Protocol for correlating the gut microbiome and metabolomics in patients with intracranial aneurysms

Summary

Gut-microbiome-combined metabolomics studies in cerebrovascular disease highlight the microbiota-gut-brain axis in neurological disorders. Here, we present a protocol for correlating the gut microbiome and metabolomics in patients with intracranial aneurysms. We describe steps for sample collection, fecal genomic DNA extraction, rRNA PCR amplification, sequencing library construction, and rRNA sequencing. We then detail procedures for metabolite extraction, liquid chromatography-tandem mass spectrometry (LC-MS/MS) non-targeted metabolomics sequencing, and ELISA for cerebrospinal fluid and plasma samples. Finally, we perform combined multi-omics analysis.

For complete details on the use and execution of this protocol, please refer to Xu et al.¹

Graphical abstract

Highlights

• •Customizable process for stool collection and storage in intracranial aneurysm patients
• •Normalized process for fecal DNA and metabolite extraction and sequencing data analysis
• •Joint analysis of multi-omics data and correlation with clinical characteristics

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

Gut-microbiome-combined metabolomics studies in cerebrovascular disease highlight the microbiota-gut-brain axis in neurological disorders. Here, we present a protocol for correlating the gut microbiome and metabolomics in patients with intracranial aneurysms. We describe steps for sample collection, fecal genomic DNA extraction, rRNA PCR amplification, sequencing library construction, and rRNA sequencing. We then detail procedures for metabolite extraction, liquid chromatography-tandem mass spectrometry (LC-MS/MS) non-targeted metabolomics sequencing, and ELISA for cerebrospinal fluid and plasma samples. Finally, we perform combined multi-omics analysis.

Before you begin

The protocol delineated below provides a detailed description of the specific steps for gut microbiome and metabolomics correlation analysis using fecal samples from intracranial aneurysm (IA) patients. The recent application of 16S rRNA sequencing of fecal samples in conjunction with UPLC-MS analysis has facilitated a deeper understanding of the mechanisms underlying gut dysbiosis and the development of various diseases.² Of particular interest is the pivotal role exerted by the MGBA in the onset and progression of neurological disorders.^3,4,5 Compared to constructed animal models, most human samples from patients exhibit considerable inter-individual variability, which is primarily attributable to factors such as dietary habits, geographical location and age. Accordingly, the sample size of each group should be sufficiently large, generally comprising at least 15 individuals. Concurrently, in order to minimize the impact of geographic diet and age, it is imperative to select healthy family members with a similar age profile who reside and consume a similar diet to the patient as normal controls. For 16S rRNA sequencing, the preferred methodology was full-length 16S rRNA gene sequencing, which offers the advantage of obtaining more abundant and accurate microbial species data compared to traditional methods targeting the V3-V4 variable region. Furthermore, a discovery cohort and a validation cohort are established concurrently for cross-validation purposes, enhancing the precision and reliability of the findings. Importantly, the combined microbiome and metabolomics analysis in the same fecal samples, ultimately coupled with the determination of key metabolite concentrations in cerebrospinal fluid (CSF) and peripheral blood (PB), can provide a comprehensive insight into the integrated role of key differential metabolisms in neurological diseases through the MGBA.

Institutional permissions

This research was screened and endorsed by the Ethics Committee at Zhongnan Hospital of Wuhan University (2024072K). All participants or legal representatives in this research have signed an informed consent form, with strict adherence to good clinical practice and the Declaration of Helsinki. In addition, gender should not affect the experimental results in the research.

Fecal sample collection and storage

To prevent contamination during sample collection, fresh stool samples of approximately 2–4 g were obtained by nursing staff using sterile containers and while wearing sterile gloves. Following collection, fresh fecal samples were first placed at −20°C later moved to a −80°C freezer within the same day, using an ice box. Based on the stage of disease progression of IAs, we categorized the subjects into two groups: unruptured intracranial aneurysms (UIAs) and ruptured intracranial aneurysms (RIAs). Stool samples from family members of UIA patients were collected to serve as a normal control group. This approach aimed to thoroughly investigate the relationship between the formation and progression of UIA and the MGBA axis by comparing the three groups. Each group exhibited differences in the timing of sample collection and screening conditions for fecal samples compared to healthy family controls.

For family controls and UIA patients, stool samples were collected within 24 h. However, unlike the family controls, samples from UIA patients were collected prior to cerebral angiography and aneurysm surgery. Specifically, for the collection of samples from RIA patients, all procedures were conducted under anesthesia and sedation to minimize the risk of triggering aneurysm re-rupture during the sampling process.

Note that the alterations in microbiota and metabolite levels associated with the use of antibiotics and probiotics can be significant; therefore, all participants must fulfill the requirement of not having used such medications within the previous two months. UIA and RIA patients should be sampled prior to initiation of antibiotic treatment. In addition, subjects with gastrointestinal diseases should be excluded from the study. For sample preservation, we recommend that all fresh stool samples be stored at −20°C immediately after collection (within 30 min) and then move to a −80°C freezer within one week.

PB and CSF collection and storage

A peripheral blood sample of 5 mL from fasting IA patients was collected using a vacuum blood collection tube within 24 h, then placed at 4°C during transportation to the laboratory for preservation and subsequent processing. For CSF, samples were collected during aneurysm clipping surgery. A 2 mL CSF sample was safely obtained in an ice box without compromising surgical outcomes and was immediately transferred to a −80°C freezer.

