PFKFB3-Driven Macrophage Glycolytic Metabolism Is a Crucial Component of Innate Antiviral Defense

Higher organisms have evolved a powerful innate immune system to defend against viral infections. The activation of innate antiviral immunity is based on recognition of virus-associated molecular patterns, including specific nucleic acid species, by cell/endosome surface TLRs in professional innate immune cell types or by another group of cytosolic viral sensors expressed in virtually all cell types (1). Viral sensing by these receptors engage different, yet overlapping signaling pathways, which eventually converge on the production of the type I IFN as well as a number of proinflammatory cytokines. Type I IFN, in particular, trigger a Stat1/2- and IRF9-dependent transcriptional program in a broad range of cell types to exert powerful antiviral effects (2, 3).

It has been increasingly recognized that innate immune activation is associated with significant metabolic changes (4, 5). Such cross-talks may represent a strategic shift of cellular metabolism to support the immune responses. As far as glucose metabolism is concerned, it was found that in macrophages and dendritic cells, TLR activation results in significant increases in oxygen-independent glycolysis and often decreases in the aerobic oxidative phosphorylation (OXPHOS) metabolism (6–12). Such metabolic changes have been proven essential for some key immunoregulatory events downstream of TLR activation. However, few studies have investigated the features of cellular metabolic changes in the context of innate antiviral immune activation by live viruses and have dissected the contributing signaling pathways. Moreover, whether immune cells and nonimmune cells undergo differential metabolic regulation under viral challenges is not known.

To fill such gaps in knowledge, we examined virus-elicited regulation of glycolytic metabolism in macrophages and mouse embryonic fibroblasts (MEFs), as representatives of innate immune cells types and nonimmune cell types, respectively. We established that type I IFN triggered by viral infection acts to upregulate glycolytic activator 6-phosphofructose-2-kinase and fructose-2,6-bisphosphatase (PFKFB3) preferentially in macrophages. We provided further evidence to demonstrate PFKFB3 as a novel, metabolic effector of innate antiviral immune response.

In this study, all experiments involving mice usage by the collaborating laboratories were approved by the Institutional Animal Care and Use Committee of the Model Animal Research Center at Nanjing University (MARC-NJU project license LJH12) and by the Institutional Animal Care and Use Committee of the School of Medicine at NJU (project license LiE-02). The animal care and use protocols were in strict accordance with Regulation for Management of Laboratory Animals (1988) and Guidelines for Care and Use of Laboratory Animals (2006), both issued by the Ministry of Science and Technology of PR China.

All mice were of C57/BL6 background. They were produced and maintained on ad libitum food and water by Nanjing Biomedical Research Institute-NJU, an American Association for the Accreditation of Laboratory Animal Care–accredited specific pathogen-free animal facility. For experiments involving viral infection, the mice were transferred to a BSL2-level animal facility at NJU School of Medicine.

The Pfkfb3 heterozygous knockout mice were generated by clustered regularly interspaced short palindromic repeats/CRISPR-associated 9 technique that caused a 73-bp indel in exon 2 of Pfkfb3. The mice were genotyped by PCR using two primers (forward: 5′-GGTATAGGACTCACACCATTAAG-3′; reverse: 5′-GACCTGGCTTACCTTTCGTTGGA-3′) to amplify a 202-bp fragment for the wild type (WT) allele and 129-bp fragment for the deletion mutant allele. In all experiments involving Pfkfb3^+/− mice, WT littermates of the same sex were always used as controls.

Unless otherwise indicated, all chemicals were purchased from Sigma-Aldrich. Polyinosinic–polycytidylic acid [poly(I:C)] was purchased from Invivogen. Mouse IFN-β (12405-1) and human IFN-α were from PBL and Sangon, respectively. Wortmannin (9951) was from Cell Signaling Technology. JAK Inhibitor I (420099) was from Merck Millipore. PFK15 was purchased from Abcam. Lipofectamine 2000 (Invitrogen) and INTERFERin (Polyplus) were used to mediate poly(I:C) transfection of nonimmune cells and macrophages, respectively.

Primary Abs were purchased from Sangon (Stat1, AB55186), Proteintech (PFKFB3, 13763-A-AP), Santa Cruz (GAPDH, A1713), Genscript (VSV-G, A00199) Sigma-Aldrich (Flag, F3165), Abcam (PFKFB3, AB181861), and Cell Signaling Technology (Abs for AKT pathway). Neutralizing Abs against mouse IFNAR1 (127302), blocking Ab against mouse TIM4 (130004), the fluorophore-conjugated Abs against CD11b (101217), CD11c (117311), or F4/80 (123122), and the ELISA kit for mIFN-β (439407) were purchased from Biolegend.

