Comprehensive Metabolomics Study To Assess ... - ACS Publications

Subscriber access provided by UNIVERSITY OF ADELAIDE LIBRARIES

Article

A Comprehensive Metabolomics Study to Assess Longitudinal Biochemical Changes and Potential Early Biomarkers in NOD Mice that Progress to Diabetes Peter Buchwald, Alejandro Tamayo-Garcia, Sivapriya Ramamoorthy, Marta Garcia-Contreras, Armando J. Mendez, and Camillo Ricordi J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/acs.jproteome.7b00512 • Publication Date (Web): 11 Aug 2017 Downloaded from http://pubs.acs.org on August 12, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Proteome Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

A Comprehensive Metabolomics Study to Assess Longitudinal Biochemical Changes and Potential Early Biomarkers in NOD Mice that Progress to Diabetes

Peter Buchwald1,2,*, Alejandro Tamayo-Garcia1, Sivapriya Ramamoorthy3, Marta GarciaContreras1, Armando J. Mendez1, and Camillo Ricordi1

1

Diabetes Research Institute and 2Molecular and Cellular Pharmacology, Miller School of Medicine, University of Miami, Miami, FL, USA; 3Metabolon, Inc., Durham, NC.

*

Correspondence to Peter Buchwald ([email protected]), Diabetes Research

Institute, University of Miami, Miller School of Medicine, 1450 NW 10th Ave., Miami, FL 33136, USA; Tel. (1) 305 243-9657.

1

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ABSTRACT A global non-targeted longitudinal metabolomics study was carried out in male and female NOD mice to characterize the time-profile of the changes in the metabolic signature caused by onset of type 1 diabetes (T1D) and identify possible early biomarkers in T1D progressors. Metabolomics profiling of samples collected at five different time-points identified 676 and 706 biochemicals in blood and feces, respectively. Several metabolites were expressed at significantly different levels in progressors at all time-points, and their proportion increased strongly following onset of hyperglycemia. At the last time-point, when all progressors were diabetic, a large percentage of metabolites had significantly different levels: 57.8% in blood and 27.8% in feces. Metabolic pathways most strongly affected included the carbohydrate, lipid, branched-chain amino acid, and oxidative ones. Several biochemicals showed considerable (>4×) change. Maltose, 3-hydroxybutyric acid, and kojibiose increased, while 1,5anhydroglucitol decreased more than 10-fold. At the earliest time-point (6-week), differences between the metabolic signatures of progressors and non-progressors were relatively modest. Nevertheless, several compounds had significantly different levels and show promise as possible early T1D biomarkers. They include fatty acid phosphocholine derivatives from the phosphatidylcholine sub-pathway (elevated in both blood and feces) as well as serotonin, ribose, and arabinose (increased) in blood plus 13-HODE, tocopherol (increased), diaminopimelate, valerate, hydroxymethylpyrimidine, and dulcitol (decreased) in feces. A combined metabolic signature based on these compounds might serve as an early predictor of T1D-progressors. 2


Page 2 of 56

Page 3 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Keywords: 1,5-anhydroglucitol; beta hydroxybutyric acid; branched-chain amino acids; diabetes mellitus; gender differences; hyperglycemia; kojibiose; metabolomics; non-obese diabetic (NOD) mice; sorbitol.

INTRODUCTION People affected by type 1 diabetes (T1D) lose their ability to produce insulin, as their βcells are destroyed by an autoimmune process

1-4

, and they require life-long administration of

exogenous insulin. T1D affects about 1 in every 400–600 children and adolescents in the U.S., and it incidence rate ranges from 0.1 (China, Venezuela) to >40 (Sardinia, Finland) per 100,000/year 4-7. As with most other autoimmune diseases, the incidence of T1D is increasing worldwide with an overall rate of about 3% per year, and the age of onset has decreased by 3–5 years over the past decade for yet unknown reasons 5, 8-10. Even a century after the introduction of insulin, T1D still represents a considerable therapeutic need: partly, because of the inconvenience of the ongoing need for administration of exogenous insulin, and partly because chronic and degenerative complications still occur in a considerable fraction of T1D patients, despite considerable improvements in management and care 11, 12. The etiology, pathogenesis, and onset mechanism of T1D are complex. It is now wellrecognized that an interplay of genetic predisposition, yet unknown environmental factors, and probably other stochastic events is needed

13

. It is also becoming increasingly clear that early 3



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

metabolic perturbations are involved in T1D development and progression, and they might even serve as early disease indicators (14-17 and references therein). To further elucidate the connections among metabolic perturbations, mounting of the autoimmune attack, and destruction of pancreatic β-cells, we carried out a comprehensive global non-targeted longitudinal metabolomics study in non-obese diabetic (NOD) mice, a common animal model of type 1 diabetes (T1D), to characterize the time-profile of the changes in the metabolic signature caused by onset of diabetes and identify possible early biomarkers separating T1D progressors from non-progressors. The NOD model, an inbreed strain developed from a line originally intended as a cataract-prone mouse strain

18

, is the most widely used animal model of T1D, and it has

contributed significantly to the study of this disease

19-23

. These mice develop spontaneous

diabetes in a manner that reproduces many crucial aspects of the human disease (e.g., the presence of islet-specific autoantibodies, inflammation of pancreatic islets, and dependence on MHC alleles), but there are also important differences (e.g., more severe insulitis, gender bias, etc.)

21, 23

. In NODs, about 60–90% of females, but only about 20–50% of males develop

diabetes between the ages of 12 to 30 weeks. Diabetes development can be altered by many factors, and specific pathogen-free conditions are needed for such high incidence rates. Insulitis, the invasion of the pancreatic islets by a mixed population of leukocytes, starts around 4 weeks of age and amplifies progressively until the onset of overt diabetes involving various immune cells (T and B cells, macrophages and dendritic cells), which can take on the organization of typical tertiary lymphoid structures 24. Peri-islet infiltration is minimal in young 4


Page 4 of 56

Page 5 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


NODS, but it becomes prominent around 9–10 weeks of age

24, 25

. By most estimates, at the

time of hyperglycemia onset, NOD mice have already lost 80–90% of their insulin-producing βcell mass

26, 27

, a proportion similar to that in humans at the time of clinical manifestation of

T1D 1, 28. The progressive loss of β-cell function is a central aspect of T1D, and declining insulin responses have been observed in prediabetic NOD mice studied until the animals were 14 or 18 weeks of age

29, 30

, around the onset of hyperglycemia. Recently, in a longer study of up to 42

weeks of age, we have shown that β-cell function (insulin secretion) decreases with age in all NODs up to 28 weeks of age paralleling diabetes onset, and then it essentially disappears in diabetic animals, while it stabilizes at still adequate levels to maintain normoglycemia in nonprogressors 25. Some metabolomics studies on T1D onset 31-33 and the metabolic perturbations it causes in children 34-37, as well as in animal models such as the NOD mice 38-41, have been performed in the last decade. Most, including studies on pancreatic islets, have been summarized relatively recently 14-17. An important limitation of these studies was the need for relatively large amounts of samples, e.g., several million cells for LC-MS type analyses (e.g., >3 million)

16

, but with

improved analytical methods this is no longer a serious issue. The method employed herein required 150 µL of blood and 200 mg (wet weight) of fecal samples making possible the longitudinal collection of samples from the same animals. This allowed more clear separation of the differences between the progressors and the non-progressors as changes could be tracked within the same animals, and also made it more likely to be able to identify early biomarkers than previous approaches. There are already more than 75,000 metabolites registered in the 5



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 56

Human Metabolome Database (HMDB; http://www.hmdb.ca) of which more than 20,000 have been detected in the blood

42

. Hence, with current methods, very detailed metabolomic

signatures can be obtained. As mentioned, NOD mice have been used in some previous metabolomic studies. One of the earliest studies had been reported by Sysi-Aho, Oresic, and co-workers

38

. It involved, a

longitudinal lipidomics study using serum samples, which indicated that female NOD mice that later progress to autoimmune diabetes exhibit the same lipidomic pattern as prediabetic children. The study also involved a separate metabolomics analysis of isolated pancreatic islets, and found several metabolites showing significant differences between animals judged as being at high versus low risk of developing diabetes. Madsen and co-workers compared blood samples from female NOD and C57BL/6 (wild type) mice and found that NOD mice had a distinct metabolic profile, which included increased glutamate and decreased TCA cycle metabolites, and partly resembled what was previously observed in children progressing to T1D 39

. More recently, the group of Hara and co-workers

40, 41

performed global metabolomics

analyses in two consecutive studies, but they compared a group of diabetic vs. non-diabetic NODs without doing longitudinal monitoring of the same mice. Hence, none of these were complete non-targeted metabolomics studies on longitudinal samples in NODs. In the present study, we collected detailed biochemical data from two matrices (blood and feces) in parallel and tracked changes with time in each animal for several hundred metabolites.

