1. IntroductionAccording to the tenth edition of IDF Diabetes Atlas 2021, 537 million people are suffering from diabetes, and this number is projected to be 783 million by 2045 [1]. DPN is one of the most prevalent chronic complications and the cause of limb amputations in diabetes mellitus (DM) [2]. Pain and numbness are typical and serious symptoms of patients with diabetic peripheral neuropathy (DPN). However, it shows no obvious clinical symptoms or manifestations in the inchoate stages. At present, the gold standard methods for diagnosing DPN are usually based on the electroneuromyography examination [3]. In practice, these diagnostic methods are difficult and impractical to implement as they are time-consuming and labor-intensive. Thus, there is still a lack of precise early diagnostic indicators of DPN. To improve the quality of life of patients with DPN, prevention by tight glucose control and lifestyle intervention is the best current treatment for DPN. Therefore, the biomarkers for early diagnosis are critical in improving the early diagnosis of DPN patients, which may also improve the prognosis of DPN.Previous investigations have demonstrated that both ischemic and metabolic factors play a key role in DPN [4]. Among those mechanisms in DPN, oxidative stress and chronic neuroinflammation have been highlighted by multiple reviews and research articles [5,6,7]. However, the multifactorial and complex pathogenetic mechanisms in DPN have not yet been fully elucidated. To further explore the specific molecular mechanism, transcriptomics analyses have been performed in several studies, most of them utilizing the gene microarray [8]. Microarrays are commonly used in performing gene expression studies to clarify the relationship between multiple different genes and the disease. In a recent study, a microarray was performed on the sciatic nerve tissues from control rats and DPN rats. The results identified a pool of candidate biomarkers involved in the early phase of experimental DPN [9]. Another study found that the neurotrophin-MAPK signaling pathway was a key signaling pathway in the development of DPN [10]. Based on these previous studies, differentially expressed genes (DEGs) were identified in our study using published datasets in the GEO database which contains DPN and normal sciatic nerve samples.Biomarkers can provide accurate early diagnosis and guidance in clinical decision-making. Additionally, they have contributed to the objective evaluation of pathogenic processes. The neurophysiological methods found that some electrophysiological indicators are expected to be widely used as diagnostic and predictive biomarkers [11]. However, traditional biomarkers have shortcomings and limitations, and few have been used clinically [12]. Recently, molecules involved in several metabolic and signaling pathways associated with DPN have been suggested as predictive biomarkers [13].A growing number of studies have revealed that neuroinflammation serves an important function in the occurrence and development of DPN [14]. For example, dorsal root ganglia are infiltrated by T-cells and neutrophils in chronic DPN [15]. Therefore, from the perspective of the immune system, evaluating the infiltration of immune cells in peripheral nerves and determining the differences in immune cell infiltrate composition would be valuable for elucidating the molecular mechanisms of DPN and developing new immunotherapeutic targets. CIBERSORT is an algorithm used to evaluate gene expression data from microarrays and analyze various immune cell proportions inside the samples [16]. It has been extensively used in the immune cell infiltration analysis in many diseases such as rheumatoid arthritis, lupus nephritis, idiopathic pulmonary fibrosis and human cancers [17,18,19]. To date, no prior studies have yet analyzed immune cell infiltration using CIBERSORT in DPN.In the present study, we obtained gene expression microarray data of DPN from the GEO database. Then, freely available and open-source bioinformatic tools were used to identify differentially expressed genes in DPN samples and normal samples. Functional and pathway enrichment analysis and protein–protein interaction (PPI) network analyses were conducted. We aimed to unravel the specific molecular mechanisms by which these DEGs contribute to the development and progression of DPN. Next, the CIBERSORT algorithm was applied to analyze the difference in immune infiltration between DPN and normal nerve tissues for the first time. Subsequently, machine learning algorithms were used to further screen and determine the potential biomarkers of DPN. Finally, to further understand the immune mechanisms during DPN development, the relationship between the biomarkers and the infiltrating immune cells was studied. In addition, 15 mRNAs were confirmed as differentially expressed by qRT-PCR. The complete workflow is shown in Figure S1. 