Contribution of TREM2 Signaling to the Development of Painful Diabetic Neuropathy by Mediating Microglial Polarization in Mice

Background: Painful diabetic neuropathy (PDN) is a common and intractable complication of diabetes mellitus, with little effective treatment. PDN has been associated with spinal neuroinammation characterized by microglial activation. Recently, the triggering receptor expressed on myeloid cells 2 (TREM2), specically localized on microglia, has been identied as a vital factor in modulating neuroinammation and microglial phenotypes in neural diseases. Therefore, we hypothesized that spinal TREM2 might contribute to PDN and neuroinammation by regulating microglial activity and phenotypes. Methods(cid:0)Type I diabetes mellitus was elicited by a single intraperitoneal administration of streptozotocin (STZ) in mice. The pain behaviors were reected by paw mechanical withdrawal thresholds (PMWT) and thermal withdrawal latency (PTWL). Results(cid:0)We demonstrated that up-regulation of microglial TREM2 and amplication of both microglial M1 and M2 response was along with the presence of diabetes-related mechanical allodynia and thermal hypersensitivity. Moreover, we found that overexpression of TREM2 in microglia aggravated the symptom of PDN, amplied microglia M1 response, and suppressed microglia M2 polarization in the lumbar spinal cord of diabetic mice. However, inhibition of TREM2 with anti-TREM2 neutralizing antibodies attenuated mechanical allodynia and thermal hyperalgesia in diabetic mice. Besides, we identied Galectin-3 (GLT-3) as the potential ligand of the TREM2 receptor in facilitating the progression of PDN. Conclusions: TREM2 could be a critical microglial membrane molecule that modulates microglial phenotypes pain hypersensitivity in PDN. GLT-3 might act as a specic ligand to trigger TREM2 signaling in PDN or other neuropathic pain. To induce the experimental type 1 diabetes (T1DM), we injected the mice i.p. with a single dose of STZ (150 mg/kg, Sigma). Then, we determined the behavioral changes in mice receiving STZ (T1DM) or saline injection (Saline). The results showed that compared with the Control group, the PMWTs of the STZ-injected mice began to decrease 3 days after STZ injection (D3) and reached the bottom on D14 (Fig. 1B, F 1, 10 = 144.4, P<0.0001, n=6, two-way ANOVA; mean ± SD: T1DM vs. Saline on D14, 2.70 ± 0.33 g vs. 5.12 ± 0.33 g, n=6, P<0.0001, Bonferroni post hoc analysis), indicating the presence of mechanical allodynia in the STZ-treated mice. We also tested the thermal nociception behaviors after STZ injection using a thermal paw stimulation system. The results suggested that, compared with the Control, the PTWL of the STZ-injected mice dropped after STZ injection, with the statistically signicant difference on D7 and D14 (Fig. 1C, F 1, 10 = 43.93, P<0.0001, n=6, two-way ANOVA; mean ± SD: T1DM vs. Saline on D14, 1.95 ± 0.22 s vs. 3.63 ± 0.16 s, n=6, P<0.0001, Bonferroni post hoc analysis), suggesting evident thermal hyperalgesia in the STZ-injected mice.


