Targeted up-regulation of Drp1 in dorsal horn attenuates neuropathic pain hypersensitivity by increasing mitochondrial fission

Mitochondria play an essential role in pathophysiology of both inflammatory and neuropathic pain (NP), but the mechanisms are not yet clear. Dynamin-related protein 1 (Drp1) is broadly expressed in the central nervous system and plays a role in the induction of mitochondrial fission process. Spared nerve injury (SNI), due to the dysfunction of the neurons within the spinal dorsal horn (SDH), is the most common NP model. We explored the neuroprotective role of Drp1 within SDH in SNI. SNI mice showed pain behavior and anxiety-like behavior, which was associated with elevation of Drp1, as well as increased density of mitochondria in SDH. Ultrastructural analysis showed SNI induced damaged mitochondria into smaller perimeter and area, tending to be circular. Characteristics of vacuole in the mitochondria further showed SNI induced the increased number of vacuole, widened vac-perimeter and vac-area. Stable overexpression of Drp1 via AAV under the control of the Drp1 promoter by intraspinal injection (Drp1 OE) attenuated abnormal gait and alleviated pain hypersensitivity of SNI mice. Mitochondrial ultrastructure analysis showed that the increased density of mitochondria induced by SNI was recovered by Drp1 OE which, however, did not change mitochondrial morphology and vacuole parameters within SDH. Contrary to Drp1 OE, down-regulation of Drp1 in the SDH by AAV-Drp1 shRNA (Drp1 RNAi) did not alter painful behavior induced by SNI. Ultrastructural analysis showed the treatment by combination of SNI and Drp1 RNAi (SNI + Drp1 RNAi) amplified the damages of mitochondria with the decreased distribution density, increased perimeter and area, as well as larger circularity tending to be more circular. Vacuole data showed SNI + Drp1 RNAi increased vacuole density, perimeter and area within the SDH mitochondria. Our results illustrate that mitochondria within the SDH are sensitive to NP, and targeted mitochondrial Drp1 overexpression attenuates pain hypersensitivity. Drp1 offers a novel therapeutic target for pain treatment.


Introduction
Neuropathic pain (NP) is a kind of pain caused by peripheral or central nervous system lesions or injuries, and the patients with NP are often accompanied with a variety of symptoms such as paroxysmal pain, allodynia and hyperalgesia [1,2]. Due to the complex mechanism of NP, treatment of NP is still a challenge in clinical medicine [3][4][5]. SNI, ligation and axotomy of the tibial and common peroneal nerves but sural nerve intact, was reported to be a classic NP model [6,7]. The symptoms of allodynia and hyperalgesia in SNI mice were similar to the patients with NP. So we applied SNI model in the present study.
The spinal dorsal horn (SDH) receiving pain signal input from peripheral sensory neurons plays a pivotal position in the integration of pain signals and central pain sensitization [8,9]. Long-lasting changes in the processing of nociceptive information within the SDH contribute to pain-related pathological changes [9,10]. Our group have devoted to the study of the synaptic connectivity that underlies the noxious transmission and modulation within the SDH [11]. Here we focused on the plasticity changes of SDH in SNI model.
Mitochondria are crucial in the nervous system, especially in neurons [12,13]. They are the most sensitive device to detect the stability of neuron microenvironment. It is reported that mitochondria dysfunctions are closely related to various nervous system diseases via different mechanisms including oxidation respiration, mitochondrial dynamics, mitophagy, mitochondrial biogenesis and calcium homeostasis [14,15]. Importantly, neuronal mitochondrial function exceptionally depend on stable mitochondrial dynamics of fission and fusion, which shapes mitochondrial morphology and architecture [15][16][17][18]. Excessive mitochondrial fragmentation or swollen mitochondria with expanded vacuole space and fewer cristae, or disruption of the mitochondrial network, has been indicated to be associated with neurological diseases [15,18,19]. Mitochondrial division is mediated by the conserved dynamin-related GTPase Drp1 which assembles onto the surface of mitochondria and constricts this tubular organelle [20]. Drp1 has been shown to be required for the development of nervous system and Drp1 deletion causes embryonic lethality in mice [21,22]. Moreover, mitochondrial division regulated by Drp1 is essential for the suppression of oxidative damage and therefore for the survival of postmitotic neurons both in vivo and in vitro [22,23]. However, it is not known how the mitochondrial component within SDH is responsive under the SNI pathological conditions and what effect of up-regulation or down-regulation of Drp1 is on pain behaviors.
Herein, we explored the effects of up-and down-regulation of Drp1 within the SDH on animal gait and pain behaviors. We also studied parameters associated with the mitochondrial ultrastructure in different groups. The present results suggest that Drp1 is the potential molecule attenuating pain in SNI model.

SNI induced pain behavior and anxiety-like behavior
Pathophysiology of NP is sophisticated, and a desire to understand the underlying pathophysiological mechanisms entails the construction of preclinical models that mimic, as closely as possible, human clinical NP symptoms [24]. Here, we concentrated on the Spared Nerve Injury (SNI) model that was developed by Decostered and Woolf [6], which is highly characterized on both the sensory (injurious) and affective (emotional) dimensions, and induces long-term painful behavior and dysfunctions. With von Frey test, the golden standard for algesimetry test, changes in mechanical sensitivity of the paw (paw withdrawal threshold, PWT) have been assessed by our lab [25]. Here we firstly confirmed SNI induced adult mice to show persistent hypersensitivity and allodynia in the lateral paw skin area innervated by the sural nerve (Fig. 1A). PWT scores decreased markedly after SNI (D1, control: 0.93 ± 0.16 g; SNI: 0.36 ± 0.10 g; P < 0.0001) and remained significantly weaker than those of control mice for 2 weeks (D14, control: 0.870 ± 0.21 g; SNI: 0.06 ± 0.02 g; P < 0.0001) (Fig. 1B). In addition, the reaction latency also showed the similar decreased trend (D1, control: 10.21 ± 0.52 s; SNI: 5.62 ± 0.49 s; P < 0.0001. D14, control: 10.54 ± 0.34 s; SNI: 4.99 ± 0.23 s; P < 0.0001), suggesting that SNI caused a significant reduction in heat pain threshold (Fig. 1C).
In addition, the mice with NP also showed anxiety-like behavior ( Fig. 1D and E), which was consistent with previous report [26]. Rodents with anxiety are afraid to explore surroundings and tend to stay in a safer place, which is the periphery of the open field and the enclosed arms of EPM [27,28]. By using open field test (OFT), our results showed that the number of activities in the central region for mice subjected to SNI (control: 49.33 ± 12.97; SNI: 8.83 ± 6.55; P < 0.0001), accompanied by the percentage of distance (control: 15.39 ± 2.83%; SNI: 5.79 ± 2.93%; P = 0.0002) and time (control: 9.54 ± 2.46%; SNI: 2.38 ± 2.63%; P = 0.0007) in central area of OFT were significantly reduced, which supported an increase in anxiety-like behavior (Fig. 1D). And the average speed in OFT was decreased remarkably in SNI mice, compared with the control group. By using elevated plus-maze (EPM), the SNI mice also showed anxiety-like behavior (Fig. 1E).  (Fig. 1E). These results showed SNI induced pain behavior and anxiety-like behavior in mice.

