G9a participates in nerve injury-induced Kcna2 downregulation in primary sensory neurons

Nerve injury-induced downregulation of voltage-gated potassium channel subunit Kcna2 in the dorsal root ganglion (DRG) is critical for DRG neuronal excitability and neuropathic pain genesis. However, how nerve injury causes this downregulation is still elusive. Euchromatic histone-lysine N-methyltransferase 2, also known as G9a, methylates histone H3 on lysine residue 9 to predominantly produce a dynamic histone dimethylation, resulting in condensed chromatin and gene transcriptional repression. We showed here that blocking nerve injury-induced increase in G9a rescued Kcna2 mRNA and protein expression in the axotomized DRG and attenuated the development of nerve injury-induced pain hypersensitivity. Mimicking this increase decreased Kcna2 mRNA and protein expression, reduced Kv current, and increased excitability in the DRG neurons and led to spinal cord central sensitization and neuropathic pain-like symptoms. G9a mRNA is co-localized with Kcna2 mRNA in the DRG neurons. These findings indicate that G9a contributes to neuropathic pain development through epigenetic silencing of Kcna2 in the axotomized DRG.

Scientific RepoRts | 6:37704 | DOI: 10.1038/srep37704 transcriptional repression 14 . Although a recent study showed an involvement of G9a in nerve injury-induced downregulation of some DRG potassium channel genes 15 , whether and how G9a regulates Kcna2, a highly expressed potassium channel in the DRG 4,5 , under neuropathic pain conditions is unknown. Here, we report that peripheral nerve injury increases the expression of G9a and H3K9me2 in the axotomized DRG neurons. These increases contribute to neuropathic pain development through epigenetic silencing of DRG Kcna2.
We further examined whether the activity and expression of G9a were altered in DRG following peripheral nerve injury. A preclinical mouse model of unilateral fourth lumbar (L4) spinal nerve ligation (SNL) 16 was carried out. The levels of euchromatic histone-lysine N-methyltransferase 2 (Ehmt2) mRNA (encoding G9a protein, Fig. 2A) and the amounts of G9a's two protein isoforms (including the 165 kDa long isoform and 140 kDa short isoform; Fig. 2B) were time-dependently increased in the ipsilateral L4 (axotomized) DRG on days 3, 7, and 14 post-SNL, but not after sham surgery. SNL also led to a time-dependent increase in the level of H3K9me2 in the ipsilateral L4 DRG (Fig. 2C). Neither SNL nor sham surgery altered basal expression of G9a's two protein isoforms or H3K9me2 in the contralateral L4 DRG, the ipsilateral L3 (intact) DRG, and ipsilateral L4 spinal cord ( Fig. 2D-F). Immunofluorescent staining showed that the numbers of G9a-labeled neurons in the ipsilateral L4 DRG on day 7 post-SNL were increased by 2.7-fold as compared to the corresponding sham groups (Fig. 2G). We also observed the increased G9a expression in the axotomized DRGs after chronic constriction injury of sciatic nerve (CCI; Fig. 2H), another preclinical neuropathic pain model 17 . The levels of G9a's two protein isoforms and H3K9me2 on day 7 post-CCI were higher than those on day 7 post-sham surgery (Fig. 2H). To test whether these increases were a specific response to peripheral nerve injury, we injected complete Freund's adjuvant (CFA) into the plantar side of one hindpaw to induce peripheral inflammation in mice 18,19 . The amounts of G9a and H3K9me2 proteins were not significantly altered in either L3/4 DRGs (Fig. 2I) or L3/4 spinal cord (Fig. 2J) on the ipsilateral side from 2 h to 7 days post-CFA. Our findings indicate that induction of G9a expression and increased level of H3K9me2 occurs exclusively in the axotomized DRG neurons and specifically in response to peripheral nerve injury.
