Newromage 189 (2019) 171-179 Contents lists available at ScienceDicect Neurolmage ELSEVIER, Journal homepage: wwnw.elsevier.comilocate/neuroimage Quantifying normal human brain metabolism using hyperpolarized [1-"°C]__ @® pyruvate and magnetic resonance imaging sa James T. Grist*, Mary A. McLean”, Frank Riemer*, Rolf F. Schulte“, Surrin S. Deen”, Fulvio Zaccagna°, Ramona Woitek “", Charlie J. Daniels", Joshua D. Kaggie °, Tomasz Matys Ilse Patterson“, Rhys Slough“, Andrew B. Gill", Anita Chhabra", Rose Eichenberger ", Marie-Christine Laurent", Arnaud Comment”, Jonathan H. Gillard’, Alasdair J. Coles’, Damian J. Tyler“, Ian Wilkinson“, Bristi Basu', David J. Lomas", Martin J. Graves“, Kevin M. Brindle’, Ferdia A. Gallagher * ppc fy Unter of Cantri, Care. UK © Caner Reach UK Canine ie, Umer of Con, Camb, UK Geno ee eater, Manch, Comey * Rady, Cant Uber Hep Candi, UK Dearne of Matin, Uren of Cardigan Cane Cina! ri Une Cate Unie apa MS eatin Tras, Cant UK “Dearne of Onan Ue of Cant, Cant Pharmacy, Carve Diner Hops Cambri, OR * ceo Canbre, RC Eel Un Cane, “GF Hetare Chom 8 Gile L \Depermene of lea arco, Unio Cari, Cai, UR * rte Pl, Arron, al Gee, Unt ef Oxford Oxo oper yperpoarized 5 Magnetic Resonance imaging (CMRI provides a highly sensitive tool t© probe tse Metabolism ‘metal in vvo and has recenly been translated into clinical aie. We report the cereal etal of yperoliend ‘ntravenoualy injected hyperpolarized (1-!Cpyrvate inthe brn of ealhy human volneers forthe fs tine. DDymamie acquisition of Fc images demonsrated "Clbeling of both Ictate and bicarbonate, cated by cyto lntate dehydrogenase and mitochondrial pyruvate dehydrogenase respectively, Ths demonstrates that both enzymes can be probed in vv in the presence ofa intact blcod- ain barrie the measured apparent ex change rte constant (iy) for exchange of the hyperpolarized 1 label between (1-C}pymsvate and the fexlogenos hctate pool as 0.012 0.006 "and the apparent rate constant (ky) for the reversible x of [1 Pechpyravate to cbicarbonate was 0.002 0.0025" Imaging also revealed that {1-CIpyravate, [1-°C) lactate and (CTbicartonate were significantly higher in gray mater compared to white mater. Imaging normal brain metabolism with hyperpolarized (1-C}pyruate and subsequent quantification, have important imple ‘on for interpreting pathological cerebral metabolism in fate ties. 1. Introduction (PEN, is one approach to imaging this cercbral metabolism in patients. Despite the sensitivity of PET, the signal aequired from "F-FDG repre: Cerebral metabolism is important for normal brain function and be- sents ux in only part ofthe glycol pathway measuring a combination ‘comes deranged in a number of pathological processes, such as inflam- of delivery t the tissue, uptake by glcose transporters and subsequent mation infection, ischemia, traumatic brain injury and in tumors GJalloh phosphorylation in the reaction eatalyzed by the glycolytic enzyme, ct al, 2015; Mathur et al, 20145 Matz ct al, 2006). "Fluorodeox- hexokinase. As the technique cannot detcet downstream products of glucose (FDG) uptake, detected using positron emission tomography glucose metabolism, such as lactate and CO>, it provides no direct * Corresponding author. Department of Radiology, School af Clinical Medicine, Box 218, University of Cambwidge, Cambridge, CB2 09, UK mal adres 3g 1000.e3m ek (EA. Gallagher). baps://doLorg/10.1016 neuroimage2019.01.027 Received 12 September 2018; Received in revised frm 8 Januiey 2019; Accepted 10 January 2019 ‘Availabe online 11 anvary 2019 10538119/0 2019 The Astor. Published by Hever Inc. This fan apen acess atic under the OC BY liens (hp://reativesommons.og/loenses/by/4.07.information on glycolytic fluxes and mitochondrial oxidative mete bolism, The work presented here uses a new imaging method to inves- tigate cerebral metabolism of pyruvate, a breakdown product of glucose. Pyruvate is transported across both the intact blox-bran barrier and the plasma membrane by the monocarboxylate transporters (MCTS). Pyruvate can either be metabolized to lactate, catalyzed by cytosolic lactate dehydrogenase (LDHD, oF is metabolized by oxidative decarbox- ‘ylation to acetyCoA, catalyzed by mitochondrial pyruvate dehydroge nase (PDH). Proton ('H) magnetic resonance spectroscopy (MRS) of the healthy brain has demonstrated steady state cerebral lactate concentra ons in the region of 0.6-1 mM, although it may be up to 2.2mMt in neonates, where glucose metabolism is altered compared to the adult brain Prichard et al, 1991; Tomiyasu et al, 2016). The metabolic shift from mitochondrial oxidative metabolism to glycolysis and lactate for mation occurs in a number of pathological processes, such as ischemia, {nglammation, and in tumors (alloh etal, 2015; Mathur etal, 2014; Matz e al, 2006). However, imaging of the spatial and temporal dist bution of lactate by "H-MRS is inhibited by a low signal-tonlse ratio (GNF) at clinical eld strengths. Therefore, alternative non-invasive ‘methods to image lactate would be valuable to monitor glycolysis in vivo. Hyperpolarized carbon-13 Magnetic Resonance Imaging ('°C-MRD, has emerged as promising technique for studying issue metabolism in hhumans (7ecagna et ly 2018). The method increases the SNR of liquid state carbon-13 MRS ("™GMRS) by more than four orders of magnitude (rdenkjacr-Larsen etal, 2011; Hurd et al, 2012). Tis substantial in ‘crease in SNR has been Wed to non-invasively image the spatial dst bution of intravenously injected 'Clabelled molecules in vive such as {1"erpyruvate, Importantly, dynamic '°C-MRI acquisition allows the nected hyperpolarized [1-*C]pyruvate o be differentiated from its metabolic produets such as [1-cHlactate, (1-!CJalanine and ¢labelled carbon dioxide/biearbonate as they form in real-time. There ‘are a number of challenges for imaging hyperpolarized carbon-13, metabolism, particularly the short halflife of the hyperpolarized ‘signal, which is typically 25-30 s for {1-"'clactateand (1-"'CIpyruvate in vivo but as low as 105 for (!