Comments by Dr. Nicola J. Smith, National Heart Foundation Future Leader Fellow & Group Leader, Molecular Pharmacology Laboratory, Victor Chang Cardiac Research Institute, Australia
As is often the case with orphan GPCRs, assigning the endogenous ligand has been controversial for the closely related peptide family orphans, GPR37 and GPR37L1. In 2013, Randy Hall and his team (PubMed: 23690594) first reported an association between both centrally-expressed orphan GPCRs and prosaposin (PSAP) and prosaptide (TX14A), the synthetic active epitope of PSAP. Since that time there has been much debate in the field about whether this pairing is correct, with some authors corroborating the findings (PubMed: 24371137; 30010619, 28795439) and others not (PubMed: 23396314; 27072655; 28688853). Note that Head Activator, found in Hydra, was earlier reported as a ligand (PubMed: 16443751) but was quickly discredited (PubMed:28688853; 23686350).
A recent paper by Sergey Kasparov’s laboratory in Bristol has added further fuel to the fire. In a series of well controlled experiments, Liu et al. (PubMed: 30260505) provided convincing evidence that prosaptide is cyto- and neuro-protective and promotes chemotaxis. They are also the first group to demonstrate an effect of prosaptide at a more physiologically plausible potency. At the same time, Bang et al. (PubMed: 30010619) published a ground-breaking paper linking GPR37 expression to macrophage function. Moreover, they proposed a second, more potent ligand for GPR37 (GPR37L1 was not studied): the pro-resolving mediator neuroprotectin D1 (NPD1). Using HEK293 cells expressing GPR37, NPD1 was a potent stimulator of Gαi/o-dependent calcium flux; findings that were corroborated in macrophages isolated from wild type, but not GPR37 knock-out, mice (PubMed: 30010619). Thus, it may be that the endogenous ligand for GPR37 (and perhaps GPR37L1?) is not a peptide after all, but a lipid.
These two studies, while exciting, do little to help us resolve the conundrum that is PSAP/prosaptide and GPR37/GPR37L1. At the very least, it seems likely that prosaptide, if not the highest affinity endogenous ligand at GPR37, is certainly capable of signalling through GPR37 to stimulate Gαi/o signal transduction (whether it is the most potent endogenous agonist will be shown in time as independent groups seek to validate the actions of NPD1).
But what of GPR37L1? This is harder to answer as a number of studies linking GPR37L1 to PSAP/prosaptide have been performed in double GPR37/GPR37L1 knock-out backgrounds or inappropriate tissue models. For example, in the original paper connecting prosaptide to the receptors, the authors claimed prosaptide acted through both GPR37 and GPR37L1 in primary astrocytes, despite the fact that their Western blots demonstrated marked GPR37 expression in comparison to GPR37L1 in the cells (PubMed: 23690594). More recently, they failed to recapitulate this initial pairing in a HEK293 model (PubMed: 28688853). The absence of GPR37L1 in primary astrocytes is consistent with the animal knock-out work of Marazziti et al. (PubMed: 24062445), who showed that GPR37L1 protein was barely detectable before post-natal day 15, which is after the window for isolating primary astrocyte cultures (confirmed by PubMed: 28795439). Coleman et al. (PubMed: 27072655) overcame this expression issue, with difficulty, by using cerebellar slice cultures in vitro to examine Gαs, but not prosaptide, signalling in wild type and knock-out tissue.
In the neuroprotection paper by Liu et al. (PubMed: 30260505) it is clear that prosaptide or PSAP are exhibiting an effect on the cells. By depleting astrocytes of PSAP and then reintroducing prosaptide exogenously, there is an obvious effect on cell migration, cytotoxicity and neuroprotection – phenotypes that are all lost when shRNA knocking down expression of both GPR37 and GPR37L1 are used. Frustratingly, though, the use of a double knock-down approach makes it impossible to ascribe a specific effect to GPR37L1. While GPR37 and GPR37L1 are very closely related by phylogeny and have highly similsar binding sites (PubMed: 27992882), this does not mean that prosaptide is a ligand at both receptors, nor that both receptors signal via the same G proteins (another area of controversy for GPR37L1, where contradictory studies including two by the same team show either Gαi/o or Gαs signaling: PubMed: 23690594, 30260505 vs 27072655, 28688853). Thus, the failure to use single receptor knock-out/knock-downs, or isolate cells with endogenous expression of GPR37L1, represent major limitations in these studies.
