Genetic loss-of-function of activating transcription factor 3 but not C-type lectin member 5A prevents diabetic peripheral neuropathy

We investigated the mediating roles of activating transcription factor 3 (ATF3), an injury marker, or C-type lectin member 5A (CLEC5A), an inflammatory response molecule, in the induction of endoplasmic reticulum (ER) stress and neuroinflammation in diabetic peripheral neuropathy in ATF3 and CLEC5A genetic knockout (aft3−/− and clec5a−/−, respectively) mice. ATF3 was expressed intranuclearly and was upregulated in mice with diabetic peripheral neuropathy (DN) and clec5a−/− mice. The DN and clec5a−/− groups also exhibited neuropathic behavior, but not in the aft3−/− group. The upregulation profiles of cytoplasmic polyadenylation element-binding protein, a protein translation–regulating molecule, and the ER stress-related molecules of inositol-requiring enzyme 1α and phosphorylated eukaryotic initiation factor 2α in the DN and clec5a−/− groups were correlated with neuropathic behavior. Ultrastructural evidence confirmed ER stress induction and neuroinflammation, including microglial enlargement and proinflammatory cytokine release, in the DN and clec5a−/− mice. By contrast, the induction of ER stress and neuroinflammation did not occur in the aft3−/− mice. Furthermore, the mRNA of reactive oxygen species–removing enzymes such as superoxide dismutase, heme oxygenase-1, and catalase were downregulated in the DN and clec5a−/− groups but were not changed in the aft3−/− group. Taken together, the results indicate that intraneuronal ATF3, but not CLEC5A, mediates the induction of ER stress and neuroinflammation associated with diabetic neuropathy.


INTRODUCTION
Diabetic peripheral neuropathy is a complicated condition caused by hyperglycemia-induced endoplasmic reticulum (ER) stress [1] and neuroinflammation, and its treatment is a clinical challenge [2]. Diabetic peripheral neuropathy also shares the neuropathological characteristics of small-fiber neuropathy (SFN), such as reduction in the intraepidermal nerve fiber (IENF) density (skin denervation) and development of neuropathic behavior. Skin denervation after diabetes is correlated with neuropathic pain [3], indicating IENF densities as a predictor of diabetic peripheral neuropathy progression [4], and neuropathic behavior in diabetic patients [5][6][7]. However, why some diabetic patients with fewer IENFs experienced no pain during quantitative sensory testing remains unclear [8,9]. The small nociceptors became sensitized after their terminal IENFs degeneration. Particularly, activating transcription factor 3 (ATF3), a potential neuronal marker under pathology, was activated on small nociceptors in concert with skin denervation [10], suggesting that ATF3 is a critical marker in addition to noxious transduction by IENF. ATF3 has also been suggested to be a marker of ER stress, and it negatively affects ER stress in obesity-related diabetes [11,12] and in renal tissue failure [13] due to obesity-lipotoxicity-induced ATF3 activation and ER stress. ATF3 has paradoxical regulatory roles; ATF3 deficiency suppresses transplant rejection by ameliorating the inflammatory response [14]. By contrast, tissue-specific ATF3 deficiency exacerbates cardiomyopathy induced by obesity-related inflammation [15]. Moreover, ATF3 induction has been reported to promote inflammation [16] and to maintain energy homeostasis [17]. Despite the diverse roles of ATF3 in pathology and cellular homeostasis, the cascade signal between ER stress and ATF3 in diabetic peripheral neuropathy remains elusive. Furthermore, proinflammatory cytokines induce both ER stress and ATF3 upregulation [18], suggesting that diabetic peripheral neuropathy is a complicated neuropathology underlying neuroinflammation.
Studies have demonstrated that in bacterial or viral infection, CLEC5A-expressing macrophages and microglia are required to induce lethal inflammation, which fails to occur under CLEC5A deficiency [20,23,24,28]. The mechanisms underlying lethal inflammation include systemic microglial and macrophage activation, excessive inflammatory cytokine release [20,23,27], neutrophil extracellular trap formation [21,24,27], and platelet activation through exosome CLEC5A activation [21,27]. However, the role of CLEC5A in diabetic peripheral neuropathy remains unclear. For example, whether CLEC5A is also a responding molecule that mediates diabetes-associated neuroinflammation is unknown. If it is not, the upstream molecule that modulates neuroinflammation and ER stress must be determined.