Plasma sample collection

Timing: 3 h for 30 samples

This section describes the collection, separation, and storage of plasma from whole blood samples for ELISA.

• 1.
  Sample preparation and centrifugation.

  • a.
    Collect whole blood samples using vacuum blood collection tubes or centrifuge tubes on ice.
  • b.
    Place the samples directly into the centrifuge and centrifuge under the following conditions: 1,000 × g for 20 min at 4°C.
  • c.
    After centrifugation, stratify the blood, with the upper liquid layer representing plasma and the lower layer consisting of red and white blood cells.
• 2.
  Plasma separation and storage.

  • a.
    Aspirate the upper plasma layer with a sterile, enzyme-free pipette tip, ensuring not to disturb the bottom cell layer. Transfer the plasma to a sterile, enzyme-free EP tube.
  • b.
    Label the EP tubes with the sample information and store them at −80°C for subsequent freezing prior to ELISA analysis.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Stool samples from healthy family members and IA patients	Xu et al.1	https://doi.org/10.1016/j.isci.2024.111184
CSF and peripheral blood from IA patients	Xu et al.1	https://doi.org/10.1016/j.isci.2024.111184
Chemicals, peptides, and recombinant proteins
Methanol	Merck	67-56-1
Acetonitrile	Merck	75-05-8
2-Chloro-L-phenylalanine	Aladdin Biotech	103616-89-3
Formic acid	TCI	64-18-6
Critical commercial assays
TGuide S96 magnetic soil/stool DNA kit	TIANGEN BIOTECH	DP812
Monarch DNA gel extraction kit	Monarch	T1020
SMRTbell template prep kit	Pacific Biosciences	100-222-300
DNA/polymerase binding kit	Pacific Biosciences	100-372-700
Qubit dsDNA HS assay kits	Thermo Fisher Scientific	Q32851
Human linoleic acid (LA) ELISA kit	MEIAO Biotech	MO-P38041R
VAHTS DNA clean beads	Vazyme Biotech	N411-03
KOD FX Neo	TOYOBO	KFX-201S
KOD One PCR master mix	TOYOBO	KMM-101
Deposited data
16S rRNA sequencing data	Xu et al.1	https://ngdc.cncb.ac.cn/gsa-human: accession no. HRA007513
Untargeted metabolomics data	Xu et al.1	https://ngdc.cncb.ac.cn/omix: accession no. OMIX007574
Oligonucleotides
16S rRNA primers pair 27F/1492R	Klindworth et al.6	https://doi.org/10.1093/nar/gks808
Software and algorithms
BMKCloud platform	Beijing Biomarker Biotech	www.biocloud.net
Cutadapt	Martin7	https://cutadapt.readthedocs.io/en/stable/
BLASTN	Altschul et al.8	https://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/
Usearch	Edgar et al.9	https://drive5.com/usearch/
QIIME2	Bolyen et al.10	https://qiime2.org
PERMANOVA	Anderson11	https://www.primer-e.com/software
Metastats	White et al.12	https://cbcb.umd.edu/software/metastats
LEfSe	Segata et al.13	https://huttenhower.sph.harvard.edu/lefse/
PICRUSt2	Douglas et al.14	https://github.com/picrust/picrust2
STAMP	Parks et al.15	https://beikolab.cs.dal.ca/software/
BugBase	Tonya et al.16	https://github.com/knights-lab/BugBase
Other
250 bp DNA ladder marker	Takara Bio	3424A
ACQUITY UPLC HSS T3 1.8 μm 2.1 × 100 mm	Waters	186003539
RNase A	TIANGEN	RT405-02
Agarose	Beijing Bomei Fuxin Technology	A011-2
Gradient thermal cycler	Thermo Fisher Scientific	1154J63

Step-by-step method details

The workflow chart illustrates the entire workflow of multi-omics experiments in this protocol (Figure 1). Steps 1–16 detail the procedure for DNA extraction from fecal samples of IA patients and full-length 16S rRNA gene sequencing. Steps 17–30 outline the procedure for metabolite extraction from fecal samples and non-targeted metabolomics sequencing. Steps 31–41 describe the procedure for determining the concentrations of key metabolites by ELISA in peripheral blood and cerebrospinal fluid samples. Steps 42–44 describe the process of integrating microbiome and metabolomics data analysis. The remaining sections of the protocol highlight key points for fecal collection and workflows for microbiome and metabolome data analysis.

Figure 1.

The workflow chart in this protocol

Part 1: Fecal genomic DNA extraction

Timing: 3 h for 24 samples

This section describes the application of TGuide S96 Magnetic Soil/Stool DNA Kit (Tiangen Biotech, DP812, Beijing, China) for fecal genome DNA extraction following the manufacturer’s instructions.

• 1.
  Sample preparation and lysis.

  • a.
    Weigh 0.25–0.5 g of each fecal sample into a 2 mL centrifuge tube. If the sample is liquid, transfer 200 μL to a centrifuge tube.

    Note: To obtain an adequate DNA yield and to ensure sample integrity, a minimum of 50 mg of fecal sample is advised.

  • b.
    Add 500 μL of Buffer SA, 100 μL of Buffer SC, along with 0.25 g of grinding beads to the sample.

