Calcitriol Exerts Prophylactic Anti-Mycobacterium Effect In A Dose-Dependent Manner

Vitamin D was empirically applied for Tuberculosis (TB) treatment in the past, and is currently used as an adjuvant for TB therapy. Although an increasing pile of evidences suggests that vitamin D has no therapeutic effect against TB infection, the prophylactic effect of vitamin D in preventing TB remains largely undetermined. To experimentally valuate the potential prophylactic effect of calcitriol (the active form of vitamin D) against mycobacterium infection, we performed dose-gradient calcitriol soaking in 30-day-old zebrash before Mycobacterium marinum (M. marinum) challenge through tail vein injection. 1H-NMR metabolomics analysis was further performed for illustration of potential mechanisms underlying the prophylactic effect of calcitriol against M. marinum. The results suggested that calcitriol exerts dose-dependent prophylactic anti-mycobacterium effects, i.e., the bacterial load and the corresponding inammatory factors (IL-1β, TNF-α, and IFN-γ) expressions in M. marinum challenged zebrash were reduced by low-dose (25 µg/L) or high-dose (2500 µg/L) calcitriol soaking, rather than by moderate-dose (250 µg/L) calcitriol soaking. Body weight of the M. marinum challenged zebrash was recovered by high-dose prophylactic calcitriol soaking rather than by low-dose or moderate-dose calcitriol. The 1H-NMR metabolomic proling identied 29 metabolites with altered abundance among the dose-gradient calcitriol groups, among which 22 metabolites were co-varied with the dose of calcitriol, the rest 7 metabolites were co-varied with the bacterial load and the inammatory response in term of cytokine expression. Further pathway analysis indicated that the glycine, serine, and threonine metabolism pathway was the activated in both of the two metabolite groups, indicating that the pathway was altered by dose-gradient of calcitriol and was in response to M. marinum infection in zebrash. The results of the present study suggested that the activation of glycine, serine and threonine metabolism pathway may play a potential role for the dose-dependent anti-mycobacterium effect induced by prophylactic calcitriol soaking.


Introduction
Vitamin D, with calcitriol (1α,25-Dihydroxyvitamin D3) as its major active form, is well known for calcium homeostasis [1,2]. De ciency of vitamin D has been linked with establishing tuberculosis (TB) [3][4][5], and is highly prevalent in pulmonary TB patients [6,7]. Supplementation of Vitamin D in the medication of pulmonary TB is thus believed to be e cacious and extensively popularized especially before the application of antibiotics in controlling TB [8,9]. However, results of recent large-cohort trials did not support the effectiveness of Vitamin D in TB treatment [10][11][12][13][14]. Similar outcomes were obtained between Vitamin D and the placebo, suggesting limited effects of Vitamin D on TB treatment [15,16].
Although the therapeutic effect of Vitamin D on TB is suspicious, its prophylactic anti-TB effect and the underlying mechanism remain largely unknown. Meanwhile, dozens of researches have reported the dose-dependent effects of Vitamin D on oxidative stress and apoptosis [17], immunomodulation[18], volumetric bone density and bone strength [19], as well as trans-differentiation of muscle to adipose cells [20]. Whether Vitamin D effectively exert a dose-dependent prophylactic anti-TB effect is undetermined.
To experimentally valuate the potential prophylactic anti-mycobacterium effect of Vitamin D, we performed dose gradient calcitriol soaking in adult zebra sh before mycobacterium marinum (M. marinum) challenge. 1H-NMR metabolomics analysis was further applied to reveal the pathways activated in M. marinum infected zebra sh receiving prophylactic calcitriol soaking.

Materials And Methods
Study design and Calcitriol soaking.
AB strain zebra sh (Danio rerio) was bought from China Zebra sh Resource Center. Embryos were collected from natural crosses and raised in E3 media (0.33 mM MgSO4 7H 2 O, 0.33mM CaCl 2 2H 2 O, 0.17mM KCl, 5mM NaCl), with a 14h light/10h dark cycle at 28.5℃. One-month old (30 days post fertilization, 30 dpf) zebra sh was raised under a soaking media of calcitriol (25,250, 2500 µg/L) in E3 media with a replacement frequency of 12h for one month. The M. marinum strain ATCC 927 was injected to the 2-month-old (60 dpf) zebra sh through tail vein with a dose of 100 CFU/ sh. The M. marinum infected zebra sh was further raised for half a month and suicided at 75 dpf. The suicided shes were weighted and stored in a -80℃ deep freezer. A detailed schematic diagram of the experimental design was shown in Fig. 1a.

