The gene regulatory network in different brain regions of neuropathic pain mouse models

Neuropathic pain is the direct result caused by lesions or somatosensory nervous system diseases that are associated with emotional regulation. The incidence of neuropathic pain in the general population is 7-10% and the mechanisms of neuropathic pain are largely unknown. It is often related to structural and functional abnormalities in multiple brain regions. The forebrain, including nucleus accumbens (NAc), medial prefrontal cortex (mPFC) and periaqueductal gray (PAG) have been shown to correspond with the regulation of neuropathic pain. To investigate the molecular mechanism of neuropathic pain across different brain regions, we identified the differentially expressed genes between the spared nerve injury model (SNI) mice of neuropathic pain and the control Sham mice in NAc, mPFC and PAG and mapped these genes onto comprehensive functional association network. With Random Walk with Restart (RWR) analysis, we identified more novel neuropathic pain genes in NAc, mPFC and PAG, such as Asic3, Cd200r1 and MT2, beside well known Capn11 and CYP2E1. What’s more, we discovered their interactions or cross talks. Our results provided novel insights of neuropathic pain and provided therapeutic targets for treating neuropathic pain.

The gene regulatory network in different brain regions of neuropathic pain mouse models CURRENT  Currently, recommended treatments for neuropathic pain is pharmacological, such as the 3 use of antidepressants, anticonvulsants and topical anesthetics 7,8 . In some cases, however, medical therapy alone cannot fully control chronic pain. Some nonpharmacological approaches, including psychological approaches, physical therapy, interventional therapy and surgical procedures have been shown to be effective for neuropathic pain 8 . In addition, the distinction between nociceptive and neuropathic pain is also important because different treatment methods are usually required for different types of pain.
Chronic pain is often related to structural and functional abnormalities in the brain 9,10 .
Previous studies have shown that an increase in activity of the forebrain neurons results in enhanced inflammatory and neuropathic pain 11,12 . The forebrain, including nucleus accumbens (NAc), medial prefrontal cortex (mPFC) and paraventricular nucleus (PVN) have been shown to correspond with the regulation of neuropathic pain 13,14 . Nucleus accumbens (NAc) is known to be related to emotional dysfunctions following neuropathic pain regulation as a key component of the brain reward system [15][16][17]  Patients with chronic pain also exhibit brain abnormalities in descending modulation of pain 25,26 , especially in the periaqueductal gray (PAG), which may be associated with dysfunctions of pain regulation 27-29 . Neuropathic pain activates neurons in the periaqueductal gray (PAG), the neurons projecting to the rostral ventromedial medulla (RVM) and then projected to the spinal cord to inhibit or facilitate the pain 30 .
Since the brain region corresponding to chronic pain is between the ventromedial PFC and PAG in humans 31 , and mPFC-PAG in rodents 32 , we would like to identify the genes that repose to neuropathic pain and investigate the molecular mechanisms of neuropathic pain. The differentially expressed genes in the nucleus accumbens (NAc), the medial prefrontal cortex (mPFC), and the periaqueductal grey (PAG) of the spared nerve injury model (SNI) were mapped onto gene regulatory network. Novel neuropathic pain gene in different brain regions were identified using Random Walk with Restart (RWR) algorithm and their interactions or cross talks were analyzed.

Methods
The differentially expressed genes between SNI and Sham mice in NAc, mPFC and PAG The network expansion of NAc, mPFC and PAG neuropathic pain genes based on RWR analysis As we mentioned before, we would like to investigate the interactions or cross-talks among different brain regions that were responsive to neuropathic pain. The differentially expressed genes between SNI mice and Sham mice in NAc, mPFC and PAG were a good start for such network analysis. Therefore, we mapped these three gene lists onto the comprehensive functional association network of STRING 34 , a widely used network for bioinformatics studies [35][36][37][38] . Only the high confidence interactions of STRING were included, in other words, the confidence score of the interaction must be greater than 0.900.
To explore the cross talk between brain regions, we applied Random Walk with Restart (RWR) algorithm 35,[39][40][41][42] . To illustrate how RWR can reveal the cross talk, let us denote the STRING network as a graph comprised of a set of genes and a set of interactions .
The whole interaction network can be represented with an adjacency matrix. The total number of genes was . The value in row and column was 1 if gene and gene had interactions and was 0 if they did not interact.
(1) Normalization. The adjacency matrix will be column-wise normalized Due to technical limitations, Equation 1 has been placed in the Supplementary Files section.

