ABHD12 controls brain lysophosphatidylserine pathways that are deregulated in a murine model of the neurodegenerative disease PHARC

Jacqueline L. Blankmana,b, Jonathan Z. Longa,b, Sunia A. Traugerc, Gary Siuzdakc, and Benjamin F. Cravatta,b,1

aThe Skaggs Institute for Chemical Biology and Departments of bChemical Physiology and cMolecular Biology, The Scripps Research Institute, La Jolla, CA 92037

Edited by David W. Russell, University of Texas Southwestern Medical Center, Dallas, TX, and approved November 30, 2012 (received for review October 1, 2012)

Advances in human genetics are leading to the discovery of new could contribute to the of the endogenous canna- disease-causing at a remarkable rate. Many such muta- binoid 2-arachidonoylglycerol (2-AG) in the nervous system. tions, however, occur in that encode for proteins of unknown Nonetheless, the physiological metabolites regulated by ABHD12 function, which limits our molecular understanding of, and ability to in vivo, and the molecular and cellular mechanisms by which this contributes to PHARC, are unknown. Here, we have devise treatments for, human disease. Here, we use untargeted −/− metabolomics combined with a genetic mouse model to determine addressed these important questions by generating ABHD12 α β mice and analyzing these animals for metabolomic and behavioral that the poorly characterized / -hydrolase do- −/− < main-containing (ABHD)12, mutations in which cause the human . Young ABHD12 mice ( 6 mo old) were mostly normal in their behavior; however, as these animals age, they neurodegenerative disorder PHARC (polyneuropathy, hearing loss, develop an array of PHARC-related phenotypes, including de- ataxia, retinosis pigmentosa, and ), is a principal lysophos- −/− fective auditory and motor behavior, with concomitant cellular phatidylserine (LPS) in the mammalian brain. ABHD12 mice indicative of a neuroinflammatory response. Metab- − − display massive increases in a rare set of very long chain LPS lipids olomic characterization of brain tissue from ABHD12 / mice that have been previously reported as Toll-like 2 activators. revealed striking elevations in a series of lysophosphatidylserine We confirm that recombinant ABHD12 protein exhibits robust LPS fl NEUROSCIENCE − − (LPS) lipids that occur before the onset of neuroin ammatory lipase activity, which is also substantially reduced in ABHD12 / fi − − and behavioral defects. We con rmed that LPS lipids are direct brain tissue. Notably, elevations in brain LPS lipids in ABHD12 / substrates for ABHD12 and that this enzyme is a major contri- mice occur early in life (2–6 mo) and are followed by age-dependent butor to bulk brain LPS lipase activity. These data, taken to- − − increases in microglial activation and auditory and motor defects gether, demonstrate that ABHD12 / mice are a suitable animal that resemble the behavioral phenotypes of human PHARC patients. model for investigating the mechanistic basis of PHARC and Taken together, our data provide a molecular model for PHARC, point to a functional role for an ABHD12-regulated LPS pathway where disruption of ABHD12 causes deregulated LPS metabolism in this disease. and the accumulation of proinflammatory lipids that promote microglial and neurobehavioral abnormalities. Results Targeted Disruption of the Abhd12 . To generate mice lacking lipidomics | neuroinflammation ABHD12, a targeting construct was designed to delete exons 8–10 of the Abhd12 gene, including the predicted catalytic serine S246 contained within a canonical GXSXG motif, upon homologous enome mapping and sequencing have facilitated determi- recombination in C57BL/6-derived embryonic stem (ES) cells Gnation of the genetic basis for over 3,500 inherited diseases (Fig. 1A). One targeted ES clone, 96, was identified by Southern in humans (1), with additional pathogenic mutations still being blotting (Fig. 1B) and used to generate chimeric mice. These discovered. For many Mendelian disorders, however, the mo- chimeras gave germ-line transmission of the targeted , as lecular and cellular mechanisms that link genotype to disease determined by Southern blotting (Fig. 1C) and PCR genotyping − − remain poorly understood. This problem is perhaps (Fig. 1D). Mice homozygous for the targeted allele (ABHD12 / most apparent for diseases where the causative mutations occur mice) were viable and born at the expected Mendelian frequency. in genes that encode for proteins