Peritoneal macrophages (PMs) and bone marrow–derived macrophages (BMDMs) were prepared from the mice (same sex, 8–12 wk old) according to standard protocols (13). MEFs were obtained from 12.5-d-old embryos as described previously (14). The cell lines used in this study (L929, Raw264.7, Hela, Vero, U937, 293T) were all previously obtained from American Type Culture Collection. The lentiviral tet-on shRNAmir expression system pTRIPZ (Open Biosystems) was modified to drive inducible expression of PFKFB3 as described previously (15). The packaged recombinant lentivirus was used to transduce RAW264.7 cells. Stable transductants were selected with 1 μg/ml puromycin. For transient transfection of PFKFB3 into MEFs, a Flag-tagged construct based on the pcDNA plasmid (Invitrogen) was used.

DMEM (Life Technologies) supplemented with 10% FBS, 1% penicillin/streptomycin were used to culture most of the cells used in the study. These include primary PMs and MEFs. For differentiation of BMDMs, mouse bone marrow aspirates were cultivated in the presence of 30% L929 supernatant (containing endogenous M-CSF) for 7 d. U937 and L929 cells were maintained in RPMI 1640 (Invitrogen) supplemented with 10% FBS, 1% penicillin/streptomycin. All cultures were maintained using a 37°C humidified incubator supplied with 5% CO₂.

Vesicular stomatitis virus (VSV; Indian strain) was propagated using Vero cells. For in vitro infection, cells were inoculated (1 h) with the virus at indicated multiplicity of infections (MOIs). The culture supernatant and cell samples were harvested at different time points afterward. For UV inactivation, the viral stock was exposed under the UV lamp in a tissue culture hood for 30 min. For coculture experiments, infected MEFs (MOI of 1 for 14 h to achieve significant cytopathic effect) were added at a ratio of 2:1 to BMDMs and incubated further. In some experiments, MEFs were added to an upper transwell apparatus (3-μm pore size) separated from the BMDMs in the lower chambers. VSV titer was determined using plaque assay in Hela cells (16). In experiments to visualize cell engulfment, MEFs were initially labeled with fluorescent probe CFSE (eBioscience) according to manufacturer’s instructions. After incubating the infected MEFs with macrophages, the cells were fixed, stained using F4/80 Ab (macrophage), and analyzed by immunofluorescence microscopy (17). In some experiments, engulfment of infected MEFs was assessed via flow cytometry. After incubation with the infected MEFs, unfixed phagocytes were stained with anti-F4/80 and subsequently subjected to flow cytometry analyses. Median fluorescent intensity was used for quantitation.

In brief, the culture medium samples were deproteinated (equal volume of 0.5 M of metaphosphoric acid). The supernatant was then neutralized with 1/20 volume of 5 M of K₂CO₃. After centrifugation to remove the precipitated salts, the samples were analyzed using the lactate assay kit according to manufacturer’s instruction (Cayman).

The cells were incubated for 30 min to 1 h in DMEM without serum and glucose. Cells were next incubated with 150 μg/ml of a fluorescent glucose analog 2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose (2-NBDG; Invitrogen) for 15 min. After washing, cells were collected by trypsin solution (MEF) or PBS containing 0.5% EDTA (macrophage). Cells were maintained on ice and promptly assayed by flow cytometry on a FACSCalibur platform (BD).

PMs or BMDMs were plated at 100,000 cells/well in a specialized 24-well plate. On the next day, untreated or IFN-treated cells were subjected to analysis following instructions bundled with the instrument. In brief, cells were washed with assay media and incubated for 1 h at 37°C without CO₂. Glycolytic flux and cellular respiration were then quantitated by recording extracellular acidification rate (ECAR; mpH/min) and oxygen consumption rate (OCR; pmol/min), respectively. The indicated stress reagents were injected at preset time points. After the assay, cell lysates were harvested. Results were normalized by protein concentrations.

Total RNA was extracted using the RNAiso Plus (Takara) according to the manufacturer’s protocol. Reverse transcription and SYBR-based quantitative PCR (qPCR) analyses were performed as previously described (15). The 18S rRNA or GAPDH was used as internal controls for normalization. Primer sequences were generally derived from the publicized database of PrimerBank. Sequences of more frequently used primers are listed in Supplemental Fig. 1A. Others can be distributed upon request.

When not indicated specifically, cells lysis was performed in buffer containing 1% Triton-X 100, 0.5% Nonidet P-40, 150 mM of NaCl, 10 mM of Tris-HCl pH 7.5, 1 mM of EDTA, 1 mM of NaF, 1 mM of sodium orthovanadate, protease inhibitor mixture (1:500 dilution; Sigma) and freshly added 1 mM of PMSF. Western blots were performed as previously described (17).

BMDMs and PMs were dissociated from the culture vessels using PBS containing 0.5% EDTA and were incubated with fluorophore-conjugated Ab against F4/80, CD11b, or CD11c and the corresponding isotype controls. The analyses were performed on a FACSCalibur platform.

Ten-week-old mice were intranasally instilled with 200 μl of PBS or clodronate-containing liposomes (ClodronateLiposomes.com). Forty-eight hours later, cells were injected i.p. with 50 μg of poly(I:C). Twelve hours later, mice were sacrificed and lung tissues were harvested for analyses. Histological processing of the tissue and immunofluorescence microscopy were performed as previously described (17, 18).