Introduction
Painful diabetic neuropathy (PDN) is a common and intractable complication of diabetes, affecting approximately one in four diabetic patients [1]. The prevalence of PDN is even rising over time [2], owing to the growing diabetic population [3]. The typical clinical manifestation of PDN includes various unpleasant symptoms such as mechanical allodynia [4], substantially lowering life quality and increasing health care costs. Although opiates and antidepressants are applied to treat neuropathic pain, their effects on PDN are limited due to the intolerable side effects and poor response rates [5,6]. Therefore, effective therapies speci c to PDN are urgently needed.
Neuroin ammation in the spinal cord has been identi ed as an essential mechanism in neuropathic pain [7,8]. The hallmarks of neuroin ammation are the presence of activated microglia and reactive astrocytes that intensify the production of cytokines, chemokines, or neurotoxic compounds [9,10]. Studies on PDN also reveal a similar surge of neuroin ammation in the spinal cord and attribute these neuroin ammatory responses to microglia activation rather than astrocytes [11,12], suggesting a particular role of microglia in PDN.
Although historically microglia were thought of as solely an in ammatory effector, multiple aspects of microglia are becoming established during the last decades. The activated microglia comprise many functional phenotypes, including the classically activated (M1) phenotype and the alternative activated (M2) phenotype [13,14]. In response to pathogens or tissue injury, the initially quiescent microglia transits into M1 microglia that release multiple pro-in ammatory cytokines or toxic substances in the early phase of in ammation [13,14]. In contrast, M2 microglia acquire an amoeboid morphology and obtain a promoted phagocytic function, allowing them to remove apoptotic neuronal debris and attenuate in ammation [13,14]. Studies have discovered the two opposing functional phenotypes of microglia in the spinal cord in the neuropathic pain induced by nerve injury [15] and bone cancer pain [16]. The neuropathic pain development was concomitant with the increase of M1 microglia, which express CD86 or iNOS, release pro-in ammatory cytokine such as IL-1β [17,18]. However, M2 microglia expressing CD206 and arginase-1 (Arg-1) in the spinal cord was associated with increased anti-in ammatory mediators and neuropathic pain relief [19,16]. Thus, it is promising to relieve PDN by improving microglia polarization toward the M2 phenotype in the spinal cord.
The microglia functional plasticity depends on the transduction of stimulus signals through the cellsurface receptors. The triggering receptor expressed on myeloid cells 2 (TREM2), a cell-surface receptor predominantly expressed on microglia, has been implicated in numerous cerebral diseases, including Parkinson's disease (PD) [20,21], Alzheimer's disease (AD) [22,23], and ischemic brain injury [24,25]. Cells highly expressing TREM2 reportedly mediate immune surveillance and in ammation resolution [26]. By contrast, the deletion of TREM2 results in the impairment of phagocytic capacity and the release of proin ammatory cytokines [27]. Moreover, overexpression of TREM2 confers neuroprotection against Parkinson's disease [20], Alzheimer's disease [28], and cerebral ischemia/reperfusion injury [25] by promoting microglial M2 polarization. However, the conclusion on TREM2 should await because multiple studies provide evidence that inhibition of TREM2 or its downstream DNAX-activating protein of 12 kDa (DAP12) could reduce the microglial number and pro-in ammatory cytokine production in the animal models of PD [29] (induced by 6-OHDA), traumatic brain injury [30], and peripheral nerve injury [31].
Therefore, the present study aimed to investigate the role of spinal TREM2 in the pathogenesis of PDN and its underlying mechanism related to microglia M1/M2 polarization.
In the present study, we reveal a detrimental role of TREM2 in exacerbating neuropathic pain and neuroin ammation in a mouse model of diabetic neuropathy. Our ndings demonstrated that microglial TREM2 activation could aggravate neuroin ammation in the spinal cord by suppressing microglial M2 polarization. Furthermore, we also identi ed galectin-3 (GLT-3) as the speci c ligand of TREM2 in facilitating the progression of PDN. The block of TREM2 or GLT-3 exerted potent analgesic and antiin ammatory effects in PDN.

Animals
Male C57BL/6J mice, weighing 22~25 g, were purchased from Laboratory Animal Center of Guangdong Province (Guangzhou, China), and kept in a standard lab housing with a 12-h light/dark cycle at a temperature of 21 ± 2 °C and 60-70% humidity and allowed access to normal standard diet and water ad libitum. Determination of mechanical withdrawal thresholds (PMWT) and thermal withdrawal latency (PTWL) On the experiment day, the mice were placed individually in transparent test compartments with a wire mesh bottom and habituated for one hour. Paw mechanical withdrawal thresholds (PMWTs) in response to mechanical stimuli were evaluated using the electronic von Frey unit (Bioseb, Montpellier, France) with a exible metal lament applying increasing force (from 0 to 10 g) against the plantar surface of the hind paw of the mouse [33]. The nocifensive paw withdrawal response automatically turned off the stimulus, and the mechanical pressure that evoked the response was recorded.
Thermal hyperalgesia was assessed by measuring paw thermal withdrawal latency (PTWL) to thermal stimuli using PL-200 Plantar Analgesia Tester (Chengdu Technology & Market Co., Ltd., Sichuan, China) as described previously [34,35]. Brie y, the mice were placed on a glass plate and were allowed to adjust to the apparatus for 30 min. The radiant heat lamp source was positioned right beneath the hind paw's plantar surface, vertically projecting a light spot with a diameter of 5 mm. The PTWL was calculated by averaging three individual trials with 5-min intervals to prevent unexpected thermal sensitization. A cutoff time of 12 s was set to avoid tissue injury.

