Abstract Background Cell metabolism and long noncoding RNA (lncRNA) played crucial roles in cancer development. However, their association in colon adenocarcinoma (COAD) remains unclear. Methods The COAD gene expression data and corresponding clinical data were retrieved from The Cancer Genome Atlas (TCGA) database. Differential expression of metabolic genes and lncRNA were identified by comparing tumor and normal colon tissues. Pearson correlation analysis was performed to identify metabolism‐associated lncRNA. COAD patients were divided into training cohort and validation cohort by randomization. Then, a univariate Cox regression analysis was introduced to evaluate the correlations between metabolism‐related lncRNAs and overall survival (OS) of the patients in the training cohort. The least absolute shrinkage and selection operator (LASSO) method was introduced to determine and establish a prognostic prediction model. Subsequently, survival analysis, receiver operating characteristic (ROC) curve analysis, and Cox regression analysis were generated to estimate the prognostic role of the LncRNA risk score in training, validation, and entire cohorts. Results We identified 152 differentially expressed metabolism‐associated lncRNAs (MRLncRNAs). A prognostic prediction model involving four metabolism‐related lncRNAs were established using LASSO. In each cohort, COAD patients in the high‐risk group had worse OS compared to those in the low‐risk group. The ROC analyses demonstrated that the lncRNA signature performed well in predicting OS. Uni‐ and multivariate analysis indicated that the lncRNA signature as an independent prognostic factor. Furthermore, a correlation analysis demonstrated that LINC01138 was the most closely lncRNA related to metabolic genes. In vitro assays demonstrated that LINC01138 affects tumor progression in COAD. Conclusions In summary, we established a metabolism‐associated lncRNAs model to predict the prognosis in COAD patients. Keywords: colon adenocarcinoma, lncRNA, metabolism, prognosis 1. INTRODUCTION Colon adenocarcinoma (COAD) is the most prevalent histopathological type of colorectal cancer with more than a million newly diagnosed cases in 2018.[36] ^1 With the advancement of early screening and treatment strategies, the survival of COAD patients has improved.[37] ^2 However, for some patients with advanced tumors, the prognosis is still poor.[38] ^3 Additionally, patients with similar clinical characteristics have significant differences in survival outcomes and quality of life.[39] ^4 Some risk biomarkers, including CA199 and CEA have been shown to be associated with COAD, but it is not suitable for all patients.[40] ^5 Presently, TNM staging, a tool based on clinical features is widely used to evaluate the prognosis of patients in clinical, but it is limited to reveal the biological heterogeneity of tumors.[41] ^5 Therefore, it is meaningful to explore reliable and convenient indicators of prediction in COAD. Cancer cells exhibit a distinct metabolism profile compared with normal cells, which is an important hallmark of cancer. In the last few decades, it had become evident that alteration of metabolism in cancer cells contributes to oncogenesis and cancer progression.[42] ^6 It has been established that cancer cells prefer to perform glycolysis even under sufficient oxygen tension, a phenomenon known as Warburg effect or aerobic glycolysis, which supports the growth and proliferation of cancer cells.[43] ^7 Altered aerobic glycolysis has been validated in many tumors, and targeting intervention of Warburg effect can inhibit tumor progression. Hexokinases 2 (HK2) is the first and rate‐limiting enzyme of glycolysis, which is responsible for the first step of glycolysis. In human glioblastoma multiforme (GBM), Amparo Wolf et al. showed that HK2 is aberrantly expressed in GBM and correlated with worse survival of patients. HK2 depletion resulted in restored oxidative glucose metabolism and increased sensitivity to apoptosis of GBM cells.