Mapping glycine uptake and its metabolic conversion to glutathione in mouse mammary tumors using functional mass spectrometry imaging

Abstract

Although glutathione plays a key role in cancer cell viability and therapy response there is no clear trend in relating the level of this antioxidant to clinical stage, histological grade, or therapy response in patient tumors. The likely reason is that static levels of glutathione are not a good indicator of how a tissue deals with oxidative stress. A better indicator is the functional capacity of the tissue to maintain glutathione levels in response to this stress. However, there are few methods to assess glutathione metabolic function in tissue. We have developed a novel functional mass spectrometry imaging (fMSI) method that can map the variations in the conversion of glycine to glutathione metabolic activity across tumor tissue sections by tracking the fate of three glycine isotopologues administered in a timed sequence to tumor-bearing anesthetized mice. This fMSI method generates multiple time point kinetic data for substrate uptake and glutathione production from each spatial location in the tissue. As expected, the fMSI data shows glutathione metabolic activity varies across the murine 4T1 mammary tumor. Although glutathione levels are highest at the tumor periphery there are regions of high content but low metabolic activity. The timed infusion method also detects variations in delivery of the glycine isotopologues thereby providing a measure of tissue perfusion, including evidence of intermittent perfusion, that contributes to the observed differences in metabolic activity. We believe this new approach will be an asset to linking molecular content to tissue function.

Graphical Abstract

1. Introduction

Glutathione is one of the primary cellular antioxidants and plays a central role in neutralizing oxidative stress in tumor tissue imparted by reprogrammed metabolism, the harsh hypoxic and acidic environments and the additional stress produced by therapy [1–4]. For these reasons, a number of clinical studies sought to link glutathione levels in patient tumors to tumor clinical stage, histological grade or therapy response but have yielded mixed results [5]. These results are likely due to the difficulty in relating the static level of a single biomarker to the functional capability of tumor tissue to replenish and maintain antioxidant capacity in response to stress. This functional capability can be assessed by measuring metabolic rates as these are the net result of the interactions of an entire network of molecular components in the tissue under regulatory control and environmental influences [6–8]. Since tumor tissue is notoriously heterogeneous in molecular content and in microenvironmental conditions, the metabolic activity will vary substantially across the tissue. Therefore, functional imaging is needed to map these variations across the tumor. In prior work, we used magnetic resonance methods to noninvasively detect the heterogeneity of isotope incorporation into tumor glutathione, but the low sensitivity limited the amount of kinetic information obtainable from these experiments [9, 10].

We developed a new method of functional mass spectrometry imaging (fMSI) [11] that used timed administration of isotopologues of glycine that detected glutathione metabolic variations at each spatial location in ex vivo liver tissue sections [12]. This single-sample isotope tracer method to provide detailed kinetic data is based on the work of Dudley et al., [13] and later dubbed multiple-infusion start time (MIST) [14] and is particularly advantageous to studying heterogeneous tumor tissue. This method can provide multiple timepoint kinetic data from a single tissue sample. In this report, mice bearing orthotopically implanted 4T1 mammary tumors were infused with three isotopologues of glycine, [2-¹³C,¹⁵N]-glycine, [1,2-¹³C₂]-glycine, [1,2-¹³C₂,¹⁵N]-glycine, at timed intervals followed by a single tissue harvest. Mass spectrometry analysis of the tissue shows isotope labeling patterns that reflect the time-dependent uptake of each glycine isotopologue and its conversion into glutathione allowing kinetic analysis of metabolic activity. The results show that glycine uptake is heterogeneous, and glutathione metabolic activity is highest at the tumor periphery.

