Genome-wide transcriptomics identifies an early preclinical signature of prion infection

The clinical course of prion diseases is accurately predictable despite long latency periods, suggesting that prion pathogenesis is driven by precisely timed molecular events. We constructed a searchable genome-wide atlas of mRNA abundance and splicing alterations during the course of disease in prion-inoculated mice. Prion infection induced PrP-dependent transient changes in mRNA abundance and processing already at eight weeks post inoculation, well ahead of any neuropathological and clinical signs. In contrast, microglia-enriched genes displayed an increase simultaneous with the appearance of clinical signs, whereas neuronal-enriched transcripts remained unchanged until the very terminal stage of disease. This suggests that glial pathophysiology, rather than neuronal demise, could be the final driver of disease. The administration of young plasma attenuated the occurrence of early mRNA abundance alterations and delayed signs in the terminal phase of the disease. The early onset of prion-induced molecular changes might thus point to novel biomarkers and potential interventional targets.

3 creating longitudinal maps of the transcriptional equivalents of neurotoxicity from the administration of prions to the development of terminal disease.
Transcriptional maps of prion-inoculated mice were previously performed using microarrays [9][10][11][12] , and more recently also by RNA sequencing (RNAseq) 13 . Most of these studies analyzed whole-brain homogenates and have focused on alterations occurring in the late phase of the disease. Yet no antiprion compounds could reverse the progression of disease 14 , perhaps because intervention was too late. Instead, focusing on the preclinical changes occurring in the affected brain may uncover alternative targets for early diagnosis and therapeutic intervention in pre-symptomatic subjects. Here we performed a comprehensive analysis of transcriptional alterations in the hippocampus of prion-infected mice over time. We identified an unexpected wealth of changes in the early phase of prion replication, long before any clinical signs of disease. Neuronal expression changes became evident only at the terminal stage of the disease, whereas the appearance of clinical symptoms coincided with microglia alterations. Prion-induced molecular changes were largely unaffected by ageing, yet the administration of young plasma attenuated early prion-induced changes and improved the health span of diseased mice.

Identification of mRNA expression changes during prion disease
To identify molecular alterations associated with the progression of prion disease, we administered intracerebrally RML6 prions or non-infectious brain homogenate (NBH, for control) to C57BL/6 mice (Fig. 1a).
Mice were sacrificed at 4,8,12,14,16,18 and 20 wpi (weeks post inoculation), as well as the terminal stage (the last time point during disease progression when mice can be humanely euthanized). We specifically focused on the hippocampal region, which plays a central role in memory formation and consolidation and is strongly affected in multiple neurodegenerative disorders including prion disease. RNA from ipsi-and contralateral hippocampi was extracted and subjected to RNA sequencing (n=3 per prion and control samples Hierarchical clustering analysis of total transcriptomes revealed a progressive segregation of prion-inoculated and control mice starting at 18 wpi ( Supplementary Fig. 1a). However, principal component analysis (PCA) showed a separation of the two groups already at 8 wpi ( Supplementary Fig. 1b). Next, we identified 3,723 differentially expressed genes (DEGs) between prion-infected and control mice (absolute log2 fold changes |log2FC| > 0.5 and false discovery rate (FDR) < 0.05 at least at one of the time points; Fig. 1a Furthermore, a striking ~60% and ~80% of the 8 and 16-wpi decreasing genes, respectively, overlapped with each other or the terminally downregulated genes (Fig. 2c). Interestingly, the initial downregulation at the 8and 16-week timepoints was transient, indicating that compensatory mechanisms might act during the earlier stages of prion disease.
To validate these changes, we analyzed the 8 wpi and terminal time point in a second, validation cohort of mice, which were inoculated and analyzed independently from the first cohort ( Supplementary Fig. 3a,b and   4). Consistently, we observed that cluster 4 genes (oscillating genes) decreased at 8 wpi and the terminal stage, and that cluster 3 genes (mainly neuronal genes) decreased at the terminal stage. Furthermore, genes belonging to clusters 1 and 2 increased at the terminal stage ( Supplementary Fig. 3c-f). Collectively, our data shows that hundreds of genes are downregulated at a very early disease stage, suggesting that this early response is linked to primary pathogenic events rather than reactive changes.

