Introduction

The current assessment of vitamin B12 (B12) deficiency is based on a plasma or serum-B12 concentration based cut-off, or on cut-offs derived from the concentration of B12-dependent plasma metabolites, such as homocysteine (Hcy) or methylmalonic acid (MMA)1,2. These are not very specific and their use is further limited by several confounding factors, including renal impairment, aging, and inflammatory states, which reduce their specificity and sensitivity3. Recently, a cut-off derived from a combination of these biomarkers to derive combined B12 (cB12) index along with the active fraction of circulating B12 bound to transcobalamin (holotranscobalamin; HoloTc), has been suggested as a more specific diagnostic4. However, all these biomarkers have varying turnover times and pool sizes and may be relatively static in response to a more dynamic body B12 status. In contrast, the propionate oxidation based breath test directly measures the functional activity of the B12-dependent mitochondrial enzyme, methylmalonyl-CoA mutase. This enzyme5,6 converts methylmalonyl-CoA, which is formed from propionyl-CoA, to succinyl-CoA, allowing for its subsequent oxidation to CO2. Therefore the isotopic tracing of the oxidative metabolism of [13C]-propionate to 13CO2 release in the breath provides a non-invasive, yet functional and dynamically responsive, test of the B12 status in the body6. The [13C]-propionate breath test has been shown to be both sensitive and specific5 for B12 status based on cross-sectional evaluations against serum B12. However, this functional breath test has not been used to define the serum B12 cut-off for deficiency. Equally, there are many B12-dependent plasma metabolites and patterns in untargeted metabolomic analyses, which might also offer more sensitive, responsive, and functional indicators of B12 status in deficiency and repletion state.

We therefore aimed first to evaluate a cut-off to diagnose B12 deficiency by the functional [13C]-propionate breath test through a breakpoint analysis against serum B12 concentrations, in comparison with the currently used cut-off of < 156 pMol/L7. Since earlier studies8,9 have demonstrated age and gender-based differences, with males10 being more susceptible to vitamin B12 deficiency, we aimed to study the novel and functional B12 status indicators through untargeted serum metabolomic profiling based analyses, in a range of deficient and replete male participants (as defined by the present recommended serum B12 cut-off11,12), and to correlate these with the functional [13C]-propionate breath test and classical B12 status indicators like cB12, MMA, Hcy, HoloTc, folate, ferritin and acylcarnitines. We further tested these functional profiles in B12 deficient male participants on 3 and 15 days after replenishing their B12 stores by intramuscular administration of a large dose (3 mg) hydroxocobalamin based on the standard guidelines recommending 1 mg of hydroxocobalamin to be administered intramuscularly every other day during the repletion phase, equating to 3–4 mg dose weekly for rapid tissue and plasma repletion. An evaluation period of 15 days was chosen to accommodate redistribution of vitamin B12 across physiological compartments following acute high-dose repletion, and to enable stabilization of downstream metabolic pathways reliant on cobalamin-dependent enzymatic activity, thereby facilitating assessment of steady-state functional responses. Alterations in metabolomic and classical B12 indicators were compared against changes in the functional [13C]-propionate breath test before and after treatment.

Methods

Subjects and experimental protocols

Healthy male participants aged 20 to 40 years, with a normal body mass index (BMI, < 25 kg/m2), were recruited from in and around St. John’s. The Institutional Ethical Review Board of St. John’s Medical College, Bengaluru, India, approved the study protocols and written informed consent was obtained from all the participants at enrolment. All experiments were performed in accordance with relevant guidelines and regulations. The study was registered at the Clinical Trials Registry of India (CTRI/2018/04/012957), on 03/04/2018. Each participant underwent a complete clinical examination, followed by measurement of their complete blood count (CBC, ABX Pentra 60 C + , HORIBA ABX Diagnostics, Japan) and serum total vitamin B12 (B12) concentration (Siemens Advia XPT immunoassay platform). Trilevel immunoassay quality controls (Biorad Lyphochek) were used for quality control, with inter- and intra-assay coefficient of variation of 8.0% and 5.0% respectively. Of the 91 participants enrolled, 37 had normal serum B12 concentration (≥ 156 pmol/L, ‘normal B12’), while 54 had low serum B12 concentrations (< 156 pmol/L, ‘low B12’)11,12.

On the day of experiment, participants reported to the metabolic ward in the fasted state at 7 AM. A baseline blood sample was collected and separated for serum and plasma as per standard protocols and stored immediately at – 80 °C for later biochemical and metabolomic profiling. In addition, a baseline breath sample was collected, to measure 13CO2 abundance and subsequently all participants were orally administered with sodium [1-13C]-propionate (~ 30 mg; Cambridge Isotope Laboratories, Andover, MA). This dose was prepared gravimetrically on a calibrated weighing scale in a sterile stable isotopic laboratory and was administered along with a small piece (~ 10 g) of whole-wheat unsweetened bread. After the propionate dose was consumed, breath samples were collected every 10 min for during the first hour and every 15 min thereafter until the end of the 2nd hour (120 min). The breath samples were later analyzed for 13CO2 enrichment, to calculate the index of propionate oxidation.