with unannotated or poorly Loss of ABHD12 mRNA (Fig. 1E), protein (Fig. 1F), and activity − − characterized functions. Here, we focus on one such disease - the (Fig. 1G) were confirmed in the brains of ABHD12 / mice by neurodegenerative disorder PHARC (polyneuropathy, hearing RT-PCR, liquid chromatography–mass spectrometry (LC-MS) loss, ataxia, retinosis pigmentosa, and cataract). proteomics, and activity-based protein profiling (ABPP) (6), re- PHARC [Mendelian Inheritance in Man (MIM) no. 612674; spectively. Levels of other protein constituents of the endocanna- Online Mendelian Inheritance in Man database, http://omim.org] binoid system were unaffected by ABHD12 disruption (Fig. 1F). is a rare, autosomal recessive disorder that causes polymodal − − sensory and motor defects associated with demyelination of sen- ABHD12 / Mice Develop an Age-Dependent, PHARC-Related Phenotype. − − somotor neurons, retinal dystrophy, and cerebellar atrophy (2). ABHD12 / mice were fertile and largely indistinguishable from + + + − PHARC symptoms are slowly progressive and begin in the child- their WT (ABHD12 / ) and heterozygous (ABHD12 / ) litter- + + + − hood or teenage years; heterozygous carriers are unaffected (3). In mates at early stages of life. ABHD12 / , ABHD12 / , and 2010, PHARC was reported to be caused by homozygous muta- tions in ABHD12, which codes for the poorly characterized enzyme α/β-hydrolase domain-containing (ABHD)12 Author contributions: J.L.B. and B.F.C. designed research; J.L.B. and S.A.T. performed (3). To date, five distinct ABHD12 mutations have been identified research; J.L.B. and J.Z.L. contributed new reagents/analytic tools; J.L.B., G.S., and B.F.C. in patients with PHARC, all of which are expected to lead to analyzed data; and J.L.B. and B.F.C. wrote the paper. complete loss of ABHD12 expression (3, 4). PHARC, therefore, The authors declare no conflict of interest. likely represents a human ABHD12 null model. This article is a PNAS Direct Submission. We demonstrated previously that ABHD12 is a membrane- 1To whom correspondence should be addressed. E-mail: [email protected]. bound enzyme that exhibits monoacylglycerol (MAG) lipase This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. activity in vitro (5). These initial data suggested that ABHD12 1073/pnas.1217121110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1217121110 PNAS Early Edition | 1of6 Downloaded by guest on September 26, 2021 Fig. 1. Generation and initial characterization of − − ABHD12 / mice. (A) ABHD12-targeting vector designed to replace Abhd12 exons 8–10 with a neomycin (Neo) selection cassette upon homolo- gous recombination. (B) Targeted clone no. 96 was identified by Southern analysis of EcoRI-digested ES cell genomic DNA with external 59 and 39 probes. Germ-line transmission was confirmed by Southern (C) and PCR (D) genotyping of genomic tail DNA. (E) RT-PCR confirmed loss of ABHD12 mRNA in ABHD12−/− mice. (F) Untargeted MS-based proteo- mics confirmed absence of ABHD12 protein in − − ABHD12 / brain tissue. Protein levels of other measured constituents of the were equivalent between genotypes. Values represent the mean ± SEM (n = 3). (G)Lackof ABHD12 activity in ABHD12−/− mice was confirmed by ABPP of brain membrane proteomes with the serine hydrolase probe fluorophosphonate (FP)- rhodamine. Other serine hydrolase activities were − − unaltered in ABHD12 / brain proteome. Fluores- cent gel shown in grayscale.

− − ABHD12 / mice were assessed in a battery of behavioral tests auditory information, auditory brainstem responses (ABRs) were potentially relevant to the symptoms of PHARC disease. Initial measured in a cohort of 9- to 10-mo-old mice after confirming − − behavioral studies were performed with relatively young mice (5– that ABHD12 / mice exhibit a reduced auditory startle response 6 mo old), but considering that neurodegenerative phenotypes at this age (Fig. S1). The sinusoidal waveform of the brainstem − − often only become apparent in aged mice, even for disease auditory evoked potential (BAEP) was intact in ABHD12 / models with childhood onset in humans (7–10), this panel of tests mice (Fig. S1), and no genotypic differences were detected in was repeated when mice were 11–12 and 17–18 mo old. peak amplitude (Table S1). However, absolute latencies of waves − − Young ABHD12 / mice behaved similarly to control litter- IV, V, and VII (Table S1) and interpeak latencies of waves