LNB is a nervous system disorder caused by spirochete Bb infection. At present, the efficacy of antibiotics in the treatment of LNB has been defined. Early detection and prompt anti-pathogen treatment can improve the prognosis of patients(Rauer et al. 2018). However, due to the lack of specific clinical manifestations, sporadic cases in non-endemic areas are usually missed and often misdiagnosed. Patients with LNB can experience serious consequences such as dementia or personality disorders in the absence of timely treatment, which greatly impacts on their quality of life. Therefore, it is crucial to study the pathogenic biomarkers of LNB that may provide evidence for early diagnosis and treatment.
Recently, bioinformatics analysis has become the conventional means for medical research into disease diagnosis or treatment, and it has significantly accelerated the utilization and integration of public biomedical resources. Nonetheless, to date, only a few omics studies have pointed to models of LNB. Our previous study established a dataset of the transcriptional changes induced by Bb in fresh rhesus frontal cortex tissues(Ding, Ma, Tao, Peng, Han, Sun, Dai, Ji, Bai, Jian, Chen, Luo, Wang, Bi, Liu and Bao 2019). In the present study, for the first time we conducted a comprehensive analysis of transcriptional mRNA changes induced by Bb both in rhesus brain tissue and primary human astrocytes(Casselli et al. 2017). Similarly, DEGs from astrocytes co-cultured with Bb and control astrocytes were identified and integrated with the gene expression dataset GSE85143 from the GEO database.
Since the two datasets generated on different sequencing platforms (Illumina HiSeq 2500 & Illumina HiSeq 2000), the DEGs associated with Lyme spirochete infection to CNS were identified by taking the intersection analysis on each dataset. Some original data used in the current study have been used in prior publications, however, we generated a set of DEGs by taking the union from all time points. The sample species are different between our rhesus dataset and GSE85143, the DAVID Gene ID Conversion Tool was used. After deleting duplicate and invalid genes, we finally identified 3,075 differentially expressed genes in this study. In our previous study, the number of DEGs identified were 2,249 (6 h post-exposure), 1,064 (12 h post-exposure), and 420 (24 h post-exposure)(Ding, Ma, Tao, Peng, Han, Sun, Dai, Ji, Bai, Jian, Chen, Luo, Wang, Bi, Liu and Bao 2019). In the study conducted by Casselli et al., GSE85143 is a part of superseries, GSE85143 is RNA-seq dataset which reveals changes in the astrocyte transcriptome following Bb infection and the other parts were about non-coding RNA profiling. We only focused on the analysis of mRNA expression profile for comprehensive analysis. We used the samples infected with Bb at 24 h and 48 h as test group and the three uninfected as control. Accordingly, there are 1450 DEGs in GSE85143 which could be compared to previously published results(Casselli, Qureshi, Peterson, Perley, Blake, Jokinen, Abbas, Nechaev, Watt, Dhasarathy and Brissette 2017).
Herein, we screened and identified 80 upregulated DEGs (Supplemental File 2) and 32 downregulated DEGs (Supplemental File 3). Among the upregulated DEGs, most GO terms in the BP group were associated with cell adhesion, while for downregulated DEGs, the significantly enriched transcripts in BP GO terms were related to extracellular matrix organization. The KEGG pathway analysis indicated that these upregulated DEGs were significantly enriched in pathways involved in the complement and coagulation cascades, hematopoietic cell lineage, Staphylococcus aureus infection and ARVC. There were no significantly enriched KEGG pathways associated with downregulated DEGs. We constructed a PPI network by screening 11 hub genes. To obtain more reliable results, the 112 DEGs of 12 algorithms were investigated using the CytoHubba plugin. The results of top 30 DEGs of each algorithm were sorted by rank and the following intersection genes were considered as hub genes (Supplemental File 4): SREBF1, LDLR, SELP, CD93, ANGPT1, TLR6, MATN3, SERPIND1, TNC, ITGA2, and CSF1R. Among these hub genes, SELP, CD93, ANGPT1, TLR6, and SERPIND1 were upregulated, while SREBF1, LDLR, MATN3, TNC, ITGA2, and CSF1R were downregulated. Furthermore, we performed KEGG pathway enrichment analyses on hub genes and the results revealed that the PI3K-Akt signaling pathway showed the most significant difference (P-value < 0.01). Finally, we validated these results using ex vivo samples of rhesus frontal cortex brain explants and in vitro in human U251 cell lines using qPCR at the transcription level. These results suggested that TLR6, ANGPT1, SREBF1, LDLR, TNC, and ITGA2 might be involved in the pathogenesis of LNB.
