1D ¹H NMR as a Tool for Fecal Metabolomics

Abstract

Metabolomic studies allow a deeper understanding of the processes of a given ecological community than nucleic acid–based surveys alone. In the case of the gut microbiota, a metabolic profile of, for example, a fecal sample provides details about the function and interactions within the distal region of the gastrointestinal tract, and such a profile can be generated in a number of different ways. This unit elaborates on the use of 1D ¹H NMR spectroscopy as a commonly used method to characterize small-molecule metabolites of the fecal metabonome (meta-metabolome). We describe a set of protocols for the preparation of fecal water extraction, storage, scanning, measurement of pH, and spectral processing and analysis. We also compare the effects of various sample storage conditions for processed and unprocessed samples to provide a framework for comprehensive analysis of small molecules from stool-derived samples.

Basic Protocol 1:

Extracting fecal water from crude fecal samples

Alternate Protocol 1:

Extracting fecal water from small crude fecal samples

Basic Protocol 2:

Acquiring NMR spectra of metabolite samples

Alternate Protocol 2:

Acquiring NMR spectra of metabolite samples using Bruker spectrometer running TopSpin 3.x

Alternate Protocol 3:

Acquiring NMR spectra of metabolite samples by semiautomated process

Basic Protocol 3:

Measuring sample pH

Support Protocol 1:

Cleaning NMR tubes

Basic Protocol 4:

Processing raw spectra data

Basic Protocol 5:

Profiling spectra

Support Protocol 2:

Spectral profiling of sugars and other complex metabolites

INTRODUCTION

The gut microbiota is the ecosystem of microorganisms—bacteria, fungi, archaea, protists, and viruses—residing along the gastrointestinal tract (Kho & Lal, 2018). This complex ecosystem is known to be metabolically active and is widely associated with many aspects of mammalian health and disease (Weir et al., 2013). As a result, efforts are being made to further understand the functional output of the ecosystem to provide greater depth of knowledge than can be gained from determining taxonomic composition alone. The net metabolic output of the gut microbiota is reflected in the stool and encompasses both microbe-microbe interactions and potential interactions between the gut microbiota and its host (Yen et al., 2015). Host stool samples are an excellent source of gut microbial metabolites (Jacobs et al., 2008; Lamichhane et al., 2015) and can be obtained without any invasive procedures; thus, stool represents the most commonly used biospecimen type for the study of the gut microbiota (particularly in health). Gut microbial metabolites of importance include short-chain fatty acids (SCFAs), alcohols, amines, sulfur compounds, phenols, and indoles (Oliphant & Allen-Vercoe, 2019).

One of the most popular techniques for analyzing human and animal fecal samples is one-dimensional proton nuclear magnetic resonance (1D ¹H NMR) spectroscopy (henceforth shortened to NMR). Compared to mass spectrometry (MS), NMR requires less sample handling and preparation, and also is directly quantitative (Emwas, 2015). However, NMR is best used as a complementary approach to MS for the study of stool samples, because NMR and MS together cover a broad range of biologically important molecules (Wissenbach et al., 2016). For stool metabonomics, NMR is often used as a screening tool because, compared to MS, NMR is quicker, simpler, and cheaper to carry out, if the infrastructure is available. The use of an internal standard in NMR also allows quantification (Weljie, Newton, Mercier, Carlson, & Slupsky, 2006). Although the repertoire of detectable compounds is smaller for NMR as for MS, NMR can detect ~80 small-molecule metabolites in typical fecal samples (Weljie et al., 2006; Yen et al., 2015), which adequately resolves broad clinical and research questions, despite trade-offs in sensitivity and selectivity in NMR (Wissenbach et al., 2016). The following series of protocols is presented as a guide for NMR-based analysis of fecal samples, and is aimed at laboratories that would like to gain experience in the methods involved. Although the methods have been specifically developed for use with human fecal samples, they are also applicable to fecal samples obtained from other sources (e.g., murine models).

Basic Protocol 1 and Alternate Protocol 1 describe how to extract fecal water from large (≥250 mg) and small (<250 mg) crude fecal samples, respectively. Basic Protocol 2 describes the acquisition of NMR data from prepared fecal water samples for subsequent profiling using Chenomx NMR Suite software. Basic Protocol 3 describes how to measure the pH of prepared fecal water samples. Support Protocol 1 describes how to properly clean NMR tubes used in these protocols. Alternate Protocol 2 describes the aquisition of NMR data from prepared fecal water samples, specifically using Bruker Spectrometer Running TopSpin 3.x software. Alternate Protocol 3 describes how to semiautomate Alternate Protocol 3. Basic Protocol 4 describes how to process acquired fecal water sample data with a focus on Chenomx NMR Suite Software, version 7.7 or later (Chenomx Inc., Alberta, Canada). Basic Protocol 5 describes how to profile peaks from an acquired spectrum using Chenomx NMR Suite software, and Support Protocol 2 details profiling of sugar peaks specifically.

STRATEGIC PLANNING

Human fecal samples can generally be obtained from donors or patients as required, and ideally should be placed on ice as soon as possible after voiding, and processed within 24 hr of collection. We recommend sample collection using a “toilet hat” type of collection device, designed for this purpose, for example Fisherbrand cat. no. 02–544-208. If immediate processing is not practical, samples can be stored frozen at −80°C for several months, although all samples for a given experiment should be treated similarly. Ethical oversight is appropriate for the collection of human samples and should be sought through relevant institutional review boards before the start of the experimental work. Fecal samples collected from animals should take into account time of voiding as far as possible, and should be obtained using appropriate animal care guidelines under institutional oversight.

BASIC PROTOCOL 1

EXTRACTING FECAL WATER FROM CRUDE FECAL SAMPLES

The stool of an average healthy adult human contains between 63% and 85% water, with the remaining dry matter consisting of 25%−54% bacterial biomass, colonic epithelial cells, undigested food stuffs, various macromolecules, and many metabolites (Rose, Parker, Jefferson, & Cartmell, 2015). Fecal water (i.e., the liquid component extracted from feces) is rich in gut microbial metabolites and is often used experimentally as a starting point for the analysis of such metabolites (Yen, Bolte, Aucoin, & Allen-Vercoe, 2018). To extract fecal water, a given fecal sample must first be homogenized using a solvent that will maintain sample pH, as changes in pH can alter the ionic charge of compounds present in the sample, resulting in changes to the acquired spectrum. Phosphate-buffered saline (PBS, pH 7.35 ± 0.1) is regarded by many as a suitable solvent in fecal water extractions because it maintains physiological pH and helps maintain stability of metabolites (Deda et al., 2017). However, this typically only holds true for samples of mass ≥250 mg. When sample mass is <250 mg, deuterated water is recommended as the solvent in order to reduce the amount of water in the final fecal water extract (as described in Basic Protocol 2). In addition to the choice of solvent, the volume of solvent should be carefully considered, as increasing the solvent volume results in dilution of metabolite concentrations in the final fecal water extract. For samples of mass ≥250 mg, a 25% weight of fecal sample per volume (w/v) of buffer/fecal slurry is recommended to yield satisfactory metabolite profiles (Yen et al., 2018). In general, the amount of solvent used to homogenize the stool samples should be kept to a minimum to avoid diluting metabolites beyond the limit of detection by NMR. Next, fecal slurries must be centrifuged and filtered to remove all dense matter that would otherwise disturb sample acquisition. Direct filtration will typically clog the filter, and so centrifugation is generally preferred and allows a primary separation. However, the effects of centrifugation on the sample metabolites should be noted. To this effect, a study by Yen et al. demonstrated the effects of ultracentrifugation on fecal samples and concluded that metabolite profiles in fecal water decrease in a linear manner by <5% per 10,000 × g increase (Yen et al., 2018). However, this metabolite decrease was not observed for p-cresol, a potentially important biomarker in various human diseases (Yen et al., 2018); instead, p-cresol increased by ~50% per 10,000 rpm increase (Yen et al., 2018). It is therefore advisable to be cautious when interpreting this compound’s concentration in cases where a given sample has been subjected to large centrifugal force. Optimization studies have also shown that centrifugation runs under 30 min at 14,000 × g are adequate at extracting metabolites from resuspended fecal samples (Gratton et al., 2016; Monleón et al., 2009). This further minimizes sample handling for NMR scanning. Overall, a lower centrifugation speed of 14,000 × g for 10 min is usually sufficient for fecal water extraction.