4. DiscussionDPN is a common, serious and troublesome chronic complication of DM [46]. Chronic hyperglycemia and oxidative stress lead to major structural and functional abnormalities of the peripheral nerves. In addition, neuroinflammation plays an important role in the development of DPN. Currently, large numbers of DPN patients complain of pain, fatigue, reduced quality of life and disability. Unfortunately, early diagnosis is difficult due to the lack of specific diagnostic indicators. Therefore, finding novel diagnostic biomarkers and analyzing the pattern of DPN immune cell infiltration is useful for improving the outcomes of patients with DPN. Previously, multiple studies have found that the signaling pathways comprised of some genes may play an important role in the development of DPN. However, few systematic analyses and comparisons of the transcriptome data have been made. More importantly, the exact mechanism underlying the progression of DPN driven by key genes remains to be fully elucidated.In this study, bioinformatics techniques were used to analyze microarray data acquired from the GEO database isolated from sciatic nerves of T2DM mouse models to identify potential biomarkers. A total of 628 upregulated and 680 downregulated DPN-related DEGs were identified in the GSE70852 dataset and GSE27382 dataset. GO enrichment analysis showed that upregulated DEGs were mainly enriched in leukocyte migration (GO:0050900), leukocyte chemotaxis (GO:0030595), cell chemotaxis (GO:0060326), neutrophil migration (GO:1990266) and granulocyte migration (GO:0097530). Additionally, downregulated DEGs were mainly enriched in neurotransmitter transport (GO:0006836), regulation of membrane potential (GO:0042391) and axonogenesis (GO:0007409). From the above results, it was revealed that a DPN upregulated immune response and was significantly associated with the impairment of neurological function. Furthermore, DO enrichment analysis showed that immune-mediated diseases such as inflammation, fibrosis and arthritis were enriched. The IL-17 signaling pathway, p53 signaling pathway and Toll-like receptor signaling pathway were identified to be associated with DEGs by pathway enrichment analysis. Several investigators have suggested that upregulated IL-17 possesses a crucial role in the inflammatory process and the development of DM [47]. Ben Y et al. found that astragaloside IV could reduce the occurrence of mitochondrial-dependent apoptosis by regulating the SIRT1/p53 pathway in DPN rats [48]. Other studies have found that Toll-like receptor4 could be a potentially sensitive diagnostic biomarker for DPN in type 2 diabetic patients [49]. Our analysis data were also consistent with the findings above.Through PPI network construction, genes that have high scores in eight algorithms were considered as key hub genes, such as CCL2, TGFB1, MMP9 and CD68. It is of note that abnormal expression of some genes has been reported to be related to DM or DPN in the past few years. As an example, C-C chemokine ligand 2 (CCL2) and its receptor are key players in the attraction of monocytes to sites of injury and inflammation and it was proposed to be a major cause of diabetic neuropathic pain [41,50]. Previous studies have demonstrated that Triphala churna acted as a neuroprotective agent in DPN via the downregulation of inflammatory cytokines such as TGFB1 [51]. Moreover, downregulation of MMP9 could improve peripheral nerve function via promoting Schwann cell autophagy in DPN [52]. A previous study found that CD68, a macrophage marker, was higher in the DRGs of patients with DPN, demonstrating that the upregulated inflammatory markers may contribute to the inflammatory response, potentially stemming from diabetes related neuronal pathology [53]. Overall, inflammation can be an important factor following peripheral nerve injury, as activated macrophages are needed to engulf myelin debris and apoptotic cells. However, sustained and low-grade inflammation is generally known to be linked to diabetes [54]. This impairs the cell viability in the peripheral nerve.In order to explore the role of immune cell infiltration in DPN, CIBERSORT analysis was applied to estimate the fractions of immune cells in sciatic nerves. We found that an increased infiltration of M1 macrophages and resting CD4 memory T cells, and a decreased infiltration of M2 macrophages, resting mast cells, monocytes and follicular helper T cells may be related to the development of DPN. Macrophages are professional phagocytes belonging to the innate immune system that can be activated by a variety of external stimuli. Based on their function, macrophages can be differentiated into two phenotypes: M1 (pro-inflammatory) and M2 (anti-inflammatory) macrophages [55]. M1 macrophages are able to secrete a broad range of inflammatory factors, such as IL-6, IL-1β and TNF-α. Previous studies have shown that M1 macrophages increased significantly in DPN patients [56]. This means that M1 macrophages might play a pivotal role in the onset and development of DPN [57]. Moreover, it was found that increased expression of TLR4 in monocytes could be related to systemic inflammation in peripheral neuropathy in T2DM [58]. These findings further support the important role of immune cell infiltration and inflammation in the development of DPN.