MATERIALS AND METHODS Diabetic peripheral neuropathy induction and animal groups
We used a modified diabetic peripheral neuropathy mouse model [29] with a single dose of streptozotocin (STZ, 200 mg/kg, Sigma, St. Louis, MO, USA) on three mouse strains: (1) 8-week-old C57/B6 mice, (2) age-matched atf3 −/− mice (gifted by Dr. Tsonwin Hai, The Ohio State University, Columbus, OH, USA), and (3) clec5a −/− mice (gifted by Dr. Shie-Liang Hsieh, Genomic Research Center, Academia Sinica, Taiwan). The atf3 −/− and clec5a −/− mice have a C57/B6 genetic background. The criteria of including in this study by mice exhibiting hyperglycemia (glucose level > 400 mg/dL) in 7 days after STZ treatment, and this study employed five groups: the (1) citrate (mice that received an equal volume of citrate solution, which served as the sham control), (2) DN (mice with glucose level > 400 mg/dL), (3) hypoDN (glucose level < 400 mg/dL, which served as the positive control), (4) atf3 −/− , and (5) clec5a −/− groups. The mice were housed in plastic cages under a 12-h light-dark cycle with access to food and water ad libitum. After neuropathic behavior evaluations at one month after treatment in each group, the mice were sacrificed with intracardiac perfusion, and the related tissues were harvested for subsequent pathological examinations. We made all possible efforts to minimize animal suffering and performed all procedures in a coded and blinded manner and in accordance with ethical guidelines related to laboratory animals. To confirm the genotypes of the atf3 −/− and clec5a −/− mice used in this study, genotyping was performed through polymerase chain reaction (PCR) of the genomic DNA extracted from the tail. The same method employed and primer sequences in another study were used for atf3 [10] and clec5a [20] genotyping.

Neuropathic behavior evaluation
The activities and appearances of the mice were first evaluated before performing neuropathic behavior evaluation by assessing the thermal (hotplate test) and mechanical (von Frey monofilament test) responses of the mice. We used the same methodology as that employed in a previous study [10]. For von Frey hair test [30], the up-and-down method was used [31] and the mechanical thresholds were calculated according to a published formula [32].
Evaluation and quantitation of protein gene product 9.5(+) intraepidermal nerve fibers IENFs were demonstrated by using a pan-axonal marker, protein gene product (PGP) 9.5, in immunohistochemical studies. Briefly, we employed anti-PGP9.5 (rabbit, 1:1000; UltraClone, Isle of Wight, UK) antiserum and the same methodology as that employed in another study [33].

Ultrastructural examinations of ER stress
We used DRG tissues to investigate the pathology of ER stress. Briefly, we dissected lumbar DRG tissues and postfixed the tissues in 2% osmium tetroxide for 2 h, dehydrated them through a graded ethanol series, and embedded them in Epon 812 resin (Polyscience, Philadelphia, PA, USA). Thin sections (50 nm) were stained with uranyl acetate and lead citrate, after which we observed them using an electron microscope (Hitachi, Tokyo, Japan) and photographed them.

Investigation and quantitation of ionized calcium-binding adapter molecule 1 (Iba1)(+) microglia
We examined the pathologies of spinal microglia using an anti-Iba1 antiserum (rabbit, 1:1000; Wako, Osaka, Japan) with the immunostaining protocol; spinal cord tissues were subjected to cryosection, resulting in sections of 50-µm thickness. To ensure adequate sampling, we selected every sixth section of lumbar cord tissues, with six sections chosen in total. To quantify the Iba1(+) microglia, we measured the density, process number, and cell size of Iba1(+) microglia on laminae I and II of the dorsal horn [34]. The Iba1(+) microglial density was defined as the glia number divided by the dorsal horn area (glia/mm 2 ).
ER stress was prevented in atf3 −/− mice after STZ-induced diabetic peripheral neuropathy Proper protein folding and assembly in rough ER (rER) lumen is crucial for normal cellular physiology; thus, we performed ER ultrastructural examinations of DRGs to confirm the altered profile of ER stress-related molecules indicated by the immunostaining study (Fig. 5). Regarding its ultrastructural morphology, the rER comprised stacks of flattened membrane-bound cisternae with a lucent lumen in the citrate (Fig. 5A1, A2) and hypoDN (Fig. 5C1, C2) groups. By contrast, the rER morphology in the DN group exhibited the typical characteristics of ER stress, such as dilated rER (drER in Fig. 5B1, B2) lumen with amorphous masses implying the accumulation of misfolded or unfolded proteins; additionally, we observed some autophagosomes with double limiting membrane [40] (As in Fig. 5B1) within DRG soma (Fig. 5B1, B2). The atf3 −/− mice exhibited less ER stress; that is, the rER comprised normal stacks with flattened membrane-bound and Fig. 5 Ultrastructural examination of the ER in streptozotocin (STZ)-induced diabetic peripheral neuropathy. The lumbar dorsal root ganglia of the citrate (A), DN (blood glucose > 400 mg/dL; B), hypoDN (blood glucose < 400 mg/dL; C), activating transcription factor 3 knockout (atf3 −/− ; D), and C-type lectin member 5A knockout (clec5a −/− ; E) groups at one month after STZ treatment, with samples prepared for electron microscopy examinations. Ultrastructural examinations of the ER (5000×) next to the cell nucleus (N; A-E), which is ER-rich, were performed. Bar, 1 µm. (A1-E2) Higher magnification (15000×) in the insets. The ER in the citrate (A1, A2) and hypoDN (C1, C2) groups appeared as a stack of flattened membrane-bound cisternae with a lucent lumen. The atf3 −/− mice had a similar ER ultrastructural profile with some dilated lucent ER lumen (D1, D2). By contrast, the DN (B1, B2) and clec5a −/− (E1, E2) groups had dilated ER lumen (drER) that contained amorphous or granular substances. Some autophagosomes with a double limiting membrane (arrowheads in B1 and E2) were also observed in the DN and clec5a −/− groups. Other autophagosomes had a double limiting membrane with an away-from-each-other fashion (arrows in E2). Bar, 250 nm. m mitochondria, rER rough endoplasmic reticulum, drER dilated rough endoplasmic reticulum, As autophagosome.