2. Materials and Methods

2.1. Animal Studies

Female Balb/c mice (6–8 weeks old) were obtained from Charles River (Wilmington, MA). Approximately 10⁵ 4T1 tumor cells (ATCC) were injected into the mammary pad of the mice. After 12–21 days, the tumors range from 0.5 to 1 cm³ in size. Mice were anesthetized (isoflurane/oxygen) and the tail vein was cannulated. The three stable isotope-labeled (SIL) isotopologues of glycine, [2-¹³C,¹⁵N]-glycine (abbreviated: ¹gly), [1,2-¹³C₂]-glycine (²gly), and [1,2-¹³C₂,¹⁵N]-glycine (³gly) (Fig. 1A), and [¹³C₂,¹⁵N]-urea (*urea) were obtained from Cambridge Isotope Laboratories (Tewksbury, MA) and were dissolved in phosphate-buffered saline. The isotopologues were administered using the MIST protocol to the anesthetized mice by bolus injection (0.05 mL) at doses of 300 μmoles/kg via the tail vein cannula at 0-, 40- and 60-minute timepoints. At the 119-minute timepoint, a 50 μmol bolus of isotope-labeled urea (*urea) was administered as a perfusion marker and mice were euthanized at 120 min (Fig. 1C). Control experiments infused unlabeled glycine (⁰gly) over the same timing sequence. Tissues were harvested and immediately frozen in isopentane cooled in liquid nitrogen. Tissues were stored at −80 °C until sectioning for MSI studies. The use of animals and all associated procedures were reviewed and approved by the NC State University Institutional Animal Care and Use Committee. The 4T1 cell line was authenticated by short tandem repeat profiling by the ATCC.

Figure 1.

(A) Glycine isotopologues used in this study with listed exact masses. (B) Glutathione is biosynthesized in two steps from three amino acids catalyzed by glutamate cysteine ligase (GCL) and glutathione synthetase (GS). The three glycine isotopologues (*gly) yield the corresponding glutathione isotopologues (*GSH). (C) Timing of the MIST administration of glycine isotopologues (*gly) and isotope labeled urea (*urea).

2.2. Sample Preparation

Microscope slides were evenly sprayed with a solution of 1 mg/mL homoglutathione (Bachem) in 50% methanol using a pneumatic sprayer as described earlier [12]. The tumor tissues were sliced to a 20 μm thickness with a cryostat prior to being thaw mounted on the prepared slide for mass spectrometry imaging. Adjacent 10 μm thick sections were prepared for histological examination.

2.3. Mass Spectrometry Imaging, Data Analysis and Processing

All tissue analyses were performed using a home-built infrared matrix-assisted laser desorption electrospray ionization (IR-MALDESI) source coupled with an Exploris 240 Orbitrap Mass Spectrometer (Thermo Fisher Scientific, Bremen Germany) for MSI by the methods previously described [12]. Recorded mass spectra and isotopic distributions were directly viewed and analyzed in XCalibur. Specifics of data analysis and the software tool used to generate metabolite isotopic enrichments appear in our recent publication [12]. Glutathione was detected in negative ion mode whereas glycine and urea were collected in positive ion mode. In positive ion mode, the m/z range 45–225 was recorded. Internal calibration was used to achieve sub-ppm mass accuracy. For glutathione detection, ions were analyzed in negative ion mode with a m/z range of 286 – 346. We used an ESI solvent of 50% acetonitrile and 0.2% formic acid. The amount of isotope incorporated into metabolites was performed using the Percent Isotope Enrichment (PIE) tool in the MSiReader software [12]. The MSI data is collected from tissue 20 μm thick tissue section with resulting sampled volume elements (voxels) of 150 × 150 × 20 μm size. Functional heatmaps displaying voxel-by-voxel slopes were generated from MSiReader PIE tool output in RStudio using the nlme, viridis, and ggplot2 packages.

2.3. Histology and Immunohistochemistry

Ten-micron tissue sections were obtained adjacent to those used for mass spectrometry analysis. One section was stained with hematoxylin-eosin stain. This tissue section was analyzed to detect regions of necrosis and neutrophil infiltration by a board-certified veterinary anatomic pathologist (DM) using light microscopy.

