Construction of a Global Pain Systems Network Highlights Phospholipid Signaling as a Regulator of Heat Nociception

The ability to perceive noxious stimuli is critical for an animal's survival in the face of environmental danger, and thus pain perception is likely to be under stringent evolutionary pressure. Using a neuronal-specific RNAi knock-down strategy in adult Drosophila, we recently completed a genome-wide functional annotation of heat nociception that allowed us to identify α2δ3 as a novel pain gene. Here we report construction of an evolutionary-conserved, system-level, global molecular pain network map. Our systems map is markedly enriched for multiple genes associated with human pain and predicts a plethora of novel candidate pain pathways. One central node of this pain network is phospholipid signaling, which has been implicated before in pain processing. To further investigate the role of phospholipid signaling in mammalian heat pain perception, we analysed the phenotype of PIP5Kα and PI3Kγ mutant mice. Intriguingly, both of these mice exhibit pronounced hypersensitivity to noxious heat and capsaicin-induced pain, which directly mapped through PI3Kγ kinase-dead knock-in mice to PI3Kγ lipid kinase activity. Using single primary sensory neuron recording, PI3Kγ function was mechanistically linked to a negative regulation of TRPV1 channel transduction. Our data provide a systems map for heat nociception and reinforces the extraordinary conservation of molecular mechanisms of nociception across different species.


Introduction
Although studies in inbred mouse strains and in human twin cohorts have indicated that pain has a strong genetic component [1][2][3][4][5], with an estimated heritability of ,50%, little is known about the specific genes involved in regulating pain sensitivity across phyla. Further, conservation of gene function between species and across evolutionary time acts as a useful tool to develop an understanding of core genetic mechanisms relative to more specialized programs and how they influence behavior [6].
Drosophila is an excellent model organism for characterizing genetic regulators of behavior such as nociception [7]. Use of Drosophila genetics has highlighted a conserved role for multiple genes in the detection and avoidance of noxious heat [8,9], and recent work on mechanosensation suggests the genetics of this process is likely also highly conserved across phyla [10]. We have previously reported a global in vivo RNAi screen for avoidance of noxious heat in Drosophila, and identification of hundreds of novel genes required in the adult fly, for its manifestation [8]. To interrogate this resource, we have now constructed a global systems network of heat pain. Our goal was to identify potentially conserved genes and pathways involved in pain perception, to provide a tool for focusing research on key pain molecular pathways.

Author Summary
Nociception is the perception of noxious, potentially damaging stimuli; and this pain or its equivalent behavioral readout is evolutionarily conserved from fruit flies to humans. Using genetic techniques in the fruit fly, we have been able to evaluate the potential functional contribution of every gene in the fruit fly genome for a role in avoidance of high noxious temperatures (heat pain-like responses). Using this functional genomics data set, we have developed a conserved network map of heat pain/ nociception that predicts numerous conserved genes and pathways as novel pain pathways, including phospholipid signaling. Studies in multiple mutant mice confirmed a role for lipid signaling in pain perception, and more specifically we identify the critical lipid kinase (PI3Kc) as a negative regulator of TRPV1 (receptor for noxious heat and capsaicin, the active component in chili peppers) signaling. This finding shows that our fly-based genetic pain network map is a valuable tool for the discovery of novel ''nociception genes'' in mammals. Figure 1. A global network map of thermal nociception. The systems network includes data from significantly enriched Drosophila KEGG pathways and GO processes, mouse and human KEGG pathways and C2 gene sets. Pathways, processes and gene sets that share a role in a biological process were pooled into functional classes while the underlying genes that constitute them are depicted with a connection to their respective functional class. Functional classes (gold), genes representing direct hits with a thermal nociception phenotype (red), their first degree binding partners (green), and developmental lethal genes (blue) represent the nodes in the network. Only select KEGG pathways, biological processes and C2 gene sets were used to build systems map. For the entire list of individual pathways, gene sets, and processes see Tables S5, S6,

