Dissecting cell type-specific metabolism in pancreatic ductal adenocarcinoma

Tumors are composed of many different cell types including cancer cells, fibroblasts, and immune cells. Dissecting functional metabolic differences between various cell types within a mixed population can be limited by the rapid turnover of metabolites relative to the time needed to isolate cells. To overcome this challenge, we traced isotope-labeled nutrients into macromolecules that turn over more slowly than metabolites. This approach was used to assess differences between cancer cell and fibroblast metabolism in pancreatic cancer organoid-fibroblast co-cultures and in pancreatic tumors. In these contexts, we find pancreatic cancer cells exhibit increased pyruvate carboxylation relative to fibroblasts, and that this flux depends on both pyruvate carboxylase and malic enzyme 1 activity. Consequently, expression of both enzymes in cancer cells is necessary for organoid and tumor growth, demonstrating that dissecting the metabolism of specific cell populations within heterogeneous systems can identify dependencies that may not be evident from studying isolated cells in culture or bulk tumor tissue.


Introduction
Tumors are composed of a heterogeneous mix of cell types, including cancer cells and stromal cells such as fibroblasts, macrophages, and other immune cells.
How these different cell types interact to enable tumor growth is poorly understood.
Environmental context plays an important role in determining how cancer cells use nutrients to proliferate and survive, and non-cancer cells within a tissue can alter nutrient availability (Lyssiotis and Kimmelman, 2017;Mayers and Vander Heiden, 2017;Muir et al., 2018;Pavlova and Thompson, 2016;Sullivan and Vander Heiden, 2019). There is evidence that different cell populations within tumors can compete for limiting nutrients (Chang et al., 2015;Ho et al., 2015;Zecchin et al., 2017), and metabolic cooperation between different cell types can also influence tumor phenotypes (Linares et al., 2017;Sousa et al., 2016;Valencia et al., 2014;Vander Heiden and Deberardinis, 2017). Nevertheless, the technical challenges associated with studying the metabolism of individual cell types within a mixed population have limited a complete understanding of the metabolic interactions between cells in tumors. More broadly, this challenge has been a barrier to study how cells use nutrients differently within tissues to support both normal and disease physiology.
Cancer cell metabolism in culture can differ from the metabolism of tumors in vivo (Biancur et al., 2017;Cantor et al., 2017;Davidson et al., 2016;Mayers and Vander Heiden, 2015;Muir et al., 2017;Sellers et al., 2019;Voorde et al., 2019). This can at least partially be ascribed to changes in cancer cell metabolism that are driven by different nutrients present in the extracellular environment; however, another major difference between cell culture and tumors is the presence of additional cell types within tumors that are absent from most culture systems. The presence of many different cell types complicates the ability to characterize cancer cell metabolism in tumors, particularly in cases where a minority of the tumor is composed of cancer cells. For instance, cancer cells are a minority cell type in pancreatic ductal adenocarcinoma (PDAC) tumors (Feig et al., 2012), and an understanding of cell metabolism in these tumors requires de-convolution of cancer-specific and stromaspecific phenotypes. Furthermore, there is evidence that the metabolism of cancer cells and different stromal cells isolated from these tumors can be different from each other when studied in culture (Francescone et al., 2018;Halbrook et al., 2019;Sousa et al., 2016), and it is unknown whether the metabolic programs used by different cell populations in culture are also used within PDAC tumors in vivo where environmental conditions are different.
Studies of bulk tumor metabolism fail to capture information about metabolic heterogeneity with regard to different cell types (Xiao et al., 2019), and existing approaches are limited in their ability to study functional metabolic phenotypes in different cell populations in intact tissue. A major limitation arises from the fact that metabolic reactions take place on time scales that are faster than the turnover of many metabolic intermediates, complicating metabolite analysis after tumor digestion and cell sorting (Shamir et al., 2016). Furthermore, cell sorting exposes cells to conditions that are different from those experienced by cells in tissues, and can change metabolism in many ways. For example, sorting can induce mechanical and oxidative stress and reduce the levels of certain metabolites (Binek et al., 2019;Llufrio et al., 2018;Roci et al., 2016), and even including small amounts of bovine serum to the buffer was not sufficient to prevent changes to metabolite level changes during cell sorting (Llufrio et al., 2018). Indeed, metabolite labeling patterns can be more robust than metabolite levels when assessing metabolites in flow cytometry sorted cells, although labeling can also be affected by cell sorting, particularly in metabolites with rapid turnover such as lactate (Roci et al., 2016). Nevertheless, interpretation of metabolite labeling patterns in sorted cells is influenced by whether cells are at metabolic steady state (Buescher et al., 2015). Thus, the rapid timescale of metabolism can limit analyses of sorted cell populations and suggests that new approaches are needed to better understand metabolism of individual cell types within mixed cell populations such as tumors.
To overcome the challenges associated with studying cell metabolism within intact tumors and organoid co-cultures, we adapted an approach based on endproduct biomass labeling (Green et al., 2016;Hosios et al., 2016;Le et al., 2017;Lewis et al., 2014;Mayers et al., 2016;Shankaran et al., 2016). This technique has been applied in studies of microbial metabolism and metabolic engineering to better understand mixed populations of bacteria (Gebreselassie and Antoniewicz, 2015;Ghosh et al., 2014;Rühl et al., 2011;Zamboni et al., 2005). Unlike short-lived metabolic intermediates, turnover of end-product macromolecules such as protein is slow relative to the time period needed to isolate tumor cell populations. By assessing the isotope-labeling pattern of end-product biomass generated when cells are exposed to labeled nutrients in a mixed cell population, metabolic differences in nutrient use by different cells can be inferred. We used this approach to uncover a difference in glucose metabolism between cancer cells and fibroblasts in PDAC.
Specifically, we find that, relative to fibroblasts, cancer cells within PDAC tumors have increased use of glucose for tricarboxylic acid (TCA) cycle anaplerosis through increased flux through pyruvate carboxylase (PC). This phenotype is not evident when cancer cells and fibroblasts are studied as separate populations in mono-culture, even though PC is necessary for tumor growth in vivo. Furthermore, deletion of PC was insufficient to account for all pyruvate carboxylation activity within cancer cells in a mixed population, revealing that malic enzyme 1 (ME1) also contributes to pyruvate carboxylation in cancer cells when fibroblasts are present and is required for PDAC tumor growth. These data argue that tracing labeled nutrients into stable biomass can be used to reveal metabolic differences between subpopulations of cells in a mixed cell system and to identify phenotypes that depend on the co-existence of multiple cell types.

