G protein‐coupled receptor 37‐like 1 modulates astrocyte glutamate transporters and neuronal NMDA receptors and is neuroprotective in ischemia

Abstract We show that the G protein‐coupled receptor GPR37‐like 1 (GPR37L1) is expressed in most astrocytes and some oligodendrocyte precursors in the mouse central nervous system. This contrasts with GPR37, which is mainly in mature oligodendrocytes. Comparison of wild type and Gpr37l1–/– mice showed that loss of GPR37L1 did not affect the input resistance or resting potential of astrocytes or neurons in the hippocampus. However, GPR37L1‐mediated signalling inhibited astrocyte glutamate transporters and – surprisingly, given its lack of expression in neurons – reduced neuronal NMDA receptor (NMDAR) activity during prolonged activation of the receptors as occurs in ischemia. This effect on NMDAR signalling was not mediated by a change in the release of D‐serine or TNF‐α, two astrocyte‐derived agents known to modulate NMDAR function. After middle cerebral artery occlusion, Gpr37l1 expression was increased around the lesion. Neuronal death was increased by ∼40% in Gpr37l1–/– brain compared to wild type in an in vitro model of ischemia. Thus, GPR37L1 protects neurons during ischemia, presumably by modulating extracellular glutamate concentration and NMDAR activation.


| I N TR ODU C TI ON
Activation of receptors on astrocytes is increasingly thought to modulate the activity and function of neurons. Release of astrocyte-derived "gliotransmitters" such as glutamate and D-serine, triggered by activation of receptors on astrocytes by signals from neurons or other cells, can alter synaptic transmitter release and the excitability of neurons (Bazargani and Attwell, 2016). However, the functions of most glial receptors are poorly understood. Here, we examine the function of one such glial-restricted receptor, the G protein-coupled receptor GPR37-like 1 (GPR37L1).
We identified GPR37L1 as a potential astrocyte-specific receptor during a visual screen of the Allen Brain Atlas gene expression database (N.P.P. unpublished). GPR37L1 belongs to the Class A rhodopsin-like receptor subfamily of GPCRs. The Gpr37l1 coding sequence was first identified by sequence similarity to the endothelin type B receptor gene, but GPR37L1 is unable to bind endothelin or related peptides (Leng, Gu, Simerly, & Spindel, 1999;Valdenaire et al., 1998). It is highly expressed in the central nervous system (CNS), heart and Sarah Jolly and Narges Bazargani contributed equally to this work. gastrointestinal tract (Freeman, 2010;Ito et al., 2009;Min et al., 2010;Valdenaire et al., 1998). GPR37L1 shares >40% amino acid sequence similarity with its close relative GPR37, which is also expressed in the CNS. Transcriptome studies suggest that Gpr37l1 is expressed mainly in astrocytes, oligodendrocyte precursors (OPs) and newly formed oligodendrocytes (OLs) in humans and mice (web.stanford.edu/group/ barres_lab/cgi-bin/geneSearch.py?geneNameIn 5 gpr37l1), while Gpr37 is mainly in newly formed and myelinating OLs (web.stanford. edu/group/barres_lab/cgi-bin/geneSearch.py?geneNameIn 5 gpr37) (Imai et al., 2001;Zhang et al., 2014). GPR37 is a substrate of Parkin, an E3 ubiquitin ligase that might regulate the dopaminergic system (Imai et al., 2001). It also regulates OL differentiation and myelination (Yang, Vainshtein, Maik-Rachline, & Peles, 2016). In contrast, little is known about the function of GPR37L1 in the CNS, except that it might modulate development of the cerebellum by regulating sonic hedgehog signalling (Marazziti et al., 2013). Recently, the polypeptide "prosaposin" (also known as PSAP) was identified as a potential ligand for both GPR37 and GPR37L1 (Meyer, Giddens, Schaefer, & Hall, 2013). Prosaposin can be secreted into the extracellular space and this is enhanced following conditions of cellular stress such as ischemia (Costain et al., 2010;Hiraiwa et al., 2003;Yokota, Uchijima, Nishizawa, Namba, & Koide, 2001). Prosaposin and prosaptide (an active fragment of prosaposin) have neuroprotective and glioprotective properties (Meyer, Giddens, Coleman, & Hall, 2014;Morita et al., 2001;Sano et al., 1994) by acting on GPR37 and GPR37L1. However, a separate study has suggested that GPR37L1 is constitutively active and that its activity is regulated by proteolytic cleavage near the N-terminus (Coleman et al., 2016). It is therefore unclear whether GPR37L1 activation is triggered by binding of an extracellular ligand (like prosaposin) or by posttranslational modification or cleavage.
We report that GPR37 and GPR37L1 are expressed in the postnatal CNS in non-overlapping cell populations. While GPR37 is expressed mainly in differentiated OLs, GPR37L1 is expressed in astrocytes and OPs. We found that (1) GPR37L1 expression does not change the basic membrane properties of hippocampal astrocytes or neurons, (2) GPR37L1 mRNA expression is upregulated in ischemia in vivo, (3) GPR37L1 expression and signalling activated by its ligand prosaptide are neuroprotective in ischemic brain slices, and (4) prosaptide-evoked GPR37L1-signalling inhibits glutamate transporters in astrocytes and reduces neuronal NMDAR activity. We suggest that the latter two effects combine to confer neuroprotection during ischemia.