3. Results

3.1. Distribution and Labeling of Glutathione

All fMSI experiments used heavy SIL glycines, designated as *gly, with control experiments infusing unlabeled, naturally abundant glycine (⁰gly). The labeling patterns of the three isotope-enriched isotopologues of glycine (¹gly, ²gly, ³gly), are shown in Fig. 1A and yield distinct species with increasing exact masses that can be resolved by high resolution mass spectrometry [12]. These *gly species can be incorporated into glutathione in a two-step biosynthetic process catalyzed by the enzymes glutamate cysteine ligase (GCL) and glutathione synthetase (GS) to yield isotopologues of glutathione (*GSH) that can also be resolved by high resolution mass spectrometry (Fig. 1B). The MIST protocol was used to administer bolus doses of ¹gly, ²gly, ³gly at 120, 80, and 60 minutes, respectively, prior to tissue harvest as shown in Fig. 1C. Isotope labeled urea (*urea) was administered one minute before harvest as a perfusion marker [15]. Incorporation of the three glycine isotopologues for three different time intervals resulted in the enrichment of three isotopologues of glutathione ¹GSH, ²GSH, ³GSH, after 120, 80 or 60 minutes, respectively. Therefore, each tissue sample will contain unlabeled glutathione, ⁰GSH, and ¹GSH, ²GSH, ³GSH. The percent isotope enrichment (PIE) of each isotopologue in the sample, obtained from the mass spectrometry imaging data, represents the ratio of isotope-labeled to total glutathione from the three time points.

A harvested 4T1 tumor tissue sample that was used to obtain thin sections for mass spectrometry and histological analyses is shown in Fig. 2A. The overall cross-sectional dimensions of the tissue shown in Fig. 2A was approximately 11 × 6 mm. Thin 20 μm sections of this tumor were analyzed by MSI. The steady-state levels of unlabeled glutathione (⁰GSH), detected at m/z 306.0765, and normalized to the homoglutathione standard, is shown in Fig. 2B. This image shows high glutathione levels at the periphery of the tumor with a large central region with low/no detectable glutathione. Similarly, heatmaps of the steady-state abundances of the three glutathione isotopologues detected at m/z’s 308.0769 (¹GSH), 308.0832 (²GSH), 308.0803 (³GSH) are shown in Figs. 2C–2E, respectively. As expected, the ¹GSH, synthesized from ¹gly over a 120-min exposure is highest in concentration, followed by ²GSH (80-min exposure) and ³GSH (60-min exposure) to the other glycine isotopologues. The PIE heatmaps of all glutathione species are presented in Figs. 2F–2I. Due to the concentration of the administered *gly and the 120 min timeframe, most (~95%) of the glutathione in the tissue is unlabeled resulting in Fig. 2F that shows a high PIE of unlabeled glutathione. The PIE maps for isotopologues ¹GSH, ²GSH and ³GSH are shown in Figs. 2G–2I. Similar data was obtained from 4T1 tumors excised from three mice treated with the identical MIST protocol and all show high levels of glutathione along the outer edge of the tumors and a central core with little to no glutathione detected (Supplementary Material, Fig. S1). The PIE data for each of these three additional tumors are shown in Figs. S2. In all tumors, the distributions of the three glutathione isotopologues are also heterogenous. The relative levels of glutathione isotopologues detected in the four tumors administered ¹gly, ²gly, ³gly are summarized by the box and whisker plots in Fig. 3 showing the PIE data collected from all tissue voxels in which isotopologues ¹GSH, ²GSH, ³GSH were detected. All tumors show relative PIE levels ¹GSH > ²GSH > ³GSH as expected for tissue exposed for 120, 80 and 60 min to the labeled glycine substrates. These data illustrate the ability of this fMSI method to generate three timepoint kinetic data from a single tissue sample. The GSH heatmap for a negative control tumor, treated with three doses of unlabeled glycine, also shows high levels of glutathione at the edges of the tumors, and a central region with low glutathione (Fig. S3A). Abundances of ¹GSH, ²GSH, ³GSH were not detected above signals expected from the natural abundance of these isotopologues (Fig. S3B–S3D). The PIE data for this tumor shows high levels of unlabeled ⁰GSH (Fig. S3E), as expected, and no ¹GSH, ²GSH, ³GSH (Figs. S3F–S3H).

