A New Genetic Signature of Lactate Metabolism-Associated Genes Predicting Clinically Distinctive Features and Tumor Microenvironment in Colorectal Cancer

Abstract

Background

Colorectal cancer (CRC) is characterized by its high malignancy and challenging prognosis. A significant aspect of cancer is metabolic reprogramming, where lactate serves as a crucial metabolite that contributes to the development of cancer and the tumor microenvironment (TME). Current studies have indicated that lactate plays a significant role in the progression of CRC. However, the relationship between lactate and the tumor microenvironment remains understudied, underscoring the potential of lactate as a novel biomarker.

Methods

We sourced transcriptomic data for colorectal cancer (CRC) patients from The Cancer Genome Atlas (TCGA), the International Cancer Genome Consortium (ICGC), and the Gene Expression Omnibus (GEO) portals, along with the corresponding clinical information. Utilizing univariate Cox regression in conjunction with LASSO regression analysis, we identified genes involved in lactate metabolism that are associated with CRC prognosis. Subsequently, we developed models based on multi-factor Cox regression. To evaluate the correlation between tumor mutational burden (TMB), tumor microenvironment (TME), and lactate scores with patient survival, we conducted gene set enrichment analysis (GSEA) and immunogenic signature analyses.

Results

3 lactate metabolism-related genes (LMRGs) (SLC16A8, GATA1, and PYGL) were used to construct models that categorized patients into 2 subgroups based on their lactate scores. The function of the differential genes between the 2 subgroups was mainly enriched in cell cycle and mRNA division, and the prognosis of patients in the high score subgroup was poor. Furthermore, a significant positive correlation was observed between TMB and LMRGs scores in the high-scoring group (P = 0.003, r ² = 0.12). Lastly, LMRGs also reflected the characteristics of TME, with differences in immune cells and immune checkpoints between the 2 subgroups.

Conclusions

LMRGs may serve as a promising biomarker for predicting prognostic survival in CRC patients and to assess the TME.

Plain language summary

不适用.

Introduction

Colorectal cancer is currently 1 of the most prevalent malignancies within the digestive system. The incidence and mortality rates of cancer are among the top 3 and remain alarmingly high. According to estimates, there will be nearly 1.9 million new cases of colorectal cancer and 900,000 fatalities globally in 2020.^1,2 High mutational load has now become a marker of immunotherapy responsiveness and immune checkpoint therapy is currently an available treatment for CRC. ³ Tumor metabolism is the foundation of TME and an essential player of anti-tumor immune response, ⁴ where changes driven by oncogenes affect TME, thereby limiting the immune response and impeding cancer therapy. However, by altering cancer metabolism, it is possible to modulate immune metabolism, thereby enhancing the efficacy of cancer therapy. ⁵

Alterations in metabolic pathways are a universal characteristic of colorectal cancers, essential for providing the requisite energy and nutrients to fuel cancer cell proliferation. ⁶ Lactate, a key metabolite, is emblematic of the Warburg effect that is distinctive of cancer metabolism. Aerobic glycolysis culminates in elevated lactate concentrations both intra- and extracellularly, ⁷ and the accumulation of lactic acid within the tumor microenvironment is pathognomonic of diseases such as cancer. ⁸ Investigations have elucidated that lactate and its metabolic derivatives can stimulate cell migration, exhibit antioxidant properties, and contribute to immune evasion and angiogenesis. ⁹ For example, the production of lactic acid by cancer cells activates immunity and GPR81 on cells such as endothelium to promote angiogenesis, immune evasion and chemoresistance to promote cancer progression. ¹⁰ In addition, lactic acid may also play an antitumor role by increasing the stemness for CD8⁺ T cells.^11,12 Given the important role of lactate and its metabolic processes in tumor progression and TME, targeting this pathway holds promise as a potent strategy in cancer therapeutics.

In this study, we screened for lactate metabolism-related genes (LMRGs) and developed predictive models based on patient survival outcomes. We then performed an extensive analysis of the tumor mutational burden (TMB) across various subgroups and examined the relationship between the tumor microenvironment and LMRGs. Our findings indicate that LMRGs hold significant prognostic value for CRC and may influence TME.

Materials and Methods

Data Acquisition

LMRGs were obtained by consulting the MSigDB (V7.4) dataset and searching for lactate-related genes. The data related to TCGA CRC were downloaded from the UCSC XENA database and used as the training set (https://xenabrowser.net/datapages/). For validation of the results of the training set, the CRC dataset downloaded from the ICGC website (https://dcc.icgc.org/) and GEO website (https://www.ncbi.nlm.nih.gov/geo/) was used as the validation set. Baseline information for CRC patients in both cohorts is shown in Table 1. The flow of the study is shown in Figure 1.

