Therapeutic targeting of SPIB/SPI1‐facilitated interplay of cancer cells and neutrophils inhibits aerobic glycolysis and cancer progression

Abstract Background As a metabolic reprogramming feature, cancer cells derive most of their energy from aerobic glycolysis, while its regulatory mechanisms and therapeutic strategies continue to be illusive. Methods Integrative analysis of publically available expression profile datasets was used to identify critical transcriptional regulators and their target glycolytic enzymes. The functions and acting mechanisms of transcriptional regulators in cancer cells were investigated by using in vitro and in vivo assays. The Kaplan–Meier curve and log‐rank assay were used to conduct the survival study. Results Salmonella pathogenicity island 1 (SPI1/PU.1), a haematopoietic transcription factor, was identified to facilitate glycolytic process, tumourigenesis, invasiveness, as well as metastasis of colon cancer cells, which was interplayed by tumour‐associated neutrophils. Mechanistically, neutrophils delivered SPI1 mRNA via extracellular vesicles, resulting in enhanced SPI1 expression of cancer cells. Through physical interaction with SPI1‐related protein (SPIB), SPI1 drove expression of glycolytic genes within cancer cells, which in turn induced polarization of neutrophils via glycolytic metabolite lactate. Depletion of neutrophils or SPIB–SPI1 interaction in cancer cells significantly inhibited glycolytic process, tumourigenesis and aggressiveness. Upregulation of SPI1 or SPIB was found to be associated with poor prognosis in patients suffering from colon cancer. Conclusions Therapeutic targeting of SPIB/SPI1‐facilitated interplay of cancerous cells and neutrophils suppresses aerobic glycolysis and progression of cancer.

Conclusions: Therapeutic targeting of SPIB/SPI1-facilitated interplay of cancerous cells and neutrophils suppresses aerobic glycolysis and progression of cancer.

K E Y W O R D S
aerobic glycolysis, cancer progression, extracellular vesicles, neutrophil, Salmonella pathogenicity island 1, SPI1-related protein

BACKGROUND
Glucose metabolism reprogramming is a feature of cancer cells to sustain rapid proliferation and aggressiveness. 1 Cancer cells utilize glucose to produce large amounts of adenosine triphosphate (ATP) and lactate in spite of sufficient oxygen supply, a process termed as Warburg effect or aerobic glycolysis. 1 For several decades, Warburg effect has been confirmed to be essential for tumour growth and aggressiveness, 1 while pharmacological inhibition of essential glycolytic enzymes, including pyruvate kinase M2 (PKM2) or hexokinase 2 (HK2), is a potential approach for inhibiting aerobic glycolysis of cancers. 2 Recent evidence shows that intratumoural neutrophil infiltration is a prognostic factor for certain cancers, such as renal cancer 3 or squamous cancer of head and neck. 4 However, this phenomenon is not universal, since certain lung cancer presents moderate or no neutrophils. 5 Tumour-associated neutrophils (TANs) are almost entirely recruited from circulating blood and play essential roles in tumour initiation, growth and metastasis. 6,7 Similar to macrophages, neutrophils possess dual roles in cancer progression, including anti-tumourigenic (N1) or protumourigenic (N2) phenotype. 8,9 As a major proximal cytokine within tumours, transforming growth factor beta 1 (TGFβ1) decreases cytotoxicity of neutrophil and mediates polarization of N2 neutrophils from N1 type. 7 In addition, estradiol prolongs survival and promotes N2 polarization of neutrophils by upregulating lymphocyte function-associated antigen 1 integrin. 10 Meanwhile, inhibition of C-X-C motif chemokine receptor 2 or clearance of neutrophils significantly impairs tumour development and progression. 11,12 However, mechanisms underlying interplay of cancer cells and neutrophils during aerobic glycolysis and cancer progression warrant further investigation. As transcription factors of ETS-transformation specificity (ETS) family, Salmonella pathogenicity island 1 (SPI1/PU.1) and SPI1-related protein (SPIB) share overlapping expression pattern and execute a critical role in B-cell development. 13 SPI1 is required for myeloid and lymphoid lineage commitment and maturation, and its deregulation leads to development of leukaemias or lymphomas. 14 SPI1 activates transcription of target genes through direct binding to their promoters via its ETS-domain 15 or cooperation with other DNA binding proteins, such as GATA binding protein 1 (GATA1) 16 or p53/p73. 17 Human SPIB was identified as a nuclear protein highly homologous with SPI1. 18 Adjacent DNA binding sites of SPIB and SPI1 within genome implicate their similar functions or mutual promoting effects in regulating gene expression. 19 SPIB cooperates with transcription factor SPI1 to participate in maturation and proliferation of B lymphocytes through regulating p50 expression. 20 Meanwhile, the roles of SPI1 and SPIB in glucose metabolism reprogramming or reciprocal interplay of malignant tumour cells with microenvironment still remains elusive.
In the current study, we discover that SPI1 is substantially expressed within tumoural tissues as well as stroma of colon cancer, while neutrophils deliver SPI1 mRNA to cancer cells via extracellular vesicles (EVs), leading to abnormal expression of SPI1 in cancer cells. Through physical interaction with its homologous partner SPIB, SPI1 is activated to promote aerobic glycolysis of cancer cells via upregulating HK2 or phosphoglycerate kinase 1 (PGK1), which in turn induces N2 polarization of neutrophils via glycolytic metabolite lactate. Depletion of neutrophils or blocking interaction between SPIB and SPI1 dramatically reduces glycolytic process, growth, invasiveness and metastatic capabilities of cancer cells, suggesting essential functions of SPIB/SPI1 for interplay of cancer cells and neutrophils in aerobic glycolysis and cancer progression.