°C]bicarbonate (Gallagher et a, 2008). ‘This limits the range of metabolites that can be probed using this tech nique. Furthermore, the hyperpolarized signal is ieeversbly depleted as images are acquired and therefore efficient imaging strategies are required to maximize the data that can be obtained, There ae a number ‘of seqtences that have been used in human studies and here we have ublzed IDEAL (erative Decomposition with Echo Asymmetry and Least ‘squares estimation) spiral chemical shift imaging (CSD), ait allows for spatial averaging to increase the SNR for shor lived metabolites (Geox ‘tal, 2016; Wiesinger eta, 2012), In order to quantify the dynamics of pyruvate metabolism, a number ‘of quantitative approaches for describing the exchange of hyperpolarized ‘carbor-13 label between pyruvate and lactate have been proposed, including both model-based and model-fee methods (Géme2 Damiin ct al, 2014; Khegai et al, 2014 Schulte et al, 2013). Modeling the process as a two-site exchange system, which gives the apparent ex: ‘change rate constant (ky) fr label flux between pyruvate and lactate, is the most accurate approach. However, time-to-peak (TTP) forthe latate signal intensity and the rato of the integrals ofthe lactate and pyruvate signals (area under the curve, AUC) are simple model-ree approaches that can also be used to estimate label Mux (Daniels et al, 2016). ‘Quantifying these metrics in normal tissue is important so that changes in disease tissue ean be understood and monitored over time. Frequency ‘domain kinetic modelling has been shown to be robust in low SNR en ‘vironments, such as hyperpolarized CMRI data. Furthermore, the frequency domain benefits from incorporating the arterial input function (AIF into the data, which would otherwise be challenging to acurately ‘estimate fom the low spatial resolution hyperpolarized images (kh etal, 2014), Previous studies in rodent and non-human primate brains have ‘demonstrated cerebral lactate labeling following injection of hyper- polarized (1-"lpyruvate, with a ky, of 0.0026 s-" reported in the Narain 1892019) 171.179 ‘macaque brain, albeit usinga higher dose than currently use for humans (-0.38mmoV/kg compared to ~0.11 mmol/kg) (Park et al», 2014) Although formation of hyperpolarized {"Clbicarbonate has also been observed in some rodent studies using a high pyruvate dose, it has not been reported in non-human primates. This suggests that cerebral py ruvate metabolism may be dose dependent, that there may be interspe- cies variation in pyruvate metabolism, or that metabolism may be affected by anaesthesia (Josan etal, 2013). Hyperpolarized [1-*Cpyruvate as recently been applied to patient studies with the first report in prostate cancer (Nelson etal, 2013), More recent studies have demonstrated the technique in normal human heart and as a treatment response marker in prostate cancer (Agarwal eal 2017; Cunningham et aly 2016). Lactate labelling has also been demonstrated in patients with brain tumors following therapy, where there is also preliminary evidence for the formation of corebal bicar- Donate (Park etal, 2018; Miloushev et al, 2018). However, as these tumors are highly invasive and have undergone therapy, the metabolism ‘of normal brain has not yet been established. Here we have used '°C-MRI {o image the conversion of hyperpolarized [1!*¢}pyruvate into both U-Mcjlaciae and [!Cjbicarbonate in the normal human brain and have quantified pyruvate metabolism in gray and white mater 2. Method and materials 21. Subject recrutment and sreening Local ethical approval was obtained for this prospective study (NRES. Commitee Fast of England, Cambridge South, REC number 15/EE/ (0255). Betwoen September 2017 and March 2018, four volunteers (mean age 27 2 years, one male, three female) were consented and sereened Prior to imaging; this included assessment of blood pressure, oxygen saturation, heart rat, electrocardiogram (ECG), and blood analysis (urea & clecrolyes fll lood count, serum lata, serum glucose, and latate dehydrogenase (LDH)). Only volunteers with normal serening, tests were included in the study. Blood sampling was undertaken prior to imaging and 30 min after. Oxygen saturation and heart rate were moni tored during pyruvate injection and throughout the examination. The volunteers were observed for up to 30min after the end of the 22. "%C pynvate preparation Pharmacy kits/fluid paths for insertion into the clinical hyper polarizer (SPINIab, ST, Research Cirele Technology, Niskayuna, NY) Were filled under sterile conditions. 