Web crawling and mRNA sequencing analyze mechanisms of photobiomodulation

Background Photobiomodulation (PBM) is praised as a promising physical therapy, which has many advantages, such as noninvasive, painless. However, the mechanisms are not fully elucidated. Methods Using web crawling, mRNA sequence and bioinformatics analysis to select genes, functional annotation, and mechanisms. The expressions of inammatory cytokines were measured using RT-qPCR. Results A total of 146 human genes and 57 pathways were identied about PBM. The 630nm LED-stimulated-MH7A cells were sequenced for further analyzing the mechanism of PBM. 2950 differentially expressed genes were identied, and the gene ontology term enrichment analysis and Kyoto encyclopedia of genes and genomes pathway analysis were performed to better understand functions and pathways. The 12 pathways were matched with the KEGG results of PBM and MH7A cells. A protein-protein interaction network was performed among genes in 12 pathways, and 10 outstanding proteins were identied. Importantly, the 9 genes were predicted with potential research value.And we also demonstrated that expression of inammatory factors (IL-6, IL-1β, IL-8, and MMP-3) was reduced; meanwhile, the expression of anti-inammatory factor IL-10 was promoted after 630nm LED. In conclusion, using web crawling, bioinformatics analysis, and mRNA sequence, we obtained 9 key genes and 10 important pathways about PBM. Importantly, we demonstrated the anti-inammatory effect of 630nm LED red light by RT-qPCR. differentially gene. Sequencing results showed that there were 2950 differential genes, and KEGG enriched 29 pathways. The top ten enriched pathways were Cytokine-cytokine receptor interaction, Neuroactive ligand-receptor interaction, Protein processing in the endoplasmic reticulum, Cell cycle, Viral protein interaction with cytokine and cytokine receptor, Complement and coagulation cascades, Legionellosis, DNA replication, Prion diseases, Bladder cancer. Among the 29 signaling pathways, 12 of them coincide with those obtained by PBM analysis, which are Cytokine-cytokine receptor interaction, Legionellosis, Prion diseases, Viral protein interaction with cytokine and cytokine receptor, Complement and coagulation cascades, Bladder cancer, Fluid shear stress, and atherosclerosis, Prostate cancer, MAPK signaling pathway, TNF signaling pathway, IL-17 signaling pathway, Hepatitis B. These 12 signaling pathways may be important in HBEGF, IL1B, IL1A, PLAU, CCL20, represent the part of differentially expressed genes in LED stimulate-MH7A. It means that they not have been studied and have research value and potential in PBM. For the accuracy of the PPI network and the effect of 640nm LED red light ,the RT-qPCR was used to analysis mRNA levels of inammatory. There was a signicant decrease in the gene expressions of inammatory cytokines IL-6, IL-8, and IL-1Βand MMP-3 expression; meanwhile, the expression of antiinammatory factor IL-10 was promoted with the radiation doses increased.These results demonstrate that the gene expression and production of inammatory cytokines by MH7A cells can be inhibited by 630-nm LED radiation.