6


Page 7 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


MATERIALS AND METHODS Animal Care and Treatment Female non-obese diabetic (NOD) mice (NOD/ShiLtJ; 4 weeks old; n = 30, 15 M + 15 F) were obtained from Jackson Laboratories (Bar Harbor, ME). Starting from week 10, glycosuria was monitored one or two times a week. In animals that turned positive, blood glucose levels (glycemia) were monitored two to three times a week. Animals with elevated glucose levels (nonfasting glycemia >300 mg/dL) on three consecutive days were identified as hyperglycemic and were designated as T1D progressors. Hyperglycemic animals were implanted with sustained release insulin pellets (LinBit, LinShin Canada, Inc., Toronto, Canada) to avoid severe hyperglycemia and minimalize animal suffering. One or two pellets were used per animal and renewed as needed after 2–3 weeks as indicated by the glucose levels. All mice were sampled for whole blood and fecal pellets at 6, 11, 16, 21, and 26 weeks of age. Blood (150 µL blood, EDTA) and fecal material (200 mg wet weight) were collected and stored at –80°C. At the end of the study, all samples from 6 (3F + 3M) non-progressors, which remained non-diabetic, and 8 (4F + 4M) progressors, which developed T1D, selected as representative were sent for analysis to Metabolon (Durham, NC) (total samples: 5 time-points × 2 samples × 14 animals = 140). At the end of the study, all animals were euthanized, and pancreas and other organ samples were collected and stored for analyses. All animal studies were reviewed and approved by the University of Miami Institutional Animal Care and Use Committee. All procedures were conducted according to the guidelines of the Committee on Care and Use of Laboratory Animals, Institute of Laboratory Animal Resources (National Research Council, Washington DC). 7



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 56

Animals were housed in microisolated cages in Virus Antibody Free rooms with free access to autoclaved food and water at the Department of Veterinary Resources of the University of Miami. Immunofluorescence Immunostaining was performed as described previously

25

. Briefly, harvested tissues

were embedded in Tissue-Tek O.C.T. (Sakura Finetek, Torrance, CA) and snap frozen by placing the tissue mold in dry ice. Tissue sections (5–10 µm) were collected on glass slides and kept at – 80°C until analysis. Samples were fixed for 10 min in 4% paraformaldehyde, washed three times with PBS 1× and permeabilized with PBS plus 0.1% Triton X100 for 1 h at room temperature. After blocking for at least 1 h in PBS with 10% Power Block (BioGenex, Fremont, CA) or normal goat serum as required by antibody detection, samples were incubated overnight at 4°C with the primary antibodies following the manufacturer’s recommendations. Immunofluorescence staining was performed for insulin (1:100; Dako, Carpinteria, CA) and glucagon (1:100; Dako). Sections were washed five times with PBS plus 0.05 Tween20 and then incubated with the secondary antibodies, washed another five times with PBS plus 0.05 Tween20 and twice with PBS 1×, and mounted for analysis via confocal microscopy. Secondary antibodies were AlexaFluor-conjugated: goat anti-rat 568; goat anti-rabbit 488; goat anti-guinea pig 647 (1:800; Life Technologies, Carlsbad, CA). Nuclei were detected by DAPI staining (Life Technologies). After final washes, slides were mounted using Vectashield (Vector Laboratories, Burlingame, CA). Immunofluorescence imaging was performed at the Diabetes Research Institute Analytical Imaging Core Facility. Images were captured using a Leica SP5 inverted confocal microscope 8


Page 9 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


with motorized stage as tiled single-layer (5× air) for whole pancreatic sections. Representative images were merged and edited for contrast in Adobe Photoshop (Adobe Systems, Inc., San Jose, CA). Metabolomic Analysis Metabolomics analysis was performed in a manner similar to that of our previous study with human pancreatic islets

43

. In brief, individual blood and fecal samples were loaded in

equivalent volumes and weights across the platform with no additional normalization performed prior to statistical analysis. Small molecules were extracted in a methanol solution containing process assessment standards as described previously

44

. The resulting clarified

supernatant extract was divided into five aliquots, one for each of the four individual liquid chromatography – mass spectrometry (LC–MS) analyses and one spare, briefly evaporated to remove the organic solvent and stored overnight under nitrogen before preparation for analysis. The global biochemical profiling analysis was composed of four unique arms consisting of reverse phase chromatography positive ionization methods optimized for hydrophilic compounds (LC–MS Pos Polar) and hydrophobic compounds (LC–MS Pos Lipid), reverse phase chromatography with negative ionization conditions (LC–MS Neg), as well as a HILIC chromatography method coupled to negative (LC–MS Polar) 45. All of the methods alternated between full scan MS and data-dependent MSn scans. The scan range varied slightly between methods but generally covered 70–1000 m/z. Experimentally

detected

metabolites

were

identified

by

matching

the

ion

chromatographic retention index, accurate mass, and mass spectral fragmentation signatures 9



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 56

with a reference library consisting of over 3,200 entries created from authentic standard metabolites under the identical analytical procedure as the experimental samples 46. In-house peak detection and integration software was used. The data output from the software was a list of m/z ratios, fragmentation spectra, retention indices, and area under the curve (AUC) values. All proposed identifications were then manually reviewed and hand curated by an analyst who approved or rejected each identification based on the criteria above 46. Statistical Analysis Two types of statistical analyses were performed:

(1)

significance tests and

(2)

classification analysis. Standard statistical analyses were performed in ArrayStudio on logtransformed data. For analyses not standard in ArrayStudio, the programs R (http://cran.rproject.org/) and JMP were used. Following log transformation and imputation of missing values, if any, with the minimum observed value for each compound, ANOVA contrasts with a repeated measures component were used as significance test to identify biochemicals that differed significantly (p < 0.05) between experimental groups. An estimate of the false discovery rate (q-value) was calculated to take into account the multiple comparisons that normally occur in metabolomic-based studies. Classification analyses used included principal components analysis (PCA), hierarchical clustering, and random forest. For the scaled intensity graphics, each biochemical on the original scale (raw area count) was rescaled to set the median across all animals and time-points equal to 1. For the visualization of interaction networks and biological pathways of the significant metabolites, Metabolon’s MetaboLync plugin to Cytoscape (http://cytoscape.org) was used, an open-source software platform 47. 10


Page 11 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


RESULTS AND DISCUSSION Overview of the Metabolomics Data Blood and fecal pellet samples were collected at predefined time-points (6, 11, 16, 21, and 26 weeks of age) from NOD mice (15 males and 15 females) that were monitored and designated as progressors (type 1 diabetic, T1D) or non-progressors (control) depending on whether they became hyperglycemic or not by the end of the study (Supporting Information Figure S1). Diabetes onset in these mice followed a pattern typical for NOD mice, with diabetes starting to occur around week 12, and approximately 60% of female and 30% of male mice becoming diabetic by week 30. The corresponding Kaplan-Meier survival curves are shown in Figure 1 with a couple of onset rates from our previous studies 25, 48 included for comparison. At the end of the study, all samples from 6 (3F + 3M) non-progressors (control) and 8 (4F + 4M) progressors (T1D) selected as representative were subjected to global metabolomics analysis, as described in the Methods section (n = 14 mice × 2 matrices × 5 time-points = 140 samples). Following data acquisition and curation, a total of 676 and 706 compounds of known identity (named biochemicals) were identified in blood and fecal samples, respectively. At each timepoint, there were several metabolites that had significantly different concentrations in progressors as compared to non-progressors (for detailed data, see Supporting Information Table S1). The number of metabolites expressed at significantly different levels increased strongly in the week 21 and 26 samples, when the progressors were diabetic while the nonprogressor controls remained normoglycemic. Figure 2 provides a summary comparison of the metabolites (biochemicals) that exhibited statistically significant changes (p < 0.05) for these 11



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 56

comparisons focusing on the difference between progressors vs. non-progressors at each of the five time-points. In blood, analysis by three-way ANOVA with repeated measures identified biochemicals exhibiting significant interaction and main effects for experimental parameters of gender, group, and time point. In feces, analysis by two-way ANOVA with repeated measures identified biochemicals exhibiting significant interaction and main effects for experimental parameters of group and time point. Overall Data Segregation: Principal Component, Hierarchical Cluster, and Random Forest Analyses Principal component analysis (PCA) projects differences across a large set of variables onto a smaller set of principal variables, and, by reducing dimensionality, it allows the visualization of the clustering of individual samples

49

. Transformations are done so that the

first principal component (PC1) has the largest possible variance (meaning that it accounts for as much variability as possible along a single dimension), and then each additional component has the highest remaining variance possible, while also being orthogonal to all preceding components. Here, PCA was used to assess whether progressors could be segregated from nonprogressors at the various time-points based on the differences in their overall metabolic signatures. The first two principal components account for a relatively modest portion of the variability in the data (18.1 + 14.5 = 32.6% and 23.5 + 10.7 = 34.2% in blood and feces, respectively); nevertheless, they show a significant clustering of both the blood and fecal data (Figure 3).

12


Page 13 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


PCA showed clear segregation of the week 26 time-points from the rest of the timepoints (weeks 6, 11, 16, and 21), especially in the blood samples, suggesting that by this time there is already a profound impact on global metabolism associated with diabetes progression despite the use of sustained release insulin implants (Figure 3). Note that there is little separation among the samples collected at different time points in NOD control mice suggesting stable metabolic phenotypes with aging in these normoglycemic controls. Interestingly, there is a clear separation by gender (male vs. female) in the blood samples along principal component 2 (PC2, vertical axis) in both progressors and non-progressors. Nevertheless, the delineation of the week 26 time-points from the rest is present in both females and males mainly along PC1 (horizontal axis) indicating that diabetes induces comparable changes in the global metabolism profile of male and female mice despite their gender-related metabolic differences. PCA of the fecal samples also revealed good separation between the early (week 6 and 11) and late (week 21 and 26) time-points in both progressors and non-progressors reflecting age- and T1D-related differences in global fecal metabolites. A good segregation of the week 21 and 26 fecal samples of the progressor NODs from those at the other time points is clearly present suggesting profound diabetes-mediated differences in global metabolism at these time points (Figure 3B). Hierarchical cluster analysis (HCA), which groups metabolically similar samples close to one another, was also performed on these data to obtain a different perspective into potential group segregation. Overall, HCA showed similar trends as PCA with a clear segregation of the later time-point samples (i.e., week 26) of the progressors (Figure 4). 13



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 56

To identify biomarkers that allow separation among groups, random forest analysis (RFA) was also performed. This is a statistical tool that utilizes a supervised classification technique based on an ensemble of decision trees

50

. Using primary grouping of progressors

(type 1) and non-progressors (control) at the five different time-points, RFA of the blood and fecal metabolic profiles resulted in 47% accuracy in differentiating the groups in both matrices (Supporting Information Figure S2). These are much better than random chance alone (10% accuracy for 2×5 = 10 groups) indicating that the differences in the biochemical profiles between the two groups were pronounced. However, the predictive accuracy for the blood samples was mainly driven by the low class error at the week 26 time-point between progressors and non-progressors, and the one for the fecal samples was mainly mediated by the low class errors at the week 6 and 26 time-points in progressor mice (Figure S2). RFA of the week 26 blood samples was able to classify T1D from control group with a predictive accuracy of 100% (data not shown) implying profound differences in the metabolic phenotype at this time point when all progressors were diabetic. RFA also ranks biochemicals by their importance to the classification scheme, and the 30 top-ranking metabolites in each matrix are summarized in Figure S2. As indicated, they span several different metabolic pathways. In blood samples, glucose and other biochemicals related to carbohydrate metabolism as well as established markers of glycemic control (1,5-anhydroglucitol) and lipid metabolites were consistently identified as contributing the most to the separation of groups.

