Knockout of the non-essential gene SUGCT creates diet-linked, age-related microbiome disbalance with a diabetes-like metabolic syndrome phenotype
SUGCT (C7orf10) is a mitochondrial enzyme that synthesizes glutaryl-CoA from glutarate in tryptophan and lysine catabolism, but it has not been studied in vivo. Although mutations in Sugct lead to Glutaric Aciduria Type 3 disease in humans, patients remain largely asymptomatic despite high levels of glutarate in the urine. To study the disease mechanism, we generated SugctKO mice and uncovered imbalanced lipid and acylcarnitine metabolism in kidney in addition to changes in the gut microbiome. After SugctKO mice were treated with antibiotics, metabolites were comparable to WT, indicating that the microbiome affects metabolism in SugctKO mice. SUGCT loss of function contributes to gut microbiota dysbiosis, leading to age-dependent pathological changes in kidney, liver, and adipose tissue. This is associated with an obesity-related phenotype that is accompanied by lipid accumulation in kidney and liver, as well as “crown-like” structures in adipocytes. Furthermore, we show that the SugctKO kidney pathology is accelerated and exacerbated by a high-lysine diet. Our study highlights the importance of non-essential genes with no readily detectable early phenotype, but with substantial contributions to the development of age-related pathologies, which result from an interplay between genetic background, microbiome, and diet in the health of mammals.
KeywordsGlutaric aciduria type 3 (GA3) C7orf10 Sugct Gut microflora Metabolomics Lipids Obesity
Driven by the ambiguous symptoms in patients with GA3, we generated Sugct knockout mice (hereinafter referred to as SugctKO) to determine in vivo functions of SUGCT in mammals. We used non-targeted global metabolomic profiling of mouse kidney and plasma, which is a powerful tool for assessing biochemical metabolites in various biological contexts. Loss of Sugct evoked changes of the gut microbiome-dependent metabolism resulting in acylcarnitine and bacterial metabolite accumulation in kidney. These symptoms are aggravated with age or when mice are fed high-lysine diet, which resulted in an obesity-related phenotype that is accompanied by lipid accumulation in kidney and liver, as well as “crown-like” inflammatory structures in adipocytes. Our findings suggest that mutations in Sugct causing GA3 disease need to be studied in the context of the gut microflora, diet, and age.
Generation of Sugct knockout mice
To establish the role of the C7orf10 gene in vivo in mammals, we generated Sugct knockout mice by inserting loxP sites flanking the third exon of the Sugct gene (Figure S1a, Table S1). The targeting vector was electroporated into embryonic stem cells (ESCs) to generate heterozygous Sugct+/flox ESCs via homologous recombination (Figure S1b). Sugct+/flox ESCs were then injected into mouse blastocysts and targeted mice were obtained by standard procedures . Sugctflox/flox were bred with β-actin-cre mice to generate heterozygous null (Sugct+/null) animals, which were then intercrossed to get homozygous null animals (Sugctnull/null; hereinafter referred to as SugctKO). PCR genotyping of Sugct+/flox and SugctKO mice revealed bands at the expected sizes (Figure S1c). Homozygous SugctKO mice were obtained at the typical Mendelian frequencies (Figure S1d) and were viable.
To investigate Sugct/SUGCT expression levels in various tissues and verify the efficiency of the constructed SugctKO, we isolated RNA and proteins from several WT and SugctKO organs. Unlike in the mutant animals, we detected the expression of Sugct/SUGCT in WT kidney and liver at the mRNA and protein levels, respectively (Figure S1e, f). Both tissues are known to contain a high number of mitochondria due to their elevated metabolic rate and are often challenged by toxins and pathogens. Interestingly, it has been shown using single-cell transcriptomics in mouse kidney that Sugct is mainly expressed in proximal tubules, which are immune responders to toxic injuries . Accumulation of urinary metabolites in kidney, where SUGCT is mostly expressed could cause deleterious phenotypes, and therefore, we started our study with investigations of renal mouse tissue.