In participants with low serum B12 concentration, the effect of parenteral B12 supplementation on their serum metabolome and the functional vitamin B12 deficiency index through breath test was also evaluated. Subjects were administered a single intramuscular dose of hydroxocobalamin (3 mg; Trineurosol-Hp, Zen Pharma Pvt Ltd, India), and the blood sampling and the breath test was repeated on day 3, after measuring the plasma vitamin B12 concentration, to ensure the replenishment of vitamin B12 stores. These subjects were also followed-up on day 15 (those who were willing to return, n = 11) to confirm if the elevated plasma vitamin B12 concentrations were maintained and to characterize the alterations in plasma metabolome. Thus, in this part of the experiment, untargeted metabolomics analysis was performed on Day 0, 3 and 15, while the functional breath test was performed on Day 0 and 3.

Quantification of methylmalonic acid and homocysteine concentrations

Blood plasma samples were also analyzed for assessing methylmalonic acid (MMA) and homocysteine (Hcy) concentrations by gas chromatography mass spectrometry based method. Briefly, plasma and pooled quality control samples (200 µL) were treated with dithioerythritol (20 µL; 200 mmol/L) containing mixture of D,L-d8-homocystine (DL-[3,3,3′,3′,4,4,4′,4′-2H8] and d3-MMA (> 98% purity; Cambridge Isotope Laboratories, Massachusetts, USA) as an internal standard (IS) and incubated at room temperature for 20 min to reduce the disulfide bonds of homocysteine and deproteinized with ethanol (400 µL), and derivatized and extracted in a single-step by the addition of methylchloroformate and toluene. The derivatives N(S)-methoxycarbonyl ethyl esters13 were analyzed in the selected-ion monitoring (SIM) mode by Gas chromatography mass spectrometry (GCMS-SQ, 5975, Agilent Technologies, CA, USA).

The analytes were separated on a VF-17MS capillary column (30 m, 0.25 mm inner diameter, 0.25 µm film thickness, J&W Scientific, CA, USA) and samples were injected in splitless mode. GC oven temperature was set at 75 °C initially which was held for 5 min and increased at a rate of 30 °C/min to 195 °C and was maintained for 2 min, and increased at 10 °C/min to 290 °C and maintained for 2 min. Helium was used as carrier gas and delivered at a flow rate 1.1 mL/min. Injector temperature maintained at 200 °C and MS was operated in electron ionization mode with following conditions; trap temperature, 190 °C, manifold temperature, 40 °C transfer line temperature, 270 °C, with electron energy of 70 eV. The calibration curves for MMA and Hcy quantification were obtained by plotting the peak area ratio for the m/z 233:237 and 174:177 ion pairs vs the concentration of Hcy and MMA in the range of 0.02–100 µmol/L and, and the resulting slope and intercept from regression equation were used to quantify Hcy and MMA concentrations in unknown samples. Two levels of in-house quality control samples prepared from pooled plasma were injected every day; intra-assay CV (coefficient of variation) were < 4% and < 2% and inter-assay CV were < 6% and < 3% for MMA and Hcy respectively. In addition, BioRad quality control samples for Hcy were also run along with each analytical batch daily for validation of the assay.

Plasma Active B12 (HoloTc), concentrations were also measured by Abbott Architect Analyzer. Biorad Lyphochek trilevel immunoassay quality controls were used, and inter- and intra-assay CVs (coefficient of variation) were 8.0% and 5.0% respectively.

The concentrations of ferritin and folate were measured by electrochemiluminescence based assay using the Cobas E411 immunoassay analyzer (Roche Diagnostics, USA). Biorad Lyphochek trilevel immunoassay quality controls were used, and inter- and intra-assay CVs were below 1.0% and 5.0% for all the controls, respectively.

Quantification of acylcarnitine species

Plasma aliquots (100 µL) were spiked with 4 μL of 2H-labelled internal standards (labelled and unlabeled standards; NSK-B and NSK-B-G1, NSK-B-US and NSK-B-G1-US were procured from Cambridge Isotope Laboratories, Andover, MA) and deproteinized with 800 μL of ice-cold methanol followed by centrifugation at 14,000 rpm for 20 min at 4 °C. The dried extracts were reconstituted with freshly prepared 100 μL of 0.2 M OBA (o-benzylhydroxylamine; prepared in 50:50 methanol:10 mM ammonium acetate, pH 5.5) and 10 μL of 2 M EDC (N-(3-dimethylaminopropyl)-N-ethylcarbodiimide hydrochloride prepared in water). After 10 min of incubation at room temperature, samples were analyzed by an Agilent 6495 iFunnel Triple Quadrupole Liquid chromatography mass spectrometry (LCMS/MS) equipped with a 1290 Infinity binary pump, autosampler, and a thermostatted column compartment (ZORBAX Extend C18 column; 2.1 × 100 mm ID, 1.8 micron particle size). The MS was operated in heated-electrospray mode with positive polarity in a dynamic reaction monitoring (DMRM) based method. The concentrations of unknown samples were quantified by the respective regression equations for individual carnitine species generated by plotting peak-area ratios of respective carnitines against their concentrations in standards (five-levels for each of the carnitine species).