1.478 [1-"']pyruvie acid (Sigma Aldrich, St Louis, Missouri, USA) containing 18mM{ of an electron paramagnetic agent (EPA, Syneom, Groningen, Netherlands) was sealed ina vial; 38 mL sterile water was used for dissolution; 19 mL sterile water with 17SmL NaOH/THs/EDTA (2.4%, 4.03%, and 0.083% w/v respectively, Royal Free Hospital, London) was used as a butler for neutralisation, Pharmacy kts were stored in a freezer at ~20°C for at least two weeks prior to use (Zaccagia etal, 2018). The vial containing the frozen pyrvate/EA mix was defrosted in the helium pressurised airlock in the hyperpolarizer for one hour. The sample was radiated at 139GHzat ~0.8K for approximately three hours. Following. rapid dissolution, the pyruvic acid was neutralised with the buffer and Quality Control (QC) checks were performed by an integrated QC module which measured: pyrwate and EPA concentration, pH, temperature, sample Polarisation and volume of dissolute. The release criteria for injection Were: pyruvate concentration 220-280mM; radical. concentration -<34M; pH 6.7-8.1; and temperanure 25-37°C. After release, the sample was passed through hatch into the adjacent MRI scanner room and O44 mi/kg of the final ~250 mM hyperpolarized pyruvate solution was injected at Sms using a syringe driver (Medrad, Warrendale, Pen sylvan, USA) followed by a saline Mush of 25mL at Sms23. Phantom imaging protocol Imaging was undertaken using a 3T MR system (MR750, GE Healtheare, Waukesha WD, using a dua-taned 'H/!°C quadrature head ‘oll (Rapid Biomedical, Rimpar Germany). ‘Transmit Gain (TG) and center frequency (f) were determined using, Bloch Siegert method (Schulte etal, 2011). To assess the "°C wansmit ‘and receive By sensitivity ofthe col, a uniform 16cm diameter sphere filled with pare polydimethyisloxane vas placed within a coilloading ring fled with saline (GE, GE Healtheare, Waukesha WI inside of the coll this was used to obtain signal ftom natural abundance carbon (Supplementary Fig, 14). ‘Transmit B, was determined by acquiring "¢ IDEAL spiral Cl images with 9 nominal Mipangles between 60° and 180° ‘and fiting a sine function to each voxel in the resulting images to determine the ratio of nominal to aetial Alp angle (@-step eyele inter. leaving one slice-selectve spectrum and seven spirals, Held of View (FOV) 400mm, slice thickness 40 mm, acquired resolution 26 « 26 mm", reconstructed resolution 8.3 mm, ip angles 70-180" in 10° increments, repetition time (TR) 15, echo shift 1.1 ms) (Wiesinger et al, 2012) 24, Clinica imaging protocol ‘The clinical imaging wos performed with the same MR system and ‘oll st up as for the phantom imaging. ‘The clinical "imaging protocol comprised: Ty-weighted volumetric Imaging (30; inversion prepared gradient echo; inversion time ~ 450 ms; FOV=240mm; TR=86ms; echo time (TE)=33ms; ip angle revolution =0.9%0.9 «1 mm}; By field map (COV=240mm, TE=7, 14s, TR=100ms, FA=20°, spatial resolu on =1 1x5 mm’) 56 transmit gain (TG) and center frequency (() were set using al ‘enriched urea phantom (8M, Sigma-Aldrich, UK) attached to the ear defenders worn by the subject. °C imaging was undertaken using a dy namic IDEAL spiral acquisition (pulse bandwidth = 2500 Hz, TR= 0.5 s time resolution — 45; FA 15"; FOV ~ 240 my acquired spatial resolu 12% 12mm? reconstructed resolution —§ » 5 mm sce thick: ness=30mm, acquired voxel volume=4.32em, total imaging time 160). Images nd spectra were reconstructed with 15 Hine broadening. Data acquisition began 10s after the end of injection, Summed images at the acquired resolution without zero-filling ze shown in Supplementary Fig, 2C to demonstrate the tue resolution. "5c image reconstruction, postprocessing, and analysis were per- formed in Matlab (The Mathworks, Natick, MA). Data and post processing code are available upon request to the authors. 25. Quantitative post processing Imaging data were reconstructed by explicitly caleulatng the IDEAL Fourier matrix, prior to inversion, By maps were summed over each "C Imaging slab. An example Bp map fs shown in Supplementary Fig. 28. By correction was applied during inversion, using an additional frequency ‘demodulation component (Moriguchi et al, 2003). ‘The slice spatial offset between metabolites was defined by equation a a 3 @ Where az is the spatial shift), af the frequency diference between metabolite (Fz), 7 the gyromagnetie ratio of '°¢ (MHzT™~, and Gy the strength of the sliceselect gradient (mT). This offset was used to deter mine the separate range of thin axal imaging slices contributing signal to ‘each metabolite individually Imaging and spectroscopic data were summed in the complex and magnitude domains respectively, and ratio maps of lactate-to-pyruvate, bicarbonate-to;pyruvate, and bicarbonatecorlactate were calculated Narain 1892019) 171.179 Total pyruvate, lactate, and bicarbonate maps were generated by: normalizing all the voxels to the peak pyruvate signal inthe brain. The rate constant, was calculated using a two site exchange mode using a frequency-domain approach and linear least-squares fiting, with any back conversion (ky) and spin lattice relaxation effects combined as an effective relaxation term, T; «(Kegan etal, 2014). The rate constant, em, was also evaluated using a two-site model in the frequency domain, representing