Conditional targeting in mice reveals that hepatic homogentisate 1,2-dioxygenase activity is essential in reducing circulating homogentisic acid and for effective therapy in the genetic disease alkaptonuria

Abstract Alkaptonuria is an inherited disease caused by homogentisate 1,2-dioxygenase (HGD) deficiency. Circulating homogentisic acid (HGA) is elevated and deposits in connective tissues as ochronotic pigment. In this study, we aimed to define developmental and adult HGD tissue expression and determine the location and amount of gene activity required to lower circulating HGA and rescue the alkaptonuria phenotype. We generated an alkaptonuria mouse model using a knockout-first design for the disruption of the HGD gene. Hgd tm1a −/− mice showed elevated HGA and ochronosis in adulthood. LacZ staining driven by the endogenous HGD promoter was localised to only liver parenchymal cells and kidney proximal tubules in adulthood, commencing at E12.5 and E15.5 respectively. Following removal of the gene trap cassette to obtain a normal mouse with a floxed 6th HGD exon, a double transgenic was then created with Mx1-Cre which conditionally deleted HGD in liver in a dose dependent manner. 20% of HGD mRNA remaining in liver did not rescue the disease, suggesting that we need more than 20% of liver HGD to correct the disease in gene therapy. Kidney HGD activity which remained intact reduced urinary HGA, most likely by increased absorption, but did not reduce plasma HGA nor did it prevent ochronosis. In addition, downstream metabolites of exogenous 13C6-HGA, were detected in heterozygous plasma, revealing that hepatocytes take up and metabolise HGA. This novel alkaptonuria mouse model demonstrated the importance of targeting liver for therapeutic intervention, supported by our observation that hepatocytes take up and metabolise HGA.


Introduction
Alkaptonuria (AKU; OMIM #203500) is a rare metabolic recessive disease where the enzyme homogentisate 1,2-dioxygenase (HGD; EC 1. 13.11.5), which is mainly found in the liver, is deficient (1). Garrod in 1908 unveiled the term inborn error of metabolism and proposed that AKU was caused by the lack of an enzyme that in normal individuals split the aromatic ring of homogentisic acid (HGA) (2). Biochemical evidence of the defect in AKU was provided by La Du in 1958, where he demonstrated the absence of HGD activity in a liver homogenate prepared from an AKU patient and established that the failure to synthesize active enzyme was the sole cause of AKU (1).
HGD deficiency leads to HGA accumulation in the blood and tissues, despite urinary excretion. It has been proposed that excess HGA undergoes oxidation and polymerization to form a dark brown ochronotic pigment (3) that deposits in connective tissues such as the skin, sclera, spine and articular cartilage, as well as in heart valves (4,5), where it causes aortic stenosis (6). AKU patients suffer from early-onset severe osteoarthropathy based on premature degeneration of articular cartilage, and disease manifestations worsen with age. Despite liver deficiency of HGD, the main pathophysiological manifestation of AKU relates to the function of non-metabolized HGA in the joints.
Treatment of other inborn errors of metabolism related to the phenylalanine/tyrosine pathway by enzyme replacement has been attempted with some success (19). In the past decade, liverdirected gene therapy has emerged as a promising alternative to transplantation in monogenic liver disorders such as AKU (20). The level of HGD required to rescue the disease, if expression outside the liver can affect the phenotype and whether circulating HGA can be metabolised by HGD-expressing cells are essential questions that must be addressed before such treatments are investigated.
To answer these questions, a new targeted knockout-first AKU mouse model was generated. This mouse harbours a LacZ reporter gene within the HGD locus for localizing gene expres-sion and was conditionally manipulated to obtain an inducible and liver-specific knockout. This study provides compelling evidence that targeting hepatic HGD plays an indispensable role in any enzyme replacement or gene therapy for AKU.

Generation of the conditionally targeted Hgd mouse
ES cells from clone C10 resulted in chimeras which achieved germline transmission. This knockout-first allele (Hgd tm1a) contained an IRES:LacZ gene trap cassette and a promoterdriven neo cassette inserted into the fifth HGD intron with the sixth exon flanked by loxP sequences (see Fig. 1A) (21,22). Homozygous Hgd tm1a mice showed an AKU phenotype based on HGD gene disruption. In C10/tm1a mice, Lox-F/Hgd-R primers amplified the loxP sequence (257 bp) showing the allele was floxed (Fig. 1). Hgd-F/Hgd-ttR primers produced a 561 bp band in the wild-type allele; the gene trap cassette sequence in the modified allele was too large to be amplified. Homozygous tm1a therefore had only the 257 bp floxed band, heterozygotes had both the floxed 257 bp and wild-type 561 bp bands and wild-type had only the 561 bp band (Fig. 1B).