Figure 2.

(A) Excised 4T1 tumor from MIST labeled mouse mounted on specimen disc for cryosectioning. MSI heatmaps depicting steady state abundances of (B) glutathione (⁰GSH), (C) ¹GSH at m/z 308.0769 (D) ²GSH at m/z 308.0832 and (E) ³GSH at m/z 309.0803. SIL abundances are corrected for isotopologue signal overlap. All heatmaps are normalized relative to the homoglutathione standard. PIE heatmaps of (F) unlabeled glutathione, ⁰GSH, (G) ¹GSH, (H) ²GSH and (I) ³GSH detected. Region of interest analyzed with PIE tool is outlined. Scale bar = 1mm.

Figure 3.

(A) Box and whisker plots summarizing the PIE data from all detectable labeled isotopologues in the tumor shown in Fig. 2. The number over each plot is the median PIE for the isotopologue data. (B)-(D) Box and whisker plots summarizing the *GSH PIE data for the three 4T1 tumors shown in Supplementary Material Fig. S3–S6.

3.2. Distribution and Uptake of Labeled Glycines

The relative rate of incorporation of isotope labels from glycine into glutathione is dependent upon delivery of the glycine isotopologues to the tissue which, in turn, is a function of how well the tissue is perfused. Perfusion of tumor tissue is known to be heterogeneous so mapping the uptake of the glycine species across the tumor provides an indication of the variation of the delivery of these substrates. Unlike the glutathione data shown in Fig. 2 and Figs. S1, S2, mass spectrometry detection of glycine is most sensitive in positive ion mode. Therefore, a 20 μm tumor section collected within 100 μm to the one used to collect the data in Fig. 2, was used to collect the positive ion mode glycine data in Fig. 4. Fig. 4A shows the PIE distribution for unlabeled ⁰gly in the tissue. The MSI PIE maps of the glycine isotopologues are shown in Figs. 4B – 4D. Similar to the *GSH data, the glycine isotopologue uptake shown in the tumor section in Figs. 4B – 4D is heterogeneous with the highest levels in the lower region of the tissue. To confirm that *gly uptake represents a measure of tissue perfusion, the positive ion MSI map of *urea uptake is shown in Fig. 4E. Since *urea was administered one minute prior to tissue harvest, the uptake indicates the regions of highest perfusion in the tissue which correlates well with the regions indicating high glycine uptake. The data in Fig. 4 shows more extensive uptake of ¹gly followed by ²gly and then ³gly. The PIE maps of the glycine isotopologues and *urea for the three other tumors studied are shown in Fig. S4.

Figure 4.

PIE heatmaps of: (A) unlabeled glycine ⁰gly, (B) ¹gly, (C) ²gly (D) ³gly detected in a thin section of the 4T1 tumor adjacent to the section shown in Figure 2. (E) MSI data showing the abundance heatmap of perfusion marker SIL-urea (*urea) detected in the same thin-section as in (A-D). Scale bar = 1mm.

3.3. Time Dependence of Label Uptake and Incorporation

The preceding data indicates that glycine isotopologue uptake and conversion into glutathione is heterogeneous resulting in differences in the total number of isotopologues detected in each tissue voxel. Fig. 5A shows the variation in the number of glutathione isotopologues found in each voxel. From the data mapped in this figure, the bottom half of the tumor shows a high concentration of voxels in which all three isotopologues were detected. The upper half, show a high level of voxels in which fewer than three isotopologues were detected. There is a large central region in which no glutathione isotopologues are detected. Fig. 5B shows the map of the number of *gly isotopologues that were detected in the adjacent tissue section that was extracted from the data shown in Fig. 4. The glycine isotopologues are detected across more of the tissue than for glutathione when you compare Fig. 5A to 5B showing substrate delivery was not limiting for the lack of glutathione production in some regions of the tumor. Fig. 5A shows many voxels in which only one isotopologue was detected. Primarily ¹GSH was detected in these voxels. Voxels containing only one glycine isotopologue are less frequently observed (Fig. 5B).