SNI induced morphological changes and ultrastructure damage of mitochondria in SDH
Mitochondria are kept in vigilant thought to be a multi-layered quality control system that shields them from various stresses and ensures the maintenance of healthy mitochondria [29,30]. There are no studies that evaluate the changes of mitochondrial dynamics following the pain. Our lab has investigated the mitochondrial distribution and subcellular morphology in SNI mice by using the mitochondrial marker, Mitotracker Red® CM-H2XRox (Mito-Red) [31]. In order to exactly visualize the architectural changes of SDH mitochondria following SNI, the transmission electron microscopy (TEM) was used to measure seven parameters which is the combined method from previous reports [19,32]. In detail, the seven important mitochondrial parameters were used, including: a. mitochondria density (mito density; N/μm 2 ); b. mitochondria perimeter (mito perimeter; μm); c. mitochondria area (mito area; μm 2 ); d. mitochondrial circularity; e. vacuole density (N/mito); f. vacuole perimeter (μm); g. vacuole area (μm 2 ) ( Fig. 2A).
In addition, the present study was to analyze the mitochondrial vacuoles which reflect the damage state of the inner mitochondrial membrane into many separate vesicular matrix compartments [19]. The statistical results of mitochondrial vacuoles showed that the perimeter and area of mitochondrial vacuoles increased (Perimeter, control: 244.0 ± 108.5 nm; SNI: 269.6 ± 141.3 nm; P = 0.0386. Area, control: 2893.1 ± 2555.2 nm 2 ; SNI: 3676.4 ± 4855.2 nm 2 ; P = 0.0361) in mice with NP, and the density of mitochondria vacuoles was higher (control: 2.39 ± 1.10/mito; SNI: 3.90 ± 1.45/mito; P = 0.0024) (Fig. 2G-J). The results showed that SNI induced damage of the inner membrane structures within the SDH mitochondria.
Here we specially established three-dimensional reconstruction of mitochondria in the ipsilateral SDH from control (Fig. 3) and SNI (Fig. 4) mice, and analyzed the plot types captured in the volume of 16 × 15.5 × 1.75 μm from 35 electron microscope images of continuous sections, respectively. In the Gallery format in the control group (Fig. 3), there were 144 mitochondria captured from one neuronal soma in SDH. We could see 106 mitochondria (74%) had small volume ranged from 0.001 μm 3 to 0.050 μm 3 , 32 mitochondria (22%) had moderate volume ranged from 0.051 μm 3 to 0.088 μm 3 , and 8 mitochondria (4%) had large volume ranged from 0.089 μm 3 to 0.127 μm 3 . However, in the SNI group a larger number of mitochondria were captured, 259 from one neuronal soma in the ipsilateral SDH (Fig. 4). It should be noted that these mitochondria tended to be much smaller. Of 259 captured mitochondria, 83% mitochondria (216) had small volume ranged from 0.001 μm 3 to 0.062 μm 3 , 14% mitochondria (35) had moderate volume ranged from 0.063 μm 3 to 0.010 μm 3 , and only 3% mitochondria (8) had large volume from 0.011 μm 3 to 0.156 μm 3 .
These results suggested the mitochondria within the SDH were sensitive to NP stress and tended to have the increased number and the decreased volume.
2.3. SNI led to the elevation of the mitochondrial fission factor Drp1 expression in SDH, but did not change the expression of two mitochondrial fusion factors of Mfn1 and Mfn2 Our previous study has shown that Drp1 is highly expressed at the spinal cord level [33]. And many studies have shown different NP contributed to the increased expression of Drp1 [34][35][36][37][38][39][40][41][42]. Here we observed the changes of Drp1 in the SDH of the mice during NP by Western blot and qRT-PCR and also confirmed the increased Drp1 expression (Fig. 5A).
Mitochondrial fusion is the reverse regulation of fission, and the factors, Mitofusin 1/2 (Mfn1 and Mfn2), located on the outer mitochondrial membrane (OMM), are responsible for fusion process [43,44]. So we observed the changes of Mfn1 and Mfn2 in the SDH of mice during NP. The results showed that there was no significant difference in the expression of Mfn1 or Mfn2 in the spinal dorsal horn after SNI, either on the ipsilateral or the contralateral side ( Fig. 5B and C).
These results suggested that elevated Drp1 may contribute to the number and morphological changes of mitochondria after SNI.

Targeted up-regulation of Drp1 within SDH alleviated pain symptoms in SNI mice, while targeted down-regulation of Drp1 did not have the analgesic effect
To verify the contribution of Drp1 to SNI-induced mechanical pain sensitization, two complementary methods were used to modulate the function of Drp1, as follows ( Fig. 6A): (1) up-regulation of Drp1 repression with intraspinal injection of Drp1 overexpression (OE) AAV into the ipsilateral SDH 2 weeks prior to SNI surgery; and (2) downregulation of Drp1 repression with intraspinal injection of Drp1 RNA interference (RNAi) AAV into the ipsilateral SDH 2 weeks prior to SNI surgery. Another 2 weeks after SNI, the spinal cord was isolated and studied by confocal microscopy observation, Western blot and qRT-PCR (Fig. 6A).