Thymocytes were induced to undergo apoptosis (∼70%) using dexamethasone (6 h). After extensive washing, the apoptotic thymocytes were then added to PMs at a 20:1 ratio. Four hours later, cells were fixed and stained (H&E). The numbers of engulfed thymocytes/uptaking macrophage and the percentage of macrophages undergoing phagocytosis were determined in at least 10 random fields by light microscopy (>500 total cells). As expected, there was often a positive correlation between the two measurements. Phagocytic indices were calculated as the product of the latter two measurements. For measuring the extent of cargo adhesion, macrophages were incubated with a great excess (in a 1:100 ratio) of apoptotic thymocytes for 15 min (19).

The assay was adapted from an established protocol based on the use of potato-derived phosphofructose-1-kinase (PFK) that is highly sensitive to the allosteric activation by fructose-2,6-bisphosphate (F2,6BP) (20). In brief, 50 μg of cell lysates (as a source of F2,6BP) was added to an assay buffer containing the pyrophosphate-dependent, potato PFK (PPi-PFK; Sigma), aldolase, triosephosphate isomerase, GAPDH, as well as NAD⁺. The assay was initiated by adding a mixture of substrates including glucose-6-phosphate/fructose-6-phosphate and pyrophosphate. The conversion of NAD⁺ to NADH was indicated by increases in OD at 340 nm. No activity was recorded when PPi-PFK was omitted from the reaction mix.

Seven-week-old mice were instilled intranasally with 5 × 10⁶ respiratory syncytial virus (RSV; A2 subtype) in a volume of 100 μl. Five days after inoculation, mice were sacrificed and lung tissues were harvested for further analyses. To prepare bone marrow chimeras, we lethally irradiated (10 Gy) 6-wk-old C57/BL6 mice according to an established protocol (21). A total of 5 × 10⁶ WT or Pfkfb3^+/− donor bone marrow cells were injected to the recipients via tail vein. After 6 wk of recovery, the mice were subjected to RSV challenges. RSV titers were determined by plaque assay using Hela cells (22).

When not specifically indicated, data presented in this study are representatives of at least two independent experiments. Every sample for quantitative measurements was analyzed in replicates. For representative results, mean values were presented (qPCR and metabolic flux: quadruplicates [± SD]; lactate: triplicates [± SD]; viral titer and F2,6BP: duplicates [± range]; the extent of efferocytosis: 10 microscopic fields [± SD]). Some experiments were repeated more robustly to show average values (± SEM) from independent experiments, and statistical analyses (unpaired t tests) were performed. In some assays where the primary measurements tend to fluctuate while the trends of change were similar between independent experiments, paired t tests were performed. In such cases, the graphs presented are still based on representative results. Furthermore, for some experiments, statistical analyses were performed to compare the values of fold changes. For such type of data presentation, the average fold changes from multiple repetitions (± SEM) are marked on top of the graphs and p values are provided in the figure legends. In experiments involving RSV infection, data from each mouse were determined individually (± SEM) and Student t tests were performed.

Because the oxygen-independent glycolytic pathway represents a major point of regulation in cellular glucose metabolism (23, 24), we reasoned that antiviral immune signaling events may influence the rate of glycolysis. We tested this hypothesis in mouse thioglycollate-elicited PMs and MEFs, representing an innate immune cell type and nonimmune cell type, respectively. PMs and MEFs were infected with VSV (Fig. 1A, 1B), a negatively stranded RNA virus. Both cell types potently engaged innate immune signaling and the ensuing transcriptional response, as evidenced by the time-dependent induction of Ifnb1, and of Mx2, a downstream target of type I IFN (Fig. 1B). Indeed, clear accumulation of VSV-G protein was observed in PMs and more readily in MEFs, indicative of effective infection (Supplemental Fig. 1B). Notably, at both 5 and 8 h postinfection, we observed time-dependent increases of glycolysis in PMs, measured by the levels of glycolytic end-product lactate (Fig. 1A). In contrast, no such upregulation of glycolysis was observed in parallel using VSV-infected MEFs (Fig. 1A). It is plausible that the differential glycolytic responses to VSV infection in PMs and MEFs may be attributed to the macrophage-restricted TLR4-TRAM pathway trigged by the VSV-G envelope protein (25). To test this possibility, we stimulated PMs using UV-inactivated VSV. In contrast with the live VSV, its inactivated counterpart was not capable of inducing the mRNAs of IFN (Supplemental Fig. 1C), possibly because of the insufficient amount of VSV-G to engage TLR4 under the experiment condition (MOI of 1). Likewise, UV-inactivated VSV failed to trigger glycolytic enhancement in PMs (Fig. 1C). These results suggested that the observed metabolic response in PMs by live VSV was likely to be dependent on cytosolic viral RNA-sensing pathway (26). Consistently, transfection of macrophages using poly(I:C), an analog of virus-associated dsRNA (27), also resulted in increased glycolysis (Fig. 1D, Supplemental Fig. 1D).

