University of Birmingham Tracer-based metabolic NMR-based flux analysis in a leukaemia cell line

high glycolytic activity, although this is not linked to the Krebs cycle, but rather towards the production of ribose sugars in the PPP and lactate. More-over, we have observed glutamine as an anaplerotic source for the Krebs cycle. Our analysis also shows the effects of BaP treatment, specifically changes in succinate and a -ketogluta-rate levels, and a reduction of pyrimidine synthesis intermedi-ChemPlusChem intensity. In such cases, the peak intensity was set to the estimated noise level in that spectrum. That noise level is evaluated by searching for the maximum of the absolute value of the intensity seen in a region devoid of real signals. When calculating percentage incorporations of 13 C, it is the 12 C reference spectrum peak intensity that might be missing. Substituting the dummy peak intensity will cause a tendency of the percentage incorporation of 13 C to be (conservatively) underestimated“.


Introduction
Our previousw ork [1][2][3] and that of others [4] showedt hat high levels of reactive oxygen species( ROS) have ap rofound impact on acute myeloid leukaemia (AML) cells, and that susceptibility to ROS represents an Achillesh eel that can be used to specifically target AML cells with novel therapies.
We have previously shown the anti-leukaemic and anti-lymphoma activity of ac ombination (denotedB aP) of two redeployed drugs,t he contraceptive steroidm edroxyprogesterone (MPA) and the lipid-regulating drug bezafibrate (BEZ). [1] Consistently associated with BEZ treatment is the rapid and sustained generation of ROS. [1] We have also shown that BaP treatment has ap rofound effecto nt he metabolome of AML cell lines (KG1a,K 562, and HL60). [2] The most pronouncedc hanges were observedf or metabolites of the Krebs cycle. HL60 cells showed reducedl evels of fumarate compared to succinate, which was attributed to the formation of malonate under BaP treatment, which inhibits succinate dehydrogenase.
Given these findings, and the success of these NMR studies in elucidating mitochondrial insult by ROS,w eu sed at racerbased metabolic analysis (metabolic flux analysis) to characterise the effect of this combination of drugs on central carbon metabolism. Tracer-based metabolic analysis represents af orm of targeted metabolomics,i nw hich the distribution of 13 C from an isotopically labelled metabolic precursor is traced among various metabolites,a llowing for am ore quantitative interpretation of metabolic changes than is currently afforded by traditional 1D 1 Hs pectra. Moreover,N MR spectroscopy yields information on site-specific label incorporation. Tracerbased studies enable specific assignments to metabolic pathways, depending on the choice of the isotopically labelled metabolites. NMR spectroscopya nd mass spectrometry have been commonly used to determine the incorporation of isotopic precursorss uch as 13 C-labelledf orms of glucose and glutamine(reviewed in Ref. [5]).
Here we report at racer-based metabolic analysiso fK 562 cells with and without BaP treatment, in which we probe the use of glucose and glutamine in centralc arbon metabolism. We ascertained the early effectsa rising from BaP treatment by monitoring label incorporation, as well as alterations to metabolites observed only in natural isotopic abundance. We used [1-13 C]glucosea nd [1,[2][3][4][5][6][7][8][9][10][11][12][13] C]glucose as tracers to probe label incorporation in glycolysis and transfer into the Krebs cycle, and [3-13 C]glutaminet of urtherp robe the Krebs cycle and glutaminolysis. Label incorporationi nto metabolites was examined using 1 H- 13 Ch eteronuclear single quantum coherence (HSQC) NMR spectra. HSQC spectra show as ignal per CH group, and in addition, if acquired with sufficient resolution in the incre-High levels of reactive oxygen species (ROS) have ap rofound impact on acute myeloid leukaemia cells and can be used to specifically target these cellsw ith novel therapies. We have previously shown how the combination of two redeployed drugs, the contraceptive steroid medroxyprogesterone and the lipid-regulating drug bezafibrate exert anti-leukaemic effects by producing ROS. Here we report a 13 C-tracer-basedN MR metabolic study to understand how these drugs work in K562 leukaemia cells. Our studys hows that [1,[2][3][4][5][6][7][8][9][10][11][12][13] C]glucose is incorporated into ribose sugars,i ndicating activity in oxidative and non-oxidative pentosep hosphate pathways alongside lactate production.There is little label incorporation into the tricarboxylic acid cycle from glucose, but much greater incorporation arises from the use of [3-13 C]glutamine. The combinedm edroxyprogesterone and bezafibrate treatment decreases label incorporation from both glucose and glutamine into a-ketoglutarate and increased that for succinate, which is consistent with ROS-mediated conversion of a-ketoglutaratet os uccinate. Most interestingly,t his combined treatment drastically reduced the production of severalpyrimidine synthesis intermediates. mented 13 Cd imension, show couplings between adjacent carbons. The former datap rovide direct information abouts itespecific label incorporation rates, the latter provides information about label incorporation at adjacent carbon positions. [5] Results HSQC spectra of K562 cell extracts allowed unambiguous identification of approximately 40 metabolites, some of which showedi sotope enrichmentf or cells grown for 24 hw ith [1,[2][3][4][5][6][7][8][9][10][11][12][13] C]glucose or [3-13 C]glutamine.C ells wereg rown with labelled precursors for 24 h, as this is the time frame in which label incorporation can be observed into aw ide range of metabolites. Reference spectra from cells grown with glucosea nd glutamine of natural isotopic abundance were prepared as controls. Although apoptosis or differentiation is usually not evident until about 96 h, our previouswork showedadistinct metabolic response after 24 h. Therefore, we also chose a2 4h time frame for the exposure to BEZ and MPAr unning in parallel with label incorporation. For many of the observed metabolite resonances, correspondingNMR signals could also be observed in unlabelled samples grown with natural isotopica bundance glucoseorg lutamine.
Glucose mainly feeds the pentose phosphate pathway and glycolysis  Figure 2a nd Ta ble S1 (in the Supporting Information) summarise the experimentally observed incorporation of labels for some key metabolites arising from [1,2-13 C]glucose, observed and quantified from K562 cells by HSQC spectra (see Table S1 for further details).