Functional breath test: breath 13 CO 2 analysis

Breath samples were analyzed for their 13CO2 abundance by monitoring ions at m/z ratios of 44 and 45 by IRMS (isotope ratio mass spectrometry, Delta V Advantage, Thermo Scientific, Germany). The ratio of 13CO2 to 12CO2 (m/z 45 to 44) were measured in the sample and compared to a reference gas (99.9999% CO2) which was calibrated with international standards. The atom % excess of 13CO2 for each time point, in breath samples were used to calculate the percent of the dose recovered (PDR) in the breath during the experimental protocol. The area under the curve (AUC) for each time point was calculated by the linear trapezoid method.

The total 13C excreted was calculated as: Total 13C Excreted (mmol) = % 13C (AUC) × CO2 production (mmol/min) × Time (min).

In this case CO2 production estimated by indirect calorimetry was used after adjusting for physical activity level (PAL) of 1.5 for this population. The calculated PDR was expressed as; PDR120: PDR over the entire test period of 120 min, PDRnormalized; normalized to a per hour, and PDRmaxima; the highest PDR observed during the study period.

Global untargeted metabolomics analysis

Plasma samples (100 µL) were spiked with an isotopically labeled internal standard (IS), 2H labeled amino acid mix (20 µL; 1 ng/mL; U-2H labeled amino acid mix > 97% purity; Cambridge Isotope Laboratories, Massachusetts, USA) along with threefold of chilled organic solvent (8:1:1, acetonitrile: methanol: acetone) and vortex-mixed. Samples were incubated at 4 °C for 30 min and centrifuged at 20,000 × g for 20 min in a refrigerated centrifuge (5810 R, Eppendorf, USA). Supernatants were transferred to another vial and dried at 40 °C in a vacuum concentrator (Labconco, USA). Dried extracts were reconstituted in acetonitrile/water (1:1) and analysis was performed on a high-resolution accurate-mass platform consisting of an ultra-high performance liquid chromatograph (UHPLC, Thermo Scientific, Vanquish Flex Binary, MA, USA) coupled to an orbitrap based mass spectrometer (Q Exactive, Thermo Scientific, San Jose, USA). The mass spectrometer was calibrated by using a positive and negative ion calibration solution (Thermo Scientific™ Pierce LTQ Velos ESI) as an external calibration on daily basis for positive and negative data acquisition respectively. Separation of the metabolites was achieved by using a Zorbax Eclipse plus-C18 column (2.1*150 mm, 1.8 µm), Agilent Technologies, CA, USA) at 40 °C. The mobile phase was delivered in a reversed-phase gradient elution at 0.35 mL/min, using water (eluent A) and acetonitrile (eluent B), both containing 0.1% formic acid. The following gradient profile was used: 0–3 min: 1% B and increased to 95% B at 14 min and held for 3 min then decreased to 1% B at 17.5 and equilibrated for another 3 min. Reconstituted extracts were loaded on an autosampler where the injection volume was set at 5 μL for each of the solvent blanks, pooled quality control (QCs) samples, which included four technical replicates of a pool of aliquots from plasma samples (from each of the study groups14), along with biological samples in batches daily.

The Q Exactive mass spectrometer was operated under electrospray ionization (HESI-II) positive and negative polarity mode in full scan (m/z 67–1000) and used resolution 70,000 (FWHM) at m/z 200, with automatic gain control (AGC) target of 1 × 106 ions and a maximum ion injection time (IT) of 100 ms. Data-dependent MS/MS were acquired on a “Top5” data-dependent mode using the following parameters: resolution 35,000; AGC 1 × 105 ions; maximum IT 50 ms; 1.0 amu isolation window; combined NCE 25%, 35% and 50% and dynamic exclusion time was set at 10 s; Spectrum data type, Centroid. Source ionization parameters were; sheath gas flow rate, 40; spray voltage, 3.80 kV; capillary temperature, 330 °C; heater temperature 350 °C and S-Lens level, 50. For negative polarity mode source ionization parameters were; Sheath gas flow rate, 45;

spray voltage, 3.50 kV; capillary temperature, 330 °C; heater temperature 350 °C and S-Lens level, 50.

Positive and negative polarity mode acquired data was analyzed by Thermo Scientific™ Compound Discoverer™ (version 3.3.0.550) software as described earlier15. Metabolites were visualized and mapped to pathways and automatically identified by mzCloud within Compound Discoverer software. Manual inspection of each annotated compound was done for each metabolite having MS/MS library match.

