In preclinical studies, neuropathic pain is usually induced by peripheral nerve injury through a surgical intervention (CCI, chronic constriction injury; SNI, spared nerve injury; SNL, spinal nerve injury; PSNL, partial sciatic nerve injury; SNT, spinal nerve transection).

In a model of diabetes type I peripheral neuropathy induced by injection of streptozotocin, T cells infiltrated the DRG at a very late stage. Significant presence of T cells in the DRG was not detected before 19 weeks post-injection, although mechanical pain and spontaneous pain were evidenced earlier (from 8 weeks post-injection; Agarwal et al., 2018). Interestingly, peripheral nerves from patients with diabetic neuropathy showed massive T cell infiltration of the endoneurial and epineurial compartments. In diabetic patients (type I and II) with peripheral neuropathy, approximatively 25 times more CD3+ T cells were counted per section in sural nerve biopsies compared to control patients. The infiltrated T cells were mostly CD8+ T cells and CD25+ cells, an indication of CD4+ or CD8+ Treg (Younger et al., 1996). However, the contribution of T cells to diabetic painful neuropathy has not been investigated yet.

CIPN is a common side effect of cancer treatment and is often associated with pain. The role of T cells in CIPN has been studied in models of systemic injection of chemotherapeutic agents such as paclitaxel, cisplatin, or oxaliplatin. In a model of paclitaxel-induced neuropathic pain, Liu reported that intrathecal anti-CD8 reduced mechanical allodynia on day 5 and 6 after paclitaxel. This study also showed that intrathecal injection of CD8+ T cells worsened pain hypersensitivity, while injection of Treg cells briefly reduced mechanical allodynia (Liu et al., 2014). These effect might result from the specific route of injection used here, as T cells are not present (or are at a very low level) in the spinal cords of control and CIPN animals (Janes et al., 2015; Denk et al., 2016; Gattlen et al., 2016; Krukowski et al., 2016), in contrast to the experimental autoimmune encephalomyelitis (EAE) model in which a substantial infiltration of T cells is observed in the spinal cord (Rothhammer et al., 2011; Duffy et al., 2019). These beneficial and detrimental effects of Treg and CD8+ T cells, respectively, were not reproduced in transgenic mice. Treg depletion, using the DEREG mice, did not affect pain hypersensitivity after oxaliplatin (Makker et al., 2017). In CIPN induced by either paclitaxel or cisplatin, Rag1^−/− or Rag2^−/− female and male mice develop mechanical allodynia with similar intensity to WT mice. Strikingly, the resolution of chemotherapy-induced mechanical allodynia was significantly delayed in the absence of T cells (Krukowski et al., 2016; Laumet et al., 2019). Reconstitution with CD8+, but not CD4+, T cells restored the resolution of mechanical allodynia (Krukowski et al., 2016). While most studies cited above focus only on evoked-pain behaviors, our studies showed that the absence of T cells also impairs the resolution of spontaneous pain assessed by conditioned place preference, and reconstitution with CD8+ T cells normalized the resolution of spontaneous pain (Laumet et al., 2019).

Interestingly, the adoptive transfer of CD8+ T cells into T-cell-deficient mice after CIPN had fully developed failed to promote resolution of pain (Laumet et al., 2019). These findings indicate that the CD8+ T cells have to be exposed to cisplatin in order to be capable of promoting resolution of CIPN. In other words, the CD8+ T cells need to be “educated” to acquire the capacity to promote resolution of CIPN by exposure to cisplatin. In support of this idea, adoptive transfer of CD8+ T cells from cisplatin-treated WT mice did indeed promote resolution of established CIPN in Rag2^−/− mice. This T cell education appears to be independent of antigen recognition by the TCR because reconstitution of Rag2^−/− mice with CD8+ T cells with mutated TCRs that recognize and respond only to one irrelevant antigen (chicken ovalbumin) retained the capacity to induce CIPN resolution (Laumet et al., 2019). Interestingly, the neuroprotective effects of T cells after brain trauma was also independent of antigen recognition by the TCR (Walsh et al., 2015).