the metabolism of [1-1"C}pyruvate to [!"Clearbon dioxide eatalyzed by pyruvate dehydrogenase (PDH, followed! by exchange with [clbiearbonate, catalyzed by the enzyme carbonic anhydrase (Géine2 Damian etal, 2014). on Where Mg(t) is the time dependent bicarbonate signal, y is the effective relaxivity of the bicarbonate signal (the inverse of Ts) and kre 's the metabolic conversion rate of pyruvate to bicarbonate ants ey —PaMa)+ beat) 26, Image analysis ‘Segmented white, gray and whole brain matter masks were produced from the 3D T weighted acquisition using statistical parametse mapping (SPM112, Welleome Trust Centre for Neuroimaging, UCL, London). two: Stage approach was used to account forthe chemical shift displacement between different metabolites. Fisly, gray and white matter probability ‘maps were calculated by summing over different ranges of thin axial imaging slices to match the thickness ofthe ¥C imaging sies, offset for cach metabolite by its chemical shift displacement (ig. 1A-B). Secondly, binary maps were produced from these images which contained >60% gray mater, white matter or brain forall three metabolites (Chard eta, 2002). 27. Region ofiuerest analysis [A number of regions of interest (ROIS) within the brain were eval: ated to determine if there was spatial heterogencity i issue metabolism: basal ganglia, dep white matter, corpus callosum, cortical gray matter, and the brain stem (Fig. 2) to assess for spatial metabolic heterogencty Several ofthese regions contained a combination of both gray and white matter. Analysis was performed on the averaged voxels from all the volunteers Tissue probability maps for each "C slice were first calculated by summing over a range of thin axial slices for each metabolite determined by its chemical shift displacement from the transmitted frequency, as ustrated for white matter in the superior slice of one volunteer. A: lactate; B: pyruvate; C: bicarbonate. D: The final binary masks were calculated where the average probability for gray matter or white matter \was >60%6 forall three metabolites; see text for deta F:!°C-pyruvate, NC iactate and “bicarbonate distribution derived from these se mentation maps shosting signal in white (unfilled) and gray (fled) matter, signals are normalized to the peak pyruvate signal in the whole brain, °p < 0.0. ‘8, B, and C (left to right: Example ROIs containing deep white matter, basal ganglia, cortical gray matter, corpus callosum, and the brain stem. ‘Whole brain, gray matter and white matter analyses were performed with segmented tissue masks for total pyruvate, total lactate, total bi cathonate, Kyu. hrm, lactateo-pyruvate, biearbonateto-lactate, and bicarbonate-to-pyrivate ratios by averaging all voxels acquired from all volunteers using the segmented masks Intr-slice gray and white matter analyses were performed by averaging voxels from all volunteers on a slice by slice basis. Further comparisons between metabolic parameters were made between tissue regions of interes, a well as analysis with and without Bp correctionNormalized signal (a.u.) Pyruvate Lactate oe Bicarbonate The assbaton of hyperpolarized signs fom the three metabolites within gray’ and white miter. Fig. 2. Region of interest analysts ond» eld map of the heathy bran 28, Staistcal analysis Statistical analysis was undertaken by comparing paired values be ‘owen gray and white matter on an interslice basis using the Wilcoxon sign rank est in the Matlab Statistics and Machine Leaning Toolbox. All statistical results were corrected for multiple comparisons using a Ben- ferroni correction, Statistical significance was defined as p< 0.05 3. Results 3.1 Coll radiofrequency hamogenity The assessment of radiofrequency (RE) excitation (B;) uniformity using the polydimethylsiloxane phantom demonstrated a highly homo- geneous By Beld, The mean ratio of the nominal tothe actual flip angle Within the central slice wae 84 =3% (mean -+S.D.), The imaging and slice profile results are shown in Supplementary Fig. 28 and 2C. 3.2. Hyperpolarized imaging ‘The time taken for dissolution and QC was 35; The time between the release of the [1-!*Clpyruvate filled syringe and the start ofthe intra ‘venous injection was 11 +2s (mean +S.D.). The levels of polarisation achieved in all four subjects, 8 measured in the liquid state by the QC module, was 25-396 (mean-+S.D.), Summed spectra from the entire time course demonstrated {1-6} pyruvate signal (171 ppm) in the three axial slices acquired, which extended fom the brain vertex tothe cerebellum (ig. 3). These spectra also demonstrated both [1—!"Cillactate (183 ppm) and {!Ctbicarbonate (61 ppm) inal our volunteers and a small quantity of pyruvate hydrate (177 ppm) was observed at early time points. Fig, 4A is an example spectral time course with a time resolution of 4s, demonstrating the axrival/formation ofthe three metabolites over time in a single volun teer. Fig. 4B demonstrates the mean signal from all four volunteer, normalized to the peak 1-"Clpyruvate signa in each case. On average signal from (1"Cpyruvate,(1—"Cllatate and (!*CTbiarbonate were observed 4, 8, and 16s after the stat of imaging respectively, which commenced 105 after the start of the hyperpolarized {1C}pyruvate injection ‘A-C: The spatial location of the 3em !