Results/Discussion
To construct a global systems network of heat pain we first identified potential mouse and human orthologs of fly candidate pain genes ( Figure S1). Of the 580 candidate fly thermal nociception genes we had previously identified [8], 399 had human orthologs (Table S1), many of which are known mammalian pain genes (Table S2). Gene ontology (GO) analyses of the human and mouse orthologs of the fly thermal nociception hits showed a marked enrichment of genes involved in neurotransmission and secretion, housekeeping systems such as mitochondrial structure, ATP synthesis, metabolism, or calcium signaling (Figure S2A-S2C and Table S3). We next generated an interaction map based on first-degree binding partners for the fly thermal nociception hits (Table S4). All binding partners were identified in yeast-2-hybrid screens and reported in the biomolecular interaction network database BIND, i.e. binding partners experimentally confirmed to interact with the candidate genes. Among the first degree binding partners in this interaction network, we found fly allatostatin C receptor 1, which has homology with mammalian opioid receptors, a fly homolog of Lmx1b, which regulates central serotonergic responses to opioids [25], a fly gene related to the nuclear factor-erythroid 2-related factor 2, which has anti-nociceptive effects [26] by inducing upregulation of heme oxygenase-1, and tyrosine hydroxylase, the enzyme required for dopamine and catecholamine production [27]. Thus, our genome-wide functional screen for thermal nociception in flies has generated a human gene network that includes orthologs of several known mammalian pain genes, in addition to numerous uncharacterized genes and pathway not previously associated with pain perception.
To construct a mammalian systems map of thermal nociception, we performed an enrichment analysis (using KEGG pathways and Broad Institute C2 gene sets) on the mouse and human orthologs of the fly pain genes and their first degree binding partners ( Figure  S1; Table S5 and Table S6). We found significant enrichment of genes (hypergeometric enrichment .90%) involved in mitochondria, metabolism, calcium signaling, inflammation, cell adhesion, RNA processing, and neurotransmission. Finally, to generate a comprehensive conserved network map of thermal nociception, the KEGG pathways from Drosophila, mouse and human were combined with relevant gene sets from the C2 annotations to create a global putative ''nociception network'' ( Figure 1; Table  S7). From this combined systems network, we identified gene sets or pathways known to play key roles in many major neural systems.
The connectivity of the entire systems map remains intact even after omitting all binding partners ( Figure S3), i.e., expansion of the ''pain'' network map by including binding partners does not introduce a bias. Moreover, all except one pathway in the network map has .50% representation from the Drosophila heat pain hits alone. To test whether this approach can predict mammalian pain genes, we measured the overlap of the direct candidate pain hits and their binding partners with previous rat microarray expression profiling data generated from pain studies [28] and all ''pain'' annotated genes from OMIM (Online Mendelian Inheritance in Man, NCBI). Intriguingly, we found a 38.55% overlap of our direct pain hits and a 42.33% overlap of their binding partners with the microarray and OMIM data for pain. In contrast, 100 random gene lists gave a maximum of 6% overlapping genes and a minimum of 0% (Table 1; Table S8). Thus, our hypothesis-free systems map is markedly enriched for genes known to be associated with rodent and human pain. Considering the complexity of the neuronal network involved in generating Table 1. Comparison of systems map with microarray and OMIM data for pain.

