Macropinocytosis in Cancer-Associated Fibroblasts Is Dependent on CaMKK2/ARHGEF2 Signaling and Functions to Support Tumor and Stromal Cell Fitness

Macropinocytosis is a form of endocytosis that mediates nonselective fluid-phase uptake. In cancer, macropinocytosis functions as a nutrient acquisition pathway by mediating the uptake of extracellular proteins and other nutritious cargo, which are transported to lysosomes and degraded by acid hydrolases, releasing their constituents (1–4). In pancreatic ductal adenocarcinoma (PDAC) cells, the macropinocytosis of extracellular albumin generates protein-derived amino acids that support metabolic fitness by contributing to the intracellular amino acid pools and to the biosynthesis of central carbon metabolites (1). Because the PDAC tumor microenvironment is starved of many amino acids, macropinocytosis represents a survival strategy that PDAC cells use to circumvent the stresses of a nutrient-scarce tumor milieu (2). Most PDAC tumors harbor oncogenic mutations in KRAS, and it is thought that through activation of RAC1, RAS signaling drives actin-dependent plasma membrane protrusions known as membrane ruffles (5, 6). When these membrane ruffles fuse with each other or back-fuse with the plasma membrane, they form macropinosomes that encapsulate the surrounding fluid and its constituents. Macropinosome labeling in live PDAC tumor tissue has demonstrated that macropinocytosis is a prevalent feature of tumor cells found in murine autochthonous tumors, xenograft tumors, and human tumors (1, 2, 7); however, it is not clear whether macropinocytosis is employed by other cell types found within the tumor.

The most depleted amino acid in human PDAC tumors is glutamine (2). Glutamine is critical to the synthesis of macromolecules, such as lipids and proteins, and plays an instrumental role in de novo nucleotide production, hexosamine biosynthesis, the urea cycle, and glutathione production (8). Glutamine also serves as a nitrogen and/or carbon source for the biosynthesis of other amino acids, including glutamate, proline, aspartate, alanine, threonine, glycine, serine, and cysteine (8). A recent finding that underscores the importance of glutamine in PDAC is that in PDAC tumors, glutamine contributes the most to tricarboxylic acid cycle metabolites relative to other nutrient sources (9). This may be a unique feature of PDAC because lung tumors use glutamine to a much lesser degree and in vitro studies suggest that lung tumor cells derive biomass mostly from the catabolism of other amino acids (9, 10). There is recent evidence indicating that the availability of glutamine can modulate the extent of macropinocytosis in PDAC cells. For example, glutamine starvation has the unique ability to drive elevated levels of macropinocytic induction in PDAC cells that would otherwise have low levels of basal uptake (11). This boost in macropinocytic capacity occurs through the potentiation of EGFR-PAK signaling via the glutamine deprivation-dependent induction of EGFR ligands, such as HB-EGF and epiregulin. In tumors, regional deficiencies in glutamine and other amino acids are associated with elevated levels of macropinocytosis relative to tumor areas with substantial nutrients (11). Through a poorly understood maturation process, macropinosomes fuse with lysosomes and are acidified, acquiring their degradative properties (12). By quantitatively monitoring protein catabolism in PDAC cells, it was shown that glutamine starvation leads to enhanced protein scavenging to much greater extents than starvation of other amino acids (13). Whether glutamine deprivation affects macropinocytosis in other cellular contexts remains to be explored.

Here, we demonstrate that depletion of glutamine drives macropinocytic induction in cancer-associated fibroblasts (CAF). In contrast to PDAC cells, stromal macropinocytosis is not reliant on EGFR signaling and instead is dependent on a CaMKK2-AMPK-RAC1 signal that is potentiated by elevated cytosolic Ca²⁺. Although glucose starvation causes robust AMPK activation, it does not substantially increase macropinocytosis, suggesting that glutamine depletion leads to a unique molecular environment that allows for AMPK-dependent uptake. We identify that the RHO/RAC guanine nucleotide exchange factor (GEF) ARHGEF2 preferentially binds to RAC1 upon glutamine starvation and is required for macropinocytosis in CAFs. Functionally, we elucidate that stromal macropinocytosis serves two purposes—it supplies CAFs with intracellular amino acids that sustain their fitness, and it provides secreted amino acids that nourish the PDAC tumor cells under conditions where glutamine is limiting. CAFs have been shown to support tumor growth, and we establish that suppressing stromal macropinocytosis via the depletion of CaMKK2, ARHGEF2, or AMPK compromises the ability of stromal fibroblasts to enhance the growth of PDAC tumors. Our work sheds light on the mechanisms driving macropinocytosis in response to nutrient scarcity and identifies a critical role for macropinocytosis within the PDAC tumor stroma.

In human PDAC specimens, macropinocytosis is stimulated in both tumor and stromal cells (2); therefore, we hypothesized that stromal macropinocytosis may have a functional role in PDAC. To determine the identity of the stromal cells that display macropinocytic uptake, we employed an orthotopic syngeneic mouse model of PDAC in which murine tumor cells originating from a genetically engineered autochthonous mouse model (KPC cells) were implanted into the tail of the pancreas (14). Using this approach, we found that the stromal cells exhibiting robust levels of macropinocytosis within these tumors were marked by α smooth muscle actin (α-SMA), indicating that they were CAFs (Fig. 1A). In contrast to the tumor setting, we found that PDGFRα-expressing fibroblasts residing in normal murine pancreata did not display appreciable macropinocytosis (Fig. 1B). A model has emerged where the tumor microenvironment comprises a heterogeneous population of CAFs with different functions within PDAC tumors (15). This model was established by the discovery of three distinct subtypes of CAFs in PDAC: myofibroblastic CAFs (myCAF) expressing high levels of α-SMA, inflammatory CAFs expressing low levels of α-SMA but high levels of cytokines and chemokines, and, more recently, antigen-presenting CAFs that express MHC class II (MHC-II)–related genes (16, 17). To examine macropinocytosis in the different CAF subtypes, we employed established flow cytometry procedures (16, 17) in conjunction with a TMR-Dextran assay to mark macropinosomes. We found that approximately 45% of total PDPN⁺ CAFs displayed macropinocytosis, which was inhibited by 5-(N-ethyl-N-isopropyl) amiloride (EIPA), a selective inhibitor of macropinocytosis (Supplementary Fig. S1A). This macropinocytic profile represented a distribution across the different CAF subtypes, which all exhibited some macropinocytic uptake (Fig. 1C). However, most macropinocytic CAFs in the orthotopic tumors were myCAFs, and this was reflective of the overall subtype distribution within the tumors (Fig. 1D; Supplementary Fig. S1B).

