Deletion of mFICD AMPylase alters cytokine secretion and affects visual short-term learning in vivo

Fic domain-containing AMP transferases (fic AMPylases) are conserved enzymes that catalyze the covalent transfer of AMP to proteins. This posttranslational modification regulates the function of several proteins, including the ER-resident chaperone Grp78/BiP. Here we introduce a mouse FICD (mFICD) AMPylase knockout mouse model to study fic AMPylase function in vertebrates. We find that mFICD deficiency is well tolerated in unstressed mice. We also show that mFICD-deficient mouse embryonic fibroblasts are depleted of AMPylated proteins. mFICD deletion alters protein synthesis and secretion in splenocytes, including that of IgM, an antibody secreted early during infections, and the proinflammatory cytokine IL-1β, without affecting the unfolded protein response. Finally, we demonstrate that visual nonspatial short-term learning is stronger in old mFICD−/− mice than in wild-type controls while other measures of cognition, memory, and learning are unaffected. Together, our results suggest a role for mFICD in adaptive immunity and neuronal plasticity in vivo.

Fic domain-containing AMP transferases (fic AMPylases) are conserved enzymes that catalyze the covalent transfer of AMP to proteins. This posttranslational modification regulates the function of several proteins, including the ER-resident chaperone Grp78/BiP. Here we introduce a mouse FICD (mFICD) AMPylase knockout mouse model to study fic AMPylase function in vertebrates. We find that mFICD deficiency is well tolerated in unstressed mice. We also show that mFICDdeficient mouse embryonic fibroblasts are depleted of AMPylated proteins. mFICD deletion alters protein synthesis and secretion in splenocytes, including that of IgM, an antibody secreted early during infections, and the proinflammatory cytokine IL-1β, without affecting the unfolded protein response. Finally, we demonstrate that visual nonspatial shortterm learning is stronger in old mFICD −/− mice than in wildtype controls while other measures of cognition, memory, and learning are unaffected. Together, our results suggest a role for mFICD in adaptive immunity and neuronal plasticity in vivo.
The posttranslational regulation of protein function is a fundamental concept in biology. To manage protein activity, dedicated enzymes attach specific chemical modifications to individual proteins, the presence of which affects the behavior and activity of the modified proteins. These modifications, called posttranslational modifications (PTMs), govern essential biological processes. They are implicated in cancer, neurodegeneration, and cardiovascular diseases, among others.
The covalent addition of an AMP moiety to the side chain of exposed threonine and serine residues has emerged as a new paradigm to control the activity of the essential ER-resident chaperone BiP. This process, AMPylation, is catalyzed by metazoan AMP transferases (AMPylases) that contain a filamentation induced by c-AMP (fic) domain. Fic domaincontaining AMPylases (fic AMPylases) are highly conserved and are present in a single copy in most metazoans, including Caenorhabditis elegans (FIC-1), Drosophila melanogaster (dfic), Mus musculus (mFICD), and Homo sapiens (FICD) (1)(2)(3).
Metazoan fic AMPylases are bifunctional: using a single active site, these enzymes catalyze both the transfer of AMP to surface-exposed threonine and serine hydroxyl groups and the removal of AMP groups from modified residues (deAMPylation) (4)(5)(6)(7)(8). The switch between AMPylation and deAMPylation is proposed to involve enzyme dimerization, the exchange of Mg 2+ with Ca 2+ ions in the active site, and the subsequent switch from an open to a closed conformation (4)(5)(6)(7)(8)(9). The latter is stabilized by interactions between an inhibitory glutamate and a nearby arginine residue, which aligns an inhibitory α-helix such that the catalytic core preferentially binds AMP over ATP, catalyzing deAMPylation (6). If the interactions between these residues are prevented or resolved, fic AMPylases adopt an open conformation that favors Mg 2+ and ATP recruitment to the active site, enabling AMPylation of target proteins (10). Thus, replacing the critical inhibitory glutamate residue with a glycine (FICD(E234G)) converts the enzyme to a constitutively active AMPylase (10)(11)(12).
Changes in cellular AMPylation levels affect cellular fitness and organismal survival: Overexpression of constitutively active fic AMPylases is toxic and kills human (17,29,30) and yeast cells (20), as well as worm (C. elegans) embryos (19) and flies (D. melanogaster) (4). In contrast, fic AMPylase deficiency is well tolerated in unstressed human cells but impairs the activation of the unfolded protein response (UPR) under stress (17) and reduces neuronal differentiation (25). Further, Fic-1 deficient worms show enhanced sensitivity to the presence of aggregation-prone poly-glutamine repeat proteins in neurons (19). Perhaps the most significant in vivo fic AMPylase knockout (KO) phenotype is found in dfic-deficient flies, which show significant defects in visual signaling and suffer from light-induced blindness caused by BiP deregulation (5,31). Despite the emerging role of fic AMPylases in proteostasis, our understanding of how these enzymes affect mammalian physiology is lacking.
Here we describe the generation and characterization of an mFICD-deficient mouse strain. mFICD −/− mice are viable and are not visually impaired. We further show that mFICD deletion alters IgM synthesis and perturbs IL-1β secretion. Finally, we provide evidence that mFICD is involved in regulating nonspatial short-term memory in vivo. Together, our results support a modulatory role for mFICD function in adaptive immunity and neuronal plasticity in vertebrates.