Verification of up-regulation and down-regulation of Drp1 targeted to SDH in mice
The confocal imaging (Fig. 6B) confirmed the Drp1 OE AAV indeed fused into the ipsilateral SDH of SNI mouse. The RNAi AAV was also confirmed to be fused into the target of the ipsilateral SDH of SNI mouse by the confocal observation (data not shown).
In addition, Western blot results (Fig. 7A) verified the increased Drp1 expression at protein level induced by targeted Drp1 OE treatment, and the decreased Drp1 expression at protein level and mRNA level induced  (B) Representative electron microscopic images of mitochondria for spinal dorsal horn neurons in control (top, n = 94) and SNI (bottom, n = 73) mice. Bright dotted lines indicate mitochondria. Scale bar, 500 nm (C) Quantitative analysis showed that the mitochondrial density of SNI mice (n = 17) was lower than that of control mice (n = 12). Two-tailed unpaired separate variance estimation t-test. (D) Diagram of distribution pattern of mitochondrial perimeter (left) and quantitative analysis showed that the mitochondrial perimeter of SNI mice (n = 73) was shorter than that of control mice (n = 94) (right). Two-tailed unpaired separate variance estimation t-test. (E) Diagram of distribution pattern of mitochondrial area (left) and quantitative analysis showed that the mitochondrial area of SNI mice (n = 73) was smaller than that of control mice (n = 94) (right). Two-tailed unpaired separate variance estimation t-test. (F) Diagram of distribution pattern of mitochondrial circularity (left) and quantitative analysis showed that the circularity of mitochondrial of SNI mice (n = 73) was higher than that of control mice (n = 94) (right). Two-tailed unpaired separate variance estimation t-test. (G) Representative electron microscopic images of mitochondria vacuoles for spinal dorsal horn neurons in control (top, n = 224) and SNI (bottom, n = 279) mice. Bright dotted lines indicate vacuoles in the mitochondria. Scale bar, 200 nm (H) Quantitative analysis showed that the density of mitochondria vacuoles of SNI mice (n = 16) was higher than that of control mice (n = 16). Two-tailed unpaired separate variance estimation t-test.
(I) Diagram of distribution pattern of vacuoles perimeter in mitochondria (left) and quantitative analysis showed that the vacuoles perimeter in mitochondria of SNI mice (n = 224) was longer than that of control mice (n = 279) (right). Two-tailed unpaired separate variance estimation t-test. (J) Diagram of distribution pattern of vacuoles area in mitochondria (left) and quantitative analysis showed that the vacuoles area in mitochondria of SNI mice (n = 224) was larger than that of control mice (n = 279) (right). Two-tailed unpaired separate variance estimation t-test. (B) The Vantage module of Imaris was used to establish three-dimensional reconstruction of mitochondria in the SDH of control mice, and the plot type was presented in XYZ color format. Scale bar, 2 μm (C) Three-dimensional mitochondria were presented in Gallery format and arranged according to volume. Scale color was set according to the volume from purple indicating the minimum volume of mitochondria of 0.000 μm 3 to the red marking the maximum volume of mitochondria of 0.127 μm 3 . In the Gallery format in the control group, there were 144 mitochondria captured from one neuronal soma in SDH. We can see 106 mitochondria had small volume ranged from 0.001 μm 3 to 0.050 μm 3 , 32 mitochondria had moderate volume ranged from 0.051 μm 3 to 0.088 μm 3 , and 8 eight mitochondria had large volume ranged from 0.089 μm 3 to 0.127 μm 3 . Scale bar, 3 μm. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) by targeted Drp1 RNAi treatment (OE, contralateral: 1.17 ± 0.09; ipsilateral: 1.73 ± 0.21; P = 0.0017. RNAi, contralateral: 2.22 ± 0.55; ipsilateral: 0.99 ± 0.34; P = 0.0027) (Fig. 7A).

Targeted up-regulation of Drp1 within SDH alleviated pain symptoms induced by SNI, analyzed by CatWalk gait method
Then we studied the effects of targeted up-regulation or downregulation of Drp1 within SDH on pain symptoms in SNI mice. We have noticed that Ferrari LF et al. (2011) [41] have first reported that intrathecal administration of antisense oligodeoxynucleotide (ASO) against Drp1 markedly attenuated neuropathic mechanical hyperalgesia caused by anti-HIV/AIDS or anticancer chemotherapy in rats. In addition, Kanda H et al. (2015) [42] have reported that intrathecal administration of Drp1 ASO decreased mechanical allodynia in HIV NP in mice.   [45] have reviewed that block of Drp1 could be beneficial for pain animals.
In order to get the meticulous results about the item of the effect of drp1 on NP, we specifically introduced the CatWalk gait analysis which could provide exquisite and reliable observations for evaluating spontaneous hyperalgesia in pain model rodents [46,47]. CatwalkXT 10.6 automated gait analysis was used here.
We first performed Catwalk gait analysis of control and SNI group. Compared with the foot print data in control group ( Fig. 6C-E), SNI pain-bearing mice showed remarkable changes ( Fig. 6F-H). All seven parameters had significant differences between control group and SNI (A) Imaris software was used to reconstruct SDH in the volume of 16 × 15.5 × 1.75 μm from 35 electron microscope images of continuous sections. Scale bar, 2 μm (B) The Vantage module of Imaris was used to establish three-dimensional reconstruction of mitochondria in the SDH of SNI mice, and the plot type was presented in XYZ color format. Scale bar, 2 μm (C) Three-dimensional mitochondria were presented in Gallery format and arranged according to volume. Scale color was set according to the volume. In the SNI group, there were more number of mitochondria of 259 captured from one neuronal soma in the ipsilateral SDH. It should be noted that these mitochondria tended to be much smaller. Of 259 captured mitochondria, 216 mitochondria (83%) had small volume ranged from 0.001 μm 3 to 0.062 μm 3 , 35 mitochondria (14%) had moderate volume ranged from 0.063 μm 3 to 0.010 μm 3 , and 8 mitochondria (3%) had large volume from 0.011 μm 3 to 0.156 μm 3 . Scale bar, 3 μm. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) group ( Fig. 6O-V), including: (1) regularity index (%) (Fig. 6O), a measure of inter-limb coordination; (2) stand (s) (Fig. 6P), the duration of a paw touching the glass plate; (3) standing on three (%) (Fig. 6R), the percentage of time spent walking with three paws; (4) print length (cm) (Fig. 6S), the length of the paw print (horizontal direction); (5) print width (cm) (Fig. 6T), the width of the complete paw print (vertical direction); (6) max contact area (cm 2 ) (Fig. 6U), the maximum print area during paw contact; and (7) max contact mean intensity (Fig. 6V), the mean intensity at maximum paw contact; only except body speed (cm/s) (Fig. 6Q), which was calculated by counting the distance of the mouse's body (paws) from the initial contact with the glass plate to the next contact divided by the time required to move this distance. In detail, compared with the control group, SNI mice showed the significant increase in standing on three (control: 14.40 ± 16.39%; SNI: 37.13 ± 25.39%; P = 0.0053) (Fig. 6R), as well as the significant decrease in regularity index (control: 83.88 ± 19.02%; SNI: 65.86 ± 23.79%; P = 0.0489) (Fig. 6O), stand (control: 0.25 ± 0.17 s; SNI: 0.07 ± 0.05 s; P = 0.0022) (Fig. 6P), print length (control: 0.85 ± 0.15 cm; SNI: 0.33 ± 0.15 cm; P < 0.0001) (Fig. 6S), print width (control: 0.73 ± 0.12 cm; SNI: 0.28 ± 0.10 cm; P < 0.0001) (Fig. 6T), max contact area of the left posterior paw (ipsilateral) (control: 0.29 ± 0.08 cm 2 ; SNI: 0.06 ± 0.04 cm 2 ; P < 0.0001) (Fig. 6U), and max contact mean intensity (control: 128.6 ± 11.3; SNI: 105.4 ± 15.0; P = 0.0006) (Fig. 6V). So SNI painbearing mice showed remarkable painful behavioral changes analyzed (V) Max contact mean intensity, the mean intensity at maximum paw contact. We could find SNI induced significant decreases of the six parameters of regularity index (O), stand (P), print length (S), print width (T), max contact area (U), and max contact mean intensity (V); as well as induced significant increase of one parameter of standing on three (R). Unexpectedly, the targeted up-regulation of Drp1 by overexpression (OE) virus within SDH significantly alleviated decreased values of stand (P), print width (T), and max contact mean intensity (V) induced by SNI. On the contrary to the analgesia effect of Drp1 OE treatment, the targeted downregulation of Drp1 by RNA interference (RNAi) virus within SDH could not make any improvement in seven parameters altered by SNI including regularity index (O), stand (P), print length (S), print width (T), max contact area (U), and max contact mean intensity (V), as well as standing on three (R). Behavioral tests were performed 2 weeks after the SNI surgery and 4 weeks after AAV delivery in Drp1 OE or RNAi group. by Catwalk method (Fig. 6F-H).
Secondly, we analyzed the effect of up-regulation of Drp1 by OE AAV on SNI-induced gait due to hyperalgesia situation. Compared that in SNI group, up-regulation of Drp1 could significantly improve the gait on three items: prolonged duration of stand (SNI: 0.07 ± 0.05 s; SNI + OE: 0.17 ± 0.11 s; P = 0.0185) (Fig. 6P), increased ipsilateral width (SNI: 0.28 ± 0.10 cm; SNI + OE: 0.46 ± 0.21 cm; P = 0.0188) (Fig. 6T), and increased max contact mean intensity (SNI: 105.4 ± 15.0; SNI + OE: 120.7 ± 21.2; P = 0.0441) (Fig. 6V). Drp1 OE treatment recovered the value of regularity index from 65.86 ± 23.79% in SNI condition to 81.58 ± 17.75% (Fig. 6O) which was close to the value at normal state, as well as recovered the value of standing on three from 37.13 ± 25.39% in SNI condition to 17.12 ± 13.47% (Fig. 6R) which was close to the value at normal state; there was no significant difference between SNI and SNI + OE. Similarly, the value of print length increased (SNI: 0.33 ± 0.15 cm; SNI + OE: 0.65 ± 0.43 cm) (Fig. 6S), and the value of ipsilateral max contact area also increased (SNI: 0.06 ± 0.04 cm 2 ; SNI + OE: 0.11 ± 0.11 cm 2 (Fig. 6U); there was no significant difference between SNI and SNI + OE. These results indicated the effects of targeted upregulation of Drp1 within SDH could be beneficial for SNI pain symptoms in mice. Our present results were unexpectedly opposite with previous three reports mentioned in the last paragraph.