Labelling in ribose moieties
The highest isotopice nrichment levels observed were of the ribose sugar resonances, due to pentose phosphate pathway (PPP) activity.Interestingly,labellingofthe ribose sugars is indicative of am ix of both oxidative and non-oxidative PPP activity.I nt he oxidative PPP,t he label from the C1 positiono fg lucose is lost when CO 2 is released by the action of 6-phosphogluconate dehydrogenase, yieldingr ibose sugars labellede xclusively at C1. This results in as ingle ribose signal for C1 in the HSQC spectrum.I nn on-oxidative PPP,t he 13 C1À 13 C2 moiety is retained and gives labelling of C1ÀC2 and C4ÀC5 moieties in the resulting ribose sugars. These two 13 Cf ragments produce doubletsa rising from scalar J CC couplings observed at C2, C4 and C5. [6] From C1, three signals arise:t he middle singlets ignal arises from the oxidativeP PP with C2 unlabelleda nd the outer doublet arises from the non-oxidative PPP.A sa ne xample, Figure 3s hows the well-resolved adenine ribose C1 signal from nicotinea denined inucleotide (NAD + ). The signal arising from the natural abundance sample is very weak (shown in black) relative to the signals measuredi nt he spectrum obtained from the [1,2-13 C]glucose-fed cell extracts (blue). Figure S1 shows all the 13 Cs ignals for uridine diphos-  www.chempluschem.org phate (UDP). Here, the C1 regions suffer from substantial signal overlap of the variousu ridine nucleotide species. However,d oublets are clearly observed for C2 and C4, indicative of non-oxidative PPP activity.

Labelling in glycolysis intermediates
Other large enrichments wereo bserved for metabolites produced from pathways branching from glycolysis. For example, glycerol3 -phosphatea nd glycerophosphocholine are labelled predominantly at the C2 and C3 positions of the glycerol moiety due to processing from [2,3-13 C]glyceraldehyde 3-phosphate. Likewise, lactate and alaninea re labelled at the C2 and C3 positions due to the formation from [2,3-13 C]pyruvate.