mFICD-deficient mice are viable and fertile
To investigate the role of mFICD-mediated protein AMPylation in vivo, we attempted to generate both mFICDdeficient and constitutively active mFICD(E234G)expressing transgenic mouse strains using CRISPR/Cas9 technology. We used an sgRNA that targets a site adjacent to the coding sequence of the regulatory motif (TVAIEG) (Fig. S1A) and a double-stranded repair template to introduce the E234G substitution. The injection of approximately 80 blastocysts resulted in more than 20 independent animals carrying insertions or deletions in the mFICD gene that often resulted in frame shifts. Notably, not a single animal carrying the constitutively active mFICD(E234G) mutation was recovered. Parallel injections using identical experimental conditions but targeting different genes efficiently produced transgenic knock-in strains (32). These results suggest that embryonic expression of constitutively active mFIC-D(E234G), particularly in the absence of a wild-type mFICD copy, may be lethal.
For this study, we backcrossed a mouse strain carrying a deletion in mFICD. This deletion introduced a frame shift resulting in a premature stop codon (Fig. S1, B and C). To characterize the mFICD −/− animals, we assessed 6-month-old female control and mFICD −/− animals using the SHIPRA method. SHIPRA is a rapid, comprehensive screening approach, which provides a qualitative behavioral and functional profile for each animal (33). We found that mFICD −/− mice performed similarly to control animals for all 14 assessed features ( Fig. 1A; feature by feature results in Table S1). We found no significant differences in body weight (Fig. 1B), rotarod performance (Fig. 1C), and life span (Fig. S1D) between control and mFICD −/− animals. Together, these results establish that mFICD deficiency is well tolerated by mice under normal growth conditions. mFICD is required for AMPylation of BiP, EEF-1A, and HSC70 in mouse embryonic fibroblasts (MEFs) AMPylation in vertebrates is conferred by at least two evolutionarily unrelated enzymes: FICD and the mitochondrial pseudokinase SelO (34). To define how mFICD deficiency alters the vertebrate AMPylome, we supplemented mFICD −/− and control mouse embryonic fibroblast (MEF) lysates with 8-azido-ATP. Following AMPylation, a click reaction was used to install a PEG-biotin handle on the modified proteins. We then recovered AMPylated proteins with streptavidin-coated beads and identified AMPylated proteins by mass spectrometry ( Fig. 2A). Comparing results from treated MEFs, mFICD −/− KO MEFs and a cell-free control, we identified 108 proteins that were AMPylated only in wild-type MEFs (Table S2). Among these proteins were several known FICD targets, including BiP (HSPA5), HSC70 (HSPA8), and translation elongation factor EEF-1A ( Fig. 2A and S2A). In contrast, only two proteins (transketolase (TKT); protein disulfide isomerase (P4HB)) were identified in both wild-type and mFICD −/− lysates. Gene ontology analysis showed that AMPylated proteins were significantly enriched in metabolic, protein (re-)folding, and stress response processes (Fig. S2B). Together, these results confirm that FICD is required for the AMPylation-mediated regulation of BiP, HSC70, EEF-1A, and other proteins.

