Available via license: CC BY 4.0
Content may be subject to copyright.
1 / 46
1ldentify potential feature genes and immune cell infiltration of
2HIRI based on branched-chain amino acid-related genes by
3machine learning and experimental validation
4
5Jiahui Zhao1, ¶，Min Wu1, ¶, Shuangjiang Li1, ¶, Shuang Li1, Mingjuan Li,1
6Qiuyun Li,1 Guangdong Pan1, 2, 3, *, Guoqing Ouyang1, 2, 3, *, Honglai Xu1, 2, 3, *
71Department of General Surgery, Liuzhou People's Hospital Affiliated to Guangxi
8Medical University, Liuzhou,Guangxi, China
92Liuzhou Hepatobiliary and Pancreatic Diseases Precision Diagnosis Research Center
10 of Engineering Technology, Liuzhou People ’s Hospital Affiliated to Guangxi
11 Medical University, Liuzhou, Guangxi, China
12 3Liuzhou Key Laboratory of Liver Cancer Research, Liuzhou People ’s Hospital,
13 Liuzhou, Guangxi, China
14 * Corresponding author:
15 Guandong Pan, Department of General Surgery, Liuzhou People’s Hospital Affiliated
16 to Guangxi Medical University, Liuzhou 545006, Guangxi, China
17 Email: pgdhx@126.com
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
2 / 46
18 Guoqing Ouyang, Department of General Surgery, Liuzhou People ’s Hospital
19 Affiliated to Guangxi Medical University, Liuzhou 545006, Guangxi, China
20 Email: ouyangguoqing@stu.gxmu.edu.cn
21 Honglai Xu, Department of General Surgery, Liuzhou People’s Hospital Affiliated to
22 Guangxi Medical University, Liuzhou 545006, Guangxi, China
23 Email: xuhonglai1973@163.com
24
25
26 ¶ These authors have contributed equally to this work as co-first authors.
27 * These authors have contributed equally to this work as corresponding authors.
28
29
30 Abstract
31 Background: Branched-chain amino acid metabolism is involved in the pathogenesis
32 of various liver diseases. In this study, we investigate the potential role of
33 branched-chain amino acid metabolism-related genes in the pathogenesis of hepatic
34 ischemia reperfusion (HIRI).
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
3 / 46
35 Methods: The gene Expression profiles of HIRI were obtained from the Gene
36 Expression Omnibu database. To determine the differential expression of
37 branched-chain amino acid metabolism-related genes between HIRI and normal
38 tissues. Then, the GO and KEGG analyses were performed, and the protein-protein
39 interaction network was constructed. Next, the random forest and LASSO algorithms
40 were used to screen hub genes, and machine learning techniques were used to build
41 diagnostic models. immunoinfiltration was analyzed in both HIRI patients and
42 controls and the ceRNA network was established. Finally, quantitative real-time PCR
43 was used to verify the expression of hub gene.
44 Results: Based on data set GSE23649, three central DEGs (SLC7A5, SLC1A5,
45 SLC43A2) were determined by the intersection of three machine learning algorithms
46 and used to establish a nomogram that yielded a high predictive performance (area
47 under the curve 0.733−0.922). In the external GSE15480 dataset, AUC value for three
48 key genes is as high as 1.000. Further analysis of nomogram, decision curve and
49 calibration curve also confirme the predictive efficacy of diagnosis. GSEA and GSVA
50 suggest that these three marker genes were involved in multiple pathways associated
51 with HIRI progression. Immunoinfiltration analysis suggest that the proportion of
52 macrophages, neutrophils, aDCs, Treg, and Th1 cells in HIRI group is higher than
53 that in control group, with statistical significance(P<0.05). The ceRNA network
54 demonstrates the complex regulatory relationships among the three hub genes and
55 these mRNA levels were further confirmed in mouse HIRI liver samples.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
4 / 46
56 Conclusions: Our study have provided a comprehensive understanding of the
57 association between branched-chain amino acid and HIRI, may provide potential
58 target for HIRI treatment and diagnosis. And provide new insights into the
59 mechanisms of HIRI.
60 Keywords: hepatic ischemia-reperfusion injury, branched-chain amino acid
61 metabolism, machine learning, Biomarkers, Bioinformatics analysis.
62 Abbreviations
63 AUC, area under the curve; CC, cellular component; DEGs, Differentially expressed
64 genes; DCA, decision curve analysis; GEO, Gene Expression Omnibus; GO, Gene
65 Ontology; GSEA, gene set enrichment analysis; GSVA, gene set variation analysis;
66 HIRI, Hepatic ischemia and reperfusion injury; KEGG, Kyoto Encyclopedia of Genes
67 and Genomes; LASSO, least absolute shrinkage and selection operator; PPI, protein
68 protein interaction; RF, random forest; RFE, support vector machine recursive feature
69 elimination; ROC, receiver operating characteristics; SVM-ssGSEA, single-sample
70 gene set enrichment analysis; TCA, tricarboxylic acid;
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
5 / 46
71 Introduction
72 During liver resection, ischemia-reperfusion (I/R) is a necessary procedure to
73 minimize intraoperative bleeding. However, this process invariably causes liver
74 damage and can lead to severe postoperative complications, including liver failure1.
75 The widespread use of liver transplantation and hepatectomy has heightened interest
76 in strategies to prevent and mitigate hepatic ischemia-reperfusion injury (HIRI). The
77 mechanisms underlying HIRI involve both direct cellular damage due to ischemia and
78 delayed injury caused by the activation of inflammatory pathways2. HIRI progresses
79 in two distinct phases: ischemia and reperfusion. In the ischemic phase, reduced blood
80 flow and oxygen deprivation in liver tissues lead to the accumulation of reactive
81 oxygen species (ROS). This increase in ROS triggers oxidative stress, contributing to
82 endothelial dysfunction, DNA damage, and localized inflammation. These
83 inflammatory cascades and oxidative stress can provoke a cytokine storm, leading to
84 cell death and tissue damage3. In the reperfusion phase, the restoration of blood flow
85 to ischemic tissues results in the release of phospholipase A2, TNF-α, IL-1β, IFN-
86 γ, and angiotensin II, which activate NADPH oxidase. Phospholipase A2 promotes
87 the production of platelet-activating factor, increasing thromboxane and leukotrienes,
88 which exacerbate local inflammation4–6. Additionally, angiotensin II stimulates its
89 receptors, upregulating NADPH oxidase expression and contributing to
90 ischemia-reperfusion injury through the angiotensin-converting enzyme pathway7,8.
91 Despite extensive basic and clinical research, the precise mechanisms underlying
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
6 / 46
92 ischemia-reperfusion injury remain to be further clarified.
93 Branched-chain amino acids (BCAAs), comprising valine, leucine, and
94 isoleucine, are essential for human nutrition. BCAAs are known to promote skeletal
95 muscle protein synthesis, enhance protein metabolism, and inhibit protein
96 degradation9. Previous study have demonstrated that BCAAs stimulate mTOR
97 mRNA translation, thereby regulating protein synthesis at the molecular level10.
98 BCAAs also mitigate hepatic steatosis and liver injury associated with nonalcoholic
99 steatohepatitis by inhibiting the expression of the FAS gene(1)11. Furthermore,
100 BCAAs can block the Akt2-INSIG2a signaling pathway through the mTOR pathway,
101 reducing intrahepatic adipogenesis and redistributing lipids in the liver, muscles, and
102 kidneys12. Recent research also indicate that fructose ingestion induces the hepatic
103 transcription factor ChREBP- β, which activates BCKDK transcription, thereby
104 linking fructose metabolism with BCAA metabolism13. Although significant progress
105 has been made in understanding the relationship between BCAA metabolic regulation
106 and disease, the potential role of BCAAs in HIRI remains unclear, further research is
107 needed to explore the potential role of BCAA as a therapeutic target for HIRI.
108 In recent years, bioinformatics and machine learning strategies have advanced
109 rapidly, enabling a deeper exploration of underlying mechanisms and potential
110 biomarkers. In this study, we analyze the differential expression and immunological
111 characteristics of BCAA-related genes(BRGs) in 33 cases of HIRI and 33 control
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
7 / 46
112 cases. Machine learning algorithms were employed to identify key genes for
113 diagnostic prediction. We validated the predictive model using a nomogram, decision
114 curve analysis (DCA), calibration curve, and receiver operating characteristic (ROC)
115 curve. Additionally, we explored the relationship between hub BRGs and immune
116 infiltration and constructed a ceRNA network.
117 Materials and methods
118 Data collection and processing
119 The GSE23649 dataset was retrieved from the NCBI Gene Expression
120 Omnibus(GEO; https://www. ncbi.nlm.nih.gov/geo/) using the keywords "Hepatic
121 ischemia reperfusion." This dataset, used as the training set, comprises 33 hepatic
122 ischemia-reperfusion samples (9 living donor samples, 8 cardiac death donor samples,
123 8 brain death donor samples with early allograft dysfunction, and 8 brain death donor
124 samples without early allograft dysfunction) and 33 control samples. The GSE15480
125 dataset served as the validation set, consisting of 6 control samples and 6 HIRI
126 samples. The flowchart of this study is shown in Figure 1.