°C slices used inthis study are shown in green on a sagittal T; weighted image through the brain: three slices were imaged containing the cerebellum (inferior slice, A), basala Interior sce: . Central stice . ‘Superior slice: aon I ao j ie j i I| Toe Fo I| Pe A ee ‘Coane sn oom (Chemie sit bom) ‘Sromeat oun oe) Fig. 3, %C spectra acquired through the healthy brain following injection of hyperpolarized pyruvate ‘ . 7 3 tin, i ' i fe goa! lI 5 } Bos 5 z 2 + fos ue °. — Fig. 4. Dynamic "spectra from the heathy brain shwsng the time course of (-*Chpyrvate,(1-"ellatte and (CIbicarbonate ganglia (contra sce, B) and corona radiata (superior slice, C).D-F:show —_{1"¢]Pyravate (171 ppm) inflow is seen with subsequent exchange into the summed "C magnitude spectra from the total time course acqusi-_[1-¥CHlactate (183 ppm) at approximately 8 afer pyruvate arival and tion, demonstrating signal from {1-!'Clpyruvate (171 ppm), (1-!"C] formation of {''Cibicarbonate (161 ppm) beginning at approximately Factae (183 ppm) and (!"¢}biearbonate (161 pp) in all slices. 12s, B: Average signal intensities (SD) for al three metabolites from A: Dynamie spectra acquired every 4 from the central slice of a all four volunteers demonstrating the temporal dynamics; signal hasbeen volunteer, following injection of hyperpolarized (1-"CIpyruvate. normalized tothe peak [1-!'C]pyruvate signal in each eas. Lactate and SW aaW tae 5: mone ane SA pp aay Fig. 5. IDEAL spiral C imaging demonstrating metabolite dnsibation in the helthy human branbicarbonate have been shown witha fivefold increase in sign for ease of viewing. IDEAL spiral CMRI demonstrated the spatial distribution of the hyperpolarized metabolites throughout the brain (ig. 5). Pyruvate signal was observed inthe cerebrum and cerebellum and in both gray and white matter. The pyruvate and lactate signals were particularly high in the cerebral venous sinuses (@g. the superior sagittal sinus and at the confluence of the sinuses) demonstrating that LDH activity and lactate transport were sufficiently rapid to allow tissue washout during the timescale ofthe experiment See Supplementary Fig. 24). ‘A. Example summed images from the brain of a healthy volunteer (umber 1) demonstrating (1!"CIpyruvate, (1-"'ojactate and (!%} bicarbonate signal from three axial slices: superior, central and inferior. ‘The Ty-weighted images have also been shown, as have the quantitative rape of the exchange of pyruvate to lactate (ky, in =). Bs Similar im aging shown asin (A) from the central slice ofthe three other volunteers. intensity 2.3, Tisue segmentation Segmentation analysis of the whole brain is shown in Fig. 1B, Significantly higher signal from all three metabolites was observed in ‘gray matter compared to white matter: [1-""C]pyruvate, 0.47 = 0.24 vs. 0.25 0.12; [1-"'llaciate, 0.09 +-0.04 vs. 0.06 = 0.03; ['cjbicarbon ‘ate, 003-4 0.01 vs. 0.02 0.01 respectively; p< 0.05). Since the dif- ference in metabolites between gray and white matter was largest for Pyruvate, the ratios of lactate-to-pyruvate and bicarbonate-to-pyruvate tended to be higher in white matter than gray, but this only reached signifieance for lactate. There was no significant difference in kp, ky, OF bicarbonatetolactate ratio between the two tissue types (Pie. 6, Table D. ‘The mean ky, derived from the whole brain forall four subjects was (0.012 + 0.006 5+. Similar results were obtained for both segmented gray ‘and white matter (Table 1). The mean value for yg derived from the ‘whole brain for all fou subjects was 0.020.002 5 "Results from each volunteer are shown in Supplementary Table 1, The effective mean py- ruvate and lactate relaxation times (T eq) for the whole brain vas 2610s. [A: Mean latateo pyruvate ratios derived from the segmented i ‘aging data forthe thrce brain slices showing the signal from all voxels in both white (unfilled) and gray (fled) matter averaged across all vol lunteers (mean + SD), B: Mean bicarbonate-to-pyruvate ratios. C Apparent exchange rote constants madelle from the time course data. D: “Mean biearbonate-to-laetate ratios BA. Region of imerest analysis In comparison, region of interest analysis revealed signifcant difer- ‘ences in ky, between deep white matter and regions containing basal ganglia o the brainstem (0,008 0.002 vs. 0.024 + 0.08; 0,008 + 0.002 s,0.020 0.004 5", respectively; p< 0.05 in both cases). This result suggests that there are regional variations inky, across the bain The C imaging data was wsed to derive quantitative parameters fom the whole brain, as well a segmented white and gray matter. Mean SD) values for ker Kew, lactate-o pyruvate rato, biearbonate-o- pyruvate ratio and biearbonate-to-lactate ratio are shown, 3S, Serum blood results Serum analysis revealed an increase in lactate concentration between baseline and 30 min after pyruvate injection: +24 . 