Figure 5.

(A) Map of the number of *GSH isotopologues found at each tissue voxel. (B) Map of the number of *gly isotopologues detected in each tissue voxel. (C) *GSH PIE levels over time from sample voxels in the MSI data (D) Heat map showing the slopes of the lines fitted to *GSH PIE data from all voxels containing ≥2 data points. (E) Heat map showing the slopes of the lines fitted to *gly PIE data from all voxels containing ≥2 data points. Region of interest analyzed with PIE tool is outlined.

Fig. 5C shows a random sampling of *GSH PIE data from individual voxels (comprising the heat maps shown in Fig. 2 and boxplots in Fig. 3) as a function of the time interval over which each isotopologue was synthesized (i.e., difference between infusion and time of tumor harvest.) All three *GSH isotopologues were detected in Scans 375 and 376 but fewer in other voxels (e.g., Scans 370 and 371 in Fig 5C.) Even with only the one species detected (Scan 370), the PIE level of this isotopologue indicates active metabolism at this location. In many voxels, no enriched glutathione was detected. Regions in which <3 isotopologues detected, are likely due to lower/intermittent tissue perfusion or low metabolic rate. For voxels in which 2 or more *GSH species were detected, a line was fit to the data where the slope of this line provides an estimate of the rate of production of glutathione, i.e., PIE/min, from the labeled metabolites in that voxel.

Heatmaps of the slope data, reflecting the PIE per minute for glutathione uptake of label from glycine in every tissue voxel are shown in Fig. 5D. In most tissue voxels in this tumor, positive slopes of varying magnitude were observed for the *GSH data. The data in Fig. 5D show greatest label uptake on the central right and lower left portion of the tissue. Interestingly, the upper left corner of the tumor showed both high steady-state glutathione levels (Fig. 2B) and perfusion of all three labels (Fig. 5A), but slopes of lower magnitude than other areas of the tumor. The slope data for glutathione PIE versus time for the other tumors data are shown in Fig. S5. Similar to Fig. 5D, the data for label uptake into glutathione in these other tumors show positive slopes indicating labeling of the glutathione pool was increasing during the 120 min experimental interval in all tumors.

Slope data, reflecting the rate of glycine uptake into the tissue in all voxels are shown in Fig. 5E. In this case, some regions of the tumor exhibit a negative slope for the glycine data, and these are mainly localized to the lower part of the tumor. Glycine slope data from the other tumors are shown in Fig. S6. Interestingly, the data in Fig. S6A primarily displays positive glycine slopes, whereas the data for tumor tissue in Figs. S6B are dominated by negative slopes illustrating the intertumoral heterogeneity in tissue perfusion. Since these are bolus injections, the negative slope indicates the glycine substrates are in the washout phase in certain locations in the tissue. Theoretically, bolus wash-in and wash-out kinetics are not linear and more complex line fits are needed to account for the actual behavior of the dosed isotopic substrates.

The data in Fig. 5A show many voxels in which zero, one or two *GSH isotopologues are detected which is likely due to no or low perfusion (zero isotopologues detected) or intermittent perfusion (one or two isotopologues detected. In support of this conclusion, we extracted data from our earlier study on liver tissue [12]. Since healthy liver is a well-perfused tissue, a MIST protocol using the same glycine isotopologues should detect all three glutathione isotopologues across the tissue. The data in Fig. S7 confirms that all three *GSH isotopologues are found across most of the tissue.