127 Differential gene expression analysis of DEGs
128 Differentially expressed genes (DEGs) related to branched-chain amino acids
129 (BCAA) between the hepatic ischemia-reperfusion and control groups were identified
130 using the Wilcoxon signed-rank test. To enhance the visualization of these DEGs, the
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
8 / 46
131 "ggplot2" and "pheatmap" R packages were employed to generate boxplots and
132 heatmaps. The "VennDiagram" R package was used to determine the intersection of
133 BRGs, which were defined as the BRGs for subsequent analyses. Statistical
134 significance was set at p < 0.05.
135 Correlation analysis and PPI network construction
136 Landscape plots of the 23 chromosomes and a heatmap of the 13 DE-BRGs were
137 generated using the "RCircos" and "heatmap" R packages, respectively. Pearson’s
138 correlation analysis was performed to create Circos plots of DE-BRGs using the
139 “circlize ” package. A protein-protein interaction (PPI) network for the 13
140 DE-BRGs was constructed using the STRING database(https://string-db.org/) , with a
141 medium confidence threshold of 0.4.
142 Pathway and functional enrichment analysis
143 The Kyoto Encyclopedia of Genes and Genomes (KEGG) database was utilized
144 as a comprehensive resource for the systematic analysis of gene functions14. Gene
145 ontology (GO) analysis identified the biological processes (BP), cellular components
146 (CC), and molecular functions (MF) associated with the 13 DE-BRGs. The
147 "clusterProfiler" R package was used for both GO and KEGG pathway analyses, with
148 visualizations generated using the "enrichplot" R package. The significance threshold
149 for enrichment was set at p < 0.05.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
9 / 46
150 Construction of DEGs diagnostic model
151 The GSE23649 dataset was used as the training set, while the GSE15480 dataset
152 was used as the validation set for the machine learning model. Three
153 algorithms—random forest (RF), support vector machine-recursive feature
154 elimination (SVM-RFE), and least absolute shrinkage and selection operator
155 (LASSO) regression—were employed to identify the most robust hub genes for
156 diagnosing hepatic ischemia-reperfusion injury (HIRI). LASSO regression,
157 implemented using the "glmnet" R package, reduced data dimensionality while
158 preserving valuable variables15,16. SVM-RFE is a method for detecting optimal
159 variables in machine learning by pruning support vectors. If there are many features,
160 halve the features for each round17. The "e1071" R package was used to construct the
161 SVM-RFE models for screening the best variable genes. The RF model is an
162 ensemble machine learning method for determining the optimal number of variables
163 using various independent decision trees18. The result can be obtained using the
164 "randomForest" R package. The intersection of the RF, LASSO algorithm, and
165 SVM-RFE results was taken to yielded the most significant hub genes.
166 ROC evaluation and nomogram construction
167 To assess the predictive accuracy of each hub gene, receiver operating
168 characteristic (ROC) curves were generated, and the area under the curve (AUC)
169 along with 95% confidence intervals (CI) were calculated using the "pROC" R
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
10 / 46
170 package. The external GSE15480 dataset was used to validate the diagnostic model. A
171 nomogram was constructed using the "rms" R package, where each hub gene was
172 assigned a score. The total score predicted the likelihood of HIRI occurrence. Clinical
173 impact curves and decision curves were also generated to evaluate the model's clinical
174 efficacy.
175 Gene set enrichment analysis and gene set variation analysis
176 To explore the potential functions of the hub genes, we conducted gene set
177 enrichment analysis (GSEA) using the KEGG gene set (c2.cp.kegg.symbols.gmt) and
178 GO gene set (c5.go.symbols.gmt) from the Molecular Signatures Database. The
179 "clusterProfiler" R package was used for GSEA, while gene set variation analysis
180 (GSVA) was performed using the "GSVA" R package to identify differentially
181 enriched pathways between high and low expression subtypes based on hub gene
182 expression. Differential expression pathways were further identified using the
183 "limma" R package.
184 Analysis of immune cell infiltration
185 The CIBERSORT algorithm was applied to convert normalized gene expression
186 matrices into immune cell components19. We analyzed the proportion of 22 immune
187 cells in the GSE23649 dataset, with a p-value < 0.05 defining accurate immune cell
188 fractions. Correlation heatmaps of immune cells in HIRI pathogenesis were generated
189 using the "corrplot" R package, and boxplots were used to compare the expression
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
11 / 46
190 levels of different immune cells. Spearman correlation analysis was conducted to
191 determine associations between hub genes and immune infiltrating cells using the
192 "GSVA" R package.
193 ceRNA network construction
194 We use miRanda (http://www.microrna.org/), miRDB (http:// www.mirdb.org/)
195 and TargetScan (https://www.targetscan.org/ The vert 80/) database to predict
196 miRNA-mRNA interactions between these three hub genes. Sponge scan
197 (http://spongescan.rc.ufl.edu/) was used to predict the target of the direct interaction
198 between miRNA and lncRNA. Then, the mRNA-miRNA-lncRNA ceRNA network
199 was established and visualized by Cytoscape software (version 3.9.0).
200 HIRI mouse model establishment and histological procedure
201 C57BL/6J mice were obtained from Hunan Slack Jingda Laboratory Animal Co.,
202 LTD. Mice had unrestricted access to sterile water and food. In 8-week-old mice,
203 hepatic ischemia was induced by clamping the hepatic artery and portal vein for 30
204 minutes, followed by reperfusion. Mice were euthanized after 6 hours of reperfusion,
205 and serum and liver samples were collected. Serum ALT and AST levels determined
206 by a fully automated biochemical immunoassay system (cobas® 8000, Germany).
207 Fresh liver tissue was immersed in a 4% paraformaldehyde solution and used for
208 paraffin embedding.4 mm thick sections were obtained using a microtome and
209 subsequently stained with H&E according to the manufacturer's protocol. All animal
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
12 / 46
210 experiments were approved by the Animal Care and Use Committee of Liuzhou
211 People's Hospital Affiliated to Guangxi Medical University.
212 RNA isolation and real-time quantitative PCR
213 RNA was extracted from HIRI mouse liver tissues using TRIzol reagent
214 (15596026CN, thermofisher, Guangzhou, China). RNA was reverse transcribed using
215 the ReverAid™ Master Mix with Dnase I-LBID(M16325,thermgfisher, Guangzhou).
216 Real-time quantitative PCR (qRT-PCR) was performed using the lightCysler @96
217 automatic fluorescence quantitative PCR instrument (Roche Diagnostic Products
218 Co.Ltd., Shanghai, China) and PowerUp ™ SYBR ™ Green (A25742, thermofisher,
219 Guangzhou). Primer sequences are listed in Supplementary Table S1.
220 Statistical analysis
221 Statistical and data analyses were performed using R software (version 4.2.1).
222 Continuous data are expressed as mean ± standard deviation. For comparisons
223 between two groups, the Student’s t-test was applied to normally distributed variables,
224 while the Wilcoxon rank-sum test was used for non-normally distributed variables. A
225 p-value less than 0.05 was considered statistically significant. ns, P > 0.05; *, P <
226 0.05; **, P < 0.01; ***, P < 0.001.
227 Results
228 Differential expression genes identification in HIRI
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
13 / 46
229 With the “limma” R package, a total of 5110 Differentially expressed genes
230 (DEGs) were identified based on the HIRI dataset GSE23649, of which 2,701 were
231 downregulated and 2,409 were upregulated (Figure 2A). The gene expression patterns
232 of 5110 DEGs were shown in the heatmap (Figure 2B). The intersection of 50110
233 DEGs and 33 genes related to BCAA metabolism revealed 13 BCAA-related DEGs
234 (DE-BRGs) (ABAT, ACAD8, ACADSB, AUH, BCKDHB, DLD, ECHS1, IVD,
235 MCCC1, SLC1A5, SLC3A2, SLC43A2, SLC7A5) with significant differences
236 between HIRI and control group (Figure 2C). The chromosomal locations of the 13
237 DE-BRGs are shown in Figure 2D. We found that SLC7A5, SLC1A5, SLC3A2, and
238 SLC43A2 were up-regulated in HIRI, while DLD, AUH, MCCC1, ACAD8, IVD,
239 ABAT, ACADSB, ECHS1, and BCKDHB were down-regulated (Figures 2E, F). The
240 correlation among the 13 DE-BRGs is shown in Figure 2G-H. MCCC1 was positively
241 correlated with ABAT，ACAD8 and BCKDHB, negatively correlated with SLC1A5
242 and SLC3A2. To investigate the potential relationship between these 13 DE-BRGs,
243 we performed PPI analysis as shown in Figure 2I.
