Increased hyperpolarized [1‐13C] lactate production in a model of joint inflammation is not accompanied by tissue acidosis as assessed using hyperpolarized 13C‐labelled bicarbonate

Arthritic conditions are a major source of chronic pain. Furthering our understanding of disease mechanisms creates the opportunity to develop more targeted therapeutics. In rheumatoid arthritis (RA), measurements of pH in human synovial fluid suggest that acidosis occurs, but that this is highly variable between individuals. Here we sought to determine if tissue acidosis occurs in a widely used rodent arthritis model: complete Freund's adjuvant (CFA)‐induced inflammation. CFA robustly evoked paw and ankle swelling, concomitant with worsening clinical scores over time. We used magnetic resonance spectroscopic imaging of hyperpolarized [1‐13C]pyruvate metabolism to demonstrate that CFA induces an increase in the lactate‐to‐pyruvate ratio. This increase is indicative of enhanced glycolysis and an increased lactate concentration, as has been observed in the synovial fluid from RA patients, and which was correlated with acidosis. We also measured the 13CO2/H13CO3 − ratio, in animals injected with hyperpolarized H13CO3 −, to estimate extracellular tissue pH and showed that despite the apparent increase in glycolytic activity in CFA‐induced inflammation there was no accompanying decrease in extracellular pH. The pH was 7.23 ± 0.06 in control paws and 7.32 ± 0.09 in inflamed paws. These results could explain why mice lacking acid‐sensing ion channel subunits 1, 2 and 3 do not display any changes in mechanical or thermal hyperalgesia in CFA‐induced inflammation.

Acid evokes pain in humans, [15][16][17] and thus inhibiting acid-induced activation of sensory neurones is an appealing strategy for treating arthritic pain.
Sensory neurones express several acid sensors: acid-sensing ion channels (ASICs), transient receptor potential vanilloid 1 (TRPV1), proton-sensing GPCRs and certain background K + channels. 18 Furthermore, response to low pH has been shown to be sensitized by inflammatory mediators. [19][20][21][22][23][24] We and others have demonstrated expression and function of TRPV1 and ASICs within articular sensory neurones, [25][26][27] and both TRPV1 and ASIC3 are upregulated in different inflammatory models, suggesting their involvement in inflammatory pain. 26,28,29 A common model for inflammatory arthritis involves intraplantar/intraarticular injection of complete Freund's adjuvant (CFA), 30 which although not replicating autoimmunity does provide a robust model of RA-like arthritis: T cell-mediated pathogenesis, leukocyte invasion, synoviocyte hyperplasia, pannus formation and disrupted gait. [30][31][32] However, it is unknown if tissue acidosis occurs in this commonly used model. In CFA studies investigating animal pain behaviour, TRPV1 knockout mice show diminished thermal hyperalgesia, but no change in mechanical hyperalgesia, 33 and animals lacking either ASIC1, ASIC2 or ASIC3 show no alleviation of thermal/mechanical hyperalgesia. 34 Hyperpolarization of 13 C nuclei can increase their sensitivity to detection in a magnetic resonance experiment by more than 10 000-fold. 35 This enormous increase in sensitivity has enabled real-time imaging of tissue metabolism in vivo following intravenous injection of hyperpolarized 13 C-labelled substrates, 36 including in humans. 37,38 Previous magnetic resonance spectroscopic imaging (MRSI) studies in rats injected with hyperpolarized [1-13 C]pyruvate showed raised lactate-to-pyruvate ratios in CFA-induced inflammation, 39 which, considering the relationship observed between lactic acid and acidosis, 9,12,13 suggested that acidosis must occur in this widely used model. The aim of this study was to determine if tissue acidosis actually does occur in regions where there was increased lactate labelling by using hyperpolarized [ 13 C]bicarbonate to measure tissue extracellular pH. 40 welfare ethical review bodies also approved procedures. Female C57BL/6 mice (aged 10-12 weeks and weighing 18-20 g, Envigo, Huntingdon, Cambridgeshire, UK) were housed in groups of up to four mice per cage with nesting material and a cardboard tube; the holding room was temperature controlled (21°C) and mice were on a standard 12 h light/dark cycle with food and water available ad libitum.
All chemicals were purchased from Sigma-Aldrich (Gillingham, Dorset, UK), unless stated otherwise.