Metabolic changes in SugctKO mouse kidney
To investigate whether we can identify other metabolites related to GA3 in SugctKO mice, we expanded the analysis of our metabolomic data. The vast majority of significantly altered kidney metabolites (80%) were upregulated in SugctKO mice with only few downregulated, suggesting that SUGCT is a “repressor” of those compounds (Fig. 2c, Table S2). To get a more expanded view of the metabolic changes, we lowered the FDR to 25%, which allowed us to detect besides tryptophan co-metabolites (5.5%), also differentially regulated acylcarnitines (25.5%) and lipids (20%) (Figures S2b, S2c, Table S2). Most of these metabolites were only modestly increased, but these data still give an impression of the direction of the metabolic rewiring in SugctKO kidney. 30% of detected lipids were prenols (arnamiol, armillarin, valtrate, isopetasoide, icariside B8), while the rest, where medium/long-chain fatty acids, one glycerophospholipid, and acyl choline (Table S2). Intriguingly, among the detected lipids were those associated with gut microflora metabolism, such as adipic acid , as well as those contributing to dysbiosis (palmitic acid, oleic acid, and linoleic acid) . In addition, 24% of detected metabolites either did not have a eukaryotic origin and/or are co-regulated by gut bacteria (Fig. 2c, S2c, S3, Table S2, S3), indicating a prokaryotic contribution to the results that we observed. Considering that the kidney is held in a sterile environment, any presence of the non-host-derived metabolites was surprising. Interestingly, it has been shown that a substantial amount of the dietary tryptophan in the human gut is metabolized by bacteria . In accordance with this, we detected in SugctKO mouse kidney increased indoleacrylic acid, which derives from indole-propionic acid , and is known as a suppressor of commensal inflammation . It is important to keep in mind that gut bacteria-derived metabolites in kidney are pathological and linked to an early decline in renal function [20, 21] (Figure S3).
16S rRNA microbiome sequencing from stool DNA
Our results indicate that loss of Sugct affects the microbiome with changes closely resembling observed microbiome disbalance in metabolic diseases like diabetes.
Antibiotic treatment reverses the metabolite profile of SugctKO mice
To investigate the potential contribution of intestinal bacteria to the metabolic changes detected in the SugctKO mouse kidney, we eliminated gut bacteria residing in the intestines by treatment with broad-spectrum antibiotics. The efficiency of antibiotic administration (hereinafter referred to as “abx”) on intestinal bacteria clearance was tested by 16S rRNA gene amplification of bacterial nucleic acids extracted from feces (Figure S4a, see also “Materials and methods”) . We were not able to detect bacterial DNA after antibiotic treatment, indicating that the number of bacteria in the intestines of our mice was below the detection limit.
Altogether, our data suggest that the antibiotic clearance of gut microflora in SugctKO mice alleviates alterations in the levels of lipids, acylcarnitines, bacterial metabolites, glycine conjugates, and dipeptides in SugctKO mouse plasma. This indicates that the absence of the gut microbiome restores the metabolic homeostasis in the animals harboring the Sugct mutation and may also suggest that the gut microbiome plays an important role in the GA3 disease.
Age-associated obesity and glucose intolerance in SugctKO mice
The significant increase of body weight in SugctKO mice could indicate metabolic dysfunction due to the loss of Sugct. Since SUGCT is highly expressed in kidney (see Figure S1e, f), where we observed metabolic changes in SugctKO mice (see Fig. 2, S2, S3), we collected kidneys from 52- to 58-week-old WT and mutant animals. We detected elevated number of vesicles in cytoplasm of renal tubular epithelial cells in WT mice , which in a subset of experimental SugctKO animals were further elevated (Fig. 5b, top panel). In addition, we noticed an increase of interstitial mononuclear cell infiltrate in SugctKO in comparison with WT mice (Fig. 5b, lower panel), an indication of inflammation. Therefore, we investigated the number of macrophages by tissue staining with F4/80 antibodies, best known as a marker of mature mouse macrophages and microglia  (Fig. 5c). We detected a threefold increased staining of macrophages in kidneys of aged SugctKO mice when compared to WT animals (Fig. 5d), which supports our hypothesis that metabolic changes in SugctKO mice may promote an inflammatory response.