I–IV mates except for modest hypermotility, which was transitory and were delayed in ABHD12-disrupted mice (Fig. 2D), confirming not observed during the later testing sessions (Fig. S1). A slight impaired auditory signaling in these animals and suggesting that reduction in the auditory startle reflex was also observed in 5- to the hearing loss induced by ABHD12 disruption may be caused − − 6-mo-old ABHD12 / mice, although this effect failed to reach by reduced auditory conductance. − − statistical significance (Fig. 2A). However, when retested at 11–12 Seventeen- to 18-month-old ABHD12 / mice exhibited addi- − − mo, ABHD12 / mice exhibited a significantly blunted startle tional deficits in the accelerating rotarod (Fig. 2E) and hanging wire − − response (Fig. 2B). At 17–18 mo old, ABHD12 / mice failed to (Fig. 2F) tests that were not observed at younger ages. The pro- + + + − − − startle above baseline, whereas ABHD12 / and ABHD12 / pensity of ABHD12 / micetofalloffoftherotarodatslower mice maintained a measurable startle response (Fig. 2C). De- speeds than control mice is indicative of mild motor incoordination, − − creased auditory startle response in ABHD12 / mice could be mirroring the ataxia observed in human PHARC patients (3). caused by hearing loss [a known PHARC symptom (3)] in these Similarly, decreased latency to fall in the hanging wire test sug- − − − − animals. To investigate the ability of ABHD12 / mice to process gests muscle weakness in aged ABHD12 / mice, consistent with

Fig. 2. ABHD12−/− mice develop age-dependent PHARC-related behaviors. (A–C) Eleven- to 12-month- old (B) and 17- to 18-mo-old (C), but not 5- to 6-mo- − − old (A), ABHD12 / mice (black triangles) showed significantly reduced auditory startle reflex com- + − + + pared with ABHD12 / (gray squares) and ABHD12 / (white diamonds) littermate controls (n = 8–19 per group). (D) The auditory brainstem response (ABR) − − was impaired in 9- to 10-mo-old ABHD12 / mice (black bars), which showed delayed interpeak latency + − of waves I–IV. ABHD12 / mice (gray bars) also showed a modest defect in ABR compared with ABHD12+/+ mice (white bars) that was only observed at the lowest auditory stimulus (n = 5 per genotype). − − (E and F) Seventeen- to 18-mo-old ABHD12 / mice displayed mild ataxia (E) and a reduced latency to fall (F) in the rotarod and hanging wire test, respectively (n = 16–19 per genotype). Data represent the averages ± SEM. Variance among genotypes was assessed by two-way ANOVA and Fisher’s protected least-significant difference (PLSD) test. The Student’s t test was used to determine significance for each individual dB + + pulse tested in A–C.*P < 0.05; **P < 0.01; ***P < 0.001 vs. ABHD12 / mice.

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1217121110 Blankman et al. Downloaded by guest on September 26, 2021 Identification of Brain Metabolites Deregulated by ABHD12 Disruption. − − Characterization of ABHD12 / mice confirmed a minor defect in brain 2-AG metabolism. Protein homogenates prepared from + + − − ABHD12 / and ABHD12 / brains hydrolyzed 2-AG to similar − − extents, but ABHD12 / homogenates showed a significantly greater reduction in 2-AG hydrolysis following inhibition of the principal brain 2-AG hydrolase, MAG lipase (MAGL) (Fig. S3). Bulk levels of 2-AG and other MAGs, however, were unchanged − − − − in ABHD12 / brains (Fig. S3). Additionally, ABHD12 / mice did not exhibit any of the classical signs of enhanced signaling through the central receptor CB1: analgesia, hypo- motility, or hypothermia (Fig. S3). These results, taken together, suggested that the behavioral and cellular pathology elicited by ABHD12 disruption may be attributable to deregulation of nonendocannabinoid metabolites. To more fully explore this + + possibility, we performed a metabolomic analysis of ABHD12 / − − and ABHD12 / brains by untargeted LC-MS (11, 12). LC-MS profiling of organic-soluble brain fractions from 2- to + + − − 6-mo-old ABHD12 / and ABHD12 / mice combined with data analysis using the XCMS program (13) identified several − − metabolites that were significantly altered in ABHD12 / brains (Fig. 4A and Table S2). The largest fold-change differences were observed for metabolites with m/z values of 552.3, 578.3, 580.4, Fig. 3. ABHD12 disruption causes age-dependent increases in microglial and 608.4, all of which were detected in the negative-ion mode − − activation in the mouse brain. (A–C) Representative