B and T cell development is unaffected by mFICD deletion
Having established that WT and mFICD −/− mice are phenotypically similar in overall physiological and morphological terms, and given mFICD's role in the UPR, an important pathway in B and T cell development (35), we asked whether their humoral immune system was affected by mFICD deficiency. Using flow cytometry, we examined the distribution of B and T cells in the spleen. We found no difference between WT and mFICD −/− mice in either the distribution of B and T cells (

Deletion of mFICD perturbs protein secretion in splenocytes
BiP, an essential molecular chaperone, discovered as an immunoglobulin-binding protein and regulated by FICD, is required for antibody assembly and maturation. We thus examined the impact of mFICD deficiency on protein secretion. Splenocytes were isolated from WT and mFICD −/− mice and stimulated with lipopolysaccharide (LPS), heparan sulfate (HS), or thapsigargin to induce cytokine secretion, which was then assayed by ELISA (Fig. 4, A-C). While mFICD deletion significantly reduced LPS-induced IL-6 secretion (Fig. 4A), it had no effect on secretion of TNFα (Fig. 4B). We also observed a significant difference in IL-1β production between WT and mFICD −/− mice (Fig. 4C). This is striking because unlike IL-6 and TNFα, which traffic through the ER, IL-1β folds in the cytoplasm, and is secreted by a nonclassical pathway. Immunoblots showed that intracellular levels of the IgM heavy chain (μ) were elevated in mFICD −/− B cells, while the levels of other proteins that fold and traffic through the ER were unchanged (Fig. S4A).
To further explore the consequences of mFICD deletion on protein folding and secretion, we focused on immunoglobulins because of their requirement for BiP activity in the course of folding and assembly (36). Naïve B cells were purified from the spleens, activated using LPS, and cultured for 3 days to allow differentiation into IgM-secreting plasmablasts. We then followed IgM folding by pulse-chase analysis. Briefly, plasmablasts were labeled with 35 S-methionine/cysteine for 15 min and chased in the absence of radioactive label for various times to follow protein maturation. IgM was immunoprecipitated from detergent lysates and media samples and analyzed by SDS-PAGE followed by autoradiography (Fig. 3D). mFICD −/− mice showed increased levels of both the soluble μ heavy chain (μ s ) and the membrane-bound B cell receptor (μ m ) in detergent lysates. This was accompanied by increased levels of both μ s and μ m mRNA (Fig. 4, E and F). Similar amounts of IgM were recovered from the media for both WT and mFICD −/− samples, suggesting that the increased levels of μ in the mFICD −/− cells did not pass ER quality control for secretion. These observations were consistent across the five pulse-chase experiments performed. We did not observe differences in the synthesis or glycosylation of other ER-folding proteins (Fig. S4B).

mFICD deletion does not impair induction of the unfolded protein response
AMPylation of BiP exerts fine control over the level of active BiP present in the ER (7,12,13,17). To examine how abrogation of such control affects the folding capacity and stress tolerance of cells, we examined the physiological UPR induced during B cell activation. As before, we isolated B cells from total splenocytes and stimulated them with LPS. Immunoblots showed no change in the levels of UPR sensors IRE1α or PERK, both before and after stimulation (Fig. 5A). There was also no change in the activation of either receptor as monitored by XBP1 splicing or eif2α phosphorylation or in the expression of downstream targets BiP and Grp94 (Fig. 5A). We further verified this by examining downstream targets of the UPR by qPCR (Fig. 5B). We observed no changes in the expression levels between WT and mFICD −/− samples for any of the genes that are downstream targets of the IRE1α, PERK, and ATF6α pathways.
The physiological UPR induced upon B cell activation is anticipatory of enhanced antibody production and differs from a UPR induced by an accumulation of misfolded proteins. To ascertain whether mFICD −/− cells responded differently to the latter type of UPR, we incubated B cells with several chemical initiators of the UPR. As with the physiological UPR from B cell activation, we found no significant difference in any of the UPR receptors or downstream targets assayed by immunoblot (Fig. 5C).
In D. melanogaster, the mFICD ortholog dfic regulates reversible photoreceptor degeneration, which is critical for visual neurotransmission and adaptation to constant light exposure (5,31). We thus tested whether mFICD −/− mice show signs of visual impairment. As part of the SHIRPA test (see Table S1), we assessed visual placing and the pupillary light reflex, which was normal in all tested 6-month-old control as well as mFICD −/− animals. To confirm these results, we performed optometry tests, in which mice were presented a rotating grating, a condition that elicits head movements in mice with intact eyesight (37). mFICD deficiency did not impact the animal's response to the moving grating early (6 months old) or late (18 months old) in life (Fig. 6, A and B). These results suggested that mFICD deficiency does not affect vision in mice.
mFICD deficiency does not impair learning, cognition, and memory FICD activity modulates neurogenesis and neuronal differentiation in human cerebral organoids (25). To determine if mFICD −/− animals may have defects in learning, cognition, and memory, we tested control and mFICD −/− animals at 6 and 18 months of age in Morris water maze tests (38). These tests are designed to assess impairments in visual short-term and long-term memory and visual-spatial abilities by observing and recording escape latency, distance moved, and velocity (38). There was no significant difference in visual nonspatial short-term learning comparing wild-type and mFICD −/− animals at 6 months of age (Fig. 7, A and B). However, 18-monthold mFICD −/− mice showed a significantly decreased escape latency between trial 1 and 2 while the performance of wildtype mice did not (Fig. 7, C and D). This suggests that visual nonspatial short-term learning is stronger in old mFICD −/− mice than in wild-type controls. Next, we assessed learning and memory capacity of mFICD −/− and wild-type mice at 6 and 18 months of age using a submerged platform setup. Both control and mFICD −/− mice performed equally well (Figs. 7E and S5A) phase. Upon removal of the platform, wild-type and mFICD −/− showed similar recall behaviors, spending the most time in the maze quadrant that contained the platform in the preceding learning phase (Fig. 7, F and G). Swimming speed was similar between control and mFICD −/− cohorts (Fig. S5B), suggesting that mFICD deficiency does not affect agingdependent decline in rough muscle function. Together, these results indicate that cerebellar function, learning, and memory are not affected by mFICD loss. Finally, to test for cognitive flexibility, we reintroduced the platform to a new position (quadrant) in the water maze. We found that both control and mFICD −/− mice at 6 and 18 months of age adapted to the new situation and memorized the new position of the platform equally well (Fig. 7H).