Next, to examine the potential glycolysis-enhancement activities by viral RNA-induced secreted factors, we harvested conditioned medium from poly(I:C)-transfected MEFs. An aliquot of conditioned medium was then added, respectively, to PMs and MEFs, and analyzed for its activity (8-h treatment) to regulate cellular glycolysis (uptake of a glucose analog 2-NBDG and lactate production; Supplemental Fig. 1E, 1F). Notably, mirroring the effects by VSV infection (Fig. 1A), the extracellular fraction from poly(I:C)-transfected MEFs sufficed to enhance glycolysis in PMs, but not in MEFs. Consequently, this led us to examine the involvement of type I IFN. All type I IFNs act via one receptor composed of subunits IFNAR1 and IFNAR2. Therefore, PMs were infected in the presence of a neutralizing Ab against type I IFN receptor (anti-IFNAR1). Importantly, VSV-induced macrophage glycolysis was abrogated in IFNAR1-blocked cells (Fig. 1E, Supplemental Fig. 1G, 1H), whereas such inhibition of IFN pathway expectedly caused greater VSV-G protein accumulation. These results demonstrate a role of type I IFN, rather than cytosolic viral RNA signaling, directly responsible for VSV-induced glycolysis in macrophages.

Consistently, rIFN-β treatment of PMs for 8 h increased glucose uptake (Fig. 1F), lactate production (Fig. 1G), as well as the ECAR measured by Seahorse metabolic analyzer (Fig. 1H). Contrastingly, IFN failed to elicit a hyperglycolytic state in MEFs (Fig. 1F, 1G). To ensure that IFN-induced increase of glycolysis in PMs was not caused by secondary effects arising from changes in OXPHOS, we also assayed IFN-treated macrophages for the OCR indicative of mitochondria respiration. Despite apparent changes in glycolytic rates, the levels of OCR were comparable between treatment groups (Fig. 1I). Collectively, our results have demonstrated that innate antiviral signaling engages a type I IFN–dependent, direct induction of a hyperglycolytic state that occurs preferentially in macrophages. This represents a previously unknown, intriguing phenotypic link between antiviral immunity and cellular metabolism.

Our analyses of macrophage metabolism were carried out at a time point when a transcriptional response to IFN is expected to be prevalent (8 h). We therefore reasoned that the observed increase in macrophage glycolysis (Fig. 1F–H) is likely to be attributed to changes in the expression of glycolytic enzymes. Indeed, when the mRNA levels of a series of glycolysis-associated proteins/enzymes were analyzed, those of PFKFB3, PFK1 (platelet isoform), and lactate dehydrogenase A were found to be markedly upregulated in IFN-treated PMs, but not MEFs (Fig. 2A). The correlation between gene expression analysis (Fig. 2A) and the metabolic measurements of these two cell types (Fig. 1F, 1G) suggest increased levels of PFKFB3, PFK1, and/or LDH underlying IFN-dependent glycolytic activation in macrophages.

The dual-activity, PFKFB family of enzymes catalyze the bidirectional conversion between fructose-6-phosphate and F2,6BP. F2,6BP is an allosteric activator of PFK1, which in turn mediates a key rate-limiting step of glycolysis (Fig. 2B). In the PFKFB family, PFKFB3 exhibits a much higher kinase/phosphatase ratio than other family members (28), essentially mediating net synthesis of F2,6BP to enhance glycolysis (Fig. 2B). Therefore, we placed our attention on PFKFB3 because of its upstream regulatory status relative to PFK1 and LDH in glycolysis. When the levels of all four members of the PFKFB family were examined in macrophages and MEFs at basal states, the relative expression of Pfkfb3 mRNA was found to be particularly high in PMs (Fig. 2C). It is interesting to note that in contrast with that of Pfkfb3, the levels of Pfkfb1, 2, 4 mRNAs were downregulated by IFN in PMs (Supplemental Fig. 2A). Nevertheless, no further analyses regarding the latter PFKFB members were carried out in this study.

Consistent with the patterns of mRNA expression, PFKFB3 protein levels were also notably induced by IFN or VSV in PMs and BMDMs, but not in MEFs (Fig. 2D, Supplemental Fig. 2B), whereas the induction of a canonical IFN-stimulated gene (ISG), that is, Stat1, was comparable in all cell types. A similar macrophage-preferential induction pattern of PFKFB3 by IFN was observed using human cell lines (Supplemental Fig. 2C). Mechanistically, Jak activity was required for IFN-dependent induction of PFKFB3 in macrophages (Supplemental Fig. 2D), suggesting Pfkfb3 as a bona fide macrophage-restricted ISG.