244 Functional enrichment analysis of DE-BRGs in HIRI
245 To gain deeper insights into the signaling pathways involved in HIRI and
246 DE-BRGs, we conducted pathway and functional enrichment analyses using the
247 "ClusterProfiler" R package to elucidate the associated biological functions and
248 pathways. The Gene Ontology (GO) enrichment analysis revealed that the most
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
14 / 46
249 significant biological processes (BP) included the response to branched-chain amino
250 acid catabolic process, branched-chain amino acid metabolic process, amino acid
251 catabolic process, and organic acid catabolic process. Cellular component (CC)
252 analysis highlighted significant involvement in the mitochondrial matrix, basal plasma
253 membrane, basal part of the cell. Molecular function (MF) analysis identified strong
254 associations with neutral L-amino acid transmembrane transporter activity, L-amino
255 acid transmembrane transporter activity, amino acid transmembrane transporter
256 activity, and carboxylic acid transmembrane transporter activity (Figure 3A).
257 Additionally, KEGG pathway analysis indicated that the 13 DE-BRG were notably
258 enriched in pathways such as valine, leucine, and isoleucine degradation, propanoate
259 metabolism, lipoic acid metabolism, and 2-oxocarboxylic acid metabolism (Figure
260 3B). These findings align with the results of the GO enrichment analysis.
261 Construction of the diagnostic hub genes for HIRI
262 To identify the hub genes with the highest diagnostic value, we employed
263 LASSO regression and two validated machine learning models—support vector
264 machine-recursive feature elimination (SVM-RFE) and random forest (RF)—to filter
265 the 13 DE-BRGs for their predictive potential in diagnosing HIRI. LASSO logistic
266 regression identified 8 key genes from the DE-BRGs (Figure 4A-B). The RF model
267 selected all 8 DE-BRGs (Figure 4C-D). The SVM-RFE model achieved the smallest
268 error and highest accuracy (minimal error = 0.391, maximal accuracy = 0.609) when
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
15 / 46
269 selecting 5 features: SLC7A5, SLC43A2, BCKDHB, SLC1A5, IVD (Figure 4E-F).
270 The intersection of genes identified by the RF, LASSO, and SVM-RFE models
271 revealed three hub genes—SLC7A5, SLC1A5, and SLC43A2—that were considered
272 to possess the highest diagnostic value (Figure 4G).
273 Evaluation of the predictive value and nomogram construction
274 ROC curve analysis demonstrated robust predictive performance for each of the
275 hub genes, with area under the curve (AUC) values as follows: SLC7A5 (AUC:
276 0.791), SLC1A5 (AUC: 0.729), and SLC43A2 (AUC: 0.761)(Figure 5A). The overall
277 model achieved an AUC of 0.834 (95% CI: 0.733−0.922) (Figure 5B). To further
278 assess the predictive efficiency of these three hub genes, we constructed a nomogram
279 model for HIRI patients using the "rms" R package (Figure 5C).In this nomogram, the
280 relative expression of each gene corresponds to a score, and the total
281 score—calculated by summing the scores of all genes—predicts the risk of HIRI. The
282 calibration curve indicates the relative relationship between the predicted and actual
283 probabilities, with closer solid and dashed lines indicating higher model accuracy.
284 The predicted curve matches the standard curve well, indicating that the nomogram
285 model for HIRI is accurate (Figure 5D). DCA indicated that the “Model ” curve
286 was higher than the gray line, suggesting a high level of accuracy and offered a better
287 clinical benefit(Figure 5E). The clinical impact curve further confirmed this result
288 (Figure 5F).
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
16 / 46
289 To further validate the model's ability to identify HIRI patients, we tested its
290 performance using the GSE15480 dataset. The result showed that SLC7A5, SLC1A5,
291 and SLC43A2 in HIRI group was upregulated compared with control group (Figure
292 5G). The AUC value of ROC for SLC7A5, SLC1A5, and SLC43A2 were 1.0, 0.944,
293 0.861, respectively. Additionally, the AUC value of model was 1.0, confirming the
294 model's strong predictive accuracy (Figure 5H). These findings suggest that the three
295 identified hub genes—SLC7A5, SLC1A5, and SLC43A2—may serve as reliable
296 diagnostic markers for HIRI.
297 Profile of GSEA and GSVA
298 To elucidate the key signaling pathways associated with hub genes in hepatic
299 ischemia-reperfusion injury (HIRI), we performed single-gene Gene Set Enrichment
300 Analysis (GSEA) using KEGG and GO pathway datasets. The GSEA of KEGG
301 pathways revealed that high expression of SLC43A2, SLC1A5, and SLC7A5 was
302 significantly associated with antigen processing and presentation. Additionally,
303 SLC43A2 and SLC7A5 were linked to cytokine-cytokine receptor interaction, the
304 MAPK signaling pathway, and the NOD-like receptor signaling pathway (Figure
305 6A-B). Both SLC1A5 and SLC7A5 were associated with systemic lupus
306 erythematosus. In contrast, high expression of SLC1A5 was connected to the mTOR
307 signaling pathway, N-Glycan biosynthesis, the proteasome, ECM-receptor
308 interactions, and systemic lupus erythematosus (Figure 6C). SLC7A5 was also
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
17 / 46
309 notably associated with complement and coagulation cascades. The GO enrichment
310 results from GSEA are presented in Figures 6D-F.
311 We then employed GSVA to identify differentially active pathways between
312 low- and high-expression subtypes based on the expression levels of the three hub
313 genes. The results indicated that the expression of both SLC7A5 and SLC43A2 were
314 associated with Type I diabetes mellitus and graft versus host diseases. The low
315 expression of SLC1A5 is implicated in pathways such as nitrogen metabolism, ECM
316 receptor interaction, hedgehog signaling pathway, and basal cell carcinoma.
317 Downregulated of SLC7A5is associated with graft versus host disease, intestinal
318 immune network for IGA production, and NOD-like receptor signaling pathway.
319 Whereas, decreased expression of SLC43A2 were correlated with autoimmune
320 thyroid disease, leishmania infection (Figure 6G, I). High expression of SLC7A5 was
321 significantly involved in the phenylalanine metabolism (Figure 6H). Additionally,
322 high expression of SLC1A5 was correlated with taste transduction and
323 glycosaminoglycan biosynthesis keratan sulfate (Figure 6I).
324 Immunoinfiltration analysis
325 To explore whether BRGs could influence HIRI progression by modulating
326 immune infiltration, we employed single-sample GSEA (ssGSEA) to investigate the
327 correlation between these hub genes and immune cells and functions infiltration in
328 HIRI. The heatmap show the distribute of immune cells in control and HIRI group
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
18 / 46
329 (Figure7A). As for immune cells, the ssGSEA algorithm revealed that aDCs,
330 Macrophages, Neutrophils, Th1 cells, and Treg were upregulated in HIRI group. In
331 term of immune functions, the ssGSEA scores of APC co-inhibition, CCR,
332 inflammation promoting, MHC class I, T cell co-inhibition, Type I IFN response were
333 higher in HIRI group than control group, while cytolytic activity and Type II IFN
334 response were higher in control group (Figure7C).
335 Figure 7B and Supplementary Table S2 showed the correlation between 3 hub
336 genes and immune infiltration. The result showed that SLC1A5 was significantly
337 related to T helper cells, T cell co-stimulation, APC co-inhibition. SLC43A2
338 associated with aDCs, CCR, MHC class I, Type I IFN response, Parainflammation,
339 APC co-stimulation. In addition, SLC7A5 significantly correlated with aDCs, CCR,
340 Neutrophils, Parainflammation, Treg. Figure 7D shows the correlation between
341 immune cells. These findings suggest that the three hub genes are closely linked to the
342 immune infiltration microenvironment in HIRI.