Intrathecal administration of lentiviral vectors and neutralizing antibodies
The intrathecal (i.t.) injection method was performed as previously described [36]. In brief, mice were covered with a soft towel and held gently but rmly by the hip bones via the thumb and index nger of the

Western blot
The mice were sacri ced by anaesthetic overdose (2.5% Avertin, 1600 mg/kg, i.p.). The lumbar enlargement (L4-L5) of the spinal cord was rapidly removed and homogenized in ice-cold RIPA buffer (Beyotime, Shanghai, China). The lysates were kept for 30 min and then centrifuged at 14,000 g for 10 min at 4 ℃. Supernatants were collected, and the protein concentration was determined by the BCA protein assay kit (Boster, Wuhan, China). Equal amounts of protein samples (30 mg) from each group were separated using 10% sodium dodecyl sulphate-polyacrylamide gel electrophoresis and transferred onto polyvinylidene di uoride (Millipore, Bedford, MA, USA) membranes. Afterwards, the membranes were blocked with 5% BSA in TBST for 1 h at room temperature and then incubated with the primary antibodies overnight at 4 ℃. On the next day, the membranes were rinsed three times with TBST and incubated with secondary antibody (anti-mouse or rabbit IgG, 1:5000; Boster) for 2 h at room temperature.
After washing three times with PBST, the protein bands were detected with enhanced chemiluminescent reagents (Boster) and analyzed by densitometric quanti cation using Fluochem HD2 Imaging System (Alpha Innotech, USA).
The following primary antibodies were used: anti-Apolipoprotein E (

Immuno uorescence analysis
For immuno uorescence analysis, mice were transcardially perfused with 4% iced formaldehyde (n=4 or 5/group). The lumbar spinal cord was then removed, embedded in para n, and cut into 5-μm thick serial sections. After depara nation, rehydration, heat-induced antigen retrieval with microwave oven (microwave method), the sections were incubated with 10% (vol/vol) normal goat bovine serum for 60 min at room temperature and were then incubated overnight at 4 °C with primary antibodies against ApoE (

Coimmunoprecipitation Assay
Total protein extracts were prepared from the spinal cord as the Western blot, followed by homogenization with immunoprecipitation buffer (Beyotime) supplemented with 1 mM PMSF protease inhibitor for 1 hour. After centrifugation at 12,000 rpm for 10 minutes, 200 mg of the protein extract was incubated with 5 mg of goat anti-TREM2 polyclonal antibody (Abcam) overnight at 4 ℃, followed by precipitation with protein A/G-agarose (Millipore, Billerica, Mass) for 3 hours [39]. After washing the immunoprecipitation buffer and heating at 95 ℃ for 10 minutes, the immunoprecipitates were subjected to SDS-PAGE and immunoblotting, followed by Densitometric quanti cation (n = 4 per group).

Statistical analysis
All data are presented as mean ± standard deviations (SD). One-way ANOVA followed by Dunnett's multiple comparison test or two-way ANOVA followed by Bonferroni post hoc test was performed using GraphPad Prism 5.0 (GraphPad Software Inc., USA).