Microglia activation drives symptomatic prion disease progression
We next aimed at correlating the gene expression changes with the progression of neuropathological changes.
Mice reached the terminal stage of the disease at 170-180 days, at which point they were sacrificed (median survival: 176 days; Supplementary Fig. 5a). Brain sections of prion-inoculated mice were assessed for morphological changes (haematoxylin/eosin), for astro-and microgliosis (GFAP and IBA1 staining) and for the presence of protease-resistant prion protein (PrP Sc ; protease treatment followed by SAF84 staining).
Spongiosis, astro-and microgliosis became evident at 16 wpi and increased at the terminal stage, whereas PrP Sc started to accumulate already at 12 wpi ( Supplementary Fig. 5b). These results are in line with the transcriptomic data discussed above. 6 We next determined the concentration of prion propagons using a quaking assay. We found that the lag phase of the assay, which measures templated nucleation of PrP fibrils, continually decreased during disease progression ( Supplementary Fig. 6a), suggesting that the amounts of infectious units increased up to the terminal stage. In parallel, we analyzed prion infectivity using the scrapie cell assay in endpoint dilution format, which yielded similar results as the quaking assay and confirmed that infectivity continuously increased from 8 wpi to the terminal stage. Prion titers were significantly higher at the terminal stage compared to 16 wpi ( Supplementary Fig. 6b), which is in contrast to previous reports stating that infectivity would reach a plateau at 16 wpi 7 .
To relate gene expression changes to neurological dysfunction, we assessed the motor performance of prioninoculated and control mice using a rotarod test. We observed a progressive decline in motor performance starting at 18 wpi (Fig. 3a). The onset of impairment was synchronous to increased expression of microgliaenriched genes and was unlinked to decreased expression of neuronal genes which became evident only at the terminal stage (Fig. 3b). A linear regression analysis between rotarod performance and the expression change of each of the 3723 DEGs, identified a significant correlation for 347 DEGs (p < 0.001). Examples of correlated non-enriched, microglia-enriched and neuronal-enriched genes are shown (Fig. 3c). While the linear regression slope of microglia genes was negative, neuronal genes exhibited a positive linear regression slope.
This was confirmed when we compared the average slope between DEGs enriched in different cell types (Fig.   3d). We conclude that motor decline, neuropathological changes and an increase in microglia gene expression occur simultaneously, long before terminal changes in neuronal gene expression and neuronal loss become evident. This suggests that microglia, rather than neurons, are the final drivers of prion disease progression.
However, the transcriptional changes observed at 8 wpi precede both the onset of clinical signs and any changes in microglial gene expression. The early 8 wpi changes might thus hierarchically control microglia changes and ultimately induce pathogenesis.

Identification of post-transcriptional changes during prion disease progression
Many RNA binding proteins are exclusively expressed in neurons, and aberrant splicing has been linked to multiple neurodegenerative diseases 16 . We identified a total of 426 isoforms that were differentially expressed in prion-inoculated mice at one or more time points (FDR < 0.05; Fig. 4a; Supplementary File 8). 102 of these splice isoforms mapped to DEGs, indicating that differential splicing of these transcripts might impact their abundance (Supplementary File 8). The 426 isoforms correspond to 239 splicing events, which mapped to 228 genes. Most of these genes showed just one significantly changing splicing event, except App, Chl1, Evl, Fus, Neo1, Olfm1, Picalm, Ppfia4 and Sorbs1. The majority of splicing changes consisted of exon 7 inclusion/skipping followed by alternative transcript starts/ends (Fig. 4b). And while most splicing changes occurred only at the terminal stage, many splicing events showed a similar trend (p < 0.05) at multiple time points during disease progression. Most strikingly, we observed an oscillating pattern of isoform expression at 8 wpi, 16 wpi and the terminal stage (Fig. 4c), indicating that these timepoints are marked by a characteristic RNA expression and splicing signature. Select differentially spliced isoforms are shown in Supplementary Fig. 7 (indicated in Supplementary File 8).
Skipping of exon 17 of Picalm, a susceptibility gene for late-onset Alzheimer's disease 17 was increased at the terminal stage of prion disease. Skipping of exon 17 introduces a premature stop codon, which leads to the production of a truncated protein, and is thought to affect clathrin-mediated synaptic endocytosis 18,19 . The inclusion of exons 7 and 8 of App, a well-described splicing event 20 increases both at 8 wpi as well as the terminal stage. Similarly, alternative splicing of the synapsin genes, Syn1 and Syn2, and of Ctsa, a disease associated microglia (DAM) gene 21 changes at both 8wpi and the terminal stage. While mRNA levels of Ctsa increased at the terminal stage, they remained unchanged at 8 wpi, indicating that the regulation of splicing and mRNA abundance of this gene are unlinked.
A further post-transcriptional mechanism that is essential for nervous system homeostasis is RNA editing, whose dysregulation has been implicated in neurodegeneration 22 . The most common form of RNA editing, adenosine (A) to inosine (I) conversion, is mediated by the adenosine deaminase acting on RNA (ADAR). The expression levels of the ADAR enzymes, Adar1 and Adarb2 did not change during prion disease progression, whereas Adarb1 was significantly downregulated at the terminal stage (Supplementary File 3). We did not observe major editing changes at any of the analyzed time points, even upon merging main and validation datasets (Supplementary File 9). Only two analyzed sites were differentially edited in terminally diseased mice ( Supplementary Fig. 8). The two sites are in the 3' untranslated regions of Sh2d5 and Padi2, and while the mRNA expression of Sh2d5 at that timepoint did not change, Padi2 expression increased in terminally diseased mice.