To verify that IFN induces PFKFB3 in a macrophage-preferential manner in vivo (experiment scheme in Fig. 2E), we focused on the lung, a tissue where the macrophages are present in a significant portion, together with a diverse array of nonimmune cell types. i.p. injection of poly(I:C) can elicit a robust, systemic type I IFN response in mice (29). Such treatment led to a notable upregulation of PFKFB3 in the lung tissue, together with Stat1 (Fig. 2F). To deplete lung macrophages, we administered clodronate liposomes intranasally to the mice. Immunofluorescence staining of macrophage marker F4/80 and qPCR analysis of Cd68 confirmed the efficiency of macrophage depletion in the lung tissue (Fig. 2G, Supplemental Fig. 2E). Importantly, the liposome treatment markedly reduced i.p. poly(I:C)-induced upregulation of PFKFB3 (Fig. 2F). In comparison, upregulation of Stat1, a canonical ISG product, was not affected in the lung tissue. Similar contrasting patterns were observed between the levels of Pfkfb3 mRNA and those of another canonical ISG, that is, Isg15 (Supplemental Fig. 2E). Collectively, the earlier results verified macrophages as a major cell type that uses the IFN-PFKFB3 axis.

To functionally validate the role of PFKFB3 in macrophage metabolism, we first used a stable macrophage cell line (Raw264.7) transduced with a tetracycline-inducible PFKFB3 expression cassette (Fig. 3A, 3B). Doxycycline treatment induced PFKFB3 expression to a similar extent achieved by IFN in primary macrophages (see Fig. 2D) and led to upregulation of glycolysis (Fig. 3A). To examine the metabolic function of endogenous PFKFB3, we generated heterozygous Pfkfb3-knockout mice. Cas9-mediated cleavage caused a 73-bp indel in exon 2 of Pfkfb3 that led to frameshift in the mutant allele (Supplemental Fig. 3A). Accordingly, the levels of Pfkfb3 mRNA and its encoded protein in multiple tissues from the heterozygous mutant mice were visibly reduced (Fig. 3C, 3D). Pfkfb3^+/− mice were morphologically normal, consistent with a previous report (30). In addition, equivalent numbers of BMDMs or PMs could be derived from the WT and the Pfkfb3^+/− mice, and the cells showed similar patterns of macrophage-associated cell-surface markers (Supplemental Fig. 3B, 3C). Furthermore, compared with the WT counterparts, Pfkfb3^+/− BMDMs engaged similar induction of Ifnb1 and two ISGs (Mx2 and Isg15) in response to transfected dsRNA, showing an intact IFN system (Supplemental Fig. 3D). Nevertheless, Pfkfb3^+/− macrophages exhibited a visible reduction in glycolysis, measured by lactate levels (Fig. 3E, left panel, Supplemental Fig. 3E) and by ECAR (Fig. 3F, Supplemental Fig. 3F). These metabolic phenotypes correlate with the levels of PFKFB3 protein (Fig. 3E, right panel, Supplemental Fig. 3E). Despite such an apparent change in glycolysis (∼25% in lactate levels), the Pfkfb3^+/− macrophages did not exhibit evident changes in cellular respiration (Fig. 3G, Supplemental Fig. 3G). Therefore, our results confirm PFKFB3 as an important glycolytic regulator in macrophages. Compared with the macrophages, Pfkfb3^+/− MEFs showed a more modest (∼13%) decrease in overall glycolytic rates (Supplemental Fig. 3H), likely attributed to a lack of specific enrichment of PFKFB3 (among all PFKFBs) (Fig. 2C). Conversely, forced expression of PFKFB3 in MEFs increased the basal glycolytic rates, while the cells were still unable to enhance glycolysis in response to IFN treatment (Supplemental Fig. 3I).

The macrophage-preferential, IFN-dependent induction of PFKFB3 hinted its role in regulating viral resistance of this particular cell type. In contrast, PFKFB3-driven glycolysis may paradoxically play a proviral role via fueling biosynthesis and other critical steps of viral life cycles (31–35). Interestingly, Pfkfb3 heterozygosity in BMDMs or MEFs did not substantially affect their viral burdens (Fig. 4A, 4B, Supplemental Fig. 4A), indicating that a moderate reduction of glycolysis did not suffice to markedly impact either VSV replication or the intrinsic antiviral activities in these two cell types.

Compared with the nonimmune MEFs, macrophages may additionally exert cell-extrinsic antiviral function, where they can restrict viral load in a non–cell-autonomous manner (36, 37). We therefore considered an additional possibility that PFKFB3 may regulate the cell-extrinsic antiviral activities of macrophages. To this end, the WT or Pfkfb3^+/− BMDMs were incubated directly with replicates of virus-infected C57/BL6 MEFs. After overnight coculture, the supernatant was subsequently analyzed for viral titer. As a control, incubation of naive MEFs with virus-infected MEFs fueled secondary infections, causing further viral production (data not shown). In contrast, coculture of BMDMs with infected MEFs leads to a substantial decrease of released virus, reflecting an extrinsic antiviral activity by the BMDMs (Fig. 4C). Importantly, WT BMDMs exhibited a greater cell-extrinsic antiviral activity compared with the Pfkfb3^+/− BMDMs (Fig. 4C). Contrastingly, the induction of Ifnb1 and several ISGs was comparable between the two coculture groups (Fig. 4D). When the IFN responses of VSV-infected BMDM monocultures were directly measured, the WT and Pfkfb3^+/− BMDMs were found to secrete comparable levels of IFN-β protein (Fig. 4E) and to engage similar upregulation of ISG mRNAs (Fig. 4F). Furthermore, despite previous suggestions of NO as a potential effector mediating macrophage cell-extrinsic antiviral activity (38), we did not detect its enhanced production by BMDMs in the coculture system (data not shown).