Glucose Metabolism in Pancreatic Tumors
PDAC involves tumors where cancer cells can be a minority cell population (Feig et al., 2012). To better understand glucose metabolism of PDAC tumor tissue in vivo, we infused U-13 C-glucose into conscious, unrestrained mice (Davidson et al., 2016;Hui et al., 2017;Marin-Valencia et al., 2012) bearing PDAC tumors from autochthonous models that are driven by activating mutations in Kras and disruption of Tp53 function (Bardeesy et al., 2006;Hingorani et al., 2005). Similar to what has been observed with other mouse cancer models and in humans (Davidson et al., 2016;Fan et al., 2009;Hensley et al., 2016;Sellers et al., 2015), extensive labeling of multiple metabolic intermediates is observed from U-13 C-glucose in pancreatic tumors and normal pancreas ( Figure 1, Figure S1).
To assess U-13 C-glucose labeling of metabolites in autochthonous pancreatic tumors arising in the LSL-KRas G12D ; p53 fl/fl ; Pdx-Cre (KP -/-C) (Bardeesy et al., 2006) and the LSL-Kras G12D ; p53 R172H/+; Pdx-Cre (KPC) (Hingorani et al., 2005) mouse models, we first confirmed that plasma glucose levels were not changed over the course of the experiment ( Figure 1A). At this rate of glucose infusion, enrichment of 13 C-glucose in plasma was around 40% in non-tumor-bearing mice, and in both KP -/-C and KPC animals ( Figure 1B). Under these conditions, labeling of pyruvate and lactate, as well as the glucose-derived amino acids alanine and serine, from U-13 C-glucose was observed in normal pancreas tissue and in tumors arising in both models ( Figure 1C-F).
Labeling of TCA cycle metabolites was also observed in normal pancreas tissue and in PDAC tumors from both models ( Figure 1G-L, Figure S1A-B). Carbon from glucose-derived pyruvate can contribute to the TCA cycle via reactions catalyzed by pyruvate dehydrogenase (PDH) or pyruvate carboxylase (PC), and the relative use of these routes of TCA cycle labeling has been inferred from the labeling pattern of TCA cycle intermediates from 13 C-labeled glucose (Davidson et al., 2016;Fan et al., 2009;Hensley et al., 2016;Sellers et al., 2015). PDH decarboxylates three-carbon pyruvate to two-carbon acetyl-CoA, and therefore if 13 C-labeled pyruvate is metabolized via PDH, a two carbon (M+2) labeling pattern is observed in TCA cycle metabolites as well as in the amino acid aspartate ( Figure 1G). In contrast, PC carboxylates three-carbon pyruvate to four-carbon oxaloacetate, and therefore if unlabeled CO 2 is added to 13 C-labeled pyruvate via this enzyme, a three carbon (M+3) labeling pattern is observed ( Figure  1G). We observed an increase in both M+2 and M+3 labeling of TCA cycle metabolites and the TCA cycle-derived amino acids aspartate and glutamate in KP -/-C tumor tissue compared to normal pancreas ( Figure 1H-N, Figure S1A-C). Proline was not extensively labeled from glucose in pancreas or PDAC tumor tissue from either model, suggesting that glucose carbon contributes minimally to the synthesis of this amino acid ( Figure   1O). Of note, in KPC mouse PDAC tumors, we observed higher M+2 labeling of only some TCA cycle metabolites compared to normal pancreas, and did not observe an increase in M+3 metabolite labeling compared to pancreas ( Figure 1H-O, Figure S1A-C) illustrating that glucose labels metabolites differently in bulk tumors arising in each model. These data suggest that glucose contributes to labeling of TCA cycle carbon via reactions that involve PDH and PC in PDAC tumors, although the extent of labeling depends on the PDAC model used.
To further study glucose metabolism in PDAC tumor tissue, we infused control and tumor-bearing KP -/-C mice with U-13 C-glucose at a higher rate in an attempt to increase plasma enrichment of labeled glucose (Davidson et al., 2016) ( Figure S1D).
Plasma glucose levels were increased as a result of this higher infusion rate ( Figure   S1E), and resulted in extensive labeling of glycolytic metabolites and glucose-derived amino acids (Figure S1F-I), as well as an increase in M+2 and M+3 labeling of TCA cycle intermediates and related amino acids in pancreatic tumor tissue relative to normal pancreas (Figure S1F-Q). These data further support that glucose carbon can contribute label to TCA cycle intermediates via pathways that involve PDH and PC in PDAC tumors, but the relative contribution varies based on plasma glucose levels and the PDAC model examined.
Despite both being driven by mutant Kras and loss of normal p53 function, differences in the autochthonous KP -/-C and KPC PDAC models are known and have been attributed to differences in tumor latency and p53 status, as well as differences in how stromal cell populations interact with cancer cells to support tumor growth (Rosenfeldt et al., 2013;Vennin et al., 2019). Thus, a difference in relative abundance of cancer and non-cancer cells, or in interactions between non-cancer cells and cancer cells, are one explanation for why differences in glucose labeling are observed across these models. To explore whether the relative abundance of cancer cells in the tumor might affect labeling, we infused mice with pancreatic tumors derived from orthotopic injection of a syngeneic KRas G12D ; p53 -/pancreatic cancer cells derived from tumors arising in the KP -/-C model (Danai et al., 2018), since cell line transplantation models are thought to result in tumors with a less dense, desmoplastic stroma compared to autochthonous models (Baker et al., 2016;Olive et al., 2009). When compared to adjacent normal pancreas, we observed an increase in M+2 and M+3 labeling of TCA metabolites and aspartate in this orthotopic tumor model (Figure S1R-U). Regardless, tumors arising from orthotopic transplantation of murine PDAC cells still contain stroma (Danai et al., 2018). Thus, tumors in all models considered consist of multiple cell types, and cancer cells are known to be a minority cell population in both autochthonous PDAC tumor models. In all cases, metabolite labeling will reflect a weighted average of labeling in all cell types present in the tissue sample and this heterogeneity in cell types in all tissues is a limitation to the use of labeled nutrient infusions to understand the metabolism of cancer cells, or any individual cell population, in tumors or other tissues.