Introduction
PBM is a new physical therapy, and it utilizes endogenous chromophores with low dose illumination delivered at the target site, which is the opposite of photodynamic therapy by stimulating exogenous chromophore [1] . PBM is also called low-level laser therapy(LLLT), which is a non-invasive form of phototherapy that utilizes wavelengths of light between 650 and 1000 nm to deliver low irradiance and doses to the target tissue [2] . So far, low-level laser therapy has been recognized as a light therapy that may be used for anti-in ammation, promoting wound healing and bone regeneration [3][4][5] .
Although many articles have put forward many opinions, the mechanisms of PBM are not fully elucidated. Studies have demonstrated that NF-kB was signi cantly activated on 0.003J /cm 2 by irradiating rat embryonic broblasts with 810nm laser [6] . Some studies have found that LLLT can inhibit apoptosis thl rough the PI3K-AKT signaling pathway. LPLI inhibits staurosporine (STS)-induced cell apoptosis by inactivating the GSK-3b/Bax pathway [7] . However, it has also been demonstrated that LLLT enhances the invasivity of tumor cells by regulating the Akt/ mTOR signaling pathway [8] .
Photodynamic therapy(PDT) is a highly effective and safe treatment method of phototherapy [9] .
Photodynamic therapy is rstly reported in the 1970s and is honored as the most promising treatment model of the 20th century in the 1980s [10] . It used chemical reactions between light and photosensitizers in speci c wavelengths to produce reactive oxygen species that kill target cells for therapeutic effect [11] .
Both light and photosensitizers were harmless and can be activated when exposed to a speci c wavelength of light in a speci c area. Therefore, this treatment has many advantages, including strong controllability, low toxicity, and non-invasive, etc [12] . To demonstrate whether the mechanism of PBM and PDT is different, this study compares the similarities in the genes and pathways affected by PBM and PDT.
Light sources are laser and light-emitting diode(LED) in phototherapy. LED, as the fourth generation light source, has the advantages of energy-saving, environmental protection, and high electro-optical conversion e ciency. Therefore, LED is becoming more application in LLLT. However, the two light sources have different properties [13] . Due to the laser is coherent light, but LED is incoherent light, thus one of our aims is to compare the effects of laser and LED.
In conclusion, we crawled 3352 literature, identi ed 146 human genes from the abstracts of the literature, and performed bioinformatic analyses to get 9 key genes and important 12 pathways about PBM. The research design was created in the form of a owchart( Figure 1). In the NCBI PubMed database, with 'photobiomodulation' or 'low level laser therapy' as keywords in [Title and Abstract] to search PBM-related literature. The same way that keywords were 'photodynamic therapy' to search PDT-related literature. And crawled through R language to get the abstracts of the relevant literature.

Crawling LED, laser-related literature
In the NCBI PubMed database, due to the research on LED started late, there were few articles related to it.
To expand the retrieval range, we searched relevant literature directly with the 'light-emitting diode' as a keyword. Similarly, in [title and abstract], 'laser' was used as the keyword to search laser-related literature.
And crawled literature through R language to get the abstracts of the relevant literature.

Identi cation of genes in the relevant literature
Adopting the gene_atomization function in the pubmed. mineR package of the R language, the genes in the abstracts in the relevant literature can be directly extracted. The gene_atomization can be used to identify genes in relevant literature abstracts with PBM, PDT, LED, and laser.
2.2 Con rming the related genes of LED and laser phototherapy The genes of the PBM group, PDT group, LED group, and laser group was identi ed, and their respective intersections were obtained through the Venn diagram. The intersection between LED genes and (PBM+PDT) group genes was LED -related phototherapy genes. Similarly, laser-related phototherapy genes were con rmed.

Enrichment of functions and pathways
To study the genes that have been identi ed from the relevant literature, we carried out a comprehensive set of functional annotations for them and used the "Clusterpro ler" package for gene ontology (GO) term enrichment analysis and Kyoto genes and Encyclopedia of Genomics (KEGG) pathway analysis [14] . All KEGG pathway and GO enrichment analysis was based on a threshold of p-value <0.05.

PPI network
Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (http://www.strin-db.org/) is a systems biology tool to assess protein-protein interaction (PPI) information [15] . To evaluate the interrelationships behind the genes, we used the STRING database for analysis. Meanwhile, the MCODE [16] plug-in of Cytoscape software was used to visualize and select hub modules of the PPI network.The degree = 5, node score = 0.2, k-core = 2, and maximum depth = 100 were used as cut-off criteria.

3.Cell culture
Human broblast-like synoviocyte line, MH7A was purchased from GuanDao Biological Research Center of Shanghai. Cells were cultured in DMEM supplemented with fetal bovine serum (10%) and penicillin/streptomycin solution(1%). The cultured cells were maintained at 37℃ in a 5% CO2 atmosphere.