Once centrifugation is complete, the supernatant (i.e., fecal water) is prepared for sample acquisition. To account for any chemical shifts in the fecal water, a reference standard, also known as a chemical shape indicator (CSI), should be added to the fecal water solution. 3-(Trimethylsilyl)-1-propanesulfonic acid sodium salt (DSS) is commonly used, added to each sample to a concentration of 0.5 mM (Yen et al., 2018). DSS produces four NMR signals: the CSI reference peak at 0 ppm, and three other peaks that are not diagnostically useful and may overlap with genuine metabolite peaks. A commercially available partially deuterated analogue, DSS-d6, produces only the 0-ppm reference peak. Furthermore, deuterated water (D₂O) must be added to a final concentration of 10% (v/v) or higher for correct operation of the NMR spectrometer, and sodium azide may be added to a concentration of 0.2% (w/v) for stability (Gretan et al., 2016; Wu, An., Yao, Wang, & Tang, 2010). The following protocol describes fecal water extraction for fecal samples ≥250 mg using DSS-d6 as a CSI.

Materials

Fecal sample of interest (≥250 mg)

1 × PBS (see recipe)

Benchtop centrifuge with refrigeration capability: e.g., Microliter 30 × 2 ml Fixed Angle Rotor (ThermoFisher Scientific, cat. no. 75003652), or similar

2-ml screwcap tubes with O-ring caps (for samples 250–500 mg; Fisher Scientific, cat. no. 12222621) or 15-ml conical tubes (for samples 500 mg–3.75 g; Fisher Scientific, cat. no. 0553912)

1.7-ml microcentrifuge tubes

Disposable Luer-Lok syringes (BD, cat. no. B309646)

~5 mM DSS-d6 in deuterated water (D₂O), containing 0.1% sodium azide (Chenomx Inc., Edmonton Canada, IS-2 Chenomx Internal Standard—DSS-d6)

0.2-μm-pore-size polyethersulfone (PES) membrane syringe filters, 25 mm (Fisher, cat. no 13100106)

0.8-μm- and 0.45-μm-pore-size PES filters (optional)

Analytical balance

1. Aseptically transfer an appropriate mass of fecal sample to a screwcap conical test tube to make up a 25% (w/v) fecal slurry in 1 × PBS.

2. Vortex the fecal slurry for 10 min on a high setting or until the slurry appears homogenized.

   In order to get an accurate representation of the microbial metabolites, the fecal sample must be homogenized in the solvent buffer.

3. If more than one sample is being prepared, repeat steps 1 and 2 for each sample.

4. Set a benchtop centrifuge to 4°C. Once cooled, centrifuge the fecal slurry/slurries at 14,000 × g for 10 min at 4°C.

5. Using a pipette, carefully transfer the supernatant from each tube into a fresh, appropriately sized tube using aseptic technique.

   The supernatant contains the gut microbial metabolites (i.e., this is the fecal water.). The pellet contains microbial biomass and other dense material from the fecal sample.

   In general, stool samples should be considered as biohazardous. It is recommended that steps 1–5 be performed in a biosafety cabinet, and that appropriate steps be taken to handle and dispose of biohazardous material.

6. Filter-sterilize the supernatant through a 0.2-μm polyethersulfone filter using a disposable Luer-Lok syringe. First, remove the plunger from the syringe barrel, screw the filter onto the barrel, place the assembled filter over an open, appropriately sized tube, and carefully pipette the supernatant into the syringe barrel. Replace the plunger, and steadily push the supernatant through the filter. If there is too much resistance using when a 0.2-μm PES filter, a serial filtration may need to be performed (i.e., filter the sample sequentially through a 0.8-μm PES filter, followed by a 0.45-μm PES filter and finally a 0.2-μm PES filter). Repeat this step for each sample.

   The final volume of fecal water extracted will be considerably reduced from the original volume of fecal slurry (up to 50%). To minimize this lost volume, use a micropipette to aseptically remove any volume from the bottom of the filter end. A minimum of 540 μl of filtrate is required for scanning with 5-mm NMR tubes, as in Basic Protocol 2.

7. Prepare 60-μl aliquots of 5 mM DSS-d6 stock solution (in D₂O) in 1.7-ml microcentrifuge tubes for each NMR sample.

   The DSS concentration need not be exactly 5 mM, but should be close to this value and precisely known. Commercially prepared stock solutions typically come with a certificate of analysis stating the DSS concentration to three significant figures.

8. Transfer 540 μl each of filtered supernatant into the 1.7-ml microcentrifuge tube containing DSS-d6 stock solution.

9. Mix the samples thoroughly by pipetting up and down.

   600 μl is the minimum volume that can be acquired using a 5-mm NMR tube (see Basic Protocol 2). If scanning volumes >600 μl, the DSS-d6 stock solution must be maintained at a 10% (v/v) concentration.

10. Store samples at 4°C, protected from light (see Alternate Protocol 1, step 10, for more details about storage), until they are ready to be scanned (Basic Protocol 2 or Alternate Protocol 2 or 3).

ALTERNATE PROTOCOL 1

EXTRACTING FECAL WATER FROM SMALL CRUDE FECAL SAMPLES

For a sample with a mass of <250 mg or in which the metabolite concentrations are expected to be very low (e.g., mouse pellets), the following protocol is recommended for fecal water extraction. The greatest obstacle with these small or dilute samples is obtaining adequate water suppression by the NMR instrument. Therefore, the use of a buffer such as PBS is not acceptable, as the water content in the sample would then be too high. Instead, deuterium oxide (D₂O) is added as the solvent to create a fecal slurry. These samples should also be diluted to a 25% weight of fecal sample per volume of D₂O fecal slurry; however, this may not always be possible as very small fecal samples may need to be diluted further in order to yield an adequate fecal water volume. The minimum volume of filter-sterilized and prepared fecal water that can analyzed by NMR is 200 μl. Thus, 180 μl of filter-sterilized fecal water is required to maintain a 10% DSS-d6 (v/v) and fecal water solution. This alternate protocol explains how to extract fecal water from fecal samples that are <250 mg or contain a low concentration of metabolites.

Materials

Fecal sample of interest (<250 mg)

99.9% deuterated water (D₂O; Sigma Aldrich, cat. no. 151882)

~5 mM DSS-d6 in deuterated water (D₂O), containing 0.1% sodium azide (Chenomx Inc., Edmonton Canada, IS-2 Chenomx Internal Standard—DSS-d6)

Benchtop centrifuge with refrigeration capability

2-ml screwcap tubes with O-ring caps (Fisher Scientific, cat. no. 12222621)

1.7-ml microcentrifuge tubes

0.2-μm-pore-size PES membrane syringe filters, 25 mm (Fisher Scientific, cat. no 13100106)

5-ml disposable Luer-Lok syringe (BD, cat. no. B309646)

Analytical balance

1. Pre-weigh enough 2-ml screwcap tubes with O-ring caps such that there is one tube ready for each sample.

2. Using aseptic technique, weigh out an appropriate mass of each fecal sample to make up a 25% (w/v) fecal slurry into a 2-ml screwcap tube with an O-ring cap.

   Fecal slurry of 25% (w/v) is the minimum recommended concentration, noting that dilute metabolite concentration is the trade-off for adequate sample volume and ease of syringe filtration.

3. Add the appropriate volume of D₂O to the fecal sample tube to maintain a 25% (w/v) fecal slurry. If this is not possible, dilute the sample with D₂O until at least 180 μl of filter-sterilized fecal water extract is obtained.

   As fecal water extract volume can be lost within a filter during the filter-sterilization step, it is best to overdilute the sample rather than having an inadequate volume to acquire for Basic Protocol 2. It is advisable to filter-sterilize at least 250 μl of the fecal slurry. Thus, the smallest volume of D₂O a sample can be diluted with is 250 μl.

4. Vortex the fecal slurry for 10 min or until the slurry appears homogenized.

   Some fecal samples tend to be dense and dry (e.g., mouse pellets). In these cases, it is recommended that the sample be mixed aseptically and manually with a pipette tip by repeatedly aspirating and dispensing, and pushing the pellet against the side of the tube to break the sample apart. Once this is done, continue with vortexing to homogenize completely.

5. Set a benchtop centrifuge to 4°C. Once cooled, centrifuge fecal slurry 10 min at 14,000 rpm, 4°C.

6. Using a pipette, carefully transfer the supernatant from the tube into a fresh, appropriately sized tube using aseptic technique.