SPI1 is elevated and facilitates glycolytic gene expression in colon cancer
To investigate potential regulators for aerobic glycolysis during cancer progression, we performed comprehensive analysis of gene expression profiles in colon cancer and normal counterparts, utilizing microarray results available from Gene Expression Omnibus (GEO). The results indicated that compared with normal colon tissues, 25 glycolytic genes were upregulated (fold change > 1.5) in colon cancer (Table S1). In addition, based on gene expression profiles among components of colon cancer tissues (GSE35602), seven glycolytic genes were found to be elevated (fold change > 1.5) in cancerous cells than stroma (Table S1). Overlapping analysis of these results revealed six glycolytic genes highly expressed in colon tissues ( Figure 1A). In a similar way, 52 transcriptional regulators were found to be upregulated (fold change > 1.5) in colon cancer and differentially expressed in cancerous cells and stroma ( Figure 1A and Table S1). Additional overlapping analyses with those governing six glycolytic enzymes determined by ChIP-X software 21 revealed four transcription factors potentially participating in the regulation of these glycolytic genes' expression ( Figure 1A), including ETS-related gene (ERG), SRY-box transcription factor 17 (SOX17), SPI1 and TAL bHLH transcription factor 1 (TAL1). SPI1 ranked first among them in terms of the number of target genes, including HK2 and PGK1, which was more enriched in stroma ( Figure 1A, B). To understand SPI1's impact on glycolysis, we chose colon cancer cell lines LoVo and HCT116 (representing middle SPI1 levels) as models ( Figure S1A). Chromatin immunoprecipitation (ChIP), quantitative PCR (qPCR), as well as dual-luciferase reporter assays revealed that in LoVo and HCT116 cells, stable transfection or silencing of SPI1 led to an increase or decrease in SPI1 enrichment as well as HK2 and PGK1 promoter activity, respectively (Figures 1C, D and S1B). In addition, forced or impaired SPI1 expression respectively enhanced or reduced the levels of HK2 and PGK1 in cancer cells ( Figures 1E and S1C). In colon cancer specimens, a positive expression correlation was noted between SPI1 and HK2 (R = 0.1067, p < 1 × 10 −4 ) or PGK (R = 0.4026, p < 1 × 10 −4 , Figure 1F). Importantly, upregulation of SPI1 (p = 2.2 × 10 −3 ), HK2 (p = 1.0 × 10 −2 ) or PGK (p = 2.9 × 10 −4 ) was found to be related with a poor survival in colon cancer patients ( Figure 1G). Above data suggested that SPI1 was elevated and facilitated glycolytic gene expression in colon cancer.

