Causal effects of genetically determined blood metabolites on multiple myeloma: a Mendelian randomization study

Previous studies have shown that metabolites play an important role in phenotypic regulation. However, the causal relationship between metabolites and multiple myeloma has not been adequately investigated. Here, we attempt to explore the causal effects of genetically determined blood metabolites on multiple myeloma. The large-scale public blood metabolites and multiple myeloma datasets from independently published genome-wide association studies (GWAS) were used to explore the causal relationship between each genetically determined blood metabolite and multiple myeloma through inverse variance weighted (IVW), weighted median, MR-Egger and mode-based estimation methods. Sensitivity tests were performed to evaluate the stability and reliability of the results by MR-Egger regression and leave-one-out methods. Metabolic pathway analysis was further explored using filtered data. Statistical analyses were all performed in R. Among 452 metabolites, ten known metabolites and three unknown metabolites had significant causal relationship with multiple myeloma (P < 0.05). Four known metabolites, 3-methyl-2-oxovalenate, oxidized bilirubin, isovalerylcarnitine and glutamine carnitine, reached statistical significance in IVW models. Metabolic pathways analysis identified four significant pathways. The occurrence of multiple myeloma may have a causal relationship with these four metabolites, and there are four metabolic pathways that are also related to the occurrence of multiple myeloma. This can provide new ideas for exploring early screening and treatment of multiple myeloma.


Selection of instrumental variables
Uniform criteria were set for SNP screening: ① P < 5 × 10 -8 as statistically significant for inclusion in the study; ② linkage disequilibrium analysis with reference to the genotype of the European population (EUR) of the Thousand Genomes, which also needed to meet an LD threshold of r 2 < 0.1 within 500 kb, retaining the single nucleotide polymorphism with the smallest P value.

Statistical analyses for MR
Investigating the causal relationship between each blood metabolite and multiple myeloma by using the Two-SampleMR package (version 0.5.6) in the R 19 .In this study, inverse variance weighting (IVW) was used 20 as the primary causal association effect assessment method.IVW is a method for MR to Meta-summarize the effects of multiple loci when analyzing multiple SNPs.IVW is used to ensure that all SNPs are valid instrumental variables and are completely independent of each other.In addition, we used the weighted median method (WME) 21 , MR-Egger regression 22 , simple mode-based estimation 23 and weighted mode-based estimation 23 to test the reliability and stability of the results.When the estimates of the causal association effects obtained from the above five different MR models were similar, we could conclude that the causal association between the metabolite and multiple myeloma was reliable and stable.If only one SNP remains, use the Wald ratio method.We also performed multiple hypothesis testing, using P < 1.10 × 10 -4 (after Bonferroni correction) as the threshold indicating the presence of a direct causal association 24 and 1.10 × 10 -4 < P < 0.05 as a potential risk predictor for multiple myeloma.Heterogeneity tests and genetic pleiotropy tests were performed for causality for all metabolites with P IVW < 0.05.At the same time, we calculated odds ratios as well as the power to detect a significant result across a range of odds ratios to help interpret the results and improve reproducibility 25 .

Heterogeneity and sensitivity tests
Q-test for IVW and MR-Egger was used to detect potential violations of the assumption by the heterogeneity of the association between individual IVs.The included instrumental variables will be considered not heterogeneous when P > 0.05.The default fixed-effects model was employed if no substantial heterogeneity (P > 0.05) was observed; otherwise, the random-effects model was utilized 26 .MR-Egger was applied to estimate horizontal pleiotropy according to its intercept, ensuring that genetic variation was independently associated with exposure and outcome.When P > 0.05, it will be considered that there is less likely genetic pleiotropy in the causal analysis.This study used the leave-one-out method to assess the likelihood of associations observed by individual SNP driver.

Information on instrumental variables (SNP)
Of all 2.1 million SNPs from 452 metabolites, we found 880 SNPs that met the selection criteria for instrumental variables, which was listed in Table S1.After overlapping these SNPs with multiple myeloma GWAS data, 839 SNPs were ultimately included in the follow-up analysis.These could be seen in Table S2.The flow chart of MR study is shown in Fig. 1.The minimum F-statistic for all SNPs included in the follow-up analysis was 28.81, indicating that the instrumental variables for metabolites were sufficiently plausible (F > 10).