Labelling in Krebs cycle intermediates
Lower levels of label incorporation were observed for metabolites arising from pyruvate enteringt he TCA cycle, predominately by pyruvate dehydrogenase( PDH)-catalysed condensation of [1,2-13 C]acetyl-CoAw ith oxaloacetate. The glutamate labelling pattern is consistentw ith the incorporation by PDH catalysis of 13 Ci nto aC 4C5 fragment from [1,2-13 C]glucose, along with lower label incorporation att he C2 and C3 positions, possibly arising from pyruvate carboxylase activity or further turns of the TCA cycle. The labelling patterns in both glutathione and pyroglutamate in the g-glutamyl cycle, and in proline are all consistent with flow of labelf rom glutamate ( Figures 1A and 2).
Similarly,a spartate, produced from the TCA cycle intermediate oxaloacetate, is clearly labelled at the C2 and C3 positions. Interestingly,t his label is not transferred into asparagine.H ow-ever,t he label in aspartate is fed into pyrimidine-base ring synthesis, as shownb yt he presence of multiplets in the spectra arising from the aspartate-derived portion of pyrimidine rings ( Figure S2).
Four-carbon TCA cycle intermediates such as succinate, fumarate, malate and aspartate showedlabelling of only approximately 5%,s uggesting that glucose is not the primary source of these TCA cycle intermediates.F ollowing this observation, additional experiments were conducted to probe the importance of glutamine as ap ossible nutrient source for the TCA cycle. Other resonancesarising from natural abundance metabolites were also observed in HSQC spectra but are not specifically labelled from glucose: glutamine is the main anaplerotic substrate Considering that few glucose-derived 13 Cc arbon atoms entered the Krebsc ycle we wondered whether glutamine could serve as an anaplerotic substrate in K562 cells. This was investigated using [3-13 C]glutamine as at racer.I ndeed, we observed much higherl abel incorporation in Krebs cycle intermediates compared to that observed if glucosew as used as at racer ( Figure 2, Table S1). The most intenselyl abelled carbon positions in metabolites (Table S1) corresponded to the expected labellingp attern that would arise from the conversion of [3-13 C]glutamine to glutamate, catalysed by mitochondrial glutaminase, and subsequently to a-ketoglutarate ( Figure 1B). The overall label incorporation into the cellular pool of the fourcarbon Krebs cycle intermediates was typically fourfold higher than that derived from [1,2-13 C]glucose (Table S1).
Outside of the mitochondria, pyrimidine carboxyltransferase catalyses thea ddition of aspartate to carbamoylp hosphate as part of the first step in pyrimidine nucleotide synthesis. Perhaps because of the much greater extent of labelling in aspartate, label incorporation was observed in severali ntermediates in this pathway;n amely N-carbamoylaspartate, dihydroorotate, orotate and orotidine, as well as in the pyrimidine rings of various uridine and cytidinen ucleotide species.
Curiously,n ol abel incorporation was seen in lactate, alanine or acetyl groups (by observing signals arising from the acetyl groups of N-acetylated amino acids and acetic acid). This contrasts sharply with work by DeBerardinis et al.,w ho found that 60 %o fg lutamine was converted to lactate and alanine in glioblastomac ells, ap rocess that involves malate exiting the TCA cycle and mitochondriaw ith subsequent conversion to pyruvate by am alic enzyme, thereby generating sufficient NADPH to support fattya cid biosynthesis. [7] Althoughi ti sc urrently not understood what regulates this process, it is increasingly clear that it is different between cancerc ell types.