| In situ hybridization
Our in situ hybridization (ISH) procedure has been described (Jolly,

| Electrophysiology
Electrophysiological recordings from wild type and knockout mice were performed with the experimenter being blind to the mouse genotype.

| Whole-cell patch-clamp recording
Neurons and astrocytes were selected visually for patch-clamping, and their identities were confirmed from their morphology after diffusion of Alexa Fluor 488 into neurons, or Alexa Fluor 488/594 into astrocytes. When voltage steps were applied, observing a large voltagegated sodium current, or a passive current-voltage relation of low resistance, confirmed that cells were pyramidal neurons, or astrocytes, respectively. Data for the drug responses presented were sampled at 1 kHz and filtered at 500 Hz.

| Field excitatory postsynaptic current recordings
Thick-walled glass electrodes, filled with HEPES-based aCSF, were connected to a stimulator, and stimuli (in 20 V steps from 0-100 V) were applied with the electrode tip close to the CA3 pyramidal axon initial segments in hippocampal slices, to evoke field excitatory postsynaptic currents (fEPSCs, recorded in voltage-clamp mode) that were recorded using an aCSF-filled pipette near the apical dendrites of the CA1 pyramidal neurons.

| Image analysis
Sections were examined in a LEICA SPE confocal microscope and micrographs were analyzed with ImageJ software (NIH), unless otherwise stated.
Protein levels were assessed with a Bradford assay with bovine serum albumin as the standard.

| Chemical Ischemia
Hippocampal slices (270 lm) from P14-P16 Gpr37l1 1/1 or Gpr37l1 -/littermates were allowed to recover for 40 min before being incubated for 30 min at 378C in (1)  (2) ischemic solution with the glucose replaced by 7 mM sucrose, gassed with 95% N 2 /5% CO 2 , and with 2 mM iodoacetate and 25 lM antimycin added to block glycolysis and oxidative phosphorylation, respectively. Propidium iodide (PI, 7.5 lM) was added to label dead cells by binding to DNA/RNA. Slices were then fixed for 1 hr in 4% PFA and immunohistochemistry for NEUN and GFAP was performed.
Two slices per condition were analyzed. Four z-stacks were generated per slice and PI-labelled cells were counted in the stratum radiatum and pyramidal cell layer, with experimenters blind to the mouse genotype. Images were 275 mm square and the z-stack depth was 25 mm (zstep 5 0.5 mm). Gain and offset settings were identical for all slices in each experiment.

| Middle cerebral artery occlusion
Brains from mice that had experienced middle cerebral artery occlusion (MCAO) were kindly donated by Kaylene Young. The ISH signal for Gpr37l1 was quantified using ImageJ ( Figure 8).