MR analyses results
Of the 452 metabolites, thirteen metabolites were causally associated with multiple myeloma at a significant level (P < 0.05), including ten known metabolites and three unknown metabolites, as detailed in Fig. 2.After Bonferroni correction (P < 1.10 × 10 -4 ), no metabolites were found that still had a significant effect.
Of the ten known metabolites, five may be associated with an increased risk of multiple myeloma: glutamine carnitine, isovalerylcarnitine, cysteine-glutathione disulfide, leucine and pelargonate; five may be associated with a decreased risk of multiple myeloma: 3-methyl-2-oxovalerate, oxidized bilirubin, 2-methylbutyroylcarnitine, indolelactate and lysine.

Heterogeneity and sensitivity analysis
Four metabolites that reached statistical significance in IVW models were tested for heterogeneity and gene pleiotropy.P-values for heterogeneity and gene pleiotropy were greater than 0.05, indicating that none of the metabolites had heterogeneity or gene pleiotropy.The relevant results are shown in Table 1.
Scatter plots of MR analysis results for the four metabolites are shown in Fig. 3.The sensitivity analysis of the above four metabolites by the leave-one-out method was robust, with no single nucleotide polymorphisms in any of the metabolites significantly affecting the results.Forest plots of the leaveone-out results for the metabolites are shown in Fig. 4.

Metabolic pathway analysis
The ten metabolites screened were subjected to metabolic pathway analysis using MetaboAnalyst5.0software and the results are shown in Fig. 5A.The metabolites were further subjected to KEGG pathway enrichment analysis and the results are shown in Fig. 5B, C. The metabolic pathway analysis showed that there were six metabolic pathways affecting multiple myeloma in the serum, four of which had statistically significant differences (P < 0.05), as shown in Table 2.