4.Light-emitting diode bio-stimulation
Light-emitting diodes (CP-YL3213-630, Truwin Optoelectronic Medical, Beijing, China) were used to deliver 630-nm radiation. Cells were exposed through the bottom of plastic culture plates to the 630-nm radiation. The time it takes to radiation once was 15 min to achieve radiation energies of 39 J/cm 2 . TNF-α stimulation groups in which TNF-α(10 ng/mL, PeproTech) was added before the radiation. Cells of experimental groups were radiated at room temperature. The control groups (non-irradiated cells) were wrapped with aluminum foil and kept at room temperature while the experimental samples were radiated.

5.Gene sequencing
mRNA-seq was performed as reported in Genesky Biotechnologies Inc., Shanghai, China. We used the "limma" R package to screen the differentially expressed genes between LED+TNF-α samples and normal samples [17] . P-value < 0.05 and |log2fold change (FC)| >1, were chosen as the cut-off threshold.

2.Gene identi cation
A total of 146 human genes related to PBM, 268 human genes related to PDT were identi ed, and the frequency of top10 genes was shown in Figure 2a, 2b. The 142 human genes related to LED and 3188 human genes related to laser were identi ed.

3.Con rm the genes in LED/laser-related phototherapy
In this experiment, the intersected genes of the LED group genes and (PBM + PDT) group genes were LED-related phototherapy. In the same way, the genes in laser-related phototherapy were con rmed. By using the Venn diagram, as shown in Figure 3, it was clear that there were 161 human genes related to LED, of which 50 genes were intersected with PBM, 54 genes were intersected with PDT, and 29 genes were repeated intersecting. The 3188 human genes related to laser light therapy, of which 138 genes were intersected with PBM, 141 genes were intersected with PDT, and 59 genes were repeated intersecting. In summary, by crawling the literature and identifying genes, a total of 75 phototherapy-related genes with LED and 220 phototherapy-related genes with laser were found. Moreover, the genes were deserved to be attention in the public intersection of LED, laser, PBM, and PDT. We proposed that the genes in the public intersection may be the key genes that play a role in phototherapy. For this reason, enrichment analysis was performed on these genes.

4.Comparison of PBM and PDT
In the R language, the clusterpro ler package was used to enrich functions and pathways. Among them, a total of 57 pathways in the PBM group were enriched, and the results of enrichment were shown in Figure   4a, b; a total of 109 pathways in the PDT group were enriched. The results of pathway and function enrichment were shown in Figure 5a, b; Compare the two results, a total of 45 pathways were in the intersection, which intersection is 78.9% of the KEGG result of PBM.

5.Comparison of LED and laser
By crawling the literature and identifying genes, a total of 75 phototherapy-related genes with LED and 220 phototherapy-related genes with laser were found. No KEGG result in the LED-related phototherapy, which was caused by the number of identi ed genes was not enough. A total of 104 pathways were enriched in the laser group, as shown in Figure 6. However, it can be known that 74 genes were intersected with laser in 75 LED-related phototherapy genes from the Venn diagram. The intersected genes account for up to 98.7% of the 91 LED genes. Therefore, it can be concluded that the genes affected by the LED in phototherapy are the same as the laser light source.
6.The 29 genes that were located in the common intersection of PBM, PDT, LED, and laser Known from the Venn diagram that a total of 29 genes were a common intersection in these four groups, and so we speculated that these genes may be the keys in phototherapy. Therefore, we also performed KEGG and GO analysis for these 29 genes. A total of 7 pathways were enriched. The results were shown in Figure 7a. At the same time, we also performed a GO enrichment analysis for these genes, and a total of 365 results were enriched, as shown in Figure 7b. It was obvious that these genes were related to tyrosine, ultraviolet, light, and oxygen stress, etc.