Furthermore, there was lipid accumulation in the SugctKO kidney (Fig. 5e, f). Beside kidney, we analyzed histopathological changes in the liver (Figure S5a) and epididymal white adipose tissue (ewat; Fig. 5g, S5b) due to the observed weight gain in mutant animals (Fig. 5a). We observed micro- and macrovesicular steatosis in SugctKO mouse liver (Figure S5a) and we detected lipids by Oil Red O staining (Figure S5c, quantification shown in S5d). In addition, SugctKO mice displayed a greater degree of inflammation in adipose tissue, often forming “crown-like structures”  (Fig. 5g). This type of adipose tissue pathology indicates adipocyte death, which is often associated with macrophages surrounding dying adipocytes . Despite the histopathological changes detected in kidney, liver, and epididymal white adipose tissue (ewat), no additional gross abnormalities were found in aged mutant animals.
In summary, we uncovered that ageing significantly contributes to the phenotype in SugctKO mice through increased body fat accumulation and progressive renal tubular vacuolation, which was accompanied with increased macrophage levels, fat accumulation in liver, and adipocyte death.
Lysine-enriched diet aggravates histopathological changes in SugctKO mouse kidney
Diet composition has far-reaching effects on mammalian physiology . Certain diet-induced pathologies that are severe in humans might only appear later or never in mice, since they are not exposed to varied diets. As previously shown in a mouse model for GA1, GcdhKO mice despite accumulating glutaric and 3-hydroxyglutaric acid, develop only mild motor deficits, unlike humans . However, 4-week-old GcdhKO mice exposed to high-lysine diet display severe striatal degeneration typical of the human GA1 disease and 75% of the mice die within 3–12 days .
We investigated the pathophysiology of the GA3 disease by performing metabolomics in kidney and plasma using our SugctKO mouse model. We found that Sugct loss of function is not only correlated with increased glutarate levels but also lipid/acylcarnitine imbalances and dysbiosis. In addition, we uncovered significant changes in the gut microbiome of SugctKO compared to WT mice. To further investigate the influence on the gut microflora metabolism caused by the SugctKO mutation, we expanded our study to mouse plasma, whose biochemistry is known to be strongly affected by bacterial metabolites that enter host’s circulatory system . We found that the clearance of the gut microflora eliminated the metabolic differences between WT and SugctKO mouse plasma, indicating that the metabolism of SugctKO mice is regulated indirectly or directly by the microbiome. Furthermore, metabolic changes in SugctKO mice were connected to age-dependent susceptibility to excessive weight gain, as well as a potential decline in kidney function accompanied with pathological changes in liver and white adipose tissue. Last but not least, high-lysine diet was an important factor in aggravating and accelerating the severity of the observed pathologies in SugctKO mouse kidney. Overall, our results indicate that outcome of genetic mutations in Sugct is modulated by the microbiome, age, and diet.