[44] ^8 In addition, several glycolysis inhibitors have been introduced to target the Warburg effect. For example, acting as an inhibitor of hexokinase, administration of 3‐BrOP decreased ATP levels, induced cell death of neuroblastoma in vitro, and reduced neuroblastoma weight in vivo.[45] ^9 In COAD, a study reported that machilin A (MA) can impair activity of LDHA, an important enzyme involvement in aerobic glycolytic, which led to decrease lactate production and ATP levels in vitro, and suppressed tumor growth in vivo.[46] ^10 Besides, other than the aberrant metabolism, for example, amino acids metabolism has been validated to be associated with cancer progression.[47] ^11 Thus, clarifying the role of metabolism may improve the prediction of prognostic for COAD patients and finding new potential targets for COAD. LncRNA is a class of noncoding RNA with more than 200 nucleotides in length.[48] ^12 It has been proposed that lncRNAs widely involved in diverse biological processes, including cell proliferation, development, and apoptosis.[49] ^13 , [50]^14 , [51]^15 Moreover, the dysregulation expression of lncRNA has been demonstrated to be associated with many diseases,[52] ^16 including COAD.[53] ^17 For instance, LncRNA BLACAT1 by epigenetically silencing of p15 affects colorectal cancer cells proliferation, which resulted in a poor outcome.[54] ^18 lncRNA SNHG15 through interacting with and maintaining slug stability to promote colon cancer progression.[55] ^19 lncRNAs are a key factor in regulate cancer metabolism. A good example is lncRNA ANRIL, abnormally expressed in acute myeloid leukemia (AML). ANRIL by inhibiting the level of adiponectin receptor (AdipoR1), a key regulator of glucose metabolism, thus affected glucose uptake and cell maintenance in AML.[56] ^20 However, the association between lncRNA and metabolism in COAD is unclarified. Herein, we intended to explore the association between lncRNA and metabolism and establish a reliable metabolism‐associated lncRNAs signature for predicting the prognosis of COAD patients. Finally, we validated the role of a representative metabolism‐associated lncRNAs in COAD. 2. MATERIALS AND METHODS 2.1. Dataset processing RNA sequencing data of 398 tumor and 39 normal colon tissues, and corresponding clinical features were obtained from TCGA database ([57]https://www.cbioportal.org) on February 10, 2022. Due to the incomplete clinical characteristics of some patients in COAD, we reanalyzed the data and excluded COAD patients without complete clinical characteristics. Overall, a total of 364 COAD patients were included. 2.2. Identification of metabolism‐related lncRNAs Cell metabolism genes were extracted from ccmGDB ([58]http://bioinfo.mc.vanderbilt.edu/ccmGDB). LncRNA annotation was obtained from Gencode ([59]www.gencodegenes.org/). Differentially expressed metabolic genes and lncRNAs were performed by “limma” package (|log2FoldChange| >1 and false discovery rate [FDR] < 0.05 as statistical significance). Based on the criteria of |correlation coefficient| ≥ 0.3 and p < 0.001 by Pearson correlation analysis, we identified metabolism‐related lncRNAs. 2.3. Establish the metabolism‐related lncRNAs prognostic signatures All eligible patients were randomly divided into training cohort with 183 patients and validation cohort with 181 patients by the R package “caret”[60] ^21 (Table [61]S1). Table [62]1 displays clinical features of COAD patients in all cohorts. First, R package “survival” was introduced to characterize the significant MRlncRNAs in training cohort. Subsequently, the LASSO Cox regression analysis was introduced to refine genes by R package “glmnet.” Ultimately, four lncRNAs were selected to establish a metabolism‐associated lncRNAs model. The risk score of each sample were calculated for each subject as follows: Risk Score = βlncRNA1 × ExplncRNA1 + βlncRNA2 × ExplncRNA2 + … βlncRNAn×ExplncRNAn (β: the coefficient from LASSO Cox regression analysis, Exp: gene expression).