Targeted down-regulation of Drp1 could not alleviate pain symptoms induced by SNI, analyzed by CatWalk gait method
In order to draw a clear conclusion about the effect of regulation of Drp1 on pain, we further packaged RNA interference (RNAi) rAAV against Drp1 and analyzed the effect of down-regulation of Drp1 by RNAi AAV on SNI-induced abnormal gait.
Thirdly, gait analysis data showed that down-regulation of Drp1 by RNAi could not make any improvement in seven parameters altered by SNI, including: regularity index (Fig. 6O), stand (Fig. 6P), standing on three (Fig. 6R), print length (Fig. 6S), print width (Fig. 6T), max contact area (Fig. 6U), and max contact mean intensity (Fig. 6V). The values of the seven parameters of the mice in SNI + RNAi group were: 70.07 ± 18.43% of regularity index (Fig. 6O), 0.09 ± 0.07 s of stand (Fig. 6P), 21.31 ± 15.26% of standing on three (Fig. 6R), 0.37 ± 0.30 cm of print length (Fig. 6S), 0.29 ± 0.14 cm of print width (Fig. 6T), 0.04 ± 0.05 cm 2 of max contact area (Fig. 6U), and 108.28 ± 11.78 of max contact mean intensity (Fig. 6V). Based on these behavioral data, we could see the RNAi treated SNI mice avoided putting the ipsilateral paw on the plate, which was an indicator for being in the pain hypersensitive state.

Targeted up-regulation of Drp1 within SDH alleviated pain symptoms induced by SNI, analyzed by von Frey and hot plate methods
To reconfirm the effect of Drp1 regulation on SNI pain, we carried out von Frey and hot plate in our further study, which are the classic and credible algesimetry of mechanical pain and heat pain, respectively [48]. Von Frey and hot plate were also conducted to detect whether silencing Drp1 aggravated the pain induced by SNI in spite of no difference shown in Catwalk gait analysis.
In the present study, the up-regulation of Drp1 remarkably increased the value of paw withdrawal threshold (PWT) reduced by SNI (Fig. 7C). In the first part of Results section (2.1), we have shown all SNI mice withdrew the ipsilateral hind paw once stimulated by the smallest von Frey filament equal to or below the threshold of 0.07 g, only within 5 s (Fig. 1B). However, compared with those in SNI group, the values of PWT of the mice in SNI + OE group significantly increased ( Fig. 7C; SNI: 0.06 ± 0.02 g; SNI + OE: 0.60 ± 0.22 g; P < 0.0001). In detail, all SNI mice with the targeted treatment of Drp1 OE within the SDH did not withdraw the ipsilateral hind paw at the stimulation level of the smallest von Frey at 0.07 g (Fig. 7C). Until evoked with the smallest von Frey filament at 0.4 g, 2 SNI mice with the targeted Drp1 OE treatment began to withdraw the ipsilateral hind paw within 5 s. Moreover, 3 SNI mice with the targeted Drp1 OE treatment began to withdraw the ipsilateral hind paw within 5 s when evoked with the smallest von Frey filament at 0.6 g. In addition, 1 SNI mouse with Drp1 OE treatment withdrew the hind paw within 5 s, even when evoked with the smallest von Frey filament at 1.0 g.
Consistently, the reaction latency tested by hot plate also contributed to illustrating the effect of pain relief of the treatment of Drp1 overexpression (Fig. 7D). In the first part of Results section (2.1), all SNI mice were observed to hop from the 52 • C plate surface and lick the ipsilateral hind paw with the reaction latency of approximately 5 s. However, compared with that in SNI group, the reaction latency of the mice in SNI + OE group significantly increased ( Fig. 7D; SNI: 4.99 ± 0.23 s; SNI + OE: 8.42 ± 0.79 s; P < 0.0001). In detail, all SNI mice with the targeted Drp1 OE hopped from the hot plate surface and licked the ipsilateral hind paw with the reaction latency longer than 6 s (Fig. 7D). Even 5 of 6 SNI + OE mice hopped and began to lick with the reaction latency longer than 8 s.
These results showed the targeted up-regulation of Drp1 within SDH could alleviate pain symptoms in SNI mice, when analyzed by von Frey and hot plate methods.

Targeted down-regulation of Drp1 could not alleviate pain symptoms induced by SNI, analyzed by von Frey and hot plate methods
Compared with that in SNI group, the values of PWT in SNI + RNAi group increased, yet with no statistical difference between SNI group and SNI + RNAi group ( Fig. 7C; SNI: 0.06 ± 0.02 g; SNI + RNAi: 0.15 ± 0.04 g; P = 0.4862). Further in thermal hypersensitivity testing, compared with those in SNI group, the values of reaction latency in SNI + RNAi group did not have any increase ( Fig. 7D; SNI: 4.99 ± 0.23 s; SNI + RNAi: 4.73 ± 1.03 s; P > 0.5887). Even 2 of 6 SNI + RNAi mice hopped and began to lick with the reaction latency shorter than 3.74 s.
These results showed the targeted down-regulation of Drp1 within SDH could not make any improvement in alleviating pain symptoms in SNI mice, when analyzed by von Frey and hot plate methods.

