Acupuncture is a method of stimulating acupoints or parts of the body with needles to prevent and treat disease. Electroacupuncture is a method in which needles are inserted into acupoints to Deqi, and then low-frequency pulse currents are introduced to stimulate and adjust the Qi of the meridians using both needle and electrical stimulation to prevent and treat diseases. Transcutanclus electrical acupoint stimulation (TEAS) is a method that combines transcutanclus electrical nerve stimulation (TENS) with acupoints to deliver specific low-frequency pulse currents through the skin to the body to treat pain. Acupuncture, Electroacupuncture and TEAS are the most commonly used acupuncture therapies for pain. Therefore, we chose them as the keywords for searching.

PubMed database was surveyed for the studies published from January 2017 to December 2021, by using the following keywords: [“acupuncture” or “electroacupuncture” or “transcutaneous electrical acupoint stimulation (TEAS)”] and [“analgesia” or “pain”]. The search formula used was: ((((acupuncture) OR (electroacupuncture)) OR (transcutaneous electrical nerve stimulation)) OR (TENS)) AND ((Analgesia) OR (pain)) AND (2017/1/1:2021/12/31[pdat]).

References that met the inclusion criteria were manually selected using Excel software before a close reading of the full text. Of 5,857 studies, 4,383 articles were not related to acupuncture and analgesia, 41 articles did not include full text, 740 were clinical research articles, 519 were reviews or meta-analyses, and 50 articles were not related to mechanistic studies, all of these were excluded. From the remaining 124 basic studies 75 articles were non-purely neuropathic pain such as inflammatory pain, cancer pain, and muscle or fascial pain were also excluded. In the end, 49 articles were selected for this review. The flow chart of the search process is shown in Figure 1.

For the mechanism studies, design data were extracted and classified using a predefined data extraction table that specified the type of neuropathic pain model, intervention (method, acupuncture point, acupuncture parameters), and outcome measures (pain-related behaviors, mechanism indicators). One author extracted the data and the other cross-checked it. Due to the similarity of some studies, the table enlists information only from some representative and recently published studies (Table 1).

NP can be divided into peripheral neuropathic pain, central neuropathic pain or mixed (peripheral and central) neuropathic pain, which is caused by damage to the peripheral or central nervous system. Commonly used animal models for NP include central nervous models such as spinal cord injury (SCI), spinal nerve ligation (SNL), and peripheral nervous models such as chronic constriction injury (CCI), spared nerve injury (SNI), brachial plexus avulsion injury (BPAI), diabetic neuropathic pain (DNP), postherpetic neuralgia (PHN), neck-incision pain, chemotherapy-induced peripheral neuropathy (CIPN) and recurrent migraine. These animal models provide a good guarantee for the smooth mechanism study of acupuncture in the treatment of NP.

Upon reviewing animal studies, it was noticed that 29 studies used ST36 (Zusanli) as the site of acupuncture intervention, 15 used GB34 (Yanglingquan), nine used BL60 (Kunlun), eight used SP6 (Sanyinjiao) and PC6 (Neiguan), five used GB30 (Huantiao) LI18 (Futu) and LI4 (Hegu), three used GB20 (Fengchi), two used PC4 (Ximen), one used GV26 (Shuigou), LR3 (Taichong), GV20 (Baihui), GV14 (Dazhui), SJ1 (Guanchong), BL13 (Feishu), BL20 (Pishu), BL23 (Shenshu), BL25 (Dachangshu), BL40 (Weizhong), ST2 (Sibai), JLSXX-QX (Jiachengjiang), C5–C7 Jiajixue, T10–T12 Jiajixue, L4 and L6 Jiajixue and L5 Jiajixue (Figure 2). These acupoints were selected mainly from the proximal and distal parts of the pain site, which is consistent with the principles of clinical acupuncture point selection. In EA, the following frequencies are most commonly applied: low (2, 10 Hz), high (100 Hz), and variable frequency (2/100, 2/15, 2/15/50 Hz), with the stimulation intensities of 0.5, 1, 1.5, 2, 3, 6–9 mA and stimulation times of 10, 15, 20, 30 min, with 30 min being the most frequently used stimulation time. The selection of EA parameters for animal models is similar to that applied in clinical treatments, which allows a better clinical translation of animal experimental results.