   The supernatant (the fecal water) contains the gut microbial metabolites). The pellet contains microbial biomass and other dense material within the fecal sample.

7. Filter-sterilize the supernatant through a 0.2-μm PES filter using a disposable Luer-Lok syringe. First, remove the plunger from the syringe barrel, screw the filter onto the barrel, place the assembled filter over an open, appropriately sized tube, and carefully pipette the supernatant into the syringe barrel. Replace the plunger, and steadily push the supernatant through the filter. If there is too much resistance using a 0.2 μm PES filter, a serial filtration may need to be performed (i.e., filter the sample sequentially through a 0.8-μm PES filter, followed by a 0.45-μm PES filter and finally a 0.2-μm PES filter). Repeat this step for each sample.

   As these sample volumes are very low, it is best to salvage as much filtrate as possible. To minimize the amount of filtrate lost, use a micropipette to aseptically remove any volume from the bottom of the filter end.

   In general, stool samples should be considered as biohazardous. It is recommended that steps 1–7 be performed in a biosafety cabinet, and that appropriate steps be taken to handle and dispose of biohazardous material.

8. For each sample, transfer a 20-μl aliquot of DSS-d6 stock solution into a labeled microcentrifuge tube, and then transfer a 180-μl aliquot of filter-sterilized fecal water into the respective microcentrifuge tube with DSS-d6. Mix the sample thoroughly by pipetting up and down.

   200 μl is the minimum volume that can be acquired using a 3-mm NMR tube (see Basic Protocol 2).

9. Store samples at 4°C, protected from light, until they are ready to be scanned (Basic Protocol 2 or Alternate Protocol 2 or 3).

It is recommended that crude fecal samples be prepared within 30 days of being collected, as metabolite peaks decrease with time. To demonstrate this, we collected material from bioreactors modeling the conditions of the human colon and supporting the growth of a complex derived fecal community from a healthy human donor (McDonald et al., 2013). 2-ml samples were taken and scanned immediately or after 24 hr, 30 days, and 60 days of storage at −80°C. Two samples were processed and acquired for each storage condition. All steps were followed as explained in Basic Protocol 1, except for the first step, in which 2 ml of bioreactor vessel content was directly centrifuged at 14,000 × g for 10 min at 4°C. Acetate, butyrate, and propionate were profiled as they have the highest concentrations, respectively, in this in vitro bacterial consortium. The results are shown in Figure 1.

Figure 1.

Changes in acetate (red), butyrate (purple), and propionate (blue) concentrations as a result of fecal sample storage. Cryogenic sample storage for 24 hr resulted in a decrease in acetate, butyrate, and propionate concentration by 20%, 17%, and 18%, respectively. There was <3% difference observed for each metabolite when comparing cryogenic storage for 24 hr and 30 days. Cryogenic sample storage for 60 days resulted in the largest decrease in acetate, butyrate, and propionate by 28%, 22%, and 26%, respectively.

BASIC PROTOCOL 2

ACQUIRING NMR SPECTRA OF METABOLITE SAMPLES

The concentration of many metabolites present within a biofluid may be determined from its ¹H NMR spectrum (Sokolenko et al., 2013). Submillimolar metabolite concentrations can be readily detected, with typical throughputs of 2–6 samples per hour. Although numerous NMR experimental procedures (referred to as “pulse sequences”) have been developed for the collection of ¹H NMR spectra from aqueous samples, the METNOESY sequence (Wu, An, Yao, Wang, & Tang, 2010), which is based on the first increment of the ¹H NOESY experiment, provides sufficient suppression of the intense water signal such that minimal spectrum post-processing is required. This protocol describes how to collect ¹H NMR spectra of metabolite samples for subsequent profiling with Chenomx NMR Suite software, an NMR profiling software. Most steps of the protocol can be adapted to other NMR profiling software. The protocol is described at two scales: using 5-mm NMR tubes when using a total sample volume (including DSS and D₂O) of 600 μl, or 3-mm NMR tubes with a sample volume of 200 μl.

NOTE: The sample temperature regulation of the probehead should be periodically calibrated (Findeisen, Brand, & Berger, 2007).

Materials

Prepared metabolite samples to be studied (containing, e.g., short chain fatty acids, organic acids etc.), each containing a known amount of DSS and ≥10% D₂O (Basic Protocol 1 or Alternate Protocol 1)

Deionized water

90% (v/v) acetone

9-in. Extended Tip NMR Pipet (Kimble, cat. no. 897085–0009)

Pipette balloon

500-ml graduated cylinder

NMR tubes and caps: this protocol was developed using 5-mm-diameter, 7-in. (Wilmad, cat. no. 506-PP-7) and 3-mm-diameter, 7-in. NMR tubes (Wilmad, cat. no. 307-PP-7)

Kimwipes

Nuclear magnetic resonance (NMR) spectrometer (an operating field of ≥600 MHz is recommended)

NMR probehead with an “inverse” coil configuration (i.e., ¹H inner coil) and gradient shimming capability, as well as automated tuning and matching, for automation (a cryogenically cooled probehead is recommended but not required)

1. Invert the microcentrifuge tubes containing the samples a minimum of five times. Alternatively, pipette the sample up and down a minimum of five times.

   If scanning a previously prepared frozen sample, allow the sample to thaw slowly by letting it sit at room temperature.

2. Wipe the exteriors of the NMR tubes with a Kimwipe. Carefully examine the NMR tubes under a bright source of light to ensure that they have no smears or scratches.

   If visible defects in the tube, such as scratches, are present, it is best not to use that tube.

3. Slowly transfer 600 μl of the sample to a 5-mm NMR tube (or 200 μl of sample if using a 3-mm NMR tube) using a glass transfer pipette and balloon. Avoid touching the interior sides of the tube with the glass pipette, as this may cause scratches.

   If the sample is not thoroughly mixed before being transferred to the NMR tube, there may be residual concentration gradients that lead to shimming issues and distorted line-shapes. Note that vortexing alone is not sufficiently turbulent to eliminate concentration gradients.

4. Ensure that there are no bubbles present in or at the surface of the solution once the sample is in the NMR tube. Cap the tube.

   1. If the sample will not be acquired immediately, store at 4°C, protected from light, until needed. Be sure to allow samples to return to room temperature immediately before scanning.

   If the sample contains bubbles, gently aspirate air onto the sample using the glass pipette and balloon or gently flick the NMR tube to remove the bubbles. Bubbles distort the spectrum when the sample is being scanned. Samples should be scanned within 24 hr of preparing the tubes, as metabolite concentrations are affected by freezing the sample (Fig. 1).

5. Place used glass pipettes into a 500-ml graduated cylinder containing deionized water such that they are fully submerged. Store in water until they can be cleaned.

   1. To clean the glass pipettes, flush them twice with running deionized water and then once with 90% acetone to aid with water evaporation.

      An empty pipette tip box can be used as a drying rack, setting the pipettes inverted and on an angle.

6. If acquiring more than one sample, keep a record of which sample is in each tube, as well as the associated scan number.

7. Set the sample temperature of the NMR probehead to 298.15 K.

   If the probehead was previously set to a temperature other than 298.15 ± 1 K, wait 30 min before proceeding to minimize thermal gradients inside the probehead.

8. Insert the sample in the spectrometer and wait for the temperature to fully equilibrate across the sample.

   Although the temperature reported by the spectrometer software may stabilize within seconds after sample insertion, it can take 5 min or longer for the temperature to fully equilibrate across the sample volume, especially when using cryogenically cooled probeheads or samples kept on ice immediately before insertion. A lack of thermal equilibration will produce a poor shimming result and will manifest as broader lines in spectra, complicating analysis.

9. Within the NMR spectrometer control software, create a new dataset. Load the pulse sequence and default parameters for the METNOESY experiment, then set the following key parameters:

   1. Relaxation delay: 2.0 s
   2. Mixing time: 0.1 s
   3. “Dummy” scans: 4 scans
   4. Number of scans: 32 scans
   5. Transmitter offset (in ppm): 4.7 ppm
   6. Sweep width: 14 ppm
   7. Acquisition time: 3.0 s.

   To ensure compatibility with the spectra in the Chenomx library, it is critical that the sum of the relaxation delay and the acquisition time be equal to 5.0 s.

10. Set the spectrometer lock. Select “H2O + D2O” as the locking solvent if the sample contains 10% D₂O or select “D2O” as the locking solvent if the sample contains >90% D₂O.

