Participants and their sample collection in the cohort
In this study, we first collected feces samples from MM patients at the time of initial diagnosis without any clinical treatment. All patients had similar ages and BMI and were without gastrointestinal diseases. The sampling procedure was approved by the Cancer Research Institute of the Central South University Medical Ethics Committee. Here, 19 newly diagnosed MM patients including 14 males and 5 females were selected. Their average age and BMI index were 60.4 ± 6.2 and 22.9 ± 4.2, respectively. To match the numbers, 18 healthy controls were selected with the age of 57.9 ± 4.4 and BMI index of 24.3 ± 2.1, and the number of males and females was 13 and 5, respectively. Meanwhile, the age and BMI index of 19 patients and 18 controls were verified, and no significant difference was found (two-tailed Welch’s t test, P = 0.49). Similarly, the gender distribution of the two groups was also validated (Fisher’s exact test, P = 1) (Table S4a, Additional file 15).
Fresh fecal samples from each participant were collected and frozen in liquid nitrogen. Subsequently, samples were stored at – 80 °C until the time of transport on dry ice to Beijing Genomics Institute Genomics Co., Ltd for metagenomic sequencing. In the meantime, a fasting serum sample from each participant was also obtained. All samples were stored at – 80 °C until submission to Metabo-Profile Biotechnology (Shanghai) Co., Ltd for untargeted metabolomics detection.
Moreover, as described above, additional feces samples were also collected from 17 MM patients and 21 healthy controls (Table S4b, Additional file 15), which along with the metagenome sequencing samples was treated as an expanded cohort for the qPCR experiment. And, two serum samples from MGUS patients were also collected in this study (Table S4c, Additional file 15).
Metagenomic sequencing, taxonomic classification, and functional annotation
The total DNA of each fecal sample was extracted using E.Z.N.A Stool DNA extraction kit (Omega Bio-tek, Norcross, Ga.) following the manufacturer’s instructions. The quality of DNA was analyzed using Qubit (Invitrogen, USA) and 1% agarose gel electrophoresis. All samples were sequenced on the Illumina platform (paired-end; insert size, 350 bp; read length, 150 bp). Raw sequences were processed to remove low-quality sequences using Trimmomatic (version 0.36), and human sequences were filtered out with the human reference genome (hg38) using Bowtie2 (version 2.2.4). The high-quality reads for the 37 samples were acquired, with an average of 25 Gb per sample.
Subsequently, the remaining high-quality reads were used for taxonomic classification using Kraken (version 1.0) and the standard Kraken database with default settings [14]. Although all samples were sequenced with a similar depth, the read counts table of several levels (e.g., phylum, class, order, family, genus, species) were rarefied to the minimum read counts to reduce the effects of uneven sampling in the cohort.
In addition, Metagenome Composition Vector (version 2.3.0) was used for functional annotation based on the clean sequences [29]. In this study, our study was focused mainly on alterations of functional proteins, and the KEGG level 4 (KEGG ontology) was therefore rarefied accordingly for subsequent identification of different functional proteins.
Bacterial diversity in the cohort
To evaluate the sequencing depth and species richness in all subjects of the cohort, rarefaction analysis was performed by using R (version 3.5.0) and package vegan (version 2.5-2). To estimate the bacterial diversity, the Shannon–Weaver index at both species and genus level was also calculated by using R and package vegan.
In addition, the Bray-Curtis dissimilarity indices between samples at the species level were calculated by using R and package vegan, after the species read counts were subjected to square root transformation and Wisconsin double standardization. To estimate the between-sample (β) diversity, permutational multivariate analysis of variance (PERMANOVA) was performed based on the Bray-Curtis distances. Meanwhile, principal coordinate analysis (PCoA) was performed based on the Bray-Curtis dissimilarity matrix to visualize β diversity. Equally, the β diversity between samples at the KOs level was executed accordingly.