images of Iba1 immu- and were dramatically (∼6- to 40-fold) elevated in ABHD12 / nostaining for in the of 18-mo-old (A), 12-mo-old (B), A m z +/+ −/− – brains (Fig. 4 and Table S2). Based on the differences in / and 6-mo-old (C) ABHD12 and ABHD12 mice. (D F) Quantifying the values and LC retention times of these metabolites, we hypoth- number of enlarged cells (>200 μm2) revealed significantly more activated microglia in cerebellar sections from 18-mo-old (D) and 12-mo-old (E) esized that they might represent related members of a common − − + + ABHD12 / mice compared with ABHD12 / mice. No genotype differences lipid class differing in acyl chain length and degree of unsatu- NEUROSCIENCE were observed at 6 mo of age (F). Data represent the averages ± SEM. ***P < ration. Targeting these metabolites by Fourier transform ion 0.001 (n = 3–4 per genotype per age group). (Scale bar, 100 μm.) cyclotron resonance (FT-ICR)-MS allowed mass determination within 2 ppm accuracy and calculation of their molecular for- mulas, which matched those of very long chain lysophosphati- the peripheral neuropathy experienced by PHARC patients (3). dylserine (VLC-LPS) lipids (Table S3). − − On the other hand, ABHD12 / mice appeared to possess normal Manual inspection of the LC-MS data revealed that an addi- ocular structure and function, because we did not detect signifi- tional metabolite with an m/z value of 524.3, corresponding to −/− cant alterations in visual performance in the optomotor tracking 18:0 LPS, was also elevated in ABHD12 brain metabolomes test (Fig. S1) or lens opacity in these animals. Taken together, (Fig. 4B). Supporting this molecular assignment, the endogenous − − these results showed that ABHD12 / mice develop several, but m/z 524.3 metabolite and a synthetic 18:0 LPS standard displayed not all, of the main features of PHARC disease, including hearing the same LC elution time (Fig. 4C) and fragmentation spectra loss, ataxia, and symptoms of polyneuropathy, indicating that (Fig. 4D). Fragmentation spectra for the m/z 578.3, 580.4, and these animals are a suitable model to study the endogenous 608.4 metabolites, which, like the spectrum for 18:0 LPS (Fig. function of ABHD12 and the molecular basis for PHARC disease. 4D), all possessed a common peak at m/z = 153, corresponding to the glycerophosphate–H2O moiety, and a peak at the parent Enhanced Microglial Activation in Brain Tissue from ABHD12−/− Mice. m/z minus 87, corresponding to the neutral loss of serine (Fig. + + We next studied brains from old (18–21 mo) ABHD12 / and S4), permitted their structural assignment as 22:1, 22:0, and 24:0 − − ABHD12 / mice by histological analysis to investigate the cel- LPS species, respectively. lular basis for the late-onset neurobehavioral abnormalities ob- In addition to the striking accumulation of VLC-LPS species − − − − served in ABHD12 / mice. Hematoxylin and eosin (H&E) identified in ABHD12 / brains, more modest alterations were staining, Luxol Fast Blue (LFB) staining, Nissl staining, calbindin also found in other brain metabolites by lipidomic analysis (Fig. D28k immunohistochemistry, and glial fibrillary acid protein 4A and Table S2), including select phosphatidylserine (PS) (Fig. (GFAP) immunohistochemistry revealed no obvious differences S4), phosphatidylglycerol (PG) (Fig. S4), and lysophosphatidy- in gross brain organization, , neurons, Purkinje neurons, linositol (LPI) lipids (Table S2). In contrast, all other (lyso) + + − − and astrocytes, respectively, between ABHD12 / and ABHD12 / phospholipid and free levels examined were equivalent between genotypes (Table S2). Finally, we detected two brain brains (Fig. S2). On the other hand, immunostaining for ionized − − calcium-binding adaptor molecule (Iba)1, which is specifically metabolites elevated in ABHD12 / mice that remain structur- expressed in microglia and , revealed a dramatic ally unidentified (m/z values of 718.6 and 814.6; Table S2). genotypic difference in the morphology of microglia in the cer- We used targeted multiple reaction monitoring (MRM) MS ebellum (Fig. 3 A–C). Quantitation of Iba1 immunoreactivity methods to confirm our untargeted lipidomic findings, including identified significantly more activated microglia (i.e., Iba1-re- the dramatic elevations in VLC-LPS lipids caused by ABHD12 active cells with enlarged soma size and retracted processes) in disruption (Fig. 4E and Tables S4 and S5). Targeted