Detection of ochronosis and its progression
Knee joints from Hgd tm1a −/− mice aged 7-40 weeks were examined for pigmentation. Ochronosis was found in calcified articular cartilage ( Fig. 2A), first appearing at 9 weeks (Fig. 2B). The pigment was initially pericellular (9-11 weeks) and very infrequent. At 26 and 40 weeks ( Fig. 2C and E, respectively), the number and intensity of pigmented chondrons were increased and showed advancement to the intracellular compartment. Clusters of pigmented chondrons were seen at ligament attachment sites (not shown). At 40 weeks, pigmentation was still confined to calcified cartilage. Heterozygous controls showed no pigmentation at 26 and 40 weeks ( Fig. 2D and F, respectively).

Adult HGD expression
Adult tissues from Hgd tm1a −/− were stained for LacZ to visualize HGD expression as blue staining. Positive staining was present in the liver and kidney cortex after 2 h which intensified when left overnight (Fig. 3A). Heterozygous staining was less intense (not shown) and was slower to develop. Hgd tm1c wild-type-like liver and kidney did not stain (Fig. 3A). All other Hgd tm1a tissues investigated including the brain, heart, lung, muscle, spleen, intestine, skin, bone, cartilage, eye and prostate were negative. HGD mRNA analysis via qPCR (HGD1 primers spanning Exons 3-4 before gene trap cassette) confirmed this staining pattern, with HGD expression only present in the liver and kidney (Fig. 3B). LacZ in the whole liver and kidney was limited by the penetration of substrate; therefore, frozen section staining was undertaken to demonstrate that LacZ and therefore HGD is expressed in the cytoplasm throughout the liver parenchyma ( Fig. 3C). In the kidney cortex, glomeruli were LacZ-negative, and only certain tubules stained positive (Fig. 3D), which were identified using periodic acid-Schiff for brush border staining in proximal convoluted tubules (PCTs) ( Fig. 3E and F).

Embryonic HGD expression
To determine when HGD is expressed, LacZ staining of timemated Hgd tm1a −/− embryos was carried out. Whole embryo staining showed positive LacZ staining in the liver at E14.5 and onwards (Fig. 3G). However, histological sections revealed punctate staining at E12.5 and onwards in the liver (Fig. 3H). Positive LacZ staining was seen in the kidney at E15.5 in some of the developing kidney tubules (Fig. 3H). All other embryonic tissues examined in frozen sections, including brain, eye, bones and other internal organs, were LacZ-negative.

Inducible and liver-specific HGD knockout
To investigate the effect of liver-specific HGD gene deletion, double transgenic Hgd tm1d MxCre +ve mice (n = 5) and wildtype (n = 3) and AKU (n = 4) controls were injected with pIpC. Blood and urine samples were collected according to the scheme in Figure 4A. Fifteen days after the first pIpC injection, MxCre +ve mice showed a 77.6% decrease in liver HGD mRNA in contrast with wild-type controls (Fig. 4B). Kidney HGD expression in MxCre +ve mice did not change and was comparable to wildtype controls (Fig. 4C). AKU mice had no HGD mRNA expression as expected (primers span Exons 9-10 after gene trap cassette).

Long-term liver-specific HGD knockout
MxCre +ve mice (n = 15) and AKU controls (n = 5) were injected with pIpC. MxCre +ve mice were injected with PBS (n = 5) as mice. Injections and sampling of blood and urine were carried out according to the scheme in Figure 5A. MxCre +ve mice injected with pIpC were culled at 9 (n = 5), 15 (n = 5) and 20 (n = 5) weeks post-injection, with wild-type and AKU controls culled at 20 weeks. Liver and kidney mRNA was taken, and knee joints were taken to assess ochronosis.
As with the previous study, liver HGD mRNA was reduced in pIpC-injected MxCre +ve mice, which was sustained to 20 weeks post-injection (see Fig. 5B). In contrast to wild-type controls, liver HGD expression was 11.7, 17.7 and 18.4% at 9, 15 and 20 weeks, respectively. Kidney HGD expression in the MxCre +ve mice remained comparable to wild-type controls (Fig. 5C). The reduction of liver HGD mRNA in the MxCre +ve mice subsequently caused plasma HGA to increase (Fig. 5D) to a level comparable with AKU controls. Urinary HGA (Fig. 5E) was increased in the MxCre +ve mice but not to that of AKU controls. Knee sections stained with Schmorls' stain were scored to obtain the number of pigmented chondrons found in a representative knee joint section (Fig. 5F). Few or no pigmented chondrons were found in the MxCre +ve knee joints 9 weeks post-pIpC, increasing in number at 15 and 20 weeks. Wild-type controls showed no pigmentation.