    If “H2O + D2O” is not an option on your spectrometer, use “D2O” instead.

11. Tune and match the radiofrequency coil in the probehead.

12. Shim the magnetic field. At a minimum, the shimming process should optimize the z1-z6 “on-axis” shims and the x, y, xy, and yz “off-axis” shims.

13. Calibrate and set the 90° pulse width. Record the value for use with subsequent samples.

    The calibrated value depends on both the spectrometer and the sample. Saltier samples will result in longer pulse widths and correspondingly lower signal-to-noise in the resulting dataset. As a rule of thumb, if the calibrated pulse width exceeds double the nominal pulse width for a given probehead, consider efforts to reduce the salt concentration or the use of a narrower diameter NMR tube to obtain more favorable pulse widths.

14. Calculate and set the water-suppression power level corresponding to an 80-Hz RF field. Record the value for use with subsequent samples.

    For spectrometers that report power levels in watts, the power level is calculated as:

    $$\text{Water}\text{Suppression}\text{Power}=\text{PLW}1\times {\left(\frac{4\times \text{P}1\times 80}{1,000,000}\right)}^2$$

    where P1 and PLW1 are the duration in microseconds and power in watts of the pulse calibrated in step 13.
    For spectrometers that report power levels in decibels, the relative attenuation of the water suppression pulse is calculated as:

    $$\text{Water}\text{Suppression}\text{Power}\text{Attenuation}=-20.0\times {\log }_{10}\left(\frac{4\times \text{P}1\times 80}{100,000}\right)$$

    where P1 is the duration in microseconds of the pulse calibrated in step 7.

    The calculated attenuation is relative to the power level of the hard pulse.

15. Optimize the frequency of the water suppression pulse such that the water signal is most strongly attenuated, to within 0.1 Hz of the nominal value. Record the value for use with subsequent samples.

    A well-suppressed water signal provides two benefits: the “twist” of the spectrum baseline around the water signal is reduced (allowing better quantification of metabolites in this spectral region) and the spectrometer will be able to utilize a higher receiver gain (thereby improving the signal-to-noise ratio of the spectrum).

    During this optimization, it is critical that the spectrometer reach a steady state before the effectiveness of each attempted frequency is assessed. How this is achieved will depend on the method of optimization: if using a “real-time” acquisition mode (e.g., Bruker “gs” command), allow two to three transients to be acquired before assessing a given frequency; if collecting an array of separate experiments (e.g., Agilent array), use four “dummy” scans for each experiment in the array.

16. Calibrate the receiver gain.

17. Acquire the spectrum.

18. Process the acquired data and reference the chemical shift axis by setting the DSS singlet peak to 0 ppm.

19. Assess the quality of the shimming by measuring the linewidth at half-height of the DSS singlet peak. If the linewidth exceeds 1.0 Hz or the peak is noticeably asymmetric, re-shim the magnetic field and reacquire the spectrum.

    DSS may bind to some biomolecules, which can result in an “artificial” broadening of the DSS peak. No amount of re-shimming will remove this broadening. If your sample contains another sharp and well-resolved singlet peak, such as acetate or formate, check its linewidth as well. If it is <1.0 Hz, the shims are acceptable, and the spectrum can be used for further profiling.

20. Eject the sample and repeat this protocol on subsequent samples.

    For samples that are of similar composition, it may be acceptable to reuse the transmitter offset frequency calibrated in step 15 as opposed to calibrating it for each sample.

    It is still advisable to wait at least 5 min after inserting subsequent samples in the NMR spectrometer, to ensure the sample is fully temperature equilibrated and no small temperature gradients remain.

ALTERNATE PROTOCOL 2

OPERATION OF BRUKER SPECTROMETERS RUNNING TOPSPIN 3.X

This protocol provides the computer commands necessary to execute Basic Protocol 2 on a Bruker NMR spectrometer running TopSpin version 3.x. To perform the protocol, first carry out steps 1–6 of Basic Protocol 2 and then proceed with the steps described below.

Additional Materials (also see Basic Protocol 2)

Bruker NMR spectrometer running TopSpin version 3.x

1. Enter edte on the command line, and set the target temperature to 298.15 K.

2. Set the sample in the spinner turbine using the Bruker-supplied gauge. Enter ej on the command line to turn on the lift air. Place the sample and spinner turbine at the top of the magnet bore, then enter ij on the command line to insert the sample.

3. Enter edc on the command line to create a new dataset. Enter rpar NOESYPR1D all to load the METNOESY and default parameters. Enter d1 2.0; d8 0.1; ds 4; ns 32; olp 4.7; sw 14; aq 3.0; digmod baseopt to set all the key parameters.

   The final command uses the “baseopt” digitization mode, which should result in a better baseline and should produce a spectrum requiring a first-order phase correction close to 0°.

4. Enter lock h2o+d2o on the command line, and wait for the locking process to complete.

5. For probeheads equipped with an automated tuning and matching module, enter atma on the command line, and wait for the auto-tuning process to complete. For other probeheads, enter wobb on the command line, and tune and match the ¹H coil using the knobs at the base of the probe.

6. Enter topshim z6 tunebxyz tunea on the command line, and wait for the auto-shimming process to complete.

7. Enter the command pulsecal on the command line to begin the pulse calibration routine. Once it is complete, the routine will report the 90° pulse width in a pop-up dialog box—the relevant value is that at the bottom right. The routine will also update parameter p1, the 90° pulse width.

8. Enter the command pulse 80 Hz to calculate the pulse power attenuation. The rightmost value on the second line (calc) is the power of the water suppression field. Type pldb9, press Enter, and update the pop-up box with this value. On typical Bruker solution probes, this value should be between 35 and 50 dB.

   When inputting the value for pldb9, a transcription error can result in catastrophic damage to the NMR spectrometer. After inputting pldb9, one may double-check the applied power level in watts by querying parameter plW9, which should be considerably less than 0.001 W.

9. Enter the command gs to enter “Real-Time Go” mode. The spectrometer will repeatedly execute the METNOESY sequence and show the result onscreen. Wait for the system to reach a steady state (i.e., let two to three scans acquire) and then select the Offset tab on the left side of the screen. Optimize the value of o1 using the slider, as judged by the lowest achievable FIDAREA value, shown at the right of the “Real-Time Go” display.

10. Enter the command rga to perform an automatic gain calibration.

11. Enter the command zg to acquire the dataset.

12. Process the dataset using the command efp, and then perform a baseline correction using the command absn. Zoom in on the DSS peak, type .cal to enter the calibration interface, click on the center of the peak, and update the chemical shift to 0 ppm.

13. Zoom in on the DSS peak such that it is the only peak visible on screen, and type the command peakw to measure its linewidth.

ALTERNATE PROTOCOL 3

SEMIAUTOMATED ACQUISITION OF NMR SPECTRA OF METABOLITE SAMPLES

This protocol describes semiautomated acquisition of ¹H NMR spectra from a series of metabolite samples, using a Bruker NMR spectrometer equipped with a 24-slot “SampleCase” carousel. It is recommended that samples be randomized in order to improve quantification precision (Sokolenko et al., 2013). Although the vendor software includes an automation suite, this protocol will assess the shimming quality of each spectrum in real time, with the option to re-shim and re-acquire any spectrum that does not meet the 1.0-Hz DSS linewidth specification recommended for profiling in Chenomx.

This protocol requires a one-time installation of files into the correct directories in a TopSpin installation. It is recommended this installation be performed by an individual with advanced knowledge of TopSpin, e.g., an NMR facility manager.

The primary script file is metab_autorun.py, a Python script. This script is designed to operate in the foreground; therefore, it is imperative that no one touch or operate the NMR computer while the script is running (e.g., if a user loads a dataset while the script is running, the script may overwrite that dataset instead of writing the data in the intended folder).

The script references three ancillary files: topshim_adjust, a TopSpin “AU” script for shimming the sample, proc_baseopt, a TopSpin macro for processing the acquired data, and metab_DSScal, a TopSpin “AU” script for measuring the width of the DSS peak and storing the result as parameter userp5 in the processed dataset. The metab_DSScal script is derived from the “humpcal” and “tmscal” scripts provided by Bruker.

This protocol also uses a modified METNOESY pulse sequence file, noesypr1d_ug. The actual pulse sequence is unchanged, but the file introduces a new parameter (CNST47) that controls the desired strength of RF field used for water suppression. Upon updating CNST47 (recommended to be set to 80 Hz) and inputting the calibrated 90° pulse width and power, the pulse sequence will calculate the appropriate water suppression power level and automatically apply that power during acquisition.