To further analyze stromal macropinocytosis, we used CAFs isolated from human PDAC tumors (18). Using CAF64 and CAF1424 cells, we found that in nutrient-replete media, CAFs displayed low levels of uptake (Fig. 2A–D). Because glutamine levels regulate macropinocytosis in PDAC cells (11), and it is among the most depleted amino acids in both human and murine PDAC tumors (2, 19), we assessed whether glutamine deprivation affects macropinocytosis in the CAFs. We observed that upon glutamine depletion, CAF macropinocytosis was significantly enhanced (Fig. 2A–D; Supplementary Fig. S2A and S2B). Interestingly, normal human pancreatic stellate cells (hPSC) had low levels of basal uptake that were not significantly enhanced by glutamine depletion (Supplementary Fig. S2C and S2D). Recently, CAFs isolated from prostate tumors have been found to exhibit macropinocytosis due to the epigenetic silencing of RASAL3, a RAS signaling inhibitor (20). We found that these prostatic fibroblasts have high baseline levels of uptake, consistent with what has been reported, but that glutamine depletion only marginally increased uptake (Supplementary Fig. S2E and S2F). Normal human fibroblasts from prostate did not display macropinocytosis under any of the conditions assessed (Supplementary Fig. S2G and S2H). Altogether, these data suggest that the role of glutamine in regulating CAF macropinocytosis might be especially important in PDAC.

In PDAC cells, we have previously delineated how EGFR signaling controls glutamine depletion–induced macropinocytosis (11); however, this mechanism does not seem to be employed in CAFs. EGF stimulation fails to induce macropinocytosis in CAFs (Supplementary Fig. S2I and S2J). Moreover, in CAFs, EGF stimulation does not trigger EGFR phosphorylation, nor does glutamine starvation drive the activation of downstream effector pathways, such as ERK (Supplementary Fig. S2K and S2L). These data suggested that CAFs employ an alternative mechanism to drive macropinocytic stimulation in response to glutamine deprivation. A major nutrient sensing pathway that cells employ to respond to bioenergetic stress is AMPK signaling (21). AMPK signaling has been previously implicated in the regulation of macropinocytosis in prostate cancer cells; however, within that context, uptake was nutrient independent, and the mechanistic underpinnings driving the process remain unclear (4). AMPK activation is best characterized in the setting of glucose starvation and has not been broadly examined in the context of glutamine depletion; therefore, we starved CAFs of glutamine and assessed AMPK activation by immunoblotting for the phosphorylated forms of the catalytic α subunits (pAMPKα). In both CAF64 and CAF1424 cells, we observed an enhancement of the pAMPKα/AMPKα ratio upon glutamine deprivation (Fig. 2E). To investigate whether AMPK plays a role in glutamine depletion–induced macropinocytosis, we genetically and pharmacologically suppressed AMPK signaling in CAFs. Knockdown of the AMPKα1 subunit or treatment with Compound C, an AMPK inhibitor, robustly suppressed macropinocytosis in CAFs depleted of glutamine (Fig. 2F–J; Supplementary Fig. S3A–S3C). At the protein level, the AMPKα1 isoform accounts for most of AMPKα protein in CAFs, and consistent with this observation, knockdown of AMPKα2 had no effect on macropinocytosis (Supplementary Fig. S3D–S3F). Taken together, our data demonstrate that glutamine levels can modulate AMPK phosphorylation and that glutamine depletion–induced macropinocytosis in CAFs is dependent on AMPK.

AMPKα phosphorylation can be mediated by LKB1 in response to metabolic changes in the cellular AMP/ATP ratio or by CaMKK2 in response to elevated cytosolic Ca⁺² levels (21). To discern the mechanism underlying AMPK activation in the context of glutamine depletion, we used shRNA-mediated knockdown of either LKB1 or CAMKK2 and assessed the extent of AMPKα phosphorylation in CAFs. We determined that only knockdown of CaMKK2 led to a decrease in the pAMPKα/AMPKα ratio (Fig. 3A). Consistent with this observation, macropinocytic induction in response to glutamine starvation was unaffected by LKB1 depletion but was strongly suppressed by knockdown of CaMKK2 (Fig. 3B; Supplementary Fig. S3G). Importantly, upon glutamine depletion, both the pAMPKα/AMPKα ratio and macropinocytosis were significantly decreased in CAFs by STO-609, a specific inhibitor of CaMKK2 (Fig. 3C–E; Supplementary Fig. S3H). CaMKK2 becomes activated by increased cytosolic Ca²⁺ levels; therefore, we assessed whether glutamine starvation was triggering an enhancement of cytosolic Ca²⁺. We performed a time course experiment in which CAFs were starved of glutamine for 0, 1, 2, 4, or 8 hours and subsequently stained cells with Cal-520, a calcium indicator dye (22). Relative to nutrient-replete control cells, CAFs that were depleted of glutamine displayed enhanced Cal-520 fluorescence that peaked after 4 hours of starvation and was then sustained (Fig. 3F and G). Calcium ionophores directly facilitate the transport of Ca²⁺ across the plasma membrane, and consistent with a role for cytosolic Ca²⁺ in the modulation of AMPK-dependent macropinocytosis, calcium ionophores, such as A23187 and ionomycin, enhanced both the pAMPKα/AMPKα ratio and macropinocytosis in CAFs cultured in nutrient-replete media (Fig. 3H–J). These findings demonstrate that glutamine limitation has the capacity to drive elevated cytosolic Ca²⁺, which in turn potentiates CaMKK2-AMPK signaling to amplify macropinocytic induction in CAFs.

RAC1 is a small GTPase that regulates macropinocytosis via the formation of actin-dependent plasma membrane ruffles (6). Because AMPK signaling has been previously demonstrated to activate RAC1 (23), we assessed RAC1 localization and activity in CAFs deprived of glutamine. Upon activation, RAC1 protein becomes enriched in membrane ruffles. Consistent with heightened RAC1 activity, we observed enhanced membrane ruffling and increased localization of RAC1 to these ruffles in CAFs cultured in glutamine-free conditions (Supplementary Fig. S3I). Importantly, glutamine deprivation–induced macropinocytosis in the CAFs was abrogated by EHOP-016, a specific RAC1 inhibitor (Supplementary Fig. S3J and S3K). To directly assess RAC1 activity, we used a genetically encoded fluorescence resonance energy transfer (FRET)–based RAC1 biosensor (24). Relative to nutrient-replete control cells, we detected enhanced RAC1 activity in CAFs that were starved of glutamine (Fig. 4A). Importantly, both RAC1 activity and RAC1 recruitment to membrane ruffles were attenuated by inhibition of AMPK or CaMKK2 via Compound C or STO-609 administration, respectively (Fig. 4A and Supplementary Fig. S3I). AMPK signaling has been mainly studied in the context of glucose limitation; therefore, we tested whether glucose starvation might also enhance macropinocytic capacity in CAFs. Interestingly, although we detected robust phosphorylation of AMPK upon glucose withdrawal, the effect on macropinocytosis was minimal relative to glutamine deprivation (Supplementary Fig. S4A–S4C). Consistent with these findings, glucose depletion failed to activate RAC1 (Supplementary Fig. S4D). These data suggested that although AMPK activation is required for glutamine depletion–induced macropinocytosis in CAFs, it is not sufficient.