Blocking increased G9a in the axotomized DRG attenuates neuropathic pain. To examine if increased G9a in the axotomized DRG has a functional effect on neuropathic pain, we next observed whether blocking increased G9a in the axotomized DRG through microinjection of AAV5-Cre into the ipsilateral L4 DRG of G9a fl/fl mice affected SNL-induced pain hypersensitivity. AAV5-GFP was used as a control. Mice were subjected to SNL or sham surgery 5 weeks after viral microinjection. Basal paw withdrawal responses to mechanical, thermal, and cold stimuli were similar between the two virus-injected groups (Fig. 3A-G). As expected, SNL Figure 2. Nerve injury-induced increase in euchromatic histone-lysine N-methyltransferase 2 (Ehmt2) mRNA, G9a's two protein isoforms and H3K9me2 in the axotomized DRG neurons. (A) Ehmt2 mRNA expression in the ipsilateral (Ipsi) L4 (axotomized) DRG after SNL or sham surgery. Con: contralateral. n = 12 mice/time point/group. *P < 0.05 or **P < 0.01 vs the corresponding control group (0 day), one-way ANOVA followed by post hoc Tukey test. (B,C) Expression of G9a's two protein isoforms (B) and H3K9me2 (C) in the ipsilateral L4 (axotomized) DRG after SNL or sham surgery. Representative Western blots (left panels) and a summary of densitometric analysis (right graphs) are shown. L: long isoform. S: short isoform. n = 6 mice/time point/group. *P < 0.05 or **P < 0.01 vs the corresponding control group (0 day), one-way ANOVA followed by post hoc Tukey test. Full-length blots are presented in Supplemental led to mechanical allodynia and thermal or cold hyperalgesia on the ipsilateral (but not contralateral) side of the AAV5-GFP-injected group from days 3 to 14 after SNL (Fig. 3A-D). These pain hypersensitivities were abolished or ameliorated on the ipsilateral side of the AAV5-Cre-injected group (Fig. 3A-D). Paw withdrawal frequency to mechanical stimulation did not change compared to basal values and paw withdrawal latencies to thermal or cold stimulation were higher compared to the AAV5-GFP-injected mice during the observation period ( Fig. 3A-D). DRG microinjection may cause cell damage although the injected DRGs retained their structural integrity and displayed no significant changes in the numbers of neurons or satellite cells (data not shown). To exclude the possibility that the observed effects above were produced by DRG microinjection, we examined SNL-induced pain hypersensitivity in the conditional G9a knockout (G9aKO) mice. Like AAV5-Cre-injected G9a fl/fl mice, G9aKO mice exhibited the reduced paw withdrawal frequency to mechanical stimulation and increased paw withdrawal latency to thermal or cold stimulation compared to G9a fl/fl mice on the ipsilateral side from days 3 to 14 after SNL ( Fig. 3H-K). Also, mechanical, thermal, or cold stimuli-induced responses did not change in sham virus-injected and sham genetic KO mice ( Fig. 3A-N). Collectively, our findings indicate that G9a in the axotomized DRG may be required for neuropathic pain induction.

DRG G9a overexpression induced pain hypersensitivities.
To further examine whether the increased G9a in the DRG was sufficient for neuropathic pain induction, we microinjected HSV that expressed full-length G9a protein (HSV-G9a) into unilateral L3/4 DRGs of naïve adult mice. HSV-GFP was used as a control. Mice injected with HSV-G9a, but not HSV-GFP, exhibited significant increases in paw withdrawal frequencies in response to mechanical stimuli and marked decreases in paw withdrawal latencies in response to thermal and  (Fig. 4A,B). These findings indicate that the increased G9a in DRG leads to mechanical allodynia and thermal and cold hyperalgesia.
We also asked whether increased G9a in the DRG altered spinal centralization. The levels of phosphorylated extracellular signal-regulated kinase 1/2 (p-ERK1/2, a marker for neuronal hyperactivation) and glial fibrillary acidic protein (GFAP, a marker for astrocyte hyperactivation) significantly increased in the ipsilateral L3/4 dorsal horn of spinal cord on day 7 after microinjection of HSV-G9a compared to those after microinjection of HSV-GFP (Fig. 4D). These findings further support our behavioral observations described above.
G9a regulation of Kcna2 expression in the axotomized DRG following SNL. To determine how increased G9a in the axotomized DRG is involved in neuropathic pain induction, we carried out microarray analyses to identify the downstream targets of G9a in the DRG. The changes in gene expression profiles induced by G9a overexpression in the DRG were observed. The expression levels of 5,234 genes out of a total 38,851 identified genes were significantly altered in the injected DRG from the HSV-G9a-injected group compared with the HSV-GFP-injected group. Approximately 55.46% of these altered genes were downregulated and 44.54% upregulated (Fig. 5A). These altered genes are involved in multiple cell functions including immune response, cellular developmental process, tumor genesis, cell apoptosis, proliferation, migration, metabolic process, epigenetic modification, nervous system development, neurogenesis, and nerve-nerve synaptic transmission. More importantly, the expression of Kcna2 and Kcna4 genes was reduced by 43% and 45%, respectively, in the HSV-G9a-injected group compared to the HSV-GFP-injected group in the DRG.