SPI1 promotes cancer progression via facilitating aerobic glycolysis
We further explored the SPI1's impacts on glycolytic process and biological behaviours of cancer cells. In LoVo and HCT116 cells, stable transfection or silencing of SPI1 led to increased and decreased extracellular acidification rate (ECAR), a glycolysis indicator, while oxygen consumption rate (OCR) was reduced and elevated, respectively, along with elevation and reduction in glucose uptake, lactate generation and ATP synthesis (Figures 2A, B and S1D). Stable over-expression or silencing of SPI1 promoted and attenuated the growth and invasion of cancer cells, respectively ( Figure S1E, F). Notably, knocking down HK2 or PGK1 rescued the upregulation of HK2 and PGK1 (Figure S2A), and augment in glucose uptake, lactate generation, ATP synthesis, proliferation and invasiveness of LoVo cells following steady ectopic expression of SPI1 ( Figure  S2B-D).
For investigating the involvement of aerobic glycolysis in SPI1-promoted tumourigenesis and aggressiveness, glycolysis inhibitor or activator was applied in cultured cancer cells. The glycolysis inhibitor 2-deoxy-glucose (2-DG) 22 prevented the ectopic expression of SPI1-induced enhancement in glucose uptake, lactate generation, ATP synthesis, proliferation and invasiveness of LoVo cells ( Figures S2E  and 2C, D). Meanwhile, treatment of insulin-like growth factor-1 (IGF-1) 22 rescued the decrease of aerobic glycolysis, proliferation, and invasiveness of HCT116 cells stably silencing SPI1 ( Figure S2E-G).
To assess in vivo effects of SPI1 on cancer progression, LoVo cell line was injected subcutaneously or into tail vein of athymic nude mice. Small animal imaging experiment indicated that fluorescence intensity was significantly higher within subcutaneous tumours generated using cancer cells stably transfected with SPI1 ( Figure 2E). There was a significant elevation in growth, weight, Ki-67 expression and CD31-staining microvessel density within subcutaneous xenograft tumours generated by cancer cells stably over-expressing SPI1 in athymic mice ( Figure 2E, F). In experimental metastasis assay, higher fluorescence signals, increased number of lung metastasis, as well as lower survival possibility were noted in athymic mice receiving administration of LoVo cells with stable SPI1 over-expression via tail vein ( Figure 2G). Meanwhile, administration of 2-DG counteracted the oncogenic roles of SPI1 in driving growth and metastasis of LoVo cells in vivo ( Figure 2E-G). Collectively, these data revealed that SPI1 promoted cancer progression via facilitating aerobic glycolysis.

Neutrophils facilitate SPI1-mediated aerobic glycolysis, tumourigenesis and aggressiveness
Since above results revealed abundance of SPI1 in tumour stroma, and considering its roles as a master regulator of neutrophil differentiation, 23 we further investigated the impact of neutrophils on aerobic glycolysis and cancer progression. Immunohistochemical staining revealed increase of Ly6G + neutrophils within xenograft tumours and metastatic lungs formed by LoVo cells stably over-expressing SPI1 ( Figure 3A). In TANs derived from human colon cancer tissues, CD66b and CD11b levels were significantly increased than those of peripheral neutrophils (PNs, Figure S3A). The CD66b levels in TANs were increased and decreased by co-culture with LoVo cells stably over-expressing or silencing SPI1, respectively (Figure S3B). Interestingly, lactate was able to induce CD66b expression and polarization of TANs ( Figure S3C, D), which was attenuated by knockdown of SPI1 in co-cultured LoVo cells ( Figure S3D). Higher SPI1 levels were noted in TANs, than those of PNs or colon cancer cells ( Figure S3E). On the other hand, co-culture with TANs led to increase of SPI1 expression and its enrichment on HK2 or PGK1 promoter in LoVo and HCT116 cells ( Figures 3B, S3F and S4A), resulting in increase of HK2 or PGK1 promoter activity and transcript levels ( Figure S4B, C), while silencing of SPI1 in cancer cells reduced these effects (Figures 3C and S4A-C). Glycolytic capacity and metabolite measurement assays indicated that conditional medium of TANs significantly facilitated glycolytic process of cancer cells (Figures 3D and S4D, E), while silencing of SPI1 suppressed these effects in cancer cells ( Figures 3D and S4D, E). Conditional medium from TANs significantly facilitated the proliferation and invasiveness of LoVo and HCT116 cells, which was abolished after knockdown of SPI1 ( Figure S4F, G). To further reveal the functions of TANs during SPI1facilitated cancer progression, LoVo cells with stable transfection of either empty construct or SPI1 were injected subcutaneously or via tail vein into athymic nude mice, which were subsequently treated with anti-Ly6G blocking antibody, an established approach for depleting neutrophils. 6 In vivo tumour growth, weight, glucose uptake, lactate generation, ATP synthesis, downstream gene expression, Ly6G + neutrophils, Ki-67 expression and CD31-staining microvessel density of hypodermic xenograft models generated by cancer cells were enhanced upon stable SPI1 over-expression, which was prevented by anti-Ly6G antibody ( Figures 3E, F and S5A). In experimental metastasis assay, increased number of lung metastasis as well as lower probability of survival was found in athymic mice received injection of LoVo cells stably over-expressing SPI1 via tail vein, whereas administration of anti-Ly6G antibody abolished these changes ( Figure 3G). These results indicated that neutrophils facilitated SPI1-mediated glycolytic process, growth and invasiveness of cancer cells.