Discussion
Of the 452 blood metabolites involved in this study, ten are potential predictors of multiple myeloma risk.Five metabolites were included that may be associated with an increased risk of developing multiple myeloma, namely glutamine carnitine, isovalerylcarnitine, cysteine-glutathione disulfide, leucine and pelargonate.The other five may be associated with a decreased risk of multiple myeloma: 3-methyl-2-oxovalerate, oxidized bilirubin, The present study identified bilirubin oxide as a potential protective substance to reduce the risk of multiple myeloma.Li Volti et al. 28 found that hematological malignancies exhibit an altered homeostasis of the redox balance, which can lead to the activation of various survival pathways that, in turn, lead to disease progression and chemoresistance.The heme oxygenase-1 (HO-1) pathway is thought to play an important role among these pathways.HO catalyzes enzymatic breakdown of heme, releasing carbon monoxide (CO), ferrous iron (Fe 2+ ) and bilirubin oxide.As a degradation product of heme, bilirubin oxide inhibits the degradation of heme and thus    the heme oxygenase-1 (HO-1) pathway.Raninga et al. 29 also found that concurrent inhibition of HO-1 would improve therapeutic outcomes in MM patients.Another metabolite that may reduce the risk of multiple myeloma is 3-methyl-2-oxovalerate, a branched-chain alpha-keto acid (BCKA) produced by the catabolism of isoleucine, which has been reported to be associated with type 2 diabetes and insulin resistance 30 .In a study on the catabolic pathway of branched-chain amino acids, 3-methyl-2-oxovalerate was the strongest predictor of IFG among the identified intermediate metabolites, independent of glucose 31 .Abnormal amino acid metabolism is one of the important features of MM.The important metabolic pathway of amino acids participates in protein synthesis as basic raw materials 32 .It has been shown that the metabolism of branched-chain amino acids influences the prognosis of multiple myeloma 33 .When catabolism of branched-chain amino acids is impaired, blood levels of 3-methyl-2-oxovalerate are reduced, potentially increasing the risk of multiple myeloma.
Glutamine carnitine and isovalerylcarnitine are both acylcarnitines 34 .Carnitine and acylcarnitine are key substances in cellular energy metabolism and can be synthesized from amino acids in the human kidney and liver.Their physiological roles include acting as sole carriers of long-chain fatty acids, transporting long-chain fatty acid classes into the mitochondria for beta-oxidation, and regulating the intracellular balance between free and acyl coenzyme A. Characteristic changes in one or more acylcarnitines indicate abnormal beta-oxidation of fatty acids and abnormal metabolism of branched-chain amino acids 35 .It has been shown that carnitine abnormalities are associated with metabolic diseases such as isoleucine and leucine metabolism disorders, isovaleric acidemia and type 2 diabetes 36 .It has also been suggested that carnitine can stimulate neuroprotective factors 37 .In the past study, isovalerylcarnitine has been proved to be able to activate the calpain system, producing an early and marked increase in apoptosis and cell killing 38 .MM cells were found to be dependent on glucose and glutamine metabolism in the first metabolic analysis of MM cells 39 .Higher levels of isoleucine and lower levels of glutamine   www.nature.com/scientificreports/and some lipids have been observed in myeloma patients at diagnosis, but not after remission 40 .Bajpai et al. 41 were able to show that targeting the glutamine metabolism sensitizes MM cells to the bcl-2 inhibitor venetoclax.
Studies of cellular metabolism have identified LDHA and HIF1α as novel targets for drug resistance in MM under hypoxic conditions in the bone marrow.Inhibition of LDHA and HIF1A can restore sensitivity to therapeutic drugs such as bortezomib.This suggests a correlation between branched-chain amino acid metabolism and the development of multiple myeloma, which is confirmed by metabolic pathway analysis.This study is innovative in many ways: firstly, it combines metabolomics and genomics to investigate the causal relationship between blood metabolites as exposure factors and multiple myeloma using a Mendelian randomization approach, which has important clinical research value; secondly, this study uses multiple MR models and sets strict quality control conditions to make the results reliable and stable; finally, the large number of exposure factors involved in this study were metabolites in the blood, thus the analytical workload was huge and posed analytical challenges.There are also some limitations to this study.One limitation of our study is that the GWAS data we used were all from European populations, so generalization to other populations may be limited; secondly, although we identified a number of metabolites that were causally associated with multiple myeloma in our study, some of these were unknown metabolites and could not be studied for further analysis.Additionally, while MR analysis provides valuable insights into etiology, we must consider that blood metabolites can be influenced by various factors such as diet, host genetics, and the gut microbiome 42 .We can just conclude that blood metabolites are associated with multiple myeloma, but causation is not necessarily direct, thus it is important to note that our findings should be validated through rigorous RCTs and basic research before application in the clinic.
In summary, this study used a Mendelian randomization approach to explore the possible causal link between blood metabolites and multiple myeloma.Although no direct causal relationship was found, a number of potential risk predictors for multiple myeloma were identified, which has the potential to provide new insights into the influence of genetic-exposure interactions in the disease process of multiple myeloma.Furthermore, the analysis of both potential risk factors and associated metabolic pathways suggests that the metabolism of branched-chain amino acids may provide a new reference for the early screening and treatment of multiple myeloma.

Figure 1 .
Figure 1.The flow chart of MR study revealing the causal relationship between human blood metabolites and the risk of MM.MR Mendelian randomization, MM multiple myeloma, SNPs single nucleotide polymorphisms.

Figure 2 .
Figure 2. Forest plot for the causality of blood metabolites on multiple myeloma.CI confidence interval, OR odds ratio, nSNP number of single nucleotide polymorphism.

Figure 5 .
Figure 5. Metabolic pathway analysis and KEGG pathway enrichment analysis of ten metabolites by MetaboAnalyst5.0.

Table 1 .
Results of 5 MR models of known metabolites that reached statistical significance in IVW models and the heterogeneity and pleiotropy tests.MR Mendelian randomization, IVW inverse variance weighting, WME weighted median method.