1H-NMR metabolomics
Fish sample preparation for 1H-NMR metabolomics was performed according to a previously reported procedure with some modi cations [21]. Brie y, one sh was mixed with 1 ml D 2 O containing 0.05% TSP (3-trimethylsilyl-[2,2,3,3-D4]-propionate, as internal standard), and homogenized in an ice-water bath with the IKA T10 Basic ULTRA-TURRAX disperser (IKA, Germany), A total of 550 µl supernatant from each sample was obtained through centrifuging of the homogenate at 4℃, 15 000 rpm for 15 min, and transferred into a 5 mm NMR tube for further 1H-NMR spectral pro ling. 1H-NMR spectrometry was pro led by using a Bruker 600 MHz AVANCE III NMR spectrometer (Bruker BioSpin, Germany) with a NOESYGPPR1D pulse sequence and the following parameters: spectral width 12345.7 Hz, spectral size 65536 points, pulse width 40.5 µs, relaxation delay 1.0 s, 64 scans. Spectra processing was performed by MestReNova (v8.0.1, Mestrelab Research, Spain) with manually corrected phase and baseline. The chemical shift of TSP was calibrated at 0.00ppm. and the spectral region of δ 0.16-9.58 was segmented at 0.01 ppm width after exclusion of the region of residual water δ 4.60-5.20. The obtained NMR data was normalized to the total sum of spectra, log-transformed and auto scaled (mean-centered and divided by the standard deviation of each NMR feature) before further analysis.
Multivariate statistical analysis was performed by using SIMCA-P (v14.1, Umetrics AB, Sweden). Principle Component Analysis (PCA) was applied for investigation of the natural separation among sample and for exclusion of the outliers using the Hotelling's T 2 statistic. Orthogonal Projection to Latent Structures Discriminant Analysis (OPLS-DA) was performed to interpretate types of variation with between-group discriminating power by incorporating the grouping information. The best-t OPLS-DA model was validated by a cross-validation of all models using a 200-step permutation test. The parameters R2Y and Q2 were applied for assessment of the tting validity and the predictive ability of the selected OPLS-DA model, respectively. Metabolites were identi ed from the 1H-NMR features by searching the human metabolome database (HMDB, https://www.hmdb.ca) and previously published articles with the following parameters: chemical shift, coupling constant, and peak type. Altered metabolites were de ned as metabolites with altered between-group abundances, and simultaneously meet the following criteria: Importance for the Projection (VIP) > 1 in the OPLS-DA model and false discovery rate (fdr)-adjusted P < 0.05 in an independent-sample T-test. Metabolic pathway analysis was performed by the pathway analysis module implemented in the web portal of MetaboAnalyst (https://www.metaboanalyst.ca). The Danio rerio (zebra sh) (KEGG) library was applied for pathway analysis with the following algorithms: hypergeometric test for over representation analysis, relative-betweenness centrality for pathway topology analysis.

Statistical analysis
Between-group statistical analyses (implemented in GraphPad Prism v8.0) were performed by two-tailed paired t-test for body weight and bacterial load, non-parametric Mann-Whitney test for the relative expression of in ammatory factors. P < 0.05 was de ned as statistically signi cant for testing of body weight, bacterial load, and relative expression of in ammatory factors. P values from the t-test of 1H-NMR spectra pro ling were adjusted by false discovery rate (fdr) with R package vegan for multiple testing correction. The fdr-adjusted P value < 0.05 was de ned as statistically signi cant.