Liver-specific HGD knockout: dose response
In order to investigate the effect of varying liver HGD mRNA expression levels on the AKU phenotype, a short-term dose response study was carried out. Figure 6A demonstrates the study design. MxCre +ve mice were given two injections of pIpC at the following doses: 3.33 μg/g (n = 3), 1 μg/g (n = 3), 0.33 μg/g (n = 3), 0.1 μg/g (n = 5), 0.03 μg/g (n = 3) and 0.01 μg/g (n = 3) body weight. AKU (Hgd tm1a −/−; n = 4) and wild-type (MxCre WT; n = 7) controls were given the highest dose of 3.33 μg/g pIpC. With the exception of the 0.01 μg/g group, liver HGD mRNA in MxCre +ve mice was reduced at 15 days post-injection (Fig. 6B) in all groups, with a dose response observed at the lower pIpC doses. The mean liver HGD expression in MxCre +ve mice 15 days post-pIpC in contrast to wild-type controls was lowered to 17.0, 20.5, 20.5, 53.4 and 54.5% with decreasing pIpC doses from 3.33 to 0.03 μg/g. The mean liver HGD expression in the 0.01 μg/g group was comparable to the wide-type controls. Kidney HGD expression was unchanged (Fig. 6C).

Discussion
Management of inborn errors of metabolism such as AKU has traditionally consisted of diet and supportive therapy. However, other treatment options have become available, including enzyme inhibition (17,24), enzyme replacement (19), cell and organ transplantation (20), gene therapy (20) and CRISPR technology (25).
The pathophysiology of AKU has been investigated in ex vivo tissue samples (14,26,27) and in vitro models (13,28), but to investigate the metabolic consequences of AKU and novel therapeutic approaches, we have generated a well-characterized animal model. This model will be used to investigate all aspects of AKU pathophysiology including the mechanism of ochronosis and any associated tissue changes including amyloidosis (29). We have generated a HGD knockout-first mouse model that included a beta galactosidase (LacZ) gene trap within the HGD gene locus, which has enabled precise localization of HGD expression. Targeted gene disruption in Hgd tm1a removes any potentially confounding mutations that could be present in an existing ENU (N-ethyl-N-nitrosourea) AKU mouse model (30), as ENU mutagenesis causes a high frequency of genomic mutations (31). This new mouse recapitulated the human disease. Manipulation of this Hgd tm1a knockout-first allele by FRT/flp and Cre/loxP recombination enabled liver-specific HGD deletion in double-transgenic Hgd tm1d MxCre +ve mice, highlighting important considerations for future therapy in AKU.
Plasma HGA in Hgd tm1a −/− is comparable to that previously reported values in the ENU AKU mouse (30). This mutagenesis model exhibited the first signs of ochronosis at 15 weeks (30). Knee joints were therefore examined from 7 to 11 weeks in Hgd tm1a −/− mice, with pericellular pigmentation identified at Figure 5. Long-term follow-up of Hgd tm1d MxCre +ve mice injected with pIpC. A shows the timescale of blood and urine sampling and pIpC injections. Hgd tm1a −/− were injected with pIpC (AKU controls, n = 5). Hgd tm1d MxCre +ve were injected with PBS (wild-type controls, n = 5). Hgd tm1d MxCre +ve mice were injected with pIpC (MxCre +ve, n = 15). B and C show relative HGD