DISCUSSION
Application of IENF assessment in clinical diagnosis of diabetic peripheral neuropathy and unresolved issues The assessment of the IENF density through the combination of skin biopsy and PGP 9.5 immunohistochemistry is a reliable clinical diagnostic tool and can be used to predict neuropathy progression [4]. However, the present study demonstrated that the reduced IENF density after diabetic peripheral neuropathy was hyperglycemia-dependent but was independent of neuropathic manifestation. The evaluation of neuropathic manifestation in patients with diabetic peripheral neuropathy requires different approaches [8], and because it is a systemic metabolic disorder, the underlying mechanism may involve different neuronal-related cells [41]. Our previous study demonstrated that ATF3 expression by the neuronal ε isoform of protein kinase determined neuropathic behavior in a mouse model of diabetic peripheral neuropathy [10]. The current finding further demonstrated the participation of ATF3 in intranuclear signal modulation by the genetic deletion of ATF3 resulting in the reduction of neuroinflammation and ER stress without protective effects on skin denervation, suggesting that the neuronal soma may be a critical No intergroup differences were discovered in density or mean process of Iba1(+) microglia. Expression changes of tumor necrosis factor-α (Tnf-α) (I), interleukin-6 (Il-6) (J), superoxide dismutase (Sod) (K), heme oxygenase-1 (Ho-1) (L), and catalase (Cat) (M) mRNA assayed using quantitative PCR normalized to glyceraldehyde-3-phosphate dehydrogenase (Gapdh). Tnf-α and Il-6 were increased in the DN and clec5a −/− groups, whereas Sod, Ho-1 and Cat levels in the atf3 −/− group were higher than those in the clec5a −/− group. Group labels are indicated on each graph. *p < 0.05, **p < 0.01, ***p < 0.001: DN, hypoDN, atf3 −/− , or clec5a −/− group versus citrate group. # p < 0.05, ### p < 0.001: atf3 −/− or clec5a −/− group versus DN group. •p < 0.05, ••p < 0.01, •••p < 0.001: atf3 −/− versus clec5a −/− group. response center for nociceptive modulation than for nociceptive transduction by IENFs.
Neuronal injury-dependent ATF3 upregulation mediated neuroinflammation and ER stress ATF3 is a susceptible intranuclear molecule associated with neuropathic manifestation in diabetic peripheral neuropathy [10], and injury-dependent ATF3 upregulation is associated with neuroinflammation [42], implying that neuroinflammation is a critical neuropathological process underlying pain development [43]. The present study demonstrated intranuclear ATF3 upregulation only in DN and clec5a −/− mice, which is associated with neuropathic manifestation, ER stress, and neuroinflammation induction. Although CLEC5A is essential for inflammatory responses [20,23,24,28], this study revealed that CLEC5A molecules are not required for diabetic peripheral neuropathy development because microglial activation and proinflammatory cytokine release still occurred in clec5a −/− mice. Proinflammatory cytokines induce ATF3 upregulation and ER stress [18]. The present study revealed a further link between ATF3 upregulation and ER stress induction; in particular, this study demonstrated that ATF3 mediated both neuroinflammatory responses and ER stress induction. For example, these neuroinflammation-related pathologies and ER stress were not observed in atf3 −/− mice, suggesting that ATF3 plays a more crucial role than CLEC5A in neuroinflammation mediation (i.e., there is an intraneuronal response [42] rather than an extracellular ligand-induced systemic inflammatory response).