6
(2) Iteration. Then, a rank walk step will be iterated. In each round of iteration, the state where was previous state probabilities at time , was the restart probability and was the initial state probabilities which was a column vector with for the seed genes (NAc, mPFC, PAG neuropathic pain genes, respectively) and to 0 for other genes on the network.
(3) Converge. The iteration process will stop when the difference between two states was smaller than 1×10 -6 .
After the RWR analysis, each gene on the network will be given a probability of being visited by the seed genes.
The NAc, mPFC and PAG neuropathic pain genes were considered as seed genes, respectively.
To evaluate how significant the probability was, we randomly chose the same number of seed genes 1000 times and calculated the RWR probabilities. If there were more than 50 times that the permutation proverbialities were greater than the actual proverbiality, the permutation p value for that gene was greater than 50/1000=0.05 and that gene will be excluded.
With the permutation p value, we identified the novel neuropathic pain genes in NAc, mPFC and PAG based on RWR analysis. Such expanded NAc, mPFC and PAG neuropathic pain genes can be overlapped to show the cross talk among NAc, mPFC and PAG under neuropathic pain.

Results
The neuropathic pain genes in NAc, mPFC and PAG identified by differential expression analysis The gene expression profiles of NAc, mPFC and PAG in SNI and Sham mice were analyzed and the differentially expressed genes between SNI and Sham mice in NAc, mPFC and PAG were identified. There were 123, 89 and 795 differentially expressed genes in NAc, mPFC and PAG, respectively.
We compared these three differentially expressed gene lists (Table S1) and plotted their Venn Diagram in Figure 1. It can be seen that only two genes, Capn11 and Cyp2e1, were overlapped. Obviously, these two genes played important roles in neuropathic pain.
Capn11 (Calpain 11) encodes an intracellular calcium-dependent cysteine protease that has protease activity and calcium-binding capacity 43 . Calpains were reported to participate in some neuronal processes, including synaptic plasticity, neurodegeneration, signal transduction and enhancement [44][45][46] . The expression of calpains can be observed in several cells types in the central nervous system (CNS), such as spinal cord neurons, cortical neurons, and glial cells 47,48 . According to previous studies, the activities of calpains were markedly increased in neurodegenerative diseases 49 , traumatic brain injury 50 and neuropathic pain 51  As shown above, these two genes functions through complex pathways and regulatory mechanisms, there were many missing genes that facilitate the neuropathic pain responses in different brain regions. To find these hidden genes, we mapped these differentially expressed genes onto functional association network of STRING.
The novel neuropathic pain genes in NAc, mPFC and PAG identified by RWR analysis on the functional association network To identify more novel neuropathic pain genes in NAc, mPFC and PAG and find their hidden links or cross talks, we mapped the differentially expressed genes onto network and performed RWR analysis on the network. By considering the differentially expressed genes in NAc, mPFC and PAG as seed genes and permuting them 1000 times, we identified the significant novel neuropathic pain genes in NAc, mPFC and PAG with permutating p value smaller than 0.05. There were 623, 888 and 507 novel neuropathic pain genes in NAc, mPFC and PAG, respectively. These novel neuropathic pain genes in NAc, mPFC and 9 PAG were given in Table S2.
The Venn Diagram among these novel neuropathic pain genes in NAc, mPFC and PAG on the network was shown in Figure 2. There were 25 overlapped genes and they were shown in Table 1. These 25 overlapped genes showed great promise in linking the three brain regions and revealing the potential cross talk mechanisms among NAc, mPFC and PAG for neuropathic pain. We will discuss their functions in the next section.