mRNA (HGD2 primers; relative to 18S) in the liver and kidney, respectively, of MxCre +ve mice at 9, 15 and 20 weeks after the first injection, and in AKU and wild-type controls at 20 weeks. Liver HGD mRNA was reduced, in contrast to wild-type controls, in MxCre +ve mice at 9 weeks and was sustained until 20 weeks. Kidney expression was not reduced by pIpC injection in MxCre +ve mice. D and E show plasma and urine HGA levels, respectively, pre-injection and at 2, 9, 15 and 20 weeks post-injection. Pre-injection, plasma HGA is not detected in MxCre +ve or WT mice and AKU mice showed elevated HGA. Post-pIpC, plasma HGA in MxCre +ve mice was increased at 2 weeks and remained at this level until 20 weeks. Pre-injection, urinary HGA is low in MxCre +ve and WT mice in contrast with AKU controls. Post-pIpC, urinary HGA showed a relatively small increase in MxCre +ve mice in contrast with AKU controls. The total number of pigmented chondrons in a representative section of the knee joint, stained with Schmorl's, is shown in F, at 9, 15 and 20 weeks post-injection. 9 weeks; progression was then similar to mutagenesis AKU mice. Ochronosis in the mouse appears to represent the early stages of human joint pathophysiology in AKU with pigmentation confined to individual chondrocytes and their territorial matrix in the calcified cartilage.
The LacZ reporter gene has enabled both temporal and spatial histological localisation of HGD showing that HGD was expressed throughout the liver parenchyma and kidney PCT cells. It has previously been suggested that HGD is expressed in the intestine and prostate (12), brain (14) and bone/cartilage (13), but this was not evident using this knock-in LacZ reporter gene, nor by qPCR analysis of HGD mRNA.
LacZ staining of time-mated embryos demonstrated that hepatic HGD expression begins at E12.5 (Fig. 3H), confirmed with qPCR analysis of liver HGD mRNA from E12.5 to E15.5. Hepatic cords, containing hepatoblasts, have formed by E10.0 in the developing liver alongside haematopoietic cells. The liver then expands as a result of hepatoblast proliferation and haematopoietic activity from E10.5 to E11.5 (32). Haematopoietic activity rapidly increases, peaking at E13.5, and does not start to decline until E15.5 (32). Haematopoietic cells encompass almost 75% of total liver volume at E13.0 (33) with hepatoblasts at E13.5 having limited contact with each other (32), explaining the diffuse and punctate LacZ staining and low HGD mRNA level in the embryonic liver. Murine hepatoblasts begin to differentiate into hepatocytes at E14.5 (32). Haematopoietic activity continues into the first post-natal week, which may explain why Day 1 pups have about an eighth of adult HGD expression (32). In the kidney, adult LacZ staining suggests that the LacZ-positive cells seen at E15.5 are developing PCT cells of the nephron.
HGD expression begins in embryonic development. The 3fold greater HGA level seen at birth in contrast to adulthood in Hgd tm1a −/− mice (Fig. 1G) highlights the importance of HGD, even at this very early time point, perhaps suggesting that therapeutics strategies, such as nitisinone (16) or gene/enzyme replacement, should ideally begin in early life or at birth.