The impact of aging on prion disease progression
The prevalence of neurodegenerative diseases increases drastically with age. Sporadic CJD typically manifests in 55-65 years old individuals 23 . This age in humans is comparable to approximately 12 months of age in mice 24 . To assess if age-related changes accelerate prion disease progression, we compared disease progression in young vs. aged mice. We inoculated a third cohort of 2-month old "young" mice (similar to cohorts #1/2) and 1-year old mice. We analyzed gene expression changes at 8 wpi in aged mice and at the terminal stage in young and aged mice. Similar to the previously analyzed samples, we observed a separation 8 of control and prion-inoculated samples at 8 wpi and the terminal stage (Supplementary Fig. 4 and 9a,b; Supplementary File 10). Remarkably, the median lifespan of aged prion-inoculated mice was only slightly, albeit significantly (p = 0.0001), shorter than that of young inoculated mice (Fig. 5a). This suggests that age does not strongly accelerate prion disease progression and that disease manifestation is similar in young and aged mice. Consistently, a correlation of gene expression changes was noticeable at 8 wpi (R = 0.39; Fig. 5b).
Unsurprisingly, the changes in gene expression in young and aged mice converged towards the terminal stage (R = 0.89; Fig. 5c).

Young plasma treatment ameliorates prion disease progression
Infusion of plasma from young mice can revert age-related impairments of cognition and synaptic plasticity 25 and ameliorates neuronal hippocampal dysfunctions in murine models of Alzheimer's Disease 26 . We therefore analyzed the impact of young plasma on the course of prion disease as well as on the early prion-induced changes at 8wpi. Mice inoculated with prions or control brain homogenate were treated for 8 weeks with young plasma or saline via bi-weekly intravenous injections. Hippocampal samples were subjected to high-throughput sequencing at 8 wpi and at the terminal stage (Supplementary Fig. 4 and 10a,b; Supplementary File 11). We also monitored the rotarod performance and assessed mice for neurological symptoms from 18 wpi onwards by scoring for the absence/presence of a hunched posture, rigid tail, piloerection, hind limb clasping and ataxia ( Fig. 6a). Remarkably, plasma administration reduced essentially all prion-induced changes at 8 wpi (Fig. 6b).
Neurological deficits, which usually occur at 20 wpi, became evident only at 23 wpi in plasma-treated animals ( Fig. 6c). Similarly, rotarod performance decreased at a later stage in plasma-treated samples ( Supplementary   Fig. 10c). In contrast, we only observed a mild, non-significant increase in the median lifespan upon plasma administration (6d), and, as expected, the terminal prion-induced changes were very similar between plasma and saline-treated mice (Fig. 6e). This suggests that young plasma treatment might improve the health span but not the lifespan of prion diseased mice.