    • i.
      Mix the contents by vortexing or using a tissue homogenizer.
    • ii.
      Heat the mixture at 70°C for 15 min to improve lysis efficiency.
  • c.
    Centrifuge the mixture at 13,400 × g for 1 min.
  • d.
    Carefully pipette approximately 500 μL of the supernatant into a fresh 2 mL tube.
• 2.
  Add buffer and centrifuge.

  Note: RNA may persist in the fecal sample. If RNA removal is necessary, it is recommended to add an additional 10 μL of RNase A (Tiangen Biotech, RT405-02, Beijing, China).

  • a.
    Add 200 μL of Buffer SH to the supernatant. Mix thoroughly, vortex for 5 s, and incubate at 4°C for 10 min.
  • b.
    Centrifuge at 13,400 × g for 3 min. Carefully collect the supernatant into a fresh 2 mL tube.
  • c.
    Add 500 μL of Buffer GFA (confirm that isopropanol is included) and gently invert the tube 10 times to mix.
• 3.
  Magnetic bead processing.

  • a.
    Introduce 10 μL of the magnetic bead suspension (G) and mix thoroughly by vortexing for 5 min.
  • b.
    Place the centrifuge tube on a magnetic rack and allow it to remain undisturbed for 30 s to ensure the beads are fully captured.
  • c.
    Carefully aspirate the supernatant.
• 4.
  Deproteinization and washing.

  • a.
    Take the tube off the magnetic stand and introduce 700 μL of the deproteinization solution (RD). Mix thoroughly by vortexing for 5 min.
  • b.
    Return the tube to the magnetic stand and leave it for 30 s to ensure the beads are fully bound. Carefully aspirate the supernatant.
  • c.
    Take the tube off the magnetic stand again and introduce 700 μL of the washing solution (PWD). Vortex for 3 min to mix.
  • d.
    Return the tube to the magnetic stand for another 30 s to ensure the beads capture fully. Gently aspirate the supernatant.
  • e.
    Repeat the washing procedure one more time.
• 5.
  Drying and elution.

  • a.
    Place the centrifuge tube on the magnetic stand and allow it to air dry at 25°C for 5–10 min.

    Note: Ensure complete evaporation of ethanol to prevent inhibition of subsequent enzymatic reactions, but avoid excessive drying to facilitate DNA elution.

  • b.
    Take the tube off the magnetic stand and introduce 50–100 μL of elution buffer (TB).
  • c.
    Mix thoroughly by vortexing, then incubate at 56°C for 5 min, gently vortexing three times during the incubation with 3–5 vortexing motions each time.
  • d.
    After the incubation, return the centrifuge tube to the magnetic stand for 2 min to ensure complete bead capture.
  • e.
    Carefully place the DNA solution into a fresh tube for proper storage.
  • f.
    Measure and record the DNA concentration using the Nanodrop (troubleshooting 1).

    Pause point: The extracted DNA can be stored at −20°C for up to 6 months.

Part 2: 16S rRNA PCR amplification and sequencing library construction

Timing: Approximately 15–16 h for 24 samples

This section details the workflow for amplifying the full-length 16S rRNA gene and constructing sequencing libraries. The described steps ensure that the DNA samples meet the required quality and quantity for successful downstream sequencing.

• 6.Measure nucleic acids for concentration using a microplate reader (GeneCompang Limited, Synergy HTX).

Note: Verifying that the nucleic acid concentration meets the necessary criteria (OD260/280 value between 1.8 and 2.0, and concentration diluted to 1 ng/μL).

• 7.Prepare master mix for 16S full-length reaction system PCR (Table 1).
• 8.Dispense 18.5 μL of the prepared Master Mix into each well of the 96-well plate. Pipette up and down multiple times.
• 9.Add 1.5 μL fecal genomic DNA (1 ng/μL) to each well and mix by blowing several times.
• 10.Seal the plate and centrifuge for 1 min to gather the liquid from the wells.
• 11.Perform PCR amplification of the mixture using a gradient thermal cycler (1154J63, Thermo Fisher Scientific Inc.) according to the settings specified in Table 2.

Pause point: The PCR products can be stored at −20°C for up to 2 weeks before proceeding with electrophoresis.

• 12.
  Quality determination of PCR amplification products.

  • a.
    Evaluate the integrity of the amplification products by electrophoresis on a 2.0% agarose gel.

    CRITICAL: Quality control of PCR amplification products is essential for the subsequent library construction. Accordingly, Table 3 presents the pass indicators for DNA quality control. In the context of full-length 16S rRNA gene sequencing, achieving a product length of 1500 bp is the primary focus. The typical concentration of amplified product ranges from 1 ng/μL to 15 ng/μL, with a total mass generally between 0.3 μg and 3 μg. Typically, the 1500 bp target band of the amplified product should be brighter than the marker band at the corresponding position, indicating that the concentration of the amplified product meets the mixing requirements. Figure 2A presents the electrophoretic images of several samples, demonstrating that the products are suitable for subsequent experiments. The 1500 bp band represents the target product, while the ∼1000 bp bands are non-specific amplification products that were excluded from further analysis.

  • b.
    Use a spectrophotometer (Thermo Scientific, NanoDrop 2000, USA) to measure the concentration and purity of the DNA.

Table 1.