Neutrophils deliver SPI1 mRNA into cancer cells via extracellular vesicles
We further extracted EVs from PNs or TANs, which were validated by electron microscopic observation, particle size analysis ( Figure 4A), and western blot assay of surface markers (CD9 and CD63, Figure 4B). There were increased levels of SPI1 mRNA in EVs extracted from TANs than those of PNs, without detectable SPI1 protein (Figure 4B, C). Meanwhile, knockdown of SPI1 in TANs via short hairpin RNA (shRNA) led to reduced SPI1 transcript levels within secreted EVs ( Figure 4C). Immunofluorescence observation showed that Dil-labelled EVs derived from neutrophils were transferrable to LoVo cells (Figure 4D). Importantly, administration of TANs-secreted EVs increased the levels of SPI1 and its binding to target gene promoter regions, leading to upregulation of HK2 and PGK1 in LoVo and HCT116 cells, which was reduced upon silencing of SPI1 in TANs ( Figures 4E and S5B-E). Similarly, treatment with LoVo cells with EVs secreted by mouse xenograft tumours-isolated TANs, but not with those from murine PNs, elevated the expression levels of HK2 and PGK1, while silencing of Spi1 in TANs prevented these changes ( Figure S5F, G). In addition, treatment with TANs-secreted EVs promoted aerobic glycolytic process ( Figure 4F), glucose uptake, lactate generation, ATP synthesis, proliferation and invasiveness of cancer cells ( Figure 4F-H). These data suggested that neutrophils delivered SPI1 mRNA into cancer cells via EVs.

SPIB physically interacts with SPI1 in cancer cells
For clarifying mechanisms for oncogenic functions of SPI1, we screened its protein partner by immunoprecipitation, Coomassie blue staining and mass spectrometry assays ( Figure S6A), which revealed 577 proteins pulled down by SPI1-specific antibody in LoVo cells ( Figure 5A and Table S2). Overlapping analysis of SPI1-interacting protein derived from BioGRID 24 and IID 25 databases identified SPIB and GATA binding protein 2 (GATA2) as potential partners ( Figure 5A). Validating co-immunoprecipitation (co-IP) assay indicated endogenous interaction of SPI1 with SPIB, but not with GATA2, in cultured cancer cells ( Figure 5B). Immunofluorescence staining indicated nuclear co-localization of SPI1 and SPIB in LoVo cells (Figure 5C). Deletion-mapping experiments using tagged or recombinant proteins revealed that ETS domains of SPI1 and SPIB were essential for their interaction (Figures 5D and S6B, C). Based on three-dimensional structure analysis using ZDOCK program, 26 amino acid residues of SPI1 (185th aspartic acid, 189th serine and 190th isoleucine) and SPIB (183th aspartic acid, 187th cysteine and 188th proline) protein were predicted to mediate their interaction. Mutation of these residues abolished the interaction between SPI1 and SPIB in cancer cells ( Figure 5E). For direct visualization of their interaction, bimolecular fluorescence complementation (BiFC) assay [27][28][29][30] was performed, which revealed obvious fluorescence in cancer cells co-transfected with vectors of SPI1 and SPIB, but not with their mutant constructs ( Figure 5F). These findings suggested that SPIB bound to SPI1 via physical interaction in cancer cells.

2.6
SPIB coordinates with SPI1 to promote aerobic glycolysis and cancer progression We further explored cooperative roles of SPIB and SPI1 in aerobic glycolysis and cancer progression. In LoVo and HCT116 cells, forced expression or silencing of SPIB boosted or lowered SPI1 transactivation ( Figure 6A). There was increase and decrease in SPI1 enrichment, promoterluciferase reporter activity, as well as levels of HK2 and PGK1 in cancer cells stably over-expressing or silencing SPIB, which was eliminated by silencing or ectopic expression of SPI1, respectively (Figures 6B-E and S7A-C). Next, we examined cooperative effects of SPI1 and SPIB on aerobic glycolysis. In HCT116 and LoVo cells, stable overexpression or knockdown of SPIB boosted or reduced the glycolytic process, glucose uptake, lactate generation and ATP synthesis, respectively, which were eliminated by silencing or over-expression of SPI1 ( Figures 6F and S7D-H). Forced expression or silencing of SPIB rescued the changes in proliferation and invasiveness of cancer cells caused by SPI1 knockdown or over-expression (Figures 6G, H and S7I, J). To assess in vivo interplay of SPIB and SPI1, HCT116 cells with stable expression of red fluorescent protein were injected subcutaneously or into the tail vein of athymic mice. Stable over-expression of SPIB led to elevation in tumour growth, weight, glucose uptake, lactate generation, ATP synthesis, downstream gene expression, Ly6G + neutrophils, Ki-67 expression and CD31-staining microvessel density of cancer cells-generated hypodermic xenograft models, which was abolished by knockdown of SPI1 (Figures 6I and S8A-C). Strong fluorescence signals, higher number of lung metastasis and neutrophil infiltration and poorer survival were noted in athymic nude mice following administration of HCT116 cell line over-expressing SPIB, which was conversed by silencing of SPI1 ( Figure 6J). Taken together, these data revealed that SPIB coordinated