Discussion
Here we have correlated the temporal sequence of transcriptional, splicing, and RNA editing events with neuropathological changes, prion infectivity, and clinical symptoms, and have generated a searchable genome-wide database of prion-induced changes during disease progression. As expected, the histological alterations (microgliosis and astrogliosis, spongiform changes and PrP Sc deposition) and clinical signs were mirrored by pronounced gene expression changes at terminal disease. Surprisingly, onset and progression of motor dysfunctions correlated precisely with the onset of glial changes and occurred long before neuronal loss was detectable. This suggests that glial perturbation, rather than neuronal demise, is the driver of prion disease progression. Moreover, we observed pronounced gene expression changes at 8 wpi, when neither neuropathological changes, prion infectivity, nor clinical symptoms are detectable. These early gene expression changes might provide new starting points for the development of novel therapeutics and diagnostics.
We further detected the differential expression of 426 splicing isoforms. Most splicing differences occurred both at terminal disease and at the preclinical 8 wpi timepoint. More than 100 of these splicing isoforms mapped to genes that were also differentially expressed, indicating that splicing and expression of these genes is linked. Nine genes were differentially spliced at multiple sites, including App, Chl1, Evl, Fus, Neo1, Olfm1, Picalm, Ppfia4 and Sorbs1. Several of these genes have been linked to neurodegeneration, suggesting that splicing dysregulation of these genes might be common to multiple neurodegenerative diseases. In contrast, only two previously identified editing sites in 3'UTRs showed a significant change in our dataset, at the terminal disease stage. These results were inconsistent with a recent study 13 , identifying a prion-induced editing signature in mouse cortex, which could at least partially be confirmed in human autopsy brain. It remains to be seen if prion-induced editing changes differ between brain regions, or if distinct murine prion disease models and different editing analyses account for this discrepancy.
Our experiments also show that prion titers never reach a plateau but continue to rise until the terminal stage of disease. Gene expression changes occur in parallel with initial prion replication, while amounts of infectious units progressively increases over disease incubation time. Earlier investigations [6][7][8] have hypothesized that scrapie neurotoxicity is only triggered when a "lethal" form of prion protein (denoted "PrP L ") accumulates in late pathogenesis, in a second phase of the disease, and drives toxicity and clinical disease. Conversely, prion replication would initially occur without associated toxicity. However, no physical evidence of "PrP L " has ever surfaced. In addition, the two-phase hypothesis rests on the quantification of prion infectivity and histological analyses fraught with analytical imprecision and poor sensitivity. The discovery of early, specific, robust and reproducible disease-associated gene expression changes before the saturation of prion infectivity refutes the two-phase hypothesis and suggests that pathology starts as soon as prions enter the brain.
Surprisingly, our analysis indicates that early alterations can be initially overcome. The existence of compensatory mechanisms is suggested by the oscillating gene expression patterns that were observed. The early downregulation of cluster-4 genes was transient and underwent complete recovery. Another peculiar expression profile was observable for certain microglial markers which were monotonically upregulated from 16 wpi. This included phagocytosis-related genes (Aif1, Dock2, Fgr, Fcgr1, Fcgr2b, and Fcgr3) and diseaseassociated microglia (DAM) genes (Itgax, Clec7a, Cxcl10 and Lag3) 27 . Remarkably, the upregulation of these genes correlates precisely with the onset of motor deficits whereas no such correlation existed with neuronal genes. We therefore hypothesize that it is glial rather than any neuronal alterations that cause clinical disease manifestation. This aligns with gliosis being observed in symptomatic but not in pre-symptomatic CJD patients 28 . Alternatively, the dysfunction of a small number of neurons at 16-18 wpi may drive clinical signs while remaining below the detection threshold of RNAseq.
Clinically overt Creutzfeldt-Jakob disease progresses much more quickly than most other neurodegenerative syndromes, making the development of effective therapies extremely difficult. However, our data suggest that the critical period for potential intervention may be much earlier than previously estimated 6,10 . Accordingly, infusion of young plasma during the first 8 weeks after prion inoculation attenuated the early downregulation of genes involved in neuronal functions, and also delayed the manifestation of neurological signs as observed in models of Alzheimer's disease 26 . Interestingly, the beneficial effects of young plasma did not translate into a significantly extended overall survival, perhaps because primary prion replication was not modified despite attenuated downstream neurotoxicity. Similarly, the survival curves of young and middle-aged prion inoculated mice were similar, suggesting that ageing has only minor effects on early and late transcriptional changes as well as lifespan. And yet the plasma experiment suggests that treatment aimed at rescuing the molecular alterations identified in the clinically silent phase might impact the consequent entire progression of the disease. This idea is also in line with emerging results of novel potential therapies which seem to be effective only when started either prophylactically or within the first 11 weeks after prion inoculation (Vallabh S., et al., from Minikel E. cureffi.com; May 23, 2019). The fact that young plasma seems to improve the health span, rather than the lifespan, of prion disease may be relevant to improving the life quality of patients with neurodegenerative diseases.
It remains to be seen whether the molecular phenomena and kinetics detected in mouse models of prion disease reflect those of the corresponding human disorders. If so, the results detailed here may instruct the development of early diagnostic and therapeutic approaches. Finally, the molecular mechanisms underlying prion diseases may also be of relevance for other common protein misfolding disorders, such as Alzheimer's and Parkinson's disease.