Genes in system map
Genes matching with pain report in OMIM Genes matching with pain report in Microarray data A comparison of genes in the global network map of thermal nociception with genes with a pain annotation in the OMIM database (http://www.ncbi.nlm.nih.gov/omim) and previous rat microarray expression profiling data generated from pain studies [28] showed the global network map of thermal nociception is enriched for genes implicated in mammalian pain diseases. doi:10.1371/journal.pgen.1003071.t001 nociception in the periphery and CNS, these pathways may operate across several neurons; however, our genome-wide functional fly pain screen and in silico data mining provide a road map for conserved molecular components and pathways putatively involved in heat nociception globally across phyla. We next wanted to validate whether this ''nociception network'' has the power to identify conserved pathways involved in nociception, and if this pathway information can then help to pinpoint key mammalian pain genes. To this end we focused our first efforts on phosphatidylinositol signaling, one of the major nodes predicted from the pain systems map, and a heat pain precedented pathway. Phosphatidylinositol signaling has been implicated in heat nociception and regulation of TRPV1 by multiple groups [11,13,16,18], however its precise role has been controversial [16,18,23], and the specific participation of different phospholipid kinases has never been evaluated genetically, which we now decided to do.
Phosphatidylinositol signaling involves the generation of PIP2 via PIP5K, and then phosphorylation of PIP2 via PI3K to generate PIP3. PIP5Ka is highly expressed in the nervous system but no neuronal function for this kinase has been established [29]. PIP5Ka mutant mice are viable and exhibit an exaggerated anaphylactic immune reaction in response to Fc-receptor engagement [30]. We find that PIP5Ka mutant mice exhibit a significant hyper-responsiveness to radiant heat ( Figure 2A) and contact heat, when compared to littermate controls ( Figure 2B). In mammals, TRPV1 is the prototypical noxious heat thermo-receptor, and is also the receptor for capsaicin, the active ingredient in chili peppers [31]. We therefore tested whether PIP5Ka mutant mice exhibit exaggerated TRPV1 agonist responses. Indeed, following capsaicin injection, PIP5Ka mutant mice display heightened reactivity compared to littermate controls ( Figure 2C) but exhibited no difference in mechanical pain threshold using the von Frey test ( Figure 2D).
PI3Kc, the only G-protein coupled PI3K, is expressed in TRPV1-positive peripheral sensory neurons in both rats and mice [32,33] and has been implicated in morphine-induced peripheral analgesia [32] and morphine tolerance [33]. Use of non-specific inhibitors, like wortmannin, has suggested a role for PI3' kinases in producing NGF-mediated TRPV1 sensitization [21], [34]. We therefore tested thermal nociception in PI3Kc (p110c) mutant mice [35] and found that these mice, like the PIP5Ka null mice, also exhibit an exaggerated behavioral response to radiant heat plantar stimulation using the Hargreaves test ( Figure 3A). This enhanced pain sensitivity was confirmed using the hot plate assay ( Figure 3B). PI3Kc mutant mice also exhibit an enhanced pain response to a capsaicin challenge ( Figure 3C). Similar to PIP5Ka null mice, the mechanical pain threshold using the von Frey test ( Figure 3D), and the behavioral responses to acetone application (a cooling sensation) ( Figure 3E) were comparable between control and PI3Kc 2/2 littermates. Of note we found a similar thermal hyperalgesia phenotype in a second independent PI3Kc 2/2 mutant mouse strain [36] (not shown). Thus, genetic loss of PI3Kc and PIP5Ka results in enhanced pain responses to heat and capsaicin, providing evidence that phosphatidylinositol signaling acts as a negative regulator of heat pain perception and TRPV1 reactivity in vivo.