Statistical analyses

Anthropometric and biochemical data were presented as median (QR-Interquartile range). The continuous variables were examined for normal distribution using Shapiro-Wilks test. If the variables were normally distributed, paired t-test, else the Wilcoxon Rank Sum test was performed to compare the targeted metabolomes before and after treatment. The Spearman Rank correlations based analysis was performed to assess the interrelationship between PDR indices before and after treatment.

To assess the functional breakpoint for B12 deficiency, a piecewise linear regression model was performed. The level of serum B12 that maximized the response on [13C]-propionate, cB12, HoloTc and minimized the response on MMA and Hcy were identified by the breakpoint changes in slope16. The statistical significance of these breakpoints was assessed using Davies’ Test17. The Delta method18 was used to determine standard error of the breakpoint estimate and then used to calculate a 95% CI.

For untargeted metabolomic data, initial visualization and analysis such as groups area ratios, fold change (log2 scale), coefficient of variance per sample groups, trend charts, Principal Component Analysis (PCA), Partial least squares discriminant analysis (PLS-DA) as well as differential analysis by ANOVA were performed within the Compound Discoverer™ 3.0 software. The p-values (per group ratio by ANOVA and TukeyHSD post-hoc test) were adjusted by using Benjamini–Hochberg corrections for the false discovery rate. The combined analysis of targeted and untargeted metabolites including breath test indices was performed for an overall visualization, a PLS-DA based score plot was also generated.

Further ChemRICH analysis, a statistical enrichment approach, was also performed that is based on chemical similarity enrichment analysis of the metabolites by using ChemRICH tool available at www.chemrich.fiehnlab.ucdavis.edu, that uses chemical ontologies-based strategy along with structural similarity to group/classify the metabolites and represents the significantly impacted metabolite clusters accordingly by grouping the annotated metabolites into non-overlapping classes based on Medial Subject Headings (MeSH) terms and Tanimoto chemical similarity calculations19. ChemRICH tool was used to detect differential regulation of various chemical classes in the subjects with normal vitamin B12 status in comparison to the low vitamin B12 status that also underwent treatment and to identify specific alterations throughout treatment until Day 15. The input data included the annotated feature name, chemical identifiers including SMILES code, uncorrected p-values of Student t-tests between two groups of interest, and fold change between the two groups of interest for each annotated class of metabolites as ChemRICH uses the Kolmogorov–Smirnov test to calculate significance level (P-value) between two treatment groups for all detected metabolite classes.

All the analysis were performed using R (software version 4.3.1-R Foundation for Statistical Computing) and p < 0.05 was considered to be statistically significant.

Results

The anthropometric and biochemical data for all the study participants (n = 91 males) cross-sectionally, are provided in Table 1. All participants were screened for their hemoglobin and serum B12 concentrations at the baseline and were assigned to low (< 156 pMol/L, n = 54) and normal (> 156 pMol/L, n = 37) B12 groups per the recommended serum cut-off for B12 deficiency7,11,12, as shown in the flow-chart in Fig. 1a. Targeted biomarker analyses showed that serum B12 concentrations correlated significantly positively with HoloTc concentrations (n = 91, rho = 0.63, p < 0.001) and negatively with MMA (n = 91, rho = − 0.47, p < 0.001) and Hcy (n = 91, rho = − 0.48, p < 0.001) concentrations, as shown in Fig. 1b–d respectively.

Table 1 Anthropometry and biochemical parameters of all the subjects at baselinea.
Full size table
Fig. 1
Fig. 1
Full size image

Consort flow chart for screening and enrollment of the participants and vitamin B12 correlates at baseline. (a) Consort flow-chart describing the study groups and their allocation after recruitment for the participants who completed the study, (b–d) significant correlation of B12 with holotranscobalamin, methylmalonic acid and homocysteine (p < 0.05) respectively.

Identification of serum B12 deficiency cut-off from the functional [13C]-propionate breath test

The identification of a serum B12 diagnostic cut-off to define functional deficiency was based on a segmented analysis (broken-line model) of serum B12 concentrations values (ranging from 50.2 to 267.2 pMol/L, n = 31) plotted against the corresponding functional [13C]-propionate breath test values. The output of [13C]-propionate oxidation in breath samples was measured as the percent dose recovery (PDR) over 120 min (PDR120), that can be normalized to an hourly value (PDRnormalized) and can also be expressed as the highest PDR observed during the study period (PDRmaxima). However, for this analysis PDR120 was used and that gave a significant serum B12 breakpoint of 144.4 pmol/L [95% CI 106.4–182.4, p = 0.02, adjusted R2 = 0.219] (Fig. 2a). The B12 deficiency cut-offs estimated from cB12, HoloTc, MMA, Hcy and C3/C2 were 176.6 pmol/L [95% CI 140–213, p < 0.001, adjusted R2 = 0.643], 97.4 pmol/L [95% CI 71.7–123.1, p = 0.01, adjusted R2 = 0.467], 219.7 pmol/L [95% CI 151.6–287.7, p = 0.03, adjusted R2 = 0.723], 108.1 pmol/L [95% CI 76.8–139.3, p < 0.001, adjusted R2 = 0.393] and 85.1 pmol/L [95% CI 61.1–109.1, p = 0.03, adjusted R2 = 0.182] as shown in Fig. 2b–f respectively. Overall, these diagnostic breakpoints varied in approximately a two-fold range of 85.1 to 176.6 pmol/L.