We have shown that loss of Sugct leads to modestly increased accumulation of a number of different acylcarnitines in the kidney (see Figure S2c, Table S2). Acylcarnitines are acyl esters of carnitine and essential compounds involved in energy production via fatty acids in mitochondria [53, 54]. l-carnitine (l-3-hydroxy-4-aminobutyrobetaine) is either synthesized in kidney and liver by the host from two essential amino acids, lysine and methionine (25%), or is directly absorbed from food (75%). Hypothetically, in the absence of functional SUGCT, the endogenous levels of carnitine in kidney might be higher due to its elevated precursor lysine, which cannot be efficiently catabolized to acetyl-CoA (see Fig. 1). The metabolic shift towards eukaryotic carnitine production increases the amount of unabsorbed dietary carnitine and triggers its bacterial metabolism into TMA (trimethylamine), whose oxidized form (trimethylamine N-oxide; TMAO) contributes to the fatty liver phenotype in the human population [55, 56]. Considering that the major source of renal ATP production is mitochondrial β-oxidation of free non-esterified fatty acids, carnitine imbalances could result in disturbances of the carnitine : acylcarnitine ratio and consequently leads to dyslipidosis followed by renal dysfunction [57, 58] (Fig. 7). Our discovery that loss of Sugct contributes to altered metabolic homeostasis between host and gut bacteria accompanied by dyslipidosis, is in line with a recent study of the GA3 disease, where a patient was reported with gastrointestinal disturbances . Interestingly, treatment of the patient with antibiotics alleviated the symptoms, which indicates an important role of the gut microflora in the GA3 disease progression. Although this anecdotal finding supports our own results, a lot more GA3 patients will need to be treated with antibiotics before a final conclusion can be drawn.
There is growing evidence that beside environmental factors, genetic control shapes host-gut microbiota interactions [43, 59], which partially explains the emerging contribution of the microbiome to the onset of obesity [60, 61]. Until now, the GA3 disease has been poorly understood, underdiagnosed and in the majority of cases untreated. Therefore, mutations in Sugct should be considered as an important contributing factor in patients with metabolic disorders and detection of acylcarnitines/lipids/bacterial metabolites could serve as useful biomarkers for the choice of disease treatment.
Materials and methods
All animal works were done in a humane way and were approved by Biological Resource Center (BRC) of Biopolis in A*STAR (IACUC #171268). Mice were housed under standard conditions, maintained on a 12-h light/dark cycle, and were co-housed. If not stated differently, mice were fed a standard chow diet containing 6% crude fat (Altromin, #1810) and were treated in compliance with the institutional guidelines for animal care and use. The high-lysine diet was prepared by adding free lysine to a standard diet (customized from Altromin) to achieve 4.7% total lysine (5X of the normal diet).
Generation of knockout mice
Mouse genomic DNA harboring the Sugct locus was retrieved from the BAC clone RP23-451G10 (Invitrogen, PKB1129) and inserted into the pBlight-TK vector. LoxP recombination sites and Neo cassette were introduced flanking the third exon of murine Sugct genomic locus . The targeting vector (PKB1318) was linearized by NotI digestion and electroporated into ES cells. Positive and negative selection with geneticin and ganciclovir, respectively, was followed by the homologous recombination screen of genomic DNA from ES cell colonies using Southern blot technique (5′ probe: chromosome 13, 17670059–17670559 bp; 3′ probe: chromosome 13, 17675399–17675899 bp). XhoI was a restriction enzyme used for screening both 5′ and 3′ end recombination events. Correctly targeted ES clones (9002, 9022, 9030) were used for the generation of the Sugct conditional knockout mouse strain. The Sugctflox allele was obtained by crossing Sugct conditional knockout mice with β-actin-Flpe transgenic mice  [strain name: B6.Cg-Tg(ACTFLPe) 9205Dym/J; stock no.: 005703.; The Jackson Laboratory] to remove the Neo cassette. Sugctflox mice were then crossed with β-actin-Cre transgenic mice [strain name: FVB/N-Tg(ACTB-cre)2Mrt/J; stock no.: 003376; The Jackson Laboratory]  to generate Sugct+/null mice that were subsequently intercrossed to obtain Sugctnull/null mice and these were backcrossed at least 10 times to a C57BL/6 J background.