Results
The spinal TREM2 increases along with the development of PDN Next, we determined the pattern of TREM2 expression in the lumbar cord of the diabetic mice (T1DM) on baseline (BL), D7, and D14. The Western blot results showed that the protein expression of TREM2 and its downstream DAP-12 in the lumbar spinal cord markedly increased on D7 and D14 (Fig. 1D, P=0.0492 on D7 and P=0.0007 on D14, n=5). These results indicated that the spinal TREM2 was activated along with the development of PDN.
The immuno uorescence double-staining results showed that on D14, TREM2 was mostly expressed in microglia in the spinal dorsal horn of STZ-treated mice (Fig. 1E). The statistical analysis indicated that the number of Iba-1+ microglia and TREM2+Iba-1+ cells was minimal in the lumbar cord of the Control mice but substantially increased in the STZ-treated diabetic mice ( Fig. 1F The switch between M1 and M2 microglial phenotypes regulates the in ammatory response in the CNS [40]. Numerous studies have demonstrated that microglial activation, but not astrocytes, contributes to PDN [11,12]. Thus, we examined the protein expression of the M1 (CD86) and M2 (CD206 and Arg-1) microglia phenotype markers in the lumbar spinal cord. The results revealed that the protein levels of both M1 (CD86) and M2 (CD206 and Arg-1) microglia phenotype markers were markedly elevated on D7 and D14 after STZ injection ( Fig. 2A-D . We also determined the expression of IL-1β, TGF-β, and IL-10 in the spinal cord, as IL-1β is derived explicitly from M1 phenotype microglia, while the primary source of TGF-β and IL-10 is M2 phenotype. We found that the spinal protein levels of IL-1β, TGF-β, and IL-10 were increased on D7 and D14 after STZ injection (Fig. 2H-K, P<0.05, P<0.01, or P<0.0001, n=5).

Overexpression of TREM2 in microglia aggravates PDN in mice
Previous studies have demonstrated that TREM2 inhibits neuroin ammatory response in neurodegenerative diseases such as Alzheimer's disease [22,23], and this effect may be related to its promotion of M2 phenotype microglia. Given the anti-in ammatory properties of M2 phenotype microglia and the above results, we speculated that overexpression of the TREM2 gene in microglia might alleviate PDN by suppressing spinal in ammation. In this regard, we injected intrathecally (i.t.) the mice with microglia-speci c TREM2 overexpression LV (LV-TREM2, 5×10 7 TU) or the control LV (LV-control, 5×10 7 TU, 7 days before STZ injection) 7 days before STZ injection.
We found that the injection of LV-control did not alter the PMWTs and PTWLs compared with T1DM mice that only received an i.t. injection of saline ( Fig.3 A  GLT-3 is the potential ligand of the microglial TREM2 receptor in facilitating the progression of PDN Furthermore, we asked what ligand triggered the TREM2 signalling in the development of PDN. Recent studies suggested that ApoE [41,42] and GLT-3 [43] were the potential ligands of TREM2 in mediating a series of downstream actions. Therefore, we detected the protein expression of ApoE and GLT-3 in the lumbar spinal cord of the T1DM mice. We found that the spinal protein expression of GLT-3, but not ApoE, was signi cantly increased in the STZ-treated diabetic mice on D7 and D14 after STZ injection ( To corroborate these proteins' attachment with the TREM2 receptor, we performed immunoprecipitation by antibodies against TREM2 and tested its e ciency using western blot analysis. We found that the immunoprecipitated GLT-3 with TREM2 antibodies from the lumbar spinal cord of diabetic mice on D14 after STZ injection was more than that of saline-treated mice (Fig. 5D), while the difference in the blotted ApoE was not found (Fig. 5D).