Fig. 2
Fig. 2
Full size image

Breakpoint analysis of targeted metabolites for B12 deficiency. (a) The breakpoint of functional breath test against serum B12 at 144.4 pmol/L [n = 31, 95% CI 106.4–182.4, p = 0.02, adjusted R2 = 0.219], (b) cB12 at 176.6 pmol/L [n = 49, 95% CI 140–213, p < 0.001, adjusted R2 = 0.643], (c) HoloTc at 97.4 pmol/L [n = 23, 95% CI 71.7–123.1, p = 0.01, adjusted R2 = 0.467], (d) MMA at 219.7 pmol/L [n = 20, 95% CI 151.6–287.7, p = 0.03, adjusted R2 = 0.723], (e) Hcy at 108.1 pmol/L [n = 49, 95% CI 76.8–139.3, p < 0.001, adjusted R2 = 0.393] and (f) C3/C2 at 85.1 pmol/L [n = 29, 95% CI 61.1–109.1, p = 0.03, adjusted R2 = 0.182] concentrations.

Global untargeted metabolomics, targeted B12 metabolites and biomarkers in cross-sectional evaluations between low and normal B12 status participants

The untargeted metabolomics analysis (positive and negative ion mode) in low B12 status participants in comparison to normal participants, performed by Compound Discoverer (3.1.0.305), showed 9203 metabolite features (positive mode; 5484 and negative mode; 3719), of which 1451 could be assigned identifiers. The unsupervised principal component analysis (PCA) performed by Compound Discoverer demonstrated clear separations between the study groups in both positive and negative ion modes. The untargeted metabolites (from both positive and negative ion mode) were combined with targeted biomarkers (B12, MMA, Hcy, HoloTc, ferritin, folate, and acylcarnitine species), and an imputed combined index (cB12), to perform a partial least squares discriminant analysis (PLS-DA)20. The PLS-DA based score plot (Fig. 3a) showed clear discriminating clusters of low (LB12) and normal vitamin B12 (NB12) participants. This analysis revealed the top ‘481’ metabolites that were crucial for the cluster formation. Among these, the most relevant metabolites are listed in decreasing order of importance based on their VIP (variable importance in projection) score values greater than or equal to 121 in Table 2. To further specify the metabolic profile associated with B12 deficiency, ChemRICH analysis was used. This is a statistical enrichment approach that utilizes chemical similarity coefficients to cluster metabolites into non-overlapping chemical groups19. Here, 44 different classes of metabolites were altered (increased or decreased), as shown in Supplementary Table 1a. Increased levels of sulfur and branched chain amino acids (methionine, valine, isoleucine), acetylcarnitine, 3-hydroxyoctanoylcarnitine, butenylcarnitine, 9-decenoylcarnitine, 3-hydroxybutyrylcarnitine), phosphatidylcholines (2-palmitoyl-sn-glycero-3-phosphocholine, lysophosphatidylethanolamine) bile acids (cholic acid, deoxycholic acid, glycocholic acid, glycoursodeoxycholic acid), 3-hydroxybutyric acid, and various dipeptides were observed in normal B12 participants. The number of compounds (cluster size) with fold change and their significance along with the details on up or downregulation of a specific metabolite is provided in (Supplementary Table 1b).

Fig. 3
Fig. 3
Full size image

Cross-sectional evaluation of study groups and their correlation matrices at baseline. (a) PLS-DA score plot for normal B12 (NB12) and low B12 (LB12) participants, (b) Correlation matrix of targeted metabolites in NB12 and LB12 participants.

Table 2 List of metabolites in decreasing order of importance based on their VIP (variable importance in projection) score values greater than or equal to 1 in low (LB12) and normal B12 (NB12) participants by PLS-DA based score plot.
Full size table

The targeted biomarkers analysis showed that B12 correlated significantly positively with HoloTc concentrations (n = 54, rho = 0.51, p < 0.001) and negatively with Hcy (n = 54, rho = − 0.45, p < 0.001) in low B12 participants. However, none of these correlations were significant amongst the normal B12 participants.

The combined untargeted and targeted biomarker analysis showed that the ratio of propionylcarnitine to acetylcarnitine (C3/C2) correlated positively with MMA (n = 32, rho = 0.39, p = 0.030), negatively with HoloTc (n = 32, rho = − 0.49, p = 0.005), whereas the ratio of taurine (T) to conjugated form of chenodeoxycholic acid with taurine; tauroChenodeoxycholic acid (T/C), correlated positively with B12 (n = 32, rho = 0.37, p = 0.036) and with carnitine (n = 32, rho = 0.35, p = 0.047). C3/C2 negatively correlated with butyrylcarnitine (n = 32, rho = − 0.38, p = 0.035) and C3/C16 negatively correlated with myristoylcarnitine (n = 32, rho = − 0.44, p = 0.013), and octadecanoylcarnitine (n = 32, rho = − 0.40, p = 0.026) whereas C3 alone correlated positively with isovalerylcarnitine (n = 32, rho = 0.39, p = 0.029) as shown in Fig. 3b.