Discussion
Protein AMPylation in metazoans is increasingly recognized as a PTM that regulates the function of the ER-resident chaperone BiP as well as that of other proteins (15,21). While the consequences of hyper-and hypo-AMPylation are well established in tissue culture systems and invertebrate models, the impact of fic AMPylase deficiency on mammalian physiology is unknown. Here we describe a functional mFICD knockout mouse. We have examined the physiological, neurological, and molecular consequences of mFICD deletion and identified points where WT and mFICD −/− mice differ. Overall, our phenotypic and pathologic assessments of mFICD −/− mice did not identify major debilitating abnormalities or dysfunctions. These findings are in accordance with previous work in human tissue culture and C. elegans in vivo models for FIC-1 deficiency, both of which showed that the absence of FICD/FIC-1 is well tolerated in unstressed cells and animals (5,7,11,12,17). In contrast with work done in dficdeficient flies, which suffer from deficits in visual perception and light-induced blindness (5, 31), we did not observe differences or defects in vision in mFICD −/− mice. Possible explanations for these divergent observations include the presence of compensatory/alternative mechanisms to regulate proteins AMPylated by mFICD −/− in mice and/or differences in target protein profile between dfic and mFICD. Further work is required to decipher the role of mFICD in visual perception in molecular detail.
On a molecular level, mFICD deficiency depletes unstressed cells of most AMPylated proteins. In accordance with previous studies, we find evidence that mFICD AMPylates both ERresident and cytoplasmic proteins (11, 19, 20, 25-29, 31, 39). The mechanisms that underlie mFICD's ability to target topologically distinct compartments remain to be defined. AMPylation of BiP by mFICD regulates the levels of active BiP to allow a fine-tuned response to ER stress (7). We show that in B cells, abrogation of mFICD activity resulted in higher levels of μ synthesis in the ER but secretion of IgM is similar for WT and mFICD −/− B cells. BiP binds to the CH1 domain of immunoglobulins to prevent aggregation and promote folding, only releasing the heavy chain molecule once a correctly folded light chain is available for association. Increased levels of  intracellular μ could be due to a deficit in light chain levels or possibly due to strengthened interaction between the CH1 domain and BiP due to increased BiP activity.
We further discovered that deletion of mFICD compromised secretion of IL-1β. Pro-IL-1β folds in the cytoplasm where it is cleaved by caspase-1 to produce the active cytokine, which is then released from the cell via nonclassical secretion through gasdermin-D pores (40). IL-1β can enter both the type I and type III nonconventional protein secretion pathways (41). It is therefore possible that AMPylation plays a role in the regulation of nonclassical secretion. Of note, the bacterial Hsp70 (DnaK) has also been implicated in nonclassical secretion (42). mFICD −/− mice do not show signs of cognitive impairment; rather, they had a tendency to perform somewhat better in most assays than wild-type control animals, a trend we also observed in motor assays. In particular visual nonspatial shortterm learning was improved in old mFICD −/− mice as compared with wild-type controls (Fig. 7, C and D). We attribute this to the presumably continuous derepression of BiP and the dysregulation of other AMPylated proteins such as HSC70 and EEF-1A in the absence of mFICD. Future work on mFICD-dependent and independent protein AMPylation in the presence of aggregation-prone proteins or in response to stressors that cause protein unfolding should define the role of this PTM in neuronal health and aging.
In conclusion, our work shows that mFICD deficiency is tolerated in the absence of stress but can impair BiPdependent (antibody folding and maturation) and -independent (IL-1β secretion) processes. Knowing that FICD is involved in the regulation of cellular stress pathways, it appears likely that the characterization of critical in vivo roles of mFICD will require specific cellular and/or environmental stresses, as yet to be identified. Future studies focusing on the effects of protein AMPylation in chronic or acute stress and disease models should thus provide additional insights into the role of mFICD in proteostasis and protein folding.