Results
Prophylactic calcitriol soaking exerted dose-dependent anti-mycobacterium effects in adult zebra sh Calcitriol has been reported to have therapeutic effect against mycobacterium infection [22,23]. To determine if this effect is prophylactic and dose-dependent, one-month old (30 days post fertilization, dpf) adult zebra sh were rstly treated with calcitriol soaking for one month, then were challenged with M. marinum through tail vein injection (at 60 dpf), and were monitored for another half a month (61-75 dpf) ( , moderate-dose calcitriol group was far away from low-dose and high-dose calcitriol groups along the X-axis (explained 66.7% of the variation in the independent variable X (R2X = 0.677)), and clustered together along the Y-axis (explained 83.3% of the variation in the categorical variable Y (R2Y = 0.833)). While low-dose group and high-dose calcitriol group was also separated with a relatively high with-in group variation (Y-axis) in high-dose calcitriol group. These results suggested that larger metabolic shift occurred in moderated-dose calcitriol group than in high-dose calcitriol group of M. marinum challenged zebra sh.
To determine the metabolites contributing to the metabolomic shift among different dose of calcitriol, we performed pair-wised OPLS-DA (Fig. 2c, e, g validated by 200-step permutation tests in Figure S1 a-c) and S-plot analysis (Fig. 2d, f, h). A total of 895 1H-NMR features were pro led, from which 692 features were retained after removal of features with minus relative abundances in over half of the sample. Eight one features with altered between-group abundances among the three groups of calcitriol soaking were obtained, which met the criteria that VIP ≥ 1 and fdr-adjusted P < 0.05 in an independent t-test. Totally 29 metabolites from the 81 features with altered between-group abundances were structurally identi ed (Table 1), among which the abundances of 22 metabolites (Fig. 3) were down-regulated and the abundance of methylguanidine was up-regulated from low-dose across moderate-dose to high dose calcitriol, the abundances of 7 metabolites were up-regulated by moderate-dose calcitriol but down-regulated by high-dose calcitriol (Fig. 3, marked with an asterisk behind the metabolite name). These results suggested that a larger part of 22 metabolites with altered abundances were strictly in response to dose-gradient prophylactic calcitriol soaking, while a smaller part of 7 metabolites were co-varied with the bacterial load and in ammatory factors. Prophylactic calcitriol soaking induced metabolomic shift was associated with its anti-

mycobacterium/anti-in ammatory effects
Because the altered metabolites contributing to the metabolic shift induced by prophylactic calcitriol soaking in M. marinum challenged zebra sh co-varied with the dose-gradients of calcitriol or in ammatory factors, we performed spearman rank correlation analysis to con rm the associations among the altered metabolites, the weight loss, the bacterial load, and the in ammatory factors. The spearman rank correlation matrix (Figure 4a, Table S1-S2) exhibited that TNF-α was positively correlated with Dimethylglycine (ρ =0.60, P < 0.01); the body weight was positively correlated with Methylguanidine (ρ =0.70, P < 0.01) and negatively correlated with L-Allothreonine, Mannitol, N-Acetylgalactosamine, Hypoxanthine, L-Histidine, Creatine, N-Acetylneuraminic acid, Betaine, D-Glucuronic acid, Palmitoylcarnitine, 2-Hydroxybutyric acid, Putrescine, L-threonine, and Guanidoacetic acid (ρ < -0.532, P < 0.01); the bacterial load was negatively correlated with most of the altered metabolites (P < 0.05) with lower correlation coe cient (the absolute ρ value less than 0.53). Among the altered metabolites, Methylguanidine was negatively correlated with most metabolites, the above TNF-α/body weight/bacterial load associated metabolites were positively correlated with each other (Figure 4a, Table   S1-S2, ρ > 0.532, P < 0.01). These results demonstrated that the metabolites contributing to the metabolic shift by prophylactic calcitriol soaking in M. marinum challenged zebra sh correlated with each other and with the phenotypes.
Glycine, serine and threonine metabolism was correlated with prophylactic calcitriol soaking and its antimycobacterium/anti-in ammatory effects To further infer the underlying pathways of the altered metabolites, we performed pathway analysis by using the module implemented in the MetaboAnalyst web portal (Fig. 4b-c). Pathway analysis from the 22 altered metabolites co-varied with the dose-gradients of calcitriol (Fig. 4b) suggested that three pathways were signi cantly altered, including glycine, serine and threonine metabolism (hypergeometric test, P = 9.32E-5), beta-alanine metabolism (hypergeometric test, P = 0.029), and histidine metabolism (hypergeometric test, P = 0.021). Pathway analysis from the 7 altered metabolites co-varied with bacterial load and in ammatory factors (Fig. 4c) suggested that glycine, serine and threonine metabolism was signi cantly altered (hypergeometric test, P = 0.0049). Glycine, serine, and threonine metabolism was predicted by both group of metabolites, with 4 metabolites (Betaine, L-Threonine, L-Allothreonine, and Guanidoacetic acid, Fig. 4d, name in pink) co-varied with the dose gradients of calcitriol and 3 metabolites (Choline, Dimethylglycine, and Creatine, Fig. 4d, name in blue) co-varied with the bacterial load and the in ammatory factors. These results indicated that alteration in glycine, serine and threonine metabolism was induced by prophylactic calcitriol soaking and was in response to the dose-dependent anti-mycobacterium and anti-in ammatory effects, suggesting a potential role of the pathway in hostmicrobe interactions.