The effect of B12 replenishment in low B12 status participants on serial evaluations of global untargeted metabolomics, targeted B12 metabolites, B12 biomarkers and the [13C]-propionate breath test

Parenteral B12 supplementation (3 mg intramuscular) in low vitamin B12 status participants (< 156 pmol/L, n = 11), increased the propionate oxidation by 14.9% after treatment on day 3 as shown in Fig. 4a (mean: 56.22 vs 48.26, p < 0.001). Although, three PDR indices (PDR120, PDRnormalized and PDRmaxima) are computed from propionate oxidation test, the PDR120 is used for all the correlation analyses, as all the three PDR indices intercorrelated positively and significantly both before (n = 21, rho = 0.89, p < 0.001) and after treatment (n = 11, rho = 0.82, p < 0.005) as shown in Fig. 4bi,bii respectively. The PDR120 correlated positively with serum vitamin B12 concentrations before (n = 21, rho = 0.53, p = 0.011) and after treatment (n = 11, rho = 0.67, p = 0.02) and negatively with Hcy (n = 11, rho = − 0.33, p = 0.32).

Fig. 4
Fig. 4
Full size image

The Percent dose recovered 13CO2 over 120 min and its interrelationship with PDRnormalized; normalized to a per hour, and PDRmaxima; the highest PDR observed during the study period. (a) 13C-propionate dose recovered over 120 min in low vitamin B12 subjects before treatment (dotted line with diamond) i.e., before intramascular injection and after treatment (continuous line with circle) i.e., after intramascular injection. The standard error of the mean (SEM) is shown at each point for n = 11 participants, (b) Interrelation of PDR indices (PDR120, PDRnormalized and PDRmaxima) before (i) and after (ii) treatment (p < 0.05).

Parenteral B12 supplementation (3 mg intramuscular) altered various untargeted metabolites in low B12 participants. The PLS-DA based score plot (Fig. 5a) showed clear discriminating clusters of low B12 (LB12) participants before treatment and after treatment on Day 3 and Day 15. This analysis revealed the top ‘459’ metabolites crucial for the cluster formation, among these, the most relevant ones are listed in Table 3 based on their VIP score values greater than or equal to 1. ChemRICH analysis showed metabolic alterations in 43 different classes (Supplementary Table 2a). These included dipeptides, unsaturated and saturated fatty acids, carnitine, lysophosphatidylcholines, dicarboxylic acids, saturated stearic acids, indoles, phenols, amino acids, sugar acids, lysophospholipids, oleic acids, dihydroxyoctadecenoic acid, oligopeptides, bile pigments and glycerophospholipids including other compounds (supplementary Table 2b).

Fig. 5
Fig. 5
Full size image

Effect of treatment on combined metabolites including breath test in low B12 participants and their correlation matrices. (a) PLS-DA score plot of combined metabolites for LB12: low B12, LB12_D3: low B12_day 3 (after treatment), LB12_D15: low B12_day 15 (after treatment), (b) Correlation matrix of targeted metabolites in LB12 participants at baseline, (c) Correlation matrix of targeted metabolites in LB12 participants at Day 3 (after treatment), (d) Jitter plots demonstrating the significant changes in B12 biomarkers after treatment in B12, folate, cB12 (combined B12), MMA (methylmalonic acid), HoloTc (holotranscobalamin), Hcy (homocysteine), taurochenodeoxycholic acid and propionylcarntine (C3).

Table 3 List of metabolites based on their VIP score values greater than or equal to 1 in low B12 (LB12) participants before treatment and after treatment on Day 3 and Day 15 as assessed by PLS-DA based score plot.
Full size table

ChemRICH analysis revealed metabolic alterations in 36 different classes of compounds altered after treatment on day 15 in comparison to low B12 participants (Supplementary Table 3a) that had 306 metabolites up or downregulated (Supplementary Table 3b). The sulfur and branched chain amino acids, methionine, valine, iso-leucine increased after treatment on Day 3 and remained as such until Day 15 after treatment (Supplementary Table 3b).