LASSO logistic regression is a reliable method for selecting diagnostic features of DPN based on regression trees. It provides a statistically rigorous method to identify the variable λ when the predicted outcomes are best. Furthermore, SVM-RFE is a classic machine learning method based on a recursive feature elimination strategy to select important genes by training a support vector machine model. To further select feature variables and build an accurate classification model, we applied these two algorithms in this study. The overlap of the LASSO logistic regression model and the SVM-RFE algorithm was obtained. Consequently, LTBP2 and GPNMB were recognized as potential diagnostic markers for DPN.

Latent transforming growth factor beta binding protein 2 (LTBP2) is a member of the fibrillin/LTBP extracellular matrix glycoprotein family [59]. It plays a critical role in regulating the extracellular matrix glycoprotein. A growing number of studies have found that LTBP2 was associated with cardiac fibrosis, acute heart failure, glomerular filtration rate and pre-eclampsia [59]. Recent investigations have suggested that overexpression of LTBP2 facilitated inflammation in endometriosis [60]. It is regretful that the role of LTBP2 in DPN development has not been studied. Therefore, this needs further experimental verification. GPNMB is an endogenous type 1 transmembrane glycoprotein. A study has shown that GPNMB is closely related to neuroinflammation [61]. Interestingly, neuroinflammation happens to be one of the most important mechanisms in the development of DPN. It would be reasonable to speculate that GPNMB may play an important role in the disease progression of DPN. In conclusion, evidence from previous studies indicates that LTBP2 and GPNMB may play an important role in the development and progression of DPN. However, validated experiments and clinical studies are still needed to assess the diagnostic value of LTBP2 and GPNMB. A comprehensive analysis was performed including LTBP2, GPNMB and immune cells. LTBP2 was significantly positively correlated with M1 macrophages and GPNMB was significantly negatively correlated with M2 macrophages. We speculate that LTBP2 and GPNMB affect immune cells to participate in the occurrence and progression of DPN. Further experimentation is needed to validate these hypotheses, including experiments regarding the interactions between genes and immune cells. 5. Conclusions

In this study, DEGs associated with DPN were identified by analyzing previously published datasets containing DPN and normal sciatic nerve samples. Then, functional enrichment and PPI network analyses were conducted for DEGs, elucidating the detailed mechanisms and the pathogenesis of DPN. What is more, by using novel bioinformatics methods such as LASSO logistic regression algorithms and the SVM-RFE algorithm, we have identified potential DPN diagnostic markers, LTBP2 and GPNMB. This is the first time that CIBERSORT was used to analyze immune cell infiltration in peripheral nerve tissues. Nevertheless, we recognize that there were important limitations in our study which cannot be ignored. First, the current study is limited by a small sample size due to the small number of gene microarrays in DPN. Furthermore, CIBERSORT analysis is based on limited genetic data that may deviate from heterotypic interactions of cells, disease-induced disorders or phenotypic plasticity. In addition, our research needs to be further experimentally validated. In conclusion, our results present the promising potential for several diagnostic biomarkers of DPN and provide a novel strategy for DPN diagnosis and treatment.