Master mix for 16S full-length reaction system PCR

Reagent	Initial concentration	Volume (μL/sample)	Volume (μL/24 samples)
Amplicon PCR 27F Primer	0.4 μM	1.5	36
Amplicon PCR 805R Primer	0.4 μM	1.5	36
KOD ONE Master Mix	2×	15	360
Nuclease-free water	N/A	10.5	252
Total	N/A	30 (including 1.5 μL of sample)	N/A

Table 2.

Settings for index PCR

PCR cycling conditions
Steps	Temperature	Time	Cycles
Initial denaturation	95°C	2 min	1
Denaturation	98°C	10 s	25
Annealing	55°C	30 s
Extension	55°C	1 min 30 s
Final extension	72°C	2 min	1
Hold	4°C	Infinite	–

Table 3.

DNA quality control standards

Target band size	Target band intensity	Explanation 1	Explanation 2	Total amount	Result
The size and position are correct	Target band is brighter than the 1500 bp marker	Library construction can be completed	Can accommodate two or more library constructions	≥0.3 μg	Excellent
The size and position are correct	Non-specific bands are present, but they can be purified using gel extraction	Two sample are required to complete the library construction	Can accommodate one constructions	<0.3 μg	Acceptable
The size and position are incorrect	Target band intensity is low	Not available for library construction	Not available for library construction	<0.2 μg	Unqualified

Figure 2.

PCR amplification products quality control

(A) The electrophoretic images of several samples.

(B) Sequence length distribution graphs of several samples.

Part 3: 16S rRNA sequencing

Timing: 3 days

This section describes the PacBio sequencing conditions, which were performed at Biomarker Biotech (www.biocloud.net) in Beijing, China.

• 13.During library construction, subject the products to damage repair, end repair, and ligation of junctions using the SMRTbell Template Prep Kit (100-991-900, PacBio).
• 14.Measure the concentration of the final library using Qubit and analyze its size with the Agilent 2100.
• 15.Perform pre-sequencing binding with the PacBio Binding Kit (102-194-100, PacBio).
• 16.Purify the final reaction product using AMpure PB Beads, and load the sample onto the Sequel II sequencer for sequencing.

Note: Batch effect in the presence of discovery and validation set, see troubleshooting 2.

Part 4: Metabolite extraction

Timing: Approximately 6–8 h for 24 samples

This section describes the procedures for extracting metabolites from solid and liquid samples, ensuring optimal recovery and preservation of metabolite integrity for subsequent non-targeted metabolomics analysis.

For non-targeted metabolomics sequencing of fecal samples, the typical requirements are greater than 50 mg for solid samples and greater than 100 μL for liquid samples.

Solid samples

• 17.Weigh 100 mg of each solid sample for metabolite extraction. Introduce 1000 μL of an extraction solution containing internal standards (methanol/acetonitrile/water, 2:2:1, with an internal standard concentration of 20 mg/L) and vortex for 30 s.
• 18.Incorporate steel beads and process the mixture using a 45 Hz grinder for 10 min, then apply ultrasonic treatment for an additional 10 min in an ice water bath.

Liquid samples

• 19.Transfer 200 μL of the liquid sample into an EP tube.
• 20.Introduce 500 μL of extraction solution with internal standards (methanol/acetonitrile, 1:1 ratio, internal standard concentration of 20 mg/L) and vortex for 30 s to mix.
• 21.Include steel beads and treat the mixture using a 45 Hz grinder for 10 min.
• 22.Subject the mixture to ultrasonic treatment for another 10 min in an ice water bath.

CRITICAL: As clinical samples are usually collected at different points in time, the method of storage can affect the metabolites.¹⁷ It is often recommended that specimens are stored in a −20°C refrigerator immediately after collection and transferred to a −80°C refrigerator for freezing on the same day. For metabolomics sequencing, frozen stool is usually recommended for no more than 18 weeks. At the same time, ensuring uniformity in sample composition is critical to obtaining reproducible results. For solid fecal samples, proper homogenization before weighing is essential to prevent variation in metabolite extraction. For liquid samples, mix well to avoid stratification or sedimentation. Additional, solid fecal samples require longer or more intense grinding (45 Hz for 10 min), while liquid samples may not need as long. Ensure the sample temperature remains low during sonication to prevent heat-sensitive metabolites from degrading. Keeping samples at low temperatures during extraction (e.g., using ice water baths for sonication and −20°C for storage) is essential to prevent metabolite degradation. Prolonged exposure to higher temperatures may lead to loss of volatile or thermally unstable metabolites.

Pause point: The mixture can be stored at −80°C for two months.

• 23.
  Common Steps.

  • a.
    Store the samples at −20°C for 1 h and then centrifuge at 13,400 × g for 15 min.
  • b.
    Gently collect 500 μL of the supernatant. Evaporate the extract using a vacuum concentrator.
  • c.
    Add 160 μL of extraction solution (a 1:1 mixture of acetonitrile and water) to reconstitute the dried metabolites.
• 24.Vortex for 30 s, then subject to ultrasonic treatment for 10 min (ice water bath).
• 25.Centrifuge at 4°C for 15 min at 13,400 × g.
• 26.Gently pipette 120 μL of the supernatant into a 2 mL injection vial. Combine 10 μL from each sample to prepare a QC sample for analysis.

Pause point: The samples can be stored at −20°C for up to 48 h.