Populations Using Existing Methods
Dissecting the metabolism of individual cell types within a mixed cell population is a barrier to identifying cancer cell-specific liabilities via functional metabolic measurements in tumors such as PDAC. Therefore, we sought to better understand the contribution of cancer cells to the labeling of TCA cycle intermediates from 13 C-glucose in PDAC tumors. We focused on M+3 labeling of TCA cycle intermediates from glucose, reflective of pyruvate carboxylation activity, because this is a metabolic phenotype observed in tumors that is less prominent in cancer cells in culture (Davidson et al., 2016). One existing approach to determine which cell type within the tumor contributes to this activity is to evaluate expression of an enzyme known to catalyze this reaction.
Indeed, immunohistochemistry (IHC) analysis of tumors arising in KP -/-C mice revealed higher PC expression in cancer cells ( Figure 2A). Analysis of a human pancreatic tumor tissue by IHC shows that human tumors exhibit a range of PC expression levels ( Figure   S2A-D), although higher PC expression is observed in cancer cells compared to stroma ( Figure 2B), similar to findings in human lung tumors (Sellers et al., 2015). However, while IHC can be useful to determine relative expression in tumor sections, it does not prove lack of expression by non-cancer cells. Indeed, qPCR analysis of mRNA isolated from sorted cell populations derived from KP -/-C tumors trended toward higher PC expression in cancer cells relative to fibroblasts, although PC mRNA was detected in fibroblasts ( Figure 2C). Furthermore, metabolic fluxes can be more dependent on metabolite concentrations than enzyme expression levels (Hackett et al., 2016).
Therefore, increased enzyme expression may not be reflective of increased activity in tissues, and argues that while suggestive, expression analysis is not a definitive approach to identify which cell population is responsible for M+3 TCA cycle labeling from glucose in pancreatic tumors.
Another approach that has been used to determine which cell type(s) within the tumor contribute to a metabolic activity is isolating distinct cell populations and studying them in culture (Dalin et al., 2019;Francescone et al., 2018;Linares et al., 2017;Sousa et al., 2016;Valencia et al., 2014;Yang et al., 2016). Pancreatic stellate cells (PSCs) are a type of resident fibroblast in the pancreas which can become activated during tumorigenesis and can impact the tumor microenvironment (Bynigeri et al., 2017;Dunér et al., 2011). When PDAC cells or PSCs alone are cultured in the presence of 13 Cglucose in vitro, PSCs exhibit similar or higher M+3 TCA cycle metabolite labeling than cancer cells ( Figure 2D-E), even though the fibroblast cell population exhibited lower PC expression in tumors (Figure 2A-B). These data further highlight the challenges associated with ascribing functional metabolic phenotypes using enzyme expression alone. Nevertheless, isolated cell populations in culture also may not retain the same functional metabolic phenotypes found within tumor tissue where many different cells compete for available nutrients.
To develop new approaches to study the phenotype of individual cell types in a mixed cell population, we first sought to generate a more tractable system that only involves interactions between two different cell types. One approach is to use organoid cultures involving PDAC cancer cells and fibroblasts (Öhlund et al., 2017) where nutrient conditions are modified such that cancer cells rely on the presence of the fibroblasts to proliferate. To do this, we generated pancreatic cancer organoid cultures from KP -/-C and KPC tumors ( Figure S2E) (Boj et al., 2015), and found that when exposed to a more minimal medium than is commonly used (Boj et al., 2015), pancreatic cancer organoid growth becomes dependent on including PSCs in the culture system ( Figure 2F-G, Figure S2F-G). Relevant to the M+3 labeling of TCA cycleassociated metabolites from glucose observed in pancreatic tumors, when sorted from this co-culture organoid system, both cancer cells and PSCs expressed higher levels of PC mRNA compared to PSCs and pancreatic cancer cells in standard monoculture ( Figure 2H). In addition, when U-13 C-glucose is provided to whole organoid co-cultures comprised of both cancer cells and PSCs, and TCA cycle intermediate labeling is assessed after rapid quenching and extraction of metabolites, M+3 labeling of aspartate and malate was observed ( Figure 2I-J, Figure S2H-K). These data argue that this organoid co-culture system may provide a model to explore the relative contribution of each cell population to the pyruvate carboxylation phenotype observed when both cell types are present.

Extracellular Nutrients
To dissect PDAC cancer cell versus other cell type contributions to specific metabolic activities, we reasoned it would be necessary to isolate each cell type for analysis after exposure to labeled nutrients. To experimentally evaluate the effect of sorting cells from the organoid co-culture system on metabolite levels and labeling from glucose, we cultured AL1376 murine PDAC cells in U-13 C-glucose and incubated the cells on ice for various lengths of time to simulate conditions the cells would experience during separation by flow cytometry (up to 240 minutes) or other antibody based methods, which require a minimum of 10-12 minutes (Abu-Remaileh et al., 2017;Chen et al., 2016). Metabolite levels and labeling were then measured over time using mass spectrometry, allowing comparison to that observed when metabolism is rapidly quenched (the zero time point). Consistent with the known rapid turnover of metabolites (Shamir et al., 2016), both metabolite levels ( Figure 3A-D, Figure S3A-C) and labeling from U-13 C-glucose ( Figure 3E-H, Figure S3D-F) changed over the time required to separate cells using antibodies and/or flow cytometry. These changes, especially the larger changes in levels, indicate that metabolism is not at steady-state when metabolites are analyzed in sorted cell populations, and complicates the interpretation of differential isotope labeling patterns (Buescher et al., 2015). In fact, changes in metabolite levels and labeling may be even greater when using flow cytometry to sort cells in practice because temperature as well as factors such as mechanical stress are less easily controlled (Binek et al., 2019). Thus, assessment of M+3 labeling of TCA cycle intermediates in sorted cell populations from organoids or tumors may not fully portray the contribution of each cell type to the pyruvate carboxylation phenotype observed when material containing multiple cell types is analyzed.
The turnover of protein and nucleic acid is slow relative to metabolites (Shamir et al., 2016), which allows gene expression and proteomic analysis in separated cell types to better reflect the state of cells within a mixed population. Because metabolites contribute to protein, lipid, and nucleic acid biomass, and isotope-labeled nutrients can be traced into this biomass (Gebreselassie and Antoniewicz, 2015;Ghosh et al., 2014;Rühl et al., 2011;Zamboni et al., 2005;Green et al., 2016;Hosios et al., 2016;Le et al., 2017;Lewis et al., 2014;Mayers et al., 2016;Shankaran et al., 2016), we reasoned that 13 C-labeling patterns in biomass might be used to infer the contribution of glucose to different metabolic pathways within a mixed cell population relevant to pancreatic cancer. We confirmed that glucose labeling of protein was stable over the time period needed to sort cells by flow cytometry (Figure 3I-J). We also confirmed that amino acids from protein hydrolysates were detectable in sorted cells from murine PDAC tumors, and were within the linear range of detection by GC-MS even when low cell numbers of less abundant cell populations were recovered from tumors ( Figure S3G-L). Therefore, examining 13 C label in amino acids from hydrolyzed protein may be informative of the labeling of free amino acids in tumor cell subpopulations that existed prior to sorting the cells.