Other than the confounding effects of both GPR37 and GPR37L1 deletion in tested cells, what are other reasons that could explain the inconsistencies between studies? Kasparov and colleagues (PubMed: 30260505) attribute this to cellular background, stating that previous studies that failed to confirm prosaptide/GPR37L1 coupling (PubMed: 27072655, 28688853) used HEK293 cells that must be lacking in the necessary endogenous machinery for signal transduction (PubMed: 30260505). To support this claim, they turned to the PRESTO-Tango assay in HEK293 cells to demonstrate prosaptide stimulation could not lead to GPR37L1-dependent recruitment of beta-arrestin. The assay choice is surprising because previous beta-arrestin-based screens at GPR37L1 have failed to show that the receptor can indeed recruit arrestins (PubMed: 23396314, 25895059), and Liu et al. (PubMed: 30260505) did not provide evidence that recruitment was intact in a more physiologically relevant cellular background. Most puzzling though is that the original paper that identified prosaptide and PSAP as GPR37/37L1 ligands used HEK293 cells to make the original pairing (PubMed: 23690594). They also refute the physiological relevance of high constitutive Gαs signalling reported by Coleman et al., even though Coleman et al. demonstrated higher cAMP accumulation in cerebellar brain slices from wild type mice when compared to GPR37L1-/- (PubMed: 27072655).
Further muddying the waters, the physiological role of GPR37L1 itself remains enigmatic. For example, Min et al. (PubMed: 20100464) initially reported GPR37L1 null mice to have a staggering 62 mmHg increase in systolic blood pressure when compared to a cardiac-specific overexpressing model, with the presence of concomitant left ventricular hypertrophy. However, Coleman et al. (PubMed: 29625592) found a far more marginal cardiovascular phenotype, with a small increase in blood pressure evident in female mice only. Notably, male GPR37L1 knock-out mice appeared to be more susceptible to cardiovascular stressors, while females were cardioprotected (PubMed: 29625592). In terms of a developmental phenotype, Marazziti et al. (PubMed: 24062445) found that GPR37L1 null mice displayed precocious cerebellar development with enhanced performance in a rotarod test up to 1 year of age. More recently, though, Jolly et al. (PubMed: 28795439) failed to confirm a behavioural difference in their own GPR37L1 knock out mice. The links between GPR37L1 and neurological defects are also confounded by the fact that GPR37 also needs to be deleted in mice for a clear phenotype to be evident. For example, while a single point mutation in GPR37L1 (K349N) in a highly consanguineous family appeared to be causative of fatal progressive myoclonus epilepsy, the mouse phenotype was most pronounced in double GPR37/GPR37L1 knock-out animals (PubMed: 28688853). In vitro studies of the GPR37L1 K349N mutant found no difference between it and the wild type receptor in terms of receptor expression, processing, signalling or ubiquitination (PubMed: 28688853). In the absence of a transgenic K349N mutant mouse, or any confirmed synthetic agonists or antagonists, the authors then assessed seizure susceptibility in knock-out mice of either GPR37L1, GPR37 or both receptors. Interestingly, using the 6Hz-induced seizure model, the GPR37-/- mice appeared to have a more pronounced phenotype than the GPR37L1-/- mice, while double KO mice were extremely susceptible to seizures at all frequencies tested. GPR37 and double KO mice both displayed more spontaneous seizures, although curiously in the flurothyl-induced seizure model only GPR37L1-/- differed from wild type. Thus, conclusive links between GPR37L1 and a specific physiological or pathophysiological state remain to be provided and it seems in general that we are far from understanding the true biology and pharmacology of the receptor.
Response by S Kasparov and AG Teschemacher.
We have read with interest the comments posted by Prof NJ Smith and would like to contribute to the discussion on GPR37/GPR37L1 activation, as we believe that our recent data, rather than ‘adding fuel to the fire’, not only confirm the functional activation of GPR37L1 (and possibly GPR37) by PSAP/prosaptide, but also highlight several issues that may have caused the current controversy in the field.