343 CeRNA networks based on marker genes
344 To further explore the regulatory mechanisms of the 3 hub genes, we constructed
345 a competing endogenous RNA (ceRNA) network using the TargetScan, miRanda, and
346 miRDB databases. The analysis identified 3 mRNAs, 76 miRNAs, and 150 lncRNAs
347 (Figure 8A). Our results indicated that 128 lncRNAs can regulate the expression of
348 SLC43A2 through competitively binding of hsa-miR-149-3p, hsa-miR-1976,
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
19 / 46
349 hsa-miR-612, hsa-miR-1207-5p and hsa-miR-129-5p. Among them, 19 shared
350 lncRNAs targeted has-miR-149-3p and 17 shared lncRNAs targeted has-miR-1976. A
351 total of 38 lncrnas were able to competitively bind to 7 mirnas, Including
352 hsa-miR-612, hsa-miR-185-3p, hsa-miR-615-5p, hsa-miR-875-3p, hsa-miR-486-3p,
353 hsa-miR-92b-5p, and hsa-miR-194-5p, and subsequently regulate SLC7A5. Among
354 them, 16 lncrnas targeted hsa-miR-612 and 7 lncrnas targeted hsa-miR-185-3p. For
355 SLC1A5, there are two regulated mirnas, including hsa-miR-361-3p and
356 hsa-miR-2355-5p. In addition, both hsa-miR-486-3p and hsa-miR-612 were involved
357 in regulating the expression of SLC43A2 and SLC7A5. Notably, seven miRNAs,
358 including has-miR-541-3p and has-miR-31-5p, could simultaneously bind lncRNA
359 C10orf91 to regulate the expression of SLC7A5 and SLC43A2.
360 Altered expression of DE-BRGs in HIRI
361 AST and ALT measurements revealed a significant increase in the HIRI group
362 compared to the control group (Figure 9A-B). Furthermore, HE staining showed
363 pronounced liver damage in the HIRI group, characterized by a marked loss of liver
364 architecture, disintegration of hepatic cords, and red blood cell exudation from the
365 hepatic cords (Figure 9C). Collectively, these results confirm the successful
366 establishment of the HIRI model.
367 To further elucidate the role of BRGs in HIRI, we assessed the mRNA
368 expression levels of SLC7A5, SLC1A5, and SLC43A2 using qRT-PCR. The results
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
20 / 46
369 demonstrated a significant upregulation of these 3 hub genes in HIRI mice compared
370 to control mice (Figure 9D-F). These findings suggest that BRGs play a pivotal role in
371 HIRI development, thereby supporting their potential as key regulators in HIRI
372 progression.
373 Discussion
374 Ischemia-reperfusion injury is a severe condition that necessitates medical
375 intervention to limit cellular damage and maintain organ function20. It frequently
376 arises in medical procedures such as hemorrhagic shock resuscitation, liver
377 transplantation, and partial liver resection, and plays a significant role in liver
378 damage, which can ultimately result in organ failure21. From a mechanistic viewpoint,
379 mitochondrial dysfunction is the driving force behind hepatic ischemia-reperfusion
380 injury (HIRI), contributing to oxidative stress, metabolic imbalances, inflammation,
381 and immune dysregulation22. BCAAs are vital for mammalian growth, and impaired
382 BCAA metabolism has been recognized as a major risk factor for coronary artery
383 disease, diabetes, heart failure, and myocardial ischemia-reperfusion injury23–25.
384 Nonetheless, the involvement of BCAA metabolic pathways in the development of
385 HIRI has not been thoroughly investigated. Consequently, this study aims to uncover
386 BCAAs associated with HIRI and assess the potential diagnostic and therapeutic
387 relevance of DE-BRGs in HIRI.
388 In this study, we analyzed the differential expression of BRGs in HIRI and
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
21 / 46
389 control liver samples obtained from the GEO database, ultimately identifying 13
390 DE-BRGs associated with BCAA metabolism. Compared to healthy individuals, the
391 abnormally BRGs expressed in HIRI patients, suggesting their critical role in the
392 onset and progression of HIRI. Subsequently, we analyzed the correlations among
393 DE-BRGs, revealing that some exhibit significant synergistic or antagonistic
394 interactions, indicating their mutual regulatory roles in HIRI pathogenesis. At the
395 protein level, several pairs, including ECHS1-ABAT and
396 SLC1A5-SLC43A2-SLC7A5-SLC3A2, as well as the
397 AUH-ACAD8-BCKDHB-DLD-IVD-ACADSB-MCCC1-ECHS1 group, were found
398 to be closely related, suggesting extensive interactions between BRGs at both the
399 gene and protein levels during HIRI development.
400 GO and KEGG enrichment analyses were conducted to explore the potential
401 functions of DE-BRGs in HIRI. The findings primarily indicate associations with
402 amino acid catabolism, the mTOR signaling pathway, and the TCA cycle. This aligns
403 with earlier studies showing that BCAAs stimulate mTOR mRNA translation,
404 regulating protein synthesis at the molecular level.10. Meanwhile, BCKDHA, a
405 metabolic enzyme involved in BCAA catabolism, is regulated by the key enzyme
406 BCKDK26,27. In α-cell-specific BCKDHA knockout mice, Yang et al. demonstrated
407 that impaired BCAA catabolism could reduce mTOR signaling, thereby inhibiting
408 α-cell proliferation28. Zhai et al. similarly noted that APN promotes ERK signal
409 transduction in HCC cells by mediating BCKDK phosphorylation, which enhances
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
22 / 46
410 HCC proliferation and metastasis29. However, our results reveal that the key
411 regulators of branched-chain amino acid metabolism, BCKDHA and BCKDK, did not
412 exhibit significant differences between HIRI and control liver samples, diverging
413 from previous studies29,30. Therefore, further research is required to clarify the
414 relationship between DE-BRGs and BCAA metabolism in HIRI. Additionally, GO
415 analysis of the cellular component emphasizes that these 13 DE-BRGs primarily
416 function within the mitochondrial matrix. This aligns with Kitagawa's study, which
417 highlights the critical role of mitochondria in maintaining cellular homeostasis,
418 particularly in processes like apoptosis, inflammation, and ROS synthesis31. Under
419 stress conditions such as hypoxia or cytotoxin activation, mitochondrial membrane
420 integrity is compromised, triggering the pro-apoptotic Bcl-2 family and exacerbating
421 ischemia-reperfusion injury32,33. These findings suggest that mitochondrial
422 dysfunction plays a crucial role in HIRI.
423 Machine learning algorithms, which remove redundant factors and retain only
424 variables relevant to the outcome, have seen growing use in medical research34–36. In
425 our study, we applied three machine learning algorithms—LASSO, Random Forest
426 (RF), and SVM-REE—to identify three hub genes (SLC7A5, SLC1A5, SLC43A2)
427 that accurately predict the risk of hepatic ischemia-reperfusion injury (HIRI),
428 achieving an AUC value of 0.834. We further validated the model using an external
429 dataset (GSE15480), where the AUC value reached 1.000. Additionally, in the
430 validation set, the AUC values for SLC7A5, SLC1A5, and SLC43A2 each exceeded
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
23 / 46
431 0.85, highlighting their strong predictive value as sensitive biomarkers for HIRI.
432 Using these three hub genes, we constructed a nomogram model, along with a
433 calibration curve and decision curve analysis (DCA), to further validate the diagnostic
434 accuracy and clinical relevance of the model. Although the calibration curve showed
435 close alignment with the standard curve, it did not yet reach an optimal level, possibly
436 due to the limited sample size, which requires further investigation.
437 Long non-coding RNAs (lncRNAs), functioning as competitive endogenous
438 RNAs (ceRNAs), can bind to miRNAs competitively, thereby regulating mRNA
439 expression and influencing physiological interactions between various cell types37.
440 Based on the potential interaction within the lncRNA-miRNA-mRNA axis, we
441 constructed a ceRNA network for HIRI. Our findings revealed that 19 shared
442 lncRNAs target hsa-miR-149-3p to regulate SLC43A2 expression. Ma et al.
443 discovered that microcystin-LR (MC-LR) promotes the expression of hsa-miR-149-3p
444 and may contribute to MC-LR-induced hepatitis and liver cancer through the MAPK
445 pathway38. Additionally, 16 shared lncRNAs target hsa-miR-612 to regulate SLC7A5
446 expression. Lu et al. reported that circETFA and CCL5 competitively bind to
447 hsa-miR-612, promoting hepatocellular carcinoma (HCC) progression by modulating
448 the PI3K/Akt signaling pathway.39. These results suggest that lncRNAs may regulate
449 the expression of three marker genes (SLC7A5, SLC1A5, and SLC43A2), offering
450 new insights into the pathogenesis of HIRI. However, the precise mechanisms
451 involving these characteristic genes in HIRI require further validation through in vitro
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
24 / 46
452 and in vivo studies.
453 SLC7A5 encodes LAT1, a sodium-independent transporter primarily responsible
454 for transporting BCAAs40. SLC7A5 encodes LAT1, a sodium-independent transporter
455 primarily responsible for transporting branched-chain amino acids (BCAAs)41–43.
456 Additionally, increased levels of activating transcription factor 4 (ATF4) upregulate
457 SLC7A5, leading to greater uptake of isoleucine and leucine, which in turn activate
458 mTORC1 and inhibit autophagy44. In this study, we observed increased mRNA
459 expression of SLC7A5 in HIRI mice compared to controls, although further
460 protein-level validation is needed. GSEA and GSVA indicated that SLC7A5
461 participates in several immune and inflammatory processes, including
462 cytokine-cytokine receptor interactions, MAPK signaling, NOD-like receptor
463 signaling, and interleukin-21 production. Members of the MAPK family include
464 ASK1, TAK1, ERK1/2, JNK, and p3845. Deng et al. demonstrated that BPS reduces
465 inflammation, apoptosis, and autophagy by inhibiting p38 and JNK pathways, thereby
466 alleviating HIRI46. Similarly, Hou et al. showed that the MAPK signaling pathway
467 mediates the activation of the NLRP3 inflammasome, triggering pro-inflammatory
468 cytokine secretion and hepatocyte damage47. Thus, further investigation into the
469 relationship between SLC7A5 and BCAA metabolism in HIRI is essential for future
470 research.