We surmised that the selectivity of glutamine depletion in driving AMPK-dependent macropinocytosis might involve a unique ability to activate RAC1. To assess this possibility, we performed a RAC1^G15A pull-down. RAC1^G15A is a nucleotide-free RAC1 mutant that efficiently binds to the active form of RAC1-GEFs (25, 26). We used this approach to pull down proteins from lysates derived from CAFs cultured in either glutamine-free or nutrient-replete media and subjected the RAC1^G15A-purified samples to proteomics. We identified one active RAC1-GEF, ARHGEF2, that was enriched in the glutamine-starved condition (Fig. 4B). Western blot analysis validated that ARHGEF2 binding to RAC^G15A was enhanced upon glutamine depletion (Fig. 4C). Moreover, glucose starvation did not result in enhanced ARHGEF2 binding to the same extent (Fig. 4C; Supplementary Fig. S4E). Interestingly, glutamine deprivation led to an increase in ARHGEF2 at both the transcript and protein levels (Supplementary Fig. S4F and S4G). In addition, ARHGEF2 expression levels were enhanced by the calcium ionophore A23187 but were not dependent on CaMKK2, as assessed by STO-609 treatment (Fig. 4D–G; Supplementary Fig. S4H). To determine whether ARHGEF2 plays a role in regulating macropinocytic induction in response to limiting glutamine, we performed knockdown experiments and assessed uptake. We found that ARHGEF2 knockdown significantly suppressed macropinocytosis under glutamine-deprived conditions but did not appear to affect basal macropinocytosis (Fig. 4H–J). Altogether, our mechanistic interrogation of macropinocytosis in CAFs suggests a model in which glutamine starvation–induced macropinocytosis requires both Ca²⁺-activated CaMKK2-AMPK signaling and ARHGEF2 (Fig. 4K).

In tumor cells, macropinocytic cargo is targeted for lysosome-dependent degradation (1). The degradative properties of macropinosomes can be scrutinized by incubating cells with a highly self-quenched BODIPY dye–conjugated form of bovine serum albumin (DQ-BSA) that emits a bright fluorescent signal only after lysosome-dependent proteolytic digestion (1). To assess protein degradation in CAFs, we performed a pulse-chase experiment by incubating glutamine-deprived CAFs for 30 minutes with DQ-BSA and subsequently chased in DQ-BSA–free media for 2 or 4 hours. We detected DQ-BSA fluorescence only in the chased cells, indicating that the macropinocytosed cargo in CAFs is destined for proteolytic degradation (Fig. 5A). Due to the nutrient-poor tumor microenvironment associated with PDAC, we reasoned that the macropinocytosis of extracellular albumin might represent a strategy in CAFs to maintain their fitness when confronted with a paucity of nutrients. To test this idea, we cultured CAFs in glutamine-depleted media with a range of glutamine concentrations and assessed cell number with and without albumin supplementation. We found that the CAFs were sensitive to subphysiologic concentrations of glutamine and that the addition of BSA to the culture media significantly enhanced CAF cell survival (Fig. 5B). Notably, this beneficial effect on CAF survival was suppressed by EIPA (Supplementary Fig. S5A–S5D). Consistent with an effect on CAF cell fitness, we observed that glutamine deprivation compromised the ability of the CAFs to produce collagen VI and fibronectin (Supplementary Fig. S5E), both extracellular matrix molecules that have been shown to promote tumor progression (27–32). Importantly, addition of BSA to the glutamine-depleted medium restored collagen VI and fibronectin expression, an effect that was blocked by macropinocytosis inhibition (Supplementary Fig. S5E).

These observations raised the possibility that protein scavenging by CAFs could be producing albumin-derived amino acids that contribute to the intracellular amino acid pools and sustain CAF cell fitness. To explore this idea, we performed isotope tracing experiments based on a technique that we have previously employed that enables specific quantitation of protein-derived amino acids produced by albumin catabolism (2). CAFs were cultured in the presence of ¹³C-labeled glucose and glutamine for 6 days, which results in near-complete labeling of a subset of intracellular nonessential amino acids (NEAAs; Supplementary Fig. S6A and S6B). CAFs were then switched to glutamine-free media in the presence of ¹³C-labeled glucose with and without supplementation of unlabeled BSA for 2 more days. Intracellular metabolites were then extracted and polar metabolites were analyzed by gas chromatography/mass spectrometry (GC/MS). In CAFs cultured with BSA, we observed an increase in the unlabeled (i.e., protein-derived) fraction for several of the amino acids (Supplementary Fig. S6C). Importantly, the enhanced intracellular abundances of unlabeled amino acids in the presence of BSA were suppressed by inhibiting macropinocytosis via Compound C (Fig. 5C). These data demonstrate that under glutamine-depleted conditions, protein-derived amino acids supplied by the macropinocytosis pathway contribute to the intracellular amino acid pools. To determine whether the degradation of macropinocytosed protein could serve as a source of amino acids that are exported extracellularly, we used isotope tracing and quantified polar metabolites in the extracellular medium. We found that the addition of BSA to the CAF culturing medium enhanced the unlabeled (i.e., protein-derived) fractions for several amino acids in the extracellular medium (Fig. 5D). Consistent with a role for macropinocytosis in the production of these extracellular amino acids, knockdown of CaMKK2 in the CAFs led to a decrease in the extracellular abundances of several NEAAs (Fig. 5E). Altogether, these findings demonstrate that macropinocytosis in CAFs leads to the production of intracellular and extracellular protein-derived amino acids.

It is well documented that CAF-conditioned media provide a proliferative advantage to PDAC tumor cells (33). To specifically assess whether macropinocytosis-produced amino acids can enhance PDAC cell survival, we generated CAF-conditioned low-glutamine medium supplemented with or without BSA for 48 hours. The CAF-conditioned medium was then filtered using a 3-kDa cutoff filter to remove large molecules such as BSA and other proteins. The capacity of the conditioned medium to enhance PDAC cell fitness was then assessed. Indeed, filtered CAF-conditioned media that had contained BSA as a nutrient source enhanced the fitness of the PDAC cells relative to control media, and these effects were reduced by conditioning media with CAFs depleted of CaMKK2 (Supplementary Fig. S6D). We next assessed whether NEAAs could directly enhance PDAC cell survival when glutamine is limiting. To do this, we cultured PDAC cell lines, including 1334E and 779E primary PDAC cells, in glutamine-free control media or glutamine-free media supplemented with NEAAs. Interestingly, the addition of NEAAs to the growth medium had the ability to suppress the deleterious effects of glutamine withdrawal (Fig. 5F; Supplementary Fig. S6E). These data reveal that macropinocytosis in CAFs could serve as a source of amino acids that support the fitness of PDAC cells, allowing them to escape the deleterious effects of nutrient stress within the tumor microenvironment.