[63] ^22 , [64]^23 The median value was used as the cutoff value, and the patients were classified as low risk and high risk accordingly. TABLE 1. Characteristics of COAD patients No. (%) Characteristics Entire cohort (n = 364) Training cohort (n = 183) Validation cohort (n = 181) p value Age, years 0.538 ≤60 110 (30.2) 58 (31.7) 52 (28.7) >60 254 (69.8) 125 (68.3) 129 (71.3) Gender 0.073 Female 168 (46.2) 93 (50.8) 75 (41.4) Male 196 (53.8) 90 (49.2) 106 (58.6) AJCC stage 0.523 I 65 (17.8) 34 (18.6) 31 (17.1) II 144 (39.6) 67 (36.6) 77 (42.5) III 99 (27.2) 51 (27.9) 48 (26.5) IV 51 (14.0) 27 (14.7) 24 (13.3) Unknown 5 (1.4) 4 (2.2) 1 (0.6) Stage T 0.829 T1 9 (2.5) 4 (2.2) 5 (2.8) T2 65 (17.8) 34 (18.6) 31 (17.1) T3 250 (68.7) 123 (67.2) 127 (70.2) T4 39 (10.7) 21 (11.5) 18 (9.9) Unknown 1 (0.3) 1 (0.5) 0 (0) Stage N 0.864 N0 218 (59.9) 109 (59.6) 109 (60.2) N1 84 (23.1) 42 (22.9) 42 (23.2) N2 62 (17.0) 32 (17.5) 30 (16.6) Stage M 0.782 M0 279 (76.7) 139 (76.0) 140 (77.3) M1 51 (14.0) 27 (14.7) 24 (13.3) Unknown 34 (9.3) 17 (9.3) 17 (9.4) Alive status 0.939 Alive 289 (79.4) 145 (79.2) 144 (79.6) Died 75 (20.6) 38 (20.8) 37 (20.4) [65]Open in a new tab 2.4. Function prediction of the lncRNA signature In general, LncRNA has no ability of coding protein, and it plays a role by regulating the expression of coding genes.[66] ^12 We indirectly clarified the function of lncRANs by investigating the metabolic genes related to lncRANs. The R package of clusterProfiler was employed to conduct a KEGG pathway and GO enrichment analysis. 2.5. Cell culture Human colon adenocarcinoma cell lines LoVo (CCL‐229), HCT116 (CCL‐247), HT‐29 (HTB‐38), and SW480 (CCL‐228) were obtained from ATCC (Rockville, MD). Cells were cultured in the McCoy's 5A, F‐12 K, and Leibovitz's L‐15 medium (containing 12% fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin) at 37°C in a 5% CO2 incubator. 2.6. Gene overexpression and knockdown Human genes LINC01138 (Gene ID: 388685) overexpression vector with pGPU6 was purchased from by GenePharma (Shanghai, China). Oligonucleotides specific for shRNAs against LINC01138 were designed and purchased from Genechem (Shanghai, China) with pGPU6 vector. The sequences were: sh‐LINC01138 #1: 5′‐GCACATTTGAGATACTGTTCAAGAGACAGTATCTCAAATGTGCTTTTTT‐3′ (forward) and 5′‐AAAAAAGCACATTTGAGATACTGTCTCTTGAACAGTATCTCAAATGTGC′ (reverse); sh‐LINC01138 #2: 5′‐GATGGAGTCTCACTGTTCAAGAGACAGTGAGACTCCATCTTTTTT‐3′(forward) and 5′‐AAAAAAGATGGAGTCTCACTGTCTCTTGAACAGTGAGACTCCATC′ (reverse). 2.7. Quantitative real‐time PCR TRIzol reagent (Invitrogen) was used to isolate total RNA according to the manufacturer's suggestion. Then, obtained RNA (~2.0 μg) was used to synthesize cDNA with a kit (RR047A, Takara, Japan). RT‐qPCR analyses were performed on an Applied Biosystems 7500 Sequence Detector with luminaris color hiGreen qPCR master mix (Thermo Fisher Scientific). The procedures were used for the qPCR as follows: predenaturation of 95°C for 10 min, 36 cycles of 95°C for 10 s, and 60°C for 60 s. The transcript levels were normalized to β‐actin and analyzed by 2^−ΔΔCt method. The primers were shown below: for LINC01138: 5′‐TACGAAAGCTGAAAGCGTGC‐3′ (sense) and 5′‐ATGAAACAGAACTGCAAACAGGC‐3′ (antisense). 2.8. Western blotting Protein extraction and was determination based on the manufacturer's instructions (C500005 and C503051, Sangon Biotech). Subsequently, the obtained proteins (~30 μg) were analyzed by SDS–PAGE and transferred onto PVDF membranes (ImmobilonP). Thereafter, the membrane was blocked for 1 h at normal temperature with nonfatty milk (3%), and the membranes were incubated at 4°C overnight with the following primary antibodies: rabbit anti‐human antibodies against DGKH (1:1000; 13,873‐1‐AP, Proteintech), PDE4D (1:3000; 67,062‐1‐Ig, Proteintech), PYCR1 (1:1000; 66,510‐1‐Ig, Proteintech), TKT (1:3000; 66,016‐1‐Ig, Proteintech), or β‐actin (1:5000; 66,009‐1‐Ig, Proteintech). Then, the membranes were washed three times at 10‐min intervals, with PBS containing 0.1% Tween‐20. Secondary antibodies were incubated for 1 h at normal temperature (1:5000, SA00001‐2, Proteintech). Finally, special bands were determined by a commercial kit (12,630, Cell Signaling Technology, Inc.). Image‐Pro Plus 6.0 imaging software (Media Cybernetics, Inc.) was used to quantify protein expression and followed statistics by GraphPad Prism 9 (GraphPad Software). 