Given that DRG Kcna2 is a key player in neuropathic pain genesis 4,10-12,20 , we examined if DRG Kcna2 is a mediator by which G9a participates in the development of neuropathic pain. We carried out quantitative real-time RT-PCR and Western blot analysis to further examine the effect of increasing DRG G9a on Kcna2 expression in the DRGs of naïve mice microinjected with HSV-G9a. G9a regulation of Kcna4 expression in DRG neurons reported previously 15 was verified. The levels of Ehmt2 mRNA, G9a's two protein isoforms and H3K9me2 increased while the amounts of Kcna2 and Kcna4 mRNAs and proteins decreased in the injected DRGs of the HSV-G9a-injected group compared to the HSV-GFP-injected group (Fig. 5B,C). Numbers of Kcna2-and Kcna4-labeled neurons were also markedly declined in the HSV-G9a-injected DRGs compared to those in the HSV-GFP-injected DRGs (Fig. 5D). Moreover, single-cell RT-PCR revealed co-expression of Ehmt2 mRNA with Kcna2 and Kcna4 mRNAs in individual small, medium and large DRG neurons (Fig. 5E,F). It is very likely that G9a directly regulates Kcna2 and Kcna4 expression in the DRG. Indeed, overexpression of G9a through transduction of HSV-G9a into in vitro cultured DRG neurons not only significantly increased the levels of G9a's two protein isoforms as well as H3K9me2 but also reduced the amounts of Kcna2 and Kcna4 proteins (Fig. 5G). Given that peripheral nerve injury downregulates Kcna2 and Kcna4 in the axotomized DRG 10,21 and that G9a is a gene repressor 22 , G9a may be responsible for their downregulation under neuropathic pain conditions. To this end, we determined whether blocking increased G9a in DRG affected Kcna2 and Kcna4 downregulation in the axotomized DRG. We microinjected AAV5-Cre into the ipsilateral L4 DRG of G9a fl/fl mice to specifically and selectively block SNL-induced increases in the amounts of Ehmt2 mRNA, G9a protein, and H3K9me2 in the axotomized DRG on day 7 post-SNL (Fig. 6A,B). AAV5-GFP was used as a control. Injection of AAV5-Cre, but not AAV5-GFP, rescued the expression of Kcna2 and Kcna4 mRNAs and proteins in the axotomized DRG on day 7 post-SNL (Fig. 6A,B). AAV5-Cre injection also markedly increased basal amounts of Kcna2 mRNA and protein in the axotomized DRG of sham G9a fl/fl mice on day 7 (Fig. 6A  from naive DRG (Fig. 6D). These binding activities were strikingly increased in the injured DRG on day 7 after SNL compared to those after sham surgery (Fig. 6E). Similar increased bindings of H3K9me2 were also observed in the injured DRG on day 7 after SNL (Fig. 6F). The evidence described above strongly suggests the contribution of G9a to nerve injury-induced Kcna2 and Kcna4 downregulation in the axotomized DRG.

DRG G9a overexpression decreases total Kv current and increases excitability in DRG neurons.
Since G9a reduced Kcna2 expression, we finally investigated whether mimicking the SNL-induced DRG G9a increase would affect total Kv current and the excitability in DRG neurons. Whole-cell voltage-clamp recording was carried out 4-7 days after HSV injection. Given that HSV-G9a expresses both G9a and GFP (green color), the green neurons were recorded. Total Kv current density in the HSV-G9a-injected group was significantly declined in large, medium, and small DRG neurons compared to the HSV-GFP-injected group (Fig. 7A-C). Maurotoxin (MTX) at 100 nM is a highly selective Kcna2 current inhibitor 10,20,23,24 . We used it to verify whether this reduction was attributed to Kcna2 downregulation. Bath application of 100 nM MTX led to a significant reduction in total Kv current in large and medium neurons from the HSV-GFP-injected group, but not from the HSV-G9a-injected group ( Fig. 7A-C). When tested at + 50 mV, large and medium neurons in the HSV-GFP-injected group retained 71.9% and 79.2% of current, respectively, after MTX treatment, however, both large and medium neurons from the HSV-G9a-injected group retained 86.2% and 88.3% of current, respectively. In small DRG neurons, the current reduction by MTX was not significant between HSV-GFP-and HSV-G9a-injected groups. These findings indicate that DRG G9a overexpression decreases total Kv current densities in all DRG neurons and reduces Kcna2-related current only in large and medium DRG neurons.