8% (mean 4 SD. n= 3; range 0.1-0.7 mM). However, there was no change in serum _slucose oF LDH. Volunteers experienced no change in baseline vital signs ‘and no significant side effects were experienced. Narain 1892019) 171.179 4. Discussion Glucose, lactate and pyruvate all play a role as cerebral energy sources, Astrocytieend-feet have high concentrations of ghicose trans porters and cover a large proportion ofthe capillary walls to facilitate rapid glucose transport into the brain (Maistre et al, 1999), Following release by neurons, the neurotransmitter glutamate may undergo sodium-dependent transport from the synaptic cleft into astrocytes, where it stimulates glycolysis and lactate formation (Maistre et al 11999), When this lactate is transported into the extracellular space by [MCTs,itmay be taken up by neurons and converted into pyruvate, which can then be use as an energy source. This hypotHessis supported by the Aiferential distribution of LDH isoforms between the two cell types: [LDH (comprising four LDHA subunits) has been shown tobe present in astrocytes but not neurons and is found in tissues that are more glyco: yt, favoring the formation of lactate; n contrast, neurons express LDH exclusively (comprising four LDH subunits) which is present in tissues that have a predominately oxidative metabolism and favor the produc: tion of pyruvate (Bitar et al, 1996; Laughton et a, 2000). In this way, there fs # close metabolic coupling between asttocytes and neurons {volving an interplay between ghicose, glutamate, pyruvate and lactate, with astrocytes being more glycolytic and neurons being predominately ‘oxidative and consuming lactate Bitar etal, 1996). Although the roles of glucose and lactate in the brain are well, described, cerebral pyruvate transport and metabolism in the healthy human brain is less well understood, as endogenous pyruvate concen: trations are much lower and pyruvate is largely intracellular. The MCT family transports pyruvate in alton to lactate; for example, astrocytes express MCT1 and MCT4, and neurons express MCT2 (Pellerin etal 2007). MCT2 has a particularly high affinity for pyruvate with a Ky of 0.1 mM, followed by MCT1 with @ ky of 1.0 mM, Therefore, both cel types, but particularly neurons, will rapidly wansport pyruvate at the Deak "tissue pyruvate concentrations achieved in the experiments deseribed here i.e. 0-1-1 mM (Pérez seuredo et al, 2016). The kinetics of pyruvate metabolism observed inthis tudy are a function of pyruvate delivery to the brain, MCT expression, LDH activity and tissue lactate concentration, all of which may vary between regions ofthe brain. This study has quantified the metabolism of hyperpolarized pyruvate {nthe healthy human brain forthe fst ime, We have shown that [1-!"C) pyruvate is rapidly transported across the blood: brain barrier to form [U-jlactate within the lifetime of the hyperpolarized signal. The peak [1-"C}pyravateand (1-!"chlactatesignals were measured at 12 and 165 respectively, following the start of imaging. The presence of (1-"C] lactate in the cerebral venous sinuses shows that there is also rapid Washout of the labelled lactate, The presence of [!cTbicarbonate throughout the brain demonstrates that PDH activity i sufficient in the normal human brain to enable mitochondrial funetion to be probed in addition to cytosolic LDH activity the peak signal from [*CTbicarbonate was measured at 26s following the end of injection. These results ‘demonstrate the possibility of applying this technology not only to dis- ceases where lactate is elevated, but aso as a biomarker of early mito chondrial damage, which i a feature of inflammation. ‘The signals acquired from [1-!’C]pyruvate, [1-!"C)latate and (°C), bicarbonate were higher in gray matter compared to white matter. Perfusion differences between gray and white matter may partially a- ‘count forthe higher gray matter signal: gray matter perfusion has been showin tobe 1.4-4.0thnes higher than in white matter (1 ct. 2014) Given the relatively Iow temporal resolution of the metabolic imaging used here, temporal differences in the pyruvate and lactate timecourses could not be detected between gray and white matter, The lactate to Pyruvate ratio showed the only significant difference between tissue types, which may be driven by the higher perfusion of gray matter. urthermore, although not significant inthis N = 4 popalation, there was an increase in the bicarbonate-o-actate ratio between the cortical ray) matter and deep white matter regions of interest, potentially showing dilferencesin metabolism, However, partial volume effects may alo playNarain 1892019) 171.179 : 8 of go i I 3 ri t [ le 5 0.15] Boz] | Eos i i on 0s 3] i. Interior Middle | ‘Superior . Inferior Middle ‘Superior c > 03 os 2 om ge % [ies au oo | FL g : Inferior ‘Middle | ‘Superior ° Inferior Middle ‘Superior Fig. 6. Quanitative analysis of metabo in white and gray mater ‘Table 1 ‘Quanttarve metabote parameters derived fom regons within the rain without By coretion, Segmented regions were automatically derived and replons of interest O15) were manually drawn, ine ne) ucate prmateraio Barbone prune rato Rear cate ato Segoened whol bainmmk ——OOIZ=OGN —_Gomzaom ——_o2s=007 ouroor omrois Sepmened white mate do12=0007 © one2s0002 025-008, an 005, xz tot Sete ay mater do=000 © o0e220002 022000, ber soas baz so8 esl gagi RO boat soo © nomason1 aes oa oe roe fan 03 Deep white ate ROL 0080002 o0e2s0001 022-007 ns--002 an+02 ‘role as the thickness ofthe gray matter may be as small a8 3mm in places and the results here epresent a sallsamplesize (sch and Dale, 2000) n omarion region of interest analysis demonstrated a signi aifferenee inky betwen areas of