Fibroblasts in PDAC Models
To facilitate sorting of PSCs and PDAC cancer cells from organoid co-cultures and tumors, a LSL-tdTomato reporter allele was bred to the KP -/-C and KPC PDAC models as a source of tdTomato+ cancer cells for both organoid and tumor models.
PSCs were isolated from pancreata from mice bearing a β-actin-GFP allele to enable sorting of GFP + PSCs for labeling of the PSC population in the organoid co-culture model ( Figure 4A). To determine the relative contribution of 13 C-glucose to M+3-labeled aspartate in cancer cells and PSCs in the organoid-fibroblast co-culture model, we exposed organoid co-cultures containing tdTomato+ cancer cells and GFP + PSCs to U-protein for amino acid analysis from each cell population. Over time, similar M+2 protein aspartate labeling was observed between the two cell types, while higher M+3 aspartate labeling was observed in cancer cells as compared to PSCs, suggesting that while the two cell types have similar labeling via reactions involving PDH, the cancer cells appear to have higher pyruvate carboxylation activity ( Figure 4B-C). This higher M+3 level was also reflected in the other TCA cycle-derived amino acids glutamate and proline ( Figure   4D-G), whereas labeling of the glucose-derived amino acids alanine and serine was not higher in cancer cells ( Figure S4A-B). We also exposed organoid co-cultures to U-13 Cglutamine over four days and traced the fate of labeled carbon into protein in each cell population. Of note, we observed slightly higher labeling of aspartate from glutamine in protein in PSCs, matching the lower fractional labeling we observed from glucose ( Figure S4C). We did not observe other appreciable differences in fractional labeling of glutamate or proline from glutamine in protein between cancer cells and PSCs ( Figure   S4D-E). Taken together, these data suggest a differential fate for glucose in these cell types, with increased M+3 labeling of aspartate from glucose carbon in cancer cells relative to PSCs.
Because the labeling of amino acids in protein is unlikely to reach steady-state even after multiple days of labeling, one explanation for the difference in aspartate labeling from labeled glucose in the cancer cells relative to the PSCs is a higher rate of protein synthesis in the cancer cells, although this is unlikely to differentially affect only M+3 labeled species. Nevertheless, to examine this possibility, we measured protein synthesis rates in each cell type using a fluorescent protein synthesis reporter in which BFP is fused to an unstable E. coli dihydrofolate reductase (DHFR) domain. Upon addition of the DHFR active site ligand trimethoprim (TMP), the reporter is stabilized and the rate of fluorescence accumulation reflects the synthesis rate of the fluorescent protein (Han et al., 2014). Consistent with previous reports, this reporter produced similar results compared to an assessment of protein synthesis through incorporation rates of the aminoacyl tRNA analog puromycin (Darnell et al., 2018) when BFP accumulation after TMP addition was assayed over time in PDAC cancer cell and PSC mono-cultures and compared to cells with no TMP added as a negative control ( Figure   S4F-G). The BFP reporter is suitable for use in sorted cells from a mixed cell system, and thus was used to assess protein synthesis rates in cancer cells and PSCs in organoid co-cultures. Interestingly, even though cancer cells and PSCs exhibited a similar protein synthesis rate in monoculture ( Figure S4F-G), accumulation of BFP fluorescence was slower in cancer cells compared to PSCs in 3D co-cultures ( Figure   4H). This argues that protein synthesis rates are higher in PSCs in organoid co-cultures, and that the higher M+3 aspartate labeling observed in cancer cells cannot be explained by a higher rate of protein synthesis in the cancer cells in this co-culture system. M+3 labeling from U-13 C-glucose is often used as a surrogate for pyruvate carboxylation activity, but can also occur from multiple rounds of TCA cycling (Alves et al., 2015). To more directly assess pyruvate carboxylation activity, we traced 1-13 Cpyruvate or 3,4-13 C-glucose fate in organoid-PSC co-cultures. 1-13 C-pyruvate or 3,4-13 C-glucose can only label aspartate via pyruvate carboxylation, because the 13 C-label is lost as carbon dioxide if pyruvate is metabolized to acetyl-coA via PDH prior to entering the TCA cycle ( Figure 4I). Compared to U-13 C-glucose labeling, a greater difference and higher M+1 aspartate labeling in protein was observed using 1-13 C-pyruvate or 3,4-13 C-glucose in sorted cancer cells compared to PSCs from organoid-PSC co-cultures, further supporting that pyruvate carboxylation activity is higher in these cells ( Figure 4J-K, Figure S4H). The ratio of M+1 aspartate to M+1 pyruvate derived from 1-13 C-pyruvate or 3,4-13 C-glucose has been used as a way to approximate pyruvate carboxylation activity (Davidson et al., 2016). To approximate pyruvate carboxylation activity using protein hydrolysates, we normalized M+1 aspartate to M+1 alanine as a surrogate for pyruvate labeling. Of note, the ratio of M+1 aspartate to M+1 alanine was also higher in cancer cells ( Figure 4L-M, Figure S4I). Taken together, these data argue that cancer cells within PDAC organoid-PSC co-cultures have higher pyruvate carboxylation activity than PSCs.
To investigate whether PDAC cancer cells also exhibit higher pyruvate carboxylation activity in tumors in vivo, we first verified that the tdTomato fluorescence in tumors arising in KP -/-C mice bearing a LSL-TdTomato allele did not co-localize with staining for the fibroblast-specific marker alpha-smooth muscle actin (α-SMA) ( Figure   5A), but did co-localize with Cytokeratin 19 (CK19), a marker of pancreatic cancer cells ( Figure 5B). This verifies that tdTomato labeling can be used to isolate cancer cells from α-SMA-positive fibroblasts, and a combination of tdTomato fluorescence and an antibody for the pan-hematopoietic marker CD45 allowed efficient sorting of cancer cells, fibroblasts, and hematopoietic cells as verified by expression of relevant mRNAs using qPCR ( Figure 5C-F, Figure S5A-B). To label protein in PDAC tumors in vivo, autochthonous tumor-bearing mice were infused with U-13 C-glucose for 24 hours ( Figure S5C-D), with aspartate labeling observed in protein hydrolysates from bulk tumors in this time frame ( Figure 5G). Cell populations were sorted from tumors, and labeling of amino acids was determined in protein hydrolysates from each cell population as well as from protein obtained from the bulk digested tumor (unsorted). In agreement with labeling patterns from organoid-co-cultures, tdTomato+ cancer cells from PDAC tumors in mice had the highest M+3 protein aspartate labeling in protein, as well as higher M+2 aspartate labeling ( Figure 5H-I). This labeling pattern was also reflected in higher M+2 and M+3 labeling in glutamate but not in other glucose-labeled amino acids in protein ( Figure S5E-J). Taken together, these data are consistent with increased pyruvate carboxylation, as well as increased glucose oxidation via PDH, in the cancer cells relative to the stromal cell populations analyzed in PDAC tumors in vivo.