We next used a complementary pharmacological approach to target the catalytic activity of PFKFB3 in vitro. A recently established PFKFB3 inhibitor, that is, PFK15 (39), visibly inhibited 2-NBDG uptake in BMDMs and MEFs (Supplemental Fig. 4B, 4C). Although an effective dose (10 μM) of PFK15 did not affect VSV production in either cell type alone (Fig. 4G, 4H), similar treatment of BMDM/infected MEFs coculture reproducibly resulted in a higher overall viral burden (Fig. 4I). In contrast, no changes in IFN pathway were apparent (Supplemental Fig. 4D). These results correlate well with the data from genetically modified BMDMs (Fig. 4A–D), and in conjunction present a model where PFKFB3-driven glycolysis promotes macrophage-extrinsic antiviral activities without affecting IFN pathway.

Through their potent phagocytic activity, macrophages remove the pathogen particles as well as host-derived apoptotic cells (40–42). The latter is referred to as efferocytosis. Efferocytosis of virus-infected, apoptotic cells may serve as an important cell-extrinsic antiviral mechanism to prevent viral spread (43–45). Indeed, when the WT or Pfkfb3^+/− BMDMs were incubated with VSV-infected MEFs (labeled with CSFE) and subjected to immunofluorescence microscopy, a higher content of MEF corpses was found within the WT BMDMs (Fig. 5A). Consistent results were obtained when engulfment of infected MEFs were quantified via flow cytometry (Supplemental Fig. 4E). To prevent efferocytosis of infected MEFs by BMDMs, we first physically segregated the two cell types using permeable membranes (3-μm pore-sized transwell culture system) that otherwise allowed for free movement of soluble factors and viruses. Consequently, the cell-extrinsic antiviral effects by BMDMs were markedly reduced and viral burden became comparable between the WT and mutant BMDM coculture groups (Fig. 5B). Similar results were also obtained comparing the direct or transwell-separated coculture of PMs with infected MEFs (Supplemental Fig. 4F).

We next sought for a more specific method of efferocytosis inhibition. Macrophages recognize apoptotic cells via multiple redundant phosphatidylserine receptors including BAI1, TIM4, and STABILIN-2 whose expression patterns vary in different types of macrophages (41, 42). However, we took advantage of the fact that TIM4 functions as the major phosphatidylserine receptor in BMDMs (46, 47). Indeed, we verified that a blocking Ab against TIM4 notably reduced BMDM-mediated efferocytosis activity against apoptotic thymocytes (Fig. 5C), as well as against VSV-infected MEFs (Supplemental Fig. 4G). Importantly, Ab-mediated TIM4 blockage not only moderately inhibited the cell-extrinsic antiviral activity of BMDMs, but also reduced the difference in viral burden between the WT and Pfkfb3^+/− coculture groups (Fig. 5D). Collectively, our data establish that PFKFB3 promotes the cell-extrinsic antiviral effector function of macrophages in an efferocytosis-dependent manner.

Next, to further generalize the contribution of PFKFB3 to efferocytosis, the WT and Pfkfb3^+/− macrophages were added with standard cargo of apoptotic thymocytes. Indeed, Pfkfb3^+/− BMDMs exhibited a >2-fold decrease in basal efferocytosis activities compared with WT BMDMs, which was not attributed to their differences in cargo adherence (Supplemental Fig. 4H). A similar efferocytosis defect was observed in Pfkfb3^+/− PMs (Supplemental Fig. 4I). Although IFN treatment of PMs markedly increased their efferocytosis activity as shown previously (45, 48), IFN-treated Pfkfb3^+/− PMs were still less potent in efferocytosis than similarly treated WT PMs (Fig. 5E). Furthermore, the efferocytosis activities of PMs were respectively enhanced and suppressed by established hyperglycolytic (high glucose culture medium) and hypoglycolytic (2-deoxyglucose treatment) conditions (Fig. 5F), clearly establishing a role of glycolytic metabolism in efferocytosis.