14


Page 15 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Changes in Diabetic NOD Mice At the last sampling (week 26), all progressors were already hyperglycemic for at least four weeks (Figure 1) so this comparison can be considered as reflecting differences between diabetic and nondiabetic NOD mice. A large proportion of the metabolites detected had significantly different levels in the T1D group vs. the nondiabetic controls (Figure 2): 57.8% in blood (266 ↑ + 125 ↓ = 391 of the 676 detected) and 27.8% in feces (82 ↑ + 114 ↓ = 196 of the 706 detected). This indicates that the overall metabolic landscape was profoundly altered by diabetes onset, which is even more remarkable as hyperglycemia in these animals was at least partially controlled by use of sustained release insulin implants. Many of these compounds showed considerable changes. Compounds that showed statistically significant (p < 0.05) and at least four-fold changes either upward or downward versus the nondiabetic controls at the last sampling are summarized in Table S2. The longitudinal time-profiles of their average fold changes are shown in Figure 5 as an indication of their changes as T1D starts to develop. Most of these changes are likely due to the onset of hyperglycemia and not of the autoimmune reaction itself, but clear delineation is difficult with the present data. As a possible indication, we looked at the correlation of their concentrations with that of glucose across all 5×14 = 70 individual data-points; the corresponding r2 values are included in Table S2. For the majority of these strongly changed metabolites, their blood concentrations indeed correlate closely with that of glucose (r2 > 0.3); those that do not and, hence, are less likely to be hyperglycemia related changes include, for example, reduced glutathione (GSH), 4-hydroxy-nonenal-

15



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 56

glutathione, xanthosine, taurodeoxycholate, β-muricholate, and nicotinamide N-oxide (Table S2). As a further verification, we also calculated the fold changes in T1D NODs versus their corresponding values at week 6 in the same mice for both blood and fecal samples. Metabolites showing again at least 4-fold change in T1D that change no more than 1.33-fold in control animals are summarized in Supporting Information Figure S3. In agreement with the trends shown in Figure 5, compounds showing the largest fold change in blood include, for example, maltose, kojibiose, 3-hydroxybutyrate (↑) as well as 1,5-anhydroglucitol and β-muricholate (↓). Biochemicals with the largest change in feces include, among others, hippurate, cinnamoylglycine, allantoin, 1-methylnicotinamide, and phenylacetylglycine (↑) as well as ursodeoxycholate sulfate, valylglutamine, and tauro-β-muricholate (↓). Line plots illustrating the average time-profiles from week 6 to 26 for some of the compounds that showed the largest fold-change in diabetic animals (either increase or decrease) are shown in Figure 6. Corresponding spaghetti plots with all individual changes are included in Supporting Information Figure S4. Note, for example, that maltose in blood showed one of the largest increases in diabetes progressors: it was consistently increased more than two-fold at all time-points starting from week 11, and it reached more than 20-fold difference by week 26 (22×, ↑) (Figure 5A, Figure 6). In the meantime, glucose increase was essentially parallel, but it reached just a 3.1-fold difference at week 26, when all progressors are already diabetic. Note, however, that sustained release insulin pellets were used in diabetic mice to

16


Page 17 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


avoid excessive hyperglycemia and minimize animal suffering, and this somewhat limited the differences in glucose. Fewer compounds showed relative decrease in T1D animals with β-muricholate being the only one showing more than four-fold change in both matrices (B: 4.9×; F: 4.5×, ↓) and 1,5anhydroglucitol showing one of the largest fold-changes observed in blood (50×, ↓) as it drops to very low levels in diabetic animals. Overall, diabetes onset significantly affected several metabolic pathways in diabetic NOD mice compared with non-diabetic controls from the same batch; main metabolic pathways affected are illustrated using the Metabolic Pathway Classification Network view of the metabolomic data for both blood and fecal samples in Figure 7 and Figure 8, respectively. These graphs clearly highlight some of the metabolic sub-pathways that were strongly and consistently affected by T1D onset, such as, ‘fructose, mannose, and galactose’, ‘leucine, isoleucine, and valine’, ‘diacelyglycerol’, ‘polyunsaturated fatty acid’, ‘fatty acid monohydroxy’, and others. These changes are in general agreement with previous observations in diabetic NODs (e.g., by Hara and co-workers

40, 41

), and some of the main ones are discussed below. So far,

there are only a limited number of studies comparing metabolic fingerprints between T1D and corresponding normal controls in humans (children) using plasma and/or urine samples 34-37. A recent detailed metabolomics study at Mayo Clinics, which followed up on a smaller earlier study 34, found 154 known metabolites significantly altered even in well-controlled T1D children (glycated hemoglobin < 6.5%,) belonging to 26 pathways including glycolysis, gluconeogenesis, bile acids, tRNA biosynthesis, amino acids, branch chain amino acids, retinol, and vitamin D 17



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

metabolism

37

Page 18 of 56

. An earlier study by Balderas and coworkers found only relatively moderate

changes in the plasma and urine samples of T1D children that were under insulin treatment and good glycemic control

35

. A recent study by Galderisi and coworkers found 59 urinary

metabolites having a higher level in children with T1D including gluco- and mineralocorticoids, phenylalanine and tryptophan catabolites (kynurenine), small peptides, glycerophospholipids, fatty acids, and gut bacterial products 36. Gender differences in the metabolome Since PCA indicated a clear separation by gender in the blood samples along PC2 in both progressor and non-progressor NOD mice (Figure 3, vertical axis), we took a more detailed look to identify the main metabolites responsible for this separation. This is of particular interest as there is gender bias in many autoimmune diseases with females being, in general, more susceptible 51, 52. While there is no significant gender bias for T1D in humans, there is a strong one in NOD mice, where females are much more likely to develop diabetes than males (typically, 60–90% vs. 20–50%)

21, 23

, probably due to roles played by sex hormones in the

autoimmune process 51. Some of the biochemicals quantified here showed considerable gender differences having up to 2–3-fold different blood concentrations (irrespective of the T1D progressor status); illustrative line plots are shown in Figure 9. Compounds with the largest and most consistent differences in blood include, for example, β-alanine, carboxyethyl-GABA, imidazole lactate, hexanoylglycine, 4-cholesten-3-one, several sphingosine derivatives, and tryptophan. There were fewer differences in the fecal samples with 4-cholesten-3-one and 3hydroxyhexanotae showing differences in both blood and fecal samples, and some cholate 18


Page 19 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


sulfates being most different in feces. There seems to be no relationship between these differences and the progressor status: one would expect female non-progressors to have values closer to those of the generally more protected males, but this was not the case for any of these compounds (Figure 9). Changes in carbohydrate metabolism and in biochemical markers of glycemic control and insulin deficiency In addition to significant elevations in glucose (at week 21 and 26; Figure 6), significantly elevated levels of sorbitol, fructose, mannose, and maltose have also been found both in blood and feces. Part of these may be generated upon shuttling of excess glucose into the sorbitol pathway (Figure 10). Some of these compounds such as maltose and 3-hydroxybutyrate in blood (also kojibiose in feces) increased more dramatically in T1D than glucose itself showing more than 10–20-fold increases. Nevertheless, the blood concentration of these compounds correlate closely with that of glucose, the corresponding r2 values being 0.86, 0.75, 0.71, and 0.70 for maltose, fructose, sorbitol, and mannose, respectively, which are some of the highest such r2 values (Table S2). The sorbitol pathway is important for handling excess glucose, but increased activity can deplete the reserves of the anabolic cofactor NADPH resulting in increased oxidative stress, reduced nitric oxide production, and increased protein glycation to form advanced glycation end products (AGEs) 53. Together with the inability of sorbitol to cross cell membranes and the resulting induction of osmotic stress, these negative consequences of the activation of the sorbitol pathway may lead to microvascular diabetic complications. Compared with normoglycemic controls, T1D mice at week 26 also showed somewhat elevated 19



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 56

levels of pyruvate (1.4×) and the glycolytic end product lactate (1.3×) (Figure 10) suggesting possible differences in glucose utilization. Significant reductions in 1,5-anhydroglucitol (1,5-AG), a marker of glycemic control

54

,

together with significant elevations in 3-hydroxybutyric acid (β-hydroxybutyrate, betahydroxybutyric acid, BHBA), a ketone body and an indicator of diabetic ketoacidosis due to insulin deficiency in T1D

55, 56

, which are both used clinically, corroborate the findings of

elevated circulating glucose and infer a signature of insulin deficiency in diabetic animals. 1,5AG is a stable, long-term marker of glycemic control known to exhibit serum levels that are inversely proportional to circulating glucose because it competes with glucose for reabsorption in the kidney. Consequently, urinary excretion of 1,5-AG is increased under hyperglycemic conditions and the serum levels fall. Therefore, the fact that our non-targeted metabolomics approach identified 1,5-AG (50×, ↓) and BHBA (12×, ↑) among the most strongly and significantly changed blood biochemicals in T1D animals (Figure 6) as indicative of higher circulating glucose concentrations nicely confirms their clinical use in T1D. Their blood concentrations also correlate closely with that of glucose giving r2 values of 0.61 and 0.62 for 1,5-AG and BHBA, respectively (Table S2). Interestingly, we also found 2-hydroxybutyric acid (alpha-hydroxybutyrate, AHB) as elevated (3×) in diabetic NODs (week 26). Metabolomics studies have indicated AHB as a marker of insulin resistance in type 2 diabetes (T2D) progressors