To assess whether specifically suppressing macropinocytosis in CAFs has the capacity to modulate tumor growth, we employed a heterotopic xenograft mouse model of PDAC in which human PDAC cells are coimplanted with human CAFs. Previous studies have established that coimplantation of PDAC and CAF cells results in an enhancement of tumor growth (33). Indeed, tumors derived from AsPC-1 cells that were coinjected with control CAFs displayed a significant increase in tumor growth relative to tumors derived from AsPC-1 cells alone (Fig. 6A). To specifically abrogate macropinocytic induction in the CAFs, we used CAF cells that were depleted of CaMKK2. AsPC-1 tumors derived from coimplantation with CaMKK2-knockdown CAFs did not display a growth advantage (Fig. 6A–C). To confirm CaMKK2 depletion in the CAF cells within the tumors, we performed immunofluorescence on tumor sections using antibodies specific to CaMKK2 and α-SMA. We found that only tumors derived from AsPC-1 cells coinjected with control CAFs displayed appreciable levels of CaMKK2 that coincided with α-SMA staining (Supplementary Fig. S7A). The observed effects on PDAC tumor growth were not due to differential CAF cell content, as the tumors derived from coinjections with control CAFs displayed similar levels of α-SMA–positive fibroblasts compared with tumors containing CaMKK2-depleted CAFs (Supplementary Fig. S7B). We next interrogated proliferative capacity within the tumors by evaluating the extent of Ki-67 staining, which specifically labels proliferative cells. In tumors derived from control CAFs, there was an enhancement in proliferative potential, which was suppressed by CaMKK2 depletion (Fig. 6D and E). CAFs within the PDAC tumor microenvironment are thought to be critical to collagen deposition, which significantly contributes to the desmoplastic response observed in these tumors; therefore, we assessed intratumoral collagen via Masson's trichrome staining. We found that although control CAFs enhanced the extent of collagen staining within the tumors, CaMKK2-knockdown CAFs did not (Fig. 6F and G). Finally, to assess macropinocytosis in the tumors, we performed an ex vivo macropinocytosis assay (34). In control CAFs marked by α-SMA staining, we detected enhanced levels of macropinocytosis relative to CAFs depleted of CaMKK2 (Fig. 6H and I). To investigate the role of ARHGEF2 in mediating the tumor-promoting effects of CAFs in vivo, we employed the same xenograft mouse model but coinjected human AsPC-1 cells with CAFs that were depleted of ARHGEF2. Similar to CaMKK2 depletion, we found that knockdown of ARHGEF2 in CAFs led to decreased tumor growth relative to control tumors, as well as diminished CAF macropinocytosis (Fig. 7A–E). Specific depletion of ARHGEF2 in the CAFs, as validated by immunofluorescence, also decreased intratumoral proliferation (Supplementary Fig. S7C–S7E). Altogether, these data demonstrate that perturbing macropinocytic stimulation in CAFs leads to a reduction in xenograft tumor growth.

To analyze the effects of stromal macropinocytosis on tumor growth in an immunocompetent setting, we employed a syngeneic heterotopic mouse model of PDAC that has been previously demonstrated to recapitulate the histopathologic complexity of the human disease (Supplementary Fig. S8A; ref. 14). KPC cells were implanted subcutaneously with mouse embryonic fibroblasts (MEF) that originated from an AMPKα double-knockout animal (AMPKα^DKO) or MEFs from a wild-type control (AMPKα^WT; ref. 35). Similar to pancreatic CAFs, the MEFs displayed glutamine depletion–induced macropinocytosis that was dependent on AMPKα expression (Supplementary Fig. S8B–S8D). In in vivo coinjection experiments with KPC cells, tumors that included AMPKα^DKO MEFs displayed significant reductions in growth relative to control tumors that contained AMPKα^WT MEFs (Fig. 7F–H). The KPC tumors containing the AMPKα^DKO MEFs also displayed reductions in stromal macropinocytic capacity relative to the AMPKα^WT MEF-derived tumors (Supplementary Fig. S8E and S8F).

Our work implicates a model in which macropinocytosis plays an instrumental role in the biology of the PDAC tumor stroma (Fig. 7I). We find that macropinocytosis-driven protein catabolism in the stroma enables CAFs to survive glutamine depletion and produces protein-derived amino acids that are secreted extracellularly to nourish the PDAC cells under conditions of nutrient stress. Systemic inhibition of macropinocytosis via EIPA administration leads to a substantial decrease in tumor growth in a xenograft mouse model of PDAC (1). Considering our determination that, in addition to the tumor cells, the CAFs within these tumors also use macropinocytosis, the observed effects with EIPA are likely a consequence of inhibiting uptake throughout the tumor. Altogether, the data point toward a scenario where PDAC and CAF macropinocytosis both contribute to tumor growth. Interestingly, we find that protein-derived alanine is among the secreted amino acids when CAFs are cultured with serum albumin as a nutrient source. Recently, it was demonstrated that stromal fibroblasts can support PDAC metabolism through autophagy-mediated alanine production (36). In addition to macropinocytosis, PDAC tumors are also dependent upon autophagy for their growth and progression (37). Our work further underscores the potential importance of alanine in augmenting PDAC cell fitness, and it remains to be elucidated in vivo what the specific contribution of each protein-scavenging pathway is to the overall production of secreted amino acids within the PDAC tumor microenvironment. In addition to amino acids, CAFs in PDAC tumors have also been shown to secrete an abundance of lipids that support metabolic and signaling processes in the PDAC tumor cells (38). It is unclear whether the production of these lipids, particularly lysophosphatidylcholines, is dependent on macropinocytosis, and it might be interesting to scrutinize the affect of glutamine depletion on lipid production by the CAFs.

We determined that CaMKK2-AMPK signaling drives stromal macropinocytosis in response to cytosolic Ca²⁺ levels, which become enhanced when glutamine is limiting. It is unclear what specifically is driving the increase in cytosolic Ca²⁺ upon glutamine depletion; however, it could be linked to endoplasmic reticulum (ER) stress (39). We previously deciphered that many genes linked to ER stress are modulated by glutamine depletion in PDAC (11), and others have elucidated that nutrient starvation likely leads to ER stress through the inhibition of glutamine/fructose-6-phosphate amidotransferase, an enzyme in the hexosamine biosynthesis pathway that is required for the production of UDP-GlcNAc (40–42). CaMKK2-AMPK signaling is thought to regulate a variety of critical cellular functions, including energy homeostasis, inflammation, and cytoskeleton remodeling (43). It is well known that macropinosome formation is dependent upon RAC1-mediated actin cytoskeleton remodeling, and our data suggest a link between AMPK and RAC1 activation specifically in the context of glutamine depletion, although we cannot rule out that factors other than glutamine might influence macropinocytosis in vivo. We show that starvation of glutamine, but not glucose, leads to upregulation of ARHGEF2, a RAC1 GEF. In this way, glutamine deprivation might provide the contextual requirements for AMPK to regulate actin cytoskeleton dynamics, as well as macropinocytosis. It remains to be determined whether this occurs through spatiotemporal regulation of AMPK signaling and/or via direct interactions between AMPK and ARHGEF2.