Growth In Vivo
To test whether PC is responsible for the observed pyruvate carboxylation activity and is functionally important for cells to proliferate in organoid co-cultures and tumors, PC expression was disrupted in murine PDAC cell lines, organoids, and PSCs using CRISPR/Cas9. First, CRISPRi was used to knock down PC expression in a PDAC cancer cell line derived from KP -/-C mice ( Figure S6A). As expected, PC knockdown in these PDAC cells resulted in a decrease in aspartate labeling from 1-13 Cpyruvate and relative pyruvate carboxylation activity compared to control cells as assessed by the ratio of M+1 labeled aspartate to M+1 labeled pyruvate ( Figure S6B-C), but did not affect proliferation in culture ( Figure S6D). However, knockdown of PC in PDAC organoids reduced growth of these cells in organoid-PSC co-cultures ( Figure  S6E-G). PC expression level and aspartate labeling from 1-13 C-pyruvate were increased by exogenous PC expression in PDAC PC knockdown cells ( Figure S6H-J). When transplanted subcutaneously, PDAC cell lines with PC knockdown formed tumors that grew similarly to control cells ( Figure S6K); however, PC expression was similar or increased in the tumors formed from PC knockdown cells compared to control tumors ( Figure S6L). These data suggest that over time, cells that grew into tumors were selected for reversal of PC knockdown and that PC is required for tumor growth in vivo even though it is dispensable in culture, as has been observed previously in lung cancer (Davidson et al., 2016;Fan et al., 2009;Sellers et al., 2015).
To further test the requirement for PC in PDAC tumors, we generated cancer cell clones with complete CRISPR/Cas9 disruption of PC expression ( Figure S6M-N).
Similar to knockdown experiments, loss of PC had no effect on proliferation of PDAC cells in culture ( Figure 6A), whereas loss of PC reduced the growth of organoid cocultures ( Figure 6B-C). CRISPR/Cas9 was also used to knockout PC in PSCs, and despite loss of PC expression and reduced pyruvate carboxylation activity ( Figure S6O-Q), PC knockout PSCs retained the ability to enhance PDAC organoid growth or growth of PDAC cancer cells as tumors in subcutaneous transplants, although the effect was reduced compared to sgControl PSCs ( Figure S6R-T). Consistent with a requirement for PC expression in cancer cells to form PDAC tumors, PC-null cancer cells did not form tumors when transplanted into syngeneic mice subcutaneously or orthotopically ( Figure 6D-E). However, surprisingly, PC-null cancer cells still displayed M+1 aspartate labeling from 1-13 C-pyruvate with similar or only a slight decrease in pyruvate carboxylation activity compared to control cells ( Figure 6F-G). Taken together, these data argue that loss of PC in cancer cells can impact tumor growth, but another enzyme must also contribute to pyruvate carboxylation activity in these cells.

Activity in PDAC Cells and is Important for Tumor Growth In Vivo
A candidate for the pyruvate carboxylation activity observed in PC-null cells is malic enzyme, since this enzyme catalyzes the interconversion of pyruvate and CO 2 with malate, another 4-carbon TCA cycle intermediate. Malic enzyme is typically assumed to catalyze malate decarboxylation as a source of NADPH in cells (Cairns et al., 2011;Hosios and Vander Heiden, 2018), but has previously been shown to be reversible and produce malate from pyruvate and CO 2 in purified enzyme assays (Ochoa et al., 1947;1948). Thus, we tested whether malic enzyme activity could sustain M+1 labeling of aspartate in PDAC cancer cells lacking PC by using CRISPR/Cas9 to knock out malic enzyme 1 (ME1). After knockout of both PC and ME1 in PDAC cell lines, aspartate labeling from 1-13 C-pyruvate is virtually abolished, suggesting that ME1 activity can contribute to pyruvate carboxylation activity in these cells ( Figure 7A-B, Figure S7A). This aspartate labeling was also increased after exogenous expression of ME1 in PC and ME1 double knockout cells ( Figure 7A-B). We used CRISPR/Cas9 to knockout or knockdown both PC and ME1 in organoids, which also resulted in decreased M+1 aspartate labeling from 1-13 C-pyruvate and decreased pyruvate carboxylation activity ( Figure S7B-D). Interestingly, ME1 expression in KP -/-C mouse and human PDAC tumors and organoids mimics that of PC in that it is more highly expressed in cancer cells compared to stroma, suggesting that ME1 could also contribute to the higher pyruvate carboxylation seen in cancer cells compared to PSCs ( Figure 7C-E, Figure S7E-I).
We next assessed whether ME1 was essential for tumor and organoid growth using CRISPR/Cas9 to knock down or knock out ME1. Similar to loss of PC, loss of ME1 had minimal effect on proliferation of cancer cells in monoculture ( Figure 7F, Figure S7J), but reduced growth of organoid co-cultures compared to controls ( Figure   7G-H, Figure S7K). Transplantation of ME1-null cancer cells in vivo also resulted in reduced tumor growth ( Figure 7I), consistent with published data (Son et al., 2013).
Taken together, these data argue that both PC and ME1 are important enzymes for PDAC cancer cells in tumors and contribute to the pyruvate carboxylation activity observed in pancreatic cancer.

