Non-Invasive Differentiation of M1 and M2 Activation in Macrophages Using Hyperpolarized 13C MRS of Pyruvate and DHA at 1.47 Tesla

Macrophage activation, first generalized to the M1/M2 dichotomy, is a complex and central process of the innate immune response. Simply, M1 describes the classical proinflammatory activation, leading to tissue damage, and M2 the alternative activation promoting tissue repair. Given the central role of macrophages in multiple diseases, the ability to noninvasively differentiate between M1 and M2 activation states would be highly valuable for monitoring disease progression and therapeutic responses. Since M1/M2 activation patterns are associated with differential metabolic reprogramming, we hypothesized that hyperpolarized 13C magnetic resonance spectroscopy (HP 13C MRS), an innovative metabolic imaging approach, could distinguish between macrophage activation states noninvasively. The metabolic conversions of HP [1-13C]pyruvate to HP [1-13C]lactate, and HP [1-13C]dehydroascorbic acid to HP [1-13C]ascorbic acid were monitored in live M1 and M2 activated J774a.1 macrophages noninvasively by HP 13C MRS on a 1.47 Tesla NMR system. Our results show that both metabolic conversions were significantly increased in M1 macrophages compared to M2 and nonactivated cells. Biochemical assays and high resolution 1H MRS were also performed to investigate the underlying changes in enzymatic activities and metabolite levels linked to M1/M2 activation. Altogether, our results demonstrate the potential of HP 13C MRS for monitoring macrophage activation states noninvasively.


Introduction
Nearly ubiquitous throughout the body, macrophages are a critical component in maintaining our health and wellbeing, playing a central role in the innate immune response, tissue homeostasis, and facilitation of crosstalk with neighboring cell types [1][2][3]. Macrophages have typically been described as having two activation states, which have each been observed to play significant roles in various pathologies, such as autoimmune disorders, obesity, and cancer malignancy [4][5][6]. Although there is still much to be understood, these activation states were first generalized to the M1 and M2 dichotomy [7], and are associated with cell-wide changes, including modulations of signaling pathways and reprogramming of cellular metabolism [8]. Briefly, M1 describes the classical proinflammatory activation response, leading to subsequent tissue damage, whereas M2 activation is associated with upregulation of anti-inflammatory pathways promoting tissue repair. From an energetic metabolism perspective, M1 macrophages have been shown to increase anaerobic glycolysis, while M2 use primarily aerobic oxidative phosphorylation for energy generation [9]. Reactive oxygen species (ROS) have also been reported to increase significantly in M1-activated macrophages, including in J774a.1 cells, and are often used as a marker of this activation state [10][11][12][13][14][15][16]. The activity of the arginase-1 enzyme iso-

M1 and M2 Activation of J774a.1 Macrophages
First, ROS and Arginase assays were performed to confirm differential M1 and M2 activation of J774a.1 macrophages, which was performed using previously established protocols ( [33] for M1 and [37] for M2). Arginase enzyme activity was unchanged with M1 activation by LPS (p = 0.849), while a significant 57-fold increase in arginase activity was seen in M2 macrophages activated by interleukin-13 (IL-13) (p < 0.0001, N = 3, 38.255 U/L/10 5 cells in M2-activated vs. 0.669 U/L/10 5 cells for control), as previously reported [37] (Supplementary Figure S1A). On the other hand, M1-activated macrophages exhibited a 2.3-fold increase in ROS compared to control (p = 0.0009, N = 6), whereas no significant increase in ROS was detected in M2-activated macrophages (p = 0.7380, N = 3) (Supplementary Figure S1B). The ROS increase in M1-activated macrophages is in agreement with previous reports [38]. The Bradford assay showed no differences in total protein concentration per cell between groups (Supplementary Figure S1C).