57, 58

. AHB can arise from states of lactic and ketoacidosis or disrupted energy

metabolism. It is thought to be a by-product during the formation of alpha-ketobutyrate (AKB) via a reaction catalyzed by lactate dehydrogenase or alpha-hydroxybutyrate dehydrogenase. 20


Page 21 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


This increased conversion of AKB to AHB is thought to occur when either the rate of AKB formation exceeds the rate of catabolism to propionyl CoA (propionylcarnitine) for anaplerotic incorporation into the TCA cycle or when there is product inhibition of the dehydrogenase responsible for AKB conversion 57. Changes in TCA cycle metabolism The dysregulation of glucose metabolism in diabetic NOD mice relative to control mice suggested that oxidative metabolism may also be altered. T1D NODs exhibited an accumulation of TCA cycle intermediates including citrate, α-ketoglutarate, succinate, fumarate, and malate (Figure 7, bottom). The levels of these biochemicals were most strongly affected in T1D blood at week 26 and may be indicative of a disruption in the TCA cycle and/or increased anaplerotic contributions to attempt to replenish TCA cycle intermediates. Hence, these findings suggest the TCA cycle is altered in T1D NOD mice and may be indicative of mitochondrial dysfunction. Changes in lipid metabolism Substantial changes in several aspects of lipid metabolism were observed when comparing diabetic and non-diabetic animals at the 26 week time-point. In particular, accumulation of several free fatty acids, including medium-chain, long-chain, saturated, unsaturated, polyunsaturated, branched-chain, and dicarboxylic fatty acids were observed in diabetic NOD mice compared with non-diabetic control NOD mice. These changes are clearly highlighted by the accumulation of red markers in the corresponding ‘fatty acid’ sub-pathway branches on the lipid super-pathway in Figure 7. Concomitant with the elevations in free fatty acids, increased blood levels of mono- and diacylglycerols along with glycerol (non-significant 21



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 56

increase), a marker of lipolysis, were also elevated. Taken together, these findings are suggestive of insulin deficiency at the level of adipose tissue resulting in increased release of free fatty acids and glycerol. Alternatively, this could also be due to an increase in fatty acid synthesis as evidenced by increases in malonylcarnitine, citrate, and glucose. Moreover, elevations in circulating acetylcarnitine and many short-, medium-, and longchain acylcarnitines in diabetic NOD animals may be associated with increased β-oxidation of fatty acids and/or mitochondrial dysfunction in liver and muscle tissue as conjugation of fatty acids with carnitine is required for transport across mitochondrial membranes and subsequent β-oxidation. Accumulation of acylcarnitines was accompanied by considerable (12×) elevations in the ketone body BHBA, an indicator of diabetic ketoacidosis due to insulin deficiency in T1D, which is used as a clinical test as discussed earlier

55, 56

. Ketones are produced from excess

acetyl-CoA in the liver, generally from β-oxidation of fatty acids, and are utilized by peripheral tissues, such as brain and cardiac muscle, to meet energy demands when glucose is (or is perceived to be) limiting. Of note, circulating ketones are typically elevated with greater fatty acid utilization. Therefore, the observed changes are indicative of a signature of altered lipid metabolism in diabetic animals consisting of altered lipolysis, increased β-oxidation of fatty acids despite some degree of mitochondrial dysfunction, and increased production of ketones – all findings that may be related to the underlying issue of severe insulin deficiency. These changes were largely absent in the progressor animals at week 16, when they were still in prediabetic stage. These findings suggest that the observed perturbations in lipid metabolism may also be associated with the conversion to the diabetic phenotype and that the 22


Page 23 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


development of dyslipidemia is accelerated with the severity of diabetes. Conversely, a significant decrease in fatty acids and carnitine-conjugated fatty acids was observed in the feces of T1D mice at week 26 compared with controls. This suggests concomitant decrease in excretion leading to relative accumulation of fatty acids and their conjugates in the blood (note the accumulation of blue markers in the corresponding ‘fatty acid’ sub-pathway branches on the lipid super-pathway in Figure 8). Changes in branched-chain amino acid metabolism One of the most strongly and thoroughly affected pathways by diabetes onset was the branched-chain amino acid (BCAA) sub-pathway (Figure 11). The BCAAs valine, isoleucine, and leucine constitute a large portion of amino acids incorporated into proteins and also serve as carbon sources to meet energy demands in times of need. BCAAs are important nutrient signals, play a role in modulating insulin secretion at the level of the β-cell, and altered BCAA metabolism has been implicated in the pathogenesis of T2D. In particular, a signature of elevated circulating BCAAs has been reported to be associated with insulin resistance and T2D 59

. As with increased acylcarnitines, accumulation of BCAAs and their degradation products may

be reflective of mitochondrial dysfunction in the liver, muscle, and adipose tissue. However, in parallel with these elevations in blood, a significant decrease in BCAAs and their metabolites was observed in the feces of diabetic NOD mice (week 26; compared with control), suggesting simultaneous decrease in excretion of BCAAs and their metabolites leading to relative accumulation in the blood. Finally, it should also be noted that the BCAAs isoleucine, leucine, and valine together with the aromatic amino acids tyrosine and phenylalanine were the five 23



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 56

amino acids found in human studies to have highly significant associations with future T1D in a metabolite profiling study by Wang, Gerszten, and co-workers 33. This clinical study suggested that fasting concentrations of these amino acids are higher in progressors than in nonprogressors. This was not the case here as none of these amino acids were found to have elevated levels in progressor NOD mice at the earlier time-points (Figure 11). Inflammation and oxidative stress Consistent elevations in several biochemical markers of inflammation and oxidative stress were observed in T1D NOD mice. For example, polyunsaturated fatty acids (PUFAs), such as linoleate (3.7×, ↑) and arachidonate (1.3×, ↑) (Figure 7), and fatty acid-derived inflammatory mediators, such as 12-HHTrE (2.5×, ↑) and 12-HEPE (1.6×, ↑), were increased in T1D NOD mice suggestive of an inflammatory milieu in these animals. It should be noted that these are contrary to the observations of Hara and co-workers in NODs indicating reductions in circulating PUFAs and lipid signaling mediators (arachidonate and derived eicosanoids) implying impaired states of systemic inflammation 41. Kynurenate, which we have recently found to be highly induced by inflammatory cytokines in isolated human islets

43

, was also slightly, but

significantly elevated at week 26 (1.6×, ↑). It is derived from tryptophan by indoleamine 2,3dioxygenase (IDO), an enzyme that is known to be stimulated by pro-inflammatory cytokines, such as interferon-γ and tumor necrosis factor-α. Several markers of oxidative stress including reduced glutathione (GSH; 6.7×, ↑), methionine sulfone (3.6×, ↑) and sulfoxide (1.5×, ↑), 13HODE + 9-HODE (hydroxyoctadecadienoic acid; 6.1×, ↑), 12,13-DiHOME (3.2×, ↑), and 9,10DiHOME (5.2×, ↑) were altered in diabetic animals as well. These changes in biochemical 24


Page 25 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


indicators of oxidative stress were accompanied by elevations in antioxidants, including the dietary-derived compound ergothioneine (1.2×, ↑) and the vitamin E isoforms β-tocopherol (3.3×, ↑). These findings are indicative of increased oxidative stress and inflammation in diabetic NOD mice. With a few exceptions related to inflammation and antioxidant capacity, the above mentioned biochemical markers did not exhibit substantial changes in feces. Changes in gut microbiome derived metabolites Many environmental factors have been implicated in altering diabetes susceptibility in NOD mice including exposure to dietary factors 60 and changes to gut microbiota 61-63. While the exact triggers for autoimmune diabetes are still unknown, the role of the gut microbiota has been of increasing interest. Recent studies have suggested that microbiota regulates T1D through the toll like receptor

63

. Commensal microbes can trigger both pro- and anti-

diabetogenic signals; the former signals control the microbiota, while the latter induce tolerance to self-antigen. Consistent with the literature, difference in microbiome-derived metabolites was observed in diabetic versus non-diabetic NOD mice. In particular, many of the changes observed in fecal samples could be tightly connected to changes in gut microbiome. The observed large change caused by T1D onset in kojibiose (blood: 4.7×, ↑; feces: 37×, ↑) is most likely due to altered microbiome metabolism. Kojibiose is a naturally occurring carbohydrate that is a promising prebiotic (i.e., non-digestible food component selectively fermented in the colon and beneficially affecting the growth and/or activity of bacteria there). It was found to have the highest prebiotic index among disaccharides 64, and it had beneficial effects when fed to hyperglycemic rats 65. 25



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 26 of 56

The metabolism of aromatic amino acids, phenylalanine, tyrosine, and tryptophan occurs in part through the involvement of enzymes encoded within the microbiome. Significant fluctuations in microbial flora-derived tryptophan, 3-(4-hydroxyphenyl)propionate, 3-indoxyl sulfate, phenyllacetylglycine, and indolepropionate in diabetic NODs compared with controls at the later time-points could indicate changes in microbial composition and/or activity with progression of diabetes in these mice. Trimethylamine N-oxide (TMAO), a gut flora-dependent metabolite derived from choline and L-carnitine was significantly (1.7×, ↑) elevated in diabetic NOD mice at week 26. Elevated blood levels of TMAO are linked to increased risk of cardiovascular disease, and the increased levels of TMAO in diabetic NOD mice might be contributed by the altered gut microbiome. Consistent with an altered gut-microbiome, we also observed changes in the benzoate pathway, which is derived from microbiome metabolism of phenolic compounds in the diet. Certain metabolites are the product of gut microbial or mammalian metabolism; others, such as hippurate, are mammalian-microbial ‘co-metabolites’. Altered blood levels of hippurate (2.5×, ↑), catechol sulfate (2.2×, ↑), p-cresol sulfate (2.6×, ↓), hydrocinnamate (2.1×, ↑), 4-hydroxycinnamate (1.9×, ↑), and caffeic acid sulfate (4.7×, ↑) may reflect either reduced intestinal permeability or altered gut microbial metabolism with progression of diabetes in NOD mice. Coupled with changes in gut microbiome-derived aromatic amino acids and TMAO, we also observed perturbed levels of some secondary bile acids in diabetic NODs, which may also indicate alterations in microbiome composition. These T1D mice showed an increase in circulating deoxycholate (2.3×, ↑) and taurodeoxycholate (4.5×, ↑) relative to the control group. 26