SNI-induced anxiety behavior could be partially alleviated by both Drp1 OE and Drp1 RNAi treatment
It is generally believed that there is an interaction between nociceptive and anxiety-like behavior in rodents, so relieving pain in NP animals can also reverse NP-induced anxiety-like behavior [49][50][51]. For the purpose of exploring the emotional changes of targeted regulation of Drp1, we compared the Drp1 OE and Drp1 RNAi respectively with the SNI group using OFT and EPM.
In our study, we found that the ratio of distance moved in the central and total area (distance in central/total, %), one of the four indicators of OFT, showed increase in the Drp1 OE group ( It was foreseeable that SNI-induced anxiety behavior was partially alleviated in Drp1 OE mice. For the Drp1 RNAi group, OFT testing first showed it led to significant increase in the active number in central region (  (Fig. 7F), when compared with those in SNI group.
It was unforeseeable that SNI-induced anxiety behavior was partially alleviated by Drp1 RNAi treatment considering its unbeneficial effect on abnormal gait and hyperalgesia induced by SNI.
We thought these results were possible because previous reports have shown that SNI pain was independent of emotional alternation and pathway [52,53]. It was proved that the neural conduction involved in anxious behavior and the neuronal activation required for were separate. Here we proposed that the effect of Drp1 regulation on pain should mainly depend on the exact mitochondrial changes of both ultrastructure morphology and vacuole parameters. So we compared and analyzed the three-dimensional reconstruction of mitochondria in different groups in the next step.

Up-regulation of Drp1 reduced mitochondrial number, but did not change the external and internal appearance of the mitochondria induced by SNI; on the contrary, down-regulation of Drp1 led SDH mitochondria to be more swollen and with more broken cristae
It was well-known that Drp1 was involved in various neurological diseases by regulating the homeostasis of mitochondrial fission and fusion process [15,18,19]. So we further explored and analyzed the mitochondrial morphology and dynamics in detail considering the obvious behavioral changes after targeted regulation of Drp1 (Fig. 8).
Drp1 RNAi increased the vacuole density to 1.56 fold, the vacuole perimeter to 1.97 fold and the vacuole area to 4.50 fold of those at SNI state, respectively. Due to the pivotal role of mitochondrial cristae in respiration process and energy conversion [54], the present electron microscope study of Drp1 RNAi inducing the SDH mitochondria changing in morphology from normal to be swollen, with broken mitochondrial cristae and more invalid vacuole compartments, implied Drp1 RNAi treatment could deteriorate mitochondrial damages induced by SNI.

Targeted up-regulation of Drp1 alleviated ROS elevation induced by SNI, while targeted down-regulation of Drp1 aggravated oxidative damage
Previous studies showed that level of reactive oxygen species (ROS) rose in neurons after pain and injury [41,42,45]. According to previous report [55], mitochondrial ROS levels were detected by DHE-ROS assay and fluorescence microscopy in the present study. SNI significantly increased ipsilateral level of ROS at 2 folds (SNI ipsilateral : 24.33 ± 1.50, P < 0.0001) in SDH when compared with that at contralateral side (SNIcontralateral : 13.64 ± 1.73) (Supplementary Fig. 1), also when compared with that in control group at both contralateral side and ipsilateral side (Control contralateral : 12.33 ± 1.48, P < 0.0001; Control ipsilateral : 13.00 ± 1.14, P < 0.0001), which confirmed the previous reports [41,42,45]. Importantly, at the ipsilateral side, accompanied with the analgesic effect of targeted up-regulation of Drp1 within the SDH, the Drp1 OE treatment alleviated the increased level of SNI-induced ROS (SNI + OE ipsilateral : 13.23 ± 1.33, P < 0.0001) (Supplementary Fig. 1). At the contralateral side, the ROS level in SNI + OE group sustained at the level similar to that in SNI gruop (SNI + OE contralateral : 13.97 ± 1.21)  Fig. 1). More importantly, consistent with the invalid effect of targeted down-regulation of Drp1 within the SDH, the Drp1 RNAi treatment further expanded ROS increase by SNI (SNI + RNAi ipsilateral : 27.70 ± 2.72, P = 0.0152) ( Supplementary Fig. 1).

MitoQ and Mdivi-1 exerted analgesia effect on SNI-induced mechanical and thermal hyperalgesia
The mitochondrial-targeted antioxidant, mitoquinone (MitoQ), is able to modify mitochondrial function, and has analgesic effect on mechanical and thermal hyperalgesia in the NP induced by the chemotherapy drug vincristine (Vin) [36]. And previous studies have shown that the intrathecal treatment with mitochondrial division inhibitor-1 (Mdivi-1) could attenuate mechanical hyperalgesia [41,42]. But we were also aware of different voices. Zhang et al. [56] found that acute manipulation of Drp1 (mdivi-1) did not significantly change mitochondrial morphology, but played a causal role in maintaining mitochondrial respiration in adult cardiomyocytes by affecting endogenous Drp1, and played an important physiological role in cellular bioenergetics and ROS signal transduction. Bordt et al. [57] showed that mdivi-1 is not a specific Drp1 inhibitor, but acts by inhibiting mitochondrial complex I-dependent O 2 consumption and reverses electron transfer-mediated production of ROS. What's more, their conclusions confirmed that mdivi-1 influences multiple aspects of mitochondrial function respiration and ROS even in the absence of Drp1, and it has limited utility in studies aiming to demonstrate a specific role for Drp1-dependent fission in biological processes. So it was indicated that the analgesic effect of MitoQ and Mdivi-1 was by restoring the effective mitochondrial respiratory function against oxidative stress. However, the effects of MitoQ and Mdivi-1 on SNI-induced mechanical and thermal hyperalgesia have not been elucidated.
So it was firstly indicated there was the analgesic effect of MitoQ and Mdivi-1 on SNI-induced NP, likely by resisting against oxidative stress.

The analgesic effect of Drp1 up-regulation could persist at 28th day following SNI
In the section of 2.4.4, we presented the analgesic effect of upregulation of Drp1 on pain hypersensitivity at 14th day following SNI surgery. Here we further confirmed the beneficial effect of Drp1 upregulation on pain began significantly at 14th day, sustained till 21st day, and fad at 28th day following SNI (Supplementary Fig. 2C). In addition, the invalid effect of Drp1 down-regulation on pain persisted for 14 days, 21 days, and 28 days following SNI ( Supplementary  Fig. 2C). These results confirmed the targeted up-regulation of Drp1, instead of Drp1 down-regulation, alleviated pain symptoms in SNI mice.