Discussion
In this study, we used the established J774a.1 mouse macrophage cell line and applied M1 [33] and M2 [37] activation protocols described in previous reports. First, we confirmed that differential activation was successfully achieved, as shown by a significant increase of arginase activity in M2 macrophages [20] and an increased level of ROS in M1 macrophages [7]. Of note, we did not see a significant decrease in ROS levels in M2-activated macrophages compared to control, as others have reported [22], likely due to the small number of cells available for this assay. Protein concentrations per cell were unchanged with M1 or M2 activation, justifying the normalization of all HP and spectrophotometric values to number of cells throughout the study. The 1 H MRS results further confirmed differential activation of macrophages. Highly significant increases in itaconate and succinate concentrations are seen in M1 macrophages compared to control, in line with previous reports [39], as itaconate is a well-known product of M1 polarization [40] and a potent inhibitor of succinate dehydrogenase, which leads to succinate accumulation [41]. Glutamine is higher in M1 compared to control, albeit not significantly, in line with a previous report utilizing the identical cell line and activation protocol [33], and also consistent with a study of LPS-induced M1 polarization utilizing a tenfold lower dosage (10 ng/mL) [42]. The increased lactate observed in M1 compared to control and M2 is also indicative of highly upregulated lactate production and Warburg-like effect that are specific to M1 activation [22]. In M2-activated macrophages, the observed decrease in arginine levels compared to M1 could be explained by increased arginine consumption through upregulated arginase activity, though, to date, the only metabolomics study of macrophage activation reports increased intracellular arginine concentration for both M1 and M2 [43]. It should be noted, however, that this study used primary human macrophages, which have been shown to respond differently than established cell lines, as well as much longer activation protocol of 72 h, which might explain the observed discrepancy [44]. A larger metabolomics study specific to activation patterns of murine macrophage cell lines would need to be pursued for more direct comparisons.
Here, we used an innovative approach allowing for noninvasive assessment of metabolism, namely HP 13 C MRS, and evaluated its potential to differentiate between activation states in live macrophages. Our results show that significant differences in HP product-to-substrate ratios of both HP [1-13 C]pyruvate and HP [1-13 C]DHA can be ob-served in differentially activated murine macrophages. The sampling of cell slurries prior to each HP 13 C MRS acquisition allowed for paired biochemical assays to be performed, thus enabling for strong biochemical validations. The methods used may be applied in other cell types, both from culture and primary populations.
The increased HP 13 C lactate-to-pyruvate ratio observed in M1-activated macrophages is in agreeance with previous reports [33], and congruent with the expected highly increased glycolytic activity in proinflammatory macrophages [45]. This increased ratio is also associated with a significant increase in LDH activity in the M1 group detected in paired subsamples. In M2-activated macrophages, the HP 13 C lactate-to-pyruvate ratio was not significantly different from control, although a trend towards significance was observed. This trend is in line with a smaller significant increase in LDH activity, which was detected by spectrophotometric assays that are likely more sensitive than HP 13 C MRS. These results are also in agreement with previous reports showing that M2 polarized macrophages exhibit a more modest increase in glycolytic activity compared to nonactivated macrophages [46]. To further understand our HP results, it is important to look into another common pathway for HP [1-13 C]pyruvate metabolism, through PDH. In M1activated macrophages, our spectrophotometric results showing decreased PDH activity compared to controls are in line with previous reports showing significantly reduced PDH activity in M1 primary murine bone-marrow-derived macrophages [39]. Reduced PDH activity in M1 macrophages would lead to less HP [1-13 C]pyruvate entering the Krebs cycle, shuttling this HP probe towards LDH and increased HP [1-13 C]lactate production, in line with our HP results. In M2-activated macrophages, on the other hand, we observed a significant increase in PDH activity. This result is interesting, and reasonable given that expression of pyruvate dehydrogenase kinase (PDK), an inhibitor of PDH, is reduced in M2 polarization [47]. In that case, and contrary to M1, increased PDH activity may lead to more HP [1-13 C]pyruvate being shuttled into the mitochondria, thus decreasing flux towards HP [1-13 C]lactate production via LDH. This result provides an additional explanation for the fact that HP lactate-pyruvate ratio was not significantly increased in M2-activated macrophages, despite the increased LDH activity. It should be noted that HP [1-13 C]pyruvate metabolized through PDH can generate HP [1-13 C]bicarbonate via pyruvate decarboxylation, and measurements of HP [1-13 C]bicarbonate have been used to determine PDH flux in previous studies [48,49]. Although the spectral bandwidth in this study was large enough to cover the resonance of HP [1-13 C]bicarbonate at 163 ppm, the signal of this metabolite was not detected, likely due to its intrinsically low level in this cell type. Finally, in addition to enzymatic activities, the levels of steady-state lactate levels from 1 H NMR might also contribute to the HP readouts through the well-documented pool-size effect [50][51][52]. The observed significant decrease in 1 H lactate levels in M2 macrophages may also contribute to the reduced HP lactate levels in that group. In M1 macrophages, lactate levels follow an increasing trend (p = 0.128), which is also in line with the HP results. Multiple additional factors contributing to the HP readouts could be considered, including levels of membrane transporters (e.g., MCT1) or NAD cofactor availability. However, such measures could not be performed in this study due to the limited number of paired samples available.
Our results show that conversion of HP [1-13 C]DHA to HP [1-13 C]AA was increased in M1 macrophages, mirrored by increased ROS levels, but unexpectedly not by changes in GSH levels or GSH/GSSG ratio. It is well known that the GSH redox system is a mitigator of ROS [53], while also being coupled to the conversion of DHA into AA [54]: GSH is oxidized to GSSG by DHA, which, in turn, is converted to AA. Increasing ROS levels should theoretically deplete the available pool of GSH (and increase levels of GSSG), leaving less GSH available for the production of AA, and thus leading to a decreased HP AA-to-DHA ratio, as previously reported [36,55]. However, both GSH/GSSG ratios, as detected by spectrophotometric assay, and total GSH levels, as detected by 1 H NMR, were not significantly different between M1 macrophages and control or M2. A previous study of the same cell line also showed that the total pool of GSH as detected by 1 H NMR was not increased post-activation by LPS [33]. Other reports using biochemical methods also show no changes in total GSH levels with M1 activation, but do note elevated GSH/GSSG ratio [56]. Elevated GSH/GSSG ratio can explain an increase in AA production, and may be a compensatory effect unique to macrophages, as they endogenously manufacture ROS as part of the immune response. Our results show that a trend to an increase in M1-activated compared to control (p = 0.1265) and M2 (p = 0.1083) was seen, although it did not reach significance, possibly due to the limited numbers of available paired samples (N = 3) that were spread thinly across the numerous assays. Further work could be done with a larger sample size to confirm the increase in GSH/GSSG ratio in M1 macrophages. Importantly, conversion of HP [1-13 C]DHA to HP [1-13 C]AA can also be affected by the levels of the glucose membrane transporter 1 (GLUT1), which is a known facilitated transporter of DHA [57]. GLUT1 has been shown to be upregulated with M1 activation compared to quiescent [58,59], whereas its expression is comparable between control and M2-activated states [42,60]. These facts mirror our HP AA/DHA ratio results and could be another possible mechanistic explanation to our observed HP data. Due to the limited number of paired samples, however, membrane transporter expressions were not evaluated here.
The power of dynamic metabolic probing using HP 13 C MR is compelling, as it allows the measuring of previously inaccessible metabolic reactions noninvasively. HP 13 C metabolic imaging is a rapidly growing field, and is now being used in multiple clinical trials across the globe, targeting several organs, including the brain, heart, prostate, and kidney [28,29,[61][62][63][64][65][66][67][68][69][70]. It is very likely that, given the constant improvements reported, both on the acquisition and processing sides, this methodology will soon approach feasibility for widespread clinical adoption. Currently, the most popular probe for hyperpolarization is HP [1-13 C]pyruvate, with some of the best polarization characteristics [52], but other probes are continuously being investigated, opening up possibilities not explored before. While the clinical data reported so far are highly compelling, mechanistic studies are still critically needed to understand the relative contribution of each cell type to the detected HP signal, especially for cells as ubiquitous as macrophages, which are found in most diseases and most organs. Here, we performed the first study of live macrophages at a clinically relevant field strength, and compared both M1 and M2 activation patterns. Before, the only other study that used HP 3 C MRS on activated macrophages was conducted at the high magnetic field strength of 11.7 Tesla, and employed M1 activation only [33]. We showed that, at clinical field strength, the spectral resolution was sufficient to enable measurements of metabolic fluxes in live cells. Further, we demonstrated that M1 and M2 macrophages have a different HP metabolic signature, with both HP [1-13 C]pyruvate and HP [1-13 C]DHA. It is important to note that incubation with LPS and IL-13 is only one way to induce M1 and M2 activation, respectively. Other cytokines, or cytokine cocktails, could be used to induce M1 and M2 activation states (e.g., IL-14 for M2 [71], or interferon-gamma (IFN-γ) for M1 [72]), and further studies would be needed to confirm that the HP results observed in this study hold across activation protocols, as well as in vivo in preclinical models. Further, potential changes in metabolism caused by the change in cellular environment from flask to NMR tube would require additional investigation. Nevertheless, this study further establishes HP metabolic imaging as an interesting tool to assess inflammation, and our results could help increase the understanding of the metabolic readouts observed in vivo in preclinical models and patients.