Page 27 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Possible Early Biomarkers in T1D Progressors An important goal of the present longitudinal sampling-based work was to identify possible early biomarkers that might distinguish T1D progressors from non-progressors. For this, we compared the metabolic signatures of progressors and non-progressors at 6 weeks of age. This was the earliest time-point sampled here, and one when all animals were still nondiabetic, and the inflammatory and autoimmune attack should be in its earliest phase (first diabetes onset occurred much later at week 13). Islet morphology is known to change with age in NOD mice

66, 67

. Insulitis might start as early as 4 weeks of age, but peri-islet infiltration is

typically still lacking at 6 week becoming apparent around 9–10 weeks of age 24, 25. Our earlier studies also confirmed that pancreatic islets begin to show peri-islet mononuclear infiltrates around 9–10 weeks of age, just before the onset of the first cases of hyperglycemia, and that these peri-islet infiltrates are consistently present around islets in all older animals, even in those that do not progress to diabetes

25

. A set of confocal images further illustrating this is

included in Supporting Information Figure S5. The role of these peri-islet leukocytes is not entirely clear; nevertheless, they also tend to be present in humans with T1D 68, but much less prominently and homogeneously than in NOD mice. They probably contribute to the destruction of β-cells (i.e., via soluble factors) and, often, but not always, they lead to islet infiltration (insulitis) than in NOD mice

69

. Insulitis in humans is often patchy, and it is less severe and invasive

21, 68

. Interestingly, we have also shown that β-cell proliferation increases

dramatically in NOD islets with the start of this infiltration

25

in agreement with observations

suggesting that soluble factors secreted by T-cells promote β-cell proliferation 70. 27



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 28 of 56

Among all biochemicals detected, a few were found to have significantly different levels at this earliest pre-inflammatory time-point sampled here (6 week); hence, they are possible early biomarkers identifying T1D progressors. Because there is only a limited number of such compounds among the large number of total detected compounds (e.g., ~2% in blood), there is too much noise, and RFA cannot efficiently separate progressors and non-progressors based on this data (week 6) (Supporting Information Figure S6). A list of all significantly different compounds at this earliest time-point is compiled in Supporting Information Table S3 ranked by decreasing order of significance (ANOVA contrast), and average concentration time-profiles for selected compounds of potential interest are shown in Figure 12. Corresponding individual profiles are also included as spaghetti plots in Supporting Information Figure S7. The metabolic sub-families that were most different between progressors and non-progressors at this earliest sampling time-point are summarized in Supporting Information Figure S8 using metabolic pathway enrichment (PE) values. Notably, several fatty acid phosphocholine derivatives in the phosphatidylcholine subpathway showed significant difference between progressors and non-progressors at this early time-point. While the corresponding fold-changes are relatively modest in the blood (1.06– 1.16×, ↑), they are somewhat larger in feces (1.44–1.75×, ↑). More importantly, they are among the most significant ones detected, and are consistently present in blood and fecal samples for several compounds, such as 1-palmitoyl-2-oleoyl-GPC (glycero-3-phosphocholine), 1-palmitoyl-2-arachidonoyl-GPC, and 1-stearoyl-2-arachidonoyl-GPC. While this is contrary to the observation of Oresic and co-workers of lower phosphatidylcholines in children who later 28


Page 29 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


progress to diabetes in the DIPP study of 129 genetically high-risk children who progressed to T1D in early childhood

31

, it agrees with the observation of Pflueger and co-workers of

significantly higher levels of cluster of lipids dominated by polyunsaturated fatty acid (PUFA)containing phosphatidylcholines (and triglycerides) in the larger BABYDIAB study of 1,650 children of mothers or fathers with T1D

32

. Our observation also seems to be somewhat

contrary to the finding of Sysi-Aho, Oresic, and co-workers that NOD mice that did not progress to diabetes despite being insulin autoantibody positive had elevated levels of selected serum lysophosphatidylcholines species compared with progressors

38

. Further, we found no

confirmation here of the lower methionine or hydroxyproline levels observed in that study. The progressors to non-progressors ratios for these compounds were all very close to unity and far from showing statistical significance either in blood or in feces. It should be noted that while these studies all used serum samples, our study used whole blood. As already mentioned, another human study looking at metabolite profiling in the Framingham Offspring Study, suggested that five branched-chain and aromatic amino acids (isoleucine, leucine, valine, tyrosine, and phenylalanine) are associated with future diabetes with fasting concentrations being higher 33. None of these were found to be increased in early blood or fecal samples here; however, as discussed earlier, they were all highly increased in blood at later time-points when animals were already diabetic (week 21 and 26; Figure 11). In addition to these phosphocholine derivatives, other potential early biomarker candidates for T1D progressors include serotonin, ribose, and arabinose (elevated) in blood as well as tocopherol (elevated) plus diaminopimelate, valerate, hydroxymethylpyrimidine, and 29



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 30 of 56

dulcitol (decreased) in feces (Figure 12, Figure S6). Whereas none of these compounds showed highly significant differences in blood samples (for all compounds, 0.01 < p < 0.05), eight compounds showed more significant differences in feces (p < 0.01; Table S3). While these compounds by themselves may not be different enough to serve as early disease indicators (biomarkers), a combination of them could provide a sufficiently reliable early biomarker signature to distinguish T1D progressors from non-progressors. Serotonin might be of particular interest as it was significantly elevated in T1D progressors at week 6 in blood (1.5×, ↑) and it stayed elevated (even if not significantly) at all time-points (1.1–1.3×, ↑). It was also elevated at week 6 in fecal samples (1.2×, ↑). In pancreatic islet, serotonin is an autocrine signal co-released with insulin, and it has been recently shown that during heightened metabolic demands, such as pregnancy or high-fat diet feeding, serotonin is produced in the islet and can stimulate β-cell proliferation via 5-HT2B and 5-HT3 receptors

71, 72

. Finally, xanthine, hypoxanthine, and cysteine S-sulfate could be of

potential interest as they show a clear peak at week 11 in T1D progressors, a time when the autoimmune inflammatory attack begins in progressors (Supporting Information Figure S9).

CONCLUSIONS In summary, our comprehensive non-targeted metabolomics study quantified the longitudinal change of several hundred biochemicals in the blood and feces of NOD mice as they progressed to hyperglycemia. Diabetes onset was found to profoundly alter the overall metabolic landscape: in diabetic animals, a large proportion of the detected metabolome was significantly altered especially in blood (57.8%), but also in feces (27.8%), despite the presence 30


Page 31 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


of sustained-release insulin. Several biochemicals showed considerable, more than four-fold change, and the metabolic pathways most strongly affected included carbohydrate, lipid, branched-chain amino acid, and oxidative metabolisms. In T1D animals, some compounds, such as 3-hydroxybutyrate, maltose, and kojibiose, increased very dramatically showing more than 10–20-fold increases, while 1,5-anhydroglucitol decreased even more. Differences between the metabolic signatures of progressors and non-progressors at the earliest time-point (6 week) were relatively modest. Nevertheless, several compounds had significantly different levels and show promise as possible early disease indicators – even if not alone by themselves, but conceivably as part of a combined metabolic signature.

ASSOCIATED CONTENT Supporting Information The Supporting Information is available free of charge at DOI: … Table S1. Summary of statistically significantly altered biochemical. Table S2. Metabolites found to show the largest fold-change in diabetic vs. non-diabetic controls. Table S3. Metabolites found to have significantly different levels at week 6 in NOD mice that later progressed to diabetes.

Figure S1. Schematic design of the present study. Figure S2. Random forest analysis of all of the present metabolomics data. 31



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 32 of 56

Figure S3. Compounds showing the largest fold-change in T1D mice. Figure S4. Spaghetti plots for compounds showing the largest fold-change in T1D progressors. Figure S5. Representative pancreatic sections immunostained for insulin in different mice. Figure S6. Random forest analysis of the week 6 metabolomics data (progressors vs. nonprogressors). Figure S7. Spaghetti plots for compounds showing the largest and statistically significant differences between progressors and non-progressors at the early 6 week sample. Figure S8. Pathways most different between diabetes progressor and non-progressor NOD mice at the earliest time-point analyzed. Figure S9. Line plots for compounds showing large and statistically significant differences between progressors and non-progressors at the week 11 sample.

AUTHOR INFORMATION Corresponding Author *

P.B.: E-mail: [email protected]. Tel.: (1) 305 243-9657.

Author Contributions Conceived the study: PB, CR. Designed the experiments: PB, ATG, MGC. Performed the experiments: ATG, MGC. Analyzed the data: SR, PB. Interpreted the results: PB, SR, CR, AJM. Drafted and finalized the manuscript: PB. All authors read and approved the final manuscript.

32


Page 33 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Competing interests The authors declare no competing financial interest.

ACKNOWLEDGMENTS Financial support from the Diabetes Research Institute Foundation is gratefully acknowledged.

REFERENCES (1) Atkinson, M. A.; Eisenbarth, G. S., Type 1 diabetes: new perspectives on disease pathogenesis and treatment. Lancet 2001, 358 (9277), 221-229.

(2) Daneman, D., Type 1 diabetes. Lancet 2006, 367 (9513), 847-858.

(3) Bluestone, J. A.; Herold, K.; Eisenbarth, G., Genetics, pathogenesis and clinical interventions in type 1 diabetes. Nature 2010, 464 (7293), 1293-1300.