3.4. Necrosis and Neutrophil Infiltration Identified by Tumor Histology

Additional tumor tissue thin sections obtained from the tissues were stained with hematoxylin-eosin and the data for the section obtained from the tumor shown in Fig. 2 is shown in Fig. 6. A region of tissue outlined in black in Fig. 6A was identified as a region of coagulative necrosis characterized by retained cellular outlines with loss of differential staining. This region corresponds to levels of little to no detectable unlabeled glutathione and the absence of glutathione isotopologues indicating no metabolic activity (Fig. 5D). However, compared to the lack of glutathione isotopologues in this region, there are a number glycine isotopologues present, although this region has a high density of voxels with less than three glycine isotopologues (Fig. 5B). In addition to identifying tissue necrosis, several regions of the tissue had a high level of neutrophil infiltration indicative of localized inflammation and are delineated by yellow borders in Fig. 6. Two areas of increased neutrophil infiltration border on the central necrotic region. All regions displaying neutrophil infiltrations all show a low rate of isotope incorporation into glutathione (Fig. 5A).

Figure 6.

(A) Hematoxylin-eosin stained tissue thin-section obtained from the tumor shown in Fig. 2A. The black line indicates a region of coagulative necrosis. (B) The same section with yellow lines indicating regions of the tissue with increased neutrophil infiltration.

4. Discussion

Measures of metabolic activity or flux reflect the net outcome of the interactions of genes, proteins and metabolites under regulatory control operating within the local tissue microenvironments. Glutathione is a key player in the development, progression, and therapy response in cancer [4, 16–19] and as such, a measure of its functional activity will likely be a better indicator of cancer aggressiveness and predictor of treatment outcome than steady-state levels. In addition, the innate metabolic and microenvironmental heterogeneity of tumor tissue requires this functional activity be assessed across the tissue. This is accomplished through functional imaging methods which often use isotope tracking technologies. Perhaps the best-known functional imaging method used in clinical oncology is positron emission tomography (PET) to detect glucose delivery, uptake, and phosphorylation. To date there is no PET probe to detect glutathione metabolic activity. We developed a magnetic resonance method to detect glutathione metabolism noninvasively in rodent tumor tissues [9, 10] but the low sensitivity of the method limited the amount of kinetic and heterogeneity data that could be obtained. An additional advantage of some noninvasive imaging methods is the capability to provide multiple time point kinetics. However noninvasive methods such as PET and magnetic resonance can be limited in the metabolic pathways that can be probed and often yield images with poor spatial resolution.

Classic isotope tracing experiments can be used to track many more metabolic pathways, but these methods are often invasive, and a commonly used approach is to administer a single tracer and harvest tissue samples at multiple time intervals after infusion. This works well in some settings but is not readily amenable to MSI analysis as it is not possible to obtain data from identical tissue spatial locations in each serial sampling. This is especially true in tumor tissue which is comprised of multiple cell types and tissue environments with few, if any, anatomical features in which to orient similar sampling sites. The MIST protocol replaces a single tracer infusion and multiple tissue harvests approach with multiple tracer administration and a single tissue harvest to provide kinetic data. This approach has been used previously to follow protein biosynthesis [13, 14, 20] and glutathione metabolic rates [21] in tissue extracts. The combination of MIST with MSI that is the basis of this fMSI method demonstrated here is not as elegant as many of the noninvasive functional imaging methods available but offers some distinct advantages. First, the MSI method can unambiguously identify and map multiple molecular species and pathways beyond the capabilities of most other imaging methods. Second, immunohistochemistry analysis of adjacent thin sections can be used to correlate classic tissue biomarkers to functional metabolic activities. Third, tissue perfusion measures can be acquired at each spatial location that is necessary to assessing variations in metabolic activity. Fourth, many MSI methods provide much higher spatial resolution than methods such as PET or functional magnetic resonance. Finally, a single-sample MIST approach precludes the need for multiple tissue samples, or animal subjects, that is frequently needed in acquiring dynamic data.