RNA extraction and real-time PCR
RNA extraction was performed using the TRIzol® Reagent (Ambion™ Life technologies, #15596018) extraction protocol. The TRIzol reagent was added to kidney and liver samples (~ 40 mg) in 2 mL lysing matrix D tubes (MP Biomedicals™, #6913-500). Tissues were homogenized in a Precellys 24 Dual® (Bertin Technologies) at 4 °C using three cycles of 60 s at 5000 rpm. RNA concentration was measured with an eight-sample spectrophotometer (ND-800, NanoDrop®). cDNA was prepared using Maxima® Reverse transcriptase (Fermentas Life Sciences, #EP0741), followed by qPCR using Maxima® SYBR Green qPCR Master Mix (Fermentas Life Sciences, #K0251) with 10 ng of cDNA per reaction in a real-time thermal cycler (Corbett Research). Absolute quantification was obtained by a standard curve method using known concentration of serially diluted kidney RT-PCR product . Primers used in this study are listed in Table S1.
Western blot analysis
Frozen tissues were lysed in radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris pH 8.0, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 2X protease inhibitors [20 μg/mL each of leupeptin, chymostatin, and pepstatin; Chemicon, EI8, EI6, and EI10, respectively]) for 20 min on ice and clarified by centrifugation. Protein concentration was assessed using the BCA assay (Thermo Scientific, #23225). The protein extracts were separated by SDS-PAGE and transferred onto PDVF membrane (Millipore, IPVH0010) using a semi-dry trans-blot system. Buffer containing TBS with 0.1% Tween 20 (TBS-T) and 4% milk (Biorad, 1706404) was used as a blocking agent. The following primary antibodies were used: anti-C7ORF10 (Proteintech Group, #21589-1-AP) and anti-Hsp90 (BD Transduction Laboratories, #610419).
LC–MS profiling from mouse kidney and plasma
Aqueous fractions were analyzed on a 6540 Quadrupole time-of-flight mass spectrometer (QTOF-MS, Agilent Technologies). The instrument was run in both positive and negative electrospray ionization (ESI) modes with scan range 50–1700 m/z and scan rate of 2 spectra/s. 5 μL of kidney and plasma extracts were injected onto Atlantis T3 Columns (100Å, 3 μm, 4.6 mm × 150 mm, #186003729). The mobile phases A and B consisted of 100% water with 0.1% formic acid and 100% acetonitrile with 0.1% formic acid, respectively. The gradient elution was as follows: 99.5% mobile phase A starting point held for 5 min, decreased over 15 min to 0% mobile phase A and held for 8 min. The column was re-equilibrated for 10 min. Total run time per sample was 38 min including equilibration.
Kidney sample preparation for LC–MS analysis
Approximately 100 mg of kidney tissue per sample was homogenized twice with 800 µL of 1:1 methanol (MeOH):H2O and followed by centrifugation at 4 °C for 10 min. The supernatant was further partitioned twice using liquid–liquid extraction with 800 µL 4:1 dichloromethane:MeOH. The aqueous fraction was centrifuged, dried, resuspended in 200 µL 1:1 MeOH:H2O, and subsequently processed by LC–MS analysis.
Antibiotic treatment, blood collection, and sample preparation for LC–MS analysis
Streptomycin sulfate salt (Sigma-Aldrich, #S6501) and penicillin G sodium salt (Sigma-Aldrich, #P3032) were added to drinking water at the concentration of 2 g/L and 1500 U/mL, respectively. The drinking water was changed every 3 days during the 4-week long treatment. Blood from WT and SugctKO mice (5-week-old prior and 9-week-old post-treatment) was collected into lithium/heparin-coated vacutainers (Microvette 500 LH, Sarstedt, #20.1345.100) either by submandibular bleeding or by terminal cardiac puncture. Vacutainers were centrifuged for 5 min at 3000 rpm at 4 °C and supernatant was immediately frozen at − 80 °C. 200 µL of acetonitrile was added to 50 µL of blood plasma per sample. The mixture was vortexed vigorously for 1 min and incubated on ice for 10 min. Subsequently, the mixture was centrifuged at 14,000g for 15 min and the supernatant was transferred for LC–MS analysis. Blank and pooled quality control (QC) samples were prepared for instrument QC purposes.