dep white matter and the bl {ganglia and brain stom, regions known to have high glyolyie setvty (Bert et al, 2014). This may be explained by regional variations in ce- rebral LDH expression or cellular density or cel type (Laut ota, 2000) Histochemical an ins hybridisation methods have shown high levels of LDH expression in the hippocampus, pons, thalamus and neocortex of several species. Further larger stds, asesting repeat bility and reproducibility of these findings and spatial heterogeneity ‘scr the brain, wll be important ia understanding dhese preliminary findings, as well as the development of methods 0 automatically segment areas of high exchange by exploiting the multidimensional na ture of dynamic hyperpolarized data and fll integrating spatial, tem poral and spect information (Daniels al Gallagher, 2017) ‘The mean whole brain ky, measured here was 0.012 0.006 +, ‘which igrater than vase af 0.00264" measured in the anaesthetised macaque bra using a pyruvate dose which was approximately four times greater per uit body weight (Fark cal, 2014). previous study assessing the anesthetised porcine brin showed lactate formation only fater the transient opening ofthe blood-brain barser, when & ky, of (0.012 0.007 +" was measured which fs similar othe values reported here (ler el, 2018) Higher ky, values have Been measured int mors, reflecting elevated LDH and lactate concentrations i these sss. For example vale of 0.025 was reported in humat prostate cancer before treatment and 0.007 s-? following androgen ablation therapy, which i similar to the value reported here for normal tissue (Aggarwal ol, 2017), However, the value of ky, measured herein normal bral is lower than that found ina previous human brain tumour study, where a mean whole brain ky, of 0.12 s-" was detected; this included tumor as well as normal-appearing brain and therefore kr. may be dominated by the elevated tumor metabolism (Miloushev etal, 2018). ‘The mean kry derived from the whole brain was 0.002 0.002 5", Which represents flux through the reaction catalyzed by PDH and sub. Sequent exchange of the '°¢ label between carbon dioxide and bicar- Donate The latter reaction, which is catalyzed by carbonic anhydrase is rapid and assumed tobe at equilibrium (Gallagher etal, 2015). Although we were able to demonstrate higher bicarbonate signal In gray matter compared to white matter, there were no significant tissue differences found in kys or in the ratios of bicarbonate-to-pyruvate or lactate to-pyruvate. The SNR of bicarbonate was limiting factor in this analysis acquisition snd methods to increase this are important for future studios: this could be achieved by incorporating spectrl-spatil pales to selectively inerease the excitation flip angle for bicarbonate (Larson tal, 2008; Schulte etal, 2013). Am inerease in SNR will also enable acqui sidon of higher spatial resolution images of each metabolite. Future larger studies are required to assess variation between individuals which ray be affeted by faetors such as age, gender and weight ‘An important element of this study was the uniform RF excitation profile, which has allowed a comparative quantitative analysis to be undertaken across the brain, A uniform RF excitation profile is a‘consideration for selective excitation sequences as the uncertainty inthe delivered flip angles may significantly affect the results derived from [Kinetic modelling (Sux ct a, 2018), The relative lack of contrast reported Ihere between kinetic parameters in normal-appearing gray and white ‘matter was unexpected, but could assis the identification of significantly faltered metabolism above or below background in pathological ‘conditions. In conclusion, this study has demonstrated that "'C-MIRI can be used {0 acquire quantitative non-invasive measurements of hyperpolarized {1Pelpyruvate metabolism in the normal human brain and could be ‘used to measure regional variations in metabolism aross the brain, This ‘work provides evidence that the methodology may have a role in assessing disease processes where lactate Is elevated and where mito- ‘chondrial funetion may be altered, ‘Acknowledgements ‘This study was funded by the Welleome Trust, Cancer Research UK (CRUK, €19212/A16628, €19212/A911376, C197/A16465), the CRUK Cambridge Centre, National institute of Health Research-Cambridge ‘Biomedical Research Centre, Medical Research Counell, CRUK/ENgi- neering and Physical Sciences Research Council Imaging Centre in ‘Cambridge and Manchester, Addenbrooke's Charitable Trust, the Cam- bridge Experimental Cancer Medicine Centre, the Evelyn ‘Trust, the ‘Multiple Sclerosis Society, the National Insttite for Health Research {Cambridge Biomedical Research Centre atthe Cambridge University ‘Hospitals NHS Foundation Trust and the Mark Foundation Institute for Integrative Cancer Medicine atthe University of Cambridge. The views ‘expressed are those ofthe thors and not necessarily those ofthe NHS, ‘the NIFIR or the Department of Health and Social Care. ‘Thanks to Sarah Hllbome, Jackie Mason, Vicky Fernandez, Hannah Loveday, Ashley Grimmer, Emma Ward, Brian White, Amy Fray, Ronnie Hernandez, Matthew Locke, Claire Trumper, Dario Prudencio, and Bruno do Carmo, Appendix A. Supplementary data Supplementary data to this article can be found