| Mouse behaviour
Mice were handled daily for 1 week to habituate them prior to behavioural tests. They were left in their home cages in the behaviour room for 30 min before initiating tests. Trials were recorded using a top-view video camera and white noise (50 dB) was played during the tests.

| Open-field test
Mice were allowed to explore a 30 cm 2 arena for 30 min and tracked with ActualTrack software. Total distance travelled and time spent in the centre was calculated.

| Novel-object recognition
Mice were placed in the arena for 5 min before being familiarized for 10 min with two identical objects. After a 10-min delay, mice were tested for 10 min by placing them in the arena with one of the original objects replaced by a novel object (NOR). 24 hr later mice were tested again for 10 min with one familiar and one new object (NOR 1 24).
The times spent inspecting the novel and familiar objects were assessed with the ActualTrack software. The discrimination index (DI) was calculated as: DI 5 {(time spent with novel object) minus (time spent with familiar object)}/(total time spent with both objects).

| Rotarod
About 2-3 months or 6-month-old mice were familiarized with the rotarod for three trials at a constant speed of 4 rpm. They were then tested for 3 days, with three trials/day, at an accelerating speed from 4 to 40 rpm for up to 5 min. The latency to fall was recorded.

| Statistics
Statistical significance was determined with GraphPad Prism (GraphPad Software, CA, USA) and OriginPro software. Data normality was assessed using Kolmogorov-Smirnov tests. Data are presented as mean 6 SEM. Data were corrected for multiple comparisons using a procedure equivalent to Holm-Bonferroni (for N comparisons, the most significant p value is multiplied by N, the 2nd most significant by N-1, the 3rd most significant by N-2, etc.; corrected p values are regarded as significant if they are <0.05).

| R E SU LTS
3.1 | Gpr37l1 is expressed in astrocytes and some oligodendrocyte precursors We combined in situ hybridization (ISH) with immunolabelling to determine which cells express Gpr37l1 in different brain areas. Gpr37l1 is widely expressed in the hippocampus (Figure 1a,c,e), cerebral cortex ( Figure 1b,d,f) and corpus callosum (Supporting Information Figure 1).
From our ISH and the Allen Brain Atlas, expression of Gpr37l1 in the hippocampus is similar in CA1, CA3 and dentate gyrus (http://mouse.  To confirm these results, we used Gpr37l1-LacZ heterozygous mice in which a LacZ cassette was inserted into the first exon of the 3.2 | Gpr37l1 and Gpr37 are expressed in different cells GPR37L1 and its close relative GPR37 share 48% amino acid identity in human (Valdenaire et al., 1998). ISH for Gpr37 mRNA showed that Gpr37 was expressed in many cells in subcortical structures such as the hypothalamus and thalamus as well as in the corpus callosum, and in smaller numbers of cells in the cortex and hippocampus (Figure 3).
Gpr37 was mostly in OLIG2 1 oligodendrocyte (OL)-lineage cells ( Figure   3a-c) but not in PDGFRA 1 cells (Figure 3d-f), suggesting that mature OLs but not OPs express Gpr37. We observed no expression of Gpr37 in GFAP 1 astrocytes (not shown). Occasionally, Gpr37 expression was seen in some NEUN 1 neurons but not in IBA1 1 microglia (not shown).
In contrast to Gpr37, Gpr37l1 is not expressed in CC1  3.4 | Gpr37l1 KO does not alter the input resistance of astrocytes or neurons or neuronal excitability GPR37L1 can protect astrocytes against oxidative stress (Meyer et al., 2013), and we show below that it also protects neurons in ischemia.
This suggests that the membrane properties or response to glutamate of neurons and astrocytes might be modulated by GPR37L1.
Hippocampal astrocytes expressing or lacking Gpr37l1 expression, as defined by fluorescent detection of GFP in Gpr37l1-GFP mice, did not differ in input resistance (Figure 4a, p 5 0.9) or resting potential FIG URE 3 Gpr37l1 and Gpr37 are expressed in mutually exclusive cell populations. Cells expressing Gpr37 transcripts were mostly found in subcortical areas (hypothalamus and thalamus) and in corpus callosum; fewer Gpr37 1 cells were present in cortex and hippocampus. Fluorescent ISH revealed expression of Gpr37 in OLIG2 1 OL lineage cells (a-c), but not in OPs expressing PDGFRA (d-f). Conversely, immunolabelling of Gpr37l1-LacZ heterozygous mice for b-galactosidase showed that Gpr37l1 is not expressed in CC1-positive mature OLs (g-i). Fluorescent double-ISH demonstrates that Gpr37l1 and Gpr37 are expressed in different cells in hippocampus (j), cortex (k) and corpus callosum (L). Dotted lines: boundary between cortical grey matter and corpus callosum (cc). Scale: 50 mm JOLLY ET AL.