We observe that selectively compromising CAF macropinocytosis in vivo leads to a reduction in tumor growth, as well as suppressed proliferation and collagen deposition. Mounting evidence suggests that total depletion of stromal components, especially α-SMA⁺ myofibroblasts, in spontaneous PDAC mouse models leads to accelerated tumor progression with more invasive, proliferative, undifferentiated tumors (44–46). One possible explanation for this difference in phenotype is that we have not totally ablated the intratumoral CAFs by abrogating their macropinocytic capacity but have instead compromised their fitness and reduced their functionality. A second possible explanation, and a limitation of our study, is that with the implantation models that we have employed, the PDAC tumors are able to recruit wild-type host stromal components that might still serve to restrain tumor progression. In future studies, it might be interesting to assess pharmacologic inhibition of CAF macropinocytosis, alone or in combination with chemotherapies, to evaluate the therapeutic potential of targeting this feature of PDAC tumors. Interestingly, in terms of the molecular determinants that modulate uptake, there seems to be a substantial amount of variability and selectivity depending on the context and cell type. For example, although EGFR-PAK signaling orchestrates glutamine depletion–induced macropinocytosis in PDAC cells (11), we show in the present study that this mechanism is not at play in pancreatic CAFs. In addition, macropinocytosis in prostatic fibroblasts is linked to RAS signaling and ERK activation (20), suggesting that even CAFs from different tumor types might employ alternative mechanisms of uptake. As we continue to decipher the main drivers of macropinocytic induction in different settings, we will arrive closer to being able to develop targeting strategies with enhanced specificity.

Primary patient-derived PDAC-associated fibroblasts (CAF64 and CAF1424) and low-passage primary PDAC cell lines (779E and 1334E) were established by Dr. Andrew Lowy at the University of California, San Diego. Primary CAF64 and CAF1424 cells were cultured in DMEM (Corning) supplemented with 17% FBS (Sigma), 100 U/mL penicillin/streptomycin (pen/strep), and 20 mmol/L HEPES, and only cells within 10 passages were used in this study. Primary CAF64 and CAF1424 cells were immortalized with pBabe-derived retrovirus encoding simian virus (SV40) large T antigen and have been designated as CAF64S and CAF1424S cells. Only CAF64S and CAF1424S cells within 15 passages were used in this study. AsPC-1, MIA PaCa-2, Panc-1, and HPAF-II cell lines were obtained from ATCC. HEK293T cells were obtained from GE Healthcare Dharmacon. LSL-Kras^G12-D/+; Trp53^fl/+; Pdx1-Cre (KPC) mice-derived PDAC cells (KPC.4662) were provided by Dr. Robert Vonderheide (University of Pennsylvania) and subcloned in our laboratory to obtain the KPC#28 clone used in this study. AMPKα1/α2–wild-type MEFs (AMPK^WT-MEFs) and AMPKα1/α2double-knockout (AMPK^DKO)–MEFs were gifts from Dr. Reuben Shaw (Salk Institute). MIA PaCa-2, Panc-1, HPAF-II, 779E, 1334E, hPSC, KPC, AMPK^WT-MEFs, AMPK^DKO-MEFs, and HEK 293T cells were cultured in DMEM containing 10% FBS, 100 U/mL pen/strep, and 20 mmol/L HEPES. AsPC-1 cells were cultured in RPMI 1640 medium (Corning) supplemented with 10% FBS, 1 mmol/L sodium pyruvate, 100 U/mL pen/strep, and 20 mmol/L HEPES. Human prostate CAFs and normal prostate fibroblasts were gifts from Dr. Neil Bhowmick at Cedars-Sinai Medical Center and were cultured as previously described (20). All cells were cultured under 5% CO₂ at 37°C and routinely tested for Mycoplasma contamination using the Mycoplasma PCR Detection Kit (Applied Biological Materials).

For glutamine or glucose starvation, cells were treated with glutamine-free (−Q) or glucose-free (−Glc) DMEM (Corning) in the presence of 10% dialyzed FBS (dzFBS; Gibco) for 20 to 24 hours; the control cells were cultured in DMEM complete medium containing 4 mmol/L glutamine and 25 mmol/L glucose.

Compound C (Sigma-Aldrich), a selective inhibitor of AMPKα, was administrated to CAFs at a concentration of 10 μmol/L during the last 6 hours of glutamine deprivation (−Q), unless otherwise specified. STO-609 (Santa Cruz Biotechnology), a selective inhibitor of CaMKK2, was administrated to CAFs at a concentration of 10 to 25 μmol/L during the last 6 hours of −Q treatment, unless otherwise specified. Calcium ionophores A23187 (Sigma-Aldrich) and ionomycin (Cayman Chemical) were administrated to CAFs at a concentration of 1 μmol/L in the presence of 4 mmol/L glutamine (+Q) for 20 to 24 hours. EIPA (Sigma-Aldrich), an inhibitor of Na⁺/H⁺ exchange, was administrated to CAFs at a concentration of 30 μmol/L under −Q conditions for 48 hours. EHOP-016 (Sigma-Aldrich), a specific inhibitor of RAC1 activity, was administrated to CAFs at a concentration of 3 μmol/L during the last 6 hours of −Q treatment.

In vitro macropinocytosis assay was performed in triplicate wells for each condition as previously described. Briefly, CAF cells were plated onto glass coverslips in 24-well plates for 1 to 2 days and then treated with or without glutamine starvation for 20 to 24 hours along with inhibitor administration. Macropinocytic uptake of cells was assessed following a 30-minute incubation at 37°C with 70-kDa FITC-Dextran or tetramethylrhodamine (TMR)–Dextran (Invitrogen), which was added directly to the culture media at a final concentration of 1 mg/mL, followed by five-time rinse on ice with ice-cold PBS and then fixed in 3.7% formaldehyde. After nuclear staining with DAPI, coverslips were mounted onto glass slides using DAKO mounting medium (Agilent Technologies). Fluorescent images were captured at 40× magnification using the EVOS FL Cell Imaging System (Thermo Fisher Scientific) and analyzed using ImageJ software (NIH). The quantification of macropinosome particles was analyzed as previously described (47, 48). The macropinocytic index was calculated by the total particle area per cell. At least 150 cells were analyzed per condition. For the macropinocytosis assay of CAF cells treated with DMEM containing different glutamine concentrations (0, 0.1, 0.2, 4 mmol/L), fluorescent images were automatically captured at 40× magnification using Cytation 5 Imaging Multi-Mode Reader (BioTek Instruments). The quantification of macropinosome particle area was the same as described above. At least 1,300 cells were analyzed per condition.

CAF cells were plated onto coverslips and subjected to glutamine starvation for 24 hours and then incubated with self-quenched BODIPY TR-X–conjugated BSA (DQ Red BSA; Invitrogen), which was directly added to the media at a concentration of 3 mg/mL at 37°C for 30 minutes. After three-time rinse with PBS, cells were fixed immediately in 3.7% formaldehyde or following a chase for 2 hours or 4 hours. Cell nuclei were stained with DAPI, and coverslips were mounted onto glass slides with DAKO media. Upon proteolytic digestion, DQ Red BSA releases fragments that have excitation and emission maxima of ∼590 nm and ∼620 nm. Images were captured randomly at 40× magnification using the EVOS FL Cell Imaging System. Triplicate wells were performed for each condition, and at least 150 cells were analyzed per condition.

To reduce background fluorescence from the culture media for imaging, FluoroBrite DMEM media (Gibco) were prepared with 10% dzFBS, 100 U/mL pen/strep, and 20 mmol/L HEPES with or without 4 mmol/L glutamine (+Q and −Q, respectively). CAFs seeded onto 8-well chamber slides (Nunc Cell Culture) were treated with +Q or −Q media for 0, 1, 2, 4, and 8 hours. Prior to the end of treatment, cells were loaded with 6 μmol/L Cal-520 AM (AAT Bioquest), a membrane-permeable fluorescent calcium sensor, at 37°C for 60 minutes. After replacement with fresh +Q or −Q media, cells were incubated at room temperature (RT) for 20 minutes before imaging. Images were captured at 20× magnification using the EVOS FL Cell Imaging System and analyzed using ImageJ software. At least 10 images were randomly captured from different fields across each sample. Mean fluorescence intensity per cell was calculated based on a representative intracellular square area after subtracting the background mean fluorescence intensity, and at least 200 cells were analyzed.