When neurons are damaged, they will lead to the release of various factors such as glutamate, substance P (SP), calcitonin gene-related peptide (CGRP), adenosine triphosphate (ATP), etc. Primary injury sensing neurons transmit excitatory signals to the SDH and interact with their receptors, i.e., glutamate receptors (NMDAR, AMPAR, mGluR), SP receptors (NK1R), calcitonin receptor-like receptors (CRLR) and purinergic receptors (P2XR, P2YR), resulting in nociceptive hyperalgesia. These factors also alter the neuronal phenotypes, such that Aβ fibers (tactile neuronal fibers) shift from transmitting non-injurious signals to transmitting injurious signals; simultaneously abnormal neurons firing also affects surrounding non-damaged neurons, causing them to fire abnormally as well, thereby further transmitting injurious signals to the center.

Glutamate is the most common excitatory neurotransmitter in the CNS and plays a key role in the upstream excitatory transmission system with the following three receptors; NMDAR, α-amino-3-hydroxy-5-methyl-4-isoxazole-propionicacid receptor (AMPAR), and metabotropic glutamate receptors (mGluR) (70). Synaptotagmin 1 (Syt-1) is a synaptic vesicle protein which can regulate neurotransmitter release. In an SNI model, EA at bilateral ST36 and SP6 significantly reduced the expression of Syt-1 in the L1–L2 spinal cord, further inhibited neurotransmitters (such as glutamate) release, mediating analgesic effects of EA (27). The 2-Hz EA at ST36 and SP6 reduced glutamate in the rostral anterior cingulate cortex (rACC)-ventral lateral periaqueductal gray (vlPAG) loop in SNI mice model and revealed anxiety-like behavior and nociceptive hypersensitivity (32). In GV20 and GV14 EA in the CCI rat's model, hippocampal glutamate reduction was observed which produced an analgesic effect on NP (23). In addition, glutamate receptor expression plays an important role in EA analgesia. EA significantly reduces phosphorylation of calcium/calmodulin-dependent protein kinase II (CaMKII) in the spinal cord, inhibits voltage-dependent Ca²⁺ channels, and glutamate receptor NR2B (N-methyl-D-aspartic acid receptor 2B, NMDAR subunit) to reduce abnormal pain and nociceptive hyperalgesia in CIPN (54). Another study reported that EA at ST36 and SP6 acupoints in the CCI rats' model reduced NR2B in the L4–5 spinal cord, thereby exerting an analgesic effect (21). Thus, EA plays an analgesic role in NP by reducing glutamate release, inhibiting glutamate receptor expression, and suppressing the upstream excitatory system.

ATP not only provides energy to cells but also acts as a neurotransmitter in the PNS and CNS and is an important extracellular signaling molecule. In many pathological conditions, ATP is released from different cells and stimulates P2X/P2Y receptors and participate in pathological pain processes. P2X are ligand-gated cation channel receptors and P2Y receptors are the metabotropic receptors. P2X receptors bind to extracellular ATP, allow the permeation of cations such as Na⁺, K⁺, and Ca²⁺, cause depolarization of the membrane potential, and play an important role in the generation and transmission of information regarding injury (71). Each P2X receptor isoform has two transmembrane structural domains as well as a large glycosylated and disulfide-rich extracellular structural domain, with the ATP binding site located outside the cell. The homotrimer P2X3 receptor is mainly located in primary sensory neurons (small and medium-sized) which respond to injury and their peripheral nerve endings, suggesting that this receptor subtype is associated with the generation and transmission of nociceptive information (72). Studies have shown that the EA at ST36 and BL60 in SNL and DNP models can reduce P2X3 receptors in L4–6 DRGs (37, 43, 44, 46). EA also reduces P2X4 (36, 45) and P2X7 receptors (38) in L4–6 DRGs, inhibits excitatory postsynaptic potentials, neural loops excitability from the periphery to the spinal cord, abnormal dendritic/spinous synaptic reconstructions and inflammation to improve NP. EA modulates three main types of P2X receptors: the P2X3 receptors located in sensory neurons themselves, the P2X4 receptors located in microglia, and the P2X7 receptors located in glial cells (microglia, astrocytes, and oligodendrocytes) (73). Therefore, EA can regulate the function of neurons directly through P2X3Rs, or indirectly through glial neuron crosstalk by P2X4/P2X7Rs pathway. P2XR is mainly regulated in the DRGs.

SP and CGRP are neuropeptides released from sensory nerve endings and are the main mediators of neurogenic inflammation. In the neck-incision pain model, EA at acupoints: LI18, LI4-PC6, or ST36-GB34 decreases SP (49, 50) and CGRP (49), and at acupoint C3–6 it acts on DRGs to block the pain signals transmitted to the CNS. In a recurrent migraine model, EA at GB20 acupoint ameliorated migraine and associated abnormal cutaneous pain by reducing CGRP in the trigeminal ganglion (TG), the caudal nucleus of the trigeminal nucleus caudalis, and the ventroposterior medial thalamic nucleus of the thalamus (60).