Statistical analysis
On the R platform, metagenomic features (i.e., taxonomic and functional features) were analyzed statistically. The top four phyla and the top thirty genera were found to account for the majority of microbes, respectively, at the phylum and genus level. Their significance in MM and HC groups was determined by the P value of Wilcoxon rank sum test < 0.05. In addition, to focus on the more significant microbes, the species with low read-count percentages were considered as noise and removed from the matrix (threshold was set at 0.01%), and the species that were present in less than half of the subjects in the cohort were deleted. Similarly, the KOs that were present in less than half of the subjects in the cohort were excluded, and the threshold percentage of KOs was set at 0.001%. Significantly different species and KOs were all identified by R platform and using package DESeq2 (version 1.26.0) based on read counts [17], and the difference that was significant was determined with the absolute value of log base 2 of fold change > 1 and adjusted P value < 0.05. The different KOs were mapped to the corresponding modules based on KO database (https://www.kegg.jp/kegg/ko.html) [49]. Modules mapped by at least 2 KOs were presented. The Spearman’s correlation in this study was calculated by using R and package psych (version 1.8.12) and visualized by package pheatmap (version 1.0.12) or Cytoscape (version 3.4.0).
Redundancy analysis
RDA was performed by using R (version 3.4.4) and package vegan (version 2.5-4), and the interactions were visualized by using package ggplot2 (version 3.1.1).
16S ribosomal DNA qPCR
The relative abundances of the different species identified by metagenomic sequencing were further tested using qPCR in both original fecal samples and more recently accumulated samples, which were collected from newly diagnosed patients with MM. The paired primers specific for each species were designed using Primerblast based on species-specific regions on the 16S ribosomal DNA (V1 or V2), while the conserved sequences were used for amplification of total bacteria [50]. Meanwhile, the coverage and specificity of each paired primers were also evaluated using an online tool TestPrime 1.0 (https://www.arb-silva.de). Primers are listed in Table S5 (Additional file 16), with the exception of bacteria sp. or complex and three species (i.e., Butyrivibrio proteoclasticus, Clostridium botulinum, Prevotella fusca), for which specific primers could not be designed. Besides, all primers were validated using gradient PCR to detect the annealing temperature and the specificity of primers.
Appropriate fecal samples were used to extract total bacterial DNAs (Mogen stool DNA kit), whose concentrations were subsequently measured using Nanodrop. One percent AGAR gel electrophoresis was used to detect whether the extracted DNA had been degraded (100 V 30 min). The reaction mixture (20 μl) for qPCR contained Applied BiosystemsTM PowerUpTM SYBRTM Green Master Mix, forward and reverse primer (final concentration 400 nM), and the extracted DNA (10 μl). The thermocycling program was 40 cycles and consisted of 95 °C for 15 s and 56 °C for 15 s and 72 °C for 1 min with an initial cycle of 50 °C for 2 min and 95 °C for 2 min. The melting curve was constructed in the range of 60 to 95 °C after PCR was performed.
Assuming that for all templates and primers a cycle equally doubles the number of template DNA, the relative abundance of a certain strain (i) can be calculated as follows:
$$ \mathrm{Relative}\ \mathrm{abundance}\ (i)=\frac{{\left(\frac{1}{2}\right)}^{\mathrm{CT}i}}{{\left(\frac{1}{2}\right)}^{\mathrm{CT}c}}={\left(\frac{1}{2}\right)}^{\mathrm{CT}i-\mathrm{CT}c}={\left(\frac{1}{2}\right)}^{\triangle \mathrm{CT}} $$
The cycle threshold of strain i primer and common primer (total bacteria) are represented by CTi and CTc, while △CT denotes the difference between them. From the equation, the logarithm of relative abundance negatively correlates linearly with △CT.