To examine whether DRG G9a overexpression altered DRG neuronal excitability, we carried out whole-cell current-clamp recording 4-7 days after viral injection. The resting membrane potentials in the HSV-G9a-injected group significantly increased by 9.44, 9.33, and 6.47 mV, respectively, in large, medium, and small neurons compared to the HSV-GFP-injected group (Fig. 7D). The current thresholds in the HSV-G9a-injected group decreased by 37.7%, 37.1%, and 18.7% of the values in the HSV-GFP-injected group in large, medium, and small neurons, respectively (Fig. 7E). The average number of action potentials (APs) evoked by stimulation of between 600 and 900 pA in the HSV-G9a-injected group was greater than that in the HSV-GFP-injected group in large neurons, whereas the stimulation of ≥ 400 pA in the HSV-G9a-injected group was greater than that in the HSV-GFP-injected group in medium and small neurons (Fig. 7F). There are no significant changes in other AP parameters such as membrane input resistance, AP threshold, AP overshoot, AP amplitude and after hyperpolarization amplitude (Table 1). These data indicate that DRG G9a overexpression increases DRG neuronal excitability.

Discussion
This study provides the first evidence to our knowledge that nerve injury-induced increases in G9a and its catalyzed repressive marker H3K9me2 are responsible for epigenetic silencing of Kcna2 in the axotomized DRG neurons. Mimicking these increases reduces Kcna2 expression and total Kv current and increases excitability in the DRG neurons and leads to pain hypersensitivity. Conversely, blocking these increases rescues Kcna2 expression in the axotomized DRG and attenuates nerve injury-induced pain hypersensitivities. DRG G9a contributes to neuropathic pain likely by silencing Kcna2 expression in the axotomized DRG.
G9a, a major euchromatic methyltransferase, is widely expressed in most tissues, including the peripheral nervous system [25][26][27] . The human and mouse G9a proteins have two alternatively spliced isoforms, long and short 13,22 , which have been identified in mouse DRG and spinal cord in the present study and in other tissues in previous work [28][29][30] . However, a recent study did not detect these two protein isoforms in the DRG tissue 15 . The reason for this difference between the present and previous studies is unknown but may be related to distinct primary antibodies used. Although G9a's two protein isoforms are expressed at a low level in normal DRG, they can be activated at the transcriptional level. Peripheral nerve injury increased expression of Ehmt2 mRNA and G9a's two protein isoforms as well as H3K9me2 in the axotomized DRG, but not intact DRG and spinal cord. Interestingly, CFA-induced peripheral inflammation did not alter levels of G9a's two protein isoforms and H3K9me2 in either DRG or spinal cord. It appears that G9a transactivation in DRG is tissue-and nerve injury-specific. Our double-labeling immunohistochemistry further demonstrated that this transactivation occurs exclusively in DRG neurons.