7.Sequencing results of LED stimulate-MH7A
To explore the mechanisms of PBM for RA disease, we sequenced LED-MH7A cells. First, MH7A cells were stimulated with TNF-α. The experimental groups were irradiated with 630nm LED for 15min at room temperature, while control groups were wrapped with aluminum foil at room temperature. A total of 2950 differentially expressed genes were identi ed. 1885 and 1065 differentially expressed genes were upregulated,down-regulated, respectively. The volcano is shown in Figure 8. KEGG enriched to 29 signal pathways, as shown in Figure 9. Among the 29 signaling pathways, 12 pathways were matched with those obtained by PBM analysis in this paper, which are Cytokine-cytokine receptor interaction, Legionellosis, Prion diseases, Viral protein interaction with cytokine and cytokine receptor, Complement and coagulation cascades, Bladder cancer, Fluid shear stress, and atherosclerosis, Prostate cancer, MAPK signaling pathway, TNF signaling pathway, IL-17 signaling pathway, Hepatitis B. All data, including sequencing reads and single-nucleus expression matrices have been deposited in NCBI's Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/). The accession number for the mRNA-seq data reported in this study is GSE172223.

8.PPI network and cluster analysis
For further analysis, we got 12 pathways that were matched in the KEGG results of MH7A and the KEGG results of PBM. The pathways were Cytokine-cytokine receptor interaction, Legionellosis, Prion diseases, Viral protein interaction with cytokine and cytokine receptor, Complement and coagulation cascades, Bladder cancer, Fluid shear stress and atherosclerosis, Prostate cancer, MAPK signaling pathway, TNF signaling pathway, IL-17 signaling pathway, Hepatitis B. The genes were constructed a PPI network in 12 pathways. The outcomes of the string pro le obtained from STRING showed that contained 129 nodes and 1235 edges. The top 10 outstanding proteins were identi ed as hub genes, which might play critical roles in the PBM through PPI analysis of STRING software including IL6, TNF, VEGFA, IL10, EGFR, EGF, CCL2, MMP9, IL1B, FOS. Among these signi cant proteins, IL6 was the most important protein and connected with 84 nodes in the network. A total of 6 clusters were generated in MCODE, and the top one cluster was selected as hub modules by the scores evaluated in MCODE. MCODE 1, which had 32 nodes and 369 edges, had the highest score in these clusters, such as Figure 10. In this cluster, the 9 nodes, including IL6, VEGFA, CXCL2, CSF1P, ITGAM, TNF, FOS, MMP9, PGF were in both MH7A cells and the PBM group. Maybe, these 9 nodes are key genes for PBM. On the one hand, in this cluster, we can know that some protein(red circles) of PBM results obtained by the method proposed in this paper already coincide with the parts of bio-MH7A cell sequencing. It proves the feasibility of our method. On the other hand, in LED stimulate-MH7A cells, the yellow circles, including CSF3, CXCL3, SOCS3, HMOX1, HBEGF, IL1B, IL1A, PLAU, and CCL20, represent the differentially expressed genes and have not been studied. Therefore, these genes have more research value and potential.

Effects of 630-nm LED radiation on mRNA gene expressions of in ammatory cytokines
For the accuracy of the PPI network and the effect of 640nm LED red light on rheumatoid disease, we performed RT-qPCR to analysis mRNA levels of in ammatory.The result was shown in Figure 11; there was a signi cant decrease in the gene expressions of in ammatory cytokines IL-6, IL-8, and IL-1Βand MMP-3 expression; meanwhile, the expression of antiin ammatory factor IL-10 was promoted with the radiation doses increased. Collectively, these results demonstrate that the gene expression and production of in ammatory cytokines by MH7A cells can be inhibited by 630-nm LED radiation.

