus, or cortex. The levels of 5 mC and 5 hmC levels are high and exceptionally dynamic during brain development and aging (Jin et al., 2011; Szulwach et al., 2011), suggesting that they play critical roles.

Recently, we reported alterations in hippocampal DNA methylation following proton (Impey et al., 2016b; Impey et al., 2017) and ⁵⁶Fe ion (Impey et al., 2016a) irradiation. To determine the effects of ²⁸Si ion irradiation on hippocampal DNA methylation, we generated cytosine methylation (5 mC) and/or cytosine hydroxymethylation (5 hmC)-DIP-Seq libraries from the hippocampi of sham irradiated mice and mice exposed to 0.3 or 0.6 Gy of ²⁸Si ion radiation using highly specific antibodies (Impey et al., 2016a; Impey et al., 2016b). These libraries were sequenced to an average depth of ∼14 million tags of which ∼61% mapped to a unique genomic location (Supplementary Table S1). Because 5mc-DIP-Seq tends to map to repetitive areas the lower percentage of unique alignments relative to 5 hmc-DIP-Seq is expected. Genomic regions enriched for 5 mC or 5 hmC were segmented using a previously published Monte Carlo-based algorithm (Impey et al., 2016a; Impey et al., 2017; Johnson et al., 2017) and the union of these regions was tested for significant differences based on a negative binomial distribution (Love et al., 2014). This pipeline identified thousands of differentially methylated and hydroxymethylated regions (DMRs and DHRs, respectively), of which the majority were within 25 kb of a transcription start site (Figure 5A).

Consistent with our and other labs’ DIP-Seq data, DMRs and DHRs distribution showed an expected bias towards genes with high levels of expression or low levels of expression respectively (Figure 5A). Also consistent with our analyses of ⁵⁶Fe irradiation and literature indicating that 5 hmC is a stable epigenetic mark associated with gene expression in brain (Hahn et al., 2014), there was a highly significant enrichment for gene ontology annotation for increased DHRs at the 0.3 Gy dose (Figure 5B). There was weak enrichment for one gene ontology category (“glutamatergic synapse”, FDR-adjusted p < 5 × 10⁻⁴) for decreased DHRs at the 0.3 Gy while all other DHR or DMR conditions showed no significant enrichment for gene ontology categories. Therefore, in this study we focused further analyses on DHRs that showed a significant increase. There was a striking overlap between gene ontology pathways linked to synaptic function and development between the 0.3 Gy ²⁸Si dose and our previous ⁵⁶Fe 0.1 Gy data (Figures 5B, C) (Impey et al., 2016a), suggesting that these two forms of space radiation regulate similar DNA methylation responses. Consistent with this, there was only a ∼14% overlap between ²⁸Si DHR-associated gene ontology categories and our previously-reported proton-irradiation DHR-associated gene ontology analysis (Impey et al., 2017).

We next sought to compare 0.3 Gy ²⁸Si DHRs (600 MeV/n) with previously published proton (150 MeV; 1 Gy) (Impey et al., 2016b) (Impey et al., 2016b), and ⁵⁶Fe data (600 MeV/n, 0.1 Gy) (Impey et al., 2016a) by selecting the most-significant gene-associated DHRs from all data sets (top 3,000 DHRs ranked by p-value; p < 0.01 cut off; duplicate genes removed) and testing for overlap of resulting RefSeq genes. These comparisons revealed highly significant overlap between ²⁸Si DHR-associated genes and analogous ⁵⁶Fe and proton DHR-associated genes (Figure 6A). Interestingly, the overlap between the 0.1 Gy ⁵⁶Fe and 0.3 Gy ²⁸Si sets was markedly more significant (0.3 Gy ²⁸Si vs. 0.1 Gy ⁵⁶Fe, Fisher exact p < 2 × 10⁻⁶⁹; 0.3 Gy ²⁸Si vs. 0.2 Gy ⁵⁶Fe, Fisher exact p < 4 × 10⁻³⁷) than the overlap between the 0.3 Gy ²⁸Si and proton 1 Gy data (Fisher exact p < 5 × 10⁻⁴⁸) (Figure 6A). The overlapping set of DHRs led us to determine the subset of DHR-associated genes that were in common to all three forms of radiation exposure. We identified overlapping genes that were associated with increased hydroxymethylation in response to all three forms of radiation (Superexact test: Figure 6B, p < 3 × 10⁻⁶; Figure 6C p < 3 × 10⁻⁴). The intersection of all 4 data sets was very highly significant (Superexact test: Figure 6D p < 3 × 10⁻²⁰) and this 45 DHR-associated gene “signature” was significantly enriched for “glutamatergic synapse”, “learning”, and “neuron projection” gene-ontology categories which suggest that this set of DHR-associated genes denotes a pathway linked to changes in synapse function and synaptic plasticity (Figure 6E).