Metabolomics analysis of human serum samples and mouse bone marrow samples
Gas chromatography coupled to time-of-flight mass spectrometry (GC-TOFMS) system (Pegasus HT, Leco Corp., St. Joseph, MO, USA) was used to quantify the detected metabolites. Metabolites were annotated using the mammalian metabolite database JiaLibTM and employing a strict matching algorithm incorporated in the XploreMET software that uses both retention times and fragmentation patterns in the mass spectrum.
In human serum samples, a total of 180 measurable and reproducible metabolite signals were detected, 67 metabolites were unidentified. We therefore analyzed 113 identified metabolites and 28 ratios of metabolites from KEGG metabolic pathways (Table S6, Additional file 17). Statistical analysis of metabolic profiling of subjects was performed using the SIMCA software (SIMCA 14.1). Based on the results of metabolic profiling with Pareto scaling, the sophisticated multivariate statistical model, orthogonal partial least-squares discriminant analysis (OPLS-DA), was established to visualize differences in metabolite profiles. Meanwhile, the R2Y and Q2 of OPLS-DA model were 0.853 and 0.679, respectively. Metabolite signals causing the significant differences in metabolic profiling between two groups were screened on the basis of variable influence on projection (VIP) > 1 and P value of Welch’s t test < 0.05.
Mouse BM samples were prepared based on the method previously published with minor modifications [51,52,53]. All of the standards used were obtained from Sigma-Aldrich (St. Louis, MO, USA), Steraloids Inc. (Newport, RI, USA), and TRC Chemicals (Toronto, ON, Canada). All the standards were accurately weighed and prepared in an appropriate solution to obtain the individual stock solution at a concentration of 5.0 mg/mL. An appropriate amount of each stock solution was mixed to create stock calibration solutions. The raw data files generated by UPLC-MS/MS were processed using the TMBQ software (v1.0, Human Metabolomics Institute, Shenzhen, Guangdong, China) to perform peak integration, calibration, and quantitation for each metabolite. The current TMBQ is hosted on Dell PowerEdge R540 Servers operated with RedHat Enterprise Linux 7.5. The secured Java UI (User Interface) permits the user to have access to use a great variety of statistical tools for viewing and exploring project data on its own desire.
The 5TGM1 mouse model
C57BL/KaLwRij mice of 6 weeks old were purchased from Harlan Laboratories Inc. (Harlan Mice, Netherlands). C57BL/KaLwRij mice can occasionally develop MM disease late in life, in which 5TGM1 cell line can be isolated. However, after 5TGM1 cells were injected into young mice, the cells colonized the bone marrow and caused myeloma. In this study, all animal studies were approved by the Ethical Committee for Animal Experiments of Central South University. The 8- to10-weeks-old mice were injected with the phosphate-buffered saline (PBS) vehicle or 8 × 105 5TGM1-luc cells, and the clinical endpoint was achieved when mice exhibited signs of hind limb weakness.
Fecal microbiota transplantation experiment, sample collection, and detection
As indicated in Fig. 3 i, fresh stool from HC and MM patient was separately collected, and then stool (2.5 g per person) was mixed and suspended using PBS (10 mL). The suspension liquid was then filtered using 70 μm strainers, and the filtrate was centrifuged at 2000 rpm for 10 min. Subsequently, after the removal of the supernatant, the remaining pellet was resuspended using 2 mL PBS. The mixture was used for fecal microbiota transplantation by gavage (200 μL per mouse).