Additional Materials (also see Basic Protocol 2)

Bruker NMR spectrometer equipped with a 24-slot SampleCase carousel

Topspin 3.2, with the files in the supplemental information installed in the correct folders (see Supporting Information)

1. Run the first sample using the method outlined in Alternate Protocol 2 with the following modifications:

   1. It is recommended (but not required) that the sample be loaded from slot 1 in the SampleCase carousel.
   2. During step 3, when creating a new dataset, use an experiment number (EXPNO) between 1 and 9. The metab_autorun.py script may overwrite any EXPNO >9 without warning.
   3. After completing step 3, enter pulprog noesypr1d_ug; cnst47 80 to utilize the modified METNOESY pulse sequence file and to apply an 80-Hz RF field for water suppression.
   4. Skip step 8.
   5. Continue with steps 9–13, and ensure that both the water suppression and resolution are satisfactory before proceeding.

2. After completing Alternate Protocol 2 with the first sample, place the remaining samples in the carousel.

   This automation method can tolerate “wrapping” around from slot 24 to slot 1, but there must be no gaps between the samples.

3. Enter the command metab_autorun.py to start the automation script. Provided the requested information on how the samples have been loaded on the carousel, the duration for temperature equilibration between samples (recommended to be set to 5 min), and whether the DSS linewidth should be assessed for each sample.

4. Confirm the script setup when prompted. The automation will immediately commence. Do not use the NMR computer in any way until the automation script is complete.

   The automation routine stores each dataset in the EXPNO corresponding to the carousel position number multiplied by 10: e.g., the spectrum for the sample in carousel position 23 will be in EXPNO 230.

   Any datasets re-run because the DSS linewidth was >1.0 Hz will be in the EXPNO corresponding to the carousel position number times 10, plus 1 (e.g., a re-run spectrum for the sample in carousel position 12 will be in EXPNO 121).

5. Once the automation is complete, store samples at 4°C until ready to measure their pH (Basic Protocol 3).

BASIC PROTOCOL 3

MEASURING SAMPLE pH

Measuring the pH of prepared fecal extract samples is an integral component of NMR spectrum processing. pH will cause compound peaks to shift upfield or downfield within the spectrum (Lamichhane et al., 2017). This is known as a chemical shift and is the result of the shielding or de-shielding of protons within a compound. In more acidic environments, protons will be less shielded and will move downfield along a spectrum, whereas the opposite is true in basic environments (Cañueto, Salek, & Correig, 2018). These chemical shifts can then cause problems when profiling peaks within a spectrum, which will be further explained in Basic Protocol 4. The following protocol explains how to measure the pH of prepared fecal extract samples that have already been acquired in order to improve accuracy when profiling peaks within a spectrum. The pH of a fecal sample from either a healthy human or a healthy mouse is between 6.5 and 7.5 (Rose et al., 2015). Some factors may alter the pH of fecal sample (e.g., disease and diet).

Materials

Acquired fecal extract samples of interest (Basic Protocol 2 or Alternate Protocol 2 or 3)

pH indicator paper, pH range 4.5–10 (Fisher, cat. no. 2614991)

Deionized water

Kimwipes (or other generic tissue paper), to absorb sample Beaker

1. Place a pH indicator paper for each sample being measured onto generic tissue paper.

2. Uncap the NMR tube containing the sample of interest. Place the NMR tube cap into a beaker containing deionized water. Store caps here until they can be cleaned.

3. Invert the NMR tube such that the tube contents are evenly displaced onto the pH indicator paper. Gently and evenly tap onto the pH indicator paper to remove all tube contents.

4. Record the pH of the sample by matching any color change with the pH indicator chart.

It is best to use pH indicator paper that can measure values in 0.5-unit increments, as most NMR processing and profiling software requests sample pH in such increments.

SUPPORT PROTOCOL 1

CLEANING NMR TUBES

Many apparatuses are available for cleaning NMR tubes. The main components of these apparatuses are a vacuum source, a NMR tube holder with tubing extensions (composed of a material that will not scratch the tubes), and a storage vessel that traps flushed liquid (Fig. 2). The vacuum and NMR tube holder allow cleaning solvents to be flushed through the NMR tubes. Once a cleaning solvent is added into the NMR tube holder, the solvent will be suctioned through the extension tubing, creating laminar and turbulent flow through the tubes. This thoroughly cleans any remaining sample from the tube. The storage vessel collects the used cleaning solvents.

Figure 2.

Schematic of NMR cleaning apparatus. The detachable tube holder and extensions can be customized to fit 3- or 5-mm NMR tubes. NMR tubes are placed onto the NMR tube extensions within the NMR tube holder. The tube holder stores cleaning solvent and drains into the storage vessel through the tube extensions. The storage vessel has a capacity of 1.5 liters. Tubing with a 0.2-μm-pore-size filter attaches to vacuum pump.

As fecal water extracts are diluted in buffer or D₂O, a simple approach can be taken to cleaning NMR tubes. Several rinses with water followed by a rinse with an organic solvent, such as acetone, to remove organic compounds are typically sufficient for removing contaminants. A brush should never be used, as it can cause damage to the tube and thus reduce tube performance. High heat should also be avoided because this can warp the tubes. It is recommended that NMR tubes be cleaned within 5 days of use to prevent sample content from drying in the tube and leaving a residue that is then difficult to remove. The following protocol describes how to clean NMR tubes used for analysis of fecal water extracts using a cleaning apparatus developed in the Allen-Vercoe Laboratory and the Engineering Department at the University of Guelph and the Aucoin Laboratory at the University of Waterloo. This cleaning apparatus features an NMR tube holder that is detachable and customizable. The tubing extensions within the tube holder can be customized to fit both 3- and 5-mm NMR tubes (i.e., a user can create one tube holder for each NMR tube size). This apparatus is similar to the Multi Tube Jet Washer/Dryer Complete from Wildman-LabGlass (cat. no. WG-1209-J1).

Materials

Distilled, deionized (DI) water

Acetone (Fisher Scientific, cat. no. A18–500)

500-ml graduated cylinder

Empty container, to hold acetone waste.

NMR tube rack

NMR tube cleaning apparatus (Fig. 3)

Figure 3.

Ethanol peak observed at 3.66 ppm within a spectrum acquired from a mouse pellet. (A) Poor matching of ethanol, with the subtraction line (green) nearly matching the spectrum line (black), and the sum line (red) not reaching the spectrum line. (B) Good matching of ethanol peak at 3.66 ppm, with the subtraction line nearly flattened and the sum line reaching, but not surpassing, the spectrum line.

1. Attach a vacuum to the NMR tube cleaning apparatus. Place NMR tubes onto the NMR tube holders. Make sure that the storage vessel is closed and all tubing extensions are sealed.

2. Fill the graduated cylinder with 500 ml DI water. Place 250 ml DI water into the NMR tube holder. Turn on the vacuum.

3. Lift and twirl the NMR tubes as the water flushes through the tubing extensions into the storage vessel.

   The lifting and twirling actions create laminar and turbulent flow, respectively. These flows are important for thoroughly cleaning the NMR tubes. The water will travel from the tube holder, into the NMR tube, and finally into the NMR tube extensions. The extensions will drain the water into the storage vessel.

4. Once the 250 ml of DI water has been flushed through the NMR tubes and into the storage vessel, repeat with remaining 250 ml DI water.

5. Repeat steps 2–4.

6. Empty the used DI water from the storage vessel. Ensure that the vessel is closed or sealed.

7. Fill the graduated cylinder with 200 ml acetone. Fill the NMR tube holder with 100 ml acetone.

8. Lift and twirl the NMR tubes as the acetone flushes through the tubing extensions into the storage vessel.

9. Repeat with remaining 100 ml acetone.

10. Empty storage vessel by placing flushed acetone into an appropriate container.

    Acetone is hazardous. Refer to Safety Data Sheet for full hazard statements.

11. Place NMR tubes on a slight angle onto a drying rack. Leave to dry.

    Leaving the NMR tube on a slight angle will allow any residual acetone to evaporate faster and easier. Drying typically takes 24 hr. Optionally, moderate heat (e.g., 70°C) may be used to dry off the acetone.

