The significance of lactate and lactate metabolism in the development and progression of certain solid tumors has been reported. However, the role of lactate metabolism in the prognosis of multiple myeloma (MM) patients and in the tumor microenvironment (TME) remains unclear.

This study aimed to construct a prognostic risk-scoring model for MM patients based on lactate metabolism and to explore the impact of lactate metabolism on the TME.

Transcriptomic sequencing expression data and survival information of MM samples were obtained from the Gene Expression Omnibus (GEO) datasets GSE136324 and GSE4581. Lactate metabolism-related genes (LMRGs) for subsequent analysis were identified through the MSigDB and GeneCards databases. Consensus clustering and Cibersort were applied to explore the relationship between LMRGs and the TME. Differences in the TME among MM samples were analyzed using the Wilcoxon rank-sum test. A linear regression model was used to identify LMRGs associated with survival prognosis, and results were visualized using forest plots and Kaplan-Meier survival curves, with comparisons made by log-rank test. Least absolute shrinkage and selection operator (Lasso) regression was used for variable selection, and univariate and multivariate Cox regression analyses were performed to identify prognostic LMRGs for model construction. The predictive performance of the model was evaluated using Kaplan-Meier survival analysis and receiver operating characteristic (ROC) curves. Pearson correlation analysis was used to examine the relationship between the lactate metabolism-based risk score and immune score, stromal score, and tumor purity.

Based on the expression profiles of LMRGs, consensus clustering via the ConsensusClusterPlus package identified two lactate metabolism-related subtypes: Lactate.cluster A and Lactate.cluster B. Patients in Lactate.cluster A had significantly worse prognosis than those in Lactate.cluster B (χ²=19.11, P<0.000 1) . Significant differences were observed between the two subtypes in the infiltration levels of naive B cells, memory B cells, plasma cells, CD₈⁺ T cells, resting memory CD₄⁺ T cells, activated memory CD₄⁺ T cells, follicular helper T cells, gamma delta T cells, resting natural killer (NK) cells, activated NK cells, monocytes, M0 macrophages, M2 macrophages, activated dendritic cells, resting mast cells, activated mast cells, eosinophils, and neutrophils (P<0.05) . Several LMRGs were significantly associated with prognosis: high expression of PC, POLG, FASN, FABP2, SLC7A5, BCAT2, SLC16A8, HAGH, PFKL, PGK2, PKLR, CA2, and SLC2A1 was associated with better prognosis, while high expression of PDP1, AGK, PEX12, and PFKFB2 was associated with poorer prognosis (P<0.05) . Eleven independent predictive variables—SLC2A1, CA2, PKLR, PFKL, PFKFB2, SLC16A8, SLC7A5, AGK, FABP2, POLG, and PC—were included in the multivariate Cox regression analysis to construct a predictive model. Based on the median risk score, patients were divided into high-risk and low-risk groups. In the training set, the high-risk group had significantly worse overall survival than the low-risk group (χ²=59.02, P<0.05) . The area under the ROC curve (AUC) for predicting 5-year survival was 0.781 (95%CI=0.664-0.886) . In the validation set, the high-risk group also had significantly poorer survival (χ²=9.24, P<0.05) , with an AUC of 0.64 (95%CI=0.542-0.737) for predicting 1-year survival. Analysis of immunotherapy response data from the training dataset showed a significant difference in outcomes between the high- and low-risk groups (Z=-2.469, P=0.014) , with high-risk score being an independent prognostic factor (P<0.05) . Comparisons of immune cell infiltration between the two groups revealed significant differences in plasma cells, CD₈⁺ T cells, resting memory CD₄⁺ T cells, activated memory CD₄⁺ T cells, follicular helper T cells, resting NK cells, activated NK cells, monocytes, M0 macrophages, resting dendritic cells, activated dendritic cells, resting mast cells, activated mast cells, and eosinophils (P<0.05) . Significant differences were also observed in the expression of immune-related genes including HAVCR2, CD₂₇₄, PVR, CD₈₀, CTLA4, PDCD₁, CD₂₀₀R1, CD₂₇₆, CD₂₀₀, BTLA, LGALS3, and VTCN1 (P<0.05) . Pearson correlation analysis between the lactate metabolism-based risk score and enrichment scores of 50 hallmark pathways revealed positive correlations with WNT beta-catenin signaling, androgen response, and UV response, and negative correlations with KRAS signaling, pancreas beta cells, and heme metabolism (P<0.05) . The risk score was positively correlated with stromal score (P<0.05) but not with tumor purity (P>0.05) .

This study constructed a risk score prognostic model for multiple myeloma patients based on lactate metabolism and validated it with independent datasets. The new prognostic model is robust and demonstrates better performance in predicting long-term survival. Furthermore, the analysis of lactate metabolism in relation to the TME suggests that lactate metabolism may influence immune cell populations within the TME of MM, thereby affecting tumor progression and the prognosis of MM patients.