2.9. MTT assay A total of 1 × 10^3 cells (in 100 μl) was seeded into a 96‐well plate and allowed to grow for the indicated time period before analysis. Thereafter, 5 mg/mL MTT solution was added to individual wells and the 96‐well plate were incubated in a 5% CO[2] incubator at 37°C for 4 h. Then removed the medium, 100 μl dimethylsulfoxide (DMSO) was added to dissolve the formazan crystals and the absorbance at 595 nm wavelength was measured (Bio‐Tek, USA). 2.10. Soft agar assay A total of 5 × 10^3 tumor cells were mixed with 0.4% agarose (214,230, Noble) and plated onto 6‐well plates with 0.8% solidified agar in medium containing 12% fetal bovine serum. Three weeks later, colonies underwent 100% methanol fixing and 0.5% crystal violet staining. 2.11. Matrigel invasion assay The invasion ability of COAD cells was evaluated using a 24‐well Transwell plates precoated with Matrigel matrix (BD Science). Briefly, 500 μl medium (12% FBS) was added to lower chambers, and 1 × 10^5 cells with serum‐free media was added to the top chamber. After 24 h, fixed with 4% paraformaldehyde (PFA) and was stained with 0.4% crystal violet dye. Subsequently, images from five random microscopic fields were captured, and invasive cells were examined and counted. 2.12. Ethical statement This article does not contain any studies with human participants or animals performed by any of the authors. 2.13. Availability of data and materials The datasets used and/or analyzed in this study are available from the corresponding author on reasonable request. 2.14. Statistical analysis The differences of genes expression between COAD and normal colon tissues were investigated using the R with independent t‐test. R package “ggplot2” and “pheatmap” were used to plot Volcano plot and Heat maps, respectively. Survival analyzed were performed by log‐rank (Mantel–Cox) analysis. ROC curves and area under the curve (AUC) analyses were performed using “survival ROC” package. Student's test was used to compare data between two independent groups. For multiple groups comparison, one‐way analysis of variance with (ANOVA) with Tukey's correction was used. All statistics were done in GraphPad Prism 9 (GraphPad Software). p < 0.05 was considered statistically significant. 3. RESULTS 3.1. Identification of differentially expressed metabolism‐related lncRNAs Figure [67]1 shows the flowchart of this study. Different analyses in gene expression were performed on the TCGA COAD between 39 normal and 398 tumor tissues, and 506 differentially expressed metabolism‐related genes and 705 differentially expressed lncRNAs were identified. Subsequently, 152 metabolism‐related lncRNAs were generated by the R package “corrplot” (Figure [68]2A, B, and Table [69]S2). FIGURE 1. FIGURE 1 [70]Open in a new tab Research flowcharts. FIGURE 2. FIGURE 2 [71]Open in a new tab Differentially expressed metabolism‐related lncRNAs. Volcano plot (A) and heatmap (B) show 506 differentially expressed metabolism genes and 705 differentially expressed lncRNAs by comparing tumor and normal colon tissues. 3.2. Establishment and validation of MRlncRNAs prognostic signature Univariate Cox regression analysis was introduced to indentify MRlncRNAs associated with OS in training cohort. 