TLR6 is a member of the Toll-like receptor (TLR) family, which plays a key role in pathogen-associated molecular patterns (PAMPs) recognition and in innate immune responses. TLR1, TLR2, and TLR6 are all located on the cell membrane. They share high protein homology in their transmembrane and cytoplasmic regions, while their extracellular region is diverse. Thus, they can recognize specific pathogens and activate common intracellular signaling pathways(Takeuchi et al. 1999). It has been shown that TLR is a driving force behind Bb infection, and that the TLR signaling pathway plays an important role in inducing inflammatory responses(Koening et al. 2009, Singh and Girschick 2006, Thomas and Fikrig 2002). Specifically, TLR2 plays an important role in the recognition of Lyme spirochetes(Hirschfeld et al. 1999). TLR6 forms heterodimer with TLR2 and distributes along the cell surface to specifically recognize diacylated lipopeptides and partly triacylated lipopeptides(Kang et al. 2009). A previous study showed that blocking TLR6 on human peripheral blood cells after exposure to Bb did not reduce the production of cytokines, while blocking TLR1 significantly decreased cytokine production(Oosting et al. 2011). Further, the Bb-induced secretion of IFN-γ in murine cells depends on TLR6(Oosting, Ter Hofstede, Sturm, Adema, Kullberg, van der Meer, Netea and Joosten 2011). Our previous studies have also shown that the TLR1/TLR2 ratio plays an important role in initiating of proinflammatory chemokine storm in Lyme disease(Zhao et al. 2019). However, due to gene expression specificities in different species, tissues or cell lineages, the role of TLR6 in LNB needs to be further explored. In the present study, TLR6 is the only hub gene that was verified to be elevated in both the human U251 cell line and in rhesus brain explants when exposed to Bb. The transcription level of TLR6 began to increase at 12 h after human astrocytes were exposed to Bb, and the increased expression was observed at only 6 h after rhesus brain explants were exposed to Bb. This may be because other immune cells in the brain tissue, such as microglia, are also involved in the pathophysiological process in the early stages of LBN. Additionally, further analysis of KEGG enrichment analysis of hub genes suggested that the PI3K-AKT pathway may play a key role in the pathogenesis of LNB. Interestingly, we found that the TLR6 signals could activate this pathway using the KEGG pathway mapping tool (Fig. S2). Therefore, TLR6 may take part in the pathogenesis of LNB.
With regard to the upregulated hub genes, in addition to TLR6, another gene that has been verified to be upregulated in human astrocytes exposed to Bb is ANGPT1. This result is consistent with the results of the study by Casselli et al.(Casselli, Qureshi, Peterson, Perley, Blake, Jokinen, Abbas, Nechaev, Watt, Dhasarathy and Brissette 2017). They confirmed the elevation of angiopoietin-like 4 (ANGPTL4) in human primary astrocytes exposed to Bb by enzyme-linked immunosorbent assays. Both ANGPT1 and ANGPTL4 belong to the angiopoietin family and encode secreted glycoproteins that play critical roles in development and in disorders including in angiogenesis, inflammation, cell proliferation, apoptosis, lipid and glucose metabolism, cell migration, and cancer(Carbone et al. 2018, Ehrlich et al. 2019, Snipstad et al. 2010, Zhang et al. 2018). Compared with other ANGPT/ANGPTL genes which are embedded in the introns of larger genes, ANGPT1 and ANGPTL4 are stand-alone genes whose expression correlated with tissue-specific enhancer or promoter chromatin sequences(Ehrlich, Lacey and Ehrlich 2019). ANGPT1 is the ligand of the transmembrane tyrosine kinase TIE2 receptor. Phosphorylation of tyrosine kinase can activate the PI3K/AKT pathway (Fig. S3), which was a core signaling pathway identified in our previous studies(Fukuhara et al. 2008, Huang et al. 2010). ANGPT2, which is highly homologous to ANGPT1, has a completely different effect on TIE2(Parikh 2017). The involvement of ANGPT1 and ANGPT2 in pathological inflammation is clear. In several diseases that have a strong association with inflammation, such as sepsis and malignant tumors, angiopoietins can be used as prognostic indicators(Fiedler and Augustin 2006, Park et al. 2009, Seol et al. 2020). In our previous study(Zhao et al. 2018), we found that the NF-κB pathway plays a key role in the pathogenesis of LNB, and other studies have reported that ANGPT can also influence the NF-κB pathway(Fiedler and Augustin 2006, Huang, Bhat, Woodnutt and Lappe 2010). These results suggested the possible potential of ANGPT either as novel diagnostic and prognostic indicators, or as a therapeutic target for LNB. In addition, upregulation of SERPIN in the upregulated hub genes were verified in the study by Casselli et al(Casselli, Qureshi, Peterson, Perley, Blake, Jokinen, Abbas, Nechaev, Watt, Dhasarathy and Brissette 2017). Thus, they have not been discussed in detail here.