BaP treatment increases intracellular glutamine and reduces de novo pyrimidine synthesis
As we have shown previously, [1] BaP treatment causes high levels of ROS, as ac onsequence of the action of BEZ (see also Figure S3). We observed as ignificant change in label incorporation with BaP treatment for afew metabolites only.The treatment decreased label incorporation from [1,2-13 C]glucose into a-ketoglutarate and increased that for succinate. This is consis- www.chempluschem.org tent with ROS-mediated conversion of a-ketoglutarate to succinate, and with the previousr esults of Tiziani and co-workers. [2] Equally,t here was little change in label incorporation observedi ns pectra from [3-13 C]glutamine-labelled cells upon BaP treatment. Label incorporation into succinate increased and labelling of a-ketoglutarate decreased, as observed for labelled glucose. Moreover,a spartate and acetyl-N-aspartate labelling increased slightly upon BaP treatment, whereas the label incorporation in pyrimidine synthesis intermediates N-carbamoylaspartate, dihydroorotate and orotate was significantly reduced. For the latter group, there were no signals present in the reference natural abundance spectra. Therefore, we reasoned that the reference signalh ad intensityb elow the noise level in the spectrum. Consequently,w ew ere able to calculate minimum label incorporations by assuming that the reference intensity was the maximum noise in the reference spectrum. To resolve this puzzle we also measured the 1D spectra of unlabelled samples ( Figure S4). These showedachange in the balance between UDP-N-acetylglucosamine and other UDP-con-taining species. Whereas just levels of UDP-glucose,U DP-glucuronatea nd UDP-galactose were clearly reduced with BaP treatment, UDP-N-acetylglucosamine increased.

EffectsofB aP on the g-glutamyl cycle
The levelso fm etabolites from the g-glutamyl cycle werea lso perturbed by BaP treatment ( Figure S5). 1 H1 Ds pectra of unlabelled samples showed that glutamate and particularly glutamine both increasedi nl evels on treatment with BaP (data not shown). Analysis of peak intensities in HSQC spectra derived from cell extracts grown in unlabelled media showedt hat BaP treatment greatlyi ncreased the amounts of glutaminea nd pyroglutamate, somewhat increased glutamate and slightly decreasedg lutathione levels (Table1). Overall this might suggest that cells respond to high concentrationso fR OS by increasing their glutamine uptake.
In spectra arising from [3-13 C]glutamine-treated cells, label incorporation at C3 can be assessed from the ratio of the signals of J CC -coupled doublet versust he singlet C4. This overcomest he problem that C3 signals of glutamine and glutathionea re significantly overlapped.I ti si mportant to note that the relative intensities of peaks in the glutathione multiplet did not change significantly with BaP treatment (Table 2), demonstrating no change in label incorporation.  [a] Arbitraryi ntensityv aluesf rom original data.
[b] myo-Inositol C1 and C3 were used as references, that is, it was assumed thatt heir levels were not affectedb yB aP treatment.S caled intensity = metabolite intensity/myo-inositol intensity.G lu, glutamate; Gln, glutamine;G SH,g lutathione; Myo, myo-inositol; Pyroglu, pyroglutamate. Label incorporation into glutamate was also not significantly affected by BaP treatment. By contrast, label incorporation from [3-13 C]glutamine into pyroglutamate increased threefold upon BaP treatment ( Table 2). The increaseo fp yroglutamate labellingi sc onsistent with cyclisation of glutamate or glutamine to pyroglutamate mediated by ROS.

Discussion
Overall,t he NMR analysis of label incorporation into metabolites arising from labelled glucose and glutamine precursors shows the significant potential of this approach for studying metabolic mechanismsi nh uman cell lines. Althought his has originally been recognisedb yS hulman and Ugurbyl and coworkers [8][9][10] and later by Szyperski, Bailey,a nd Wüthrich, [11,12] the methodo fc hoice for most subsequent studies has been mass spectrometry,p robably owing to its much highers ensitivity.H owever,t he limitation of mass spectrometry is that it provides less information from mass increments, whereas NMR can in principle yield additional information for site-specific label incorporation. If [1,[2][3][4][5][6][7][8][9][10][11][12][13] C]glucose is used, furtheri nformation arises from the 1 J CC scalar coupling between adjacent carbons, especially as the C1 atom in glucose is abstracted in the entry step to the oxidative PPP,b ut not for the non-oxidative PPP branch. In some cases, this is also useful to determine label incorporationsi nto carbon atoms that cannot be easily resolved owing to overlap if an adjacent CH can be observed as ar eporter.T his has been shown for glutamine-related metabolites.
HSQC spectra show al arger number of metabolites than directly observed 1D 13 Cs pectra,o wing to its much higher sensitivity.T his is of course limited to carbon atoms with an attached proton, although label incorporation into CO and COO groups can often be inferred from the signals of neighbouring CHs. In order to resolve J CC couplings, at least 4096 increments need to be acquired, resulting in an overall acquisition time of more than 4hfor two incrementswith arecycle time of 2s.