Similar to our magnetic resonance results in rat fibrosarcomas [10] and mammary [9] tumors, the fMSI data shows higher glutathione labeling located near the outer periphery of 4T1 tumors. However, our magnetic resonance method, although noninvasive, had poor sensitivity and spatial resolution. Due to this poor sensitivity, label incorporation could only be detected at a single timepoint. We were later able to use magnetic resonance to noninvasively monitor the time dependence of glutathione synthesis in rat and human liver but were not able to map variations in metabolic activity differences across the tissue [22]. The increased levels of glutathione metabolic activity detected at the outer edges of the 4T1 tumor using fMSI is consistent with earlier studies of this tumor that showed increased cellular proliferation [23], glucose uptake [24] and tissue perfusion at the tumor periphery. All three factors would be expected to contribute to high glutathione metabolic activity. In contrast, the histological data show central regions of necrosis and the imaging data in this necrotic tissue showed little to no detectable glutathione but did detect the presence of both steady-state glycine and glycine isotopologue uptake. This indicates that metabolic substrates are delivered to these necrotic regions, but glutathione metabolic machinery is absent. The maintenance of glycine delivery to what would normally be considered regions of poor vasculature may be a sign of ‘early necrosis’ previously observed in magnetic resonance studies of the 4T1 tumor where the blood vessels that are present in this region are highly permeable leading to release of contrast agents [25]. These permeable vessels may also facilitate release of the glycine substrates into the extracellular space of the central necrotic region that was detected in this study by fMSI.

The value of obtaining multiple timepoints as demonstrated by this fMSI method is that the trend in the metabolic kinetics may be determined at each spatial location. For example, the data indicates that incorporation of isotope-labeled glycine is still increasing in most tissue regions in the 4T1 tumor over the 120 min time course of the experiment shown in Fig 5E but is in a washout phase in other regions of the tumor. A single timepoint reading would miss these differences. Whether the glycine isotopologues are in a wash-in or wash-out phase will also affect glutathione kinetics. Rapid washout from the tissue may not allow time for the labeled glycine to be incorporated into glutathione resulting in glutathione synthesis rate data that is lower than the actual value. Determining whether low metabolic activity is due to low perfusion or conversely, high perfusion with label wash-out, will require fitting both glutathione and glycine kinetic data to models of label transit through the metabolic network at each spatial location in the tissue. Particularly in this study, where bolus doses are administered, label wash-in and wash-out kinetics needs to reflect both substrate uptake and washout and glutathione synthesis and consumption provided by modeling. The metabolic data in this study was obtained through linear fits of the PIE data as a function of time and is a convenient way to visualize the data and a reasonable approximation of the rate. However, the uptake and metabolism of the labeled substrates would be more accurately represented by a more complex function. The MIST procedure has previously been used with both a bolus injection [20] and continuous infusion [13, 14, 21] protocols and both these approaches can provide unique data on dynamic biological processes. We are currently working on different infusion protocols and timing schemes and developing more physiologically accurate metabolic models to fit the data and incorporating this into the fMSI analysis software. However, these results also emphasize that perfusion must be considered when assessing metabolic heterogeneity.