Discussion
Metabolism can differ between cancer cells in culture and tumors, and understanding how nutrients are used by cancer cells in vivo has been an area of interest for developing cancer therapies. Tumor metabolic phenotypes have been assumed to reflect the metabolism of cancer cells within a tumor; however, in many tumors such as in PDAC, cancer cells are a minority cell population. Metabolic interactions between cell types have been described in normal tissues (Bélanger et al., 2011), and some metabolic phenotypes observed in cancer cells such as increased glucose utilization are also prominent in other cell types including fibroblasts and immune cells that can be abundant in some tumors (Lemons et al., 2010;Vincent et al., 2008;Zhao et al., 2019). Therefore, methods to deconvolute which cell types in a tumor are responsible for observed tissue metabolic phenotypes are needed.
We find that pancreatic tumors exhibit evidence of glucose metabolism, with carboxylation of glucose-derived pyruvate being more active in cancer cells than in other tumor cell types. However, because glucose will label both pyruvate and lactate, and these nutrients can be exchanged between cell types, it cannot be concluded that the cancer cells necessarily derive TCA cycle metabolites directly from glucose in a cell autonomous manner. In fact, rapid exchange of labeled intracellular and extracellular pyruvate and lactate among cell types is likely, making it difficult to address the original cellular source of labeled TCA metabolites with these methods. Thus, while this approach could be used to understand differential pathway use between cell types, in many cases it will not be able to distinguish the exact source of carbon that labels metabolites in individual cells.
Another important caveat to interpreting labeling patterns in protein or other macromolecules in cells within tissues is that labeling is unlikely to reach steady state, particularly for analysis of cells in tissues in vivo. This failure to reach steady state means that differences in label delivery or uptake could cause differences in biomass labeling even when the pathway involved in labeling is similarly active in both cell types.
Thus, controlling for variables such as biomass synthesis rates between cell types can help with data interpretation. The ability to reach a pseudo-metabolic steady state facilitates interpretation of labeling data, however this requires that both circulating nutrient levels and labeling patterns are relatively constant (Buescher et al., 2015;Jang et al., 2018). Glucose infusion rates and techniques can vary across studies (Davidson et al., 2016;Faubert et al., 2017;Hui et al., 2017;Ma et al., 2019;Marin-Valencia et al., 2012), and these differences may affect whether circulating nutrient levels are constant.
While this may be one reason why differences in labeling were observed across PDAC models evaluated in this study, additional factors such as tumor initiation and growth rates, cells of origin, p53 status, and different composition of cancer and stromal cells are known to exist as well (Rosenfeldt et al., 2013;Vennin et al., 2019).
We find that whether cells are grown in 2D cultures, in 3D organoid co-culture with PSCs, or in autochthonous tumors in vivo impacts whether pyruvate carboxylation is important for proliferation, with the organoid and autochthonous models showing a similar dependency on this activity. Tumor organoid-stromal co-cultures represent a tractable model for metabolic characterization, and thus may be useful for exploration of other symbiotic metabolic relationships between pancreatic cancer cells and fibroblasts.
However, while the difference in M+3 aspartate labeling seen in vivo was recapitulated by organoid-fibroblast co-cultures, other differences such as higher M+2 aspartate and glutamate labeling observed in vivo were not observed in the co-culture model. Therefore, some aspects of cell type-specific metabolism are not recapitulated even in co-culture organoid systems.
We find that PC and ME1 expression in cancer cells are both important for PDAC tumor growth in vivo. A dependence on pyruvate carboxylation seems to be a characteristic of both PDAC and lung tumors in vivo that is not prominent in standard cell culture systems (Christen et al., 2016;Davidson et al., 2016;Fan et al., 2009;Hensley et al., 2016;Sellers et al., 2015). Why this is the case is not known, but PC is an important anaplerotic pathway for the TCA cycle, contributing to biosynthesis of macromolecules such as protein, nucleotides, and lipids in cancer cells. Glucose metabolism and increased glucose uptake have been shown to be important for biosynthesis in PDAC tumors (Santana-Codina et al., 2018;Ying et al., 2012), but it has also been suggested that some PDAC tumors rely less on glucose for fuel and instead Glutamine is also a source of TCA anaplerotic carbon that may contribute to biosynthesis differentially between cancer cells and stroma. Previous work has suggested that utilization of ME1 to produce pyruvate from glutamine can be important for PDAC cells to maintain redox balance, specifically via NADPH generation (Son et al., 2013), and that glutamine can be a major contributor to TCA metabolites in PDAC tumors (Hui et al., 2017). A potential role for malic enzyme in pyruvate carboxylation suggests use of this enzyme to produce malate could be another pathway for TCA cycle anaplerosis. Of note, this reaction would require NADPH, and may be more favored in cancer cells that exhibit a reduced redox state (Hosios and Vander Heiden, 2018). We also considered phosphoenolpyruvate carboxykinase (PEPCK) or malic enzymes 2 and 3 as possible contributors to pyruvate carboxylation activity, although these reactions are less energetically favorable in the reverse direction in comparison to malic enzyme 1; malic enzyme 1 is cytosolic, which is thought to be a more reducing environment than the mitochondria where malic enzymes 2 and 3 are localized (Hu et al., 2008). We did not see evidence for differential glutamine utilization in our organoid-PSC co-cultures, and PDAC tumors are resistant to glutaminase inhibitors (Biancur et al., 2017), but further work is needed to assess how glutamine metabolism and other anaplerotic pathways might be differentially active in cancer cells and non-cancer cells in PDAC tumors.
In pancreatic β-cells, PC and ME are thought to be part of a coordinated metabolic cycle that regulates insulin secretion (Pongratz et al., 2007). In this pyruvate cycle, ME1 generates NADPH and produces pyruvate from malate in the cytosol, which can then be used by PC to generate oxaloacetate in the mitochondria (Pongratz et al., 2007). While loss of ME activity might be expected to impact pyruvate carboxylation activity when both enzymes are present, the fact that residual pyruvate carboxylation activity is observed in the absence of PC, and that this is lost upon ME1 disruption argues that PC and ME may have redundant metabolic functions under some conditions, such as those found in pancreatic cancer cells.
PC has been targeted with antisense oligonucleotides (Kumashiro et al., 2013) and relatively non-specific chemical inhibitors (Bahl et al., 1997;Zeczycki et al., 2010); however, inhibiting PC may have deleterious effects on whole body metabolism by interfering with gluconeogenesis or glucose-stimulated insulin secretion. Whether malic enzyme can compensate sufficiently for PC inhibition in those tissues to allow therapeutic targeting, or if malic enzyme is a viable alternative target remains to be determined. Nevertheless, our data suggest that stable isotope tracing into macromolecules can be utilized to deconvolute complex tracing patterns in mammalian tissues to identify increased pathway activity in a particular cell type. Understanding the metabolic similarities and differences between cancer cells and stroma within PDAC and other tumors will be important in further delineating cancer-specific dependencies.

Materials and Methods
Mouse models:

Glucose Infusion:
Infusion of U-13 C-glucose (Cambridge Isotope Laboratories) was performed as previously described (Davidson et al., 2016). Surgery was performed to implant a catheter into the jugular vein of animals 3-4 days prior to infusion. For 4-6 hour infusions, mice were fasted for 4 hours prior to beginning the infusion. For 24-hour infusions, mice were not fasted prior to infusion. Infusions were performed in conscious, free-moving animals for 4 or 24 hours at a rate of 30mg/kg/min. For 6 hour infusions, each animal, regardless of body weight, was infused with a fixed volume of 300 µl of a 500 mg/ml glucose solution over 6 hours, which is an infusion rate of 0.4 mg/min Tumors were either digested for FACS or rapidly frozen using a Biosqueezer (BioSpec Products) and stored at -80°C prior to metabolite extraction.