Discussion
Although Vitamin D has been proved to promote macrophage-mediated killing of Mycobacterium Tuberculosis, a puzzle still exists in the bene t of Vitamin D in terms of tuberculosis treatment outcomes in light of increasing negative evidences from randomized, double-blind, placebo-controlled trials. Nevertheless, the prophylactic effect rather than the therapeutic effect of Vitamin D in tuberculosis prevention was rarely investigated. Because human trial-based evaluation of the prophylactic effect of Vitamin D is hard to carry out, we applied M. marinum challenged zebra sh as an animal model mimicking tuberculosis infection in human to investigate the prophylactic effect of calcitriol (the active form of Vitamin D) against M. marinum infection. We observed dose-dependent anti-mycobacterium and anti-in ammatory effects of calcitriol in M. marinum challenged zebra sh. Further metabolomic investigation suggest that the glycine, serine, and threonine metabolism pathway may play a role in the prophylactic effects of calcitriol.
A previous report [24] suggested that the anti-mycobacterium activity of calcitriol might operate at helping alveolar macrophages and tissue dendritic cells in preventing the initial implantation and ingestion of bacilli. Thus, a constant circulating level of calcitriol might be important in bolstering of resistance to tuberculosis [25]. Therefore, we applied three doses of calcitriol soaking (25/250/2500 µg/L per day) to investigate its prophylactic effect against M. marinum challenge. In accordance with previous reports [26,27], low-dose and high-dose calcitriol soaking exerted anti-mycobacterium and anti-in ammatory effects. However, moderate-dose prophylactic soaking (250 µg/L per day) unexpectedly had no effect on the bacterial load and even exerted pro-in ammatory effects in M. marinum challenged zebra sh (Fig. 1d-f). This complicated dose-dependent anti-mycobacterium and anti-in ammatory effects of calcitriol may be one of the reasons for the puzzle in the bene t of Vitamin D in terms of tuberculosis treatment outcomes.
We next performed metabolomic analysis to investigate the underlying mechanism of the complicated dose-dependent anti-mycobacterium and anti-in ammatory effects of prophylactic calcitriol soaking. A total of 29 metabolites with between-group altered abundances were identi ed, including 41.8% In accordance with our ndings, Vitamin D combined with one amino acid, L-arginine, is reported to be a potential adjunctive immunotherapy in tuberculosis [28], and the administered dose mattered [29]. Purine metabolism has been correlated with cytokine production in patients with bro-cavernous pulmonary tuberculosis [30]. Fatty acid metabolism was reported to be in uenced by Vitamin D de ciency and consequently increased the risk of tuberculosis infection [31]. N-Acetylgalactosamine, a substrate of arylsulfatase B (N-acetylgalactosamine-4-sulfatase), was proved to contribute to intracellular oxygen signaling and in uence hypoxia [32], mannitol, an inhibitor of CYP2E1 [33], could improve lung function in cystic brosis [34], Deoxyuridine troposphere nucleotidohydrolase, using deoxyuridine as a substrate, is a drug target against mycobacterium infections [35], Carnosine is an endogenous antioxidant widely distributed in excitable tissues[36], owning cytoprotective properties in primary cultured rat hepatocytes [37], Methylguanidine could inhibit lymphocyte transformation in vitro [38]. The concordance between the ndings in previous reports and the present study suggested that the M. marinum challenged zebra sh is a suitable model in mimicking tuberculosis infection in human.
The glycine, serine, and threonine metabolism pathway (Fig. 4b-c) was predicted by the 22 metabolites co-varied with the doses of calcitriol and the 7 metabolites co-varied with the bacterial load and the in ammatory factors (Fig. 3), respectively. According to the variation trends of the two metabolite groups, we speculated that the glycine, serine, and threonine metabolism pathway was regulated by prophylactic calcitriol soaking and was in response to the dose-dependent anti-mycobacterium and anti-in ammatory effects. In agreement with our nding, knockout of Vitamin D receptor resulted in upregulated glycine, serine, and threonine metabolism in murine intestinal microbiome [39]. Inhibiting of Vitamin D through massive small bowel resection could also increase the expression of glycine, serine, and threonine metabolism [40]. Glycine, serine and threonine metabolism could potentiate kanamycin-mediated killing of Edwardsiella piscicida [41], and confound e cacy of complement-mediated killing of di cult-to-treat pathogenic strains of bacteria [42]. The nding of the present study and previous reports suggested that glycine, serine, and threonine metabolism potentially contribute to the anti-mycobacterium and antiin ammatory effects of calcitriol.
As a shortcoming of the present study, how the glycine, serine, and threonine metabolism pathway was altered by calcitriol soaking and responded to M. marinum challenge was not determined, which was essential for understanding the dose-dependent anti-mycobacterium effect of prophylactic calcitriol.
Further studies are recommended to reveal the underlying pathways of calcitriol against mycobacterium infection through glycine, serine, and threonine metabolism.
In summary, the present study observed dose-dependent anti-mycobacterium and anti-in ammatory effects of prophylactic calcitriol soaking in M. marinum challenged adult zebra sh, with moderate-dose calcitriol exerted opposite effects to low-dose and high-dose calcitriol in bacterial load abundances of and in ammatory factors. Further metabolomic analysis indicated that the glycine, serine, and threonine metabolism pathway was induced by prophylactic calcitriol soaking and was in response to the dosedependent anti-mycobacterium and anti-in ammatory effects of calcitriol.