LC–MS data and statistical analysis
The MassHunter Profinder software (Agilent technologies, version 6.0) was used for data extraction with the following parameters: batch recursive feature extraction; ion species allowed: H+, Na+, and K+; charge states were limited to a maximum of 2; compound ion count threshold was set to 2 or more ions; Extracted Ion Chromatogram (EIC) tolerance for mass was set to 10 ppm; retention time 2–13 min; and absolute height > 1000 counts. Peak area values from precursor ion chromatograms, extracted by the Mass Profiler Professional (MPP) software, were averaged over technical triplicate experiments by geometric mean. Compound identities for extracted precursor ion peaks were assigned by two modes of identification. First, identifications by isotopic standards were considered identified with full confidence. Second, the remainder of peaks without standard-based verification was queried against a composite database of formula-mass information from the NIST version 14 , HMDB version 3 , MassBank , and LIPIDBLAST  libraries based on the similarity of neutral mass (mass error). The distribution of mass errors against the best matching compounds was decomposed into true and false identifications, where the sub-distribution of the false identifications was learned from that of the second-best matching database entries of all peaks. The mixture deconvolution enabled us to compute the posterior probability of true identification, up to unique formula (subject to the uniqueness of mass value per structure). Compound assignments of probability 0.8 or above were taken as positive identifications, which corresponded to less than 20 ppm mass error in all cases. Finally, the identifications resulting in assignment of two or more isomers (of identical formula) were further removed from the data to avoid ambiguity of identification. Positive and negative ion mode data were processed separately throughout this process. For statistical analysis, the data were log-transformed (base 2) after adding a small fudge factor, determined by the 10 percentile point of each data set (by organ, by ionization mode) to avoid over-estimated fold change originating from noisy peaks with low peak area values. All statistical analyses were performed using R (http://cran.r-project.org), including principal component analysis, two-sample independent t tests and q value calculation, and generation of heat maps and volcano plots. In differential abundance analysis, the quantitative data matrix was obtained by summing peak area values of all peaks with identical compound identity in each sample, since abundant species tend to appear in multiple peaks, indicating long chromatographic elution. Two-sample t test followed by multiple testing correction (q value, ) was used for testing differential abundance in all comparisons. For the kidney data, as the total number of significant findings was very small, we applied q value of 0.15 (15% FDR) as the threshold of statistical significance. In the plasma data, we applied q value of 0.05 (5% FDR) to control the total number of false discoveries reasonably low. We also applied minimal 20% change as additional requirement in the selection of significant findings.
Bacterial DNA extraction from mouse fecal pellets and 16S rRNA gene PCR amplification
Feces were freshly collected and immediately snap frozen in liquid nitrogen before extraction using QIAamp Fast DNA Stool Mini Kit (#51604). Extracted bacterial DNA was amplified using the primer pair 338F* (PKO6597) and 1061R (PKO6598), as has been described . All primers are listed in Table S1.
16S rRNA gene amplification and sequencing
We used the method that has been described by Ong et al.  and recently by Ta et al. . Our comparison groups consisted of 7 WT and 7 SugctKO mice that were all co-housed for consistent exposure to the same microbial environment, which meant that always 2 WT and 2 SugctKO mice were housed in the same cage. We identified significant differences between the averages of the two groups with one-sided t test assuming equal variance (p < 0.1) and rank them by percent increase or decrease in SugctKO mice (Fig. 3a). In addition, we summarize the total composition of the microbiome as pie charts (Fig. 3b) and Sankey diagrams (Fig. 3c) generated using a sequence similarity cut off [30, 71] and the bacterial accession in the Greengenes database.