Mice
Animal experiments were performed in compliance with the Swiss Animal Protection Law, under the approval of the Veterinary office of the Canton Zurich (animal permits ZH41/2012, ZH90/2013, ZH040/15, ZH243/15).
All efforts were made to prevent or minimize animal discomfort and suffering; individual housing was avoided.
Prion-inoculated and control-injected mice were regularly monitored for the development of clinical signs, according to well established procedures. Humane termination criteria were employed. Intracerebral injections,

Prion Inoculations
Imported mice were allowed at least one week of habituation in our animal facility before experimental manipulations. Six-week or 51-week old C57BL/6 male mice (Supplementary File 1) were injected in the right hemisphere with 30 μl of RML6 (passage 6 of Rocky Mountain Laboratory strain mouse-adapted scrapie prions) at a 10 -2 dilution of a 10% homogenate, containing 9.02 log LD50 of infectious units per ml in 10% homogenate. Control inoculations were performed using 30 µl of non-infectious brain homogenate (NBH) from CD-1 mice at the same dilution. Each experimental mouse was randomly assigned a code and, based on this, to an experimental group/time point. After inoculation, mice were initially monitored three times per week. After clinical onset, mice were daily monitored. Mice were sacrificed at pre-defined time points based on their experimental group, starting always at the same time of the day and alternating between prion-inoculated and NBH-injected mice. Prion-inoculated mice allocated to the terminal group were sacrificed upon clear signs of terminal prion disease. Control-injected mice assigned to the latest time point group were sacrificed at 192 days post-inoculation, 13 days later than the last terminal prion-inoculated mouse. Transcardial perfusion with ice-cold PBS was performed in deeply anesthetized mice.

Plasma collection and administration, and neurological scoring
Plasma preparation and administration was performed as previously described 25 , with minor modifications.
Peripheral blood was collected from 150 young (6 weeks old) mice using retro-orbital bleeding in deeply anesthetized mice who were then immediately sacrificed by decapitation. Pooled mouse plasma was separated from blood collected with EDTA by centrifugation at 1,000g, followed by dialysis using 3.5-kDa Dtube dialyzers (EMD Millipore) in PBS to remove EDTA, aliquoting and storage at −80 °C until use. Systemic administration of either plasma or saline as control was performed by injection of 100µl in the tail vein of mice who had received an intra-cerebral injection with either RML6 prions or NBH. Treatment started the day after the intracerebral injection and was performed twice weekly for 8 weeks. Mice were sacrificed at 8 wpi or the terminal stage (n=5 per each group) by transcardial perfusion with ice-cold PBS in deeply anesthetized mice.
Hippocampi from both hemispheres were dissected, snap frozen and stored at -80°C until processing. RNA extraction and sequencing were performed in two separate rounds (Supplementary File 1) but were pooled for analysis.
Neurological scoring was performed weekly from 18 wpi until 24 wpi by an observer who was blinded to experimental group allocation (RML or NBH-injected, plasma-or saline-treatment), and encompassed the assignment of scores from 0 (no abnormality) to 1-2-3 (for mild, moderate or severe abnormalities) in the following domains: hunched posture, piloerection, rigid tail, hind limbs clasping, ataxia.

Immunohistochemistry
Histological analyses were performed on 2-µm thick sections from formalin fixed, formic acid treated, paraffin embedded brain tissues, as previously described 29 . Sections were subjected to deparaffinization through graded alcohols, followed by heat-induced antigen retrieval performed in EDTA-based buffer CC1 (Ventana).