Therapeutic blocking SPIB-SPI1 interaction suppresses glycolytic process and cancer progression
For elucidating treatment efficiency of blocking SPIB-SPI1 interaction, an inhibitory peptide of 11 amino acids was designed by using ROSIE program, 31 and termed as SIP-11 ( Figure 7A). Meanwhile, a peptide (CTLP) with mutation of amino acid residues was also applied as a control. Administration of SIP-11 led to its uptake and nuclear distribution within cancer cells ( Figure 7B). Biotinylatedpeptide pull-down studies indicated the ability of SIP-11, rather than CTLP, in directly binding to SPIB protein ( Figure 7C). BiFC and co-IP assays indicated that SIP-11 treatment abolished the interaction between SPI1 and SPIB in colon cancer cells ( Figure 7D, E). In addition, treatment with SIP-11, but not with CTLP, abolished direct interaction between recombinant GST-tagged SPI1 and His-tagged SPIB protein ( Figure S9A). Accordingly, administration of SIP-11 led to decrease in SPI1 transactivation and enrichment, promoter-luciferase reporter activity, as well as levels of HK2 or PGK1 (Figures 7F and S9B-F), than those treated with CTLP, resulting in reduced glycolytic process, glucose uptake, lactate generation and ATP synthesis in LoVo and HCT116 cells ( Figures 7G and  S9G). In MTT colorimetric assay, SIP-11 treatment significantly decreased the viabilities of cancer cells, rather than normal mammary epithelial cells (MCF 10A) or TANs lack of significant SPIB levels ( Figure S9H). Administration of SIP-11 inhibited proliferation and invasiveness of cancer cells ( Figures 7H and S9I). To test potency of SIP-11 in vivo, tumour-bearing athymic mice received peptide treatment via tail vein, which led to decrease in 18 F-flurodeoxyglucose (FDG) uptake, growth, weight, glycolysis, SPI1 target gene expression, Ly6G + neutrophils, Ki-67 expression and CD31-staining microvessel density of subcutaneous xenograft models generated by injection of HCT116 cells into nude mice (Figures 7I, J and S10A-D). Moreover, therapeutic SIP-11 led to lower number of lung metastasis, fewer neutrophils, as well as improved survival of athymic mice following administration of HCT116 cells via tail vein ( Figures 7K and S10E). Above findings suggested that therapeutic interfering SPIB-SPI1 interaction inhibited glycolytic process and progression of cancer.