Additionally, EA also reduces transient receptor potential ion channel vanilloid 1 (TRPV1) (26), anexelekto (AXL), p-AXL (35), and hyperpolarization-activated cyclic nucleotide-gated (HCN) in L4–6 DRGs (20) in gasserian ganglion (GG), which inhibits excitatory postsynaptic potentials and suppresses the upstream excitatory transmission system for analgesic effects.

All the above studies suggest that acupuncture produces analgesic effects by downregulating neurotransmitters and receptors such as glutamate and its NMDA receptors, P2XR, SP, CGRP, and ion channels (TRPV1 and HCN), which inhibit the upstream excitatory system and weaken the neuronal transmission efficiency.

When central inhibitory neurons die, they lose their inhibitory function and produce bursts of neuronal discharge, resulting in central NP. Decreased concentrations of inhibitory transmitters such as opioid peptides, GABA, and endogenous cannabinoids in the spinal cord and their reduced binding sites are also important causes of central NP.

Activation of opioid peptides (e.g., enkephalins, β-endorphins) and opioid receptors such as μ-opioid receptors (MOR) by acupuncture is a more recognized analgesic mechanism, which helps to inhibit the release of excitatory transmitters such as glutamate in the downstream pathway. EA suppresses NP by promoting the β-endorphins secretions from spinal microglia, spleen, and lymph nodes (34, 42). TEAS at GB34, GV26, and ST36 in CCI rats model upregulates the expression of MORs in L3-L5 DRGs (14), thereby activating the downstream inhibitory system. EA at GB30 and GB34 acupoints in PHN rats upregulates MORs and UNC5H2 in L4-L6 DRGs and downregulates DCC and Netrin-1, changes spinal cord dorsal horn's growth-permitting environment into an inhibitory environment, reducing RTX-induced primary afferent nerve sprouting, thus, acting as an analgesic (47).

GABA (inhibitory neurotransmitter) reduces pain perception through presynaptic inhibition (74). Studies have shown that EA promotes GABA and its receptors (GABA-AR, GABA-BR) in the spinal cord, DRG, and periaqueductal gray matter (PAG), inhibits excitatory neurotransmitters such as glutamate further reduces abnormal pain and nociceptive hyperalgesia in NP (22–24, 36, 50, 52). EA promotes the expression of glutamic acid decarboxylase 67 (GAD67) in C3-C6 DRGs of rats with neck-incision pain models. GAD67 up-regulates GABA expression and the upregulation of GABA inhibits excitatory postsynaptic potentials and reduces pain. Additionally, EA increases GABAergic somatostatin-positive interneurons, inhibits excitatory pyramidal neurons and vasoactive intestinal peptide-positive interneurons, and exerts analgesic effects which is dependent on the activation of endogenous cannabinoid receptor 1 (CB1R) (25). In neck-incision pain model rats, EA at LI18, LI4-PC6, and ST36-GB34 upregulates CB1 and CB1R in the C2–C5 spinal cord, and thus produces an analgesic effect (51). Alpha-7 nicotinic acetylcholine receptor (α7nAChR) is an ion channel receptor that is involved in the regulation of neurotransmitters such as GABA and 5-HT. It has been shown that EA upregulated α7nAChR in the L4–L6 spinal cord, thereby, upregulating GABA, 5-HT, etc. expression to reduce SNI-induced mechanical hypersensitivity (29).

5-HT is a very important neurotransmitter in living organisms. Under various psychological and physiological stresses, it can be accompanied by changes in synthesis and metabolism of 5-HT in the brain. Abnormal function of 5-HT in the CNS is closely related to pain, anxiety and cognition, and plays a biphasic regulatory role in the downward promotion and inhibition pathway. EA upregulates 5-HT 1A (54) and downregulates 5-HT7R to regulate PKA and ERK1/2 in the TG and trigeminal nucleus caudalis (TNC), mediating the antinociceptive hyperalgesic effects for migraine (58, 59).

Recent studies have shown that the reward effects of pain relief of EA analgesia are associated with the activation of hypothalamic appetite-modifying neurons (17, 31). EA activates appetitin neurons in the PAG, and the released appetitin then activates postsynaptic OX1Rs in the PAG, a Gq protein-coupled receptor that activates phospholipase C (PLC) to generate diacylglycerol (DAG), which can be converted to 2-arachidonic acid glycerol (2-AG; an endogenous cannabinoid, provided by DAGL). 2-AG then crosses the synapse in a retrograde manner and activates presynaptic CB1R, which stimulates the downstream pain inhibitory pathway, thus, producing analgesia. This disinhibition mechanism in vlPAG is mediated by the OX1R-PLC-DAGL-2-AG-CB1R cascade.