12 lncRNAs (p < 0.05) were retained, all genes (PCAT6, TMEM147‐AS1, ZKSCAN2‐DT, LINC01138, MELTF‐AS1, LINC02474, MMP25‐AS1, MNX1‐AS1, ATP2A1‐AS1, GABPB1‐AS1, ELFN1‐AS1, and LINC00174) were related with increased risk (HRs >1) (Figure [72]3A). To minimize the overfitting caused by univariate Cox regression, LASSO regression analysis was introduced and a 4‐gene signature was established. (Figure [73]3B–D). Based on the median risk score, the training cohort was classified as high‐risk group and low‐risk group accordingly. Kaplan–Meier plots demonstrated patients in the high‐risk group experienced worse OS than those in the low‐risk group (Figure [74]4A). The distribution patterns of risk score was shown in Figure [75]4B. A heatmap exhibited the expression of 4 MRlncRNAs in two groups (Figure [76]4C). Time‐dependent ROC curve analyses revealed that the MRlncRNAs signature has a good performance in the prediction of OS, and our result demonstrated that the AUC was 0.702, 0.735, and 0.834, for 1‐year, 3‐year, and 5‐year OS, respectively (Figure [77]4D). Consistent with training cohort, the validation cohort and entire cohort showed similar results, which verified the lncRNAs signature's accuracy (Figure [78]4E‐L). FIGURE 3. FIGURE 3 [79]Open in a new tab Establishment of metabolism‐related lncRNAs prognostic signature. (A) Univariate Cox regression analysis of OS for 12 metabolism‐related lncRNAs. (B and C) LASSO regression analysis was used to determine candidate metabolism‐related lncRNAs. (D) Multivariate Cox regression analysis of four metabolism‐related lncRNAs. FIGURE 4. FIGURE 4 [80]Open in a new tab Validation of metabolism‐related lncRNAs prognostic signature. (A, E, and I) Represent the KM plot for OS of patients in different risk groups of three cohorts. (B, F, and J) Represent the risk scores distribution of gene expression pattern in three cohorts. (C, G, and K) The heatmaps show the gene expression in three cohorts. (D, H, and L) Time‐dependent ROC curves for the prognostic value of the metabolism‐related lncRNAs signature in three cohorts. Survival analysis was examined by log‐rank (Mantel–Cox) test in A, E, and I. 3.3. Independent prognostic value of the risk score model To assess the prediction ability of the risk score model in COAD patients, Cox regression analyses were introduced. As shown in Figure [81]5A, the univariate Cox regression analyses indicated that the risk score was correlated with the OS of COAD patients, which was in line with the result of multivariate Cox regression analyses (Figure [82]5B). Additionally, the analyses in validation‐ and the entire‐ cohort showed similar results (Figure [83]5C–F). FIGURE 5. FIGURE 5 [84]Open in a new tab Independent prognostic value of the risk score model. (A, C, and E) Univariate Cox regression analysis of different factors in three cohorts. (B, D, and F) Multivariate Cox regression analysis of different factors in three cohorts. 3.4. Prognostic role of the MRlncRNAs signature in clinical subgroups To assess the prediction effciency of risk score model in clinical subgroups, stratification survival analysis was performed. The subgroups were divided according to age status (age < =60 and age > 60), gender status (female and male), tumor size status (T1, T2, T3, and T4), lymph node metastatic status (N0, N1, and N2), and distant metastatic status (not metastatic and metastatic). According to the cut‐off value of median risk score, patients were stratified into the low‐risk group and high‐risk group for OS in different subgroups, except for the N2 subgroup. Compared with the low‐risk group, patients in the high‐risk group presented worse OS (Figure [85]6A–J). Moreover, we did not conduct subgroup analysis for T1, T2, and T4 due to failing to divide patients into two risk groups. FIGURE 6. FIGURE 6 [86]Open in a new tab Prognostic role of the prognostic signature in clinical subgroups. (A) Age < = 60 year. (B) Age > 60 year. (C) Male. (D) Female. (E) T3. (F) N0. (G) N1. (H) N2. (I) Not metastatic. (J) Metastatic. Survival analysis was examined by log‐rank (Mantel–Cox) test in Figure [87]6. 