experi- − − sections from 18-mo-old ABHD12 / mice (Fig. 3D). A signifi- ments, on the other hand, failed to detect VLC-LPI or VLC- + + − − cant increase in activated microglia was also observed in cere- LPG species in ABHD12 / or ABHD12 / brains. Our targeted − − bellar sections from 12-mo-old ABHD12 / mice (Fig. 3 B and E), analyses also confirmed changes in other lipid metabolites in − − a time point at which these animals displayed normal rotarod ABHD12 / brains (Table S4), including PS lipids, which dis- behavior (Fig. 2E). In contrast, WT microglial morphology was played a notably remodeled acyl chain distribution with selective − − found in cerebellar sections from 6-mo-old ABHD12 / mice enrichment of arachidonate-containing species (Fig. 4F). An al- (Fig. 3 C and F). These data indicate that disruption of ABHD12 tered LPS and PS lipid profile was also found in brain regions − − promotes an age-dependent neuroinflammatory response in the (cerebellum, cortex, and hippocampus) from ABHD12 / mice, brain that temporally precedes the onset of cerebellar-mediated with the greatest absolute levels and accumulation of VLC-LPS behavioral abnormalities. detected in the cerebellum of these animals (Table S6). In contrast,

Blankman et al. PNAS Early Edition | 3of6 Downloaded by guest on September 26, 2021 Fig. 4. Deregulated (lyso)phosphatidylserine [(L)PS] − − metabolites in brain tissue from ABHD12 / mice. (A) Untargeted metabolomics identified several metabo- lites with altered brain levels in 2- to 6-mo-old − − + + ABHD12 / mice compared with ABHD12 / mice. The m/z values with the largest fold-change differences between genotypes are marked. (B) Relative peak intensity of a metabolite with an m/z value of 524.3, corresponding to 18:0 lysophosphatidylserine (LPS), in − − + + ABHD12 / vs. ABHD12 / brain tissue. (C and D)The m/z 524.3 metabolite (black trace) exhibited the same LC elution time (C) and MS/MS fragmentation pattern (D) as a synthetic 18:0 LPS standard (red trace). Key daughter ions in the MS/MS analysis include 153.0 (dehydro-glycerophosphate) and 437.3 [18:0 lyso- phosphatidic acid (LPA)]. (E and F) Targeted MRM measurements of LPS (E)andPS(F) species from brain + + + − tissue of ABHD12 / (white bars), ABHD12 / (gray bars), or ABHD12−/− (black bars) mice. Data are pre- sented as average values ± SEM (n = 3–5 per group). + + *P < 0.05; **P < 0.01; ***P < 0.001 vs. ABHD12 / (Student’s t test).

− − LPS and PS levels were largely unchanged in several ABHD12 / LPG, and BMP (bismonoacylglycerolphosphate) hydrolase ac- peripheral tissues, with less than twofold differences observed tivities (Fig. 5B), although we observed only modest alterations among genotypes (Table S6). (

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1217121110 Blankman et al. Downloaded by guest on September 26, 2021 by microglia (25) and can induce microglial-mediated neuronal death in primary neural cultures (26, 27) and in vivo (28). We hypothesize the VLC-LPS species that accumulate following ABHD12 disruption could serve as endogenous TLR2 ligands, leading to microglial activation and neuroinflammation in the brain. In summary, our findings provide a working model to explain the molecular basis for PHARC, wherein ABHD12 disruption produces, in a temporally ordered manner, deregulated (L)PS metabolism that includes elevations in proinflammatory VLC-LPS lipids, followed by microglial activation and the onset of neuro- inflammation and, finally, an array of neurobehavioral abnor- − − malities (Fig. 6). We should, however, clarify that ABHD12 / mice do not appear to exhibit all features of human PHARC. We did not, for instance, observe significant optomotor defects or − − cerebellar atrophy in ABHD12 / mice. Of course, difficulties in fully recapitulating human neurodegenerative diseases with genetic mouse models is a well-recognized challenge (29, 30), and we be- − − lieve, overall, that ABHD12 / mice display many of the key fea- tures of PHARC. Perhaps most importantly, these animals provide a valuable tool to decipher the molecular and cellular basis for PHARC. In this regard, although the major metabolic changes Fig. 5. ABHD12 is a major brain LPS lipase. (A) Membrane homogenates caused by ABHD12 ablation corresponded to (L)PS lipids, we from HEK293T cells transfected with a murine ABHD12 cDNA (gray bars) cannot rule out a potential role for (L)PGs, (L)PIs, or other displayed increased hydrolytic