In conclusion, acupuncture upregulates opioid peptides such as β-endorphin and MOR receptors, GABA and its receptors, and bi-directionally regulates 5-HT and its receptors (upregulates 5-HT 1A and downregulates 5-HT7R) to activate the downstream pain inhibitory pathway and produce analgesic effects. Hypothalamic appetite-modifying neurons also play an important role in the analgesic effect produced by EA.

Neuro-inflammation is a critical cause of NP development, especially, PNS and CNS neuroinflammation caused by glial cell activation, which mediates the development and maintenance of NP (67–69). Microglia and astrocytes are the main immune cells of CNS, act as cell membrane receptors to respond to different neurotransmitters, interact with neurons to induce neuro-inflammation, and ease the transmission of pain signals. This response mechanism contributes to the adaptation of the CNS to changes in the internal environment mediated by injurious stimuli, leading to long-duration sensitivity and chronic pain in peripheral and central nociceptive neurotransmission pathways.

In the CCI model, EA can decrease central neuro-inflammation to reduce nociceptive hyperalgesia in NP by inhibiting TNF-α expression in the PFC, hippocampus, amygdala, and hypothalamus (13). The inhibition of JAK2/STAT3 (30) and PI3K/mTOR (39) pathways in spinal cord by EA plays an important role in the inhibition of central neuro-inflammation. In the CIPN model, EA inhibits L4–6 DRGs' TLR4 and MyD88 expression, TRPV1 expression, channel functional activity, and spinal glial cell activation (55). Acupuncture analgesia is mainly associated with the inhibition of microglia and astrocyte activation (19, 56). Microglia are the most important glial cells. EA downregulates microglial chemokine receptor (CX3CR1) expression, inhibits microglia activation, and reduces central neuro-inflammation pain by inhibiting TLR4/MyD88 co-expression and NF-κB p65 activation (18, 28, 33). Some findings suggest that inhibition of microglia activation by acupuncture may be associated with reduced cyclooxygenase-2 (COX-2) expression in spinal cord (40).

When astrocytes are active in the spinal cord segment, neuroinflammation can occur. EA can promote the upregulation of A1Rs on astrocytes and inhibits the active state of astrocytes, thus reducing the release of inflammatory substances such as GFAP and TNF-α, producing an analgesic effect (53). In the CCI model, EA at the ST36 acupoint increased the expression of L4–6 spinal adenosine and A1Rs (16), inhibited astrocytes activation, and played an important role in the analgesic effect for NP.

Therefore, EA inhibits the JAK2/STAT3 and PI3K/mTOR pathways, downregulates the microglial CX3CR1 and upregulates astrocytes A1Rs, inhibits glial cell activation, reduces inflammatory substances such as TNF-α, suppresses neuroinflammation, and thus has an analgesic effect.

Recently it has been revealed that the analgesic effect of acupuncture for NP may be related to metabolic regulation. EA at GB30 and GB34 significantly reversed increase in glucose metabolism induced by CCI and decreased the protein expression of glucose transporter 3 (GLUT-3) in the left medial Prefrontal Cortex (mPFC), suggesting that the analgesic effect of EA may be related to the inhibition of glucose metabolism and GLUT-3 expression in the mPFC (12). In a study involving the BPAI model, the brain metabolic connections between the bilateral hemisphere: somatosensory cortex (SC), the motor cortex (MC), caudate putamen (Cpu), and dorsolateral thalamus (DLT) were decreased compared with the normal animals. EA increased the strength of brain metabolic connections between the aforementioned regions at the 4th and 16th weeks, indicating that the regulation of brain metabolic connections may be an important mechanism that produces the analgesic effect by EA for the treatment of NP (41). Therefore, the mechanism of analgesic action of acupuncture in the treatment of NP may be elucidated from the perspective of metabolic regulation, and a new understanding may be obtained.

When there is an oxidative/antioxidative imbalance, excess oxidative stress products can directly damage cell membrane structures, and organelles, and increase membrane permeability, leading to cellular dysfunction or even autolysis. A study investigated the analgesic effect of EA on paclitaxel-induced NP and found that under oxidative stress, EA upregulated the expression of Nrf2 antioxidant response element (Nrf2-ARE) and superoxide dismutase in the DRG, enhanced antioxidant signaling pathways, and inhibited oxidative stress products NOX4 and 8-iso-prostaglandin F2alpha (8-iso PGF2α), thereby attenuating paclitaxel-induced NP (57). Therefore, the regulation of EA on the level of oxidative stress of neurons may also be an important way to inhibit NP.