471 The SLC1A5 protein functions as a transporter that facilitates the absorption of
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
25 / 46
472 neutral amino acids and plays a crucial role in glutamine uptake48. As an essential
473 energy source for the enterohepatic circulation, elevated intestinal glutamine levels
474 can increase blood concentrations of αKG, which promotes macrophage M2
475 polarization by enhancing the TCA cycle to repair HIRI. In our study, we observed
476 that SLC1A5 expression was higher in HIRI samples compared to control liver
477 samples, indicating a positive correlation between SLC1A5 expression and HIRI.
478 GSEA and GSVA revealed that SLC1A5 is involved in ECM receptor interaction,
479 mTOR signaling pathways, and N-polysaccharide biosynthesis in pre-ribosomes.
480 Similar to SLC7A5, SLC1A5 regulates mTORC1 signaling activation, leading to
481 increased phosphorylation of mTORC1 and subsequent regulation of cell growth and
482 proliferation49. Wang J's study demonstrated that suberoylanilide hydroxamic acid
483 alleviated orthotopic liver transplantation (OLT)-induced IRI by upregulating
484 autophagy in Kupffer cells via the AKT/mTOR signaling pathway. Additionally,
485 SLC1A5 has been shown to be upregulated in various tumors, including melanoma,
486 lung cancer, colon cancer, and breast cancer, supporting the significantly increased
487 energy demands of tumor cells compared to normal cells50–53.
488 SLC43A2 (LAT4) is a uniporter that transports large, essential neutral amino
489 acids, primarily BCAA54. It is highly expressed in tissues such as the placenta,
490 peripheral blood leukocytes, kidneys, small intestine, and brain. Bian et al. found that
491 tumor cells outcompete T cells for methionine uptake through SLC43A2, which
492 affects histone methylation and T cell function. Inhibiting SLC43A2 restores T cell
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
26 / 46
493 function, thereby enhancing anti-tumor immunity in preclinical models55. Peng H et
494 al. discovered that in esophageal squamous cell carcinoma (ESCC), SLC43A2
495 activates the NF-κB signaling pathway by increasing methionine uptake, creating a
496 positive feedback loop that further upregulates SLC43A2 expression and promotes
497 ESCC progression56. GSEA and GSVA revealed that SLC43A2 is involved in several
498 immune processes, including the MAPK, NOD-like receptor, and chemokine
499 signaling pathways. In our study, we observed elevated SLC43A2 expression in liver
500 samples from the HIRI group compared to the control group, consistent with previous
501 reports57,58. Additionally, both SLC43A2 and SLC7A5 are associated with the
502 cytokine-cytokine receptor interaction pathway, involving the TGF-β and TNF
503 families58, which are critical components of the inflammatory process and play a
504 pivotal role in HIRI. These findings indicate that the three core genes identified play
505 multifaceted roles in the occurrence and progression of HIRI. Understanding the
506 mechanisms of these genes is essential for advancing therapeutic strategies for HIRI.
507 In the early stages of hepatic ischemia-reperfusion injury (HIRI), dysregulation
508 of liver mitochondrial function and microcirculation leads to hepatocyte injury,
509 triggering a cascade of immune responses that worsen liver damage by recruiting and
510 activating macrophages, neutrophils, T cells, and monocytes59,60. Our ssGSEA of
511 immune infiltration in HIRI patients revealed a significant increase in the proportion
512 of macrophages, neutrophils, aDCs, Treg, and Th1 cells compared to controls. These
513 results suggest that the three hub genes—SLC7A5, SLC1A5, and SLC43A2—are
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
27 / 46
514 associated with various immune cells and functions, particularly neutrophils, aDCs,
515 and Tregs, aligning with prior research61. KCs, play a key role in HIRI. In the early
516 ischemic stage, KCs are activated first, secreting cytokines such as TNF-α and IL-1β,
517 which recruit additional immune cells, thereby exacerbating liver injury62. Moreover,
518 recent studies show that DAMPs released from necrotic hepatocytes activate
519 inflammasome components such as NLRP3 and AIM2 in KCs via pattern recognition
520 receptors, further intensifying HIRI63. Our findings also confirmed a significant
521 increase in macrophage infiltration in the HIRI group compared to controls Both
522 SLC43A2 and SLC7A5 were strongly associated with aDCs. DCs are highly potent
523 professional antigen-presenting cells, and research by Castellaneta et al. demonstrated
524 that severe depletion of plasmacytoid DCs (pDCs) in mice downregulated IFN-α
525 expression in liver tissue, attenuating HIRI64. This highlights the damage-inducing
526 and pro-inflammatory role of DCs.. However, Nakamoto et al. found that loss of the
527 prostaglandin E receptor EP3 (PTGER3) in conventional DCs (cDCs) increased
528 macrophage activity and delayed liver repair after warm ischemia, indicating that the
529 roles of DCs in HIRI may vary depending on their subtypes65. This variability
530 suggests a direction for future research.. Neutrophils also play a significant role in
531 liver injury. Kaltenmeier et al. showed that DAMPs activate neutrophils, leading to
532 the release of enzymes that activate the complement system and trigger an
533 inflammatory response during HIRI66. Neutrophils also play a significant role in liver
534 injury. Kaltenmeier et al. showed that DAMPs activate neutrophils, leading to the
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
28 / 46
535 release of enzymes that activate the complement system and trigger an inflammatory
536 response during HIRII67. Targeting neutrophils may represent a promising therapeutic
537 approach to alleviating HIRI. CD4+ T cells are critical mediators of inflammation in
538 the HIRI cascade, capable of activating macrophages and amplifying the immune
539 response68,69. CD4+ T cells can differentiate into various subtypes, including Th1,
540 Th2, Th17, and Treg cells70. Th1 cells primarily eliminate intracellular pathogens by
541 secreting IFN-γ and recruiting immune cells, while Treg cells are involved in
542 immunosuppression through the secretion of IL-10 and TGF-β. Our study showed that
543 SLC7A5 is significantly associated with Treg cells. In summary, these findings
544 underscore the crucial role of the immune system in the development and progression
545 of HIRI.
546 We also constructed a prediction model for HIRI and explored the characteristic
547 genes and signaling pathways involved. However, our study has several limitations.
548 First, the data were obtained from the GEO public database, which restricts sample
549 size and introduces potential selection bias due to the absence of raw sequencing data.
550 Second, although our mice experiments partially validated the bioinformatics
551 analysis, we were unable to collect sufficient clinical samples due to time constraints.
552 Gathering adequate clinical samples is vital to fully uncover the role of BCAA
553 metabolism in HIRI. Additionally, the discrepancy between RNA sequencing and
554 qRT-PCR results highlights the complexity of BCAA metabolic regulation in HIRI.
555 Therefore, future research should aim to comprehensively investigate these
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
29 / 46
556 mechanisms to better understand the regulatory pathways involved.
557 Conclusion
558 In this study, we identified three characteristic genes through bioinformatics analysis
559 and machine learning algorithms that may serve as potential diagnostic biomarkers for
560 HIRI. Additionally, we demonstrated the critical role of the immune system in the
561 occurrence and progression of HIRI, emphasizing significant differences in immune
562 responses between HIRI and control liver samples. Our findings provide new insights
563 into the role of branched-chain amino acid metabolism in HIRI and establish a
564 foundation for future research into the pathogenesis of HIRI and potential therapeutic
565 targets.
566 Acknowledgments
567 The authors sincerely express gratitude for the insightful comments provided by the
568 reviewers on the present article. Additionally, we would like to thank the researchers
569 who generously contributed the datasets (GSE23649 and GSE15480).
570 Author contributions
571 The formal analysis and original draft of the manuscript were performed by JZ, MW.
572 Project administration was performed by SJL, SL, ML, QL. Software analysis was
573 performed by JZ, MW, GO. Data curation was conducted by JZ, SJL, GP, HX. The
574 experiment was performed by MW, ML, QL. The writing, reviewing, and editing of
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
30 / 46
575 the article were contributed by JZ, GO, GP, HX. All authors contributed to the article
576 and approved the final version for submission.
577 Ethics statement
578 All animal experiments were approved by the Animal Care and Use Committee of
579 Liuzhou People's Hospital Affiliated to Guangxi Medical University
580 (IACUC20240028). All experiments were performed in accordance with regulations
581 and complied with the ARRIVE guidelines.