Peripheral nerve injury alters the expression of genes encoding receptors, enzymes, and ion channels in the axotomized DRG 12,31,32 . G9a may be a key regulator of these changes under neuropathic pain conditions. Laumet et al. reported that G9a inhibition restored 396 genes downregulated by nerve injury and normalized 242 genes upregulated by nerve injury in the axotomized DRG 15 . They further demonstrated that G9a was required for nerve injury-induced silencing of Kcna4, Kcnd2, Kcnq2, and Kcnma1 genes in the DRG 15 . However, no direct evidence demonstrated the participation of these target K + channels in neuropathic pain due to the lack of selective and specific antagonists/inhibitors for these K + channels. Thus, whether G9a contributes to neuropathic pain through the silencing of these K + channels in the DRG neurons is still uncertain. Our gene microarray analysis showed that DRG G9a overexpression altered the expression of 5,234 genes, in which 2,903 genes were downregulated and 2,331 genes upregulated. More importantly, our microarray assay revealed that overexpression of G9a significantly reduced Kcna2 RNA expression in the DRG, which was further confirmed by our quantitative RT-PCR and Western blot assays. Although the contribution of most of these altered genes to neuropathic pain was unclear, DRG Kcna2 is a key player in neuropathic pain. Rescuing Kcna2 downregulation by overexpressing Kcna2 RNA or blocking increased Kcna2 AS RNA in the axotomized DRG ameliorated neuropathic pain 4,10,20 . Mimicking nerve injury-induced Kcna2 downregulation by overexpressing Kcna2 AS RNA in DRG reduced total Kv current, depolarized resting membrane potential, decreased current threshold for inducing AP, and increased the excitability in the DRG neurons and led to neuropathic pain-like behaviors 4,10,20 . Present data showed that blocking increased G9a rescued Kcna2 expression in the axotomized DRG and attenuated neuropathic pain. G9a overexpression in the DRG led to similar phenomena to those produced by DRG Kcna2 AS RNA overexpression. Kcna2 belongs to a family of delayed rectifiers, which allow DRG neurons to efficiently repolarize following an action potential. How Kcna2 downregulation induced by overexpression of either G9a or Kcna2 AS RNA participates in resting membrane depolarization in the DRG neurons is still unclear. The changes in the expression of other channels induced directly by G9a overexpression and/or following Kcna2 downregulation in the DRG neurons may be involved in resting membrane depolarization observed presently. Given that G9a is co-localized with Kcna2 in the DRG neurons, it is very likely that G9a participates in the mechanisms underlying neuropathic pain development by downregulating Kcna2 in the axotomized DRG. However, other potential mechanisms of G9a involvement in neuropathic pain cannot be ruled out, because G9a regulates the expression of other genes in the axotomized DRG 33 . It should be noted that multiple mechanisms are involved in nerve injury-induced Kcna2 silencing in the axotomized DRG. In addition to G9a's role revealed in the present study and Kcna2 AS RNA reported previously 10 , our recent work showed that DNMT3a-triggered de novo DNA methylation in the promoter and 5′ -untranslated regions of the Kcna2 gene was crucial in nerve injury-induced Kcna2 silencing in the axotomized DRG (data not shown). How these three epigenetic mechanisms work together to regulate Kcna2 expression and whether they interact with/affect each other under neuropathic pain conditions are unknown and remain to be further investigated. In addition, a decrease in Kcna2 mRNA levels may be the result of decreased mRNA stability and/ or reduced transcription factors-triggered transcriptional activation. These potential possibilities will also be addressed in our future studies.
In summary, our study revealed a G9a-triggered epigenetic mechanism of how Kcna2 is downregulated in the axotomized DRG under neuropathic pain conditions. Given that DRG G9a inhibition or knockout produces antinociceptive effects without affecting motor function and basal pain perception, G9a could be a novel target in neuropathic pain management.

Materials and Methods
Animal preparations. C57BL/6 J wild-type mice, G9a fl/fl mice 34 , and Advillin Cre/+ mice 35 were used in this study. G9a fl/fl mice were fully backcrossed to C57BL/6 J mice and were homozygous for a floxed G9a allele. Male sensory-specific Cre line Advillin Cre/+ mice were crossed with female G9a fl/fl mice to obtain G9a conditional knockout (G9aKO) mice. All animals were kept in a standard 12-h light/dark cycle, with water and food pellets available ad libitum. Male mice weighing 25-30 g were used for behavior testing. All procedures used were approved by the Animal Care and Use Committee at Rutgers New Jersey Medical School and are consistent with the ethical guidelines of the US National Institutes of Health and the International Association for the Study of Pain. All efforts were made to minimize animal suffering and to reduce the number of animals used. All of the experimenters were blind to treatment condition. DRG microinjection. DRG microinjection was carried out as described 10,20 with minor modification. Briefly, a midline incision was made in the lower lumbar back region, and the L 3 and/or L 4 articular processes were exposed and then removed with small ronguers. After the DRG was exposed, viral solution (0.5 μ l) was injected into two sites in the L 3 and L 4 DRGs or into one site in the L 4 DRG with a glass micropipette connected to a Hamilton syringe. The pipette was removed 10 min after injection. The surgical field was irrigated with sterile saline and the skin incision closed with wound clips. The injected mice displayed no sign of paresis or other abnormalities. The immune responses from viral injection were thus minimal. AAV5-GFP and AAV5-Cre were purchased from UNC Vector Core (Chapel Hill, NC). HSV-GFP and HSV-G9a (expressing both G9a and GFP) were provided by Dr. Eric J Nestler 34 .