Inflammatory pain is modeled by injection of Complete Freund’s adjuvant (CFA), formalin, or other inflammatory agents into the paw. In response to intraplantar CFA injection, immune cells (CD45+) infiltrate the paw. T cells represented 2%–4% of infiltrated immune cells, and their percentage remain unchanged over in the first 96 h (Rittner et al., 2001) but showed significant increases after 7 days that are maintained for at least 14 days (Ghasemlou et al., 2015). After CFA, the severity of mechanical allodynia was identical in WT and in five different strains of T-cell-deficient mice (nude, Tcrb^−/−, Tcrd^−/−, Rag1^−/− and Rag2^−/−; Ghasemlou et al., 2015; Sorge et al., 2015; Laumet et al., 2016, 2019; Petrović et al., 2019). These data indicate that inflammatory pain hypersensitivity in the CFA model develops independently of T cells. While the onset and severity of inflammatory allodynia are similar between WT and T-cell-deficient mice, several independent studies reported that the resolution of mechanical allodynia was significantly delayed in T-cell-deficient mice (Boué et al., 2011, 2012; Basso et al., 2016; Laumet et al., 2016). Reconstitution of Rag1^−/− or Rag2^−/− mice with WT T cells normalized resolution of CFA-induced mechanical allodynia. Similar findings were obtained after intraplantar injection of IL-1β (Kavelaars lab, unpublished data). The pain behavior in response to formalin was worsened in nude mice compared to WT, and reconstitution of nude mice with CD4+ T cells normalized their response to formalin in both sexes (Rosen et al., 2019). In the antigen- and collagen-induced models of arthritis, CD8+ T cell depletion worsened the pain hypersensitivity (Baddack-Werncke et al., 2017), while in a postsurgical pain model, no alteration in thermal and mechanical hypersensitivity was reported in Tcrb^−/− and Tcrd^−/− mice compared to WT mice (Ghasemlou et al., 2015; Petrović et al., 2019). The lack of contribution of γδ T cells to inflammatory pain induced by plantar incision was reported in both sexes (Petrović et al., 2019), and these mice deficient in γδ T cells have a normal pattern of αβ T cells. In conclusion, the existing literature indicates that, in inflammatory pain models, T cells are beneficial or neutral to the pain phenotype.

Like most of the preclinical research in pain (Mogil, 2012), the role of T cells has been almost exclusively studied in male rodents, but recent evidence suggests that T cells may contribute to sex differences in pain signaling. Key studies in this area showed that inhibition of microglia relieved nerve injury-induced pain only in male mice (Sorge et al., 2015; Taves et al., 2016; Luo et al., 2018). Critically, this sex difference disappeared in T-cell-deficient mice (Rag1^−/− and nude mice; Sorge et al., 2015; Mapplebeck et al., 2018). Moreover, a beneficial role of T cells became apparent when comparing pregnant WT and T-cell-deficient mice. In late pregnant WT mice, CFA- and SNI-induced allodynia are suppressed, but this does not happen in T-cell-deficient mice (Rag1^−/− and nude mice). Adoptive transfer of CD4+ T cells restored pregnancy analgesia (Rosen et al., 2017). T cells are also responsible for the reduced morphine analgesia observed in female mice, and this sex difference in morphine analgesic sensitivity was restored by adoptive transfer of male CD4+ T cells to female nude mice (Rosen et al., 2019). Notably, however, no sex difference was observed in the contribution of CD8+ T cells to CIPN resolution (Laumet et al., 2019). In summary, these data indicate complex interactions between T cells and sex in pain signaling, although the physiology of these interactions remains to be uncovered.