582 Funding
583 This work was supported in part by Guangxi Science and Technology Project (Guike
584 AB23026016); Liuzhou Science and Technology Project (2022CAC0209) and Youth
585 Science Foundation of Guangxi Medical University (GXMUYSF202451)
586 Data availability statement
587 The datasets utilized in this study were obtained from the GEO database
588 (https://www.ncbi.nlm.nih.gov/geo/) with the following data accessions: GSE23649
589 and GSE15480. The original codes used for analyses presented in the study are
590 publicly available. This data can be found here:
591 https://github.com/jiahui980702/HIRI-BCAA-.
592 References：
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
31 / 46
593 1. Hammond JS, Guha IN, Beckingham IJ, Lobo DN. Prediction, prevention and
594 management of postresection liver failure. Br J Surg. 2011;98(9):1188-1200.
595 doi:10.1002/bjs.7630
596 2. Selzner N. Protective strategies against ischemic injury of the liver.
597 Gastroenterology. 2003;125(3):917-936. doi:10.1016/S0016-5085(03)01048-5
598 3. Ornellas FM, Ornellas DS, Martini SV, et al. Bone Marrow–Derived
599 Mononuclear Cell Therapy Accelerates Renal Ischemia-Reperfusion Injury Recovery
600 by Modulating Inflammatory, Antioxidant and Apoptotic Related Molecules. Cell
601 Physiol Biochem. 2017;41(5):1736-1752. doi:10.1159/000471866
602 4. Dewald B, Baggiolini M. Activation of NADPH oxidase in human neutrophils.
603 Synergism between fMLP and the neutrophil products PAF and LTB4. Biochemical
604 and Biophysical Research Communications. 1985;128(1):297-304.
605 doi:10.1016/0006-291X(85)91678-X
606 5. Granger DN, Korthuis RJ. Physiologic Mechanisms of Postischemic Tissue
607 Injury. Annu Rev Physiol. 1995;57(1):311-332.
608 doi:10.1146/annurev.ph.57.030195.001523
609 6. Fraser PA. The role of free radical generation in increasing cerebrovascular
610 permeability. Free Radical Biology and Medicine. 2011;51(5):967-977.
611 doi:10.1016/j.freeradbiomed.2011.06.003
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
32 / 46
612 7. Granger DN, Kvietys PR. Reperfusion injury and reactive oxygen species: The
613 evolution of a concept. Redox Biology. 2015;6:524-551.
614 doi:10.1016/j.redox.2015.08.020
615 8. Wu B, Lin R, Dai R, Chen C, Wu H, Hong M. Valsartan attenuates oxidative
616 stress and NF-κB activation and reduces myocardial apoptosis after ischemia and
617 reperfusion. European Journal of Pharmacology. 2013;705(1-3):140-147.
618 doi:10.1016/j.ejphar.2013.02.036
619 9. Holeček M. Branched-chain amino acids in health and disease: metabolism,
620 alterations in blood plasma, and as supplements. Nutr Metab (Lond). 2018;15(1):33.
621 doi:10.1186/s12986-018-0271-1
622 10. Zhang S, Zeng X, Ren M, Mao X, Qiao S. Novel metabolic and physiological
623 functions of branched chain amino acids: a review. J Anim Sci Biotechnol. 2017;8:10.
624 doi:10.1186/s40104-016-0139-z
625 11. Ampong I, Watkins A, Gutierrez-Merino J, Ikwuobe J, Griffiths HR. Dietary
626 protein insufficiency: an important consideration in fatty liver disease? Br J Nutr.
627 2020;123(6):601-609. doi:10.1017/S0007114519003064
628 12. Zhao H, Zhang F, Sun D, et al. Branched-Chain Amino Acids Exacerbate
629 Obesity-Related Hepatic Glucose and Lipid Metabolic Disorders via Attenuating
630 Akt2 Signaling. Diabetes. 2020;69(6):1164-1177. doi:10.2337/db19-0920
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
33 / 46
631 13. White PJ, McGarrah RW, Grimsrud PA, et al. The BCKDH Kinase and
632 Phosphatase Integrate BCAA and Lipid Metabolism via Regulation of ATP-Citrate
633 Lyase. Cell Metab. 2018;27(6):1281-1293.e7. doi:10.1016/j.cmet.2018.04.015
634 14. Kanehisa M. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids
635 Research. 2000;28(1):27-30. doi:10.1093/nar/28.1.27
636 15. Van Egmond MB, Spini G, Van Der Galien O, et al. Privacy-preserving dataset
637 combination and Lasso regression for healthcare predictions. BMC Med Inform Decis
638 Mak. 2021;21(1):266. doi:10.1186/s12911-021-01582-y
639 16. Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear
640 Models via Coordinate Descent. J Stat Softw. 2010;33(1):1-22.
641 17. Huang ML, Hung YH, Lee WM, Li RK, Jiang BR. SVM-RFE Based Feature
642 Selection and Taguchi Parameters Optimization for Multiclass SVM Classifier. The
643 Scientific World Journal. 2014;2014:1-10. doi:10.1155/2014/795624
644 18. Rigatti SJ. Random Forest. Journal of Insurance Medicine. 2017;47(1):31-39.
645 doi:10.17849/insm-47-01-31-39.1
646 19. Liu C, Zhou Y, Zhou Y, Tang X, Tang L, Wang J. Identification of crucial genes
647 for predicting the risk of atherosclerosis with system lupus erythematosus based on
648 comprehensive bioinformatics analysis and machine learning. Computers in Biology
649 and Medicine. 2023;152:106388. doi:10.1016/j.compbiomed.2022.106388
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
34 / 46
650 20. My W, Gt Y, Wt L, et al. Current Mechanistic Concepts in Ischemia and
651 Reperfusion Injury. Cellular physiology and biochemistry : international journal of
652 experimental cellular physiology, biochemistry, and pharmacology. 2018;46(4).
653 doi:10.1159/000489241
654 21. Zhai Y, Petrowsky H, Hong JC, Busuttil RW, Kupiec-Weglinski JW. Ischaemia–
655 reperfusion injury in liver transplantation—from bench to bedside. Nat Rev
656 Gastroenterol Hepatol. 2013;10(2):79-89. doi:10.1038/nrgastro.2012.225
657 22. Teodoro JS, Da Silva RT, Machado IF, et al. Shaping of Hepatic
658 Ischemia/Reperfusion Events: The Crucial Role of Mitochondria. Cells.
659 2022;11(4):688. doi:10.3390/cells11040688
660 23. Sun H, Olson KC, Gao C, et al. Catabolic Defect of Branched-Chain Amino
661 Acids Promotes Heart Failure. Circulation. 2016;133(21):2038-2049.
662 doi:10.1161/CIRCULATIONAHA.115.020226
663 24. Hsieh PN, Zhang L, Jain MK. Coordination of cardiac rhythmic output and
664 circadian metabolic regulation in the heart. Cell Mol Life Sci. 2018;75(3):403-416.
665 doi:10.1007/s00018-017-2606-x
666 25. Lian K, Guo X, Wang Q, et al. PP2Cm overexpression alleviates MI/R injury
667 mediated by a BCAA catabolism defect and oxidative stress in diabetic mice.
668 European Journal of Pharmacology. 2020;866:172796.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
35 / 46
669 doi:10.1016/j.ejphar.2019.172796
670 26. Adeva MM, Calviño J, Souto G, Donapetry C. Insulin resistance and the
671 metabolism of branched-chain amino acids in humans. Amino Acids.
672 2012;43(1):171-181. doi:10.1007/s00726-011-1088-7
673 27. Lu G, Sun H, She P, et al. Protein phosphatase 2Cm is a critical regulator of
674 branched-chain amino acid catabolism in mice and cultured cells. J Clin Invest.
675 2009;119(6):1678-1687. doi:10.1172/JCI38151
676 28. Yang Y, Wang S, Sheng C, et al. Branched-chain amino acid catabolic defect
677 promotes α-cell proliferation via activating mTOR signaling. Molecular and Cellular
678 Endocrinology. 2024;582:112143. doi:10.1016/j.mce.2023.112143
679 29. Zhai M, Yang Z, Zhang C, et al. APN-mediated phosphorylation of BCKDK
680 promotes hepatocellular carcinoma metastasis and proliferation via the ERK signaling
681 pathway. Cell Death Dis. 2020;11(5):396. doi:10.1038/s41419-020-2610-1
682 30. Solloway MJ, Madjidi A, Gu C, et al. Glucagon Couples Hepatic Amino Acid
683 Catabolism to mTOR-Dependent Regulation of α-Cell Mass. Cell Rep.
684 2015;12(3):495-510. doi:10.1016/j.celrep.2015.06.034
685 31. Alevriadou BR, Shanmughapriya S, Patel A, Stathopulos PB, Madesh M.