A similar set of analyses that compared 0.6 Gy ²⁸Si DHR-associated genes with ⁵⁶Fe and proton DHR-associated genes also found more significant overlap for the 0.1 Gy ⁵⁶Fe data (Figure 7A: 0.6 Gy ²⁸Si vs. 0.1 Gy ⁵⁶Fe, Fisher exact p < 5 × 10⁻⁶⁸) than for the proton 1Gy (Fisher exact, p < 9 × 10⁴⁹) or 0.2 Gy ⁵⁶Fe comparisons (0.6 Gy ²⁸Si vs. 0.2 Gy ⁵⁶Fe, Fisher exact p < 4 × 10⁻³⁰). The DHR-associated gene overlap between all three forms of radiation was again highly significant (Figure 7B, p < 2x⁻¹⁰⁰; Figure 7C, p < 7 × 10⁻⁷⁰) as was the overlap between the proton, ²⁸Si, and two doses of ⁵⁶Fe radiation (Figure 7D, SuperExact test, 9 × 10⁻²⁰⁰).

The 133 gene “4-way” intersection (Figure 7D) was significantly enriched for gene ontology categories linked to synaptic function and neuron projection (Figure 7E) suggesting that 0.6 Gy ²⁸Si irradiation induced a response more closely correlated with our published ⁵⁶Fe and proton radiation responses (Impey et al., 2016a; Impey et al., 2016b). We next tested whether this 133 gene space radiation “signature” contained genes previously linked to radiation in rodent and human models (Xu et al., 2020) and found a non-significant intersection (Table 1) that included genes involved in growth-factor and small GTPase signaling (WIPF1, PIAS1, and IGFR1). Remarkably, the 45 genes in the 0.3 Gy ²⁸Si “4-way” intersection (Figure 6D) were all present in the 133 gene 0.6 Gy ²⁸Si “4-way” intersection (Figure 8A, SuperExact test, p < 7 × 10⁻⁶⁸) indicating that there is a DHR-associated gene signature shared between the two ²⁸Si doses (0.3 Gy, 0.6 Gy), the two ⁵⁶Fe doses (0.1 Gy, 0.2 Gy), and the 1 Gy proton dose (Figure 8B; Table 2) (Cekanaviciute et al., 2023). This space radiation signature of 45 genes contained only one gene, PIAS1, previously linked to radiation response in a database largely limited to “photon” radiation studies (Xu et al., 2020). To provide better context for this space-radiation hydroxymethylation-related gene signature we annotated each gene using PubMed and Google Scholar searches for “radiation”, “ionizing radiation,” or “space radiation;” 23 of 45 genes were associated with studies indicating a radiation response ranging from “radiation response” gene ontology category membership (3), changes in gene expression (19), differential methylation of genetic loci (2), and radiation-associated mutagenesis or crosslinking (3) (Table 2). Interestingly, Npas3 was identified as a significant space-radiation decreased gene (¹²C ion beam) (Iwakawa et al., 2008) while Ncald and Parkn were identified as genes that are preferentially mutated in response to space radiation-associated DNA damage (⁴⁰Ar and ⁶⁵Fe ion beam) (Cekanaviciute et al., 2023).