| CFA-induced inflammation
Mice were anaesthetized using isofluorane (2%), and two 15 μl injections of 10 mg/ml CFA (Chondrex, Redmond, WA, USA) were made using a Hamilton syringe and 27 G needle, to give a total dose of 300 μg per paw. Control injections with phosphate-buffered saline (PBS) were made in the contralateral hind paw.

| Assessment of paw swelling and clinical scores
Mice were weighed and calliper measurements of ankle and foot pad diameters were performed daily. Clinical scores were also made daily according to Reference 42, with a score of 0 for a normal paw, 1 for a slight swelling and/or erythema, 2 for a pronounced swelling and 3 for ankylosis of the paw and ankle. A two-way ANOVA was used to compare changes in calliper measurements and clinical scores between CFAand PBS-injected paws over time (n = 5 mice from days 1 to 5). Sidak's multiple comparison test was used to compare each time point.

| MRI
Images and spectra were acquired using a 7 T MR instrument (Agilent, Palo Alto, CA, USA). Proton images were acquired axially through downwards pointing feet using a volume coil. Images were acquired using a fast spin echo (FSE) sequence (40 × 40 mm 2 slices, 2 and 6-10 mm thick, covering the entire foot, 128 × 128 data points, eight echoes, effective echo time (T E ) of 48 ms, 2 s repetition time (T R )) or using a spoiled gradient echo (GE) sequence (40 × 40 mm 2 , 2 mm thick slices; 128 × 128 data points; T R , 400 ms; T E 2.85 ms). Proton imaging was used to confirm positioning and to quantify foot volumes.
2.5 | Magnetic resonance spectroscopic imaging of hyperpolarized [1-13 C]pyruvate metabolism 13 C images were acquired from both non-inflamed hind paws of two mice following injection of hyperpolarized [1-13 C]pyruvate 1 day prior to injection with CFA and PBS. CFA and PBS were then injected, as described above, and inflammation allowed to develop for five to seven days before the 13 C imaging was repeated. Each mouse was placed inside a 1 H/ 13 C volume coil (bird cage, internal diameter 42 mm) with both feet arranged vertically inside a 20 mm circular 13 C receive coil (RAPID Biomedical, Rimpar, Germany). There was no significant difference in body weight between mice with CFA-inflamed and non-inflamed paws at the time of imaging (mean 18 g). A reference FSE proton image was acquired as a single axial slice (6-10 mm thick) that encompassed both feet including the heel, up to where the top of the talus meets the leg.
[1-13 C]pyruvate (99% 13 C labelled, 44 mg) in a solution containing 15 mmol/l of trityl radical, tris(8-carboxy-2,2,6,6-tetra(hydroxyethyl)-benzo-(1-5)-bis-(1,3)-dithiole-4-yl)-methyl sodium salt (OX063; GE Healthcare, Amersham, UK) and 1.5 mmol/l gadolinium chelate (Dotarem, Guerbet, Paris, France) was polarized as described previously 43  (173 ppm) and [1-13 C]lactate (185 ppm) were used to calculate metabolite ratios. A quality control step was also applied for data display, where the acceptance threshold on the pyruvate signal amplitude was equivalent to a signal-to-noise ratio greater than 11 (intensity of a 50 Hz linewidth resonance divided by root-mean-squared noise). Voxels were selected for quantitative comparison between inflamed and non-inflamed paws by drawing a region of interest around individual feet in the reference FSE image. Voxels whose centre co-localized to this region of interest were included. A one-way t-test was performed on all voxels in a CFA-inflamed paw, comparing the mean lactate-to-pyruvate ratio to that of both paws from a mouse with non-inflamed paws. No correction was made for multiple comparisons.