Neuropathic pain models. L 4 Spinal nerve ligation (SNL) and chronic constriction injury of sciatic nerve (CCI) were carried out as described previously 10,20 . Briefly, SNL was performed by ligation of the fourth lumbar spinal nerve with 7-0 silk suture and transection at the distal site. CCI was done loosely ligation of sciatic nerve with 4-0 chromic gut thread at 4 sites with an interval of 1 mm. Sham-operative groups underwent identical procedures but without ligation of the respective nerve. Mechanical, thermal, and cold behavioral tests as described below were performed before surgery (− 35 and − 1 days) and 3, 5, 7, 10, 12, or 14 days after surgery.
Behavioral tests. Mechanical, thermal, and cold behavioral tests were carried out as described 10,36 . These behavioral tests were conducted at 1 hour intervals. Paw withdrawal frequencies in response to mechanical stimuli (calibrated von Frey filaments) were first measured. Paw withdrawal latencies to noxious heat were then measured with a Model 336 Analgesia Meter (IITC Inc. Life Science Instruments. Woodland Hills, CA). Finally, paw withdrawal latencies to noxious cold (0 °C) were measured with a cold aluminum plate. DRG neuronal culture and transfection. Primary DRG neuronal cultures and viral transfection were carried out as described 10 . Briefly, adult mice were euthanized with isoflurane and all DRGs were collected in cold Neurobasal Medium (Gibco/ThermoFisher Scientific) with 10% fetal bovine serum (JR Scientific, Woodland, CA), 100 units/ml Penicillin, 100 μ g/ml Streptomycin (Quality Biological, Gaithersburg, MD) and then treated with enzyme solution (5 mg/ml dispase, 1 mg/ml collagenase type I in Hanks' balanced salt solution (HBSS) without Ca 2+ and Mg 2+ (Gibco/ThermoFisher Scientific)). After trituration and centrifugation, dissociated cells were resuspended in mixed Neurobasal Medium and plated in a six-well plate coated with 50 μ g/ml poly-D-lysine (Sigma, St. Louis, MO). The cells were incubated at 95% O 2 , 5% CO 2 , and 37 °C. One day later, 1 μ l of virus (titer ≥ 1 × 10 12 /ml) was added to each 2 ml-well. Neurons were collected 2 days later.

Reverse transcription (RT)-polymerase chain reaction (PCR).
For quantitative real-time RT-PCR, four unilateral mouse DRGs were pooled together to achieve enough RNA. Total RNA was extracted by the Trizol method (Invitrogen/ThermoFisher Scientific), treated with DNase I (New England Biolabs, Ipswich, MA), and reverse-transcribed using the ThermoScript reverse transcriptase (Invitrogen/ThermoFisher Scientific), random hexamers, oligo (dT) primers or specific RT-primers ( Table 2). The reverse primer of each pairs of primers were used as RT-primers. Template (1 μ l) was amplified by real-time PCR by using the primers listed in the Table 2 (Integrated DNA Technologies). GAPDH was used as an internal control for normalization. Each sample was run in triplicate in a 20 μ L reaction with 250 nM forward and reverse primers, 10 μ l of SsoAdvanced Universal SYBR Green Supermix (Bio-Rad Laboratories, Hercules, CA) and 20 ng of cDNA. Reactions were performed in a BIO-RAD CFX96 real-time PCR system. Ratios of ipsilateral-side mRNA levels to contralateral-side mRNA Scientific RepoRts | 6:37704 | DOI: 10.1038/srep37704 levels were calculated using the Δ Ct method (2 −ΔΔCt ). All data were normalized to GAPDH, which has been demonstrated to be stable even after peripheral nerve injury insult 10 .