There are regions in all four tumors where glutathione is absent but glycine isotopologues are present indicating the glutathione metabolic machinery is not active in these regions. Perfusion differences detected by variations in glycine isotopologue uptake were supported by unique MSI data that mapped isotope-labeled urea uptake. Urea is a metabolite found at high concentration in most tissues and rapidly diffuses from the blood and is quickly taken up by tissue [26]. The use of isotope-labeled urea in perfusion imaging was demonstrated in hyperpolarized magnetic resonance studies [15] and was adopted in this study for use with MSI. This method readily detected well-perfused regions of tumor tissue. In contrast, poorer perfusion is detectable in some tumor regions by low uptake of the three glycine isotopologues. The data also shows that timed MIST protocol may be capable of detecting intermittent perfusion as only one or two glutathione isotopologues were detected in some tissue voxels. Re-examination of our liver data supports the conclusion that intermittent perfusion may be responsible for detection of fewer than three isotopologues in the tissue. Early experiments to detect intermittent perfusion and cycling hypoxia used timed dosing of fluorescent perfusion markers [27] and timed dosing of hypoxia markers [28] to detect the dynamics of vascular flow and oxygen delivery. The MIST dosing schedule is analogous to the timed administration of perfusion and hypoxia markers used in these early studies and supports our hypothesis that the glycine and glutathione data in some tissue voxels reflect the effects of intermittent blood flow. Intermittent perfusion, leading to cyclical variations in tissue oxygen levels, is often observed in regions of tumor tissue [29] and may be a driver of tumor progression [30] so detection and analysis of such tumor fractions is critical to assessing the metabolic consequences of limited blood flow. The detection of a single isotopologue in the data shown in Fig. 5C (i.e., Scan 370) suggests that intermittent perfusion does not stop label incorporation into glutathione. How these data are fit into a metabolic model, or visually displayed, is still under consideration. In addition, fMSI experiments to better characterize regions of intermittent perfusion are planned. Although the slopes of the linear fits to the enrichment data are an approximation of the metabolic rates, they do illustrate the heterogeneity in both glycine uptake and metabolism into glutathione and demonstrate the value of the fMSI technique using MIST infusion protocols.

The MIST infusion protocol shown in Fig. 1C worked well in mapping metabolic and perfusion variations in the 4T1 tumor. However, our initial attempts to study glutathione metabolism in the 4T1 tumor that used a MIST timing protocol identical to that used in our liver studies [12], resulted in no detectable uptake of isotope labels into glutathione. In the liver study, the glycine isotopologues were administered at time points of 60, 30 and 10 min and MSI detected all three glutathione isotopologues across the liver tissue. For the 4T1 tumor, glutathione metabolic activity is slower, so a longer MIST protocol totaling 120 min with appropriate time separation between doses was necessary to allow time for label incorporation and to generate the data shown. Therefore, the overall length of MIST infusion protocol and intervals between doses needs to be tailored to the metabolic rates to yield data reflecting the dynamics of the system. Although the 120 min interval was sufficient to detect label incorporation from glycine into glutathione, the mass range used in these studies to detect glutathione (m/z 286 – 346) precluded detection of oxidized glutathione (exact mass 612.1520 Da). The timeframe for labeling of oxidized glutathione would be expected to lag that of the reduced form and, combined with its lower concentration in cells, would be expected to be more challenging to reliably detect isotope incorporation. Future studies to incorporate a longer MIST infusion protocol and a wider mass detection range to detect this species are planned.

This study utilized an IR-MALDESI source paired with orbitrap mass analysis to generate imaging data with a spatial resolution of 150 μm. This is a far higher resolution than available from some functional imaging methods such as PET. There are a number of other MSI ionization and detection methods that can map isotope uptake with <1 μm spatial resolution [11]. Those methods, paired with MIST infusions, offer the possibility of obtaining single-cell or subcellular functional data. Advanced molecular profiling methods have already collected reams of data at the single cell to whole-tissue level and the addition of fMSI methods to the analytical arsenal over these size scales can provide unprecedented data on the spatial relationships between molecular components and biological function; the objective of the field of spatial biology [31]. We will continue to probe differences in glutathione metabolic activity in tumor tissue to establish functional measures in tumors from different tissue types and levels of aggressiveness. This communication focuses on the ability of fMSI to measure glutathione metabolism, but this method is amenable to any process in which multiple isotopologues of a metabolic substrate are available.

Highlights:

• Timed tracer infusions combined with MS imaging maps tissue dynamics

• Rates of glycine metabolism into glutathione are spatially heterogeneous in tumors

• High glutathione levels do not always correlate with high metabolic activity

• Maps of glycine and urea uptake yield perfusion images

• Timed dosing provides evidence of regions of intermittent perfusion

Data Availability Statement

Upon acceptance for publication, all tumor data will be made public in the METASPACE data repository.