We next utilized RNA-Seq to profile the transcriptional response to ²⁸Si irradiation using hippocampal RNA co-extracted from the same tissue used for DIP-Seq experiments. Although our ⁵⁶Fe irradiation study (Impey et al., 2016a) identified highly significant changes in gene expression that correlated with DHRs (but not DMRs), we did not observe a significant transcriptional response to either 0.3 Gy or 0.6 Gy ²⁸Si irradiation at the 2 weeks time point (no genes reached the FDR-adjusted p < 0.01 cut off, Supplementary Tables S1, S2). This result is consistent with our observation that ²⁸Si irradiation at the 2 weeks time point only demonstrated significant gene ontologies at the 0.3 Gy dose whereas we saw highly significant gene ontology and KEGG pathway enrichment for both down and upregulated DHRs following 0.1 Gy and 0.2 Gy ⁵⁶Fe irradiation (Impey et al., 2016a). Because gene expression pathway data is often regulated in a biologically meaningful manner even when individual genes do not achieve FDR-adjusted significance, we selected differentially-expressed genes at an unadjusted p-value cut off of p < 0.01 and assessed gene-ontology category and KEGG pathway enrichment via the Fisher-exact test. For genes increased by the 0.3 Gy ²⁸Si dose, gene ontology categories associated with mitochondrial oxidative phosphorylation were most-significantly enriched as was the KEGG pathway for oxidative phosphorylation (Figure 9A; Supplementary Figure S1). Interestingly, we also observed significant enrichment for KEGG pathways associated with Huntington’s, Alzheimer’s, and Parkinson’s diseases (Supplementary Figures S2–S4). We also saw enrichment for Huntington’s, Alzheimer’s, and Parkinson’s disease KEGG pathways for genes significantly decreased by ⁵⁶Fe irradiation in a previous study (Impey et al., 2016a), suggesting that these pathways may represent a more generalized space-radiation response. Because oxidative stress has been linked to the neuropathology of these neurodegenerative disorders (Kim et al., 2015), our data suggest that a 0.3 Gy ²⁸Si radiation dose may trigger persistent oxidative stress that adversely impacts neuronal homeostasis/metabolism. Genes decreased by 0.3 Gy ²⁸Si irradiation were significantly enriched for synaptic components and neuronal proteins (Figure 9B) which is consistent with the idea that the radiation-induced oxidative stress adversely impacts synaptic function/homeostasis.

Genes increased by 0.6 Gy ²⁸Si irradiation were enriched for gene ontology categories associated with synaptic and neuronal proteins (Figure 9C) while genes decreased by 0.6 Gy ²⁸Si irradiation showed less-significant enrichment for neuron-associated gene ontology categories (Figure 9D) and KEGG pathways (Supplementary Figure S5). The Alzheimer’s disease KEGG pathway was significantly enriched for genes decreased by 0.6 Gy ²⁸Si irradiation which is the opposite effect seen for 0.3 Gy ²⁸Si. It is conceivable that the 0.6 Gy ²⁸Si dose triggered compensatory neuroprotective gene expression changes. Gene ontology analyses of indicated intersections between gene-annotated DHRs and differentially-expressed genes are illustrated in Figure 10.

We next assessed the correlation between gene-associated differential DNA hydroxymethylation and gene expression. Although we did not detect the highly-correlated spatial correlation of DHRs and gene expression seen for ⁵⁶Fe irradiation (Impey et al., 2016a), we found significant overlap between downregulated genes and both increased and decreased DHRs at the 0.3 Gy and 0.6 Gy ²⁸Si radiation doses (Supplementary Figure S6). These data suggest that by intersecting DIP-Seq data with RNA-Seq data, we increased statistical power allowing us to detect biologically relevant epigenetic remodeling. Consistent with this idea, the highly significant intersection between decreased 0.3 Gy ²⁸Si DHRs and both up and downregulated genes were significantly enriched for gene ontology categories linked to synapses, neuronal connectivity, and neuronal development (Supplementary Figure S7). We also identified the “mitochondrion” gene ontology category when comparing increased 0.3 Gy ²⁸Si DHRs with increased 0.3 Gy ²⁸Si gene expression and the “nervous system development” gene ontology category for the intersection of decreased 0.6 Gy ²⁸Si DHRs with decreased 0.6 Gy ²⁸Si gene expression.