| MRS of hyperpolarized 13 C-labelled carbon dioxide and bicarbonate
Spectra were acquired from one non-inflamed hind paw of four animals, 1 day prior to CFA injection. Spectra were similarly acquired, from four CFA-inflamed paws in four animals five to seven days after CFA injection. Animals were culled following completion of the MR experiments. Mice were placed in the MR instrument with one foot (CFA-inflamed or non-inflamed, n = 4 feet for each group) pointing down through a 9 mm diameter custom-built solenoid coil that covered the foot from heel to toes. Foot volumes were determined using the open-source Fiji software package. 45 The tissue was outlined in 2 mm thick proton image slices and the resulting areas for each slice were summed. The volume was calculated for the whole foot from the toes to the heel, where the top of the talus joins the leg. There was no significant difference in the body weights of mice with CFA-inflamed and non-inflamed paws at the time of imaging (mean 19 g). Carbon-13 labelled caesium bicarbonate was prepared and polarized as described previously 40,46 : 0.7 mmol of CsH 13 CO 3 was dissolved in 0.54 mmol of glycerol (Sigma-Aldrich) and 63 μl of water with 15 mmol/l OX063 and 1 mmol/l Dotarem. Hyperpolarized samples were dissolved in 6 ml superheated buffer containing 80 mM phosphate at pH 7.5 and 100 mg/l EDTA, and then rapidly ion exchanged with 3 g of Chelex 100 resin in the sodium form (Bio-Rad Laboratories, Watford, UK) before injection via a tail vein. Pulse and acquire, coil-localized, spectra (6000 Hz bandwidth, nominal flip angle of 10°, 0.45 ms echo time, 1024 data points) were acquired every second from 12 s after injection. The first 28 spectra-from 12 to 39 s-were summed, phase corrected and a quadratic baseline correction applied. The 13 CO 2 /H 13 CO 3 ratio was calculated by integrating the signal intensities between 127 and 123 ppm and between 163 and 156 ppm respectively and converted to a pH value by assuming a pK a of 6.17. 40 3 | RESULTS

| Significant hind-paw inflammation is observable from 24 h after injection of CFA
Inflammation of the hind paw induced by CFA injection can be followed by measuring paw swelling, as described by Chillingworth and Donaldson, 30 and by observation of how swelling and joint damage produce anatomical changes in the foot, which is reflected in the clinical score. 42 Following injection of CFA, inflammation of the ankle and footpad was observed within 24 h and was significantly greater than in PBSinjected paws at all time points ( Figure 1A,B, p < 0.0001, n = 5 mice). The clinical inflammation score also increased gradually from 24 h onwards, being significantly greater than in PBS-injected paws at all time points ( Figure 1C, p < 0.0001, n = 5 mice). These results are consistent with published data for this model 30 and demonstrate that when the pH was measured there was significant inflammation in CFA-injected mouse paws.
These data were consistent with foot volumes estimated from MRI data. The CFA-inflamed paws in which the pH was measured were significantly larger than non-inflamed paws (non-inflamed 0.180 ± 0.009 cm 3 versus CFA inflamed 0.253 ± 0.021, p ≤ 0.01, Figure 1D, n = 4 feet). The mice used for the [1-13 C]pyruvate measurements showed similar significant swelling in CFA-injected paws (data not shown).

| Labelled lactate is elevated in the CFA-inflamed paws of animals injected with hyperpolarized [1-13 C]pyruvate
Seven days post-CFA administration, 13 C MRS images were acquired 20 s after i.v. injection of hyperpolarized [1-13 C]pyruvate. The conversion to [1-13 C]lactate was higher in CFA-inflamed paws than in non-inflamed paws. Although the images showed that there was a range of lactate-to-pyruvate ratios across the inflamed paw (Figure 2A-D), these were higher than in non-inflamed paws ( Figure 2E-H) and in contralateral PBS-injected paws ( Figure 2C,D). Analysis of all voxels within CFA-inflamed or non-inflamed paws showed that there was a significant elevation of the lactate-to-pyruvate ratio in the CFA-inflamed paws ( Figure 2I). The mean lactate-to-pyruvate ratios were 0.30 ± 0.23 for non-inflamed paws and 0.74 ± 0.32 for CFA-inflamed paws (p < 0.001, for each individual comparison: see Figure 2I for details). This is consistent with increased levels of lactate 39,47,48 and glycolysis in the inflamed tissue, where resident fibroblast-like synoviocytes, which are key contributors to synovial inflammation, show a shift to glycolytic metabolism in RA. 49