DISCUSSION
Neutrophils, the most abundant component of white blood cells, have a limited lifespan. In response to various cytokines, such as C-X-C motif chemokine ligands, neutrophils are recruited into tumour milieus and sustained by tissue-derived survival signals, resulting in neutrophil accumulation. 32 Previous studies show that TANs are observed in nude mice bearing murine breast tumours 33 or human pancreatic tumours, 34 and contribute to venous thrombosis or tumour necrosis in nude mice 34,35 TANs have differential activation/differentiation status and exert tumour promoting or suppressing effects in a context-specific way. 7 Anti-tumour activities of N1 TANs are characterized by upregulation of immunoactivating chemokines or cytokines, reduced arginase levels or increased ability to kill tumour cells. 7 However, in premetastatic lung niche, neutrophils promote initiation of breast cancer metastasis by producing leukotrienes, 36,37 while employing Ly6G-neutralizing antibodies to remove neutrophils in post-radiotherapeutic tumour-bearing mice reduces the amount of glioblastoma stem cells and prolongs their survival. 38 In the current study, our data demonstrated that transcription factor SPI1 was elevated in cancerous cells and stroma of colon cancer, especially in neutrophils, which was consistent with previous findings that SPI1 is involved in differentiation of neutrophils. 23 Depletion of neutrophils significantly suppresses growth and invasive capabilities of cancer cells in athymic mice. Our evidence shows that as a glycolytic metabolite, lactate triggers polarization of N2 neutrophils, which deliver SPI1 mRNA via extracellular vesicles to enhance SPI1 expres-sion of cancer cells. Through physical interaction with cofactor SPIB, SPI1 exerted tumour-promoting functions in glycolytic process and progression of cancer ( Figure 8G), shedding light on the SPIB/SPI1-mediated positive interplay loop of cancer cells and neutrophils as a therapeutic target against cancers.
Previous studies indicate that SPI1 is upregulated and linked with poor prognosis in breast carcinoma, 39 while SPI1 inhibits invasion of hepatocellular carcinoma cells via upregulating miR-615-5p or suppressing insulin like growth factor 2 expression, 40 suggesting its tumour promoting or suppressive roles in a context-specific way. In the current study, we demonstrate that as a transcriptional activator, SPI1 enriches on promoter regions of glycolytic enzymes HK2 and PGK1 and facilitates their expression in cancer cells. As an important enzyme catalysing irreversible step of glycolytic pathway, HK2 is elevated in various cancer tissues and maintains malignant state of tumours. 41 As the first ATP-producing enzyme, PGK1 participates in generation of 3-phosphoglycerate during glycolysis. 42 PGK1 plays crucial roles in oncogenesis and progression of human cancers and contributes to poor prognosis. 43 Since pharmacological inhibition of glycolysis or knockdown of target genes abolished the oncogneic roles of SPI1 in tumourigenesis and aggressiveness, our results demonstrate that SPI1 promotes cancer progression through facilitating aerobic glycolysis of cancer cells.
As a glycolytic metabolite, lactate is produced by most tumour cells and associated with metastasis and poor survival of cancer patients. 44 Lactate stimulates angiogenesis via promoting stabilization of hypoxia inducible factor 1 alpha, activating nuclear factor kappa B signalling or inducing secretion of vascular endothelial growth factor from tumour-associated stromal cells. 44 We found that lactate was able to induce N2 polarization of neutrophils. In co-culture and conditional medium assays, TANs facilitated glycolytic process, growth and invasion of colon cancer cells, while their derived EVs displayed a high level of SPI1 transcript. As lipid bilayer membrane vesicles released by fusion with cell membrane, EVs serve as a key factor mediating communication between cancerous and microenvironment cells. 45 Based on evidence that inhibition of polarization or SPI1 expression of TANs strongly attenuated tumourigenesis and aggressiveness induced by SPI1, we believe that oncogenic roles of SPI1 in aerobic glycolysis of cancer cells are dependent on TANs and their EVs-mediated delivery of SPI1 mRNA into cancer cells.
Recent studies show that SPIB is upregulated in certain solid malignancies including colon cancer, hepatocellular carcinoma and gastric cancer 46 and increases invasive behaviour of lung cancer cells via downregulation of claudin-2, 47 indicating its involvement in tumourigenesis. The ETS domains of SPIB and SPI1 are 70% identical with each other, and they exhibit similar DNA binding profiles in the genome. 48 BiFC is a method for directly visualizing protein-protein interaction in cultured cells. The reconstituted fluorescence can be observed under a microscope once complementary non-fluorescent fragments are brought into close proximity (a distance around 7 nm) via two interacting proteins. 28 In this study, based on evidence from co-IP and BiFC assays, we found that ETS domains of SPIB and SPI1 were essential for their physical interaction in colon cancer cells. Importantly, SPIB facilitated transactivation of SPI1 to increase expression of glycolytic genes and drove glycolytic process, proliferation and invasiveness of colon cancer cells. Administration of an inhibitory peptide blocking SPIB-SPI1 interaction was efficient in suppressing glycolytic process, tumourigenesis and aggressiveness, indicating the oncogenic activities of SPIB/SPI1 in aerobic glycolysis and progression of cancer.

CONCLUSIONS
Our results reveal that SPI1 and SPIB exert tumourpromoting functions driving glycolytic process and progression of cancer. Mechanistic studies show that neutrophils deliver SPI1 mRNA to cancer cells via EVs, while SPI1 cooperates with SPIB to facilitate upregulation of HK2 and PGK1, two glycolytic enzgymes, resulting in increased aerobic glycolysis, proliferation, invasiveness, as well as metastatic capabilities of cancer cells. Depletion of neutrophils or blocking SPIB-SPI1 interaction significantly suppresses glycolytic process, tumourigenesis, as well as aggressiveness of cancer cells, suggesting that SPIB/SPI1facilitated interplay of cancer cells and neutrophils may be a potential therapeutic target for cancers. While inhibitory peptide SIP-11 appears to be specific, off-target effects cannot be ruled out due to other potential binding proteins. Further studies are warranted to investigate the structural modelling or chemical modifications of SIP-11 for improving its preclinical feasibility. In addition, recent studies show that recruitment of neutrophils to tumour inoculation sites is inhibited by T-regulatory cells, 49 while Ly6G + neutrophils might suppress the cytotoxic effects of T cells. 7 Thus, the roles of SPIB/SPI1 in aerobic glycolysis and cancer progression are warranted by further studies using syngeneic mouse models.