(4) Atkinson, M. A.; Eisenbarth, G. S.; Michels, A. W., Type 1 diabetes. Lancet 2014, 383 (9911), 69-82.

(5) Burn, P., Type 1 diabetes. Nat. Rev. Drug Discov. 2010, 9 (3), 187-188.

(6) Dabelea, D.; Mayer-Davis, E. J.; Saydah, S.; Imperatore, G.; Linder, B.; Divers, J.; Bell, R.; Badaru, A.; Talton, J. W.; Crume, T.; Liese, A. D.; Merchant, A. T.; Lawrence, J. M.; Reynolds, K.; Dolan, L.; Liu, L. L.; Hamman, R. F.; Study, S. f. D. i. Y., Prevalence of type 1 and type 2 diabetes 33



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 34 of 56

among children and adolescents from 2001 to 2009. J. Am. Med. Assoc. (JAMA) 2014, 311 (17), 1778-1786.

(7) DIAMOND, Project Group, Incidence and trends of childhood Type 1 diabetes worldwide 1990-1999. Diabet. Med. 2006, 23 (8), 857-866.

(8) Onkamo, P.; Väänänen, S.; Karvonen, M.; Tuomilehto, J., Worldwide increase in incidence of Type I diabetes - the analysis of the data on published incidence trends. Diabetologia 1999, 42 (12), 1395-1403.

(9) Soltesz, G.; Patterson, C. C.; Dahlquist, G., Worldwide childhood type 1 diabetes incidence-what can we learn from epidemiology? Pediatr. Diabetes 2007, 8 (Suppl 6), 6-14.

(10) Patterson, C. C.; Dahlquist, G. G.; Gyurus, E.; Green, A.; Soltesz, G., Incidence trends for childhood type 1 diabetes in Europe during 1989-2003 and predicted new cases 2005-20: a multicentre prospective registration study. Lancet 2009, 373 (9680), 2027-2033.

(11) Pambianco, G.; Costacou, T.; Ellis, D.; Becker, D. J.; Klein, R.; Orchard, T. J., The 30-year natural history of type 1 diabetes complications: the Pittsburgh Epidemiology of Diabetes Complications Study experience. Diabetes 2006, 55 (5), 1463-1469.

(12) Gregg, E. W.; Li, Y.; Wang, J.; Burrows, N. R.; Ali, M. K.; Rolka, D.; Williams, D. E.; Geiss, L., Changes in diabetes-related complications in the United States, 1990-2010. N. Engl. J. Med. 2014, 370 (16), 1514-1523.

(13) van Belle, T. L.; Coppieters, K. T.; von Herrath, M. G., Type 1 diabetes: etiology, immunology, and therapeutic strategies. Physiol. Rev. 2011, 91 (1), 79-118. 34


Page 35 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(14) Oresic, M., Metabolomics in the studies of islet autoimmunity and type 1 diabetes. Rev. Diabet. Stud. 2012, 9 (4), 236-247.

(15) Overgaard, A. J.; Kaur, S.; Pociot, F., Metabolomic biomarkers in the progression to type 1 diabetes. Curr. Diab. Rep. 2016, 16 (12), 127.

(16) Gooding, J. R.; Jensen, M. V.; Newgard, C. B., Metabolomics applied to the pancreatic islet. Arch. Biochem. Biophys. 2016, 589, 120-130.

(17) Frohnert, B. I.; Rewers, M. J., Metabolomics in childhood diabetes. Pediatr. Diabetes 2016, 17 (1), 3-14.

(18) Makino, S.; Kunimoto, K.; Muraoka, Y.; Mizushima, Y.; Katagiri, K.; Tochino, Y., Breeding of a non-obese, diabetic strain of mice. Jikken Dobutsu (Exp. Anim.) 1980, 29 (1), 1-13.

(19) Delovitch, T. L.; Singh, B., The nonobese diabetic mouse as a model of autoimmune diabetes: immune dysregulation gets the NOD. Immunity 1997, 7 (6), 727-738.

(20) Anderson, M. S.; Bluestone, J. A., The NOD mouse: a model of immune dysregulation. Annu. Rev. Immunol. 2005, 23, 447-485.

(21) Roep, B. O.; Atkinson, M.; von Herrath, M., Satisfaction (not) guaranteed: re-evaluating the use of animal models of type 1 diabetes. Nat. Rev. Immunol. 2004, 4 (12), 989-997.

(22) Shoda, L. K.; Young, D. L.; Ramanujan, S.; Whiting, C. C.; Atkinson, M. A.; Bluestone, J. A.; Eisenbarth, G. S.; Mathis, D.; Rossini, A. A.; Campbell, S. E.; Kahn, R.; Kreuwel, H. T., A comprehensive review of interventions in the NOD mouse and implications for translation. Immunity 2005, 23 (2), 115-126. 35



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 36 of 56

(23) Reed, J. C.; Herold, K. C., Thinking bedside at the bench: the NOD mouse model of T1DM. Nat. Rev. Endocrinol. 2015, 11 (5), 308-314.

(24) Magnuson, A. M.; Thurber, G. M.; Kohler, R. H.; Weissleder, R.; Mathis, D.; Benoist, C., Population dynamics of islet-infiltrating cells in autoimmune diabetes. Proc. Natl. Acad. Sci. U.S.A. 2015, 112 (5), 1511-1516.

(25) Cechin, S. R.; Lopez-Ocejo, O.; Karpinsky-Semper, D.; Buchwald, P., Biphasic decline of βcell function with age in euglycemic nonobese diabetic mice parallels diabetes onset. IUBMB Life 2015, 67 (8), 634-644.

(26) Alanentalo, T.; Hornblad, A.; Mayans, S.; Karin Nilsson, A.; Sharpe, J.; Larefalk, A.; Ahlgren, U.; Holmberg, D., Quantification and three-dimensional imaging of the insulitis-induced destruction of beta-cells in murine type 1 diabetes. Diabetes 2010, 59 (7), 1756-1764.

(27) Virostko, J.; Radhika, A.; Poffenberger, G.; Dula, A. N.; Moore, D. J.; Powers, A. C., Bioluminescence imaging reveals dynamics of beta cell loss in the non-obese diabetic (NOD) mouse model. PloS one 2013, 8 (3), e57784.

(28) Matveyenko, A. V.; Butler, P. C., Relationship between beta-cell mass and diabetes onset. Diabetes Obes. Metab. 2008, 10 Suppl 4, 23-31.

(29) Ize-Ludlow, D.; Lightfoot, Y. L.; Parker, M.; Xue, S.; Wasserfall, C.; Haller, M. J.; Schatz, D.; Becker, D. J.; Atkinson, M. A.; Mathews, C. E., Progressive erosion of beta-cell function precedes the onset of hyperglycemia in the NOD mouse model of type 1 diabetes. Diabetes 2011, 60 (8), 2086-2091.

36


Page 37 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(30) Sreenan, S.; Pick, A. J.; Levisetti, M.; Baldwin, A. C.; Pugh, W.; Polonsky, K. S., Increased beta-cell proliferation and reduced mass before diabetes onset in the nonobese diabetic mouse. Diabetes 1999, 48 (5), 989-996.

(31) Oresic, M.; Simell, S.; Sysi-Aho, M.; Nanto-Salonen, K.; Seppanen-Laakso, T.; Parikka, V.; Katajamaa, M.; Hekkala, A.; Mattila, I.; Keskinen, P.; Yetukuri, L.; Reinikainen, A.; Lahde, J.; Suortti, T.; Hakalax, J.; Simell, T.; Hyoty, H.; Veijola, R.; Ilonen, J.; Lahesmaa, R.; Knip, M.; Simell, O., Dysregulation of lipid and amino acid metabolism precedes islet autoimmunity in children who later progress to type 1 diabetes. J. Exp. Med. 2008, 205 (13), 2975-2984.

(32) Pflueger, M.; Seppanen-Laakso, T.; Suortti, T.; Hyotylainen, T.; Achenbach, P.; Bonifacio, E.; Oresic, M.; Ziegler, A. G., Age- and islet autoimmunity-associated differences in amino acid and lipid metabolites in children at risk for type 1 diabetes. Diabetes 2011, 60 (11), 2740-2747.

(33) Wang, T. J.; Larson, M. G.; Vasan, R. S.; Cheng, S.; Rhee, E. P.; McCabe, E.; Lewis, G. D.; Fox, C. S.; Jacques, P. F.; Fernandez, C.; O'Donnell, C. J.; Carr, S. A.; Mootha, V. K.; Florez, J. C.; Souza, A.; Melander, O.; Clish, C. B.; Gerszten, R. E., Metabolite profiles and the risk of developing diabetes. Nat. Med. 2011, 17 (4), 448-453.

(34) Lanza, I. R.; Zhang, S.; Ward, L. E.; Karakelides, H.; Raftery, D.; Nair, K. S., Quantitative metabolomics by H-NMR and LC-MS/MS confirms altered metabolic pathways in diabetes. PloS one 2010, 5 (5), e10538.

(35) Balderas, C.; Ruperez, F. J.; Ibanez, E.; Senorans, J.; Guerrero-Fernandez, J.; Casado, I. G.; Gracia-Bouthelier, R.; Garcia, A.; Barbas, C., Plasma and urine metabolic fingerprinting of type 1 diabetic children. Electrophoresis 2013, 34 (19), 2882-2890.

37



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 38 of 56

(36) Galderisi, A.; Pirillo, P.; Moret, V.; Stocchero, M.; Gucciardi, A.; Perilongo, G.; Moretti, C.; Monciotti, C.; Giordano, G.; Baraldi, E., Metabolomics reveals new metabolic perturbations in children with type 1 diabetes. Pediatr. Diabetes 2017, ePub.