686 Mitochondrial Ca 2+ transport in the endothelium: regulation by ions, redox signalling
687 and mechanical forces. J R Soc Interface. 2017;14(137):20170672.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
36 / 46
688 doi:10.1098/rsif.2017.0672
689 32. Kalogeris T, Bao Y, Korthuis RJ. Mitochondrial reactive oxygen species: a
690 double edged sword in ischemia/reperfusion vs preconditioning. Redox Biol.
691 2014;2:702-714. doi:10.1016/j.redox.2014.05.006
692 33. Liu H, Jing X, Dong A, Bai B, Wang H. Overexpression of TIMP3 Protects
693 Against Cardiac Ischemia/Reperfusion Injury by Inhibiting Myocardial Apoptosis
694 Through ROS/Mapks Pathway. Cell Physiol Biochem. 2017;44(3):1011-1023.
695 doi:10.1159/000485401
696 34. Han N, He J, Shi L, Zhang M, Zheng J, Fan Y. Identification of biomarkers in
697 nonalcoholic fatty liver disease: A machine learning method and experimental study.
698 Front Genet. 2022;13:1020899. doi:10.3389/fgene.2022.1020899
699 35. Zheng PF, Chen LZ, Liu P, Pan HW, Fan WJ, Liu ZY. Identification of
700 immune-related key genes in the peripheral blood of ischaemic stroke patients using a
701 weighted gene coexpression network analysis and machine learning. J Transl Med.
702 2022;20(1):361. doi:10.1186/s12967-022-03562-w
703 36. Zhang Y, Xia R, Lv M, et al. Machine-Learning Algorithm-Based Prediction of
704 Diagnostic Gene Biomarkers Related to Immune Infiltration in Patients With Chronic
705 Obstructive Pulmonary Disease. Front Immunol. 2022;13:740513.
706 doi:10.3389/fimmu.2022.740513
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
37 / 46
707 37. Rohilla S, Awasthi A, Kaur S, Puria R. Evolutionary conservation of long
708 non-coding RNAs in non-alcoholic fatty liver disease. Life Sci. 2021;264:118560.
709 doi:10.1016/j.lfs.2020.118560
710 38. Ma J, Li Y, Yao L, Li X. Analysis of MicroRNA Expression Profiling Involved
711 in MC-LR-Induced Cytotoxicity by High-Throughput Sequencing. Toxins (Basel).
712 2017;9(1):23. doi:10.3390/toxins9010023
713 39. Lu C, Rong D, Hui B, et al. CircETFA upregulates CCL5 by sponging miR-612
714 and recruiting EIF4A3 to promote hepatocellular carcinoma. Cell Death Discov.
715 2021;7(1):321. doi:10.1038/s41420-021-00710-x
716 40. Bhutia YD, Babu E, Ramachandran S, Ganapathy V. Amino Acid Transporters in
717 Cancer and Their Relevance to “Glutamine Addiction”: Novel Targets for the Design
718 of a New Class of Anticancer Drugs. Cancer Research. 2015;75(9):1782-1788.
719 doi:10.1158/0008-5472.CAN-14-3745
720 41. Kaira K, Sunose Y, Ohshima Y, et al. Clinical significance of L-type amino acid
721 transporter 1 expression as a prognostic marker and potential of new targeting therapy
722 in biliary tract cancer. BMC Cancer. 2013;13:482. doi:10.1186/1471-2407-13-482
723 42. Kaira K, Sunose Y, Arakawa K, et al. Prognostic significance of L-type
724 amino-acid transporter 1 expression in surgically resected pancreatic cancer. Br J
725 Cancer. 2012;107(4):632-638. doi:10.1038/bjc.2012.310
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
38 / 46
726 43. Furuya M, Horiguchi J, Nakajima H, Kanai Y, Oyama T. Correlation of L-type
727 amino acid transporter 1 and CD98 expression with triple negative breast cancer
728 prognosis. Cancer Sci. 2012;103(2):382-389. doi:10.1111/j.1349-7006.2011.02151.x
729 44. Chen R, Zou Y, Mao D, et al. The general amino acid control pathway regulates
730 mTOR and autophagy during serum/glutamine starvation. J Cell Biol.
731 2014;206(2):173-182. doi:10.1083/jcb.201403009
732 45. Yu B, Zhang Y, Wang T, et al. MAPK Signaling Pathways in Hepatic
733 Ischemia/Reperfusion Injury. J Inflamm Res. 2023;16:1405-1418.
734 doi:10.2147/JIR.S396604
735 46. Deng J, Feng J, Liu T, et al. Beraprost sodium preconditioning prevents
736 inflammation, apoptosis, and autophagy during hepatic ischemia-reperfusion injury in
737 mice via the P38 and JNK pathways. Drug Des Devel Ther. 2018;12:4067-4082.
738 doi:10.2147/DDDT.S182292
739 47. Hou L, Zhang Z, Yang L, et al. NLRP3 inflammasome priming and activation in
740 cholestatic liver injury via the sphingosine 1-phosphate/S1P receptor
741 2/Gα(12/13)/MAPK signaling pathway. J Mol Med (Berl). 2021;99(2):273-288.
742 doi:10.1007/s00109-020-02032-4
743 48. Sui X, Zhang R, Liu S, et al. RSL3 Drives Ferroptosis Through GPX4
744 Inactivation and ROS Production in Colorectal Cancer. Front Pharmacol.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
39 / 46
745 2018;9:1371. doi:10.3389/fphar.2018.01371
746 49. Fuchs BC, Finger RE, Onan MC, Bode BP. ASCT2 silencing regulates
747 mammalian target-of-rapamycin growth and survival signaling in human hepatoma
748 cells. Am J Physiol Cell Physiol. 2007;293(1):C55-63.
749 doi:10.1152/ajpcell.00330.2006
750 50. van Geldermalsen M, Wang Q, Nagarajah R, et al. ASCT2/SLC1A5 controls
751 glutamine uptake and tumour growth in triple-negative basal-like breast cancer.
752 Oncogene. 2016;35(24):3201-3208. doi:10.1038/onc.2015.381
753 51. Liu X, Qin H, Li Z, et al. Inspiratory hyperoxia suppresses lung cancer metastasis
754 through a MYC/SLC1A5-dependent metabolic pathway. Eur Respir J.
755 2022;60(6):2200062. doi:10.1183/13993003.00062-2022
756 52. Luo M, Wu L, Zhang K, et al. miR-137 regulates ferroptosis by targeting
757 glutamine transporter SLC1A5 in melanoma. Cell Death Differ.
758 2018;25(8):1457-1472. doi:10.1038/s41418-017-0053-8
759 53. Venkateswaran N, Lafita-Navarro MC, Hao YH, et al. MYC promotes tryptophan
760 uptake and metabolism by the kynurenine pathway in colon cancer. Genes Dev.
761 2019;33(17-18):1236-1251. doi:10.1101/gad.327056.119
762 54. Bodoy S, Martín L, Zorzano A, Palacín M, Estévez R, Bertran J. Identification of
763 LAT4, a novel amino acid transporter with system L activity. J Biol Chem.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
40 / 46
764 2005;280(12):12002-12011. doi:10.1074/jbc.M408638200
765 55. Bian Y, Li W, Kremer DM, et al. Cancer SLC43A2 alters T cell methionine
766 metabolism and histone methylation. Nature. 2020;585(7824):277-282.
767 doi:10.1038/s41586-020-2682-1
768 56. Peng H, Yan Y, He M, et al. SLC43A2 and NFκB signaling pathway regulate
769 methionine/cystine restriction-induced ferroptosis in esophageal squamous cell
770 carcinoma via a feedback loop. Cell Death Dis. 2023;14(6):347.
771 doi:10.1038/s41419-023-05860-7
772 57. Inoue Y, Shirasuna K, Kimura H, et al. NLRP3 regulates neutrophil functions and
773 contributes to hepatic ischemia-reperfusion injury independently of inflammasomes. J
774 Immunol. 2014;192(9):4342-4351. doi:10.4049/jimmunol.1302039
775 58. Wu XY, Zhao MJ, Liao W, et al. Oridonin attenuates liver ischemia-reperfusion
776 injury by suppressing PKM2/NLRP3-mediated macrophage pyroptosis. Cell
777 Immunol. 2024;401-402:104838. doi:10.1016/j.cellimm.2024.104838
778 59. Heo MJ, Suh JH, Poulsen KL, Ju C, Kim KH. Updates on the Immune Cell Basis
779 of Hepatic Ischemia-Reperfusion Injury. Mol Cells. 2023;46(9):527-534.