Discussion
Based on the results of the current study, we proposed a model represented schematically in Fig. 11. Firstly, SNI mice showed painful and anxiety-like behaviors, which was associated with elevation of Drp1, as well as increased density of mitochondria within the pain control target, the spinal dorsal horn (SDH). Ultrastructural analysis further indicated the SDH mitochondria were sensitive to SNI stress by increasing their number, reducing their parameter and area, and transforming to be more spherical. Secondly, we showed the stable overexpression of Drp1 (Drp1 OE) alleviated SNI-induced pain behavior and abnormal gait. And, consistent with this, ultrastructural analysis indicated up-regulation Drp1 significantly improved the vacuole parameters including vac density, parameter and area, although Drp1 OE did not change mitochondrial density and their morphology. Thirdly, we showed the stable down-regulation by Drp1 RNA interference (RNAi) did not alleviate SNI-induced pain, and even deteriorated the SDH mitochondria to be more swollen and containing more vacuole structures. Our results supported that mitochondria were sensitive to NP and suggested Drp1 might be a novel therapeutic target for pain treatment.

Mitochondria in the nervous system are sensitive to NP
Our study strongly supports that mitochondria are sensitive to NP stress. The process of NP is the result of complex pathological changes caused by peripheral or central nerve injury [1,2]. Due to its complexity of pathogenesis and adverse reactions to drugs, its therapeutic treatment remains a challenge. Among all mechanistic studies, accumulating evidence has pointed to mitochondrial dysfunction [34][35][36][37][38][39][40][41][42]45]. Neurons are dependent on mitochondria, which not only provide energy to power cellular function through oxidative phosphorylation, but also regulate cellular oxidation-reduction status, calcium levels, signal transduction, and apoptosis [58]. Various stress conditions are known to quickly trigger the disruption within mitochondria [59]. Bennett GJ et al., in 2014 [60] and Dai CQ et al., in 2020 [45] reviewed the extensive evidence which supports the hypothesis that mitochondrial dysfunction in nervous system is the fundamental causative factor in NP. Our results have supported that the mitochondria within the spinal dorsal horn (SDH), the key target in pain control, responded to NP stress by changing their structure and morphology.

Mitochondrial fission factor Drp1 was sensitive to NP
Our study also supports that mitochondrial fission factor Drp1 was sensitive to NP. In the nervous system, mitochondrial dynamics including fission and fusion play critical roles in maintaining functional mitochondria when neurons are exposed to metabolic or microenvironmental stresses [16][17][18]. The dynamin family of large GTPases regulates mitochondrial membrane dynamics in all eukaryotes: Drp1 for fission, and mitofusins (Mfn1/2) and optic atrophy-1 (Opa1) for fusion [20][21][22][23]. Our own study has shown that Drp1 is distributed all over the whole nervous system, especially high at the spinal cord level [33]. In different NP model, Drp1 expression or its phosphorylated state is reported to either increase or decrease. It has been reported there was upregulation of Drp1 proteins at dorsal root ganglia (DRG) by immunohistochemical staining in the diabetic NP model [34], or by Western blot in the chronic constriction injury (CCI) NP model [39,40] proteins at DRG by Western blot CCI model [40]. Here we support Drp1 is sensitive to NP by the form of elevated expression, which might imply mitochondrial fission began to occur under the painful stress. But we did not perform the observation of phosphorylated Drp1 or activated Drp1 in other form, which should be studies in the next research.

Changes of mitochondrial morphology and vacuole ultrastructure following NP
Although many studies have deduced the mitochondrial dysfunction underlying NP, mainly based on the alternation of regulator factors involved in mitochondrial fission or fusion, such as Drp1, Mfn1, or Opa1, and so on, studying of mitochondria morphology in nervous system tissue is very limited. Our present studies are the first time to present very detailed description about mitochondrial changes in morphology and vacuole structure following SNI-induced NP. Seven important mitochondrial parameters were introduced, including: a, mitochondria (mito) density; b, mito perimeter; c, mito area; d, mito circularity; e, vacuole (vac) density; f, vac perimeter g, vac area ( Fig. 2A). We have found SNI could induce mitochondria to be in more number, and at much shorter and smaller shape. We further observed the increased size of mitochondrial vacuoles following SNI. Our results are consistent with previous reports. Nieto et al. [61] have shown that mitochondrial alterations, including enlarged size (swelling) and vacuolization, were observed in axons of saphenous nerve in paclitaxel-induced NP model. Interestingly, Jin HW et al. have reported significantly increased Fig. 11. A schematic model proposing how Drp1, the key mitochondrial fission protein, modulates neuropathic pain and associated mitochondrial dysfunction. Pain stress (SNI) resulted in increased mitochondrial division, increased vacuoles and decreased volume of SDH mitochondria in the spinal dorsal horn (control, 0.1-0.5 × 1-2 μm; SNI, 0.1-0.5 × 0.5-1.5 μm), and the mitochondria tended to be round, suggesting that mitochondrial function be impaired. After Drp1 RNAi was used to down-regulate Drp1, mitochondria became larger (0.5-1.0 × 1-2 μm) and were accompanied by a large number of vacuoles, which further aggravated mitochondrial dysfunction. When Drp1 was up-regulated by Drp1 OE, the mitochondrial size returned to normal (0.1-0.5 × 1-2 μm), accompanied by the reduction of mitochondrial vacuoles, mitochondrial function was restored, and neuropathic pain was alleviated SNI: spared nerve injury; SDH: spinal dorsal horn; OE: overexpression; RNAi: RNA interference.
incidence of swollen and vacuolated axonal mitochondria in A fibers and C fibers in paclitaxel-induced NP model [62].
We further believe the ultrastructural details are much more valuable than only focusing on the changes of expression level of Drp1. In the mitochondrial life cycle, Drp1 has been shown to induce distinct fates [63], of which one is the biogenesis of new mitochondria, the other is the clearance of dysfunctional mitochondria through mitophagy. Some studies have demonstrated ultrastructural damage to neurons. Some reports have revealed the increased number and decreased size of mitochondira in pyramidal neurons Alzheimer's disease (AD) brain [64,65]. And various Parkinson's disease (PD)-related neurotoxic molecules cause increased mitochondrial fission in vitro [66], suggesting that mitochondrial fusion and fission are involved in PD. More importantly, we have not only found SNI could induce more fragmented mitochondria within the SDH, but also observed the increased ROS, the byproduct of ATP synthesis, at the same time. So the mitochondrial dysfunction could be true in SNI. But further research is needed to detect whether or not these mitochondria exhibit decreased membrane potential [67,68] or the deficits of mitochondrial DNA (mtDNA) [59], which have been proved to be prominent features found in neuronal disorders.

ROS and NP
Reactive oxygen species (ROS) are generated by neurons, glia, and immune cells. Many previous studies have shown that ROS are involved in the hyperalgesia in NMDA-induced pain model [69] or capsaicin-induced pain model [70] at the spinal cord level, or in visceral pain model [71] at amygdala level, or in cisplatin-evoked [72] or CCI -evoked pain model [73] at DRG level, or in hypoxia/reoxygenation-evoked acute hyperalgesia [74] at microglia, or in spinal nerve ligation (SNL)-induced NP model at astrocytes [75]. Moreover, treatment of ROS donor t-BOOH could induce allodynia [76]. Previously Kallenborn-Gerhardt W [77] et al. and Grace PM et al. [78] have reviewed how the ROS is generated after NP enhances neuroexcitability in pain pathways and the potential therapeutic strategies for normalizing ROS. Our previous report [31] and the present results all indicate that increased ROS is generated at the spinal cord level following SNI pain.