The focus of our recent study (Liu et al 2018) is the stimulation astrocytic cell migration, cytotoxicity and neuroprotection by PSAP and prosaptide via activation of GPR37L1 and/or GPR37. As Prof NJ Smith states, GPR37L1 and GPR37 might or might not share the same pharmacological profile and we fully agree with this notion.
We do not think that further arguments against dominant expression of GPR37L1 in astrocytes will lead anywhere. Not only our data (which come from the rat, this needs to be kept in mind) show that GPR37L1 is, by far, the dominant receptor of the two in astrocytes. This observation is consistent with the “gold standard” astrocytic transcriptomes from Barres, Khakh, Sofroniev labs, and several other datasets available to us. There may, however, exist substantial regional differences in the levels of GPR37L1 in astrocytes from rat brain and, possibly age dependent changes, such as with almost any other gene found in astrocytes. In rodents, GPR37L1 is found in oligodendrocytes as well, but in humans expression is highly selective to astrocytes (Zhang et al 2016 data). We may benefit from further, unbiased experiments with well validated tools to clarify this issue once and for all. Perhaps, the most interesting issue is GPR37L1 expression in man. At least in glioblastomas which bear the closest resemblance of astrocytic lineage, GPR37L1 is definitely the dominant one: (http://cgga.org.cn:9091/gliomasdb/index.jsp).
With this in mind, we find it much more likely that GPR37L1 is the key player when we show that the specific effect of prosaptide in serum-free (or PSAP-depleted) conditions is abolished in double knock-down experiments (Liu et al 2018). Since now the discussion is focusing around possible differences between GPR37L1 and GPR37, given the resources, it would be an important next step to clarify the relative contributions of the two receptors. These are straightforward experiments which can be done by knocking down one or the other receptor, but it was not the point of our study.
We think that our evidence should put to rest the argument that has caused the field to run in circles: We confirm the earlier conclusions by Meyer et al (2013) showing pairing of PSAP/prosaptide and GPR37L1 (possibly also GPR37) in astrocytes. In the recent study by Jolly et al PM:28795439, figure 5g is a striking demonstration of an effect of prosaptide on astrocytes which completely disappears in GPR37L1 knock-outs. Interestingly, some other actions remain and to us this can only mean that there is more than one target and more than one mechanism of these effects.
It is a fact that other groups, using different cellular models, have not been able to replicate the ability of prosaptide or PSAP to activate GPR37 (and GPR37L1, where it was tested), hence casting doubt on the validity of the earlier data – but not exactly a situation conducive to further investigation. We would like to emphasise here some confounding factors that may prevent the detection of GPR37L1 (and/or GPR37) activation by PSAP or prosaptide under certain experimental conditions.
1. The receptors’ intracellular signal transduction appears to be intact in primary astrocytes but not, or at least not reliably reproducible, in many cell lines that are commonly used for expression of recombinant receptors and screening, e.g. HEK293. In our hands, HEK293 cells don’t support GPR37L1/GPR37 activation. Even in the original report (Meyer et al 2013), the difference between the magnitude of effect in HEK293 cells and astrocytes is striking – it is the astrocytes (Fig 4) where one can really see PSAP in action. We discussed this issue with Prof Hall and there may well be culture condition or clonal differences between HEK293 cells involved. Irrespective, the cell model with intact coupling, and therefore the one to use, is the astrocyte. One of the points raised by Prof Smith relates to the use of β-arrestin-recruitment based assays. We are rather surprised by this criticism given that recruitment of β-arresting is a canonical property of Class A GPCR described in every review and textbook, and GPR37L1 and GPR37 clearly map to Class A. In fact, postulation of the interaction with β-arrestin is the basis for several screening platforms, including the MRC Technologies (Discover X) orphan panel and PRESTO-TANGO (PM:25895059). Indeed, in the 2015 review PM:26635605 Prof Smith wrote “Third, like HA, TX14A was also included in the MRC Technologies β-arrestin-based orphan GPCR screen but was not detected as a ligand at either GPR37 or GPR37L1(Southern et al., 2013)”. Therefore, our experiments were not entirely ungrounded. If GPR37L1 and GPR37 don’t respond to prosaptide in β-arrestin-based assays, two of the possible reasons may be: a) GPR37L1 and GPR37 actually can’t recruit arrestins, after all, and that would be something really special about these receptors, or, b) an essential signalling co-factor or binding partner is missing in the cell line used for the assay. Neither rules out that the prosaptide-GPR37L1 pairing is valid in their native cellular background. Therefore, negative data in these screening platforms should not be taken as conclusive proof of the lack of pairing.