Design of research
A minimum of three biological replicates were used in each experiment. The size of animal cohorts was calculated using data from prior studies. No samples were excluded from the analyses. When possible, a randomization and blinding method was implemented. The primary research objectives were to explore the essential transcriptional regulators and glycolytic genes in cancer progression. The gathered findings resulted in the development of second hypothesis that therapeutic targeting of transcriptional regulators would inhibit the aerobic glycolysis and cancer progression. Standard cell culture procedures were applied to manage experimental design for laboratory research. It also included statistical examination of GEO and TCGA datasets in retrospect.

Cell culture
Short tandem repeat profiling was used to validate the human cell lines HEK-293 (CRL-1573), LoVo (CCL-229), . After resuscitating frozen aliquots, cell lines were applied for studies within 6 months. The Lookout Mycoplasma PCR Detection Kit (Sigma, St. Louis, MO) was routinely applied for examining contamination by mycoplasma. Cells were maintained at 37 • C in an atmosphere of 5% CO 2 , using Dulbecco's modified Eagle medium (DMEM) or RPMI-1640 media with 10% foetal bovine serum (Thermo Fisher Scientific, Inc., Waltham, MA), or incubated using 2-DG or IGF1 as indicated.

Co-culture assay of neutrophils
For isolation of naive neutrophils, peripheral blood of healthy donor or mice was collected and freshly isolated upon receipt. Separation of cells was undertaken by centrifugation over a three-layer discontinuous Percoll gradient. 50 Primary TANs were isolated from fresh colon cancer or xenograft specimens. 7 In brief, fresh tissues were cut into small (1-2 mm) pieces, digested at 37 • C for 2 h and subsequently filtered through cell strainers with diameters of 500, 100 and 70 µm. Cells were labelled with CD66b (ab233811) and CD11b (ab133357, Abcam Inc.) antibodies and purified by Microbeads (Miltenyi Biotec, Somerville, MA). For validating polarization of neutrophils, western blot was performed using antibodies specific for CD66b (ab233811) and CD11b (ab133357, Abcam Inc., Cambridge, MA). Neutrophils were grown in DMEM/F12 containing 0.02% bovine serum albumin, 10 mg⋅ml -1 apo-transferrin and 1 mg⋅ml -1 insulin (Sigma) for 45 min. For co-culture assay, 5 × 10 5 cancer cells were placed in lower compartment, while PNs or TANs (5 × 10 5 cells) were seeded into upper compartment of six-well plates with 1.0 µm pore size inserts (Greiner-Bio-One, Kremsmünster, Austria).

EV isolation and transfer assay
EVs were isolated from culture medium of PNs or TANs as previously described. 45 Briefly, conditional media were treated via 0.22 µm filters. EVs were extracted via ultracentrifuge at 100 000 × g for 90 min, examined by transmission electron microscopy and particle size analysis and validated by western blot using primary antibodies against CD9 (ab92726) and CD63 (ab134045, Abcam Inc.). Purified EVs were suspended in 1000 µl, incubated with 10 µl Dil (Sigma) at 37 • C for 30 min and added to culture medium.  Table S3. The efficiency and specificity of amplification were validated by exponentialbased fluorescence analysis, melting curve and Sanger sequencing ( Figure S12), while no reverse transcription or template served as negative controls. On a Mx3000P spectrofluorometric thermal cycler (Stratagene), real-time PCR assay was conducted with SYBR Premix Ex TaqII  (TaKaRa Inc. and a temperature regime (95 • C 15 min, 65 • C 120 s, 50 cycles, 95 • C 10 s). Cycle threshold (Ct) values for each sample were determined, while transcriptional levels were measured by normalization to reference gene (β-actin or Tubb5) and 2 −ΔΔCt method. 51 The results were validated by three independent experiments, with 3-4 replicates for each sample.

Luciferase reporter assay
Promoter fragments of human HK2 or PGK1 gene were obtained by PCR amplification (Table S5) using genomic DNA and inserted into pGL3-Basic (Promega, Madison, WI). By inserting oligonucleotides carrying four canonical binding sites (Table S5) into pGL4.23 (Promega), a luciferase reporter was prepared for assessing SPI1 transactivation. The dual-luciferase test was carried out as directed by the manufacturer (Promega). 30,52-55

Restoration of gene expression
Cancer cell lines were transfected by SPIB construct for restoring gene expression disrupted by SPI1 knockdown. To rescue gene expression generated by SPI1 overexpression, Genesilencer Transfection Reagent was used to transfect shRNA specific for SPIB (Table S6)

Lentiviral packaging
In HEK293T cells, lentiviral constructs were co-transfected along with psPAX2 and pMD2G (Addgene). At 36 and 60 h post-transfection, infectious lentivirus was extracted and prepared using 0.45 m PVDF filters. Ultracentrifugation was used to concentrate recombinant lentivirus 100-fold (2 h at 120 000 × g). Within 48 h, lentivirus pellets were suspended using phosphate buffer saline (PBS) and applied for use.