The pain promoting effect of T cells may result from amplification of neuroinflammation (Figure 3). For example, it has been proposed that Th1 and Th17 cells facilitate macrophage infiltration in the damaged nerve and DRG (Kleinschnitz et al., 2006; Kobayashi et al., 2015). In the spinal cord, Cd4^−/− mice showed less astrocyte activation at 14 days after SNT (Draleau et al., 2014). In the injured nerve, the infiltrated T cells (Th17 cells) produce IL-17, and this may contribute to microgliosis via stimulation of the IL-17 receptors expressed on microglia (Kleinschnitz et al., 2006). Consistent with this model, inhibition of IL-17 signaling reduced microgliosis, mechanical allodynia, and paw flinches associated with bone cancer pain (Huo et al., 2019).

In addition to cytokines, T cells produce the serine protease leukocyte elastase (LE, encoded by the gene Elane). LE is released by infiltrated T cells in the DRG after SNI, and it activates matrix metalloprotease 9 (MMP9) which facilitates neuropathic pain (Ferry et al., 1997; Ji et al., 2009). To assess the critical role of LE-producing T cells in neuropathic pain, Vicuña et al. (2015) reconstituted Rag2^−/− mice with T cells from WT or Elane^−/− mice and monitored their pain sensitivity following SNI. The lack of LE in the T cells prevented the development of neuropathic pain.

Following nerve injury, infiltrated CD4+ T cells in the dorsal horn of the spinal cord are often associated with increased pain (Cao and DeLeo, 2008; Costigan et al., 2009; Leger et al., 2011). Therefore, a potential therapeutic strategy may be to target the infiltration of the CD4+ T cells into the spinal cord. Repurposing drugs that have been developed to block the infiltration of T cells in the central nervous system in multiple sclerosis may be an attractive strategy to treat neuropathic pain induced by nerve injury. FTY720, a drug used to treat multiple sclerosis, sequesters T cells in the lymph nodes and prevents the infiltration of the nervous system. After PSNL, FTY720-treated mice showed less mechanical and thermal pain sensitivity compared to vehicle-treated mice (Kobayashi et al., 2015). An important caveat is that FTY720 may also reduce pain by mechanisms independent of T cell sequestration (Doyle et al., 2019). Approaches based on blocking α4 integrin to prevent the infiltration of CD4+ T cells into the dorsal horn of the spinal cord are attractive as well (Yednock et al., 1992; Rothhammer et al., 2011), though such antibodies have not yet been tested in chronic pain models. An alternative way to prevent the infiltration of pathogenic CD4+ T cells into the DRG and spinal cord is through surgical sympathectomy (McLachlan and Hu, 2014; Du et al., 2018). Surgical sympathectomy is effective at alleviating neuropathic and inflammatory pain (Agarwal-Kozlowski et al., 2011; Iwase et al., 2012; Xie et al., 2016), but whether this pain relief resulted from blocking T cell infiltration is unknown.

Recent studies indicate that T cells also promote the resolution of pain and prevent the transition from acute to chronic pain (Figure 3). The pathways triggered by T cells to resolve pain are not fully understood, but some mechanisms have been elucidated. The subsets of Treg cells, Th2 cells, and suppressor CD8+ T cells have been shown to reduce or resolve pain, and this is likely through their capacity to switch the milieu to an anti-inflammatory environment (Moalem et al., 2004; Austin et al., 2012; Lees et al., 2015; Baddack-Werncke et al., 2017). Importantly, promoting the anti-inflammatory activity of T cells can be achieved by activation of the anti-inflammatory reflex via electrical vagus nerve stimulation (Chakravarthy et al., 2015), suggesting a possible translational treatment.