Table 1.

Clinical and Pathological Characters of CRC Patients in TCGA and ICGC Cohort.

Characteristics	Number
TCGA cohort (N = 689)
Age	>60	432
	≤60	197
Gender	Male	335
	Female	294
T	T1	20
	T2	109
	T3	428
	T4	37
	Tis	1
N	N0	357
	N1	151
	N2	118
	NX	2
M	M0	467
	M1	89
	MX	64
Clinical stage	I	109
	II	229
	III	181
	IV	90
ICGC cohort (N = 582)
Age	>60	399
	≤60	183
Gender	Male	308
	Female	274

Figure 1.

The flowchart of the study.

Construction of LMRGs Model

Univariate Cox regression and LASSO Cox regression were performed on 278 LMRGs to determine the prognostic LMRGs. Multivariate Cox regression was used to determine the scoring coefficients for genes. The formula for calculating LMRGs score is as follows: score = individual gene expression level × sum of gene coefficients. ¹³ We classified CRC sufferers into high and low subgroups based on mediometric scores. For evaluating the prognostic value of LMRGs, we performed Kaplan-Meier (KM) survival analysis. In addition to this we also explored the relationship between LMRGs and clinical stratification, such as TNM staging. In order to predict CRC patients at 1, 3 and 5 years, we constructed a column line plot based on LMRGs and important clinicopathological parameters and demographic information. Calibration curves were applied to assess the agreement between the prediction and the reality results.

Comprehensive Analysis of TME

First, we used the ESTIMATER ¹⁴ algorithm to estimate the patient’s stroma, immune score, and tumor purity. Relative scores of tumor-infiltrating immune cells were calculated using CIBERSORT (RRID: SCR_016955) to assess differences in immune function between the 2 subgroups.

Gene Function Analysis

To determine the biological pathways involved in lactate, we performed enrichment analysis using GSEA. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted utilizing the R package, clusterProfiler (RRID: SCR_016884). This GO analysis encompassed biological processes (BP), cellular components (CC), and molecular functions (MF). Any GO terms and KEGG pathways (RRID: SCR_012773) exhibiting a P value less than 0.05 were deemed significantly enriched.

Statistical Analysis

All data analysis was completed by R software. Differences between groups were compared using Student's t test and Wilcoxon rank sum test. Univariate and multivariate analyses were performed using Cox regression models. Log-rank tests were performed to assess survival differences. P values <0.05 were considered statistically significant.

Ethics Approval

Ethics approval do not apply to this manuscript. The data utilized in this study were obtained from publicly available databases, including The Cancer Genome Atlas (TCGA), Gene Expression Omnibus (GEO), and the International Cancer Genome Consortium (ICGC). These datasets are accessible to the public and have undergone prior ethical review and approval by their respective organizations. As such, no additional ethical approval was required for the use of these data in our research. All data were handled in accordance with the data usage policies of the respective databases. The reporting of this study conforms to STROBE guidelines and RECORD guidelines.^15,16

Results

Construction of the LMRGs Model

After conducting univariate Cox regression analysis on the screened gene set to identify LMRGs associated with survival outcome, a total of 21 LMRGs were found to be significantly correlated with prognosis (Figure 2(A)). Following LASSO-COX regression selection, the LMRGs were constructed by multivariate Cox regression analysis for only 13 LMRGs (Figure 2(B) and (C)). According to the regression coefficients and the performance of the 3 key genes associated with colorectal survival, The LMRGs score was calculated for each CRC patient as follows: LMRGs risk score = 0.68425*SLC16A8-2.72668*GATA1+0.20472*PYGL (Figure 2(D)).

Figure 2.

Cox regression analysis and LASSO analysis of LMRGs. (A) Univariate Cox regression analysis screened 21prognostic LMRGs. (B) The LASSO coefficient profile of 21 prognostic LMRGs. (C) Tuning parameter(λ) selection in LASSO model using cross-validation. (D) Multivariate Cox regression analysis of LMRGs.