Among the downregulated hub genes, we verified that the expression of LDLR, SREBF1, TNC, and ITGA showed a statistically significant decrease in human astrocytes on exposure to Bb. Following exposure of rhesus brain explants to Bb, we found there was a decrease in both LDLR and TNC levels, although there the differences were not statistically significant (P-values of LDLR at 6 h, 12 h and 24 h Bb-exposed astrocytes vs controls were 0.8, 0.08, and 0.37, respectively). LDLR is a cell surface glycoprotein that can be expressed on various cell types including astrocytes in many tissues. LDLR can specifically recognize and bind lipoproteins containing apolipoprotein (Apo)E or ApoB100 to mediate cholesterol metabolism, thus it is a central component for the maintenance of cholesterol homeostasis(Brown and Goldstein 1986). Studies have elucidated that LNB may promote the neuro-inflammatory response and endow susceptibility to Alzheimer’s disease (AD), which can not only lead to dementia, but also may induce pathological features typical of AD, such as amyloid-beta (Aβ) deposits(Miklossy et al. 2004, Miklossy et al. 2006). Conversely, upregulating the expression of LDLR can improve the brain clearance of Aβ, reduce amyloid-deposition and attenuate the neuroinflammatory response(Basak et al. 2012, Yao et al. 2016). A recent study revealed that lower LDLR expression can aggravate the neuronal inflammatory response, which might occur through NF-κB signaling and NLRP3-ASC caspase-1 inflammasome assembly(Sun et al. 2020). Thus, LDLR may serve as a protective factor for LNB.
Meanwhile, the reduction of SREBF1 expression at the transcriptional level has been verified in human U251 cells at exposure to Bb. The proteins encoded by SREBFs are transcription factors that control cholesterol homeostasis. LDLR is one of the target genes regulated by sterol regulatory element-binding proteins (SREBPs)(Horton et al. 2003). We observed a decrease in TNC transcriptional expression of U251 cells after stimulation by Bb, and we also found a downward trend over time in rhesus brain explants, albeit there was no statistically significant difference. The protein encoded by TNC belongs to the tenascin family and is widely involved in pathological processes such as inflammation and malignancy. TNC plays a key role in the proliferation of primary astrocytes(Ikeshima-Kataoka et al. 2008, Roll and Faissner 2019). There have been many reports that TNC is highly expressed when the CNS is injured(Roll and Faissner 2019). However, brain injured TNC knock-out mice showed higher expression of inflammatory factors such as TNF-α, IL-6, and IL-1β than wild-type mice(Ikeshima-Kataoka, Shen, Eto, Saito and Yuasa 2008). The expression level of TNC varies under different pathological conditions. It may help regulate the production of inflammatory factors in the damaged brain. In addition, ITGA2 reduction was observed at 6 h and 48 h in U251 cells after exposure to Bb. ITGA2 is a glycoprotein of the integrin family, which mediates cell–cell interactions and those of the cell-extracellular matrix. It also participates in the pathophysiological process of inflammation and the immune response. However, studies have reported that ITGA2 can aggravate the destruction of inflammatory cartilage in rheumatoid arthritis, and has been implicated in cell growth and apoptosis in tumors(Penrose et al. 2019, Peters et al. 2012). The roles of ITGA2 in LNB remain to be clarified, therefore, further studies are needed.
All in all, in present study, we combined two different datasets related to LNB and identified a group of commonly affected transcripts. Using the commonly affected genes from two very different experimental systems to conduct detailed functional and network analyses. Among them, 11 hub genes were identified from network analysis. We also conducted validation experiments and validated some of the hub genes.
There are some limitations in our current study. First, in vivo experimental validation is needed to validate these findings. In addition, although we have verified the mRNA expression of hub genes using a cell line in vitro and in ex vivo rhesus brain explants, we did not carry out in-depth mechanism research. Additionally, due to the different sample sources and species of the two datasets, the intersection of the DEGs could provide only limited information. Moreover, the sufficient validation by alternate methods or by measuring the protein levels for biological relevance to LNB will need to be performed in future work. And the number of samples is relatively small. The underlying mechanisms need to be further studied in our future studies.
In conclusion, the present study suggests that TLR6, ANGPT1, LDLR, SREBF1, TNC, and ITGA were differentially highly expressed in Bb-infected astrocytes compared to control astrocytes. These candidate genes are clinically promising biomarkers that can be used for LNB diagnosis or treatment. In addition, LDLR was found to be significantly associated with favorable outcomes in neuroinflammatory diseases, which may provide a better understanding of molecular mechanisms and novel targets as therapeutic strategies in the future. Thus, our findings may provide valuable insights into the study of LNB pathogenesis and provides guidance for designing further follow-up studies.