Before gavage was performed, all mice were treated with a cocktail of broad-spectrum antibiotics including ampicillin (0.2 g/L), vancomycin (0.1 g/L), neomycin (0.2 g/L), and metronidazole (0.2 g/L) in the drinking water for 2 weeks [40]. Subsequently, FMT_HC mice and FMT_MM mice, respectively, were given fecal suspension for 2 weeks by gavage (twice per week). In contrast, the mice of the PBS group were given PBS by gavage for 2 weeks (twice per week). To evaluate the influence of fecal microbiota transplantation on experiment mice, the stool was collected from FMT_HC mice and FMT_MM mice at week -4 and week 0 (before and after gavage). The relative abundances of several characteristic bacteria were separately measured, and PCoA was performed to determine whether the colonization of distinct microflora was successful in FMT_MM and FMT_HC mice, including Clostridium butyricum, Streptococcus mitis, Streptococcus gordonii, Streptococcus pneumoniae, Raoultella ornithinolytica, Citrobacter freundii, Enterobacter cloacae, Klebsiella aerogenes, Klebsiella pneumoniae, and Klebsiella variicola. Notably, Anaerostipes hadrus, Clostridium saccharobutylicum, Streptococcus salivarius, and Streptococcus oralis were undetectable in most of the mice.
At week 0, all mice were induced to develop MM via tail-vein injection of 5TGM1 cells (1 × 106 cfu). Subsequently, serum sample from all mice was collected every week for the measurement of the concentration of lgG2b, NH4+, and urea; fecal sample from all mice was collected to measure the relative abundances of the species of interest. Live imaging of tumors was performed for all mice once per week from week 2. At week 6, after all mice were euthanized, cecum contents were collected for the measurement of the concentrations of l-glutamic acid, l-glutamine, NH4+, and urea, and the activities of urease and glutamine synthase [54]. In the meantime, the BMs from all mice were separately harvested from their femur and tibia by centrifugation at 4 °C (6000 rpm, 5 min). The cell-free supernatants of BM were prepared for targeted metabolomic detection. And the kidney tissue from all mice was separately collected for immunocytochemistry analysis.
ELISA was performed for the detection of mouse IgG2b in serum using Mouse IgG2b ELISA Quantification set (Bethyl Laboratories, Inc., USA). The concentration of NH4+ and urea in the serum was detected using the Blood Ammonia Assay Kit (Nanjing Jiancheng Biotechnology Co., Ltd., China) and Urea Detection Kit (diacetyl-monoxime colorimetric method) (Shanghai Zhuocai Biotechnology Co., Ltd., China), respectively. And the concentration of l-glutamic acid and l-glutamine was measured using the Glutamine Colorimetric Assay Kit (BioVision, Inc. USA).
In addition, the cecal contents (0.1 g) were dissolved using PBS (900 μL), and then shaken with glass beads (0.2 g, 425–600 μm) to break up the microorganisms inside (3000 rpm, 6 min). The mixture was centrifuged at 4 °C (5000 rpm, 6 min) to obtain supernatant for the subsequent measurement using the corresponding kits, including the concentration of l-glutamic acid, l-glutamine, urea, and ammonia and the activities of urease and glutamine synthetase. As shown in the kit instructions, before the measurement of l-glutamine, it was necessary to remove macromolecular proteins from samples by centrifuging in an ultrafiltration tube (14,000 rcf, 20 min) (UFC500396, Millipore, USA). The activities of urease and glutamine synthase were detected using Microorganism URE ELISA Kit (Shanghai Zhuocai biotechnology Co., Ltd., China) and Microorganism GS ELISA Kit (Shanghai Zhuocai biotechnology Co., Ltd., China), respectively. In the statistical analysis of the results, we excluded the dead mice and the outliers.