Targeted up-regulation of Drp1 attenuates SNI-induced pain
In 2011, Ferrari et al. reported that [41] intrathecal administration of ASO against Drp1 could markedly attenuate neuropathic mechanical hyperalgesia caused by HIV/AIDS antiretroviral and anticancer chemotherapy in rats. In 2016, Kanda et al. [42] reported that the intrathecal Drp1 ASO could markedly attenuate neuropathic mechanical hyperalgesia caused by gp120 in rats. Our present results are totally opposite to these two results. Three main different points should be taken into consideration. Firstly, the NP model is different. What we have found is that the targeted up-regulation of Drp1 by overexpression (OE) virus delivery within the SDH could alleviate SNI-induced mechanical and thermal pain symptoms. Secondly, the animal species is different. What we have used here is mouse. Thirdly, the regulation way of Drp1 is different. We have combined the up-regulation of Drp1 by AAV overexpressing (OE) Drp1 and the down-regulation of Drp1 by AAV expressing RNAi in the present study.
We think the mechanism underlying of the beneficial effect of Drp1 OE on pain could lie in its contribution to restoring the mitochondrial morphology, especially stabilizing inner mitochondrial membrane. It might be proven that Drp1 OE could alleviate the increased number of mitochondria and elevated level of ROS (Figs. 8 and 9; video 3, 4). Mitochondrial fission is an acute and adaptive response upon injury and stress [80]. Drp1 is very important for mitochondrial fission in nervous system. Drp1-deficient mice have shown severe abnormalities in neurons both at developmental stage and post-mitotic stage [81]. Accordingly, Drp1 could be a positive regulator of neuronal survival. So the increased Drp1 in highly polarized neurons ensures the rapid transport of mitochondria to sites of synaptic terminals, where local energy production is critical. Defects in mitochondrial transport can cause local energy depletion and disruption of Ca 2+ buffering, which can ultimately trigger synaptic dysfunction and loss. So we think the mitochondrial fission is crucial in maintenance of gate control in pain stress. But the significance of mitochondrial fission in neurons following NP is still unclear.

Limitations
In the present study, we should admit there are at least two limitations. The first was that, unfortunately, Drp1 OE could not significantly up-regulate Drp1 mRNAs detected by qRT-PCR method, although Drp1 OE could significantly up-regulate Drp1 proteins detected by Western Blot. Secondly, it is well known that there are many neurons within SDH and the inhibitory GABAergic neurons releasing the inhibitory neurotransmitter gamma-aminobutyric acid (GABA) are assumed to play an important role in pain control. But our present observation did not distinguish the neuronal types. So in the future, it is necessary to use GABA associated transgenic mice to make research on mitochondrial plasticity.
In conclusion, we propose that the mitochondria at spinal cord level are sensitive to NP and the targeted up-regulation of mitochondrial fission factor Drp1 could attenuate pain hypersensitivity in SNI pain model. Despite challenges, progress in the understanding of the pathophysiology of mitochondria in NP is spurring the development of new diagnostic procedures and management for NP.

Animals
Adult male C57BL/6 J mice weighing 20 ± 2.5 g on average, were provided by the Experimental Animal Center of the Air Force Medical University, Xi'an, Shaanxi, China. The animals were randomly located under a 12-h light-dark cycle (8 a.m.-8 p.m.), with unrestricted access to food and water. All animal protocols were based on the procedures approved by the Animal Ethical and Welfare Committee for Institutional Animal Care and Use Committee (IACUC) of the Air Force Medical University (Permission NO. IACUC-20190107) and followed our institutional guidelines for the use of laboratory animals. All testings were done in a double-blinded manner.

Surgical procedures of SNI model
A model of SNI in mice was induced as described previously [6]. Mice were given pentobarbital anesthesia (50 mg/kg, i.p.) and an incision of approximately 1.5 cm was made on the superior border of the right hind limb. Then, the muscle was gently separated to reveal the main trunk of sciatic nerve and its 3 branches, which were bluntly separated. Spared nerve injury (SNI) was produced by ligating and severing a 2-4 mm segments of the common peroneal and tibia nerve (Fig. 1A). The sciatic nerve and its branches were only exposed in the control mice. After the operation, all mice were monitored for any signs of infection.

Behavioral assays
The experiments were done from 09:00 to 12:00 a.m. To minimize possible transfer influences and potential visual or olfactory effects, the mice were transferred to the testing room inside their home cages and allowed to acclimatize for at least 30 min. After each test, the surfaces of instrument were carefully cleaned with 75% alcohol to avoid interference with the scent of the previous mouse.

Mechanical hypersensitivity
Baseline behavioral assessments were conducted within 2 weeks after SNI to identify the development of NP. Using von Frey (No.37450-275, Aesthesio, Italy) monofilaments, the paw withdrawal threshold (PWT) method was used to assess the hypersensitive response of mice to mechanical stimuli. Each von Frey filament was held perpendicular to the lateral aspect of the hind paw for approximately 5 s. The left hind paw was stimulated first, then the right hind paw (5 min apart). The mechanical threshold (in grams) is defined as the first filament that causes at least three withdrawals in five applications [49]. The PWT was measured using a series of calibrated von Frey filaments before SNI (day-1) and at post-operative day (SNI) 1, 2, 3, 5, 7, 14, 21 and 28.

Thermal hyperalgesia
Reaction latencies to hot plate were measured pre-(baseline) and post-SNI to indicate thermal hyperalgesia [82]. Mice were placed on the hot plate (YLS-6B, Jinan, China) and the temperature was adjusted to 52 ± 0.5 • C. The time of the first nociceptive response (licking, flinching or jumping) of the surgical side of the paw was recorded and the animal was immediately removed from the hot plate. To avoid injury to the paw, the cut-off time was maintained at 25 s.

Open field test (OFT)
The test was conducted using a 50 × 50 cm Plexiglas square box with 50 cm high walls as previously described [83]. Mice were placed in the middle of the Plexiglas arena and could freely explore the whole arena for 15 min [84].

Elevated plus-maze (EPM)
Elevated plus-maze is made of Plexiglas and consists of 2 open arms (35 × 7 cm) and 2 opposite closed arms (35 × 7 cm), the latter surrounded by 15 cm high walls. The apparatus was raised to 50 cm above the ground. Each mouse was placed in the center of the apparatus, and allowed to move freely for 5 min [85]. Anxiety-like behavior was determined by measuring the preference of mice to open-arm for 5 min [86].