BASIC PROTOCOL 4

PROCESSING RAW SPECTRAL DATA

Before profiling peaks within a spectrum, corrections to the phase and baseline of the spectrum, calibration of the CSI, and inputting of the sample pH are necessary. Once a spectrum has been acquired, two files are created that can be used for spectral processing within the NMR profiling software: the fid and lr files. This protocol will focus on processing a spectrum using Chenomx NMR Suite Software Inc., though many steps are transferrable to other NMR processing software. Although the Chenomx NMR Suite is capable of processing other file formats (e.g., phasefiles, .jdf, .jcm, .dx, .ft2, and .mnova), this protocol will focus on processing fid and 1r files only. Both can be imported into the Chenomx NMR Suite in the same manner. To compare, the fid file is the raw and unprocessed time-domain “free-induction decay,” whereas the 1r file is the processed spectrum produced by the TopSpin program, and contains any phase and baseline corrections applied in TopSpin.

Phase correction will correct any phase shifts that may occur while a sample is being acquired. This can be seen within a spectrum when there are asymmetries along the baseline at the sides of peaks; in some cases, peaks may be completely inverted (i.e., pointing downward from the baseline). Correctly phased spectra should have symmetrical peaks with no inversions, as seen in Figures 4 and 5. Next, correcting the baseline, the line that represents 0 signal, will remove distortions, as seen in Figure 6. Chenomx NMR Suite provides an algorithmic baseline correction, which applies a cubic spline smoothing method to correct for rolling baselines. The predicted baseline can subsequently be modified by manually adjusting the inserted splines. Manual assessment of baseline correction is particularly important for spectra of fecal water, as peaks tend to be heavily convoluted throughout much of the spectrum, and are subject to “over-fitting” at the baseline. The integral of the CSI peak(s) is used as a reference to calibrate the relationship between peak integral and metabolite concentration. Because a set concentration of CSI, such as DSS-d6, was spiked into each sample, this concentration is equated to the empirical measurements of the CSI. As such, metabolite signatures entered into the Chenomx metabolite database under the same pulse sequence are quantifiable. Calibrating the CSI, as seen in Figure 7, is done by checking the DSS peak width within a spectrum. For best results, the DSS peak width should be <1 Hz, as explained in Basic Protocol 2. If the DSS peak width does exceed 1 Hz, there are two likely causes: poor shims during data acquisition or transient binding of DSS to large biomolecules (e.g., proteins). To discriminate between these two causes, measure the peak width of another singlet peak, such as acetate or formate. If that peak’s width is also > 1 Hz, the spectrum was acquired with poor shims, and the solution is to reacquire the data with better shims (as an alternative, the “Shim Correction” feature in Chenomx may be able to rescue a poorly shimmed spectrum). If that peak’s width is <1 Hz, the source of the problem is the transient binding of DSS to large biomolecules, for which there is no straightforward solution; the data should be used as is. Finally, the pH of the sample must be calibrated during processing in order to account for any chemical shifts of compounds. A processed spectrum will be saved as a .cnx file, which can be used to profile compounds. The following protocol explains how to use the Processor Module in the Chenomx NMR Suite Program to process fid and/or 1r files and must be followed in the order of steps presented.

Figure 4.

Phase correction for DSS peak from a spectrum acquired from a sample extracted from a murine fecal pellet. (A) Unphased DSS peak with asymmetrical smaller peaks. Bases of the peaks are not symmetrical (red dotted line). (B) Phase-corrected DSS peak with smaller peaks symmetrical to one another along their base.

Figure 5.

Baseline correction for a spectrum acquired from a mouse pellet. (A) Incorrectly processed baseline (blue) placed over areas of no signals and areas with peaks on the spectrum line (black). (B) Correctly processed baseline overlaying the spectrum line in areas with no signal and underneath areas with peaks.

Figure 6.

Unphased and phased spectra acquired from a murine pellet. (A) Extreme unphased and asymmetrical peaks are observed at ~4.8, ~1.9, and 0 ppm. (B) Symmetrical peaks are observed after phasing spectrum.

Figure 7.

CSI calibration of a spectrum acquired from a murine fecal pellet. (A) An uncalibrated CSI peak with the proposed calibration (red) overhanging the spectral peak (black). The height of the proposed calibration is greater than that of the spectral peak, and more area is covered by the proposed calibration than by the spectral peak. (B) A calibrated CSI peak with matching heights between the spectral peak and the calibration peak. Minimal overlay on the right side matches minimal underlaying on the left side. The area covered by both peaks is nearly identical.

Chenomx keyboard shortcuts:

• Scrolling: vertical zoom

• Ctrl + scrolling: horizontal zoom

• Shift + scrolling: horizontal scrolling

• Ctrl + Shift + R: moves to next spectrum

• Ctrl + Shift + W: moves to previous spectrum

• Q and E: transition between peaks of a compound (specific to the Profiler Module)

• Space bar: fit peak automatically (specific to the Profiler Module).

Materials

fid and/or 1r files from acquired fecal extract samples (Basic Protocol 2 or Alternate Protocol 2 or 3)

Chenomx NMR Suite version 7.7 or later—Processor Module (Chenomx Inc., Edmonton, AB, Canada)

1. Using the Chenomx NMR Suite Processor Module, import either the fid or 1r file.

   1. If you are using the Professional version of Chenomx and multiple files are being imported, the “Batch Import” option, under “Tools,” can be used to import all files at once. The same steps apply as with single file import and processing.

   These files are stored in the “pdata” folder. Each acquired sample should have its own pdata folder.

2. A dialogue box will appear asking to set calibration settings for the CSI and pH. Select DSS as the CSI with the final concentration within the acquired sample (e.g., 0.5 mM if a 5-mM DSS-d6 stock solution was added to a 10% [v/v] concentration). Input the sample pH derived from Basic Protocol 3. Click the “Next” button.

   Do not select the option for more than one pH indicator being present, as pH was measured manually. Formate, TSP, or another custom CSI can be used as the CSI in the Chenomx NMR Suite; however, this protocol is optimized for the use of DSS.

3. Select both “Automatic Phase Correction” and “Automatic Baseline Correction.” Click the “Next” button.

   Basic Protocol 2 provides an adequate shim to samples; therefore, the spectra do not need to be re-shimmed. Although deleting the water region within the spectra is an option, it is advisable to keep the water region, as peaks may be present near this region. Thus, deleting the water region may also delete the peaks for some metabolites.

4. Select a folder to save the processed .cnx files.

   It is useful to keep .cnx files belonging to one project or experiment in one folder to improve the ability to jump between files while processing and scanning. This is further explained in Basic Protocol 4.

5. In the Chenomx NMR Suite Processor, open a .cnx file.

   Although the 1r file should have minimal required processing, it is still advisable to look through the phasing, baseline, and CSI calibration manually and ensure that all processing has been done correctly. Note that the pH will need to be calibrated manually in both fid and 1r files.

6. Select the Phase Correction option at the bottom of the window, located in the “Step 1: Process your Spectrum” box. Select the “Auto” button. The processor will automatically calculate phase angles. Before accepting, scroll through the spectrum to ensure automatic phasing is done correctly. If it is not, manually adjust the phase angles.

   1. Scroll to the far right of the spectrum to DSS (at 0 ppm). Select the “Zero Order” phase and adjust the peaks until the baseline on either side of the peak is leveled. If necessary, use the “Fine Tune” option to adjust in smaller increments.
   2. Use the “Pivot Point” to lock the DSS peak before phasing the rest of the spectrum.
   3. Scroll to the highest peak in the spectrum (usually acetate, ~1.9 ppm). Select the “First Order” phase to correct the angles of this peak the same way as in step 6a.
   4. Return to DSS and phase First Order once more if necessary.
   5. Once content with phasing, click Accept.

   Always phase the spectrum before processing any other components. If at any point phasing is not sufficient, you may reset the phasing. This will remove all previous phasing to the spectra. Adjusting a previously phased spectrum after other processing steps will remove all processing done to that spectrum. It is advisable to phase one spectrum at a time, as not all phase adjustments made in a spectrum are adequate for other spectra (i.e., avoid the “Batch Processing” tool found under “Tools”).

7. Select the “Baseline Correction” option at the bottom of the window, located in the “Step 1: Process your Spectrum” box. Select the “Find Baseline” button. The processor will automatically correct the baseline. Before accepting, scroll through the spectrum to ensure that automatic baseline correction is done correctly. In that case, the spectrum line (black, excluding peaks) should align with the processed baseline (blue). If it is not, manually adjust the break points along the baseline such that the two align.