HEK 293T cells at ∼40% confluency in 10-cm petri dishes (Nunc Cell Culture) were cotransfected with plasmids pMD2G, psPAX2, and pLKO-shRNA using Lipofectamine 2000 (Life Technologies). Conditioned media containing lentivirus were collected after 48 hours, filtered through a 0.45-μm polyethersulfone filter (Celltreat) to remove the cell debris, and stored at −80°C for long-term use. CAFs were then transduced with lentivirus diluted from 1:2 to 1:20, and knockdown was determined by immunoblots. The lowest dilution of lentivirus achieving >70% reduction at the protein level was used for subsequent experiments. For preparation of SV40 retrovirus, HEK 293T cells were transfected with pCL-Ampho packaging plasmid and pBabe-zero-SV40. The pLKO-based lentiviral vectors used in this study were shAMPKa1 (TRCN0000000861 and TRCN0000199831), shAMPKa2 (TRCN0000002168 and TRCN0000002171), shCaMKK2 (TRCN0000002299 and TRCN0000363122), shLKB1 (TRCN0000000409 and TRCN0000000411), and shARHGEF2 (TRCN0000003174 and TRCN0000003175). Two negative controls were used in this study: shGFP (TRCN0000072201) and shLuciferase (TRCN0000072259).

CAFs transduced with lentivirus expressing RAC1–2G FRET biosensor for 48 hours were seeded onto an 8-well chambered cover glass (Cellvis). The next day, cells were treated with DMEM complete media (+Q) or without glutamine (−Q). In a subset of experiments, cells treated with −Q were administrated with or without 10 μmol/L Compound C or 25 μmol/L STO-609 for 6 hours. Then, cells were quickly washed twice with prechilled PBS and fixed with 3.7% formaldehyde for 20 minutes. After washing three times with PBS, cells were kept in PBS at 4°C until ready for imaging. Images were captured with 40× oil objective using an LSM 710 NLO confocal microscope (Zeiss). Both donor (mTFP) emission channel and FRET channel were captured using the excitation filter of the donor, and at least 25 cells were imaged per sample. FRET ratios were analyzed using Fiji software as previously described (24). Briefly, after subtracting background, two corresponding channel images were registered and converted into 32 bits, followed by image segmentation. Then, FRET ratio images were generated using the RatioPlus plug-in, and the intensity of the FRET ratio was normalized to the maximum intensity per cell.

Immortalized CAFs in 15-cm petri dishes with ∼60% confluence were treated with DMEM complete media or without glutamine or glucose for 24 hours. Cells were washed twice with prechilled PBS and lysed with KCl lysis buffer containing 50 mmol/L Tris-HCl (pH 7.5), 25 mmol/L KCl, 1 mmol/L EDTA, 1% NP-40, 1× protease inhibitor cocktail (Roche), and 1× PhosSTOP (Roche). Cells were collected with a scraper, transferred into 1.5-mL Eppendorf tubes, and kept on ice for 30 minutes before centrifugation at 18,000 × g for 15 minutes. Afterward, clarified cell lysates were transferred into new Eppendorf tubes and kept on ice. Protein concentrations were determined using a BCA Protein Assay Kit (Pierce). Equal amounts of total protein (∼1.5 mg) were incubated with 50 μL RAC1^G15A agarose beads (Cell Biolabs) at 4°C under gentle rotation to pull down active RAC1-GEFs. After a 3-hour incubation, beads were quickly washed three times with KCl lysis buffer and followed by another three washes with 20 mmol/L Tris-HCl (pH 7.5). After careful and complete removal of remaining buffer, beads were kept temporarily at −20°C prior to analysis. To identify the active RAC1-GEF, proteins pulled down with RAC1^G15A beads were hydrolyzed on beads and further applied to LC/MS-MS analysis, which was performed by the Proteomic Core Facility at Sanford Burnham Prebys Medical Discovery Institute. To confirm the proteomics, beads were boiled with 1× SDS sample buffer, and RAC1-GEF candidates were assessed by immunoblots.

Cell samples were lysed in RIPA buffer [10 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 1% sodium deoxycholate, 0.1% SDS, 1% Triton X-100] supplemented with 1× protease inhibitor cocktail and 1× PhosSTOP. After centrifugation at 18,000 × g for 15 minutes at 4°C, cell lysates were transferred to new Eppendorf tubes, and protein concentrations were determined using a BCA Protein Assay Kit. After adding 1× SDS sample loading buffer, lysates were boiled for 5 minutes, followed by SDS-PAGE electrophoresis. Proteins were transferred to nitrocellulose membranes using Mini Gel Tank and Mini Blot Module (Life Technologies) and probed using the following primary antibodies: pAMPKα (T172) (Cell Signaling Technology, #2535, 1:1,000), AMPKα (Cell Signaling Technology, #2532 or #5831, 1:1,000), AMPKα1 (Cell Signaling Technology, #2795, 1:1,000), AMPKα2 (Cell Signaling Technology, #2757, 1:1,000), LKB1 (Santa Cruz Biotechnology, sc-32245, 1:1,000), CaMKK2 (Proteintech, 11549–1-AP, 1:1,000), ARHGEF2 (Cell Signaling Technology, #4076, 1:500), pEGFR (Y1068) (Cell Signaling Technology, #3777, 1:1,000), EGFR (Cell Signaling Technology, #4267, 1:1,000), pERK1/2 (T202/Y204) (Cell Signaling Technology, #4370, 1:1,000), ERK1/2 (Cell Signaling Technology, #4695, 1:1,000), pULK1 (Cell Signaling Technology, #5869, 1:1,000), ULK1 (Cell Signaling Technology, #8054, 1:1,000), Collagen VI (Novus Biologicals, NB120–6588, 1:500), fibronectin (ABclonal Technology, A16678, 1:1,000), α-tubulin (Sigma-Aldrich, T6074, 1:4,000), and β-actin (Sigma-Aldrich, A1978, 1:10,000). After staining with IRDye 680RD or 800CW secondary antibody (LI-COR), immunoblots were detected using an Odyssey CLx Imager (LI-COR). Quantitative analysis of immunoblots was performed using Image Studio Lite Software (LI-COR). To analyze phosphorylation of proteins, membranes were first probed with phospho-specific antibodies, then stripped with NewBlot IR Stripping Buffer (LI-COR) and reprobed with antibodies for the total proteins.