Many neuroprotective and pain resolving effects of CD8+, Th2 and Treg cells could be recapitulated by IL-10 administration and are absent in mice lacking IL-10, pointing to IL-10 as a major player in the beneficial effects of T cells (Frenkel et al., 2005; Liesz et al., 2009; Xie et al., 2015; Krukowski et al., 2016; Laumet et al., 2018; Duffy et al., 2019). IL-10 alleviates inflammation and pain in various chronic pain models (Wagner et al., 1998; Plunkett et al., 2001; Eijkelkamp et al., 2016; Krukowski et al., 2016), and it is possible that T cells act through IL-10 production. However, it is also possible that T cells do not produce IL-10 themselves but induce other cells to synthesize and release IL-10 (Xin et al., 2011; Krukowski et al., 2016). Resolution of mechanical allodynia was similar in Rag1^−/- reconstituted with WT or Il10^−/− CD8+ T cells, indicating that CD8+ T cells were not the source of the IL-10 required for resolution of pain (Krukowski et al., 2016). Likewise, in models of nerve injury and inflammation-induced depression-like behavior, CD4+ and CD3+ T cells conferred neuroprotection and facilitated resolution by inducing IL-10 production from CNS-resident cells (Xin et al., 2011; Laumet et al., 2018). After spinal cord injury, Th1 cells secrete IFNγ to trigger IL-10 production by macrophages and microglia which will promote resolution of motor deficits (Ishii et al., 2013). Alternatively, in models of acute systemic inflammation, Treg secrete IL-13 to induce IL-10 production by IL-13R+ macrophages (Proto et al., 2018). Thus, how T cells induce the production of IL-10 to resolve pain is not yet understood.

In addition to cytokines, T cells release endogenous opioids to induce analgesia (Kavelaars et al., 1991; Kavelaars and Heijnen, 2000; Sitte et al., 2007; Labuz et al., 2010; Celik et al., 2016; Basso et al., 2018). Endogenous opioids can bind opioid receptors on sensory neurons to dampen pain signaling (Stein et al., 1990, 2003; Labuz et al., 2010). The mRNAs of proenkephalin (encoding the enkephalins) and proopiomelanocortin (encoding the endorphins) can be induced in T cells (Kavelaars et al., 1991; Kavelaars and Heijnen, 2000; Labuz et al., 2010; Boué et al., 2014; Basso et al., 2016). Ex vivo, T cells from mice immunized with ovalbumin in CFA produce up to seven time more proenkephalin Penk mRNA in response to antigen stimulation than naïve CD4+ T cells (Boué et al., 2011, 2012). In vivo, T cells have a critical role in stress-induced analgesia, which is known to be mediated by endogenous opioids. Restraint stress-induced analgesia was absent in athymic nude mice and reduced in WT mice after T cell depletion (Labuz et al., 2006; Rosen et al., 2019). The release of endogenous opioids by T cells during stress-induced analgesia was partly dependent on the receptor for corticotropin-releasing factor (CRF; Labuz et al., 2010). The analgesic effects of T cell-producing endogenous opioids have been investigated in models of chronic pain as well. Infiltrated T cells and other leukocytes in the damaged nerve produce and release opioid peptides (Labuz et al., 2009). Interestingly, while T cells may represent only 11% of infiltrated leukocytes in injured nerves, they constituted approximately 50% of opioid peptide-containing immune cells (Labuz et al., 2010). As mentioned above, pregnancy analgesia (reduced pain sensitivity in the SNI and CFA models in late pregnant mice) was absent in T-cell-deficient mice (Rag1^−/− and nude) and was restored after adoptive transfer of T cells (Rosen et al., 2017). Rosen et al proposed that T cells promote pregnancy analgesia because they induce upregulation of the oprd1 expression (δ Opioid Receptor, δOR) in the spinal cord. Indeed, the lack of oprd1 impaired pregnancy analgesia (Rosen et al., 2017). Similarly, in the CFA model, δOR (but not μOR or κOR) antagonist blocked the endogenous analgesic effect of T cells (Boué et al., 2012). In chronic inflammatory pain models, both CD4+ and CD8+ T cells contribute to endogenous opioid-dependent analgesia and pain resolution (Boué et al., 2011, 2012, 2014; Baddack-Werncke et al., 2017). In contrast to WT T cells, adoptive transfer of T cells from Penk^−/− mice did not induce resolution of CFA-induced allodynia, suggesting that T cells promote resolution of inflammatory pain by enkephalin release (Basso et al., 2016). Notably, T-cell-derived enkephalins increase the number of Th2 cells and reduced the numbers of Th1 and Th17 cells (Boué et al., 2014; Basso et al., 2018). These findings indicate that in addition to their direct analgesic effects, endogenous opioids released by T cells may also suppress pain via their anti-inflammatory effects.