| Measurements of the hyperpolarized [ 13 C]carbon dioxide/[ 13 C]bicarbonate ratio showed no significant difference in extracellular pH between CFA-inflamed and non-inflamed paws
To determine whether raised labelled lactate concentrations in the CFA model were accompanied by tissue acidosis we measured the 13 CO 2 / H 13 CO 3 ratio following i.v. injection of hyperpolarized [ 13 C]bicarbonate. Serial spectra acquired from an individual animal showed that there was no change in the 13 CO 2 /H 13 CO 3 − ratio from 12 s after bicarbonate injection, demonstrating that the carbonic anhydrase-catalysed conversion of bicarbonate to carbon dioxide had reached equilibrium and therefore that the pH could be estimated from this ratio ( Figure 3). The 13 C nuclear spin polarization in hyperpolarized H 13 CO 3 and 13 CO 2 has a very short half-life in vivo of only about 10 s, 40 and by 39 s after bicarbonate injection the CO 2 signal had decayed and was no longer detectable. Therefore, the interval between 12 and 39 s after bicarbonate injection was chosen for subsequent analysis, where spectra acquired during this period were summed in order to calculate a 13 CO 2 /H 13 CO 3 − ratio and a pH using the Henderson-Hasselbach equation. Gradient echo 1 H images show the increased size of a paw injected with CFA (compare Figure 4A with Figure 4F). Summed 13 C spectra from CFA-inflamed and non-inflamed paws are shown in Figure 4B-E and G-J respectively, and the extracellular pH values calculated from the 13 CO 2 /H 13 CO 3 ratios in these spectra are shown in Figure 5. The extracellular pH values in CFA-inflamed and non-inflamed paws were not significantly different (pH 7.32 ± 0.09 versus pH 7.23 ± 0.06 respectively, n = 4, p = 0.92).