Declarations
Availability of data and materials The dataset supporting the conclusions of this article is include within the article.

Ethics approval and consent to participate
This study was conducted in accordance with the Declaration of Helsinki. The experimental procedures were approved by the ethics committee at Shanxi University.

Consent for publication
Not applicable. Experimental design diagram and phenotypic variations in M. marinum challenged adult zebra sh receiving prophylactic calcitriol soaking. (a) the experimental design of the present study was as follows: 30 tails zebra sh were randomly divided into ve groups (6 tails per group); four groups received calcitriol soaking from 30 dpf (day post fertilization) to 60 dpf with doses of 0, 25, 250, 2500 μg/L per day, respectively. The rest one group zebra sh was taken as control. At the time point of 60 dpf, the four zebra sh groups received calcitriol soaking were challenged with M. marinum (100 CFU/ sh) through tail vein injection, and monitored for 15 days. At the time point of 75 dpf, all of the ve groups of zebra sh were suicided and stored at -80 ℃ for further investigation. Body weight (b), bacterial load (c), relative abundances of IL-1β (d), IFN-γ (e) and TNF-α (f) in samples collected at 75 dpf were measured and expressed with Violin plot. The violin plot outlines illustrate kernel probability density, with the top and bottom edges de ne the 95% con dence interval. The dashed lines inside a plot indicate 75% quartile, median, and 25% quartile of the corresponding data from top to bottom. Between-group statistical signi cances were evaluated by unpaired t-test with Welch's correction (for body weight and bacterial load) or Mann-Whitney test (for the relative abundances of in ammatory factors). * P < 0.05, ** P < 0.01, *** P< 0.001.  ). The S-plot shows covariance coe cient (p) and correlation coe cient (p(corr)) of the features from 1H-NMR metabolomic pro ling. Each point in the S-plot represents a 1H-NMR feature and is colored according to the value of correlation coe cient. Cut-off values for p > |0.1| and p(corr) > |0.532| were used.

Figure 3
Relative abundances of the altered metabolites among M. marinum challenged zebra sh receiving different doses of calcitriol soaking. The altered metabolites were screened by pairwise comparison among different groups (low-/moderate-/high-dose of calcitriol soaking) of M. marinum challenged zebra sh. Metabolites with relative abundance ranging from 0.001 to 0.003 (a), from 0.002 to 0.006 (b), from 0.01 to 0.04 (c), from 0.005 to 0.015 (d), from 0.01 to 0.08 were displayed separately. Metabolite name following an asterisk indicates that the metabolite was co-varied with the bacterial load and in ammatory factors. Between-group statistical signi cances were evaluated by Mann-Whitney test. * P < 0.05, ** P < 0.01, *** P< 0.001.