780 doi:10.14348/molcells.2023.0099
781 60. Liu J, Man K. Mechanistic Insight and Clinical Implications of
782 Ischemia/Reperfusion Injury Post Liver Transplantation. Cell Mol Gastroenterol
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
41 / 46
783 Hepatol. 2023;15(6):1463-1474. doi:10.1016/j.jcmgh.2023.03.003
784 61. Nachef M, Ali AK, Almutairi SM, Lee SH. Targeting SLC1A5 and
785 SLC3A2/SLC7A5 as a Potential Strategy to Strengthen Anti-Tumor Immunity in the
786 Tumor Microenvironment. Front Immunol. 2021;12:624324.
787 doi:10.3389/fimmu.2021.624324
788 62. Yue S, Zhou H, Wang X, Busuttil RW, Kupiec-Weglinski JW, Zhai Y. Prolonged
789 Ischemia Triggers Necrotic Depletion of Tissue-Resident Macrophages To Facilitate
790 Inflammatory Immune Activation in Liver Ischemia Reperfusion Injury. J Immunol.
791 2017;198(9):3588-3595. doi:10.4049/jimmunol.1601428
792 63. Shan Z, Ju C. Hepatic Macrophages in Liver Injury. Front Immunol.
793 2020;11:322. doi:10.3389/fimmu.2020.00322
794 64. Castellaneta A, Yoshida O, Kimura S, et al. Plasmacytoid dendritic cell-derived
795 IFN-α promotes murine liver ischemia/reperfusion injury by induction of hepatocyte
796 IRF-1. Hepatology. 2014;60(1):267-277. doi:10.1002/hep.27037
797 65. Nakamoto S, Ito Y, Nishizawa N, et al. EP3 signaling in dendritic cells promotes
798 liver repair by inducing IL‐13‐mediated macrophage differentiation in mice. The
799 FASEB Journal. 2020;34(4):5610-5627. doi:10.1096/fj.201901955R
800 66. Kaltenmeier C, Yazdani HO, Handu S, Popp B, Geller D, Tohme S. The Role of
801 Neutrophils as a Driver in Hepatic Ischemia-Reperfusion Injury and Cancer Growth.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
42 / 46
802 Front Immunol. 2022;13:887565. doi:10.3389/fimmu.2022.887565
803 67. Hirao H, Kojima H, Dery KJ, et al. Neutrophil CEACAM1 determines
804 susceptibility to NETosis by regulating the S1PR2/S1PR3 axis in liver
805 transplantation. J Clin Invest. 2023;133(3):e162940. doi:10.1172/JCI162940
806 68. Kageyama S, Kadono K, Hirao H, et al. Ischemia-reperfusion Injury in
807 Allogeneic Liver Transplantation: A Role of CD4 T Cells in Early Allograft Injury.
808 Transplantation. 2021;105(9):1989-1997. doi:10.1097/TP.0000000000003488
809 69. Wang J, Xia S, Ren H, Shi X. The role and function of CD4+ T cells in hepatic
810 ischemia-reperfusion injury. Expert Rev Gastroenterol Hepatol. 2022;16(1):5-11.
811 doi:10.1080/17474124.2022.2020642
812 70. Nakayamada S, Takahashi H, Kanno Y, O’Shea JJ. Helper T cell diversity and
813 plasticity. Curr Opin Immunol. 2012;24(3):297-302. doi:10.1016/j.coi.2012.01.014
814
815 Graphical Abstract:
816 Figure 1
817 Flowchart of the present study.
818 Figure 2
819 Identification of BCAA-related differentially expressed genes in HIRI. (A) DEGs are
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
43 / 46
820 shown by volcano plots, with blue denoting the downregulated genes, red denoting
821 the upregulated genes. (B) The gene expression patterns of DEGs were shown in the
822 heatmap. (C) Venn diagram showed the intersection of genes between DEGs and
823 BCAA-related genes. (D) The locations of the 13 DE-BRGs on 23 chromosomes. (E)
824 The expression patterns of 13 DE-BRGs were shown in the heatmap. (F) Boxplot
825 showed the differential expression of DE-BRGs between HIRI and control samples.
826 (G) Correlation analysis of 13 DE-BRGs. Red and green colors represent positive and
827 negative correlations respectively. (H)Correlation analysis of 13 DE-BRGs in the
828 circle graph. Red and green colors represent positive and negative correlations
829 respectively. (I)Gene relationship network diagram of 13 DE-BRGs. p values were
830 showed as: *, p < 0.05; **, p < 0.01; ***, p < 0.001. DEGs, differential expression
831 genes; DE-BRGs, differential expression of branched-chain amino acid
832 metabolism-related genes.
833 Figure3
834 Functional analyses for the DE-BRGs. (A) Bubble diagrams of the GO enrichment
835 analysis of 13 DE-BRGs. (B) Bubble diagrams of the KEGG enrichment analysis of
836 13 DE-BRGs. DE-BRGs, differential expression of branched-chain amino acid
837 metabolism-related genes. GO, Gene Ontology; BP, Biological process; CC, Cellular
838 component; MF, Molecular function; KEGG, Kyoto Encyclopedia of Genes and
839 Genomes.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
44 / 46
840 Figure 4
841 Candidate biomarker identification for HIRI via machine learning algorithms. (A)
842 LASSO regression coefficient analysis. The solid vertical lines represent the partial
843 likelihood deviance SE. The dotted vertical line is drawn at the optimal lambda. (B)
844 LASSO regression cross validation graph. Each curve corresponds to one gene. (C)
845 Relationship between the number of random forest trees and error rates. The red line
846 represents the error of the HIRI group, the green line represents the error of the
847 Control group, and the black line represents the total sample error. (D) The rank of
848 genes in accordance with their relative importance. (E) The error of the feature
849 selection for the SVM-RFE algorithm, with the lowest cross-validation error is found
850 in 4 gens and the values is 39.1%. (F) The accuracy of the feature selection for the
851 SVM-RFE algorithm. The peak of the curve is achieved at 4 genes with an accuracy
852 of 60.9%. (G) The Venn diagram shows the overlap of marker genes between
853 LASSO, random forest, and SVM-RFE algorithms.
854 Figure 5
855 Validation of biomarker gene expression (A) The ROC results for the 3 marker genes.
856 The AUC value of SLC7A5, SLC1A5 and SLC43A2 was 0.791, 0.729, 0.761,
857 espectively. (B) The ROC curves were evaluated comprehensively by GSE23649. The
858 AUC value was 0.834(95% CI: 0.733−0.922). (C) Nomogram graph of the 3 marker
859 genes. (D) Calibration curve displaying the diagnostic ability of the nomogram model.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
45 / 46
860 (E) DCA illustrating the predictive efficiency of Nomogram models. (F) The clinical
861 impact curve showed a higher diagnostic ability of the nomogram model. (G) The
862 ROC results of 3 marker genes in validation set. The AUC value of SLC7A5,
863 SLC1A5 and SLC43A2 was1.000, 0.944, 0.861, respectively. (H) The ROC curves of
864 the three hub genes were evaluated comprehensively by the validation set. The AUC
865 value was 1.000(95% CI: 1.000−1.000). AUC, area under curve; ROC, receiver
866 operating characteristic; DCA, Decision curve analysis.
867 Figure 6
868 GSEA and GSVA analysis of three marker genes. The KEGG and GO pathway
869 enrichment analysis of (A, D) SLC43A2, (B, E) SLC7A5 and (C, F) SLC1A5 were
870 carried out by GSEA enrichment method, and the two items with the highest and
871 lowest enrichment scores are visualized according to the arrangement of enrichment
872 scores. The KEGG pathway enrichment analysis of (G) SLC43A2, (H) SLC7A5 and
873 (I) SLC1A5 was carried out by GSVA enrichment method, and the top 50 are
874 visualized according to the enrichment score.
875 Figure 7
876 (A) Heatmap of immune cell infiltration between HIRI and control samples, with red
877 and green representing high and low expression, respectively. (B) The correlation
878 between 29 immune cells and functions and three marker genes. Red and green colors
879 represent positive and negative correlations, respectively. (C) Boxplots indicated the
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
46 / 46
880 differences in immune cells and function between HIRI and control samples. (D)
881 Correlation analysis among 16 immune cells. p values were showed as: *, p < 0.05;
882 **, p < 0.01; ***, p < 0.001.
883 Figure 8
884 CeRNA networks based on marker genes. (A) The ceRNA network based on marker
885 genes. With Pink dots for mRNA, green dots for miRNA, and blue dots for lncRNA.
886 Figure 9
887 HIRI mice model construction and Altered expression of DE-BRGs in HIRI. (A, B)
888 serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) value in
889 both wildtype and HIRI mice. (C) The HE staining of wild type and HIRI mice. The
890 result of HE staining showed significant liver injury in HIRI group. (D-F) mRNA
891 expression of SLC1A5, SLC7A5 and SLC43A2 by RT-PCR. Liver tissues and blood
892 of the wildtype and HIRI mice were collected. p values were showed as: *, p < 0.05;
893 **, p < 0.01; ***, p < 0.001. DE-BRGs, differential expression of branched-chain
894 amino acid metabolism-related genes. Wildtype control mice (n=8), HIRI mice(n=8).
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted December 20, 2024. ; https://doi.org/10.1101/2024.12.17.629007doi: bioRxiv preprint