Isotope Labeling Experiments:
100,000 adherent cells were plated in 6 well plates, or organoids and organoid-PSC cocultures were plated on plastic coverslips (Thermo 174985) in 24-well plates. The following day, the cells were washed three times with PBS and then isotope-labeled media was added for the specified length of time (24-72 hours). For U-13 C-glucose or 3,4-13 C-glucose tracing, DMEM without glucose and pyruvate was used, supplemented with 25mM U-13 C-glucose or 3,4-13 C-glucose, 10% dialyzed FBS, and penicillinstreptomycin. For 1-13 C-pyruvate tracing, DMEM with glucose and without pyruvate was used, adding 2mM 1-13 C-pyruvate and supplementing with 10% dialyzed FBS and penicillin-streptomycin.

Polar Metabolite Extraction:
Adherent cells were washed once with ice-cold saline on ice and then extracted with a 5:3:5 ratio of ice-cold HPLC-grade methanol:water:chloroform. Mouse tissue or coverslips containing organoids and organoid-PSC co-cultures were washed once with saline prior to extraction. Tissue or matrigel domes containing the organoids and organoid-PSC co-cultures were then rapidly frozen using a Biosqueezer (BioSpec Products) and stored at -80°C prior to metabolite extraction. Snap frozen tissues or organoids were extracted with a 5:3:5 ratio of ice-cold HPLC-grade methanol:water:chloroform. For mouse plasma, 10µL of plasma was extracted with 600µL ice-cold methanol. All samples were vortexed for 10 minutes at 4°C followed by centrifugation for 5 minutes at maximum speed on a tabletop centrifuge (~21,000xg) at 4°C. An equal volume of the aqueous phase of each sample was then dried under nitrogen gas and frozen at -80°C until analysis. For organoid samples, 2 rounds of extraction were done to eliminate excess protein from matrigel.

Protein Hydrolysis:
Acid hydrolysis of protein was performed as described previously (Mayers et al., 2016;Sullivan et al., 2018). Frozen tissue or cell pellets were boiled for 24 hours at 100°C in 500µL (cell pellets) -1mL (tissue) 6M HCl for amino acid analysis (Sigma 84429). 50µL (tissue) -100 µL (cell pellets) of HCl solution was then dried under nitrogen gas while heating at 80°C. Dried hydrolysates were stored at -80°C until derivatization.
Helium was used as the carrier gas at a flow rate of 1.2 mL/min. One microliter of sample was injected at 270°C. After injection, the GC oven was held at 100°C for 1 min. and increased to 300°C at 3.5 °C/min. The oven was then ramped to 320°C at 20 °C/min. and held for 5 min. at this 320°C. The MS system operated under electron impact ionization at 70 eV and the MS source and quadrupole were held at 230°C and 150°C, respectively. The detector was used in scanning mode, and the scanned ion range was 100-650 m/z. Data were corrected for natural isotope abundance.

Adherent Cell Culture:
Cell lines were cultured in DMEM (Corning 10-013-CV) supplemented with 10% fetal bovine serum and penicillin-streptomycin. PSCs were isolated from β-actin-GFP mice (006567) as previously described (Apte, 2011;Danai et al., 2018): 3mL of 1.3mg/mL cold collagenase P (Sigma 11213865001) and 0.01mg/mL DNAse (Sigma D5025) in GBSS (Sigma G9779) were injected into the pancreas. The tissue was then placed into 2mL of collagenase P solution on ice. Cells were then placed in a 37°C water bath for 15 minutes. The digested pancreas was filtered through a 250µm strainer and washed with GBSS with 0.3% BSA. A gradient was created by resuspending the cells in Nycodenz (VWR 100356-726) and layering in GBSS with 0.3% BSA. Cells were then centrifuged at 1300g for 20 minutes at 4°C. The layer containing PSCs was removed, filtered through a 70µm strainer, washed in GBSS with 0.3% BSA, and plated for cell culture in DMEM with 10% FBS and penicillin-streptomycin. PSCs were immortalized with TERT and SV40 largeT after several passages.

Organoid Culture:
Organoids were isolated from mice bearing PDAC tumors and cultured as previously , and 1µg/mL R-spondin. R-spondin was purified from 293T cells engineered to produce it using a Protein A Antibody Purification Kit (Sigma PURE1A). Organoids were grown in complete media when passaging. For organoid-PSC co-culture experiments, co-cultures were grown in DMEM without pyruvate (Corning 10-017-CV) supplemented with 10% dialyzed FBS and penicillin-streptomycin.
Organoids were digested to single cells by incubating with 2 mg/mL dispase in Advanced DMEM/F12 with penicillin-streptomycin, HEPES, and GlutaMAX at 37°C for 20 minutes. Organoids were then triturated with a fire-polished glass pipette and enzymatically digested with 1mL TrypLE Express (Thermo Fisher 12605-010) for 10 minutes rotating at 37°C, followed by addition of 1mL of dispase containing media and 10µL of 10mg/mL DNAse (Sigma) and digested rotating at 37°C for 20 minutes or until single cells were visible under a microscope. Cells were counted and plated in GFR matrigel at a concentration of 2000 cells/well.

Proliferation Assays:
50,000 cells were seeded in 6-well plates in 2mL DMEM with 10%FBS and penicillinstreptomycin. The next day, cells were counted for day 0 and media was changed on remaining cells. 8mL of media was added and cells were left to proliferate for three days. On day three, cells were trypsinized and counted. Alternatively, proliferation was measured using sulforhodamine B staining as previously described (Vichai and Kirtikara, 2006). Cells were fixed on day 0 and day 3 with 500 µl of 10% trichloroacetic acid (Sigma T9159) in 1 mL media and incubated at 4 o C for at least one hour. Plates were washed under running water and cells were stained with 1mL sulforhodamine B (Sigma 230162) and incubated at room temperature for 30 minutes. Dye was removed and cells were washed 3 times with 1% acetic acid. Plates were then dried and 1mL of 10mM Tris pH 10.5 was added to each well to solubilize the dye. 100uL of each sample was then transferred to a 96 well plate and absorbance was measured at 510nm on a microplate reader.