| DISCUSSION
Although there have been several reports of tissue acidosis occurring in RA, [8][9][10][11] there is large individual variability. 9,12 In mouse models of arthritis it is often assumed that tissue acidosis occurs, but no data exists to substantiate this belief. We have shown here that intraplantar injection of CFA in the mouse paw induces inflammation and that this is accompanied by a significant increase in 13 C lactate labelling in animals injected with hyperpolarized [1-13 C]pyruvate. However, despite the apparent increase in glycolytic activity and lactate production we were unable to find any evidence of extracellular acidification of this tissue, as assessed from measurements of the 13 CO 2 /H 13 CO 3 ratio in animals injected with hyperpolarized 13 C-labelled bicarbonate.
Measurements of lactate labelling in animals injected with hyperpolarized [1-13 C]pyruvate have been used previously to assess glycolytic activity in inflamed joints. 39 MacKenzie et al. 39  invasion. While this measurement does not measure directly lactate concentration, it shows that the exchange of 13 C label between pyruvate and lactate is faster. The exchange rate is, in part, dependent on the size of the regional lactate pool 47,48 and therefore is indicative of increased glycolytic activity in inflamed tissue. MacKenzie et al. 39 showed marked infiltration of leukocytes in the majority of rats with CFA-induced hind paw inflammation, with accumulation of inflammatory cells at sites of injection. Leukocytes become increasingly glycolytic on activation as do stromal cells, such as fibroblast-like synoviocytes in RA. 49 Increased glycolysis is consistent with previous studies that have shown this more directly, for example, synovial fluid taken from patients with RA has shown elevated lactate levels. 9,12,13 The signal-to-noise ratios obtained following injection of hyperpolarized 13 C-labelled bicarbonate were not sufficient for imaging, and therefore for these experiments we acquired spectra from the whole mouse paw. Signal localization was obtained by placing a custom-made receiver coil around the paw, the improved coil-filling factor improving the signal-to-noise ratio. The measured pH corresponded therefore to that of the dominant tissue observed in MR images of this volume (see Figure 4), which was mainly muscle, tendon, ligament and connective tissue in the paw. The very short half-life of the nuclear spin polarization in hyperpolarized H 13 CO 3 and 13 CO 2 means that the ratio reflects predominantly the extracellular pH 46  The 13 C-labelled bicarbonate/carbon dioxide signal ratios obtained from individual spectra acquired from one inflamed paw every second, from 12 s after 13 C-bicarbonate injection Measurement of 13 C spectra after injection of 13 C-labelled bicarbonate in control and inflamed mouse paws. A, A gradient-echo 1 H MR image from a 2 mm thick slice through an inflamed mouse paw. B-E, Summed 13 C spectra acquired from four inflamed mouse paws between 12 and 39 s after injection of 13 C-labelled bicarbonate. The spectra show a larger 13 C bicarbonate resonance and smaller carbon dioxide resonance. F, A gradient-echo 1 H MR image from a 2 mm thick slice through a normal mouse paw. G-J, Summed 13 C spectra acquired from four non-inflamed mouse paws between 12 and 39 s after injection of 13 C-labelled bicarbonate observe a lowered pH could have been due to a lack of equilibration of hyperpolarized 13 C label between H 13 CO 3 and 13 CO 2 , which is catalysed predominantly by carbonic anhydrase. 46 However, we demonstrated that 12 s was sufficient to achieve label equilibration (Figure 3), which is less than the approximately 16 s required in a murine lymphoma in vivo, 49 but comparable to the time required for label equilibration in rat heart muscle, 51 which included the time for formation of carbon dioxide and bicarbonate from [1-13 C]pyruvate catabolism.
The hyperpolarized 13 C-labelled lactate detected following injection of hyperpolarized [1-13 C]pyruvate is predominantly intracellular, at least in animal tumour models, 52 whereas the pH determined using hyperpolarized 13 C-labelled bicarbonate is predominantly extracellular. 40 Therefore, our failure to detect a decrease in pH, despite an increase in glycolytic activity, may be because any pH decrease is largely intracellular. However, lactic acid is rapidly exported from cells, [53][54][55] and it seems unlikely that there would not have been an increase in extracellular lactic acid concentration in the inflamed joints, in which case the resulting increase in H + concentration must not have exceeded the extracellular buffering capacity of the tissue. Whatever the explanation, our results have shown that increased glycolytic activity in the inflamed joint is not accompanied by extracellular tissue acidosis.
Although we have not assessed pain behaviour here, this was inferred from measurements of joint swelling and worsening clinical scores, which are features that have been shown previously to correlate with indicators of pain, namely mechanical and thermal hyperalgesia, in CFAinduced inflammation. 29,30,34,56 The results presented here would suggest, therefore, that tissue acidosis is not itself a primary contributor to the pain observed in the CFA-induced arthritis model, which might perhaps explain the lack of relief from either mechanical or thermal hyperalgesia in mice lacking ASIC1, ASIC2 or ASIC3. 34 Although TRPV1 knockout mice display diminished thermal hyperalgesia in the CFA model, 33 this is likely due to a shift in the thermal sensitivity of TRPV1 activation resulting from inflammatory mediators such as nerve growth factor dependent removal of phosphatidylinositol 4,5-bisphosphate inhibition of TRPV1, 57 rather than due to acid-mediated modulation of the TRPV1 thermal activation threshold, which can also occur. 58

| CONCLUSIONS
In summary, we have demonstrated that in the CFA-inflamed mouse paw model of arthritis there is elevated production of lactate, but that this is not coupled with a significant decrease in extracellular pH. This result could explain the lack of phenotype observed in mice lacking different ASIC subunits and questions the validity of CFA-induced arthritis as a model for RA, in which tissue acidosis has been demonstrated.