Cell Culture
J774a.1 mouse macrophages (ATCC, Manassas, VA, USA) were grown in Dulbecco's modified Eagle's media (DMEM) containing 10% fetal bovine serum and 5% penicillin/ streptomycin (UCSF). M1 activation was achieved with 100 ng/mL LPS treatment for 24 h (E. coli; Sigma Aldrich), and M2 activation with 5 ng/mL of murine IL-13 for 24 h (Peprotech), as previously described [32,36]. A control group was established with a vehicle treatment of sterile PBS. All passage numbers used were~4-14 to reduce the risk of genetic drift, and a mycoplasma testing kit (ATCC) confirmed the culture was contamination-free. Figure 1 presents a schematic of the full experimental design.
For HP 13 C experiments, cells were studied as paired samples of either M1 or M2 activation and paired controls. The definition of paired samples were as follows: three T225 flasks were split into ten T225 flasks, five of which were activated towards either M1 or M2, and the other five with vehicle as control. The adherent macrophages were incubated with 0.04% EDTA in calcium and magnesium-free PBS solution (Cell Culture Facility, UCSF) for approximately 10 min, collected, and then centrifuged at 125× g for 5 min. The pellet was resuspended in 200 µL of fresh DMEM (no additives), and a 20 µL sample was taken and washed with PBS to be saved at −80 • C for paired spectrophotometric assays (for all assays except ROS). Another 20 µL sample was taken for cell counting (referred to as subsample in the rest of this manuscript), and the remaining slurry suspension (~20 million cells) was transferred to a 5 mm NMR tube. All injections of HP probes were done within 5 min of cell resuspension and transfer to the NMR tube. Hyperpolarized spectra were then acquired on a 1.47 T Oxford Pulsar NMR system using a 1D pulse acquire sequence with the following parameters: flip angle = 20 • , repetition time (TR) = 3 s, pulse type = WALTZ-4, pulse length = 4 µs, and number of scans (NS) = 100, for a total acquisition of 5 min. Analysis was performed with Mestrenova (Mestrelab, Santiago de Compostela, Spain) software. Signal to noise ratio (SNR) for the injected substrates and detected products were calculated from the total summed spectra as area under curve divided by standard deviation of the noise. Contaminants were identified via blank experiments with hyperpolarized substrates injected into NMR tubes with media only. Spin-lattice relaxation time T1 of each HP probe was also calculated from such blank experiments. All HP data are represented as mean ± standard error of the mean (SEM), and are normalized to cell number and volume of injection.