Discussion
Although PBM has great potential applications, the mechanisms are not fully elucidated. By crawling the abstracts from the literature, the genes that have been studied were identi ed. And the enrichment analysis of KEGG and GO were adopted to know which pathways and functions were involved. Currently, most of the research on PBM is focuses on the PI3K-Akt signaling pathway and rheumatoid arthritis disease. At the same time, in the KEGG results of the laser group, the PI3K-Akt signaling pathway and rheumatoid arthritis signaling pathway appear again, which suggests that PBM regulates the PI3K-Akt and rheumatoid arthritis signaling pathway to play a role. Possibly, we put forward some hypotheses based on these results. Through PI3K-Akt and MAPK signaling pathways, rheumatoid arthritis could be improved by PBM. Coincidentally, one study has con rmed that 630nm LED can inhibit the proliferation of synovial broblasts in rheumatoid arthritis through the PI3K-Akt signaling pathway [18] .
The PI3K-Akt signaling pathway is one of the most investigated pathways in PBM. The p53 and PI3-Akt signaling pathways mediate the up-regulation of GADD45 to induce human RPE cell apoptosis [19] ; lowlevel laser therapy induces the proliferation, migration, and tube formation of human umbilical vascular endothelial cells by activating the PI3K/Akt signaling pathway [20] ; LLLT reduces cytokines by inhibiting PI3K activation of protein kinase A16 [21] ; LLLT (660 nm or 780 nm) can modify oral dysplastic cells (DOK) and oral cancer cells (SCC9 and SCC25) growth by modulating the Akt/mTOR/CyclinD1 signaling pathway [8] .
A large number of studies have demonstrated that the MAPK signaling pathway plays a key role in PBM.
450nm PBM controls the bacterial infection of periodontitis through ROS/MAPK/mTOR pathway [22] . Studies have also shown that PBM can improve acute in ammation in mice through the cannabinoid receptor/ATP-sensitive K(+) pathway/P38-MAPK signaling pathway [23] . Melatonin-treated cells were laserirradiated, and the differentiation and mineralization of cells were found to involve p38 MAPK and PRKD1 signaling mechanisms [24] .
Crawling information through the Internet is a promising strategy. Georgia crawled obituaries online in 2015 to analyze the link between parity and cancer risk [25] . In our study, we used web crawling to obtain genes for comparing two different phototherapy methods: PBM and PDT. The result was that 146 human genes related to PBM and 268 genes related to PDT. The proportion of gene intersection is 39.9%, and the proportion of KEGG of PBM and PDT is 78.9%. Therefore, it can be concluded that the two different phototherapy methods, PBM and PDT, may regulate the same pathway to play an important biological role.
Moreover, the light sources of LED and laser were compared to verify if there are signi cant differences. As a result, 75 phototherapy-related genes with LED and 220 phototherapy-related genes with laser were identi ed. The ratio of the two intersecting genes in LED genes is 98.9%, which means that most of the phototherapy related genes in LED are covered by the laser-related phototherapy genes. Therefore, it can also be proved that their effects e have no signi cant differences in phototherapy. For the 75 LED genes related to phototherapy, there was no KEGG result caused by few identi ed genes.
We speculated that the 29 genes in the PBM, PDT, LED, and laser are the key genes for the role of phototherapy. Therefore, the enrichment analysis was performed on the 29 genes to explore the mechanism of phototherapy. The result of KEGG was shown to be related to Longevity regulation, Amphetamine addiction, Melanoma, etc. GO results show that these genes were related to tyrosine, ultraviolet, light, and oxygen stress, etc. On the one hand, the results of KEGG of these 29 genes were only 7 pathways caused by few genes; on the other hand, from GO results, the function of these genes can be understood to further clarify the role of phototherapy.