activity compared with mock-transfected cell ABHD12-regulated metabolites in the development of PHARC. membranes (black bars) toward several lysophospholipids and other singly Additionally, although our results do not support a role for acylated glycerolipids (e.g., MAG, BMP). In contrast, doubly acylated glyc- ABHD12 as a regulator of bulk endocannabinoid signaling in the erolipids (e.g., DAG, PA, PC, PE, PI, PG, PS) were not hydrolyzed by ABHD12. mammalian brain, it remains plausible that this enzyme could See Table S7 for more information on lipid species tested as ABHD12 sub- control local 2-AG accumulation in specificproinflammatory brain NEUROSCIENCE strates. (Inset) Active ABHD12 expression in transfected cells was confirmed cells, such as microglia. Cannabinoid type 2 (CB2) receptors, in −/− by ABPP with the FP-rhodamine probe. (B) ABHD12 brain membrane particular, are up-regulated in activated microglia associated with homogenates (black bars) displayed significantly reduced hydrolytic activity ’ +/+ Alzheimer s and multiple sclerosis plaques (31, 32) and are thought toward 18:1 LPS, LPG, LPI, and BMP compared with ABHD12 brain ho- to regulate microglial motility (33). Microglial/glial CB2 activation, mogenates (white bars). Data are presented as average values ± SEM (n = 3 + + per group). *P < 0.05; **P < 0.01; ***P < 0.001 vs. mock or ABHD12 / however, is generally considered neuroprotective, and it has been homogenates (Student’s t test). associated with a reduction in neurotoxic factor release (34). De- termining whether ABHD12 regulates 2-AG in primary microglia and how the loss of ABHD12 affects CB2-mediated microglial elevations in LPS lipids, including massive (>10-fold) increases activity are fertile topics for future research. Finally, we speculate in several VLC-LPS lipids. We also detected substantial changes that, if elevated LPS lipids prove to be a causative agent in in PS lipids that appear to reflect remodeled acyl chain distri- bution. We suspect that this latter metabolic phenotype may be a secondary consequence of ABHD12-mediated changes in LPS lipids, given that ABHD12 shows robust LPS lipase, but not PS lipase activity. Importantly, the alterations in (L)PS lipid levels − − occurred in young ABHD12 / mice (2–6 mo old) before the onset of microglial activation or neurobehavioral abnormalities. These data thus support a model where deregulated (L)PS me- tabolism is a cause, rather than a consequence, of PHARC-like − − syndrome in ABHD12 / mice (see below). − − How might accumulation of LPS in the brains of ABHD12 / mice contribute to neuroinflammation and PHARC-like behav- ioral defects? LPS has been shown to be a bioactive lipid that produces several cellular effects (14), including mast cell de- granulation (15–17), clearance of apoptotic cells (18, 19), leukemic cell stimulation (20), and chemotactic migration of human glioma cells (21) and murine fibroblasts (22). Although the bioactivity of LPS in the nervous system has not been well studied, our findings support a model where LPS exhibits similar immunomodulatory activities to stimulate microglial function in the brain. Reports on VLC-LPS are scarce in the literature. Interestingly, however, parasitic worms that cause schistosomiasis, a chronic blood-vascular disease common in the developing world, produce Fig. 6. Working model to explain the molecular and cellular basis for ABHD12-dependent neuropathology. Our results point to a model where many of the same VLC-LPS species that are highly elevated in ’ −/− loss of ABHD12 leads to accumulation of the enzyme s endogenous LPS ABHD12 brains (23). Parasite-derived VLC-LPS, but not syn- substrates, as well as corresponding (indirect) alterations in PS lipids. These thetic 16:0 LPS, was found to activate toll-like receptor (TLR)2 on metabolic changes occur in young mice (<6 mo), before the appearance of dendritic cells and modify their T-cell–stimulating activity. TLR2 cellular or behavioral signs of neuropathology. By 12 mo of age, neuro- recognizes many pathogen-derived molecules and is expressed in inflammatory markers, such as cerebellar microglial activation, are observed − − the CNS by glial cells (microglia, astrocytes, oligodendrocytes) and in ABHD12 / mice, which are then followed by PHARC-related behavioral neurons (24). TLR2 agonists initiate a proinflammatory response deficits that intensify with age. FFA, free fatty acid; GPS, glycerophosphate.