For single-cell real-time RT-PCR, freshly dissociated mouse DRG neurons were first prepared as described previously 10 . Briefly, four hours after plating, a single living DRG neuron was collected under an inverted microscope fit with a micromanipulator and microinjector and placed in a PCR tube with 5-10 μ l of cell lysis buffer (Signosis, Sunnyvale, CA). After centrifugation, the supernatants were collected. The remaining real-time RT-PCR procedure was carried out according to the manufacturer's instructions with the single-cell real-time RT-PCR assay kit (Signosis). All primers used are listed in Table 2.
Single-or double-labeled immunohistochemistry. Mice were anesthetized with isoflurane and perfused with 4% paraformaldehyde before being analyzed by single-or double-labeled immunohistochemistry. L3 and L4 DRGs were removed, post-fixed, and dehydrated before frozen sectioning at 20 μ m. After the sections were blocked for 1 h at room temperature in 0.01 M PBS containing 10% goat serum and 0.3% Triton X-100, they were incubated with the following primary antibodies or the regents over one or two nights at 4 °C. The antibodies and regents include: Control experiments included substitution of normal mouse or rabbit serum for the primary antiserum and omission of the primary antiserum. All immunofluorescence-labeled images were examined using a Leica DMI4000 fluorescence microscope and captured with a DFC365FX camera (Leica, Germany). Single-or double-labeled neurons were quantified manually or by using NIH Image J Software.
Western blotting. To achieve enough proteins, four unilateral mouse DRGs were pooled together.
Tissues were homogenized and the cultured cells ultrasonicated in chilled lysis buffer (10 mM Tris, 1 mM phenylmethylsulfonyl fluoride, 5 mM MgCl 2 , 5 mM EGTA, 1 mM EDTA, 1 mM DTT, 40 μ M leupeptin, 250 mM sucrose). After centrifugation at 4 °C for 15 min at 1,000 g, the supernatant was collected for cytosolic proteins and the pellet for nuclear proteins. The contents of the proteins in the samples were measured using the Bio-Rad protein assay (Bio-Rad) and then the samples were heated at 99 °C for 5 min and loaded onto a 4-15% stacking/7.5% separating SDS-polyacrylamide gel (Bio-Rad Laboratories). The proteins were then electrophoretically transferred onto a polyvinylidene difluoride membrane (Bio-Rad Laboratories). After the membranes were blocked with 3% nonfat milk in Tris-buffered saline containing 0.1% Tween-20
Chromatin immunoprecipitation (ChIP). The ChIP assays were conducted using the EZ ChIP Kit (Upstate/EMD Millipore, Darmstadt, Germany) as described 10 . The homogenization solution from DRGs was crosslinked with 1% formaldehyde for 10 min at room temperature. The reaction was terminated by the addition of 0.25 M glycine. After centrifugation, the collected pellet was lysed by SDS lysis buffer with protease inhibitor cocktail and sonicated until the DNA was broken into fragments with a mean length of 200 to 1,000 bp. After the samples were pre-cleaned with protein G agarose, they were subjected to immunoprecipitation overnight with 2 μ g of rabbit antibodies against G9a (Abcam) and H3K9me2 (Abcam), or with 2 μ g of normal rabbit serum overnight at 4 °C. Input (10-20% of the sample for immunoprecipitation) was used as a positive control. The DNA fragments were purified and identified using PCR/Real-time PCR with the primers listed in Table 2.
Briefly, 4 days after microinjection of HSV-G9a or HSV-GFP into unilateral L3 sand L4 DRGs, mice were rapidly decapitated and the injected DRGs were removed, quickly frozen in liquid nitrogen, and then transferred to − 80 °C until RNA was extracted. Injected DRGs from 2 mice were pooled together. Total RNA was extracted by the Trizol method (Invitrogen) and treated with excess DNase I (New England Biolabs, Ipswich, MA). RNA quality was checked using Agilent's Bioanalyzer. Reverse transcription, amplification, labeling and hybridization to Affymetrix Mouse Exon 1.0ST arrays were performed using standard procedures at Johns Hopkins Deep Sequencing & Microarray Core Facility. Raw data were extracted from the raw Affymetrix CEL files using Partek Genomics Suite 6.6 platform (Partek Inc., St. Louis MO, USA). The extended gene transcript set was imported and normalized using the RMA (Robust Multiarray Analysis) algorithm, then compared between two groups using a one-way ANOVA. Normalized data were further analyzed using Spotfire DecisionSite for Functional Genomics software (TIBCO Spotfire, Boston MA, USA). Log 2 ratios representing the difference in the expression level were obtained on the basis of the expression level from the GFP-injected group. Gene lists were generated using significance criteria of a linear 1.2 fold change cutoff coupled with a nonstringent p-value cutoff of P < 0.05, more than 5 probes per transcript and appropriate genomic mapping. Evaluation of lists of differentially expressed genes for enrichment in predefined categories and functional groups of genes was carried out using functional analysis tools GORILLA. To minimize batch effects and maintain a high degree of confidence in these data, all animals were handled, treated and euthanized at the same time under the same conditions and the processing of all RNAs and arrays carried out at the same time.