Prognostic Implications of LMRGs

Using the LMRGs score obtained from the above formula and the median score of all CRC patients as the cut-off point, we divided these cases into 2 subgroups with high and low scores, and used principal component analysis to see whether the 3 key genes could be accurately separated into these 2 subgroups (Figure 3(A)). We used heat maps to show the performance of the 3 key genes used to construct the model in the 2 subgroups mentioned above (Figure 3(B)) and the differential genes in the 2 groups with high and low LMRGs scores (Figure 3(C)). Based on KM analysis, we can find that patients with the upper LMRGs score have a shorter median survival and poorer prognosis, and although not statistically significant in the ICGC dataset, it was also evident that patients in the high score group had a poorer survival prognosis (Figure 3(D)). In addition to this, we collected gene expression profiles and clinical information of CRC patients through the GEO database and calculated their LMRGs scores, and found that patients with higher scores in the GEO_39582 and GEO_87211 datasets had a poorer prognosis, whereas in GEO_17536, the trends shown were consistent, although not statistically significant (Supplemental Figure 1). For the purpose of accurately predicting the probability of patient survival, we created a histogram that combined the calculated LMRGs scores with other clinicopathological characteristics sitting in a joint analysis, including age, gender and TNM stage (Figure 3(E)). We have been able to evaluate the survival rates of these colorectal cancer patients at 1, 3 and 5 years, respectively, on the basis of the summary scores obtained from the nomo plots. The research findings demonstrate the crucial role of LMRGs scoring in the progression and prognosis of CRC.

Figure 3.

Grouping and prognosis of LMRGs in CRC. (A) PCA was used to determine whether the samples could be grouped correctly based on the LMRGs score. (B) Heatmap of the expression of 3 key genes between the 2 subgroups. (C) Volcano map of differential genes between the 2 subgroups. (D) KM survival analysis in the LMRGs-low and LMRGs-high groups in both TCGA and ICGC datasets. (E) Development of a nomogram by combining LMRGs score with age, gender, and TNM stage to predict the survival probability.

Connection to TMB

Based on the TCGA dataset, we counted the TMB for each CRC patient. We found that patients in the high scoring group had a higher TMB and that there was a positive correlation between patients’ TMB and score (R = 0.12). (Figure 4(A) and (B)). Follow-up survival analysis was undertaken to assess whether there was a combined effect of LMRGs score and TMB on patient survival prognosis. We have observed that patients with low LMRGs scores and low TMB have significantly longer survival time compared to those with high LMRGs scores and high TMB. (Figure 4(C) and (D)).

Figure 4.

Tumor mutation characteristics of different LMRGs subgroups. (A) Differences in TMB in the LMRGs-low and LMRGs-high groups. (B) Relationship between TMB and LMRGs scores. (C) KM survival analysis of TMB. (D) Effect of LMRGs score combined with TMB on survival prognosis.

TME Characteristics Between the Two Subgroups

TME is composed of stromal and immune cells that have a professorial impact on tumor development and their sensitivity to treatment. We found relatively high stromal cell abundance and immune scores in the low LMRGs group, while tumor purity was higher in the high LMRGs group than in the low group (Figure 5(A)). On top of that, there were variations in immune cells between the 2 subgroups, mainly between M1 macrophages as well as CD8 T cells (Figure 5(B)). In addition to the immune cells mentioned above, we investigated further the relationship between immune checkpoints and LMRGs scores. We found that the LMRGs score was a positive correlation with the expression of some immune checkpoints, including CD200 and IDO1 (Figure 6(A) and (B)).

Figure 5.

The landscape of TME in CRC. (A) Stromal score, immunoreactivity and tumor purity between the 2 subgroups. (B) Differences in immune cells between the 2 subgroups.

Figure 6.

Correlation of LMRGs score with immune checkpoints. (A) Differences in immune checkpoint expression between the 2 subgroups. (B) Correlation of immune checkpoints with LMRGs scores.

Analysis of the Biological Function of LMRGs

For the purpose of exploring the LMRGs molecular mechanism, we conducted GSEA analysis to identify differentially expressed genes between 2 groups. GO and GSEA analysis by differential genes between the 2 groups revealed that LMRGs is mainly involved in regulating cell cycle and metabolism-related signals as well as in mRNA division (Figure 7(A) and (B)).

Figure 7.

GSEA of LMRGs-low group and LMRGs-high group. (A) Enrichment results of GO analysis. (B) Enrichment results of signaling pathways.