3.5. Functional analysis of the regulatory network As shown in Figure [88]7, metabolic genes related to lncRANs in the biological process (BP) category were enriched in carboxylic acid biosynthetic process, organic acid biosynthetic process, and small molecule catabolic process. In the cellular components (CC) category were mainly associated with Golgi lumen, lysosomal lumen, mitochondrial matrix, and vacuolar lumen. In the molecular functions (MF) category mainly participated in the regulation of coenzyme binding, transferase activity, transferring glycosyl groups, and carboxylic acid binding. For KEGG pathway, the main significant pathways included purine metabolism, drug metabolism, and chemical carcinogenesis, etc. FIGURE 7. FIGURE 7 [89]Open in a new tab Functional enrichment analysis. (A) Gene ontology (GO) enrichment analysis of the metabolic genes related to lncRANs. CC, cellular component; MF, molecular function. (B) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of metabolic genes related to lncRANs. 3.6. The expression of four MRlncRNAs in clinical subgroups Next, we analyzed the expression of four MRlncRNAs in different clinical subgroups. Briefly, the expression level of lncRNAs was not associated with the status of age, primary tumor size, and distant metastatic classification (Figure [90]8A, C, E). Only LINC02474 showed differences in expression between the two groups in gender classification (Figure [91]8B). Furthermore, the expression level of LINC01138, LINC02474, and PCAT6 showed statistical difference in expression for lymph node involvement status and alive status classification (Figure [92]8D, F). However, this phenomenon was not observed in ELFN1‐AS1 (Figure [93]8D, F). FIGURE 8. FIGURE 8 [94]Open in a new tab The expression of metabolism‐related lncRNAs in clinical subgroups. (A) Age. (B) Gender. (C) T: primary tumor status. (D) N: lymph node involvement status. (E) M: distant metastatic status. (F) Alive status. Unpaired two‐tailed t‐test was used to compare in Figure [95]8. *p < 0.05, **p < 0.01. ***p < 0.001. 3.7. LINC01138 modulates metabolic genes in COAD cell lines We next investigated the links between these four lncRNAs and the metabolic genes (|coefficient| > 0.3). As shown in Figure [96]9A, LINC01138 is the most closely lncRNA related to metabolic genes. Therefore, we initially selected LINC01138 as a candidate gene to explore its role in COAD. First, based on endogenous expression level of LINC01138 in cancer cell lines, we chose SW480 and HCT116 as the follow‐up research object (Figure [97]9B). As expected, enforced or impaired the expression of LINC01138 affected metabolic genes expression in COAD cell lines (Figure [98]8C). The correlation between LINC01138 and metabolic genes is shown in Figure [99]8D. FIGURE 9. FIGURE 9 [100]Open in a new tab Correlation between LINC01138 and metabolism genes A, The alluvial diagram showing the correlation between metabolism related lncRNAs and metabolism genes. B, qRT‐PCR assays showing the levels of LINC01138 in COAD cell lines (n = 3). C, Immunoblotting assays showing the levels of EDGKH, PDE4D, PYCR1, and TK1 in COAD cells. D, The gene expression correlation between LINC01138 and DGKH, PDE4D, PYCR1, and TK1 in COAD. Pearson correlation analysis and statistics were done in GraphPad Prism 9 for Figure [101]9D. 3.8. LINC01138 affects tumor progression in COAD in vitro To characterize the effects of LINC01138 in COAD, cells with stable overexpression and knockdown of LINC01138 were established. Overexpression of LINC01138 accelerated cell viability of SW480. Moreover, a significantly increased in cell growth and invasion was observed in cultured SW480 cells following LINC01138 overexpression (Figure [102]10A–F). Meanwhile, silencing the expression of LINC01138 result in decreased in cell viability, growth, and invasion in HCT116 cells (Figure [103]10A–F). Taken together, these data showed that LINC011381 affects tumor progression in COAD. FIGURE 10. FIGURE 10 [104]Open in a new tab LINC01138 affects tumor progression in COAD cells. (A and B) MTT assay indicating the viability of SW480 and HCT116 cells treated stably transfected as indicated, respectively (n = 5). (C and D) Soft agar assays indicating the growth of SW480 and HCT116 cells treated with stably transfected as indicated, respectively (n = 3). (E and F) Matrigel invasion assays indicating the invasion of SW480 and HCT116 cells treated with stably transfected as indicated, respectively (n = 3). Unpaired two‐tailed t‐test was used to compare in Figure [105]10. ***p < 0.001. 