Discussion
Among the 25 genes in Table 1, many of them were involved in pathways or functions associated with neuropathic pain genes. The following three genes were most promising.
Asic3 (Acid-sensing ion channels, ASICs) was cationic channel expressed principally in central (CNS) and peripheral (PNS) nervous systems 64,65 . Ion channel modulation is a main approach to achieve novel neuropathic pain management 66 . Evidence from many experiments have suggested the involvement of ASICs in pain sensation 66,67 . Among the ASICs, ASIC3 is known to regulate inflammatory pain, ischemic pain and mechanical pain 64,68 . Inflammation is one of the pain symptoms that induces a significant increase of ASIC3 channel expression in sensory neurons, which demonstrate the crucial role of ASIC3 in the generation of pain associated with inflammation 69 . Therefore, inhibition of ASIC3 channel at the sensory system could obviously help to alleviate pain. In addition, Jeong et al. suggested that ASIC3 may be associated with the antinociceptive effects of amiloride and benzamil, inhibitors for ASIC channels 70,71 , in neuropathic pain and blocking ASIC3 channel may be a novel therapeutic strategy in neuropathic pain treatment 72 .
Cd200r1 encodes a membrane glycoprotein of the immunoglobulin superfamily that is highly expressed on neurons in the central nervous system while its receptor CD200R is restricted to the surfaces of myeloid lineage cells like macrophages and microglia 73,74 .
The CD200-CD200R interaction has been reported to be closely associated with the macrophage-mediated damage in autoimmune disease and various neuroinflammatory diseases [75][76][77][78] . Animal models have also shown that loss of immunosuppression through CD200 has significant impact on neuroinflammation and neurodegeneration 79,80 .
Hernangomez et al. 80 reported that the CD200/CD200R regulatory system can suppress the neuroinflammatory reactions associated with peripheral neuropathic pain.
CD200/CD200R may be a target for treating neuropathic pain.
MT2 (Metallothioneins II) is a major neuroprotective protein with a high affinity for metals 81,82 . MT2 has been found in many CNS (central nervous system) regions 83,84 , such as cortex, hippocampus, brainstem and spinal cord 85

Conclusions
As a common nervous system disease with an incidence of 7-10% in the general population, the mechanisms of neuropathic pain are largely unknown. It is a complex disease involving the structural and functional abnormalities in multiple brain regions. The forebrain, including nucleus accumbens (NAc), medial prefrontal cortex (mPFC) and periaqueductal gray (PAG), all correspond to the response of neuropathic pain. To investigate the molecular mechanism of neuropathic pain across different brain regions, we identified the differentially expressed genes between SNI mice which was a widely used model for neuropathic pain and the Sham mice which was used as control. The differentially expressed genes in NAc, mPFC and PAG were mapped onto STRING network.
Using Random Walk with Restart (RWR) analysis, more novel neuropathic pain genes in NAc, mPFC and PAG were revealed based on network structure and more overlapped genes among them had emerged. These overlapped novel neuropathic pain genes in NAc, mPFC and PAG can help us understand how different brain regions communicate with each other and coordinate the regulation of neuropathic pain. These genes worth to be further validated and investigated as therapeutic target. Declarations

Ethics approval and consent to participate
This article does not contain any studies with animals and human performed by any of the authors.

Consent to publish
Not applicable.

Availability of data and materials
The data and materials in the current study are available from the corresponding author on reasonable request.

Competing Interests
None.   Figure 1 The Venn Diagram of differentially expressed genes in NAc, mPFC and PAG Among the 123, 89 and 795 differentially expressed genes in NAc, mPFC and PAG, only two genes, Capn11 and Cyp2e1, were overlapped. These two genes played important roles in neuropathic pain, but there were many undiscovered neuropathic pain genes in the differentially expression analysis. TableS1.xlsx