Spectrophotometric Assays
ROS levels, arginase activity, LDH activity, PDH activity, and GSH levels were measured with spectrophotometry on paired samples collected from the same flasks as the ones used for HP experiments. ROS data of control (N = 6), M1-activated (N = 6), and M2activated (N = 3) were reported as fold-change from control using a commercial intracellular ROS fluorescence assay kit (Abcam, Cambridge, UK), used according to manufacturer's instructions). Arginase assay (Abcam) activity of control (N = 6), M1-activated (N = 3), and M2-activated (N = 3) was reported in units/Liter (U/L) and normalized to 10 5 cells per well. An in-house LDH assay measuring the rate of NADH (Sigma) depletion for control (N = 6) and activated groups (N = 3 for both M1 and M2) was performed and reported as µM NADH/minute normalized to protein concentration quantified by Bradford assay (Thermofisher, Waltham, MA, USA). A PDH assay kit (Abcam, Cambridge, UK) was used to measure PDH enzyme activity between groups (N = 3 for each control, M1, and M2), and reported as optical density (milliOD)/minute with normalization to protein concentration. GSH and its oxidized form GSSG were measured with a commercial kit (Biovision, San Francisco, CA, USA). The total glutathione levels were first measured, then subtracted from measured levels of GSH to obtain GSSG values, with the subsequent GSH/GSSG ratio reported with normalization to 10 5 cells per well. All spectrophotometric data are reported as mean ± standard deviation.

High Resolution 1 H NMR of Extracted Cell Metabolites
Metabolites from M1-activated (N = 5), M2-activated (N = 5), and control (N = 4) J774a.1 cells were extracted using equal parts methanol-water-chloroform, as previously described [74]. Cold 4 • C saline (5 mL) was added directly to T75 flasks 2-3 times and removed to rinse, and −20 • C methanol (3 mL) was subsequently added. The adherent macrophages were scraped off into the methanol-this mixture was transferred to a clean tube. Equal parts −20 • C chloroform and 4 • C H 2 O were homogenously mixed, and the fractions separated by centrifuging at 125× g at 4 • C. The methanol fraction was collected, 0.65 mM trimethylsilylpropanoic acid (TSP) (Acros Organics) added, and the mixture lyophilized. The resultant extracts were reconstituted in 420 µL D 2 O, and the samples were scanned on an 800 MHz NMR system (Bruker, Billerica, MA, USA) with a 1D NOESY presaturation sequence with TR = 3.31 s, mixing time = 50 ms, and NS = 384. A subsequent fully relaxed scan was performed with a relaxation delay of 30 s to ensure a time of at least 5 times the longest T1 based on literature values [75]. This spectrum was used to generate correction factors that were then subsequently applied to all spectra to enable absolute quantification of metabolites. Spectral processing was performed with Mestrenova (Mestrelab, Santiago de Compostela, Spain), and a select group of metabolites of interest previously reported in macrophage studies [33] were fitted and quantified using Chenomx NMR Suite (Chenomx Inc, Edmonton, AB, Canada) with reference to the Human Metabolomics Database [76]. Concentrations of quantified metabolites were normalized to cell number and TSP reference, and reported as mean ± standard deviation.

Statistical Analyses
The sample sizes of HP experiments were determined using an 80% power calculation of preliminary paired data between activated and nonactivated groups. All data were tested with two-way ANOVA between activated and nonactivated, with Sidak's multiple comparisons test. All tests were performed with Prism 8 (GraphPad) software.