(37) Dutta, T.; Kudva, Y. C.; Persson, X. M.; Schenck, L. A.; Ford, G. C.; Singh, R. J.; Carter, R.; Nair, K. S., Impact of long-term poor and good glycemic control on metabolomics alterations in type 1 diabetic people. J. Clin. Endocrinol. Metab. 2016, 101 (3), 1023-1033.

(38) Sysi-Aho, M.; Ermolov, A.; Gopalacharyulu, P. V.; Tripathi, A.; Seppanen-Laakso, T.; Maukonen, J.; Mattila, I.; Ruohonen, S. T.; Vahatalo, L.; Yetukuri, L.; Harkonen, T.; Lindfors, E.; Nikkila, J.; Ilonen, J.; Simell, O.; Saarela, M.; Knip, M.; Kaski, S.; Savontaus, E.; Oresic, M., Metabolic regulation in progression to autoimmune diabetes. PLoS Comput. Biol. 2011, 7 (10), e1002257.

(39) Madsen, R.; Banday, V. S.; Moritz, T.; Trygg, J.; Lejon, K., Altered metabolic signature in prediabetic NOD mice. PloS one 2012, 7 (4), e35445.

(40) Grapov, D.; Fahrmann, J.; Hwang, J.; Poudel, A.; Jo, J.; Periwal, V.; Fiehn, O.; Hara, M., Diabetes associated metabolomic perturbations in NOD mice. Metabolomics 2015, 11 (2), 425437.

(41) Fahrmann, J.; Grapov, D.; Yang, J.; Hammock, B.; Fiehn, O.; Bell, G. I.; Hara, M., Systemic alterations in the metabolome of diabetic NOD mice delineate increased oxidative stress accompanied by reduced inflammation and hypertriglyceremia. Am. J. Physiol. Endocrinol. Metab. 2015, 308 (11), E978-E989.

(42) Wishart, D. S.; Jewison, T.; Guo, A. C.; Wilson, M.; Knox, C.; Liu, Y.; Djoumbou, Y.; Mandal, R.; Aziat, F.; Dong, E.; Bouatra, S.; Sinelnikov, I.; Arndt, D.; Xia, J.; Liu, P.; Yallou, F.; Bjorndahl, T.; 38


Page 39 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Perez-Pineiro, R.; Eisner, R.; Allen, F.; Neveu, V.; Greiner, R.; Scalbert, A., HMDB 3.0 - The Human Metabolome Database in 2013. Nucleic Acids Res. 2013, 41 (Database issue), D801D807.

(43) Garcia-Contreras, M.; Tamayo-Garcia, A.; Pappan, K. L.; Michelotti, G. A.; Stabler, C. L.; Ricordi, C.; Buchwald, P., A metabolomics study of the effects of inflammation, hypoxia, and high glucose on isolated human pancreatic islets. J. Proteome Res. 2017, 16 (6), 2294−2306.

(44) Evans, A. M.; DeHaven, C. D.; Barrett, T.; Mitchell, M.; Milgram, E., Integrated, nontargeted ultrahigh

performance

liquid

chromatography/electrospray

ionization

tandem

mass

spectrometry platform for the identification and relative quantification of the small-molecule complement of biological systems. Anal. Chem. 2009, 81 (16), 6656-6667.

(45) Evans, A. M.; Bridgewater, B. R.; Liu, Q.; Mitchell, M. W.; Robinson, R. J.; Dai, H.; Stewart, S. J.; DeHaven, C. D.; Miller, L. A. D., High resolution mass spectrometry improves data quantity and quality as compared to unit mass resolution mass spectrometry in high-throughput profiling metabolomics. Metabolomics 2014, 4, 132.

(46) DeHaven, C. D.; Evans, A. M.; Dai, H.; Lawton, K. A., Organization of GC/MS and LC/MS metabolomics data into chemical libraries. J. Cheminform. 2010, 2 (1), 9.

(47) Shannon P, M. A., Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T., Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 2003, 13:2498–5204.

(48) Cechin, S. R.; Lopez-Ocejo, O.; Ricordi, C.; Buchwald, P., Smad7 is a promising therapeutic target on the TGF-β pathway to prevent or reverse new-onset type 1 diabetes. Diabetes 2015, 64 (S1) (S1), A60. 39



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 40 of 56

(49) Jolliffe, I. T., Principal Component Analysis. Springer: New York, 1986; p 271.

(50) Breiman, L., Random forests. Mach Learn 2001, 45 (1), 5-32.

(51) Sakiani, S.; Olsen, N. J.; Kovacs, W. J., Gonadal steroids and humoral immunity. Nat. Rev. Endocrinol 2013, 9 (1), 56-62.

(52) Zandman-Goddard, G.; Peeva, E.; Shoenfeld, Y., Gender and autoimmunity. Autoimmun. Rev. 2007, 6 (6), 366-372.

(53) Tang, W. H.; Martin, K. A.; Hwa, J., Aldose reductase, oxidative stress, and diabetic mellitus. Front. Pharmacol. 2012, 3, 87.

(54) Dungan, K. M., 1,5-anhydroglucitol (GlycoMark) as a marker of short-term glycemic control and glycemic excursions. Expert Rev Mol Diagn 2008, 8 (1), 9-19.

(55) Taboulet, P.; Deconinck, N.; Thurel, A.; Haas, L.; Manamani, J.; Porcher, R.; Schmit, C.; Fontaine, J. P.; Gautier, J. F., Correlation between urine ketones (acetoacetate) and capillary blood ketones (3-beta-hydroxybutyrate) in hyperglycaemic patients. Diabetes Metab. 2007, 33 (2), 135-139.

(56) Klocker, A. A.; Phelan, H.; Twigg, S. M.; Craig, M. E., Blood beta-hydroxybutyrate vs. urine acetoacetate testing for the prevention and management of ketoacidosis in Type 1 diabetes: a systematic review. Diabet. Med. 2013, 30 (7), 818-824.

(57) Gall, W. E.; Beebe, K.; Lawton, K. A.; Adam, K. P.; Mitchell, M. W.; Nakhle, P. J.; Ryals, J. A.; Milburn, M. V.; Nannipieri, M.; Camastra, S.; Natali, A.; Ferrannini, E.; Group, R. S., α40


Page 41 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Hydroxybutyrate is an early biomarker of insulin resistance and glucose intolerance in a nondiabetic population. PloS one 2010, 5 (5), e10883.

(58) Peddinti, G.; Cobb, J.; Yengo, L.; Froguel, P.; Kravic, J.; Balkau, B.; Tuomi, T.; Aittokallio, T.; Groop, L., Early metabolic markers identify potential targets for the prevention of type 2 diabetes. Diabetologia 2017, ePub.

(59) Lynch, C. J.; Adams, S. H., Branched-chain amino acids in metabolic signalling and insulin resistance. Nat. Rev. Endocrinol. 2014, 10 (12), 723-736.

(60) Schmid, S.; Koczwara, K.; Schwinghammer, S.; Lampasona, V.; Ziegler, A. G.; Bonifacio, E., Delayed exposure to wheat and barley proteins reduces diabetes incidence in non-obese diabetic mice. Clin. Immunol. 2004, 111 (1), 108-118.

(61) Yurkovetskiy, L.; Burrows, M.; Khan, A. A.; Graham, L.; Volchkov, P.; Becker, L.; Antonopoulos, D.; Umesaki, Y.; Chervonsky, A. V., Gender bias in autoimmunity is influenced by microbiota. Immunity 2013, 39 (2), 400-412.

(62) Wen, L.; Ley, R. E.; Volchkov, P. Y.; Stranges, P. B.; Avanesyan, L.; Stonebraker, A. C.; Hu, C.; Wong, F. S.; Szot, G. L.; Bluestone, J. A.; Gordon, J. I.; Chervonsky, A. V., Innate immunity and intestinal microbiota in the development of Type 1 diabetes. Nature 2008, 455 (7216), 11091113.

(63) Burrows, M. P.; Volchkov, P.; Kobayashi, K. S.; Chervonsky, A. V., Microbiota regulates type 1 diabetes through Toll-like receptors. Proc. Natl. Acad. Sci. U.S.A. 2015, 112 (32), 9973-9977.

(64) Sanz, M. L.; Gibson, G. R.; Rastall, R. A., Influence of disaccharide structure on prebiotic selectivity in vitro. J. Agric. Food Chem. 2005, 53 (13), 5192-5199. 41



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 42 of 56

(65) Laparra, J. M.; Diez-Municio, M.; Moreno, F. J.; Herrero, M., Kojibiose ameliorates arachidic acid-induced metabolic alterations in hyperglycaemic rats. Br. J. Nutr. 2015, 114 (9), 1395-1402.

(66) Plesner, A.; Ten Holder, J. T.; Verchere, C. B., Islet remodeling in female mice with spontaneous autoimmune and streptozotocin-induced diabetes. PloS one 2014, 9 (8), e102843.

(67) Pechhold, K.; Zhu, X.; Harrison, V. S.; Lee, J.; Chakrabarty, S.; Koczwara, K.; Gavrilova, O.; Harlan, D. M., Dynamic changes in pancreatic endocrine cell abundance, distribution, and function in antigen-induced and spontaneous autoimmune diabetes. Diabetes 2009, 58 (5), 1175-1184.

(68) Reddy, S.; Zeng, N.; Al-Diery, H.; Jung, D.; Yeu, C.; Joret, M. O.; Merrilees, M. J.; Wu, F., Analysis of peri-islet CD45-positive leucocytic infiltrates in long-standing type 1 diabetic patients. Diabetologia 2015, 58 (5), 1024-35.

(69) Korpos, E.; Kadri, N.; Kappelhoff, R.; Wegner, J.; Overall, C. M.; Weber, E.; Holmberg, D.; Cardell, S.; Sorokin, L., The peri-islet basement membrane, a barrier to infiltrating leukocytes in type 1 diabetes in mouse and human. Diabetes 2013, 62 (2), 531-542.