Since PIP5Ka and PI3Kc cooperate to sequentially generate PIP3, and both mutant mice exhibit a similar hypersensitivity phenotype, we focused further on the role of PI3Kc and PIP3 generation in setting the threshold for heat pain perception. PI3Kc is highly expressed in haematopoietic cells and functions as key mediator of inflammatory cell migration to the site of injury [35,37]. We therefore tested PI3Kc mutant mice for potential defects in inflammation-induced pain sensitization, i.e. thermal hyperalgesia. PI3Kc 2/2 and control mice developed comparable levels of thermal hyperalgesia ( Figure S4A and S4B) following plantar CFA injection. CFA-induced inflammation, as determined by paw swelling, was also comparable between mutant and control mice ( Figure S4C). To further exclude a potential role of haematopoietic cells, we transplanted wild type bone marrow into PI3Kc mutant mice (WTRKO) and PI3Kc mutant bone marrow into wild type mice (KORWT). The presence of a wild type haematopoietic system did not rescue the enhanced sensitivity to thermal pain in the PI3Kc mutant background ( Figure 3F), i.e. the requirement for PI3Kc in thermal sensing maps to non-haematopoeitic cells. PI3Kc has been shown to act both in a kinase-dependent fashion, through conversion of PIP2 to PIP3, and in a kinase-independent manner [38]. We therefore tested the behavioral response of PI3Kc kinasedead (KD) knock-in mice. PI3Kc KD mice exhibited a heightened reaction to noxious heat with a reduced thermal nociception latency ( Figure 3G), comparable to the enhanced heat pain responses observed in complete PI3Kc mutant mice. Thus, the kinase activity generating PIP3 modulates heat pain. We also assessed the general neurological phenotypes of PI3Kc 2/2 mice, all of which appeared normal ( Figure 3H; Figure S5A-S5G). Furthermore, the overall morphology and histology of the central nervous system appeared normal in PI3Kc 2/2 mice. These data demonstrate that generation of PIP3 through PI3Kc negatively regulates pain sensitivity in vivo.
To test if the phosphatidylinositol signaling pathway acts in primary sensory nociceptors, we employed electrophysiology on isolated wild type and PI3Kc 2/2 dorsal root ganglion (DRG) neurons. PI3Kc 2/2 DRG neurons responded to a thermal ramp ( Figure 4A) with a significantly increased steepness in the inward current response to increasing temperature when compared to control neurons. This translated into a substantial increase in the Q10 value ( Figure 4B), a measure of temperature-dependent rate change in channel conductivity, indicating that PI3Kc 2/2 DRG cells exhibit massive hyper-activation in response to noxious heat, although initiation of this response occurs at a slightly elevated temperature (44.77uC for PI3Kc 2/2 vs 42.22uC for wild type DRG neurons, Figure 4A). Since we also observed an enhanced response to capsaicin in PI3Kc 2/2 mice in vivo, we directly tested TRPV1 reactivity to capsaicin in sensory neurons in vitro. In accordance with our behavioral data, isolated PI3Kc 2/2 DRG neurons exhibited augmented sensitivity to capsaicin (Figure 4C Our data provide a conserved functional systems network map for pain behaviour. This network map revealed many pathways and gene sets previously reported to be involved in mammalian nociception, including multiple genes annotated as candidate pain genes in the human OMIM database. Thus, our systems approach, starting from a functional whole genome fly screen and bioinformatic construction of a conserved pain network map, has the power to identify regulators of mammalian nociception. Our network pointed to a key role for phosphatidylinositol signaling in noxious heat nociception. Positive as well as negative regulatory functions of phosphatidylinositol signaling on the thermal nociceptive sensor TRPV1 have been reported [15,16,19,20,23,24]. Our results provide genetic data that the phosphatidylinositol signaling pathway is relevant to heat pain sensitivity in vivo. In particular, we find that the lipid kinases PIP5a and PI3Kc are involved in regulating heat nociception responses by acting as negative modulators of thermal pain perception and TRPV1 activity. Our data reinforce the extraordinary evolutionary conservation of the neurobiological mechanisms of nociception, from its manifestation as an acute damage avoidance response in simple organisms like flies to the complex sensation of pain in mammals. When used in conjunction with additional complimentary approaches (e.g. published literature, gene expression profiling, or genetic association studies), this systems network map should be a valuable tool to further pinpoint and prioritize novel candidate nociception genes in mammals.