Due to the repeated occurrence of RA signaling pathways in our results, we also performed gene sequencing in MH7A cells under the conditions of whether there was light or not in TNF-α stimulation to explore the differentially expressed gene. Sequencing results showed that there were 2950 differential genes, and KEGG enriched 29 pathways. The top ten enriched pathways were Cytokine-cytokine receptor interaction, Neuroactive ligand-receptor interaction, Protein processing in the endoplasmic reticulum, Cell cycle, Viral protein interaction with cytokine and cytokine receptor, Complement and coagulation cascades, Legionellosis, DNA replication, Prion diseases, Bladder cancer. Among the 29 signaling pathways, 12 of them coincide with those obtained by PBM analysis, which are Cytokine-cytokine receptor interaction, Legionellosis, Prion diseases, Viral protein interaction with cytokine and cytokine receptor, Complement and coagulation cascades, Bladder cancer, Fluid shear stress, and atherosclerosis, Prostate cancer, MAPK signaling pathway, TNF signaling pathway, IL-17 signaling pathway, Hepatitis B. These 12 signaling pathways may be important in PBM.
A PPI network was constructed in the genes of 12 pathways to explore outstanding proteins. The network contained 129 nodes and 1235 edges. And we identi ed ten outstanding proteins in this network, which were IL6, TNF, VEGFA, IL10, EGFR, EGF, CCL2, MMP9, IL1B, FOS. And IL6 was the most important that connected with 84 nodes in the network. In fact, in our previous study, we have demonstrated the effect of 630nm LED on MH7A cells, and the IL6 decreased after 630nm LED by ELISA experiment and q-PCR. In this network, there were six clusters generated and the cluster with the highest score was chosen. In this cluster, a total of nine proteins appeared in MH7A and PBM results, and the eight proteins appeared in only LED stimulate-MH7A cells. The nine proteins (red circles) had a very high degree, which proves the feasibility of our method. The eight proteins (yellow circles), including CSF3, CXCL3, SOCS3, HMOX1, HBEGF, IL1B, IL1A, PLAU, CCL20, represent the part of differentially expressed genes in LED stimulate-MH7A. It means that they not have been studied and have research value and potential in PBM. For the accuracy of the PPI network and the effect of 640nm LED red light ,the RT-qPCR was used to analysis mRNA levels of in ammatory. There was a signi cant decrease in the gene expressions of in ammatory cytokines IL-6, IL-8, and IL-1Βand MMP-3 expression; meanwhile, the expression of antiin ammatory factor IL-10 was promoted with the radiation doses increased.These results demonstrate that the gene expression and production of in ammatory cytokines by MH7A cells can be inhibited by 630-nm LED radiation.
In this experiment, due to the limitations of the PubMed search engine, we tried to use different keywords to search for articles related to light therapy. Although a simple search will expand the number of literature, the accuracy will also decrease. However, according to the late start of LED research, the number of related literature was not much. To expand the number of literature, the simplest search formula was chosen: light-emitting diode. Accuracy was sacri ced, but in post-processing, the intersections of LED genes with those identi ed in PBM and PDT genes were studied. Therefore, the method can guarantee both quantity and accuracy.
In gene identi cation, we used the gene_atomization (a function) in the pubmed.mineR package(R language). But the function has one thing worth noting. In some cases, it recognizes the carbon element 3 as C3 but enters it as the gene: complement 3. Therefore, after we used the function to identify the human gene ID, the confounding genes need to be removed arti cially. During manual processing, it was found that most of the confusing genes were 2-bit, such as those of PC and TG, which could be used to focus attention on this type of gene.

Conclusion
In summary, web crawling is a promising strategy to generate study hypotheses for guiding and prioritizing future studies. Used web crawling, bioinformatics analysis, and gene sequence, we identi ed 12 pathways and 10 genes that play important roles in PBM and predicted nine genes with research value. Anti-in ammatory effect of 630nm LED red light was demonstrated by RT-qPCR. Moreover, it was also veri ed that PBM and PDT, as well as LED and laser have no signi cant difference in effect on phototherapy.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding authors on reasonable request. The top 10 KEGG results of differentially expressed genes in MH7A cells. All KEGG analysis was based on a threshold of p-value <0.05.