Experimental Procedures
Generation of FICD −/− mouse All animal procedures were performed according to NIH guidelines and approved by the Committee on Animal Care at MIT and the Institutional Animal Care and Use Committee (IACUC) at Boston Children's Hospital. CRISPR/Cas9mediated genome editing was performed exactly as described in Maruyama et al. (32). We targeted a genomic location using an sgRNA (ccacacggtggccatcgaggg) close to the sequence encoding for the regulatory FICD motif (TVAIEG), which resulted in multiple transgenic animal containing 1-300 bp long deletion or insertions. Mice were out-crossed six times with C57BL/6J animals to eliminate putative off-target mutation introduced during CAS9-based genome editing.

Identification of AMPylated proteins
Mass-spectrometry-based identification of AMPylated proteins was performed using a chemical reporter setup as previously described (43,44). Briefly, we supplemented total lysates of mFICD −/− and wild-type MEFs with N6-propargyl-ATP and incubated the lysates at room temperature for 1 h. This allowed endogenous mFICD to utilize N 6 -propargyl-ATP as nucleotide substrate. We then supplemented the reaction with biotin-(PEG) 3 -azide to covalently couple a biotin handle to the AMP-propargyl groups now found on AMPylated proteins. AMPylated (biotinylated) proteins were recovered using Streptavidin-modified agarose beads, eluted for LC/MS/MS analysis.

Mass spectrometry
Eluates from Streptavidin-modified agarose beads were reduced, alkylated, and digested with trypsin at 37 C overnight. The resulting peptides were extracted, concentrated, and injected onto a Waters NanoAcquity HPLC equipped with a self-packed Aeris 3 μ3 column. Peptides were eluted using customized reverse-phase gradients and analyzed using an Orbitrap Elite mass spectrometer (Thermo Fisher) in nanospray configuration, operated in a data-dependent manner. The resulting fragmentation spectra were matched against custom databases with Mascot (Matrix Science) 2.5.1 and PEAKS (Bioinformatics Solutions) 7.5. A detailed description of the performed mass searches is provided in Table S2.

Behavioral test
We examined ten female C57BL/6J (wild type) and ten female mFICD −/− mice setup in cages of four, each containing two control and two mFICD −/− animals. Each mouse received a unique paw tattoo to enable ID tracking throughout the experiment. All project members involved in behavioral testing of animals were blinded and unaware of cage compositions.

Morris water maze tests
Mice were placed in a circular pool (137-cm diameter) filled with water maintained at room temperature. The tank was divided into quadrants each marked on the wall of the tank with a different visual cue for spatial orientation. Swim time and path length were recorded by automated video tracking (Ethovision XT 11.5, Noldus Information Technology, The Netherlands). A black curtain surrounded the tank to prevent cues in the room aiding the spatial performance. On day 1 (visible platform), mice were given 90 s to swim to a platform elevated 1 cm above the water and marked with a flag. The position of the platform was kept constant, but the starting quadrant varied. Four starts (each from one of the four quadrants) were administered in one trial, and two visual trials were administered during day 1. On days 2 and 3 (learning), the platform was moved to the opposite quadrant from the visible position, submerged 1 cm below the water line, and the flag was removed. Mice were given 90 s to find the hidden platform and left on the platform for 5 s to orientate themselves. Mice that did not find the platform within 90 s were guided to it by the experimenter. Each mouse completed three learning trials on day 2 (Trials 1-3), and two learning trials on day 3 (Trials 4 and 5), with each trial comprising four different start positions. On day 4 (recall), the platform was removed entirely from the tank, the mice placed in the quadrant opposite the quadrant that had previously contained the platform (e.g., north east, NE) and given 60 s to swim. Time spent and path length traversed in each of the four quadrants were recorded. On day 5 (recall), the platform was placed back into the tank 1 cm below the water in a quadrant different from the visible and learning trials and mice were tested across three trials (four starts of 90 s each from different locations).