Blankman et al. PNAS Early Edition | 5of6 Downloaded by guest on September 26, 2021 PHARC, then identifying the enzymatic source for these proin- (cresyl violet) to visualize neurons. Immunohistochemistry was performed on flammatory lipids in the nervous system (i.e., the PS lipase that frozen free-floating sections as described previously (36). Sections from produces VLC-LPS lipids) could provide an attractive molecular 18- to 21-mo-old mice (n = 4) were immunostained with a GFAP antibody to target for therapeutic intervention. visualize astrocytes or a calbindin antibody to visualize Purkinje neurons. Microglia were visualized in sections from 6-, 12-, and 18-mo-old ABHD12+/+ Materials and Methods and ABHD12−/− mice (n = 3–4 per genotype per age) using an antibody An extended section is provided in SI Materials and Methods. against Iba1 and the number of enlarged microglia (>200 μm) in matching cerebellar sections were counted using ImageJ software (National Institutes − − Generation of ABHD12 / Mice. Constitutive ablation of the Abhd12 gene was of Health). achieved on the C57BL/6 genetic background using standard gene targeting techniques. All mice used in this study were generated from breeding LC-MS Metabolite Profiling. Untargeted metabolomics analysis was per- +/− ABHD12 mice. formed on organic-soluble brain extracts from 2- to 6-mo-old ABHD12+/+ and ABHD12−/− mice (n = 3–5) as described previously (11). The molecular as- fi Biochemical Studies. Activity-based protein pro ling (ABPP) and signment of ABHD12-regulated metabolites was achieved by high mass-ac- activity assays of mouse brain and transfected HEK293T cell homogenates curacy measurements, MS/MS fragmentation analysis, and coelution studies were performed as described previously (5, 35). Untargeted LC-MS proteomic − − + − with synthetic lipid standards (37). Tissue lipid levels in ABHD12 / ,ABHD12 / , analysis of brain homogenates was performed as previously described (35). + + and ABHD12 / littermates (6 mo old; n = 5) were quantified by targeted MRM methods (35). Behavioral Studies. Tests for locomotor activity, startle response, auditory brainstem response, rotarod performance, hanging wire performance, and optomotor activity were performed as described in the SI Materials and Methods. ACKNOWLEDGMENTS. We thank Sergey Kupriyanov and Greg Martin [The Scripps Research Institute (TSRI) Mouse Genetics Core], Amanda Roberts and Salvador Huitron-Resendiz (TSRI Mouse Behavioral Assessment Core), Margie Histological and Immunohistochemical Analyses. Slide-mounted sections from Chadwell and Anita San-Soucie (TSRI Histology Core), and Holly Pugh for + + − − 18- to 21-mo-old ABHD12 / and ABHD12 / mice (n = 4) were stained with technical assistance. This work was supported by US National Institutes of H&E to visualize gross brain organization, LFB to visualize myelin, and Nissl Health Grants DA017259 and DA009789.

1. Online Mendelian Inheritance in Man (2012) McKusick-Nathans Institute of Genetic 19. Frasch SC, et al. (2011) Signaling via macrophage G2A enhances efferocytosis of dying Medicine (John Hopkins University, Baltimore). Available at http://omim.org. Accessed neutrophils by augmentation of Rac activity. J Biol Chem 286(14):12108–12122. September 15, 2012. 20. Park KS, Lee HY, Kim M-K, Shin EH, Bae Y-S (2005) Lysophosphatidylserine stimulates 2. Fiskerstrand T, et al. (2009) A novel Refsum-like disorder that maps to leukemic cells but not normal leukocytes. Biochem Biophys Res Commun 333(2): 20. Neurology 72(1):20–27. 353–358. 3. Fiskerstrand T, et al. (2010) Mutations in ABHD12 cause the neurodegenerative 21. Lee SY, et al. (2008) Lysophosphatidylserine stimulates chemotactic migration in U87 disease PHARC: An inborn error of endocannabinoid metabolism. Am J Hum Genet human glioma cells. Biochem Biophys Res Commun 374(1):147–151. 87(3):410–417. 22. Park KS, et al. (2006) Lysophosphatidylserine stimulates L2071 mouse fibroblast che- 4. Eisenberger T, et al. (2012) Targeted next-generation sequencing identifies a homo- motactic migration via a process involving pertussis toxin-sensitive trimeric G-pro- zygous in ABHD12, the gene underlying PHARC, in a family teins. Mol Pharmacol 69(3):1066–1073. clinically diagnosed with Usher syndrome type 3. Orphanet J Rare Dis 7(1):59. 23. van der Kleij D, et al. (2002) A novel host-parasite lipid cross-talk. Schistosomal lyso- 5. Blankman JL, Simon GM, Cravatt BF (2007) A comprehensive profile of brain that phosphatidylserine activates toll-like receptor 2 and affects immune polarization. hydrolyze the endocannabinoid 2-arachidonoylglycerol. Chem Biol 14(12):1347–1356. JBiolChem277(50):48122–48129. 