K562 centralcarbon metabolism
This analysisp rovides new insights into the metabolism of K562 leukaemic cells. This cell line wasd erived from aB CR-ABL translocation-positive blast-crisis chronic myeloid leukaemia and is an established cell-line model for leukaemia. Our data demonstrates that large amountsofg lucose are funnelled into the PPP,a so bserved by 13 Cl abellingo fr ibose. The observed labelling pattern is indicative of am ix of oxidative and non-oxidative PPPs, as evidenced by C1-C2 and C4-C5 couplings. This was furthers upported by label incorporationi nto variousr ibose sugars arising from [1-13 C]glucose, which is only possible by the non-oxidative pathway,f or example, 36 %l abel incorporation into C1 of adenosine diphosphate.
Furtherl abeli ncorporation was observed in other glycolysisrelated metabolites such as glycerol 3-phosphate and lactate, althoughl evels were lower than those observed for ribose sugars.I nt he Krebs cycle, label incorporations in glutamate, succinate, fumarate and malate were all low ( % 5%)f rom glu-cose and larger (13-50 %) from glutamine. Thisi si na ccordance with as trong Warburg effect, by which lactatei sp roduced and entry of pyruvate into the Krebs cycle is blocked. Large amounts of glucose-derived label divertedi ntot he PPP have previously been observed for cancer cells. [13] The high levels of labeli ncorporation from glutaminei ndicate highly active glutaminolysis.
In K562 cells, glutamine-derived label was not observed in lactate or alanine. This is in sharpc ontrast to the work of De-Berardinis et al., [6] who found that 60 %o fg lutamine converted to lactatei ng lioblastoma cells. Neitherd id we observe label incorporation into asparagine, although aspartate is labelled from glucoseand, to alesser degree, from glutamine.

Effects of BaPt reatment
Tracer-based metabolic analysis of BaP-treated K562 cells confirmed that a-ketoglutarate levelsa re reduced, whereas succinate was increased after a2 4htreatment with BaP.I mportantly,B aP treatment reduced label incorporation into pyrimidine synthesis intermediates, especially those of N-carbamoylaspartate, orotic acid and dihydroorotic acid. The de novo synthesis of pyrimidines is crucially important for the cell and its perturbationi sl ikely to have huge implicationso utside of the immediate synthesis pathway as highlighted in ar ecent study by He et al. in 2014. [14] After assembly from bicarbonate, glutamine and adenosine triphosphate, uridine and cytidinen ucleotides are fuels for the synthesis of RNA, DNA, phospholipids, UDP sugarsa nd glycogen.I nt his study,w eh ave fort he first time trackeda ll stages of this process through the labelling of aspartatef rom our [3-13 C]glutamine source. It remains puzzling why label incorporation into the pyrimidine end product is alwaysh igh and not significantly affected by BaP treatment. One possible explanation is that the pyrimidine pool builds up prior to the inhibition of pyrimidine synthesis or that the pyrimidines alvage pathway becomes more active to maintain these pools.
BaP treatment causes an unexpected shift in the balance of different UDP species, showingd ecreasesi nt he amounts of UDP-glucose, UDP-glucuronate and UDP-galactose and an increasef or UDP-N-acetylglucosamine. This is apparent from the ribose signals of UDP compounds in 1D spectra of unlabelled samples, whereas the pyrimidine signals of individual UDP species are not separately resolved.