Tiglylcarnitine, 2-methylbutyrylcarnitine, butenylcarnitine, N-propionylmethionine, tauroChenodeoxycholic acid were increased after treatment whereas propionylcarnitine, O-heptanoylcarnitine, 3-[(2,6-Dimethylheptanoyl)oxy]-4-(trimethylammonio)butanoate) were decreased on Day 3 after treatment. Other classes of oligopeptides and dipeptides were also increased including carboxylic and unsaturated and saturated fatty acids detailed in Supplementary Table 2b, with cluster size and significance. Parenteral B12 supplementation significantly increased the concentrations of targeted biomarkers such as B12, cB12, HoloTc, taurochenodeoxycholic acid and decreased folate, MMA, Hcy and propionylcarnitine (p < 0.05) (Fig. 5d) with no effect on ferritin and other carnitine species. There was a significant difference (p < 0.05) in the concentrations of propionylcarnitine (C3) and its ratio to acetylcarnitine (C2) and palmitoylcarnitine (C16) on Day 3 and Day 15 after treatment, such that the median values of C3/C2 and C3/C16 were 0.07, 0.31; 0.05, 0.24 & 0.04, 0.17 at Day 0, Day 3 and Day 15 respectively. The median percentage decrease observed in these indices after treatment (Day 3) were 30.2%, (IQR 31.4) for C3, 19.8%, (IQR 39.6) for C3/C2 and 35% for C3/C16 (IQR 30.8) and 34.2%, (IQR 25.4) for C3, 31.5%, (IQR 32.4) for C3/C2 and 45.0% for C3/C16 (IQR 16.1) on Day 15.

C3 concentrations correlated negatively with both cB12 (n = 16, rho = − 0.66, p = 0.007), and HoloTc (n = 16, rho = − 0.61, p = 0.014). The ratio of propionylcarnitine (C3) to acetylcarnitine (C2) i.e., (C3/C2) correlated negatively with HoloTc (n = 16, rho = − 0.71, p = 0.003), cB12 (n = 16, rho = − 0.62, p = 0.012), butyrylcarnitine (n = 16, rho = − 0.55, p = 0.030), hydroxyisovalerylcarnitine (n = 16, rho = − 0.53, p = 0.036), and positively with MMA (n = 16, rho = 0.74, p = 0.002) whereas C3 and C3/C16 correlated positively with MMA (n = 16, rho = 0.72, p = 0.002, rho = 0.54, p = 0.030) and C3/C16 correlated negatively with cB12 (n = 16, rho = − 0.55, p = 0.026), alone in low B12 participants, whereas T/C correlated positively with ferritin (n = 16, rho = − 0.51, p = 0.048) as shown in Fig. 5b.

On Day 3 after treatment, the ratio of propionylcarnitine to palmitoylcarnitine (C3/C16) correlated positively with folate (n = 10, rho = 0.70, p = 0.031) whereas C3 correlated positively with Hcy (n = 10, rho = 0.69, p = 0.035) and with isovalrylcarnitine (n = 10, rho = 0.72, p = 0.024) as shown in Fig. 5c. However, none of these correlations were observed on Day 15, perhaps owing to the limited number of participants at this time point.

Discussion

We have demonstrated the utility of a non-invasive, functional [13C]-propionate breath test to characterize B12 deficiency and used this to derive a serum B12 diagnostic cut-off to define the latter. The cut-off derived from this analysis was slightly lower than the conventionally used value7,11,12. All other diagnostic breakpoints showed a varied range of approximately two-fold i.e., 85.1 to 176.6 pmol/L. However, the declining trend of the breakpoint line after the cut-off value of 144.4 pmol/L from [13C]-propionate breath test could be indicative of varied response of propionate metabolism. As this oxidative pathway of propionyl CoA to succinyl CoA and the Kreb’s cycle, requires at least 6 B-vitamins as co-factors including propionyl-CoA Carboxylase22. Moreover, due to the similarity in propionate metabolism of classic organic acidemias (propionic acidemia and methylmalonic acidemia) as well as defect in Cobalamin C that is involved in B12 metabolism, could lead to various phenotypes and complex metabolic effects in this population23,24. In addition, the metabolomic profiles of the low B12 status participants, before and after treatment, including their interrelationship with known B12 biomarkers and targeted metabolites, were also evaluated to identify other potential diagnostic biomarkers.

Though B12 deficiency is currently assessed from decreased serum concentrations of B12 and HoloTc, and increased concentrations of MMA and Hcy, there is no consensus on the cut-off value for diagnosing vitamin B12 deficiency25. This could explain the discrepancies in the ranges of B12 deficiency prevalences reported over the years, as a single cut-off value was used across different age and gender groups, even though age is known to be a major confounding factor in these concentrations26. A review of the serum B12 cut-offs that have been proposed for the diagnosis of B12 deficiency shows a broad range, between 100 and 350 pMol/L27. Available guidelines recommend the suspicion of B12 deficiency if the serum B12 concentration is < 156 pMol/L7 with confirmation from plasma/ serum Hcy or MMA concentrations28. However, the measurement of serum B12 concentration depends on active HoloTc (B12 bound to transcobalamin) and the biologically inactive holohaptocorrin (B12 bound to haptocorrin) fractions. There can be variability imposed from altered expression of haptocorrin as well as analytical interference to antibodies29,30,31,32,33. Serum HoloTc by itself is considered to be clinically more meaningful than the measurement of total serum B12 concentration as a biomarker of B12 deficiency34, but is limited by its higher cost and availability. With all these caveats, it is widely recognized that there is no single optimal biomarker of B12 deficiency. There is also the need to have dynamic and sensitive indicators, including serum metabolites, of B12 deficiency, as well as the need for non-invasive functional tests in clinical settings.