Liver-specific HGD deletion
To investigate the contribution of non-hepatic HGD towards HGA metabolism, double transgenic mice were generated by mating floxed Hgd tm1d mice with an MxCre recombinase line and used for liver-specific HGD deletion (23,34). Two doses of 10 μg/g body weight pIpC (Figs 4 and 5) resulted in an approximate 80% reduction in liver HGD mRNA, whilst kidney HGD mRNA was maintained at the wild-type level. This reduction of liver HGD mRNA and subsequent increase in plasma HGA to a level comparable with AKU controls suggests that hepatic HGD is crucial for HGA metabolism.
These results indicate that future gene/enzyme replacement therapy should target the liver to combat elevated plasma HGA that causes ochronosis in AKU. This is supported by a case report of an AKU patient receiving a liver transplant, after which they found no HGA in the urine and reported a halt in progressive arthropathy (35). Full liver HGD mRNA knockout was not achieved with two 10 μg/g doses in the present study. Further reducing liver HGD mRNA would not provide further insight into the level of liver HGD mRNA required to rescue the phenotype.
The data here reveals that approximately 20% liver HGD mRNA, delivered by gene therapy for example, will not rescue AKU; elevated HGA subsequently caused ochronosis in the knee joints of the mice (Fig. 5F). Determining how much liver HGD is required to rescue the phenotype is an important question for future therapy. The dose response study here (Fig. 6) intended to estimate how much liver HGD mRNA would significantly lower circulating HGA. A dose response was observed in both the plasma and urine HGA levels ( Fig. 6D and E). However, for liver HGD mRNA, there was only a dose response between 0.33 and 0.01 μg/g pIpC. One possible explanation for HGD mRNA not corresponding to plasma HGA in the other dose groups could be that the mRNA:protein ratio is not linear. Thus, we suggest that the minimum level of liver HGD mRNA required to eliminate circulating HGA must fall between the dashed lines (Fig. 6F), between 26 and 43% of liver HGD mRNA.
The intact kidney HGD mRNA did not have an impact on plasma HGA, but instead caused reduced urinary HGA in contrast to AKU controls. In 2002, an AKU patient who received a kidney transplant reportedly had normalized plasma HGA and decreased urinary HGA (both approximately half pre-transplant levels) (36). Liver-specific HGD deletion shown here in mice however suggests that kidney HGD mRNA is unlikely to rescue the AKU phenotype as blood HGA levels were elevated. The improvement reported with kidney transplantation was likely based on improved renal elimination as the patient had renal failure and subsequently very high HGA pre-transplant, rather than donor HGD expression. Indeed, in the conditional mouse model (Figs 4 and 5), we were expecting high HGA in the urine when circulating HGA increased. However, in the liver-specific knockout, the urine level did not increase, suggesting HGA reabsorption and subsequent metabolism by the intact kidney HGD, for which we do not know the mechanism. Fundamentally, the importance of this finding is that intact kidney HGD does not rescue the disease and therefore liver HGD is critical for the correction of AKU.

HGD activity and HGA metabolism: considerations for gene therapy
Restoring HGD activity in all liver cells, via gene therapy or enzyme replacement, is unachievable by any current method, but the data here suggests that it is not necessary. In heterozygous mice and humans, one functioning copy of the HGD gene in all cells is sufficient to deal with HGA metabolism. Assuming heterozygous mice possess 50% of the HGD mRNA in contrast with wild-types (adult LacZ staining suggests this assumption is correct), then 50% HGD mRNA in every hepatocyte can rescue AKU. As described previously, conditional deletion to 20% of total HGD liver mRNA did not appear to rescue the phenotype. However, we cannot determine how many cells were expressing the gene, either as one or two alleles, nor the distribution of expression which could be variable from cell to cell. The proportion of corrected cells and the level of therapeutic gene expression per cell required to rescue AKU is therefore still not clear. A recent study providing the first human genotypephenotype correlation data for the three most frequent HGD mutation variants (33 patients homozygous for these variants) identified in the SONIA-2 trial demonstrated no difference in baseline visit serum or 24 h urine HGA, or clinical symptoms such as eye pigmentation, hip bone density or degree of scoliosis between patients predicted to have 1% (16 patients) or 31-34% (17 patients) residual HGD catalytic activity as determined in vitro against wild-type recombinant human HGD enzymes (9).
Another consideration for gene therapy is that hepatocytes without sufficient HGD activity to metabolize HGA could lead to its accumulation in the bloodstream. In this study, intravenous injection of isotopically labelled HGA (Fig. 7) provides evidence that circulating HGA can re-enter HGD-expressing cells to be metabolized by intracellular HGD. Thus, it should not be necessary to repair 100% of liver cells because HGA produced by AKU hepatocytes could be taken up and metabolized by genetically repaired cells. This model represents a paradigm for inherited liver metabolic diseases, in particular, genetic disorders of tyrosine metabolism.

Conclusion
In summary, this new targeted HGD knockout mouse exhibits the characteristic traits of AKU. Both adult and embryo HGD expression has been localized to only the liver and kidney using a reporter gene within the HGD locus. More importantly, the conditional Hgd tm1d MxCre model has highlighted the importance of liver HGD expression in limiting the pathological effects of the HGA pool despite apparent reabsorption of HGA in the kidney.

Generation of HGD knockout-first mice
Two ES cell lines (clones Hgd C10, Hgd C11) were obtained from the UC Davis KOMP repository. They were grown on feeder cells as described (37). Healthy ES colonies were injected into blastocysts of C57BL/6 (Harlan, UK) and chimeras born from both lines. Only C10 chimeras achieved germline transmission. All mice were housed and maintained within the University of Liverpool Biological Services Unit in specific pathogen-free conditions in accordance with UK Home Office guidelines. Food and water were available ad libitum.