CPEB is synthesized following ER stress due to fatty-liver-related inflammation [44] and is responsible for pain development [38]. CPEB regulated the inflammatory response through intranuclear signaling [45]. This current study revealed the co-upregulation of CPEB and the ER stress-related molecules of p-eIF2α and IRE1α in DN and clec5a −/− mice, but not in atf3 −/− mice, suggesting that the pathological profiles of these molecules are regulated by neuronal injury-dependent neuroinflammation. Under injury stress, Bcl-XL, an antiapoptotic Bcl 2 family protein, has a neuroprotective effect [39]; the present study also demonstrated the co-upregulation pattern of CPEB and Bcl-XL under neuronal injury. These physiological linkages may be attributed to CPEB aggregation-induced cellular stress, resulting in protein disassembly-induced neuropathy [46]. In addition, CPEB knockdown with antisense oligodeoxynucleotide relieved neuropathic pain [47]. Recently, our research group demonstrated a reduction in Bcl-XL upregulation in atf3 −/− mice after diabetic peripheral neuropathy [10], and the current study further revealed that atf3 −/− mice did not exhibit the coupregulation of CPEB and Bcl-XL. Importantly, this is first study to suggest that ATF3 regulates CPEB pathophysiological activities under injury stress; that is, these pieces of evidence suggest that ATF3 is an upstream modulator responsible for neuronal injurydependent neuroinflammation and ER stress, which lead to a neuropathic manifestation in diabetic peripheral neuropathy.
Importance and implications of ATF3-modulated cellular oxidative stress ER stress and neuroinflammation are the major pathologies of obesity-associated diabetes [11,12] under lipotoxic-dependent diabetic neuropathy. In the current mouse model of diabetic peripheral neuropathy induced by STZ, β-islet cells, which mimic insulin-dependent and obesity-independent diabetes, were destroyed. Accordingly, ER stress and neuroinflammation were also induced through a lipotoxic-independent pathway, which may be injury-dependent [48]. Although ER stress induction (in DRGs) and neuroinflammation (in the lumbar spinal cord) are locally distinct, ROS could be used as an intermediate between injured neurons and activated microglia due to ROS attenuation, which reduced the neuropathic manifestation [49]. The relationships of ATF3 and ROS have been investigated; for example, the loss of ROS scavenger correlated to ATF3 upregulation, and ATF3 deficiency could against bacterial and fungal infection even under ROS stress [50]. In addition, ATF3 silencing by small interfering RNA enhanced the antioxidant effect and attenuated ER stress [11] as well as prevented ER stress-associated cell death [51], suggesting that ATF3 mediates oxidative stress [52]. This ATF3-dependent oxidative stress involved ROS production and inflammation during injury [53,54]. The present study confirmed that the genetic expression of ROS-removing enzyme, Sod, Ho-1, and Cat mRNA was downregulated by DN and clec5a −/− , but this was no changed in the atf3 −/− mice. Accordingly, this study employed two genetic knockout mice, atf3 −/− and clec5a −/− mice, to demonstrate that ATF3 is a crucial intraneuronal upstream molecule such as the genetic loss-of-function of ATF3 that not only prevents intracellular ER stress signals but also eliminates ROS by maintaining a steady level of the ROS-removing enzyme. This pathology of ER stress and excessive ROS were believed to induce the diabetic peripheral neuropathy (Fig. 8). Collectively, the evidence furnished by this study suggests ATF3's potential as a therapeutic target for neuropathic pain control in lipotoxicindependent diabetic peripheral neuropathy.

DATA AVAILABILITY
The data that support the findings of this study are available on request from the corresponding author upon reasonable request. Fig. 8 Mechanisms of intranuclear ATF3-mediated diabetic peripheral neuropathy. ATF3, a susceptible molecule, is activated under neuronal stress. ATF3 mediated (1) the induction of ER stress pathology and (2) ROS production, which induced (3) neuroinflammation including microglial activation and increase in TNFα and IL-6 release. Both ER stress induction and neuroinflammation resulted in the evoking of diabetic peripheral neuropathy. ATF3 activating transcription factor 3, DRG dorsal root ganglion, ROS reactive oxygen species, TNFα tumor necrosis factor α, IL-6 interleukin-6.