¹³C-labeling was performed by culturing CAF cells in glutamine-free and glucose-free DMEM (DMEM-Q-Glc; Corning) containing 20 mmol/L ¹³C₆-glucose (Sigma-Aldrich) and 2 mmol/L ¹³C₅-glutamine (Sigma-Aldrich) in the presence of 10% FBS for 6 days. Glutamine-free, glucose-free medium with similar components of DMEM, but reduced to 10% of the amino acid amounts (Medium^1/10AA-Q-Glc), was reconstituted with equivalent inorganic salts, 0.4% MEM Amino Acids Solution (Gibco), 1% MEM Vitamin Solution (Gibco), 3 mg/L glycine (Sigma-Aldrich), 4.2 mg/L serine (Sigma-Aldrich), 100 U/mL pen/strep, and 20 mmol/L HEPES. After ¹³C-labeling, CAFs were washed twice with PBS and treated with Medium^1/10AA-Q-Glc medium in the presence of 10 mmol/L ¹³C₆-glucose with or without 40 mg/mL BSA (Fraction V; Millipore) for 48 hours. Then, conditioned media were collected and centrifuged at 1,000 × g for 5 minutes for GC-MS analysis of extracellular polar metabolites. CAF cells were also extracted for GC-MS analysis as previously described (49). Approximately 0.4 × 10⁶ cells were quickly washed with prechilled PBS three times and extracted with 0.45 mL 50% methanol (in water) (−20°C) containing 20 μmol/L L-norvaline (Sigma-Aldrich, internal standard for polar metabolites) and 0.225 mL chloroform. After centrifugation, the upper methanol/water phase was collected for analysis of ¹³C-labeling and quantification of intracellular free amino acids. Extra wells of CAFs were processed side-by-side for cell counts. Total amounts of intracellular free amino acids were determined by GC-MS analysis as nmol per 1 million cells. Protocols for sample derivatization and running of GC-MS were as described before (49). Unlabeled metabolite standards were used for quantification of the amounts of unlabeled metabolites in samples. These data were then corrected for the fraction of metabolites that were ¹³C-labeled, to yield total metabolite amounts.

Immortalized CAFs or CAFs transduced with shLuciferase control or shCaMKK2 lentivirus were seeded onto 12-well plates. When cells reached 50% to 60% confluence, they were treated with Medium^1/10AA-Q medium supplemented with 40 mg/mL albumin for 48 hours. The no-cell media controls were added to empty wells and also incubated for 48 hours. Afterward, cell-conditioned media and no-cell media controls were collected as described above. Media samples (0.3 mL) were mixed with 0.3 mL methanol, 0.3 mL chloroform, and 10 μL 1 mmol/L L-norvaline and then processed for quantification of amino acids by GC-MS as described for cell samples.

CAFs seeded onto a 96-well plate with confluence of ∼60% were treated with DMEM with different doses of glutamine (0, 0.1, 0.2 mmol/L) with or without 40 mg/mL albumin and 30 μmol/L EIPA. On days 0, 2, and 4 posttreatment, cells were stained with 0.5% crystal violet for 20 minutes, extensively washed with distilled water, and dried overnight. The crystal violet plates were scanned, and the relative cell number was quantified as stained cell area using ImageJ. PDAC cells MIA PaCa-2, Panc-1, AsPC-1, HPAF-II, 779E, and 1334E cells seeded onto 96-well plates were treated with DMEM-Q with or without 100 μmol/L MEM NEAAs (Gibco) for several days (9 days for 779E and 1334E, 4–5 days for the other cells). The medium was refreshed every 3 days. At the end of treatment, relative cell number was quantified using crystal violet staining.

CAFs seeded onto 96-well plates with a confluence of ∼60% were treated with DMEM-Q with or without 40 mg/mL albumin and 30 μmol/L EIPA. Medium was refreshed after 3 days. Five to 6 days posttreatment, cells were washed twice with PBS, and the MTT assay was performed by incubating cells with 0.5 mg/mL MTT reagent (Sigma-Aldrich) for 1 hour. CAFs transduced with shLuciferase control or shCaMKK2 lentivirus were conditioned with Medium^1/10AA-Q medium in the presence of 40 mg/mL albumin for 48 hours. Conditioned media were collected and passed through 0.45-μm filters, followed by centrifugation through 3-kDa cutoff columns (EMD Millipore). The resultant conditioned media were administrated to PDAC cells AsPC-1, MIA PaCa-2, HPAF-II, and Panc-1 for 4 to 5 days, and the media were refreshed after 3 days. Relative cell fitness was determined by the MTT assay. PDAC cells (AsPC-1, MIA PaCa-2, HPAF-II, and Panc-1) were treated with Medium^1/10AA-Q medium with or without 100 μmol/L NEAA for 4 to 5 days. The media were refreshed after 3 days. Relative cell fitness was determined by the MTT assay.

Total RNA was isolated from cells using the PureLink RNA Mini Kit (Invitrogen) according to the manufacturer's instructions. cDNA was synthesized from 1 μg total RNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). qRT-PCR was then performed using SYBR Premix Ex Taq II master mix (Takara) on the LightCycler 96 Instrument (Roche). The primer sequences for ARHGEF2 are as follows: forward primer, 5′-GATCTGCTCAT CAGCCAGTTCT CAG-3′; reverse primer, 5′-GATGAATTGCTGGAA GCGTTTGTC-3′.

Six-week-old female nude mice (Foxn1nu/Foxn1nu) and C57BL/6 mice were purchased from The Jackson Laboratory and allowed to acclimatize for at least 4 days. Female nude mice were used in the xenograft experiments due to their lower aggressiveness relative to that of male nude mice. Mouse studies were conducted at the animal facility of Sanford Burnham Prebys Medical Discovery Institute in accordance with the mouse handling and experimental protocols that were approved by the Institutional Animal Care and Use Committee. This study was in compliance with all the relevant ethical standards.

For heterotopic xenografts, 2.5 × 10⁵ AsPC-1 cells suspended in PBS with 50% matrigel (Corning) were subcutaneously injected into the flanks of nude mice alone or coinjected with 7.5 × 10⁵ CAF64S cells transduced with shCaMKK2 lentivirus (TRCN0000363122, shGFP as the control) or shARHGEF2 lentivirus (TRCN0000003174, shLuciferase as the control). Tumor sizes were measured twice a week using a digital caliper. Tumor volume was calculated according to the following formula: 1/2 × length (mm) × width (mm)². Mice were sacrificed when the largest tumors reached 500 to 600 mm³. Tumor tissues were collected immediately for ex vivo macropinocytosis assay, flash frozen for immunofluorescent staining, or fixed in formaldehyde for immunohistochemistry.

For heterotopic syngeneic tumors, 2.5 × 10⁵ KPC#28 cells and 7.5 × 10⁵ AMPK^WT-MEFs or AMPK^DKO-MEFs were suspended in PBS with 50% matrigel and subcutaneously injected into the flanks of C57BL/6 mice. Tumor sizes were measured every 4 days using a digital caliper. Mice were sacrificed when the largest tumors reached ∼500 mm³. Tumor tissues were collected and processed as described above.

For orthotopic syngeneic tumors, 2.5 × 10⁴ KPC4662 cells suspended in PBS with 50% matrigel were injected into the tail of the pancreas of C57BL/6 mice. Tumors were harvested 4 weeks later for ex vivo macropinocytosis assay.