Particle engulfment after phagocytic receptor activation involves rapid extension of cellular processes to eventually enclose the cargos. Subcellularly, such a dynamic sequence involves active polymerization of actin filaments that is an energy-demanding process (49, 50). Therefore, we hypothesized that PFKFB3-driven glycolysis may be coupled to efferocytosis-associated signaling to provide instant energy support (i.e., rapid generation of ATP) for cargo engulfment. The latter hypothesis was supported by the fact that efferocytosis is delicately regulated by phospholipid signaling activators including PI3K (40, 47), which is known to engage glycolysis (23). As expected, addition of apoptotic thymocytes into a PM culture led to a time-dependent phosphorylation of AKT and its target PRAS40, indicative of PI3K activation (Fig. 5G). Importantly, such treatment of PMs also resulted in a PI3K-dependent increase in the levels of F2,6BP (Fig. 5H), possibly resulting from AKT-mediated phosphorylation/activation of PFKFB3 as suggested previously (51). Correlating with F2,6BP measurements, the levels of 2-NBDG uptake were also upregulated after initiation of efferocytosis, in a PI3K-dependent manner (Fig. 5I). Furthermore, consistent with a role of IFN-induced PFKFB3 to promote such metabolic engagement, IFN pretreatment could further increase apoptotic thymocyte-triggered 2-NBDG uptake (Supplemental Fig. 4J). These results strongly support that PFKFB3-driven glycolysis provides a metabolic basis for optimal efferocytosis and serves as a critical regulatory node. In contrast, despite a recent report suggesting a role of PFKFB3 as a feed-forward regulator of PI3K signaling (52), no significant defects in the latter pathway were observed in Pfkfb3^+/− PMs (Supplemental Fig. 4K).

Our in vitro data thus far showed the antiviral, immunometabolic aspects of PFKFB3-driven glycolysis. To establish the in vivo role of PFKFB3 under viral infection, we used a mouse model of RSV infection. It was shown previously that lung macrophages contributed to controlling the initial RSV load (53–55). Because RSV mainly replicates in epithelial cells (55), this in vivo model was conducive for examining the cell-extrinsic antiviral function of macrophages. WT and Pfkfb3^+/− mice (n = 5 for each group) were challenged with RSV and the lung tissues were harvested 5 d later. Importantly, the Pfkfb3^+/− mice had significantly higher viral burdens compared with the WT mice (Fig. 6A, 6B), clearly demonstrating the antiviral function of PFKFB3 in vivo. In comparison, the markers indicative of macrophages were elevated to similar extent in the WT and mutant lungs (Fig. 6C). Interestingly, the mRNAs of several ISGs were moderately higher in the mutant lungs (Fig. 6C), likely to be resulting from a persistent, greater viral load.

Lastly, we sought to verify the contribution of hematopoietic compartment-originated PFKFB3 to viral resistance by using WT or Pfkfb3^+/− bone marrow chimeras. The mice reconstituted with either the WT or Pfkfb3^+/− bone marrow were subsequently subjected to RSV challenge (Fig. 6D–F). Indeed, Pfkfb3 heterozygosity in the hematopoietic compartment caused a visible increase in viral burden (Fig. 6D, 6E), without apparently affecting the induction of Cd68 mRNA indicative of infiltrated macrophages (Fig. 6F). We notice that the differences in viral susceptibility between the WT and Pfkfb3^+/− bone marrow–chimeric mice were less significant than those from nontransplanted mice (compare Fig. 6A and 6D). It is possible that the chosen recovery period after bone marrow transplantation (6 wk) was not yet sufficient for complete reconstitution of resident alveolar macrophages (56), where PFKFB3 may exert initial cell-extrinsic antiviral function. Nevertheless, these in vivo results corroborate our in vitro data and supported a role of PFKFB3 in promoting cell-extrinsic antiviral effector function of innate immune cells.

Viruses hijack the host cells’ synthetic machineries to support various stages of their life cycles. Because pathways of energy metabolism can regulate biosynthesis, as well as various steps associated with viral replication/assembly/propagation, they are believed to significantly influence the infectivity of intruding viruses (31–35). However, few studies have explored whether an optimal antiviral innate immune program also requires significant metabolic inputs.

In this study, we have provided evidence that cytoplasmic viral recognition, via the action of type I IFN, induces a PFKFB3-driven hyperglycolytic state in primary macrophages, but not in the nonimmune MEFs (Figs. 1, 2). Such a macrophage-preferential effect by IFN is consistent with a model of “immune-centric” metabolic reprogramming that serves to enhance the functions of innate immune cells. Importantly, despite an apparent effect on glycolysis, the IFN-PFKFB3 axis did not affect cellular respiration. This is different from a recently described scenario in dendritic cells where a dominant inhibition of OXPHOS coupled with a corresponding increase of glycolysis upon in vivo poly(I:C) stimulation (11). Given the known functional diversity of IFN (57), the macrophage-preferential IFN-PFKFB3 axis therefore represents a unique system to study the delicate cross-talks between metabolism and immunity.