6. Simon GM, Cravatt BF (2010) Activity-based proteomics of enzyme superfamilies: 24. Okun E, et al. (2009) Toll-like receptors in . Brain Res Brain Res Rev Serine as a case study. J Biol Chem 285(15):11051–11055. 59(2):278–292. 7. Pujol A, et al. (2002) Late onset neurological phenotype of the X-ALD gene inactivation 25. Kinsner A, et al. (2006) Highly purified lipoteichoic acid induced pro-inflammatory in mice: A mouse model for adrenomyeloneuropathy. Hum Mol Genet 11(5):499–505. signalling in primary culture of rat microglia through Toll-like receptor 2: Selective 8. Wheeler VC, et al. (2000) Long glutamine tracts cause nuclear localization of a novel potentiation of nitric oxide production by muramyl dipeptide. J Neurochem 99(2): form of huntingtin in medium spiny striatal neurons in HdhQ92 and HdhQ111 knock- 596–607. in mice. Hum Mol Genet 9(4):503–513. 26. Neher JJ, et al. (2011) Inhibition of microglial phagocytosis is sufficient to prevent 9. Lorenzetti D, et al. (2000) Repeat instability and motor incoordination in mice with inflammatory neuronal death. J Immunol 186(8):4973–4983. a targeted expanded CAG repeat in the Sca1 . Hum Mol Genet 9(5):779–785. 27. Lehnardt S, et al. (2006) A mechanism for neurodegeneration induced by group B 10. Shinzawa K, et al. (2008) Neuroaxonal dystrophy caused by group VIA streptococci through activation of the TLR2/MyD88 pathway in microglia. J Immunol A2 deficiency in mice: A model of human neurodegenerative disease. J Neurosci 28(9): 177(1):583–592. 2212–2220. 28. Hoffmann O, et al. (2007) TLR2 mediates neuroinflammation and neuronal damage. 11. Saghatelian A, et al. (2004) Assignment of endogenous substrates to enzymes by J Immunol 178(10):6476–6481. global metabolite profiling. Biochemistry 43(45):14332–14339. 29. Sango K, et al. (1995) Mouse models of Tay-Sachs and Sandhoff diseases differ in 12. Gross RW, Han X (2011) Lipidomics at the interface of structure and function in sys- neurologic phenotype and ganglioside metabolism. Nat Genet 11(2):170–176. tems biology. Chem Biol 18(3):284–291. 30. Lu J-F, et al. (1997) A mouse model for X-linked adrenoleukodystrophy. Proc Natl 13. Smith CA, Want EJ, O’Maille G, Abagyan R, Siuzdak G (2006) XCMS: Processing mass Acad Sci USA 94(17):9366–9371. spectrometry data for metabolite profiling using nonlinear peak alignment, match- 31. Benito C, et al. (2003) Cannabinoid CB2 receptors and fatty acid amide hydrolase are ing, and identification. Anal Chem 78(3):779–787. selectively overexpressed in neuritic plaque-associated glia in Alzheimer’s disease 14. Makide K, Kitamura H, Sato Y, Okutani M, Aoki J (2009) Emerging lysophospholipid brains. J Neurosci 23(35):11136–11141. mediators, lysophosphatidylserine, lysophosphatidylthreonine, lysophosphatidyle- 32. Yiangou Y, et al. (2006) COX-2, CB2 and P2X7-immunoreactivities are increased in thanolamine and lysophosphatidylglycerol. Prostaglandins Other Lipid Mediat 89(3- activated microglial cells/macrophages of multiple sclerosis and amyotrophic lateral 4):135–139. sclerosis spinal cord. BMC Neurol 6(1):12. 15. Bigon E, Bruni A, Mietto L, Toffano G (1980) Lysophosphatidylserine-induced release 33. Walter L, et al. (2003) Nonpsychotropic cannabinoid receptors regulate microglial cell of intra-cellular amines in mice. Br J Pharmacol 69(1):11–12. migration. J Neurosci 23(4):1398–1405. 16. Martin TW, Lagunoff D (1979) Interactions of lysophospholipids and mast cells. Na- 34. Sánchez AJ, García-Merino A (2012) Neuroprotective agents: . Clin Im- ture 279(5710):250–252. munol 142(1):57–67. 17. Smith GA, Hesketh TR, Plumb RW, Metcalfe JC (1979) The exogenous lipid re- 35. Schlosburg JE, et al. (2010) Chronic blockade causes func- quirement for histamine release from rat peritoneal mast cells stimulated by conca- tional antagonism of the endocannabinoid system. Nat Neurosci 13(9):1113–1119. navalin A. FEBS Lett 105(1):58–62. 36. Nomura DK, et al. (2011) Endocannabinoid hydrolysis generates brain prostaglandins 18. Frasch SC, et al. (2008) NADPH oxidase-dependent generation of lysophosphati- that promote neuroinflammation. Science 334(6057):809–813. dylserine enhances clearance of activated and dying neutrophils via G2A. J Biol Chem 37. Long JZ, et al. (2011) Metabolomics annotates ABHD3 as a physiologic regulator of 283(48):33736–33749. medium-chain phospholipids. Nat Chem Biol 7(11):763–765.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1217121110 Blankman et al. Downloaded by guest on September 26, 2021