2. There is one other, very important feature of the prosaposin signalling system which may complicate interpretation of the experiments. This is the presence of the ligand (Saposin C) in culture media even in nominally “serum-free” or prosaposin-free conditions. In our experiments we detected condition-dependent accumulation of immunopositive PSAP signal in PSAP-depleted media in co-cultures of astrocytes and neurones. Evidently, it originated from the cells themselves which is not surprising, given that PSAP is a lysosomal protein. Perhaps also in the brain slices there is some prosaposin. This means that under many assay conditions, the ligand is building up and likely to constitutively stimulate or may be down-regulate GPR37L1 and possibly GPR37 under ‘control’ conditions. We think this is an important issue and needs further attention.
3. In case GPR37 and GPR37L1 do share the same pharmacological profile and activate the same signalling cascades, they may functionally compensate for each other in single knock-out approaches. That this may be the case could be suggested by the generally mild or variable phenotypes of single but drastic consequences in double knock-out scenarios. We also would like to remind that drawing far reaching conclusions about the physiological roles of the individual proteins based on observations in knock-out mice may sometimes produce rather surprising results. Those who would like a confirmation, could obtain a colony of viable acetylcholinesterase knock-out mice (https://www.jax.org/strain/005987). It is that same enzyme which is inhibited by war gases which kill humans and animals within seconds to minutes.
Moving forward from PSAP-GPR37/L1 signalling in astrocytes, one of the next most pertinent tasks is to further characterise the pharmacological profile and intracellular signalling of each of the two receptors. Here, the recent discovery of a new ligand of GPR37 – NPD1 – is of great interest (Bang et al 2018). Of note, that study confirms (Fig 3C) the activity of prosaptide, as well, at least for GPR37-evoked phagocytosis in vitro. The exact nature of the intracellular signalling elicited by NPD1 in HEK293 cells and in macrophages, however, appears quite different from GPR37L1 activation in astrocytes, and would be highly interesting to follow up: Bang et al suggest Gi-protein-dependent intracellular Ca2+ release (pertussis toxin- and thapsigargin-sensitive) as trigger for anti-inflammatory action; PI3K/AKT and ERK pathways are also implied but how these typically tyrosine kinase receptor activated pathways link with GPR37, and prototypical class A GPCR signalling, remains to be clarified. Notably, the paper reports that any Ca2+ increase, even when triggered by a non-selective agent such as ionomycin, mimics the effect of NPD1-GPR37 interaction in macrophages. It would be important to determine, in macrophages, whether NPD1 also elicits the expected decrease in cAMP in response to Gi-protein activation and also, whether NPD1 activates GPR37L1-dependent signalling in astrocytes.
But the final big pressing question – to us – takes us back to the powerful neuroprotection that PSAP-GPR37L1 signalling in astrocytes mediates. It seems that already in 1997 O’Brien and his colleagues were just one step away from the identification of the still elusive receptor: PM:9388493. Neuroprotection by PSAP, prosaptide and related short peptides has been seen, since 1994 (!), in too many studies by several laboratories, including ours – PM:7937812, followed by PM:7980569, PM:7768361, PM:8780031, PM:8780053, PM:9553960, PM:9748612, PM:10412024, PM:10400252, PM:10454138, PM:11046216, PM:10852246, PM:10773009, PM:10643816, PM:11356264, PM:11702044, PM:17466547, PM:25993033, PM:27372641, PM:28795439, PM:30260505.
It just cannot to be further ignored as some kind of artefact or curiosity. As soon as we understand this mechanism, we will open up an avenue for working towards a novel family of neuroprotective drugs.