Methodologically, we exploited a convenient Pfkfb3 heterozygotic knockout mouse model (Fig. 3), which was morphologically normal, unlike the homozygous mutants (30). Despite a moderate reduction (∼25%) in glycolysis, single allele loss of Pfkfkb3 in macrophages resulted in little compensatory changes in cellular respiration. An even lesser decrease in glycolysis was observed in Pfkfb3^+/− MEFs. These results are consistent with the role of PFKFB3 as a regulator, rather than a mediator, of glycolysis (50, 58) and in conjunction validate Pfkfb3^+/− mice/cells as a straightforward partial loss-of-function model.

Functionally, PFKFB3 selectively regulates efferocytosis-dependent, macrophage-specialized extrinsic antiviral activity (Figs. 4, 5), which correlates well with its macrophage-preferential induction by IFN. In addition to preventing viral propagation, efferocytosis of infected or damaged cells is also likely to contribute to limiting virus-induced immunopathology via physical actions and secretion of anti-inflammatory cytokines (41, 42). Pfkfb3^+/− macrophages exhibited lower efferocytosis ability, suggesting that the events leading to cargo engulfment are critically sensitive to changes in PFKFB3 levels and glycolysis. Mechanistically, the efferocytosis process involves active actin filaments assembly (41, 42), where PFKFB3-driven glycolytic reaction may be involved to supply the required ATP, similar to an earlier model for endothelial tip cell migration (50). Our data extend the latter model to further suggest that PFKFB3 actively participates in efferocytosis via establishing the temporal and spatial association between glycolytic metabolism and actin polymerization (Fig. 5G–I). In addition to metabolically supporting efferocytosis, PFKFB3 may possibly also affect other downstream events. Interestingly, efferocytosis was previously shown to interact with the autophagic pathway (46). Because autophagy is intricately connected to the cells’ metabolic states (59) and it is well-known to control viral resistance (60), PFKFB3-regulated extrinsic antiviral activity may potentially involve changes in autophagic pathway. Such a possibility awaits future investigation. Nevertheless, our study has provided strong evidence for a rather intuitive model linking PFKFB3-driven glycolysis to macrophage phagocytic behavior. Such a new immunometabolic role by PFKFB3 complements previous reports that a TLR4-PFKFB3 axis modulates the M1 macrophage gene expression circuitry (7, 61). It is also worth noting that because most TLRs can trigger IFN production to varying degrees (1), the IFN-PFKFB3 axis may be commonly engaged by TLRs. Our data based on a TLR3 activation model indeed support such a notion (Supplemental Fig. 2F).

In retrospect, it is interesting to note that the lipid metabolic pathways were previously known as critical regulators of the efferocytosis process (62). In contrast, glucose metabolism has been surprisingly underinvestigated in the field. Interestingly, different from our observation, an earlier study using a phagocytic Chinese hamster ovary cell line showed that high concentration of glucose had a negative effect on efferocytosis (63). We speculate that the cellular contexts, that is, macrophages versus other cell types, primary versus transformed cells, or even differences in macrophage activation states, may markedly influence the metabolism/efferocytosis connections. Because different cell identities/fates are characterized by distinctive metabolic landscapes (24), the earlier discrepancies in others’ and our findings may suggest an intriguing possibility that interactions among various metabolic pathways coordinately control cells’ efferocytosis behaviors. The latter appears as an exciting avenue for future research.

By using an RSV infection mouse model, we have confirmed the antiviral function of PFKFB3, and at least some contribution of its action from the hematopoietic compartment (Fig. 6), in agreement with our in vitro results. Even though the RSV susceptibility phenotype associated with Pfkfb3 heterozygosity was moderate, it demonstrates that tuning the levels/activities of PFKFB3 is sufficient to alter the virus–host balance. It is formally possible that PFKFB3 may exert other uncharacterized antiviral functions dependent on the viral and cellular contexts. Further efforts to test the latter notion in diverse viral infection models are highly warranted. Nevertheless, our data have provided genetic evidence for an antiviral aspect of glycolysis at the organism level, a significant finding considering that many previous reports have suggested glycolysis as a proviral pathway (34, 35). In this regard, it is worth investigating whether the featured expression of PFKFB3 (compared with other PFKFB isoforms; Fig. 2C) in macrophages, an innate immune cell type, may contribute to its predominant role in immunometabolism, but not virus-promoting metabolism.

Overall, our work has established PFKFB3-driven glycolysis as a critical metabolic effector mechanism of innate antiviral immunity. Importantly, the in vivo data from this study challenge the simplified, virus-central view of cellular metabolism and highlight the significance of immunometabolism in innate antiviral defense. Interestingly, PFKFB3 was also suggested previously as a positive regulator of T cell functions (64). Therefore, we suggest that PFKFB3 may represent a promising drug target whose therapeutic activation may enhance both the innate and the adaptive arms of antiviral response.

We are grateful to the laboratories of G. Liu and Q. Zhang (Nanjing Univ.) for technical help. We thank the Nanjing Biomedical Research Institute-NJU and the animal facility of NJU Medical School for excellent mouse services. We acknowledge the Collaborative Liver Disease Research Program of NJU Medical School for instrumental support.

The authors have no financial conflicts of interest.