| 53
We assessed the excitability of CA3 pyramidal neurons using whole-cell current-clamp recordings in slices from Gpr37l1 -/and Gpr37l 1/1 mice. We found no difference between the Gpr37l1 -/and (e, f) CA3 pyramidal cells in hippocampal slices from wild type and Gpr37l1 knock-out mice have similar (e) membrane resistance and (f) resting potential. (g, h) CA1 pyramidal cells in hippocampal slices from wild type and Gpr37l1 knock-out mice have similar (g) membrane resistance and (h) capacitance (used to normalise drug-evoked currents in Figure 5; resting potential was not studied as the internal solution contained Cs 1 ). (i, j) Excitability of CA3 neurons in slices from wild type and Gpr37l1 knock-out mice. (i) Latency to first action potential as a function of current injected into CA3 pyramidal neurons (Gpr37l1 1/1 n 5 15, Gpr37l1 -/n 5 15). (j) Percentage of responses in (i) that showed action potentials as a function of injected current. (k) Field EPSCs evoked in area CA1 by applying stimuli to the Schaffer collaterals of CA3 axons, in 20 V steps from 0 to 100 V. Amplitudes of field EPSCs were normalized to the maximal response (at 100 V) for each slice (Gpr37l1 1/1 n 5 8, Gpr37l1 -/n 5 9) might also alter the extracellular glutamate concentration reached in ischemia when transporters reverse and release glutamate (Rossi, Oshima, & Attwell, 2000). Such a change of glutamate release should alter NMDA receptor (NMDAR)-mediated cell death in ischemia (Brassai, Suvanjeiev, Ban, & Lakatos, 2015;Vornov and Coyle, 1991) and thus contribute to the neuroprotective effect in ischemia of GPR37L1 and prosaptide [see below and Morita et al. (2001)].
To test this hypothesis, we first compared the expression levels of the glutamate transporters, GLT-1 and GLAST, expressed in astrocytes, using hippocampal extracts from Gpr37l1 1/1 and Gpr37l1 -/mice.
Astrocytes in the stratum radiatum were whole-cell voltage-clamped (near their resting potential) and responses to D-aspartate (200 lM), a substrate for glutamate transporters (Gundersen et al., 1995), were recorded in the presence and absence of the glutamate transporter blocker TFB-TBOA (10 lM, Figure 5c). Blockers of NMDARs, AMPARs, GABA A Rs, voltage-gated Na 1 channels and inwardly rectifying potassium channels were also present throughout the experiment (see Materials and Methods).
TFB-TBOA (10 lM), which blocks both GLT-1 and GLAST transporters (Shimamoto et al., 2004), blocked the D-aspartate evoked current in both Gpr37l1 1/1 and Gpr37l1 -/astrocytes (Figure 5e, p 5 0.4), confirming that the current is generated by glutamate transporters. The lack of a difference in glutamate transporter current with GPR37L1 knocked out could reflect GPR37L1 not being activated under physiological conditions, since it is known that the expression and release of prosaposin are up-regulated following ischemia (Costain et al., 2010;Hiraiwa et al., 2003). We therefore investigated the effect of prosaptide on the glutamate transporter current evoked by D-aspartate (200 lM), to test whether it modulates the uptake current in the presence or absence of GPR37L1.
The mean current generated by prosaptide alone was 21 6 2 pA for three Gpr37l1 1/1 astrocytes and 0.2 6 1.0 for four Gpr37l1 -/astrocytes (not significantly different from zero, p 5 0.5 and 0.9, respectively). The inhibition of glutamate transporters by prosaptide in Gpr37l1 1/1 astrocytes is presumably mediated by GPR37L1 receptors in the astrocytes themselves and cannot reflect prosaptide acting on the related GPR37 receptor because it had no effect in Gpr37l1 -/slices.