4. DISCUSSION Cellular metabolic reprogramming is a common feature for cancer, allowing tumor cells to meet the demands of rapid proliferation or survival.[106] ^1 For example, O‐GlcNAcylation activates PGK1 led to enhance glycolysis, and finally promote tumorigenesis in colon cancer.[107] ^24 A new study demonstrated that UCP2 deficiency resulted in promoting phospholipid synthesis, which facilitates the transformation of malignant colon cells^. [108]^25 These available studies show that target cancer metabolism maybe a promising way for cancer therapy. LncRNA is a class of noncoding RNA with more than 200 nucleotides with no protein‐coding capacity.[109] ^12 In recent years, aberranted lncRNAs have been found in various tumors including COAD, leading to the genesis and progression of cancer.[110] ^13 , [111]^14 , [112]^15 , [113]^16 , [114]^17 However, the association between lncRNA and cancer cell metabolism in COAD is largely unknown. In this work, 152 differentially expressed lncRNAs were significantly associated with differentially expressed metabolism genes. Among them, 12 lncRNAs were associated with OS of COAD patients. Finally, based on the analyses of LASSO, a prognostic prediction model involving four metabolism‐related lncRNAs were established. Patients with COAD were randomly divided to training‐ and validation‐cohorts. According to the median risk score, the training cohort was divided to the high‐risk group and low‐risk group. Patients in the high‐risk group experienced worse OS than those in the low‐risk group. ROC curve analyses revealed that the accuracy of the MRlncRNAs signature. Univariate and multivariate analyses further suggested that the gene signature is an independent prognostic factor of COAD. Besides, in clinical subgroups, the lncRNAs signature also performed well. Finally, as a representative gene of LINC01138, our result showed that LINC011381 affected tumor progression in COAD. In summary, this metabolic signature has a good ability to predict prognosis of COAD patients. All four metabolism‐related lncRNAs (ELFN1‐AS1, LINC01138, LINC02474, and PCAT6) displayed a high expression in the high‐risk group. Among them, ELFN1‐AS1, LINC02474, and PCAT6 have been reported to be associated with COAD.[115] ^26 , [116]^27 , [117]^28 For example, ELFN1‐AS1 acts as oncogene through sponging microRNAs in colorectal cancer.[118] ^26 LINC02474 by inhibiting GZMB expression to regulate progression of colorectal cancer.[119] ^27 Since our study revealed that LINC01138 was most closely with metabolism gene expression, we initially explored the function of LINC01138. Western blot assays revealed that forced LINC01138 expression in colon cancer cells SW480 increased the expression of diacylglycerol kinase eta (DGKH) and Phosphodiesterase 4D (PDE4D), whereas resulted in decrease the expression of pyrroline‐5‐carboxylate reductase 1 (PYCR1) and transketolase (TKT). LINC01138 knockdown in colon cancer cells HCT116 led to an opposite change in metabolic gene. This result is in line with correlation analysis from dataset. Aberranted LINC01138 is found highly expression some tumors. A study showed that LINC01138 by interacting with PRMT5 to facilitate tumor progression in hepatocellular carcinoma.[120] ^29 Moreover, another report demonstrated that LINC01138 affected cell growth in clear cell renal cell carcinoma through modulating lipid metabolism.[121] ^30 However, the association between LINC01138 and metabolism in COAD is unclarified. Here, we first characterized LINC01138 is associated with the progression of COAD. The four metabolism genes (DGKH, PDE4D, PYCR1, and TKT) are most closely associated with LINC01138. DGKH belongs to the diacylglycerol kinase enzyme family that participates in regulating the intracellular concentrations of diacylglycerol and phosphatidic acid.