Part 5: LC-MS/MS for non-targeted metabolomics sequencing

Timing: Approximately 20 h for 24 samples

This section describes the metabolomics sequencing conditions, which were performed at Biomarker Biotech (www.biomarker.com) in Beijing, China.

• 27.Conduct the analysis using the Waters Acquity I-Class PLUS coupled with the Xevo G2-XS QTof (Waters Corporation, Shanghai, China).

CRITICAL: Calibration of accuracy is performed every 20 samples during the collection process, and a QC sample is examined for every 10 samples to verify the stability of the LC–MS system.

• 28.Inject 2 μL of each sample, maintaining a flow rate of 400 μL/min, using a mobile phase consisting of solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile) as described in Table 4.
• 29.
  Mass Spectrometry Conditions.

  • a.
    Low collision energy: 2 V.
  • b.
    High collision energy range: 10–40 V.
  • c.
    Scanning frequency: 0.2 s per spectrum.
• 30.
  ESI Ion Source Parameters:

  • a.
    Capillary voltage: 2500 V (positive ion mode) or −2000 V (negative ion mode), cone voltage is 30 V.

    Note: Metabolite identification in different ionic modes (troubleshooting 3).

  • b.
    High collision energy range: 10–40 V.
  • c.
    Ion source temperature and desolvation gas temperature are 100°C and 500°C.
  • d.
    Backing gas flow Rate: 50 L/h; desolvation gas flow Rate: 800 L/h.
  • e.
    Mass-to-charge (m/z) range: 50–1200.

Table 4.

LC gradient elution conditions

Time (min)	Flow rate (mL/min)	Solvent A (%)	Solvent B (%)
0	0.4	95	5
0.5	0.4	95	5
5.5	0.4	50	50
9	0.4	5	95
10.5	0.4	5	95
12	0.4	95	5

Part 6: ELISA for cerebrospinal fluid and plasma sample

Timing: Approximately 3 h for 30 samples

This section describes the application of Human Linoleic Acid ELISA kits (MEIAO Biotech, MO-P38041R) for the determination of plasma and CSF concentrations of target metabolites following the manufacturer’s instructions.

• 31.Thaw plasma and CSF samples on ice at 4°C for testing, and equilibrate the precoated microtiter wells at 25°C for 20 min.
• 32.Prepare the standard and sample wells and add 50 μL of different concentrations of standards to each standard well.
• 33.Introduce 10 μL of sample to the sample wells, followed by 40 μL of sample diluent; do not add to the blank wells.
• 34.Dispense 100 μL of HRP-labeled detection antibody into each well containing standards and samples, including the blank wells.
• 35.Seal the reaction wells with a plate sealing membrane and incubate at 37°C for 60 min in a water bath or thermostat.
• 36.After incubation, remove the liquid and gently blot the wells with absorbent paper.
• 37.Introduce washing solution to each well and let it sit for 1 min. Discard the solution, blot the wells again, and repeat this washing step five times.
• 38.Introduce 50 μL of each substrate A and B to each well. Incubate the plate at 37°C for 15 min.
• 39.Introduce 50 μL of termination solution to each well.
• 40.Within 15 min, record the optical density (OD) at 450 nm for each well.

Note: See troubleshooting 4 if the OD value of the sample to be measured is too high or too low.

• 41.Calculate the concentration of each sample using the linear regression curve of the standards.

Part 7: Combined multi-omics analysis

Timing: Approximately 1–2 months

This section describes the process for integrating microbiome and metabolomics data using Procrustes analysis, O2PLS and correlation analysis, while also integrating multi-omics data with patients' clinical information and IA-related characteristics. All analyses were conducted on the BioCloud Platform (www.biocloud.net).

• 42.
  Procrustes Analysis.

  • a.
    Use the vegan package with the Bayesian distance algorithm to compute the microbial quantitative matrix.
  • b.
    Preprocess the microbiome and metabolomics data using UV scaling.
  • c.
    Compute the metabolite quantitative matrix using the Euclidean distance algorithm.
  • d.
    Perform PCoA for distance sorting and extract the feature coordinates of the microbiome and metabolomics.
  • e.
    Extract the feature coordinates from the PCoA results of the microbiome and metabolomics.
  • f.
    Perform Procrustes analysis following the package documentation (https://rdrr.io/cran/vegan/man/procrustes.html) (troubleshooting 5).

Note: Given the complexity of clinical samples, we consider that an M² value less than 0.3 and vector residuals less than 0.5 indicate a high degree of consistency and low discrepancy between the two datasets.

• 43.
  Two-way orthogonal partial least squares (O2PLS).

  • a.
    Build O2PLS models for statistical analysis with the OmicsPLS package (https://cran.r-project.org/web/packages/OmicsPLS/index.html).
  • b.
    Calculate the scores for each sample to generate a joint score plot.
  • c.
    Select the top 15 metabolites/microbes (based on the loading values in the first two dimensions, representing the strongest associations).
  • d.
    Visualize the selected metabolites/microbes in a bar plot.
• 44.
  Correlation analysis.

  • a.
    Collect basic information and intracranial aneurysm characteristics from patients with intracranial aneurysms and their family members in the control group.