| How does astrocyte GPR37L1 regulate neuronal NMDAR responses?
Because GPR37L1 is present in astrocytes (and OPs, although these receptors may be less well positioned to regulate NMDAR responses), while the NMDAR responses recorded are from neurons, a signal must pass from astrocytes to neurons to alter the NMDAR response when GPR37L1 is activated by prosaptide. We tested whether the gliotransmitters D-serine or TNF-a mediate this effect.
The addition of prosaptide to boost GPR37L1 signaling significantly reduced cell death in the pyramidal layer of the hippocampus in Gpr37l1 1/1 slices (a 25% decrease comparing cell death in ischemia with or without prosaptide, n 5 4, p 5 0.01, Figure 9f). Prosaptide also decreased cell death in the hippocampus of Gpr37l1 -/mice (20% decrease, n 5 4, p 5 0.008, Figure 9f). However, prosaptide did not reduce ischemia-evoked cell death in the stratum radiatum in either the Gpr37l1 1/1 or the Gpr37l1 -/slices (1.1% increase in death in Gpr37l1 1/1 slices, and 0.9% increase in Gpr37l1 -/slices, data not shown).
Thus, the activation of GPR37L1 that occurs in ischemia in the absence of added prosaptide (presumably caused by release of endogenous prosaposin) is neuroprotective for pyramidal neurons of the hippocampus and this protective effect is amplified when GPR37L1 is stimulated further by bath application of prosaptide. The fact that prosaptide is also neuroprotective in the Gpr37l1 -/mice suggests that the neuroprotective effect of prosaposin might also rely partly on other receptors such as GPR37 (expressed in mature OLs, Figure 3), or on other unknown mechanisms.
We show that Gpr37l1 is expressed in most or all astrocytes and a subset of OPs (Figure 1). The expression pattern differed from that of the related receptor Gpr37, which was mainly in mature OLs and not in astrocytes (Figure 3), contradicting a report that Gpr37 is expressed in cultured astrocytes (Meyer et al., 2013) but consistent with transcriptome data (Zhang et al., 2014). In mice, Gpr37l1 expression increases over the first postnatal month and continues to be expressed in adulthood ( Figure 2) implying a role, not just in development, but in the function of the mature nervous system. Surprisingly, we could not verify an earlier claim that GPR37L1 deletion affects motor performance (Marazziti et al., 2013), possibly due to the use of different Gpr37l1 -/mouse lines with different genetic backgrounds, and we found no obvious effect on OP proliferation. Gpr37l1 expression also had no effect on the resting electrical properties of hippocampal pyramidal neurons or astrocytes ( Figure 4) but it had two potentially important effects on glutamatergic signalling.
First, although expression of Gpr37l1 did not affect expression of the astrocyte glutamate transporters GLT-1 and GLAST, activation of GPR37L1 with the prosaposin cleavage product prosaptide inhibited astrocyte glutamate uptake and this effect was abolished in the Gpr37l1 KO ( Figure 5). This suggests that prosaptide was acting through GPR37L1 receptors expressed on the astrocytes being recorded from, presumably [since GPR37L1 is coupled to G i proteins; Meyer et al. (2013)] by lowering the cyclic AMP level in the astrocyte, altering phosphorylation by protein kinase A and thereby affecting the transporter cycling rate or trafficking of the transporter to and from the plasma membrane. Deleting Gpr37l1 did not affect the uptake current in the absence of applied prosaptide ( Figure 5), suggesting that there is normally little tonic release of prosaposin (at least in brain slices) and little spontaneous activity of the GPR37L1 receptor, contradicting the suggestion (Coleman et al., 2016) that GPR37L1 is spontaneously active (although we cannot rule out the possibility of compensation in response to the knock-out). However, prosaposin expression and release are increased in ischemia (Costain et al., 2010;Hiraiwa et al., 2003) and we found that expression of Gpr37l1 is increased in the penumbra of lesions caused by MCAO (Figure 8), so it is likely that glutamate transport activity is inhibited in these conditions.