乳酸和乳酸代谢在一些实体肿瘤发生和进展中的意义已有报道。然而，乳酸代谢在多发性骨髓瘤（MM）患者预后及肿瘤微环境（TME）中的作用尚不清楚。

本研究旨在构建基于乳酸代谢的MM患者风险评分预后模型，并探索乳酸代谢对TME的影响。

首先从基因表达综合数据库（GEO）下载MM数据集GSE136324、GSE4581的转录组测序表达数据和样本生存信息。从MSigDB数据库和GeneCards数据库筛选出纳入后续分析的乳酸代谢相关基因。应用一致性聚类和Cibersort方法探索乳酸代谢相关基因与TME的关系。采用Wilcoxon秩和检验分析MM样本的TME差异。基于线性回归模型筛选出与生存预后关联的乳酸代谢相关基因，并利用森林图和Kaplan-Meier生存曲线进行可视化呈现，生存曲线的比较采用Log-rank检验。采用Lasso回归分析筛选变量，采用单因素和多因素Cox回归分析筛选与预后相关的乳酸代谢相关基因建立预后模型。采用Kaplan-Meier生存分析以及受试者工作特征（ROC）曲线评估模型的预测能力。采用Pearson相关性分析探讨乳酸代谢模型风险评分与免疫评分、基质评分和肿瘤纯度的相关性。

基于乳酸代谢相关基因的表达谱应用ConsensusClusterPlus包进行一致性聚类，识别出2个乳酸代谢相关基因亚类Lactate.clusterA和Lactate.clusterB，Lactate.clusterA亚类预后较Lactate.clusterB亚类预后更差（χ²=19.11，P<0.000 1）。初始B细胞、记忆B细胞、浆细胞、CD₈阳性T细胞、静息记忆CD₄阳性T细胞、活化记忆CD₄阳性T细胞、滤泡辅助性T细胞、γδT细胞、静息自然杀伤细胞、活化自然杀伤细胞、单核细胞、M0型巨噬细胞、M2型巨噬细胞、活化树突状细胞、静息肥大细胞、活化肥大细胞、嗜酸性粒细胞、中性粒细胞在两类亚类样本间差异有统计学意义（P<0.05）。筛选出预后相关的乳酸代谢相关基因，胆碱磷酸转移酶（PC）、聚合酶γ（POLG）、脂肪酸合酶（FASN）、脂肪酸结合蛋白2（FABP2）、溶质载体家族7成员5（SLC7A5）、分支链氨基酸转氨酶2（BCAT2）、单羧酸转运蛋白8（SLC16A8）、3-羟基酰基辅酶A脱氢酶（HAGH）、6-磷酸果糖激酶（PFKL）、3-磷酸甘油酸激酶2（PGK2）、丙酮酸激酶同工酶R（PKLR）、碳酸酐酶2（CA2）、葡萄糖转运蛋白1（SLC2A1）基因高表达预后较好；丙酮酸脱氢酶磷酸酶催化亚基1（PDP1）、乙酰辅酶A琥珀酰转移酶（AGK）、过氧化物酶体组装蛋白12（PEX12）和6-磷酸果糖-2-激酶/果糖-2，6-双磷酸酶2（PFKFB2）基因高表达预后较差（P<0.05）。筛选出SLC2A1、CA2、PKLR、PFKL、PFKFB2、SLC16A8、SLC7A5、AGK、FABP2、POLG、PC 11个独立预测因子变量纳入多因素Cox回归分析构建预测模型，根据风险得分的中位数将样本分为高风险组与低风险组，训练集高风险组生存率低于低风险组（χ²=59.02，P<0.05）。绘制ROC曲线，模型预测5年生存的ROC曲线下面积（AUC）为0.781（95%CI=0.664~0.886）。验证集高风险组生存率低于低风险组（χ²=9.24，P<0.05），模型预测1年生存的AUC为0.64（95%CI=0.542~0.737）。收集训练数据集样本的免疫治疗效果资料，结果显示高风险组与低风险组预后情况存在差异（Z=-2.469，P=0.014），高风险组是患者预后的影响因素（P<0.05）。比较高风险组、低风险组样本免疫细胞浸润水平，浆细胞、CD₈阳性T细胞、静息记忆CD₄阳性T细胞、活化记忆CD₄阳性T细胞、滤泡辅助性T细胞、静息自然杀伤细胞、活化自然杀伤细胞、单核细胞、M0型巨噬细胞、静息树突状细胞、活化树突状细胞、静息肥大细胞、活化肥大细胞、嗜酸性粒细胞丰度差异有统计学意义（P<0.05）。高风险组、低风险组样本甲型肝炎病毒细胞受体2（HAVCR2）、程序性细胞死亡1配体1（CD₂₇₄）、脊髓灰质炎病毒受体（PVR）、CD₈₀、细胞毒性T淋巴细胞相关抗原4（CTLA4）、程序性细胞死亡蛋白1（PDCD₁）、CD₂₀₀受体1（CD₂₀₀R1）、CD₂₇₆、CD₂₀₀、B和T淋巴细胞衰减因子（BTLA）、半乳糖凝集素3（LGALS3）、V域免疫球蛋白抑制剂1（VTCN1）基因表达差异有统计学意义（P<0.05）。计算50 hallmark通路的富集得分，采用Pearson相关性分析探究乳酸代谢预后模型风险得分和不同通路的相关性，发现WNT beta catenin signaling、Androgen response和UV response通路与乳酸代谢预后模型风险得分呈正相关，KRAS signaling、Pancreas beta cells和Heme metabolism与乳酸代谢预后模型风险得分呈负相关（P<0.05）；乳酸代谢预后模型风险得分与基质评分呈正相关（P<0.05），与肿瘤纯度无相关性（P>0.05）。

本研究构建了一个基于乳酸代谢的MM患者风险评分预后模型，在预测长期生存方面表现更好。而乳酸代谢与TME相关分析提示乳酸代谢可能影响MM患者TME中的免疫细胞群，从而影响肿瘤的进展，并进而影响MM患者的预后。