    Note: For the basic information of patients, we focused on factors such as age, gender, blood pressure, blood lipid levels, and diabetes history. The characteristics of the intracranial aneurysm are of greater importance, with key factors including its location, size, aspect ratio, shape, and whether it is located at an arterial bifurcation. Additionally, for patients with ruptured intracranial aneurysms, it is crucial to record the Glasgow Coma Scale (GCS) score, Hunt-Hess grade, and the presence of neurological dysfunction.

  • b.
    Use Spearman’s correlation coefficient, ranging from −1 to 1, to assess the strength and direction of the linear relationship between microbiome and metabolome data.

    Note: Depending on the experimental objectives and actual conditions, an appropriate threshold is set to identify significant correlations. For multi-omics data correlation analysis, a correlation coefficient greater than 0.4 and a p value less than 0.05 are considered suitable. For the correlation analysis with clinical data, a correlation coefficient greater than 0.3 and a p value less than 0.05 are deemed appropriate.

  • c.
    Use heatmaps or network graphs to visualize significant relationships (troubleshooting 6).
  • d.
    Visualize the selected metabolites/microbes in a bar plot.

Expected outcomes

Microbiomics sequencing data processing and analysis

Data processing

In this study, full-length 16S rRNA genes were sequenced on the PacBio Sequel platform, and the downstream data obtained were in BAM format. The CCS (Circular Consensus Sequence) file was exported using SMRT Link software (v.11.1), and LIMA software (v.1.7.0) was employed to identify barcode sequences, which allowed for the separation of CCS sequences from different samples. Ultimately, the CCSs were converted into FASTQ format for subsequent quality assessment and analyses. Subsequently, the unprocessed CCS sequences were filtered using Cutadapt software (version 1.9.1) to identify and remove any primer sequences.⁷ Furthermore, length filtering was employed to generate clean CCS sequences (Clean-CCS), free of any primer contamination. The process of removing chimeras and filtering out reference sequences was conducted using UCHIME (v.4.2) and BLASTN (v.2.9.0+) software, resulting in the acquisition of effective data (Effective-CCS).⁸

Data quality

The data quality was evaluated through the statistical processing of the number of sequences present in each sample. The data were evaluated primarily based on parameters such as the number of sequences, sequence length, and so forth. Typically, the average sequence length (avglen) should range from 1400 to 1500 bp, with a raw CCS count exceeding 10,000. The effective CCS ratio should fall between 60% and 95%, while the majority of effective CCS should be around 1400 to 1600 bp. In Figure 2B, one sample from each group was selected to show the sequence length distribution graphs. The pertinent indicators are presented in Table 5.

Table 5.

Sequencing quality metrics and reference

Metric	Reference range
Sequence count
Raw CCS	Greater than 10,000
Clean CCS Ratio	Greater than 99% of raw CCS
Effective CCS Ratio	60%–95% of raw CCS
Sequence length
AvgLen	1400-1500 bp
Sequence Length Distribution	Majority of effective CCS between 1400-1600 bp
Effective Sequence Ratio	60%–95%
Chimera detection
Chimera Sequence Proportion	Less than 5%

Operational taxonomic unit (OTU) and diversity analysis

The obtained CCSs were clustered using Usearch software at the 97.0% similarity level to obtain OTUs.¹⁸ Subsequently, the abundance of individual OTUs in the samples was calculated, and the information for each sample OTU was presented in the form of a histogram (Figure 3A).

Figure 3.

Operational taxonomic unit (OTU) and diversity analysis in microbiome

(A) Each sample OTU information.

(B and C) Alpha diversity and Shannon index in microbiome. Data are represented as mean ± SEM. p value, ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. Reprinted with permission from Xu et al., 2024.¹

(D) Beta diversity in microbiome. Reprinted with permission from Xu et al., 2024.¹

Alpha diversity of subgroups was assessed using QIIME2 software.¹⁰ Species richness was assessed using the Chao1 and Ace indices, whereas the Shannon and Simpson indices were applied to evaluate species diversity. Additionally, the coverage value was noted, with a threshold greater than 0.99 considered appropriate. Concurrently, the Shannon index was calculated by Mothur software (v.1.35.1) for each group at disparate sequencing depths to assess the adequacy of the sequencing data volume. A flat dilution curve indicates that the sequencing data is sufficiently large and rich in feature types. Figures 3B and 3C illustrates the alpha diversity and Shannon index from our previous study.

Similarly, beta diversity analyses were conducted using QIIME software for the purpose of comparing the degree of similarity in species diversity between the various groups. In analyzing fecal samples, the unweighted (Jaccard and Unweighted Unifrac) algorithm is recommended for calculating beta values between samples.^19,20 Due to the complexity and diversity of the microorganisms themselves, as well as the more drastic differences in the composition of different species, this is the most appropriate approach. Typically, we performed this through principal component analysis (PCA) as well as non-metric multi-dimensional scaling (NMDS), whereby the distance between sample points is used to assess similarity. The application of NMDS is believed to be more appropriate for the given analyses, as it better reflects the nonlinear structure inherent to ecological data (Figure 3D). PERMANOVA is employed to assess both between-group and within-group variations.¹¹ An R-value approaching 1 indicates that the discrepancies between groups are more pronounced than those within groups.

Species identification and differential gut microbiota analysis

The obtained CCS sequences were compared with those in the SILVA database to determine the corresponding species taxonomic information for each feature.²¹ Species abundance tables at different taxonomic levels were generated using QIIME software, and community structure maps were constructed to visualize the corresponding data. For presentation purposes, it is generally recommended to focus on the phylum level, as this effectively illustrates the distributions of major microbial groups (Figure 4A).