If mild ischemia inhibits glutamate uptake, there is expected to be a rise in extracellular glutamate concentration, which might desensitize AMPARs and tonically activate NMDARs, thus altering neuronal information processing. A further suppression of glutamate transport by prosaposin release in this situation will accentuate these effects. The situation is different in profound ischemia, however, when ion gradient run-down leads to glutamate transporters reversing and releasing glutamate, which reaches a concentration of 100-200 lM in the extracellular space and evokes a neurotoxic entry of Ca 21 via NMDAR channels (Krzyzanowska, Pomierny, Filip, & Pera, 2014;Rossi et al., 2000;Rothman & Olney, 1995). In this situation, inhibition of glutamate transport by prosaposin release will slow the release of glutamate. However, at least in the first few minutes of ischemia, transporter knock-out experiments measuring the latency to the anoxic depolarization (when the extracellular glutamate concentration rises dramatically) suggest that it is the neuronal glutamate transporters that reverse first rather than astrocyte transporters, probably because the intracellular glutamate concentration is higher in neurons than in astrocytes (Gebhardt, K€ orner, & Heinemann, 2002;Hamann, Rossi, Marie, & Attwell, 2002).
Second, and perhaps more importantly, prosaptide-evoked GPR37L1 signalling decreases the response of neurons to prolonged  Figure 6). Such prolonged activation will occur during the prolonged elevation of extracellular glutamate concentration that occurs in ischemia and GPR37L1 should thus decrease the neurotoxic rise of [Ca 21 ] i that occurs in neurons in ischemia. Indeed, expression of GPR37L1 was neuroprotective during ischemia even in the absence of added prosaptide (Figure 9)-an effect that presumably depends on the release of prosaposin that is induced by ischemia (Costain et al., 2010;Hiraiwa et al., 2003;Yokota et al., 2001). In vivo, upregulation of Gpr37l1 in the penumbra of an ischemic lesion (Figure 8) might promote GRP37L1-mediated neuroprotection. The mechanism by which GPR37L1 decreases neuronal responses to prolonged activation of NMDARs is mysterious. Because the GPR37L1 is located in astrocytes, to regulate neuronal NMDARs a gliotransmitter of some sort must have its release from the astrocytes modulated when GPR37L1 is activated. We have ruled out two candidates for this role-D-serine and TNF-a -which have previously been shown to increase NMDAR responses when released from astrocytes (Henneberger et al., 2010;Shigetomi et al., 2013; Figure 7). Our work suggests that a further gliotransmitter must exist that has a similar effect, and that its release is modulated by GPR37L1, but further work is needed to identify this agent.
We found that Gpr37l1 is also expressed in 25% of OPs but we did not detect any effect of Gpr37l1 knockout on OP density or myelination in healthy mice (Supporting Information Figure 3b,c). However, GPR37 and GPR37L1 might protect against demyelination caused by injury or disease, and/or stimulate remyelination (Hiraiwa, Campana, Mizisin, Mohiuddin, & O'Brien, 1999;Hiraiwa, Taylor, Campana, Darin, & O'Brien, 1997). Myelinating OLs are sensitive to ischemia (Back & Rosenberg, 2014) and are probably damaged in our in vitro ischemia experiments, but we did not quantify this. The potential glioprotective role of GPR37L1 during ischemia or other insults, and what distinguishes the GPR37L1-expressing and nonexpressing subpopulations of OPs, are interesting questions for the future.