The characteristic bacteria transplantation experiment and sample collection
Experimental mice were also gavaged with 2 × 108 cfu/200μL of Klebsiella pneumoniae (BNCC 102997 = ATCC 10031) and Clostridium butyricum (BNCC 337239 = ATCC 19398), which were purchased from Beijing BeiNa Biotechnology Institute. Klebsiella pneumoniae was cultured in nutrient broth (Product ID: 022010, Guangdong HuanKai Microbial Co., Ltd., China) at 37 °C and 200 rpm for ~ 16 h, while Clostridium butyricum was statically cultured in thioglycollate medium (Product ID: HB5191, Qingdao Hope Bio Technology Co., Ltd., China) under anaerobic conditions at 37 °C for ~ 18 h. In addition, as previously described [55], Klebsiella pneumoniae with mutant glnA (coding glutamine synthetase, EC: 6.3.1.2) was constructed by homologous recombination technique using plasmid pKO3-Km. According to the turbidity/absorbance of the fermentation broth, there was no significant difference in the growth of wild-type K. pneumoniae and mutants. Here, for the purpose, we selected Mut3 for subsequent experiments, due to less glnA expression and less remaining glutamine in broth (Figure S10, Additional file 18). Microorganism cells were obtained from fermentation broth and were resuspended with PBS for subsequent gavage. Each mouse was given by gavage 200 μL suspension (about 2 × 108 cells).
Similarly, all experimental mice were treated with a broad-spectrum cocktail of antibiotics for 2 weeks, and then each group of mice was transplanted with a specific fresh fermentation broth for 2 weeks (twice per week). At week 0, all experimental mice were induced to develop MM via tail-vein injection of 5TGM1 cells (1 × 106 cfu). To amplify the effect on experiments, microbial transplantation continued twice a week. Meanwhile, serum lgG2b and tumor fluorescence intensity from all mice were monitored every week.
Gavage experiments with ammonium and urea
We carried out mouse studies with NH4Cl (100 mM) and urea (50 mM) by gavage, with the control being with NaCl (0.9%). Each group of mice (i.e., NaCl, NH4Cl, and Urea) was given 100 μL by gavage once a day. After 1 week of gavage, all mice were induced to develop MM via tail-vein injection of 5TGM1 cells (1 × 106 cfu) at week 0 (Fig. 8a), and afterwards all mice still received gavage daily.
To see if gavage can change the abundance of bacteria, the stool was collected from all mice at week -1, week 0, week 2, week 4, and week 6. Subsequently, these stool samples were used to measure separately the relative abundances of several MM-enriched bacteria, including Streptococcus mitis, Streptococcus gordonii, Streptococcus pneumoniae, Raoultella ornithinolytica, Citrobacter freundii, Enterobacter cloacae, Klebsiella aerogenes, Klebsiella pneumoniae, and Klebsiella variicola. At week 6, after all mice were sacrificed, cecal contents were collected for the measurement of the concentrations of l-glutamic acid, l-glutamine, NH4+, and urea and the determination of the activities of urease and glutamine synthase. In addition, serum samples were collected to measure the concentrations of l-glutamic acid and l-glutamine.
Mouse studies using defective diet and sample collection
We also conducted mouse experiments, in which mice were fed with glutamine-deficient diet (Gln-) or glutamine/cysteine-deficient diet (Plus-). The mice fed with a holistic diet (Ctr) were used as controls. After 1 week of feeding, all mice were induced to develop MM via tail-vein injection of 5TGM1 cells (1 × 106 cfu) at week 0 (Fig. 9a), and their diet remained the same. During the experimental process, the MM progression of mice was monitored by the use of tumor fluorescence intensity and was found to be most significant at week 7. Thus, at week 7, serum samples were collected for the measurement of the concentrations of lgG2b, l-glutamic acid, and l-glutamine.
Immunocytochemistry analysis
The collected kidney tissue samples from all the mice were first fixed with paraformaldehyde, embedded in paraffin after dehydration, and sliced for immunohistochemistry. The slides then were subjected to dewaxing, rehydration, and hydrogen peroxide treatment. Subsequently, the tissue sections were incubated with anti-IgG2b kappa antibody in a 1:2000 dilution overnight at 4 °C. Next, the slides were incubated with HRP-conjugated secondary antibody and stained with 3,3′-diaminobenzidine tetrahydrochloride hydrate (DAB) for 3 min. Finally, cell nuclei were counterstained with hematoxylin. The stained sections were evaluated by using PerkinElmer Quantitative Pathology Imaging System with software inForm 2.4.