[122] ^31 Roles of DGKH in COAD have not been fully discussed. However, Molatore et al. proved that DGKH was upregulated in rats with pheochromocytoma and may represent a potential biomarkers of the disease.[123] ^32 Thus, DGKH may be important in COAD development. PDE4D is a subtype of metallohydrolases, which is involved in multiple tumor promoting processes^. [124]^33 A previous study uncovered that miR‐139‐5p by decreasing the level of PDE4D led to suppress the growth of xenograft colorectal tumors.[125] ^34 PYCR1 is the key enzyme responsible for proline synthesis. Previous studies demonstrated that PYCR1 is associated with the development of numerous cancers, including kidney cancer and hepatocellular carcinoma.[126] ^35 , [127]^36 Yan et al. revealed that PYCR1 is highly expressed in colorectal cancer, and reduced expression of PYCR1 inhibits proliferation of colorectal cancer, suggest an oncogenic role of PYCR1 in colorectal cancer.[128] ^37 TKT is the key enzyme controlling the metabolic flux and direction of pentose phosphate pathway, which is required for maintaining cell proliferation in various cancers.[129] ^38 , [130]^39 Meanwhile, a new study demonstrated that TKT promoted colorectal cancer cell growth and metastasis through modulating AKT phosphorylation.[131] ^40 In the present study, our preliminary study suggested that LINC01138 can regulate the level of metabolic genes, and combined with the role of LINC01138 in COAD, we reasonably speculated that LINC01138 may affect the progression of COAD by regulating metabolism. Inevitably, our study had a few limitations. First, our work based on open accessed TCGA databases, and the prognostic model should be validated in additional independent samples. Second, the coexpression mechanisms of crucial lncRNA and metabolism genes should be clarified. Third, functional studies are required to elucidate our findings in this study. 5. CONCLUSIONS In summary, we established a metabolism‐associated lncRNAs model to predict the prognosis in COAD patients. Our work may provide guidance for the treatment of COAD. AUTHOR CONTRIBUTIONS Yimin Sun: Data curation (lead); funding acquisition (equal); methodology (lead); resources (lead); validation (lead); writing – original draft (equal); writing – review and editing (equal). Bingyan Liu: Data curation (lead); formal analysis (lead); investigation (lead); software (equal); validation (equal); writing – original draft (equal). Baolai Xiao: Formal analysis (equal); investigation (equal); methodology (equal); software (equal); validation (equal); visualization (equal); writing – original draft (equal). Xuefeng Jiang: Data curation (equal); formal analysis (equal); methodology (equal); writing – original draft (equal). Jinjian Xiang: Data curation (equal); formal analysis (equal); investigation (equal); software (equal); writing – original draft (equal). Jianping Xie: Formal analysis (equal); investigation (equal); supervision (equal); validation (equal); writing – original draft (equal). Xiao Miao Hu: Conceptualization (lead); funding acquisition (lead); methodology (equal); project administration (lead); writing – review and editing (lead). FUNDING INFORMATION This work was supported by the Health and Family Planning Research Foundation of Hubei Province of China (grant no. WJ2016‐Y‐26). CONFLICT OF INTEREST The authors declare no competing interests. Supporting information Figure S1 [132]Click here for additional data file.^ (3.9MB, pdf) Table S1. [133]Click here for additional data file.^ (27.3KB, xlsx) Table S2. [134]Click here for additional data file.^ (903.5KB, xlsx) Sun Y, Liu B, Xiao B, et al. Metabolism‐related lncRNAs signature to predict the prognosis of colon adenocarcinoma. Cancer Med. 2023;12:5994‐6008. doi: 10.1002/cam4.5412 DATA AVAILABILITY STATEMENT The datasets used and/or analysed in this study are available from the corresponding author on reasonable request. REFERENCES