As previously described (34), cross sections with ∼1 mm thickness were cut from fresh tumor tissues, immediately injected at multiple sites with 20 mg/mL 10-kDa TMR-Dextran (Invitrogen), and finally immersed in 400 μL TMR-Dextran for 15 minutes. After quickly washing three times with PBS, cross sections were embedded in OCT compound and snap-frozen on dry ice. CAFs were labeled with rabbit anti–α-SMA antibody (Cell Signaling Technology, #19245, 1:500) and Alexa Fluor 555–conjugated goat anti-rabbit secondary antibody for immunofluorescence staining. Images were captured at 40× magnification using the EVOS FL Cell Imaging System and analyzed using ImageJ software as previously described.

KPC4662 orthotopic tumors were minced to 3- to 4-mm pieces and dissociated using a Tumor Dissociation Kit (Miltenyi Biotec). After passing through a 70-μmol/L cell strainer, cells were centrifuged at 500 × g for 5 minutes and resuspended in RBC lysis buffer (eBioscience) for 3 minutes to eliminate red blood cells. After quench with Hank's Balanced Salt Solution (HBSS)/2% dzFBS, cells were spun down and resuspended in HBSS/2% dzFBS. Cell numbers were counted using a hemocytometer. Then, ∼3 × 10⁶ cells were incubated with 2 mg/mL 70-kDa TMR-Dextran at 37°C for 30 minutes, followed by two washes with cold HBSS/2% dzFBS. Cells were spun down and blocked with anti-mouse CD16/32 antibody (BioLegend) at 4°C for 15 minutes and then were stained with the following antibodies at 4°C for 30 minutes: CD45-PerCP/Cy5.5 (BioLegend, clone 30-F11), podoplanin (PDPN)–APC/Cy7 (BioLegend, clone 8.1.1), PDGFRα-FITC (eBioscience, clone APA5), Ly6C–Alexa Fluor 647 (BioLegend, clone HK1.4), and I-A/I-E (MHC-II)–Brilliant Violet 785 (BioLegend, clone M5/114.15.2), as previously described (16, 17). Cells were washed twice with cold HBSS/2% dzFBS, fixed with 3.7% formaldehyde, and washed twice again to remove the fixative. CAF subtype markers and TMR-Dextran were then analyzed by flow cytometry. From each sample, 100,000 events were acquired on the BD LSRFortessa X-20 flow cytometer.

Tumor tissues frozen in OCT compound were cut into sections with a 5-μm thickness. After blocking with 10% goat serum/1% BSA for 1 hour, CAFs were stained with rabbit anti–α-SMA antibody (Cell Signaling Technology, 1:500) overnight at 4°C and then incubated with Alexa Fluor 488– or Alexa Fluor 647–conjugated goat anti-rabbit secondary antibody at RT for 1 hour. For immunofluorescent costaining of α-SMA and CaMKK2, tumor tissue sections were first stained with goat anti–α-SMA antibody (Novus Biologicals, NB300–978, 1:50) overnight at 4°C and Alexa Fluor 488–conjugated donkey anti-goat secondary antibody; afterward, sections were stained with rabbit anti-CaMKK2 antibody (Proteintech, 1:100) for 2 hours at RT, followed by a 1-hour incubation of Alexa Fluor 555–conjugated goat anti-rabbit secondary antibody. For immunofluorescent costaining of CK8 and ARHGEF2, tumor sections were stained first with rat anti-CK8 antibody (TROMO-I, DSHB, AB_531826, 1:400) and subsequently stained with rabbit anti-ARHGEF2 antibody (Abcam, ab155785, 1:100), as described above, and the sequential tumor sections were stained with rabbit anti–α-SMA (Cell Signaling Technology, 1:500) at the same time. Images were captured at 40× magnification using the EVOS FL Cell Imaging System and analyzed using Fiji software.

Tumor tissues fixed in 10% formalin were subjected to standard paraffin processing and sectioning. After deparaffinization and rehydration, heat-induced antigen retrieval of sections was performed in 0.01 mol/L citrate buffer (pH 6.0) using a microwave. At RT, sections were rinsed three times with PBS and immersed in 1% hydrogen peroxide for 30 minutes. Afterward, tumor tissues were permeabilized with 0.1% Triton-X100 in TBS containing 0.1% Tween 20 (TBST) for 10 minutes and blocked with 10% goat serum and 1% BSA in TBST containing 0.1% Triton-X100 at RT for 1 hour. Tumor tissues were stained with rabbit Ki-67 (CST, #9027, 1:400), rabbit α-SMA (1:500), or rat CK8 (TROMO-I, DSHB, 1:400) primary antibody at 4°C overnight, followed by incubation with biotinylated goat anti-rabbit or goat anti-rat secondary antibody (Vector Labs, BA-1000, BA-9400) at RT for 1.5 hours and another 30-minute incubation with horseradish peroxidase (HRP)–coupled Avidin-Biotinylated enzyme Complex reagent (ABC; Vector Labs). HRP signal was developed using the DAB HRP Substrate Kit (Vector Laboratories). After nuclear counterstaining with hematoxylin, sections were dehydrated and mounted with coverslips using Permount mounting medium (Thermo Fisher Scientific). Trichrome staining of paraffin-embedded tumor samples was performed by Sanford Burnham Prebys Medical Discovery Institutes Histology Core facility. Images were captured using a brightfield microscope installed with an INFINITY camera (Lumenera).

All experiments were independently performed at least three times. At least triplicate samples were performed for each independent experiment. All graphs were made using GraphPad Prism software (GraphPad Software). Results are shown as means; error bars represent SEM. Statistical significance was determined by the unpaired two-tailed Student t test or one-way ANOVA, and P values less than 0.05 were considered statistically significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

D.A. Scott reports grants from the National Cancer Institute during the conduct of the study. C. Commisso reports a patent #9983194 issued. No disclosures were reported by the other authors.

Y. Zhang: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M.V. Recouvreux: Resources, formal analysis, validation, investigation. M. Jung: Formal analysis, validation. K.M. Galenkamp: Formal analysis, validation, methodology. Y. Li: Investigation. O. Zagnitko: Formal analysis, validation, investigation, methodology. D.A. Scott: Formal analysis, supervision, validation, visualization, methodology, writing–original draft. A.M. Lowy: Resources, validation, methodology. C. Commisso: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We thank current and former members of the Commisso laboratory for their helpful comments and discussions, Dr. Neil Bhowmick for prostate CAFs and normal prostate fibroblasts, Dr. Robert Vonderheide for KPC cells, Dr. Reuben Shaw for AMPKα^DKO and AMPKα^WT MEFs, the Sanford Burnham Prebys Cancer Metabolism Core for performing the ¹³C-labeling and polar metabolite quantification experiments, Dr. Alexandre Rosa Campos and the Sanford Burnham Prebys Proteomics Core for performing mass spectrometry experiments, and Guillermina Garcia and Monica Sevilla as well as the Sanford Burnham Prebys Histology Core for performing trichrome staining. This work was supported by NIH grant R01CA207189 to C. Commisso. Y. Zhang is a recipient of an NIH/NCI T32 Fellowship (CA211036). Sanford Burnham Prebys core services are supported by NCI Cancer Center Support grant P30CA030199.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.