   1. Scroll to the far left of the spectrum, and zoom in such that any inclines and declines in the spectrum are more evident.

   2. While scrolling through the spectrum, be sure to add, adjust, or remove breakpoints along the baseline such that they follow the baseline.

   3. In areas where peaks are present, follow the trajectory of the baseline before the peak and after the peak, such that the baseline does not overwrite it. Be cautious of areas in which very small peaks are present (usually downstream in the spectrum) so as to not follow the trajectory of the peak and eliminate it.

      If the baseline is too high or too low, this will alter the concentration of that peak during the profiling of the spectrum. Going over or through the spectrum peak with the baseline will eliminate or reduce that peak from the spectrum. Peaks should never be completely below or partially below the baseline.

   4. It is advisable to avoid adding breakpoints in the water region (~4.8 ppm). Continue processing the baseline as though water was not present (i.e., make the baseline go through the water region), and adjust the baseline accordingly following the water region.

   5. Go through the spectrum one more time after adding in breakpoints to ensure the breakpoints follow the baseline correctly.

      In certain areas like inclines or declines, many breakpoints may need to be added to follow the curvature of the spectrum baseline. Sometimes, this can create extreme curvature in the processed baseline following the area of incline or decline. In this case, adding in one or more additional breakpoints before the region of incline or decline will minimize any excessive curvature.

   6. Accept the baseline processing.

8. To calibrate the CSI, click Calibrate CSI under “Step 2: Calibrate your Spectrum.” Select “Calibrate Automatically.” The calibration peak (red) should fill the spectrum line (black) where DSS is located, without surpassing the spectrum line or underfilling the DSS peak (i.e., there should be no white areas underneath the spectrum line at DSS), as seen in Figure 7. If the automatic calibration is not satisfactory, calibrate the CSI manually.

   1. Click on the calibration peak and drag it to the spectrum line where DSS is located.

      Do not allow the calibration peak to be higher than the spectrum line. Doing so will affect profiled peaks as it will change the peak/metabolite concentrations.

   2. Use the width control along the sides of the calibration peak to match it to the DSS spectrum width.

   In some instances, the area of the calibration peak will not completely fill the area of the spectrum peak. It may then be necessary to extend one side of the width of the calibration peak slightly over the spectrum peak on one side, but never over the top of the spectrum peak. If this is done, the area that extends over the side of the spectrum peak must closely equal the area on the opposite side that does not fill the spectrum peak.

9. To calibrate the pH, click “Calibrate pH” under “Step 2: Calibrate your Spectrum.”

10. Enter the measured pH of your sample. Click “Accept.”

All processed spectra will be saved in their associated .cnx file. This file will be used for spectral profiling (Basic Protocol 5).

BASIC PROTOCOL 5

PROFILING SPECTRA

Once processed, spectra may undergo profiling—i.e., the process in which projected peaks in a reference database are aligned with spectral peaks for the identification and quantification of metabolites. Software packages for profiling NMR data are equipped with reference databases. However, it is advisable to also import the Human Metabolome Database (HMDB) when working with human fecal metabolites. HMDB contains thousands of metabolites, of which 117 are of bacterial origin and have been found in stool (Wishart et al., 2018). Increasing the referencing repertoire will bolster the likelihood of profiling most peaks within a given spectrum. Although there remains the limitation that NMR databases do not encompass reference peaks for all possible metabolites, databases are continuously expanded. Several software packages capable of profiling NMR data include HiRes, BAYESIL, BQuant, and Wine Screener and Metabolic Profiler. This protocol will outline the process of spectral profiling using the Chenomx NMR Suite, although many steps can be applied to other NMR profiling software.

There are two approaches to profiling metabolites that will be outlined in this protocol: (1) Targeted profiling (when seeking to quantify the concentrations of a predetermined list of metabolites); and (2) untargeted profiling (when seeking to identify and quantify as many metabolites as possible within a given spectrum). Often, the first spectrum in a project may undergo untargeted profiling, and then the list of metabolites generated may be used to perform targeted profiling of the remaining spectra.

Profiling modules quantify metabolites in a sample by comparing the spectral peak sizes to that of an internal standard of known concentration. Profiler lines may be used to guide profiling. The spectrum line (black) represent the acquired spectrum after processing. The subtraction line (green) represents the spectrum after the peaks of profiled metabolites have been removed. As such, well-profiled metabolites result in a relatively flat subtraction line. The sum line (red) represents the summed effect of overlapping peaks and is continuously updated to reflect changes made to the profiles of each compound. This protocol describes how to profile processed spectra within the Chenomx NMR Suite—Profiler Module, though many steps are transferable to other NMR profiling software.

Materials

.cnx files of processed spectra (Basic Protocol 4)

Chenomx NMR Suite version 7.7 or later—Profiler Module with 600-MHz library HMDB recommended

1. Open the “Profile Module” in the Chenomx NMR Suite. Open the .cnx file of interest.

2. Select the metabolite databases to use as references in the “Compound Table.”

   Within the “Compound Table,” it is possible to save metabolite/compound sets. This is helpful if certain metabolites will be profiled over many different experiments. As well, the compound table will save a list of profiled compounds within a spectrum.

3. If taking the targeted approach, search for a compound of interest in the “Find in Table” box. Click on the compound name. This will take you to the corresponding peak for that compound. If more than one peak is present, this will take you to the most downfield peak (i.e., peak with the lowest ppm). For the non-targeted approach, right-click onto a peak of interest and search through compounds found within that region (a list of possible compounds will appear).

   If working with a set of spectra from the same experiment in which a targeted approach on a set of compounds will be taken, a “Batch Fit” can be done. In this option, select the spectra and compounds of interest, and the Profiler Module will automatically attempt to fit those compounds in the spectra. However, this fitting is not always accurate; therefore, it is advisable to go through each compound and spectrum individually to ensure they are matched correctly, as outlined in this protocol.

4. Ensure that the profiled compound’s peak (blue) corresponds to the peak present within the spectrum (black), that it is within the pH range, and that the spectrum peak shape matches that of the profiled compound. The position within the spectrum of the compound peak being profiled will be indicated by blue triangles on the x and y axes. The blue triangle on the x axis sits in the middle of the compound peak, while the blue triangle on the y axis indicates the compound concentration. Both can be used to fit the compound to the spectrum.

   If the compound has more than one peak across the spectrum, check through each peak individually to ensure that it matches the profiled compound peak and that it is in fact present in the spectrum. If not all peaks are present in the spectrum or the peaks do not match, the profiling compound is not present within the sample. As peaks can shift due to pH, the Chenomx software allows peaks in their database to be moved within their acceptable range. If the peak is moved outside of this acceptable range, i.e., pH cannot explain the shift in the peak, the profiled compound will appear red. Attributing a compound to peaks that appear outside of their acceptable range significantly lowers the confidence in that compound’s identification.

   To automatically fit the profiled compound, press the space bar. It is advisable to manually check that the fitting is correct.

5. Fit the profiled compound to the spectrum’s peak without allowing the projected compound peak or the sum line (red) to surpass the spectrum line. The objective is to fill the spectrum peak with the projected peak of the compound. A good fit will be evident when the subtraction line (green) flattens, which is indicative that the sum line aligns with the spectrum line, as seen in Figure 3.

   Even though a flat subtraction line is the ideal fit, in practice, there is usually a slight difference between the software metabolite signature and the sample spectrum. Therefore, the profiler should aim to have a subtraction line that is level, with noise levels evenly distributed across the baseline, above and below. If the profiling compound is superimposed onto the spectrum line and the sum line sits below the spectrum line, this likely indicates that there is another compound overlapping this peak. In this case, first profile the non-overlapped peaks, and then go back to profiling overlapped peaks such that the sum line reaches the spectrum line.

   If more than one peak is present for a compound, profile one peak at a time. Raising the profiled compound to the spectrum line at one peak will also raise the profiled compound at its other peaks, as metabolite concentration is a function of peak integration. It is recommended that larger and/or overlapping peaks be profiled first, followed by the smallest and/or non-overlapping peaks, as this will avoid any overestimations.

   When a compound is present in a very low concentration (<0.03 mM), a presence or absence approach should be taken. The identification of compounds with such low concentrations is not very reliable.

6. Once a compound has been profiled within a spectrum, open another spectrum and profile that same compound, as outlined in steps 3–5. If only one spectrum is being analyzed, begin profiling the next compound.