Rotarod
The rotarod test was performed as previously described 29 , with minor modifications. The rotarod apparatus

Standard Scrapie Cell Assay
CAD5 cells were grown with standard OFBS Medium (Opti-MEM containing 10 % FBS, 1 % streptomycin and penicillin, 1 % Glutamax; Gibco) in a T150 cell culturing flask. One day prior to infection, 10,000 CAD5 and CAD5 KO cells lacking PrP C expression were plated with 100 µL OFBS in 96-well cell culture plates (TPP) and incubated at 37 °C with 5% CO2. On the following day, 100 µL of brain homogenate diluted in OFBS mixed with 0.01% brain homogenate from C57BL/6J-Prnp ZH3/ZH3 mice to provide a complex matrix was added to the cells for the infection. To establish a standard curve for infection, a 1:5 serial dilution of RML6 brain homogenate (20% w/v in 0.32M sucrose, 10 LD50 units per mL) was used with a range from 1 × 10 −3 to 6.4 × 10 -8 . For each sample, three different dilutions were performed ranging from 1 × 10 −3 to 1 × 10 −5 . To control for residual inoculum, CAD5 KO cells were incubated with RML brain homogenate corresponding to the highest concentration of the standard (0.01%). CAD5 cells were incubated with (0.01%) non-infectious brain homogenate (10% w/v in 0.32M sucrose) to control for efficient proteinase K (PK) (Roche) digestion and to compute the background of the assay. Three days following infection, cells were split 1:8 into new 96 well plates containing fresh OFBS. After reaching confluence, two additional 1:8 splitting steps were performed, corresponding to days 7 and 10 post infection. On day 14 post infection, ELISPOT membranes (Millipore) were activated by adding filtered 50 µL ethanol/well, washed twice with 160 µL PBS and nearly 40,000 cells per well transferred onto the membrane and dried with a plate thermomixer (Eppendorf) at 50 °C. After drying, plates were stored at 4 °C until lysis and digestion. 50 µL of 0.5 ug/mL PK in lysis buffer (50 mM Tris-HCl pH8, 150 mM NaCl, 0.5% w/v sodium deoxycholate, 0.5% w/v Triton-X-100) was added to each well and incubated for 90 minutes at 37 °C. Following incubation, vacuum was applied to discard the contents and wells were washed twice with 160 µL PBS. To stop digestion, 160 µL of 2 mM PMSF (Sigma Aldrich) diluted in PBS was applied to the membrane and incubated at room temperature for 10 min. Tris guanidinium thiocyanate was prepared by diluting 3 M guanidinium thiocyanate in 10 mM Tris HCl pH8, and added subsequently with a total volume of 160 µL/well and incubated for 10 min. Supernatant was discarded into 2M NaOH and membrane was

Real-time quaking induced conversion assay (RT-QuIC)
RT-QuIC assays of prion-infected mouse brain homogenates were performed as previously described 30 .