Figure 4.

Species identification and differential gut microbiota analysis

(A) Species composition of the gut microbiota in groups at the phylum level. Reprinted with permission from Xu et al., 2024.¹

(B) LEfSe analysis for biomarker species. Reprinted with permission from Xu et al., 2024.¹

For the analysis of significantly different species in multiple groups, we recommend the Analysis of Variance (ANOVA) or the Kruskal-Wallis rank-sum test (no fewer than 5 samples per group). In addition, to investigate the differences between two groups, a t-test can be conducted on the species abundance data using Metastats software.¹² For key divergent species and biomarker discovery, screening was conducted using LEfSe software with a threshold of linear discriminant analysis (LDA) > 3 (Figure 4B).¹³ Regarding the construction of a diagnostic model using the receiver operating characteristic (ROC) curve, random forest analysis (based on Gini values) combined with Lasso regression was employed to screen for candidate differential species. Subsequently, ROC analysis was performed using the pROC package (v.1.18.0).

Functional gene prediction analyses

Functional gene prediction analyses were conducted using PICRUSt2 software, which annotated the feature sequences against an existing phylogenetic tree.^14,15 To evaluate the functional abundance among different samples, statistical tests were performed using STAMP software.¹⁵ Pairwise T-tests were conducted between different groups, with a significance threshold set at a p-value of 0.05.

Additionally, the composition and differential analysis of KEGG metabolic pathways were performed to observe variations in functional genes associated with metabolic pathways among different sample groups. BugBase was also utilized to predict functional pathways at the biological level within complex microbiomes.²²

Metabolomics data processing and analysis

Data processing

Raw data collected using MassLynx software (v.4.2) were subjected to initial processing using Progenesis QI software, which facilitated peak extraction, peak alignment, and normalization of the data. For compound identification, we employed multiple online databases, including the METLIN, Npatlas, and YMDB databases. The identification process involved comparing the extracted peaks against these databases to ascertain the presence of known compounds. To ensure the accuracy of the identifications, we established stringent criteria for mass deviations: a mass deviation of 100 ppm for the parent ion and 50 ppm for the fragment ion. These thresholds were set to minimize the occurrence of false positives and enhance the reliability of the identified compounds.²³ The final output comprised a comprehensive list of identified compounds along with their corresponding abundances, which were subsequently utilized for further statistical analyses and interpretation of the results.

Relevance assessment

The spearman rank correlation was employed to assess the correlation of biological repeats. Additionally, PCA was performed to gain an initial insight into the general metabolic variations across the samples in each group and to assess the extent of variability among the samples within the groups. (Figure 5A). For comparisons between two groups, orthogonal projections to latent structures-discriminant analysis (OPLS-DA) are commonly employed (Figure 5B).²⁴ The R²Y and Q²Y values represent the model’s ability to explain the X and Y matrices, respectively; values closer to 1 signify a more stable and reliable model. Generally, a Q²Y value greater than 0.5 is considered indicative of a valid model, while a Q²Y greater than 0.9 is regarded as excellent (Figure 5C). The application of both PCA and PLS-DA is recommended. PCA is primarily used to detect outlier samples and to assess batch effects due to experimental covariates to facilitate the exclusion of outlier data. The application of PLS-DA enables a more precise delineation of the distinctions between groups.

Figure 5.

Metabolomics data analysis

(A) PCA to assess the extent of variability among the samples. Reprinted with permission from Xu et al., 2024.¹

(B and C) OPLS-DA for comparisons between two groups. Reprinted with permission from Xu et al., 2024.¹

Differential metabolite identification and KEGG functional enrichment analysis

Based on the OPLS-DA results, the Variable Importance in Projection (VIP) derived from the multivariate analysis was combined with the p-value and the fold change (FC) from the univariate analysis to further identify differential metabolites. In general, the following criteria were used for screening: |fold change (FC)| > 1, p-value < 0.05, and VIP > 1. These thresholds were considered standard for identifying metabolites that exhibit significant differences across groups. To illustrate the differential metabolites and those shared between groups, both volcano plots and Venn diagrams were employed. Functional enrichment analyses were conducted to assess the enrichment of known biological functions or processes within the list of differentially expressed metabolites using over-representation analysis (ORA) in conjunction with the KEGG database. The p-values for the list of differentially expressed metabolites were calculated using a hypergeometric distribution, according to the formula:

$$p=1-\frac{m-1}{N}\sum _{i=0}^M\frac{\left(\begin{array}{l}M\\ i\end{array}\right)\left(\begin{array}{l}N-M\\ n-i\end{array}\right)}{\left(\begin{array}{l}N\\ n\end{array}\right)}$$

Limitations

While profiling the human gut microbiota from fecal samples is recognized as a non-invasive procedure, it has limitations in accurately reflecting the entire gastrointestinal flora. Under ethically appropriate conditions, biopsies of the intestinal mucosa can provide a more representative profile of the gut microbiota. Concurrently, a range of cerebrovascular disorders, such as cerebral hemorrhage, stroke, and ruptured intracranial aneurysms, can lead to the sudden onset of coma, hypotension, and other severe symptoms in affected individuals. The emergency state resulting from these conditions is likely to induce alterations in gut microbiota and metabolite. Therefore, certain species and metabolites may be altered in ways that are unrelated to the development of the disease, making them difficult to differentiate. Moreover, the differential biomarkers identified lack specificity and may be significantly dysregulated across a variety of diseases. For the same disease, study results can vary significantly across different regions, necessitating comprehensive analyses with large sample sizes from multiple centers to draw realistic and thorough conclusions. For sample collection, as most clinical samples are obtained over a long period of time (ranging from months to years). As a result, there are limitations to the preservation of fecal samples, and the duration of storage at −20°C or −80°C can significantly affect gut microbiota composition and metabolite levels.