Urine and blood collection
Tyrosine pathway metabolites in acidified urine and plasma from venous tail bleeds were analysed via high-performance liquid chromatography (HPLC) tandem mass spectrometry assays (38,39).

LacZ staining: tissues/embryos
Embryos from E13.5 to E16.5 were stained for ß-galactosidase as previously described (40). Adult tissues were stained via the same protocol with size-adjusted fixation times.

LacZ staining: frozen sections
Frozen liver and kidney sections (6 μm) were fixed (0.2% glutaraldehyde in PBS) for 10 min. After three washes in cold PBS, they were stained for ß-galactosidase as mentioned previously and counterstained with eosin. Whole embryos were fixed as mentioned previously, transferred to 30% sucrose overnight and then embedded in OCT on dry ice. Frozen sections (7 μm) were LacZ-stained at room temperature overnight. qPCR RNA was extracted using the Qiagen RNeasy Mini Kit, reverse transcription of RNA was carried out using the Applied Biosystems RNA-to-cDNA kit and qPCR was performed using the Bio-Rad iQ TM SYBR ® Green Supermix. HGD1 primers are forward 5 -TGTCCACGGAACACCAATAA-3 and reverse 5 -GCCAACTTCATCCCAGTTGT-3 . HGD2 primers are forward 5 -GACCCATCGGAGCAAATGGC-3 and reverse 5 -AGTGTAACCAC CTGGCACTC-3 . 18S (housekeeping gene) primer sequences are forward 5 -TGTCCACGGAACACCAATAA-3 and reverse 5 -AGTTCTCCAGCCCTCTTGGT-3 . 18S was not affected by pIpC administration.

Ochronosis
Coronal knee joint paraffin sections were Schmorl's-stained and counterstained with nuclear fast red (28,41) to identify pigmented chondrons. Scoring of all four joint quadrants was carried out blind to experimental conditions and genotype.

Hgd tm1d conditional knockout
To obtain the conditional Hgd tm1d line, Hgd tm1a mice were crossed with Flpo mice to remove the FRT-flanked gene trap cassette, leaving a floxed target exon (Hgd tm1c) (42). Homozygote-floxed Hgd tm1c mice were crossed with MxCre mice. The removal of the floxed 6 th exon was induced with two intraperitoneal injections of polyinosinic:polycytidylic acid (pIpC) at 10 μg/g body weight (43). Hgd tm1a −/− mice were injected with pIpC as AKU controls, and wild-type controls were either Hgd tm1d MxCre WT injected with pIpC or Hgd tm1d MxCre +ve mice with PBS. HGA was measured in plasma and urine samples collected pre-injection and at time points post-injection. Liver and kidney HGD mRNA was analysed. In the long-term study, knee joints were collected for ochronosis scoring. For the dose response study, the same protocol was followed as mentioned previously, using diluted pIpC at doses 3.33 to 0.01 μg/g body weight.

Isotopic HGA injection
Hgd tm1a −/− (n = 4) and Hgd tm1a −/+ (n = 4) mice were injected with 13 C 6 -HGA into the lateral tail vein, adjusted to body weight to achieve a final blood concentration of ∼1 mmol/L. Under anaesthesia, venous tail bleeds were collected at time points post-injection, ranging from 2 to 60 min. Whole blood was centrifuged, and the supernatant removed and immediately frozen.
Non-targeted metabolic flux analysis was performed to trace metabolism of 13 C 6 -HGA. Metabolic profiling was performed using a published mass spectrometric technique (44). Briefly, plasma was diluted 1:9 plasma:deionized water and HPLC performed on an Atlantis dC 18 column (3 × 100 mm, 3 μm, Waters, UK) coupled to an Agilent (Cheadle, UK) 6550 quadrupole time-of-flight mass spectrometer. An accurate-mass compound database with potential association to HGA was generated for data mining using Agilent Pathways to PCDL. Data were mined for these compound targets with an accurate mass window of ± 5 ppm using 'batch isotopologue extraction' in ProFinder (build 08:00, Agilent). Isotopologue extraction investigates association with the injected 13 C 6 -HGA by examining the relative abundances of the M + 0-M + 6 isotopologues for compound targets.

Statistical analysis
Statistical analysis was performed using Stats Direct 3 statistical software (UK). Significance is denoted as P < 0.05 * , P < 0.01 * * and P < 0.001 * * * . Conflict of Interest statement. The authors declare no competing interests.