Protein Synthesis Assays:
A fluorescent reporter in which BFP is fused to an unstable E. coli dihydrofolate reductase (DHFR) degron domain which is stabilized by trimethoprim (Han et al., 2014) was used to determine global protein synthesis rate as previously described (Darnell et al., 2018). Briefly, PDAC / PSC cell lines and PDAC organoids expressing the reporter were generated by lentiviral transduction and puromycin selection followed by flow cytometry-based sorting for populations that were BFP-positive after TMP addition for 24-48 hours. For each experiment, the reporter protein was stabilized upon addition of 10 uM trimethoprim (TMP) and fluorescence accumulation was measured in cells or organoids by flow cytometry over several time points within 12 hours of TMP addition.
Data were normalized to no TMP controls. Puromycin incorporation assays were performed as previously described (Schmidt et al., 2009). 10ug/mL puromycin was spiked into the medium of cells grown in 6cm plates. Plates were kept at 37°C for indicated pulse times (spanning 2.5 to 20 minutes). As a negative control, 100ug/mL cycloheximide was added to a plate of cells for 45 minutes before the addition of puromycin. At the end of the pulse, plates were washed once with ice cold PBS on ice and flash frozen in liquid nitrogen. Cells were harvested from frozen plates by scraping into RIPA buffer containing cOmplete Mini EDTA-free Protease Inhibitor Cocktail (Roche 11836170001) and PhosSTOP Phosphatase Inhibitor Cocktail Tablets (Roche 04906845001) and protein concentration was quantified using the Pierce BCA Protein Assay Kit (Pierce 23225). 2uL of lysate (approximately 2ug) was spotted directly onto 0.2 um nitrocellulose membranes and blotted with primary antibodies against puromycin (Sigma MABE343 1:25,000 dilution) and vinculin (Abcam ab18058, 1:1000 dilution) as a control.

Generation of PC and ME1 knockdown and knockout cells:
CRISPRi knockdown cell lines of PC and ME1 were generated by transfecting cells expressing modified dCas9-KRAB fusion protein, as previously described (Horlbeck et al., 2016). The target sequences used for PC sgRNAs were (PC1-GCGGCGGCCACGGCTAGAGG, PC2-GTGGAGGCAGGGGCCGTCAG), the sequence for non-targeting control was GCGACTAGCGCCATGAGCGG, and the target sequence of ME1 sgRNA was GCCGCAGTGGCCTCCCGGGT. After transfection, cells were selected under 5ug/ml puromycin. Rescue of CRISPRi knockdown cell lines of PC was performed by re-expressing the cDNA of the rescued gene under a CMV promoter using a custom lentiviral construct generated on VectorBuilder and selected in 500ug/ml blasticidin. CRISPR knockout cell lines for PC and ME1 were generated using the LentiCRISPRv2 system, as previously described (Sanjana et al., 2014), with guides against the target sequence 5' CGGCATGCGGGTCGTGCATA 3' for PC and 5' GTTTGGCATTCCGGAAGCCA 3' for ME1. After transfection, cells were selected under 5ug/ml of puromycin, single-cell cloned, and knockout validation performed using western blot. For organoids, the same vector systems and guide sequences were used.
Organoids were transfected with concentrated virus by spinfection for 45 minutes at room temperature. For CRISPR knockout organoids, organoid cultures were selected under 5ug/ml of puromycin, digested to single cells, and then single organoids were picked, expanded, and validated using western blot.
Double knockout cell lines for PC and ME1 were generated using pUSPmNG (U6 sgRNA PGK with mNeonGreen) incorporated into cells via electroporation (Amaxa VPI-1005), and selected by FACS using NeonGreen expression. For organoids, double knockout organoids were generated using the LentiCRISPRv2 system and spinfected for 45 minutes at room temperature. After transfection, cells were selected under 500ug/ml of blasticidin, digested to single cells, and then single organoids were picked, expanded, and validated using western blot.

Flow Cytometry:
Tumors were dissected, minced, and digested rotating for 30 minutes at 37°C with 1 mg/mL Collagenase I (Worthington Biochemical LS004194), 3mg/mL Dispase II (Roche 04942078001), and 0.1mg/mL DNase I (Sigma D4527) in PBS. Following digestion, cells were incubated with EDTA to 10mM at room temperature for 5 min. Cells were then filtered through a 70µm strainer and washed twice with PBS. Single cell suspensions were resuspended in flow cytometry staining buffer (Thermo Fisher 00-4222-57) and first stained with 10µL of CD16/CD32 monoclonal antibody (Thermo Fisher 14-0161-82) for 15 minutes to block Fc receptors and then stained using with antibodies to CD45-APC-Cy7 (BD 557659) and SYTOX Red Dead Cell Stain (Life Technologies S34859) to visualize dead cells. All antibodies were incubated for 15-20 minutes on ice. Cell sorting was performed with a BD FACS Aria and data was analyzed with FlowJo Software (BD).

Immunohistochemistry:
Sections from formalin fixed paraffin embedded mouse tissue or a human pancreatic cancer tissue microarray (Biomax PA961e) were stained with antibodies against PC

Quantification and Statistical Analysis:
GraphPad Prism software was used for statistical analysis. All statistical information is      with 1-13 C-pyruvate. The difference in aspartate M+1 labeling was significant (p=0.0020) using an unpaired student's t test. Mean +/-SD is shown. (K) Fractional labeling of aspartate M+1 from protein hydrolysates following three days of KP -/-CT organoid-PSC co-culture with 3,4-13 C-glucose. The difference in aspartate M+1 labeling was significant (p=0.0007) using an unpaired student's t test. Mean +/-SD is shown. (L) Aspartate M+1 isotopomer labeling was normalized to alanine M+1 labeling as a surrogate for pyruvate carboxylation activity from protein hydrolysates obtained from KP -/-CT organoid-PSC cocultures exposed for three days to 1-13 C-pyruvate. The difference in pyruvate carboxylation activity was significant (p=0.0016) using an unpaired student's t test.
Mean +/-SD is shown. (M) Aspartate M+1 isotopomer labeling was normalized to alanine M+1 labeling as a surrogate for pyruvate carboxylation activity from protein hydrolysates obtained from KP -/-CT organoid-PSC co-cultures exposed for three days to 3,4-13 C-glucose. The difference in pyruvate carboxylation activity was significant (p=0.0001) using an unpaired student's t test. Mean +/-SD is shown.   in DMEM-pyruvate with 10% dialyzed FBS alone or with murine PSCs. sgControl organoids with PSCs trended towards higher tdTomato fluorescence compared to sgME1 organoids with PSCs (p=0.0579) but was not significant based on an unpaired, two-tailed student's t-test. Mean +/-SD is shown. The sgControl data are also shown in     labeling was significant (p=0.0004) using an unpaired student's t test. Mean +/-SD is shown. (I) Aspartate M+1 isotopomer labeling was normalized to alanine M+1 labeling as a surrogate for pyruvate carboxylation activity from protein hydrolysates following three days of 1-13 C-pyruvate tracing in organoid-PSC co-cultures using murine PDAC organoids cultured from KPCT mice. The difference in pyruvate carboxylation activity was significant (p=0.0007) using an unpaired student's t test. Mean +/-SD is shown.