online at https//doi ‘org/10.1016/).neuroimage.2019.01.027, References Agarwal Viger, DB, Kerhnee,2017. pepo 1-Crpyeate ‘pc emcees eat meta rape sn eden tarsen 4, each, AM, Cathe, N Un, J, Anerson, Ss, 201, Dynamic car polation ple forsee met. NM oed. 24, ert sco, Pp, 2014 rai somal variates ad beng Bins a FOG PEI/CT taping. Po Ci 4, 129-140. tay, P Gay. Palen, ours, Mage P1996, Seecte sbaron of cae deyaogenae enzyesn nero J. Cer, lod How ‘Meabol 16 10791089. (Chr I. cia, CM. ete, MA, Kaper, P Kaper, Thompson, A. ‘Mil, 202, Bain metal anges acne oy and oral speaing hie matin lineal easly rpg reiting mule eles rain 125, ‘cunningham, CH, tau, 39°, Chen, AP. Gera, Pets, W. Rolin ih GA. may Ate peat "Cebit Behan Danny Galogher EA, 2017. Unrpendzed segmentation of ‘arbood Ml dang ay Mark anon eld mode IS rane ed Imag, 37, 340-650, Dancy C1, McLem MA, She, RE Robb J, Gill AB, Meghan, (eaves, ML, Schwager, Loma, Dl rnd, Calagbes FA, 2006. ‘coat gms ei cio wh pi cl ale, AM, 200, Meworing he hes ofthe bu ere ote ems ingot omc mages Po Nall Acad Se Ue Sates A 97, 11050-1085. alge: PA. Ketan i Day SH, Dl, Adenoerdase, JH, Zana 8 ‘ernst Rae #53 940-3 ‘Narain 1892009) 171.179 Gates, FA Sado. etme, ML, Seo, EM. Rodeges TL, Weight A (Gl ABs Mee, Dot, TC, Noe, J, Mere, A Mle, SY, Hoes. Dy. Si i, Hawa WI. Hare AL bale E2015, Crone anys sett montred nev by Hypealed c mage ‘moon spctoopy dst hs importance frp elton In us {incr Res 75,4109 4118 Geer Dame, PA Sper J Snel MLA, Koga 0. Wieseger FG, SL, ase, A Staiger, M.S el tL 2014 Mase i ‘node of Comal ME sing (Cheat Rodi Res. Prac 67619, Gordon, Viento, D8, Laon, PZ, 2016. Devapment of 2 sym echo fie tig eh nc nin oft a pie urd RE, en VPs Chen, A, Aneajber Lanse, J, 2012 typerpaaied 2 nw, 36, 13141528, Jal, by Carpenter, KLE, Hela, A, Carpenter, TA, Menon, Di, Hatin, 7.3, cement and pthopyslogia dingy Nett rain De 30, 615-632 Joe, 5 Hr linge, Senders Lack, LAL Yen VE, ederim, ShicmanD, ayer, 2013 Elects a ofan pen on hyperpolarized ‘tal measurement tras Mage en Mel 70, 7-112, eg 0 Seat och ALA Merz ML Farell Et, AM, Aedes aren Jy ise, Has, Sehwnige Bi Whig, F204 Aparentate onsant main sing hyperpolarized (1 pyre Home. 27, Lami 1, Ps Ker, A, Che, AP Li, MS, Zea, MLL (Cnaingia, Ci ay, kane. Vion, Di, 208 Sep atin exatin puso bpp 1 dynamic hemi st meg. Laughton 0, Chama, Bll Pell, Mags PJ, Boras, 2000 ‘iret nonenger RNA dtm of ite dey vgec LIN sed DIS ‘of in the tan Neate 95,619 25 1x Sitar SIN Purdy, Dis Brig EW 201 Qt cerebbum rey mater ‘nd white mater petfision wing pled atria pn abelng Med es. nt Magistet, P,P Rtas, DL, Salman, RG, 199. rg on dean Scenes 283, 495497 Matha, Hodas GL, Coot, Mar MU, 2014 Rete acse metabo: igh int multe iron pathogen rent. Nera 1-2. Dati Reet, K. Tasch C, Nowa, M, Dochenhausen, A Bana Me, ooo, J, 2008. Dordt coe mabolsn In act stoke pens an deacon poe, Dats Care 29, 792-797. ue Sere Lae J, Ly AZ, Ray, Par, KR, nse, “ong RS em, PA, Lndar J Lanse, Gl, BA yer Da ‘Si, 308 pyrace ranpr aco the nd rain aver prin Ippo MIL 8 Rep 8,182 uote, V. Gali, Ki, Boyan R Lyakeenko, SK Deane LO ‘Ming LK, Brean, CM, Tabu, V, Yang, Hada, Ale Soc, RE, bran with hyperpolarized FC pyrnate demonstrate lactate production is Bean tomer patents Cane es. 78, 5755-5760, Moriah Hew, 3, Doe, 2003 Dion eign pl ejecta wt ‘sone erection pw spec fr seal peo tot pa ‘pea pes Mag, Resm M5 915-924 Nekon 3, Kaanewi 3 Vigneon I anon, PZ, Hanes A, Femone, M “an Gekinge, MCh, Bk, Park, Ree Carve 5 aster Ps Winer. Vi. Adeje Lanen, JH, Che, AP RE ‘Opi Maroy 13 Metalic jg pts ith pi caer ‘sing ypeoliad 1-clpravre et Tans. Med 5, 198108 ark amon, P82, Top, J avai Ree, Bok Re Rob, Fans 2 Kaltes, P Bbw, igo, Dik Neb, 8 2018 Dam Tpetpelarandearboo13 MR mbes aging of noaunan pts ral, Meg Pak Lan PAZ, Case, Cal, Che, HY ik Van Ching. M, esto Sir, J, Ru, Ruanewic, J, gn, DS Chang, 5. ‘Nebon, 82, 2018 Deopment of mods and eat of wing hyppolatnd fab" aig sta fr evaluating bla mbes in ent sta. Mah Reson, Med 0 84-8 Pale, Anse Kvn Boz or, Abert Agnes Sere, Sebati, Michel Mare ail Robes, Masts Pl 07 Any dependent relat fey tnetaols ty acy: a piste, Gi 55, 1416-1425. er creo, a He, VF, Sour Mt Fakes, J, Pen, VL, Pll L, Sons P2016. Mosocarbonynte Casper in he bein odin emt Bich Bop Aa Mo. al ex 186, 2461 2497 PicharJ,Haiman,D, Nowy, Ey Pol. Kawabor,T Ans, A, owsean, A Hasoc, C, Shulnan R, 1951 taae ise detect y "HNMR in Im val ten dring plo stmt Pre Nal Ae, Se Unit Sates Seni RE Shh Dre, Ah UA aig ALM JM ‘Seger sit NMR homed 4, 106-1072. Sehutes UF Sper 1, Wei Merz MLL, Jie Ma Keg 0. Dut A ‘Are aon, Case, Has Seba, M, Wise F203. ‘Staten recovery metab cachage re mang wth hyperotred [130 ravte wing spec spatial exc. Magn Reson. Med 69, 1209-1216Sum, CY, Water, CML, Mlcbel KA. Venton, AML SY Honkson JA, 2018, Inder of parse sec) en parma Ss of ype pyre Mage son. ed. 79,3299 48, “ome Mt Ala, N Sasa, J. Tabb, YE, M. Nozawa K hima, “Tl, Oba, T 26 Noel locate concen Yang ae ooo ai ‘Mog Ren ang 34,1269 1273. Naren 1892019) 171.179 Wisige F Weld E, Menzel ML, Jani MA, Kha. Ge, Sse, A, Scge B.S, LP, 2012 IDEAL sp CS or dja Mlk ‘ang of typerptred (1"Clpyrnata Mag. Ron, hed 5,816. cag FGI. Dee, $8, Wah Re Lacan, My Meteo, MA Bas, ‘alt FA, 2018 Hyper eaibow 13 magne teomne spaces Iman allt stdyng tour metabo, BRL 1,68,