Bioinformatics analysis
Identification of fly orthologs in mouse and human was done using pre-computed orthology predictions [39]. Gene Ontology (GO) analysis was performed using GOstat. Binding partner identification was done using GeneSpring GX. Hypergeometric tests were used to identify over-represented gene lists (BROAD Institute) and pathways (KEGG) amongst the pain hits and to generate a conserved systems map. Pain genes and binding partners in the system map that have been annotated as pain genes in the Online Mendelian Inheritance in Man database or by our previous Microarray experiments [28] were also identified.

Mouse behavioral tests
PI3Kc (p110c) knock out [35,36], kinase dead PI3Kc knock-in [38], and PIP5Ka mutant mice [30] and have been previously described. Thermal and mechanical sensitivities were assessed using the Hargreaves, hot plate, and von Frey tests. Capsaicin behavior was assessed over 5 minutes following intraplantar injection of capsaicin.

Single DRG neuron recordings
Lumbar dorsal root ganglia (DRG) were harvested as previously reported [40,41]. Patch-clamp recordings were performed using the whole-cell voltage-clamp configuration of the patch-clamp technique as previously described [40,41].

Ethics statement
All mice were bred and maintained according to an ethical animal license protocol complying with the current Austrian law.
Detailed Materials and Methods are available in Text S1. Drosophila thermal nociception hits. GO terms were pooled into functional categories for data representation. Numbers indicate the counts of GO terms included in each functional category. These GO terms are grouped according to their parent termscellular components (CC), biological processes (BP), and molecular functions (MF). (C) Global C2 data set for mammalian orthologs. Functional classification of C2 gene sets (MsigDb, Broad institute) found enriched in mouse and human orthologs of our primary fly candidate thermal nociception genes and their first degree binding partners. The numbers indicate statistically significant C2 genes sets grouped into each functional category. (PDF) Figure S3 A global network map of thermal nociception based on primary hits without binding partners. The systems network includes data from the significantly enriched Drosophila KEGG pathways and GO processes, mouse and human KEGG pathways and C2 gene sets based on analysis of primary screening hits without binding partners. Pathways, processes and gene sets that share a role in a biological process were pooled into functional classes while the underlying genes that constitute them are depicted with a connection to their respective functional class. In all experiments there we no significant differences (t-test). (PDF)

Supporting Information
Table S1 Predicted mammalian orthologs for Drosophila pain genes and binding partners. To identify orthologs between Drosophila and mouse or Drosophila and human, we used precomputed orthology predictions obtained from Compara r49, Homologene (03/08), Inparanoid v6.1, Orthomcl v2 [39]. Both One-to-one and many to many mappings were observed between the Drosophila and the mammalian genes.
(XLS) , gene count, total gene count, and P value have been included. Since GO terms that lie at level 4 or below in the GO hierarchy tree convey more biological information, we excluded terms with more than 500 genes. Further, if both parent and several child terms are found significant, we manually curated the data and retained the term that contained the maximum overlapping genes with the pain hit list. These terms are highlighted in the mouse and the human lists separately. From these selected GO terms, those that denote related functions in biological systems were manually clubbed together and functionally annotated. This analysis is included in the column labeled ''Manually assigned functional category''. A P-value of 0.1 was considered significant. ''Count'' and ''Total'' indicate the numbers of hits among all genes assigned to a defined GO term. (XLS)

Table S4
Interactions between candidate pain genes and first degree binding partners. All interactions involving mammalian orthologs of Drosophila pain hit genes and their first degree binding partners are listed. Each row corresponds to a biological interaction between two proteins. Every interaction has two participating nodes: Source and Target. Gene symbols and CG IDs for all source and target nodes are listed. Classifications of all source and target nodes into pain hits (pain), binding partners (BP), and developmentally lethal phenotypes are provided. Potential mosue and human orthologs are presented for hits and binding partners. Both One-to-one and many to many mappings were observed between the Drosophila and the mammalian genes.
(XLS)  Figure 1 and Table S7.   Figure 1. Selected significant KEGG pathways and C2 gene sets were manually grouped into uniform functional categories as shown in the column ''Gene set categories''. Gene symbols, Entrez IDs and Drosophila CG IDs were extracted for each pathway or gene set, as appropriate. (B) Percentage of original pain hits and lethals that contribute to each of the significant enrichment of each gene set or pathway. (XLS)

Table S8
Overlap of pain systems map with pain OMIM and microarray data. A comparison of genes in the global network map of thermal nociception with genes with a pain annotation in the OMIM database (http://www.ncbi.nlm.nih.gov/omim) and previous rat microarray expression profiling data generated from pain studies [28] showed the global network map of thermal nociception is enriched for genes implicated in mammalian pain. Included are the Drosophila gene CGID, mammalian gene symbol, the genes status as a direct hit (pain) or predicted binding partner (BP), and if this gene was identified as significantly regulated in the dorsal horn (DH) or dorsal root ganglia (DRG) following experimentally induced chronic pain. Genes with a pain annotation in the OMIM database are also identified.

(XLS)
Text S1 This file contains Detailed Materials and Methods and Supporting References. (DOCX)