Rotarod tests
On day 1, mice were placed on a rotarod (Economex, Columbus Instruments, USA) set to revolve at 4 rpm for 5 continuous minutes. This training phase allows mice to acclimate to the movement of the rotarod. The following day, mice are placed on the rotarod revolving at 4 rpm for a 10 s acclimation before accelerating at 0.1 rpm/s. These trials were repeated four times with a break of at least 5 min between repeats. Latency to fall was recorded as readout.

Optometer tests
We used a protocol adapted from Prusky et al. (37), which employs an optomotor device (CerebralMechanics, Canada) to measure visual acuity. The device consisted of four computer monitors arranged in a square with a lid on top to enclose the mouse within. A computer program was used to project on the monitors a virtual cylinder in 3-D coordinate space. Visual stimuli were drawn on the walls of the cylinder, and from the perspective of the platform, each monitor appeared as a window on a surrounding 3-D world. The software also controlled the speed of rotation and geometry of the cylinder and the spatial frequency and contrast of the stimuli. A red crosshair in the video frame indicated the center of the cylinder rotation. Mice were placed one at a time on the platform inside the device, the lid of the box was closed, and the animals were allowed to move freely, and as the mouse moved about the platform, the experimenter followed the mouse's head with the red crosshair. When a grating perceptible to the mouse was projected on the cylinder wall and the cylinder was rotated (12 deg/s), the mouse normally stopped moving its body and would begin to track the grating with reflexive head movements in concert with the rotation. An experimenter assessed whether the animals tracked the cylinder by monitoring in the video window the image of the cylinder, the animal, and the crosshair simultaneously. If the mouse's head tracked the cylinder rotation, it was judged that the animal could see the grating. Using a staircase procedure, the mouse was tested systematically against increasing spatial frequencies of the grating until the animal no longer responded. The threshold was then calculated as the highest spatial frequency that the mouse responded to.

Pulse chase labeling and immunoprecipitation
Suspension pulse chase was performed as previously described (46). Briefly, LPS-stimulated plasmablasts were starved for methionine and cysteine for 15 min and pulse labeled for 5 min with 55 μCi/1 × 10 6 cells of Express 35 S protein labeling mix (PerkinElmer). Samples were then diluted 10× with cell culture media containing 5 mM unlabeled methionine and cysteine. At the end of each chase time, a sample was removed and diluted 1:1 into ice-cold PBS containing 20 mM N-ethyl maleimide (Sigma). Cells were collected by centrifugation at 1250g for 5 min and washed twice with ice-cold PBS. Cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (Roche). IgM was immunoprecipitated from lysates or media samples with a goat anti-mouse IgM antibody (Southern Biotech). Immunoprecipitates were washed with lysis buffer and eluted from beads with SDS-PAGE sample buffer containing 20 mM DTT and analyzed by SDS-PAGE. Gels were then dried and exposed to BioMax MR films (Carestream).
RNA extraction, cDNA synthesis, and qPCR LPS-activated plasmablasts on each day of activation were lysed using TriZol (Invitrogen) and RNA was isolated using RNA extraction kit (Zymo). RNA was checked for integrity and equivalent amounts of RNA were converted into cDNA using the Maxima H minus cDNA Synthesis master mix (Invitrogen). Samples for qPCR were prepared with primers against target genes and mixed with SSO Advanced Universal SYBR Green Supermix (BioRad) in 96-well plates and analyzed on a CFX96 qPCR machine (BioRad). Analysis was performed using the standard Δct method using actin as a reference gene. Statistical analysis was carried out on non-log-transformed values.

Statistical analysis
Data is graphed as means ± SD; where feasible, individual data points are shown. Depending on the experiment, we used unpaired/paired two-tailed t tests, Turkey's multiple comparison's test, two-way ANOVA with multiple comparison test as well as Kaplan-Meier tests to evaluate survival data. Individual p values are displayed in each figure.
Acknowledgments-We like to thank the Truttmann, Ploegh, and Hu laboratories for feedback on the manuscript. The Animal Behavior and Physiology Core at Boston Children's Hospital is acknowledged for performing the behavioral tests (IDDRC 1U54HD090255). Funding and additional information-This work was supported by grants NIH/NCI R01CA163910 to C.-C. A. H. and support from AARG and MICHR UL1TR002240 to M. C. T. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.