Whole-cell patch clamp recording. To record total potassium current in DRG neurons, we first prepared dissociated mouse L3/L4 DRG neurons as described above. Whole-cell patch clamp recording was carried out 4 to 8 h after plating. To improve recording efficiency, only green labeled neurons were recorded. Coverslips were placed in the perfusion chamber (Warner Instruments, Hamden, CT). The electrode resistances of micropipettes ranged from 3 to 5 MΩ . Neurons were voltage-clamped with an Axopatch-700B amplifier (Molecular Devices, Sunnyvale, CA). The intracellular pipette solution contained (in mM) potassium gluconate 120, KCl 20, MgCl 2 2, EGTA 10, HEPES 10, and Mg-ATP 4 (pH 7.3 with KOH, 310 mOsm). We minimized the Na + and Ca 2+ component in voltage-gated potassium current recording using an extracellular solution composed of (in mM) choline chloride 150, KCl 5, CdCl 2 1, CaCl 2 2, MgCl 2 1, HEPES 10, glucose 10 (pH 7.4 with Tris base, 320 mOsm). Signals were filtered at 1 kHz and digitized by using a DigiData 1550 with pClamp 10.4 software (Molecular Devices). Series resistance was compensated by 60-80%. Cell membrane capacitances were acquired by reading the value for whole-cell capacitance compensation directly from the amplifier. An online P/4 leak subtraction was performed to eliminate leak current contribution. The data were stored on a computer by a DigiData 1550 interface and were analyzed by the pCLAMP 10.4 software package (Molecular Devices).
To record the action potential, we switched the recording mode into current clamp. The extracellular solution consisted of (in mM) NaCl 140, KCl 4, CaCl 2 2, MgCl 2 2, HEPES 10 and glucose 5, with pH adjusted to 7.38 by NaOH. The intracellular pipette solution contained (in mM) KCl 135, Mg-ATP 3, Na 2 ATP 0.5, CaCl 2 1.1, EGTA 2 and glucose 5; pH was adjusted to 7.38 with KOH and osmolarity adjusted to 300 mOsm with sucrose. The resting membrane potential was taken 3 min after a stable recording was first obtained. Depolarizing currents of 100-1,400 pA (200-ms duration) were delivered in increments of 100 pA until an action potential (AP) was evoked. The injection current threshold was defined as the minimum current required to evoke the first AP. The membrane potential was held at the existing resting membrane potential during the current injection. The AP threshold was defined as the first point on the rapid rising phase of the spike at which the change in voltage exceeded 50 mV/ms. The AP amplitude was measured between the peak and the baseline. The membrane input resistance for each cell was obtained from the slope of a steady-state I-V plot in response to a series of hyperpolarizing currents, 200-ms duration delivered in steps of 100 pA from 200 pA to − 2,000 pA. The after-hyperpolarization amplitude was measured between the maximum hyperpolarization and the final plateau voltage, and the AP overshoot was measured between the AP peak and 0 mV. The data were stored on a computer by a DigiData 1550 interface and were analyzed by the pCLAMP 10.4 software package (Molecular Devices). All experiments were performed at room temperature.

Statistical analysis.
For in vitro experiments, the cells were evenly suspended and then randomly distributed in each well tested. For in vivo experiments, the animals were distributed into various treatment groups randomly. All of the results are given as means ± S.E.M. The data were statistically analyzed with two-tailed, paired Student's t-test and a one-way or two-way ANOVA. When ANOVA showed a significant difference, pairwise comparisons between means were tested by the post hoc Tukey method (SigmaPlot 12.5, San Jose, CA). Significance was set at P < 0.05.