   It is advisable to profile one compound at a time through a series of spectra, to maintain consistency. As profiling can be subjective, it is important to approach compound profiling in the same manner throughout each spectrum.

   If using the Professional version of Chenomx NMR Suite, a batch export of profiled compounds with concentration is possible. This will be exported as a .csv file.

SUPPORT PROTOCOL 2

SPECTRAL PROFILING OF SUGARS AND OTHER COMPLEX METABOLITES

In contrast to other small-molecule metabolites, sugars can be quite challenging to profile. As a result of their cyclic structure and atomic make-up, they contain many hydrogen atoms in close proximity to one another, resulting in elaborate chemical signatures. For example, glucose has 14 unique peaks and fucose, 24.

This presents a challenge to identifying and distinguishing sugars, especially as their common structure and shared functional groups ensure that the majority of these peaks map to the same regions on the NMR spectra. Consequently, spectral crowding and overlapping peaks are observed.

The following steps should be considered when quantifying sugars or other complex metabolites.

Materials

.cnx files of processed spectra

Chenomx NMR Suite version 7.7 or later—Profiler Module with 600 MHz library HMDB recommended

1. Select the metabolite you wish to profile from the compound table.

2. If profiling more than one spectrum, select one peak to profile throughout all spectra to ensure consistency.

   As there are many peaks to choose from, it is important to quantify by measuring to the height of the same peak. This ensures consistency across spectra. If possible, try to select a peak in the least crowded region; for sugars, this is often a peak in the 5–5.4 ppm region of the spectra, corresponding to the hydrogen in the H1 position.

3. Click through each of the other peaks to ensure their presence.

   Many sugars have very similar chemical structures and differ only by the presence or absence of a few peaks. It is important to confirm that all peaks are in fact present.

4. If using a targeted approach, it is important to check that the peaks are not eliminated during the process of water suppression. Often, this is not a problem because the remaining peaks are sufficient for identification and quantification. However, if critical identification peaks are found in the 4.5–5 ppm region, identification by NMR may be challenging.

REAGENTS AND SOLUTIONS

Phosphate-buffered saline (PBS), 1 liter

8 g NaCl

0.2 g KCl

1.44 g Na₂HPO₄

0.24 g KH₂PO₄

Dilute into 1 liter DI H₂O

Adjust the pH to 7.4 with HCl

Filter-sterilize solution before using to dilute samples.

COMMENTARY

Background Information

Studying the metabolites in a biospecimen, referred to as “metabolomics,” illustrates the biochemical activities and metabolic outputs of biological systems. This method of “omics” study is especially important when investigating the gut microbiome as it illustrates a more wholesome picture of gut-microbial interactions and aids us in understanding health and disease (Monleón et al., 2009). NMR is a tool that allows us to investigate and understand gut microbial ecosystems by identifying and quantifying compounds within these ecosystems with limited sample preparation (Ravanbakhsh et al., 2015). Compared to MS, NMR is quicker, simpler, and cheaper. Also, because NMR uses an internal standard, it allows compounds to be simultaneously identified and quantified (Weljie et al., 2006), whereas quantification is not possible in MS. However, MS is more sensitive and is able to identify smaller compounds and those in lower concentrations. Ideally, use of both NMR and MS would provide more coverage of the metabolome, but NMR is certainly more cost effective and still provides ample information (Markley et al., 2017).

Understanding Results

Once a prepared NMR sample has been scanned (Basic Protocol 2 and/or Alternate Protocol 2 or 3), an NMR spectrum is acquired. This raw spectrum needs to be processed such that the baseline, phasing, and CSI calibration are all properly adjusted, to allow the correct profiling of metabolites. Example acquired NMR spectra are presented in Basic Protocol 4, Figure 4. Once the spectrum has been processed, metabolites therein can be profiled using a reference data set. Example profiled spectra are presented in Basic Protocol 5, Figure 7. NMR allows the quantification of profiled metabolites within a sample. Typically, multiple samples are scanned during one session. Quantified metabolite datasets across multiple samples may be compared using statistical analysis, described below.

Critical Parameters and Troubleshooting

If too much filtrate is lost while preparing a fecal water extract to use a 5-mm NMR tube, a 3-mm NMR tube may be used instead (so long as there is at least 180 μl of filtrate). Proceed to preparing the fecal extract as explained in Alternate Protocol 1.

The extraction step (Basic Protocol 1) is crucial for acquiring high-quality spectra. Peaks from larger compounds such as proteins may overlay over metabolites with a smaller concentration. If this is the case, a serial filtration (as explained in Basic Protocol 1) and/or additional centrifugation step may be necessary. This, however, is uncommon in fecal water extractions.

If comparing multiple fecal samples for a single experiment, it is advisable to weigh out the same mass for each fecal sample, for consistency.

When processing raw spectra (Basic Protocol 4), it is critical to avoid any baseline distortion and/or incorrect phasing, as this will cause problems later when profiling compounds of processed spectra (Basic Protocol 5). A baseline that is too low or too high will alter metabolite quantification and lead to inconsistencies. Incorrect phasing will result in difficulties in identifying compounds, as predicted compound shaped will not match those in the spectrum. If baseline distortion and/or poor phasing is evident, it is advisable to re-process the raw spectrum.

Time Considerations

The time required to extract fecal water from crude fecal samples (Basic Protocol 1 and Alternate Protocol 1) is dependent on the number of samples undergoing extraction and their density. Extraction of fecal water from 12 samples typically takes 2 hr.

Acquiring NMR spectra of metabolite samples (Basic Protocol 2) takes 15 min per sample. Acquiring NMR spectra of metabolite samples using a Bruker Spectrometer Running TopSpin 3.x (Alternate Protocol 2) takes 15 min per sample. Semiautomated acquisition of NMR spectra of metabolite samples (Alternate Protocol 3) can take up to 4 hr for 12 samples with a 5-min temperature calibration for each sample. These time considerations for both protocols apply when using a 600-MHz spectrometer with a cryogenic probe. With a non-cryogenic probe, an additional 15 min per sample will be needed.

Measuring sample pH of samples (Basic Protocol 3) will take 2 min per sample.

Cleaning NMR tubes (Support Protocol 1) is dependent on the number of tubes being cleaned. 12 NMR tubes typically take 50 min to clean and 24 hr to dry.

Processing raw spectra data (Basic Protocol 4) is dependent on user experience. An experienced user can process a spectrum in 10 min, whereas a novice user may require 15–20 min per spectrum.

Spectral profiling, and especially and spectral profiling of sugars and other complex metabolites (Basic Protocol 5 and Support Protocol 1), can be time consuming for both experienced and novice users. The time required is dependent upon whether the user will be taking a targeted or untargeted profiling approach, the number of compounds being profiled, and how well the spectrum is processed. An experienced user is typically able to profile 12 compounds within a spectrum in 1–2 hr. A novice user may need an additional hour. Batch fitting may decrease the anticipated needed time by an hour.

Statistical Analysis

Once compounds have been identified and quantified from a stool sample, statistical analysis of the profiled spectra can be performed. In investigating more than one fecal sample, statistical analysis can help in determining whether and/or how certain compounds shift away or group together. This is especially important in a clinical setting. For example, it would be beneficial to see how the metabolic output of a gut microbial community responds to antibiotic treatment. Identifying key differences and patterns in a highly dimensional space is challenging, but performing partial least-squares discriminate analysis (PLS-DA) can be a starting point. PLS-DA is a tool for dimensionality reduction and can be especially helpful for isolating any metabolites that are critical to shifts in community metabolism. Conveniently, it also prepares the data for visualization in 2D space, providing an interpretable look at the data. Metabolite concentrations, which are exported into a concentration data matrix, can be imported into statistical computing software (e.g., R and Python) capable of conducting PLS-DA analysis. When analyzing more than two sample sets, color-coordinating sample sets may be helpful in distinguishing between shifts and sample relationships. As an example, Figure 8, from Carlucci et al. (2019), displays the independent clustering of three microbial ecosystems based on metabolic output.

Figure 8.

Example of partial least-squares discriminate analysis (PLS-DA) on metabolite datasets obtained from simplified (for clarity) gut-derived communities. Defined ecosystems were obtained from a stool sample from a healthy individual; two defined microbial ecosystems, one of which was perturbed with antibiotic, were each run in separate bioreactors using conditions mimicking the human distal colonic environment. PLS-DA analysis reveals independent metabolic clusters based on the metabolic output from each community. Control: growth medium only. Figure derived from Carlucci et al. (2019).