Discussion

The progression of CRC is not only genetically and epigenetically regulated, but also closely related to TME.^17,18 Metabolic reprogramming is indispensable for various types of immune cells to maintain their own and microenvironmental homeostasis. Cancer metabolism plays a crucial role not only in the signaling pathways involved in tumor initiation and survival, but also in regulating anti-tumor immune responses through the release of metabolic byproducts. ¹⁹ Aerobic glycolysis is an important hallmark of metabolic reprogramming in cancer cells. Under aerobic conditions, glucose can still undergo metabolism and produce lactate. Lactate contributes significantly to the regulation of TME and serves as a crucial signaling molecule for metabolic pathways, immune responses and intracellular communication.^19,20 The LMRGs model consists of 3 key genes including SLC16A8, GATA1, and PYGL. SLC16A8’s primary substrate is lactate and is a member of the SLC16 gene family, which is involved in a wide range of metabolic pathways and plays a role in cancer. ²¹ GATA1 is a spectrum-restricted transcription factor present in highly differentiated cells such as megakaryocyte lines and mast cell lines, and is essential for proper cell differentiation, proliferation and apoptosis. ²² A previous study found that GATA1 can enhance the efficiency of PI3K/AKT signaling pathway and thus promote CRC progression. ²³ It has been demonstrated that PYGL can promote cellular epithelial-mesenchymal transition (EMT) and metastasis by promoting glycolysis, establishing a connection between glucose metabolic reprogramming and EMT, and ultimately to promote tumor progression.^24,25

TMB was quantified for each patient based on mutations across the genome. In this study, we found a synergistic interaction between TMB and LMRGs, where patients exhibiting high levels of both indicators had the least favorable prognosis, possibly due to the high mutations associated with high TMB. Cancer occurs as a result of somatic mutations and cloning. ²⁶ Certain hereditary DNA mutations have been shown in previous studies to significantly increase the risk of cancer. ²⁷ Additionally, there were discernible disparities in the tumor immune microenvironment from 1 subgroup to another. We observed discrepancies between the 2 subgroups not only in overall immune scores but also concerning specific classes of immune cells, for instance, CD8 T cells and M1 macrophages. Macrophages are widely believed to be involved in inflammation, immune escape, stromal remodeling, and cancer metastasis. M1 macrophages mainly exhibit pro-inflammatory properties, ²⁸ but previous research has found that M1 macrophages can suppress anti-immune responses to promote the progression of cancers such as oral cancer and liver cancer.^29,30 CD8⁺ T cells are closely linked to the metabolic environment and changes in the environment affect their differentiation. ³¹ Their function is stringently governed by multiple transcription factors and they play an anti-tumor role in the context of cancer.^32,33 A hallmark of the tumor microenvironment is the increased concentration of lactate, which is pivotal not only for tumor growth but also for maintaining immune cell homeostasis within the TME. Previous studies have shown that lactate can inhibit T-cell differentiation and activation,^34,35 as well as promote macrophage polarization.^36,37 In addition to immune cells, we explored the relationship between LMRGs scores and immune checkpoints. Recently immune checkpoint therapies can improve the effectiveness of cancer treatment. ³⁸ Differences in immune checkpoint expression were noted between the 2 subgroups and the scores were positively correlated with IDO1 and CD200. The level of immune checkpoint expression could serve as a predictive biomarker for immunotherapy responsiveness. Therefore, the LMRGs score might prove valuable in forecasting a patient’s amenability to immunotherapy.

In this study, we have developed a novel CRC prognosis and TME prediction model based on lactate metabolism-related genes. However, several limitations exist in this study. Firstly, the exact molecular functions of the 3 genes involved in constructing the model are not known in the context of colorectal cancer, and further experiments are necessary to elucidate their roles in CRC. Secondly, the data used to construct and validate the model is retrospective, and a multicenter prospective study is required to verify its accuracy and clinical value.

Overall, we have developed a novel LMRGs model that holds significant predictive value for CRC survival and is instrumental in assessing TME. Our findings indicate a strong correlation between LMRGs and factors such as age, gender, and disease stage. Furthermore, we observed variations in TME status among patients with differing LMRGs scores, including alterations in stromal cell expression levels, the extent of immune cell infiltration, the degree of immune activation, tumor purity, and the expression of immune checkpoints. The novelty of this research lies in its comprehensive approach, as it does not focus on a specific pathway but rather examines the relationship between lactate and TME holistically. This investigation further delves into the correlation between lactate and CRC. The study substantiates the pivotal role of lactate within the TME, suggesting its potential as an innovative biomarker and therapeutic target for CRC.

Conclusion

In summary, our findings indicate that lactate significantly influences the progression of colorectal cancer and shapes the immune microenvironment. Moreover, lactate scoring may be able to serve as an innovative tool for predicting patient survival and aiding in treatment strategies.

Supplemental Material

Supplemental Material - A New Genetic Signature of Lactate Metabolism-Associated Genes Predicting Clinically Distinctive Features and Tumor Microenvironment in Colorectal Cancer

Supplemental Material for A New Genetic Signature of Lactate Metabolism-Associated Genes Predicting Clinically Distinctive Features and Tumor Microenvironment in Colorectal Cancer by Kaiwen Wang, Yu Lou, and Zhihui Tao in Cancer Control