Beside the role of T cells in endogenous analgesia, T cells play a role in pain relief induced by exogenous opioids. T-cell-deficient mice (Rag1^−/−, nude and Cd4^−/− mice) showed reduced morphine analgesia in the formalin and tail-withdrawal tests. Reconstitution with CD4+ but not CD8+ T cells restored morphine analgesia (Rosen et al., 2019). T cell-mediated endogenous analgesia is stimulated by administration of exogenous opioids, as T cells increase the production and release of endogenous opioids in response to exogenous opioid (Labuz et al., 2006; Boué et al., 2012; Celik et al., 2016; Figure 4). Finally, administration of synthetic opioid agonists in the damaged nerve produces analgesia which is dependent of infiltrated leukocytes [as mentioned above, 50% of opioid-producing leukocytes are T cells (Labuz et al., 2010)].

There are a variety of potential pathways to promote a pro-resolution phenotype in T cells, many of which are sensitive to existing compounds. T cell subsets are not stable and can be “re-fated” upon appropriate stimulation. For example, Th17 cells naturally acquire an anti-inflammatory phenotype to then become IL-10-producing Tregs to resolve inflammation in various models of chronic inflammation (Gagliani et al., 2015). This plasticity presents an attractive therapeutic opportunity to switch pain promoting Th1 and Th17 cells to a phenotype that promotes resolution of pain and inflammation such as Treg or Th2 cells.

Glatiramer acetate (GA), a drug with good safety profiles and tolerability used to reduce the frequency of multiple sclerosis relapse, has immunomodulatory properties (Dhib-Jalbut, 2003; Arnon and Aharoni, 2004; Blanchette and Neuhaus, 2008). GA increased the number of IL-10-producing CD4+ T cells in the dorsal horn of spinal cord, reduced the activation of microglia, and alleviated allodynia in models of inflammatory and neuropathic pain (Sharma et al., 2008; Leger et al., 2011).

Experimentally, Treg response can be amplified by treatment with the superagonist of the B7 receptor for co-stimulation: CD28 (supCD28). In the CCI model, supCD28 administration expanded the number of Treg cells in the injured sciatic nerve and spinal cord. SupCD28-stimulated Tregs reduced the number of macrophages in the sciatic nerve and the DRG and decreased astrocyte and microglia activation in the spinal cord as well. SupCD28 did not affect the onset of CCI-induced mechanical allodynia but accelerated its resolution (Austin et al., 2012).

Another way to stimulate the pro-resolution T cell pathway could be via vaccination with CNS-restricted self-antigens (Schwartz and Moalem, 2001). After axotomy, immunization with myelin-derived peptide (myelin oligodendrocyte glycoprotein: MOG) stimulated neuron survival by recruiting autoreactive T cells to the site of injury (Moalem et al., 1999; Hauben et al., 2000a,b). The beneficial effects MOG immunization may rely on IL-10 producing CD4+ T cells (Frenkel et al., 2005). However, despite the high incidence of chronic neuropathic pain after nerve injury, the immunization strategy has not yet been tested in chronic pain models.

CD8+ T cells are mostly beneficial in animal models of inflammatory pain and CIPN (Krukowski et al., 2016; Baddack-Werncke et al., 2017). As described above, in order to resolve CIPN, CD8+ T cells need to be educated. Interestingly, adoptive transfer of educated CD8+ T cells before chemotherapy prevented the development of pain in response to cisplatin or paclitaxel treatment (Laumet et al., 2019). If we can develop ways to educate CD8+ T cells in vitro to promote resolution of pain, one could envision that CD8+ T cells from a patient with CIPN can be educated ex vivo to acquire a pro-resolution phenotype and be re-injected as an autograft to the same patient to treat CIPN (Figure 5).