CatWalk gait analysis
Gait analysis was performed using the CatWalk XT system (Noldus, the Netherlands), which has been proven a very reliable method for measuring pain-related behaviors [87]. The mouse was placed on the open end of an enclosed glass platform and walked across the glass floor voluntarily, during which a high-speed camera underneath the device captured images of each paw and transmitted the data to gait analysis software (CatWalk XT, version 10.6; Noldus) [46,88]. In this study, eight available parameters were identified to assess the dynamic behaviors relevant to NP [88,89]: (1) regularity index as the degree of inter-limb coordination as a percentage of complete coordination; (2) stand as the duration of a paw touching the glass plate; (3) body speed is calculated by counting the distance of the mouse's body (paws) from the initial contact with the glass plate to the next contact divided by the time required to move this distance; (4) standing on three as the percentage of time spent walking with three paws; (5) print length as the length of the paw print (horizontal direction); (6) print width as the width of the complete paw print (vertical direction); (7) max contact area as the maximum print area during paw contact; and (8) max contact mean intensity as the mean intensity at maximum paw contact.

Virus injections
The AAV vector was injected into the SDH of mice as described previously [90]. Mice were anaesthetized with pentobarbital and the target vertebral column position was located using palpation. Remove the paraspinous muscle above the L4/5 vertebra and a partial laminectomy of the target side vertebral body was performed for injection. A microsyringe (RWD 79013, Shenzhen, China) was used to inject AAV with a target injection capacity of 300 nL and an injection speed of 50 nL/min. In order to inject the target into the SDH, stereotaxic apparatus (RWD 69100, Shenzhen, China) was used to move 500 nm to the side of the spinal cord, and the final depth of the tip was 300 nm. The needle was left in the spinal cord for another 10 min before it was removed. The wound was stratified with intermittent sutures and iodine disinfectant was applied to the closed wound.

Microscopic analysis
Virus-labeled Drp1 was observed in spinal cord histochemical staining sections, and co-immunostaining for DAPI (1:500; C1002, Beyotime Biotechnology, China) staining was conducted to mark the outline of sagittal spinal cord sections. The images were acquired on an Olympus IX73 (Japan)inverted microscope or an Olympus FV1200 (Japan)confocal microscope, and the images obtained were processed using the open source FIJI (NIH ImageJ, USA) software (http://fiji.sc/Fij i).

Real time-quantitative PCR (qRT-PCR)
To measure the changes of mitochondrial genes expression levels, the mice were sacrificed for qRT-PCR measurement. Spinal cords were collected and homogenized, and then Trizol reagent (Invitrogen, CA) was used to measure total RNA. The extracted RNA was used as a template to synthesize the corresponding cDNA and to perform qRT-PCR assays. The genes detected included the mitochondrial morphologyrelated genes of Drp1 (Tsingke, Beijing, China), Mfn1 (Tsingke, Beijing, China) and Mfn2 (Tsingke, Beijing, China). The internal reference is β-actin (Tsingke, Beijing, China). The primer sequences are shown in Table 1. The reaction conditions: pre-denaturation 95 • C, 15 s, followed by denaturation 95 • C, 5 s for each step, annealing, extension 60 • C for 30 s for 40 cycles. After amplification, the homogeneity of the product was checked using a lysis curve. The relative content was statistically analyzed using the 2 -△△ Ct method.

Transmission electron microscopy (TEM)
Mice were anaesthetized and perfused sequentially with 0.01 M PBS and 4% formaldehyde with 0.1% glutaraldehyde in PBS. Then, the spinal cord tissues were placed into 4% glutaraldehyde solution for tissue fixation. After fixation, each spinal cord tissue was washed twice with 0.1 M PB. This was followed by 1% osmic acid (TED PELLA, Inc. No. 18451) for staining for 2 h. Then, 50%, 70%, and 90% ethanol and 100% acetone were used respectively for dehydration, followed by acetone: embedding agent (Embed 812, DDSA, NMA, DMP-30) = 1:1 for 2 h at room temperature. Lastly, the tissue was removed from the embedding agent and stored overnight at room temperature. The following day, the tissue was placed into a special electron microscope plate and incubated at 60 • C for 48 h. A 70 nm thick section was prepared using an ultra-thin sectioning machine. The slices were placed on a copper grid and stained with lead nitrate and uranyl acetate for 10 min each. Images were recorded using a TEM (JEM 1400, Olympus, Japan). The mitochondrial morphology data were analyzed applying ImageJ-based image analysis. Finally, 30 consecutive tissue sections were selected for the TEM analysis and sections of about 1500 nm thickness were obtained.

Morphometric analysis
The size of each mitochondria was obtained by drawing the electron microscopy sagittal profile of the mitochondria in SDH. The scale was set according to the bar of the picture in ImageJ. By using the freehand selections of this software, the outline of mitochondria was carefully drawn, and then the perimeter and area of the mitochondria were measured. The circularity of mitochondrial was determined according to the method previously described [32], that is, circularity = 4πarea/perimeter 2 . Hand counted the number of mitochondria in the SDH in each image to get the total number of mitochondria, and then calculated the density of the mitochondria. The same methods were used to calculate the area, perimeter and density of the vacuoles in the mitochondria.
Three-dimensional reconstruction of mitochondria: 30 images in tif format were converted to Imarisfiles (.ims) using (ImarisFileConverter X64, Version 9.0.1, BitPlane, Switzerland). 3D reconstruction and subsequent analysis was done using Imaris software (BitPlane, Switzerland). All mitochondria surface tracings were based on the contour tracing tool and carried out manually/semi-automatically. For surface creation, a total of 35 consecutive ultrathin sections (50 nm in thickness) were traced and image data were collected by electron microscopy. The surface at maximum resolution was created to ensure that the tracing lines match the surface boundaries. This allows the 3D reconstruction of the object from 2D images.

ROS assay
To detect ROS, spinal cord tissues were promptly frozen and cut into 15-μm-thick slices at − 20 • C and then adhered to glass slides. The following steps were strictly in line with the standard protocol of manufacturer's instructions (BB-470513, BestBio, China) [55]. Briefly, the spinal cord samples were stained with ROS probe (DHE) for 30 min at 37 • C in darkness. DHE was oxidized to be ethidium oxide by intracellular ROS after permeating the cell membrane. Then DAPI (1:500; C1002, Beyotime Biotechnology, China) was applied for 30 s to sketch the outline of spinal cord. Fluorescence was detected using an Olympus IX73 inverted fluorescence microscope (Olympus, Japan). The images were observed at 610 nm wavelength.

Drug treatment
MitoQ (MCE, HY-100116 A, USA; 50 mg/kg) [92] and Mdivi-1 (MCE, HY-15886, USA; 10 mg/kg) [93] were intraperitoneally injected into the mice of SNI-induced neuropathic pain two weeks later, respectively. So did the Drp1 OE and Drp1 RNAi group. And 1 h later, all three groups of mice were tested for mechanical and thermal pain by using von Frey and the hot plate, respectively.

Statistical analysis
The statistical comparison methodologies for each study method are given in the figure legends. Quantitative data are expressed as mean ± SD. One-way ANOVA was used to assess differences between groups. Differences between groups were evaluated by one-way analysis of variance, and the Turkish post hoc test was used for multiple comparisons between groups. Statistical analyses were performed using SPSS 15.0. p < 0.05 was considered a statistically significant difference.

Conflicts of interest
The authors have declared that no conflict of interest exists.