Troubleshooting

Problem 1

Extracted DNA concentration is too low.

Potential solution

The following aspects can effectively increase the efficiency of DNA extraction: 1. Extension of the incubation time for the elution step and heating of the elution buffer to 70°C prior to elution. 2. Add less elution buffer, e.g., 30 μL. 3. Increase in inputs of fecal samples. The usual recommended input fecal sample size should be 250 mg, with a minimum of 50 mg.

Problem 2

The division of collected samples into discovery and validation cohorts is a common recommendation for interactive analysis and validation. However, when sequencing results are compared, poor homogeneity and characteristic differences are often observed. This phenomenon, particularly evident in microbiomics, is attributed to the influence of various factors, including dietary habits, geographical locations, age groups, and others, on the gut microbiota.

Potential solution

The technical challenges associated with batch differences remain a significant obstacle in 16S rDNA sequencing and LC–MS. Clinical samples are more susceptible to change and variability than animal models. Accordingly, rigorous control of variables in the selection of samples is essential. It is recommended that same-sex, age-matched family members of patients living together be used as a normal control. We recommend that the sample size of each group be increased as much as possible (>20 cases) when laboratory conditions and funding allow. For the extraction of microbes and metabolites, stable temperature, extraction time and laboratory space are required to ensure accuracy.

Problem 3

Typically, metabolites are obtained in both positive and negative ion modes, but certain metabolites can only be identified in one of the modes. In addition, the molecular weight of the same metabolite may be significantly different in the positive and negative ion modes.

Potential solution

Each metabolite is identified by the mass-to-charge ratio (m/z) of the adducts and we normally recommend the simultaneous detection of both ionic modes, which are ultimately analyzed together for more accurate results.

Problem 4

Abnormalities in OD values were mainly due to inappropriate sample dilutions, resulting in sample concentrations that were too high or too low and outside the linear range of the standard curve (step 31). At the same time, the reaction time is too long or too short will also have an effect on the OD value.

Potential solution

For different metabolites, plasma and cerebrospinal fluid concentrations differ, mainly owing to the existence of the blood-brain barrier. A 5-fold dilution is usually recommended for the assay on most metabolites. For certain substances present in low concentrations in plasma and CSF, such as most inflammatory factors, a 1:1 dilution is usually recommended. On the other hand, for substances present in high concentrations, such as cholesterol and glucose, a 1:9 dilution is usually recommended. In addition, a reaction time of 10–30 min is recommended.

Problem 5

There may be inconsistencies in the data between the microbiome and the metabolome. Microbiome and metabolome data are inherently different in terms of their biological origin, measurement techniques, and data formats. For instance, microbiome data often have a compositional structure and are subject to sequencing depth variability, whereas metabolomic data are quantitative and influenced by batch effects. This inconsistency can lead to difficulties in aligning datasets for joint analysis.

Potential solution

Preprocessing steps, such as compositional transformation (e.g., centered log-ratio transformation for microbiome data) and normalization (e.g., UV scaling for metabolomic data), are employed to standardize data formats. Additionally, dimensionality reduction methods like principal coordinate analysis (PCoA) provide a common framework for comparing features. More importantly, ensure that samples of the microbiome and metabolites originate from the same fecal sample and are sequenced simultaneously at the same point in time. This avoids the inherent differences between sample sources and the degradation of DNA and metabolites over time.

Problem 6

Correlations identified between microbial and metabolic features do not inherently imply causative relationships, complicating biological interpretation.

Potential solution

Firstly, it is crucial to set thresholds for significance in correlation analyses. Also more essential is that findings require validation by experimental methods, such as gut microbiota studies for IA patients, where identified differential metabolites should be validated in in vivo or in vitro experiments to confirm causality. In vivo experiments are mainly performed by constructing mouse models of IAs, whereas in vitro is mainly focused on vascular-related cell lines for comprehensive analyses of metabolite processing.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xiang Li (li.xiang@whu.edu.cn).

Technical contact

Technical questions on executing this protocol should be directed to and will be answered by the technical contact, Hongyu Xu (xuhongyu@whu.edu.cn).

Materials availability

This study did not generate new unique reagents.

Data and code availability

The published article (https://doi.org/10.1016/j.isci.2024.111184) includes all datasets generated or analyzed during this study.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (82271556 and 82171517). The graphical abstract and figures in this research were created using Biorender.com. We acknowledge the technical assistance provided by Beijing Biomarker Technologies Co. during the preparation of this article.

Author contributions

Conceptualization, X.L. and W.W.; writing – original draft, H.X., J.L., and Z.J.; writing – review and editing, H.X., R.L., and W.W.; funding acquisition, W.W. and X.L.; supervision, W.W. and X.L.

Declaration of interests

The authors declare no competing interests.