The functional [13C]-propionate oxidation to 13CO2 requires methylmalonyl mutase and B12 as a cofactor. Our results revealed lower PDRs (13CO2 recovery) in participants with low B12 status, that increased after a single 3 mg intramuscular hydroxocobalamin dose, which is in line with earlier findings by Wagner et al.5, that demonstrated a ~ 23% change in PDR after treatment with 1 mg of cyanocobalamin administered intramuscularly for 5 days with re-testing after 2–6 weeks. Our findings of PDRs are further validated by another Japanese study35 that used one hour protocol for propionate breath test and showed significant higher PDR in post-supplemented group in comparison to pre-supplemented group (41.93 vs 32.84) with higher recovery at 30 min after propionate dosing similar to that of our study and Wagner et al5,35 confirming the peak maxima within 30 min. Therefore, a shorter protocol of one hour with selective breath sampling at 15 min interval might be sufficient for measuring significant differences in propionate oxidation as a functional index of deficiency in response to B12 supplementation as a non-invasive approach to diagnose B12 deficiency.

In participants with low B12 status, before and after supplementation, we also used untargeted serum metabolomics as a readout based on the hypothesis that certain metabolites or combinations of metabolites could exist to indicate or assess B12 deficiency in humans. Previously, the serum B12 metabolome was studied in a Chilean elderly population by Alex et al.36, showing perturbations in serum acylcarnitine, plasmalogens, phospholipids, lyso-phospholipids, and sphingomyelins after a single intramuscular dose of 10 mg cyanocobalamin. However, the injected dose that was used also contained 100 mg of thiamine and pyridoxine, which may have contributed to the metabolomic changes observed in this study. Even so, this study revealed connections between B12 status and serum metabolic markers of mitochondrial function, myelin integrity, oxidative stress, and peripheral nerve function, that is, acylcarnitine and plasmalogens. However, the lack of a control group, a smaller sample size, and the combined supplementation of pyridoxine and thiamine, could have limited some of the possibly meaningful pre- and post-supplemented metabolomic responses to B12 supplementation alone in this population. Furthermore, analysis of targeted metabolomic data from an earlier transgenic B12 deficient mouse model study revealed that most of the downregulated metabolites were linked to protein, amino acid, betaine, bile acid, and purine metabolism37.

The branched-chain amino acids, odd-chain fatty acids, cholesterol and propionate gets catabolized to succinyl-CoA, using B12 and methylmalonyl-CoA mutase with an intermediate step of methylmalonyl-CoA, to be used in the tricarboxylic acid (TCA) cycle in the mitochondria. In this study, we detected several metabolites that were linked to known metabolic pathways such amino acids (sulfur and branched chain amino acids), peptides, acylcarnitines, phospholipids, fatty acid oxidation, that are similar to that of Chilean36 study except the bile acid metabolites that were found to be altered during deficiency as well as in response to supplementation. The increase in methionine, valine, iso-leucine, acylcarnitines and decreased propionylcarnitine levels after supplementation reflect an improvement in one-carbon and TCA cycle metabolites including mitochondrial function.

The reduced activity of methylmalonyl-CoA mutase leads to increased methylmalonic acid and propionyl-CoA, that is further converted to propionylcarnitine. The alterations in the acylcarnitine metabolism are not only linked to the fatty acid transport but also plays an important role in regulating the intracellular sugar and lipid metabolism38. Since the accumulation of propionylcarnitine (C3) which is a marker of methylmalonic aciduria and propionic acidemia, the utility of assessing ratio of C3/C2 i.e., propionylcarnitine to acetylcarnitine, is considered to provide more information about propionate metabolism during deficiency and can help differentiate between various metabolic disorders affecting propionate metabolism including inherent defects/ deficiencies in enzymes involved in propionate pathway39,40. Hence, after B12 treatment, the ratio of C3/C2 can be helpful to assess the effectiveness of intervention during B12 deficiency such that a decreased ratio would indicate improved propionate metabolism due to restored vitamin B12 function. The decreased propionylcarnitine (C3) and its ratio to acetylcarnitine (C2) and palmitoylcarnitine (C16) after treatment in this study suggests incorporating these metabolites for screening B12 deficiency as potential biomarker as an indicator of improved propionate metabolism that was also demonstrated by the propionate breath test.

Here we described how B12 deficiency alters the biochemical features interlinking functional pathways and various other metabolic pathways facilitating further understanding of B12 pathophysiology. However, the findings of this study are limited to a smaller sample size of the adult male population as studied for the functional propionate breath test and metabolomic assessment that limits the generalizability of these observations at population level and needs further evaluation to assess gender-based differences. Since this study did not include liver function tests, the ability to fully characterize bile acid and vitamin B12 metabolism is limited, warranting the inclusion of relevant hepatic assessments in future research.