Data analysis
Alignment and feature counting: RNA sequencing data analysis was performed as previously described 31 , with modifications. Quality control of reads was performed using FastQC. Low-quality ends were clipped (5': 3 bases; 3': 10 bases). Trimmed reads were aligned to the reference genome and transcriptome (FASTA and GTF files, respectively, downloaded from the UCSC mm10) with STAR version 2.3.0e_r291 32 with default settings.
Differential gene expression: Differentially expressed genes were identified based |log2FC| > 0 and FDR < 0.05 using the R package edgeR 33 from Bioconductor (version 3.0). Only genes with at least 10 counts in at least 50% of the samples in one of the groups were considered in the analysis. Differentially expressed genes (DEGs) were defined as genes changing with |log2FC| > 0.5 and FDR < 0.05. Log2FC with p ≥ 0.05 were set to 0 for unsupervised kmeans clustering, t-distributed stochastic neighbor embedding (t-SNE) and the comparison with rotarod performance (Fig. 1b, c, 3b -d, and Supplementary Fig. 2a). RNA isolated from saline and plasma-treated mice was isolated, processed and sequenced in two independent runs (indicated in Supplementary Fig. 10a, b). All saline and plasma-treated mice were analyzed together. An additional covariate was added to the model in edgeR to account for the batch effect associated with the different runs.
To assess if DEGs changed at multiple time points (Fig. 2a, c and Supplementary Fig. 2b) we lowered the cutoff to |log2FC| > 0.5 and p < 0.05.
Alternative splicing: Analysis of splice variants was performed by using SGseq 34 and DEXseq 35 R packages as described previously 36 with some modifications. SGseq was first applied to identify splicing events (e.g. cassette alternative exon) related to two or more variants (e.g. isoforms with exon included or exon skipped).
Exons and splice junctions were predicted from BAM files. Predictions for each sample were merged to obtain a common set of transcript features, and exons were partitioned into disjoint exon bins. A genome-wide splice graph was assembled based on splice junctions and exon bins, and splice events were identified from the graph. To determine differential splicing events, a single value for each variant was produced by either adding up the 5' and 3' counts, or, if these represented the same transcript features, by considering the unique value.
These counts then constituted the initial input to DEXSeq, as described in the SGSeq manual. Briefly, instead of quantifying differential usage of exons across a single gene, we analyzed differential usage of variants across a single event. Such adaptation of DEXSeq is also reported in the DEXSeq vignette. Similar to the differential gene expression analysis, we retained only variants with at least five counts in at least three samples (of any condition). After this filtering step, in events associated with a unique variant, such variant was considered to be constitutive and discarded. For most comparisons, these two filters reduced the total number of variants tested to around 6,000. Differential analysis was then performed implementing a sample+exon+condition: exon model in DEXSeq. To limit the number of tests, the first variant of each event (generally a 'skipping' variant) was discarded. As our data, possibly due to the Nugen amplification, shows high levels of intronic reads, retained intron events were also discarded. Differentially expressed isoforms were defined as isoforms changing with FDR < 0.05. Log2FC with p ≥ 0.05 were set to 0 for unsupervised kmeans clustering (Fig. 4c). To compare if splicing isoforms were differentially expressed at multiple time points (Fig.   4a) we lowered the cut-off to p < 0.05.
RNA editing: Examination of RNA editing was conducted as previously described 37 . A catalogue of loci with mismatches with respect to the reference genome was created following RNAsequencing based best practice workflow compiled for GATK version 3.4. 38 . 17,831 RNA editing sites have previously been reported in mice 39 , and were further investigated. For the identification of differentially edited loci, data from both cohort#1 and cohort#2 were analyzed together. For each locus we considered a summary statistic (the sum) of the counts of the allelic observations across the samples in each group. In the pair-wise comparisons, we then constructed a contingency table based on edited/unedited observations at the locus in the prion/ctrl groups. An exact Fisher test was performed, and the p-values were Benjamini-Hochberg adjusted for multiple testing. Neighbor Embedding (t-SNE) method was used for projecting high dimensional gene expression data in a 2D space. We used the t-SNE implementation from scikit-learn library in Python with the following parameters: learning rate=200, n iter=1,000, random state = 0, metric=Euclidean, init=pca.

Data availability
All data will be deposited in the GEO database. In addition, gene expression profiles during disease development can be easily visualized and downloaded at the following website: http://histodb12.usz.ch/iMice/public/PrionRNASeqDatabase/PrionRNASeqDatabase.php

Code availability
Code is available from the authors by request.          Summary table of 114 sequenced mice   Table including sequencing and sample ID, dataset,

Supplementary File 4: GO analysis of mDEGs upregulated 16wpi
Gene Ontology analysis of terms related to biological processes, molecular function and cellular localization.
87 genes (85 of which were in the database) were upregulated from 16wpi onwards and were assessed for GO term enrichment compared to 16,127 expressed genes (15,493 of which were in the database).

Supplementary File 5: GO analysis of mDEGs upregulated 18 wpi
Gene Ontology analysis of terms related to biological processes, molecular function and cellular localization.
440 genes (435 of which were in the database) were upregulated from 18 wpi onwards and were assessed for GO term enrichment compared to 16,127 expressed genes (15,493 of which were in the database).

Supplementary File 6: GO analysis of terminally downregulated DEGs
Gene Ontology analysis of terms related to biological processes, molecular function and cellular localization. 102 splice isoforms map to genes that are differentially expressed in the main dataset. The respective RNA expression information is included for these 102 entries: cluster number (corresponding to clusters shown in Figure 1), and all eight edgeR comparisons related to the main dataset (expression, log2FC, P value and FDR).  Figure 7). Shown are genomic coordinates (chromosome, start, end, width, strand), reference and alternative genomic sequence, the corresponding gene, the fraction of edited reads (of total reads) per sample and P value and adjusted P value of all eight comparisons (main and validation datasets were analyzed together). GeneID, cell type enrichment if applicable (_NE = not enriched; AS = astrocytes; EC = endothelial cells; MG = microglia; N = neurons; OL = oligodendrocytes), cluster number (corresponding to clusters shown in Figure   1), and information on the four edgeR comparisons related to the aging dataset. For each comparison expression, log2FC, P value and FDR are included.   immune resp.

Supplementary File 11: Plasma-induced gene expression changes
resp. to external biotic stimulus reg. of immune system proc.
resp. to biotic stimulus resp. to other organism pos. reg. of immune system proc.