Discussion
In the present study, we investigated the role of TREM2 in the development of PDN and explored its underlying mechanism in mice. We demonstrated that up-regulation of microglial TREM2 and increase of both microglial M1 and M2 response was along with the presence of diabetes-related mechanical allodynia and thermal hyperalgesia. Moreover, we found that overexpression of TREM2 in microglia aggravated the symptom of PDN, ampli ed microglia M1 response, and suppressed microglia M2 polarization in the lumbar spinal cord of diabetic mice. However, inhibition of TREM2 with anti-TREM2 neutralizing antibodies attenuated mechanical allodynia and thermal hyperalgesia in diabetic mice.
Besides, we identi ed GLT-3 as the potential ligand of the TREM2 receptor in facilitating the progression of PDN.
Neuroin ammation, an immune response that arises in the central nervous system (CNS), has been recognized as a hallmark of many neurologic conditions [44,45], including PDN [46,47] and other diabetic cerebral disorders [48]. A critical role of neuroin ammation is also found in neuropathic pain, characterized by speci c activation of microglia rather than astrocytes and increased release of various cytokines [49]. Studies on PDN also reveal a similar surge of neuroin ammation in the spinal cord and attribute these neuroin ammatory responses to microglia activation rather than astrocytes [11,12], suggesting a particular role of microglia in PDN. As a result, investigating the mechanism underlying microglial activation, especially the speci c receptors expressed on the microglial surface, is of great potential for treating neuropathic pain.
In the CNS, TREM2 is a transmembrane receptor predominantly expressed on microglia [20,21]. However, the role of TREM2 in modulating neuroin ammation is still controversial. A large body of studies in neural diseases (i.e., AD [22,23] and PD [20,21]) suggested a bene cial effect of TREM2 by decreasing neuroin ammation and aggravating cognitive impairment. On the other hand, some studies reported an opposite deleterious effect of microglial TREM2 by facilitating neuroin ammation. Another research also recently associated TREM2 with diabetic brain disorder by showing that overexpression of TREM2 alleviated neuroin ammatory response and cognitive decline in Type 2 diabetes mellitus mice [50]. These results led us to clarify the role of TREM2 in PDN.
Our results showed that TREM2 protein levels and microglial numbers simultaneously increase after the induction of T1DM with STZ injection. Since the fold changes of the protein expression in the dorsal horn were comparable to those of microglial numbers, the increase in TREM2 may be due to an increased number of microglia rather than up-regulated expression. Our results indicated that overexpression of microglial TREM2 augmented the spinal in ammatory response (production of pro-in ammatory cytokines IL-1β and TNF-α) and exacerbate pain hypersensitivity in PDN mice. Conversely, inhibiting TREM2 with anti-TREM2 neutralizing antibody decreased the spinal neuroin ammation and ameliorated the pain behavior. This nding is consistent with previous studies on peripheral nerve injury-induced neuropathic pain [31,51], demonstrating that DAP12 (downstream of TREM2 ) knockout suppresses spinal pro-in ammatory responses.
Although previous studies have unravelled the critical role of microglia in PDN, most consider microglia as a marker of in ammation (M1 phenotype) and ignore the M2 phenotype. TREM2 is thought to be an essential molecule that regulates microglial phenotype and activation status. We found that M1 and M2 microglia increased in the spinal cord of PDN mice, and the increase of spinal microglial number was largely M1 microglia [50]. This nding is partly consistent with a recent study [50], which also observed a rise in M1 and M2 microglia in T2DM mice. However, they found overexpression of TREM2 promoted microglial M2 polarization, whereas we suggested that overexpression of microglial TREM2 increased M1 response but reduced microglial M2 phenotypes. The phenomena were further supported by the result that treatment with anti-TREM2 neutralizing antibody decreased M1 polarization and enhanced M2 polarization.
Together, TREM2-mediated signals could be one of the critical pathways in microglial M1/M2 polarization.
Although the ligand that binds to and activates TREM2 remains elusive, a broad array of lipids produced directly or indirectly due to damage to primary sensory neurons has been proposed as potential TREM2 ligands. TREM2 could serve as a microglial sensor for lipid mediators released from damaged sensory nerve terminals or other cells in the DH. Recently, apolipoprotein E (ApoE) was reported as a potent ligand for TREM2. These reports suggest that ApoE could be a potent activator of the TREM2/DAP12 system in the present pain model. Furthermore, a recent study reveals Galectin-3 as a novel endogenous TREM2 ligand, which adversely regulates in ammatory response in AD [43]. Our results demonstrated that the protein expression of galectin-3 but not ApoE increased in the spinal cord of PDN mice. Furthermore, we observed that the immunoprecipitated GLT-3 with TREM2 antibodies was also elevated, indicating the GLT-3 might be the ligand for TREM2 signaling in PDN.
In conclusion, this study provides evidence that TREM2 could be a critical microglial membrane molecule that modulates microglial phenotypes and elicits pain hypersensitivity in PDN. Our results demonstrated that GLT-3 might act as a speci c ligand to trigger TREM2 signaling in PDN or other neuropathic pain. GLT-3 and TREM2 could serve as potential therapeutic targets for PDN.