Conclusion
The data presentedinthis study highlightsthe potential of isotopic tracer-based profiling in understandingc ancer cell metabolism and any subsequent metabolic changes arising from treatment. K562 cells clearly display high glycolytic activity,a lthought his is not linked to the Krebs cycle, but rather towards the production of ribose sugars in the PPP and lactate. Moreover,w eh ave observed glutamine as an anaplerotic source for the Krebsc ycle. Our analysis also shows the effects of BaP treatment, specifically changes in succinate and a-ketoglutarate levels, and ar eduction of pyrimidine synthesis intermedi-
Cells were harvested by centrifugation (800 g,1 0min) and the pellets were washed with PBS and transferred to 1.8 mL glass vials in PBS (1 mL). Cells were centrifuged (4000 g,4 8C, 1min), then the PBS was removed and the cell pellet resuspended in methanol (400 mL) that had been pre-chilled over dry ice, thus freezing the samples. Samples were kept on dry ice and then stored at À80 8C. After removal from À80 8Cs torage, the vials were placed on wet ice and distilled H 2 O( 325 mL) and chloroform (400 mL, prechilled on wet ice) were added before vortexing for 30 sf ollowed by incubation on ice for 10 min to allow the phases to separate. Following centrifugation (1500 g,48C, 10 min), the polar (upper) fraction was transferred to Eppendorf tubes and dried overnight in ac entrifugal vacuum concentrator.

NMR data acquisition
Polar extracts were resuspended in metabolomics buffer (60 mL; sodium phosphate buffer,0 .5 mm trimethylpropanoic acid, 10 % D 2 O, pH 7) with vortexing, and supernatant (40 mL) was transferred to champagne vials. Supernatant (35 mL) was then transferred to 1.7 mm NMR tubes and kept at 4 8Cp rior to measurement.
All spectra were acquired at 298 Ko naBruker 600 MHz spectrometer with aT CI 1.7 mm z-PFG cryogenic probe using ac ooled Bruker SampleJet autosampler.I na ll experiments, the 1 Hc arrier was set to the water frequency and the 1 H908 pulse was calibrated at apower of 0.326 W.
For the 1 H-13 CH SQC spectra the pulse sequence used was based on the Bruker standard pulse program hsqcetgpsp, which uses echo/anti-echo time-proportional phase incrementation gradient selection, with additional gradient pulses to improve water suppression. Key parameters for the 1 Ho bservation dimension were: as pectral width of 7812.5 Hz, 2048 complex points in the direct dimension, 4096 increments for the 13 Ci ndirect dimension with as pectral width of 159.0 ppm. Spectra were acquired with two scans and an interscan delay of 1.5 s, giving at otal experiment time of approximately 4h.A ll spectra were processed using NMRLab [15] in MATLAB. Cosine-squared window functions were applied to both dimensions.

Analysis of 2D spectra
Peaks were picked in as emi-automated manner using Metabo-Lab. [16] To calculate percentage label incorporations, the cross peaks in labelled spectra (spectra from cell extracts grown in labelled media) and reference spectra (spectra from cell extracts grown in natural isotopic abundance media) were compared. The 13 Ci sotope constitutes about 1% of naturally occurring carbon. Therefore, for more concentrated metabolites, cross peaks could be observed in the reference spectra. Peak intensities in control and reference spectra were used to calculate the percentage incorporation of 13 Cl abels into particular carbon atoms of ag iven metabolite following the equation %incorporation = 100 N/(DS)w here N is the intensity of the selected peak of the metabolite in labelled media, D is the intensity of the selected peak of the metabolite in the control spectrum, and S is the mean of as cale factor.T he scale factor = N r /D r ,w here N r is intensity of peak i from ar eference metabolite in the numerator spectrum and D r is the intensity of peak i from ar eference metabolite in the denominator spectrum. The reference metabolite was chosen because it was one of ag roup of metabolites that did not change in intensity significantly between treated and untreated spectra or between spectra for enriched and natural abundance media. Results were similar using valine, leucine or isoleucine as the reference metabolite. Labelling was not considered significantly changed unless BaP treatment changed percentage label incorporation by at least afactor of 2.
In some spectra, peaks were not observed owing to their low intensity.I ns uch cases, the peak intensity was set to the estimated noise level in that spectrum. That noise level is evaluated by searching for the maximum of the absolute value of the intensity seen in ar egion devoid of real signals. When calculating percentage incorporations of 13 C, it is the 12 Cr eference spectrum peak intensity that might be missing. Substituting the dummy peak intensity will cause atendency of the percentage incorporation of 13 Ct o be (conservatively) underestimated".