For hematoxylin and eosin (H&E) staining, tissues were fixed in 10% neutral buffered formalin (NBF, Sigma-Aldrich, HT501128) for 18–24 h, transferred to ice–cold 70% ethanol, and embedded in paraffin blocks followed by the staining of tissues sections.
Oil red O (ORO) staining
Slides with frozen tissue sections were dipped in isopropanol 60% for 5 min and stained in filtered ORO working solution [3:2—ORO stock (1% ORO in isopropanol): Dextrin (1% Dextrin in water)], followed by rinse in isopropanol 60%. Then, slides were counterstained in hematoxylin solution for 2 min, washed in deionized water for 1 min, followed by the wash in tap water for 5 min.
Slides with paraffin-embedded tissue sections were deparaffinized and rehydrated at room temperature for 5 min, followed by blocking of endogenous peroxidase in methanol/H2O2 (1:33.3) for 15 min and rehydration in water for 1 min. Subsequently, slides were incubated with 20 mg/mL Proteinase K (Invitrogen, #25530-049) for 20 min at 37 °C, washed 2× in Phosphate Buffered Saline (PBS), and blocked with 1% Bovine Serum Albumin (BSA, Sigma-Aldrich, #A7906-100G) for 45 min at RT. After blocking, slides were incubated with primary rat anti-mouse F4/80 (Serotec, #MCA497G) antibodies (1/200) at RT, washed 3× in PBS followed by incubation with AffiniPure rabbit anti-rat secondary antibodies (1:5000, Jackson ImmunoResearch, #312 005-003) for 1 h at RT. Slides were washed 3X in PBS, treated for 30 min with Dako Envision + System-HRP Labeled Polymer Anti-mouse (DAKO, #K4001) for 30 min at RT, washed 3× in PBS, incubated with Dako Liquid 3,3′-Diaminobenzidine (DAB) + Substrate Chromogen System [(DAKO, #K3468), 1 drop of the DAB Chromogen per mL of Substrate Buffer] for 5 min at RT, counterstained with hematoxylin for 5 min at RT, washed 3X in PBS, rinsed with running water, and mounted.
All experiments were performed with a minimum of three animals. For data that followed a normal distribution, statistical significance was tested using the two-way Student t test. For LC–MS plasma analysis, we used two-way ANOVA. Data were represented, as the mean value and error bars represent the standard error of the mean (SEM). p value was calculated with two-tailed paired t test with 95% level of confidence.
Open access funding provided by Lund University. We thank Zakiah Talib for animal care and all past and current members of the Kaldis lab for support and discussions. We thank Roger Low for contributions at early project stages and Keng Hwee Neo for exploratory work and controls not included here. We thank Falicia Goh from Natural Product Research Laboratory in BII for the discussions about the analysis of the metabolomics data. We thank Norman Pavelka for discussions and Mark Lewandoski for the β-actin–Cre/Flpe mice. We acknowledge the technical expertise provided by the Advanced Molecular Pathology Laboratory at IMCB. Lino Tessarollo was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research. This work was supported by the Biomedical Research Council of A*STAR (Agency for Science, Technology and Research), Singapore, and in part by a Grant from the National Medical Research Council Singapore, NMRC (OFIRG15nov120), Natural Product Research Laboratory BMRC Transition Fund (H16/99/b0/004), and the National Research Foundation Singapore (NRF-CRP17-2017-06).
Compliance with ethical standards
Conflict of interest
Authors declare no conflict of interest.
- 1.Sinha S, Eisenhaber B, Jensen LJ, Kalbuaji B, Eisenhaber F (2018) Darkness in the human gene and protein function space: widely modest or absent illumination by the life science literature and the trend for fewer protein function discoveries since 2000. Proteomics 18:e1800093PubMedCrossRefGoogle Scholar
- 34.Adibi SA (1977) Clearance of dipeptides from plasma: role of kidney and intestine. Ciba Found Symp 50:265–285Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.