(70) Dirice, E.; Kahraman, S.; Jiang, W.; El Ouaamari, A.; De Jesus, D. F.; Teo, A. K.; Hu, J.; Kawamori, D.; Gaglia, J. L.; Mathis, D.; Kulkarni, R. N., Soluble factors secreted by T-cells promote beta-cell proliferation. Diabetes 2014, 63 (1), 188-202.

(71) Kim, H.; Toyofuku, Y.; Lynn, F. C.; Chak, E.; Uchida, T.; Mizukami, H.; Fujitani, Y.; Kawamori, R.; Miyatsuka, T.; Kosaka, Y.; Yang, K.; Honig, G.; van der Hart, M.; Kishimoto, N.; Wang, J.; Yagihashi, S.; Tecott, L. H.; Watada, H.; German, M. S., Serotonin regulates pancreatic beta cell mass during pregnancy. Nat. Med. 2010, 16 (7), 804-848.

42


Page 43 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(72) Bennet, H.; Mollet, I. G.; Balhuizen, A.; Medina, A.; Nagorny, C.; Bagge, A.; Fadista, J.; Ottosson-Laakso, E.; Vikman, P.; Dekker-Nitert, M.; Eliasson, L.; Wierup, N.; Artner, I.; Fex, M., Serotonin (5-HT) receptor 2b activation augments glucose-stimulated insulin secretion in human and mouse islets of Langerhans. Diabetologia 2016, 59 (4), 744-754.

43



FIGURES AND FIGURE LEGENDS

Percent diabetic (hyperglycemic)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 44 of 56

Figure 1. Kaplan-Meier survival curves showing the overall rate of diabetes onset within the NOD mice of the present study. Data for male (M) and female (F) mice (n = 15 + 15) are shown separately in blue and red, respectively. Sample collections times are denoted with purple arrows at the bottom. For comparison, onset rates obtained by us in two previous studies with female NODs 25, 48 are also included (shown in lighter colors).

44


Page 45 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 2. Changes with age in the number of statistically significantly altered biochemicals in the blood and fecal samples of T1D progressor NOD mice. Data shown as stacked columns are from Table S1 with the number of compounds that are increased in progressors vs. non-progressors shown in red and those that decreased in blue.

45



Males

26 weeks Type 1

Blood Control-6 weeks Control-11 weeks

PC2 [14.46%]

Control- 16 weeks Control- 21 weeks Control- 26 weeks Type I-6 weeks

Females

Type I-11 weeks Type I- 16 weeks Type I- 21 weeks Type I- 26 weeks

PC1 [18.11%]

A

Feces Control-6 weeks Control-11 weeks Control- 16 weeks

PC2 [10.71%]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 46 of 56

Control- 21 weeks Control- 26 weeks 21 and 26 weeks Type 1

Type I-6 weeks Type I-11 weeks Type I- 16 weeks Type I- 21 weeks Type I- 26 weeks

B

PC1 [23.47%]

Figure 3. Principal component analysis (PCA) of the present metabolomics data in blood (left) and fecal samples (right) obtained longitudinally (weeks 6 to 26) from NOD mice that progressed (Type 1) or not (Control) to diabetes.

46


Page 47 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 4. Hierarchical clustering of the present metabolomics data in blood and fecal samples obtained longitudinally (weeks 6 to 26) from NOD mice that progressed (Type 1) or not (Control) to diabetes. Within the clusters, green hues represent elevated metabolite levels, while red hues represent decreased metabolite levels. Samples (columns) are color-coded as indicated by the figure key on the bottom left side, and rows are arranged by the corresponding metabolic super-pathways as indicated on the right edge.

47



Fold Change in Diabetic Progressors (Blood) 100

maltose 3-hydroxybutyrate 10

kojibiose Fold change (log scale)

glucose

1 5

10

15

20

25

30

beta-muricholate nicotinamide N-oxide 0.1

Age (week)

1,5 anhydroglucitol 0.01

A

maltose

linoleoyl-linolenoyl-glycerol

2-isopropylmalate

3-hydroxybutyrate (BHBA)

palmitoyl-linolenoyl-glycerol

stearoyl-linoleoyl-glycerol

2-keto-3-deoxy-gluconate

glutathione, reduced (GSH)

xanthosine

3-hydroxydecanoate

13-HODE + 9-HODE

linoleoyl-linolenoyl-glycerol

linoleoyl-linoleoyl-glycerol

stearoyl-linoleoyl-glycerol

4-hydroxy-nonenal-glutathione

3-hydroxybutyrylcarnitine

9,10-DiHOME

sedoheptulose

1,2,3-benzenetriol sulfate

palmitoleoyl-arachidonoyl-glycerol

3-hydroxylaurate

linolenate [alpha or gamma]

caffeic acid sulfate

kojibiose

taurodeoxycholate

glucose

mannitol/sorbitol

gamma-glutamyl-epsilon-lysine

1,5-anhydroglucitol (1,5-AG)

nicotinamide N-oxide

beta-muricholate

Fold Change in Diabetic Progressors (Feces) 100

kojibiose

10

glucose 3-hydroxybutyrate maltose Fold change (log scale)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 48 of 56

1 5

10

15

20

25

30

beta-muricholate I-urobilinogen ursodeoxycholate sulfate

0.1

0.01

Age (week) kojibiose

gluconate

galacturonate

hippurate

guanosine-2',3'-cyclic monophosphate

glucose

1-palmitoyl-2-oleoyl-GPI (16:0/18:1)*

tigloylglycine

indolepropionylglycine

inosine

nicotinate ribonucleoside

pimeloylcarnitine/3-methyladipoylcarnitine (C7-DC)

oleoyl-oleoyl-glycerol (18:1/18:1) [1]*

2'-deoxycytidine

5-methyluridine (ribothymidine)

ribonate

citrate

mannose

cytidine 2',3'-cyclic monophosphate

guanosine

3-hydroxybutyrate (BHBA)

uridine-2',3'-cyclic monophosphate

hydroquinone sulfate

allantoic acid

1-oleoyl-GPA (18:1)

nicotinic acid mononucleotide (NaMN)

maltose

mannitol/sorbitol N-acetylglutamine

B

beta-muricholate

myristoleoylcarnitine (C14:1)*

lithocholate

myristoylcarnitine (C14)

I-urobilinogen

ursodeoxycholate sulfate (2)

Figure 5. Longitudinal time-profile of the fold-change in the average concentration (progressors vs. nonprogressors) for compounds showing the largest fold-change in T1D vs. controls in blood (A) and fecal (B) samples at the last sampling (week 26; all compounds with >4-fold change and p < 0.05). Data shown on log scale to incorporate compounds that show increase as well as decrease vs. non-diabetic controls.

48


Page 49 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 6. Line plots illustrating the longitudinal time-profile of some of the compounds showing the largest fold-change increase (A) or decrease (B) in T1D progressors vs. non-progressor controls in both blood and fecal samples. Data (scaled intensity) shown as average ± SD (n = 8 and 6); numbers in the upper left corners of the corresponding insets indicate average fold-change values at week 26, when all progressors were already diabetic. Chemical structures are also included for illustration. Note that the scaled intensity values shown here are obtained after normalization of the original scale data to set the median for each biochemical across all animals and time-points equal to one. Corresponding spaghetti plots with individual lines for each animal are included in Supporting Information Figure S4.

49



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 50 of 56

Figure 7. Effect of diabetes onset on NOD mouse metabolism (i.e., T1D vs. control at week 26) shown using MetaboLync Pathway Classification Network for blood samples for selected super-families. Within each pathway, the size of the circle correlates with the magnitude of change, and the color indicates significant change vs. control (p < 0.05; red increased, blue decreased).

50


Page 51 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 8. Effect of diabetes onset on NOD mouse metabolism (i.e., T1D vs. control at week 26) shown using MetaboLync Pathway Classification Network for fecal samples for selected super-families. Within each pathway, the size of the circle correlates with the magnitude of change, and the color indicates significant change vs. control (p < 0.05; red increased, blue decreased).

51



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 52 of 56

Figure 9. Selected metabolites showing the most different values in male and female NOD mice. Line profiles showing change in time are color coded by progressor status (red T1D progressors vs. green non-progressors) with males in darker and females in lighter hues. As before, data are scaled intensity shown as average ± SD (n = 4M + 4F and 3M + 3F).

52


Page 53 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 10. The glycolytic and sorbitol pathways of glucose utilization with line plots included showing metabolites detected here and significantly altered in diabetic NOD mice vs. corresponding controls. All metabolites shown are significantly different (p < 0.05) at the week 26 time-point, when all progressors are diabetic (lactate is also shown; for it, the difference is only close to being significant, p = 0.077). Numbers in the upper left corner of the corresponding insets indicate the average fold-change (T1D vs. normoglycemic controls) at the last time-point. Additional related metabolites that were found to be highly (>10×) altered in T1D animals are also shown (left); they include maltose as well as 1,5anhydroglucitol and 3-hydroxybutyric acid, which are clinically used as biomarkers (see text).

53



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 54 of 56

Figure 11. The ‘leucine, isoleucine and valine metabolism’ sub-pathway with line plots included showing metabolites detected here and significantly altered in diabetic NOD mice vs. corresponding controls (week 26). All metabolites shown are significantly different (p < 0.05) at the week 26 time-point, when all progressors are diabetic. Figure and table included at right illustrate the individual fold-change values.

54


Page 55 of 56

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 12. Longitudinal time-profile of the compounds showing the largest and statistically significant (p < 0.05) differences between progressors and non-progressors at the early 6 week sample in blood (A) and feces (B). Data (scaled intensity) shown as average ± SD (n = 8 and 6); numbers in the upper left corner of the corresponding insets indicate average fold-change and p values at week 6 (gray ovals), when all animals are nondiabetic. Chemical structures are also included for illustration. Corresponding spaghetti plots with individual lines for each animal are included in Supporting Information Figure S7.

55



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

TABLE OF CONTENT (TOC) GRAPHIC

56


Page 56 of 56

Comprehensive Metabolomics Study To Assess ... - ACS Publications

Recommend Documents