Co-IP assay and mass spectrometry
Co-IP reactions were carried out, 29

Inhibitory peptide design and synthesis
Inhibitory peptides were prepared to disrupt the SPI1-SPIB interaction. The Tat protein transduction domain's 11-amino-acid (YGRKKRRQRRR) was applied for cellular penetration. Therefore, inhibitory polypeptides were produced at ChinaPeptides Co. Ltd (Shanghai, China) with purity greater than 95% by connecting with N-terminal biotin-labelled cell-penetrating peptide and C-terminal FITC.

5.16
Pull-down assay using biotinylated-peptide Using 1× cell lysis buffer (Thermo Fisher Scientific, Inc.), proteins were extracted to be treated by biotinylatedpeptide at 4 • C overnight. The cell lysis was then incubated by streptavidin-agarose for 2 h at 4 • C. After thoroughly rinsing the beads, the peptide-pulled down proteins were subjected for western blotting.

5.17
Aerobic glycolysis and extracellular flux analysis As previously reported, 22,57,58 glucose uptake, lactate generation and ATP synthesis of cancer cells were measured. ECAR and OCR were assessed by Seahorse Biosciences XFe24 Flux Analyzer (North Billerica, MA), using basal (G) Schematic depicting the mechanisms underlying SPIB/SPI1-facilitated aerobic glycolysis and cancer progression: neutrophils deliver SPI1 mRNA via extracellular vesicles, resulting in enhanced SPI1 expression within cancer cells. Through physical interaction with its homologous partner SPIB, SPI1 is activated to promote aerobic glycolysis of cancer cells via increasing expression of HK2 and PGK1, which in turn induces N2 polarization of neutrophils via glycolytic metabolite lactate. Administration of a small peptide blocking SPI1-SPIB interaction suppresses aerobic glycolysis, tumourigenesis and aggressiveness of cancer cells. Student's t test and ANOVA compared the difference in C and D. Pearson's correlation coefficient analysis for gene expression in E. Log-rank test for survival comparison in F. *p < .05 vs. P or HEK-293. Data are shown as mean ± SEM (error bars) and representative of three independent experiments in A-D XF media or that with 10 mmol⋅L −1 glucose, 2 µmol⋅L −1 oligomycin and 100 mmol⋅L −1 2-deoxyglucose.

Tumourigenesis and aggressiveness assays in vivo
All animal studies were conducted out in compliance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and were approved by Tongji Medical College's Animal Care Committee (approval number: Y20080290). As previously disclosed, 29,30,[52][53][54] in vivo tumourigenesis and metastatic investigations were carried out using blindly randomized BALB/c nude mice (4-week-old, male, n = 5 per group). Cancer cells (1 × 10 6 or 0.4 × 10 6 ) with over-expression of red fluorescent protein were administrated into the dorsal flanks or tail vein of athymic mice for in vivo treatment investigations. After 1 week, blindly randomized mice were subjected to oral gavage of 2-DG (1 g⋅kg -1 ⋅day -1 , on alternate days for 28-42 days), 22 intraperitoneal injection of anti-Ly6G antibody (200 µg per mouse, every 2 days for 28-42 days) 6 or tail vein administration of therapeutic peptides (50 mg⋅kg -1 ⋅day -1 , once per day at indicated time points). 22,29,30,[52][53][54] The In-Vivo Xtreme II equipment for small animal imaging (Bruker Corporation, Billerica, MA) was used to observe nude mice.

Animal 18 F-FDG imaging
One day before they were sacrificed, mice were fasted for 12 h and allowed to acclimate to positron emission tomography (PET) imaging facility environment in a warmed chamber for at least 1 h. Mice were treated by a single intravenous injection of approximate 200 ± 10 µCi FDG (100 µl). Animals were sedated using 2% isoflurane and placed on a scanner bed after 60 min of FDG uptake. By using TransPET Discoverist 180 system (Raycan Technology Co., Ltd, Suzhou, China), the PET/computed tomography (CT) pictures were collected in static mode for 10 min, followed by a CT scan in normal mode. The PET images were rebuilt via a three-dimensional OSEM approach. The FDK technique was used to reconstruct CT images with a 256 × 256 × 256 matrix. Carimas program (Turku PET Center, Turku, Finland) was used to display the images. The mean standardized uptake value (SUV) was derived by dividing the mean pixel value with decay-corrected regionof-interest activity (Ci⋅kg -1 ) by injected dosage (Ci)/weight (kg).

Clinical samples
The Tongji Medical College's Institutional Review Board approved human tissue investigation (protocol: 2011-S085), which was undertaken in compliance with Declaration of Helsinki's standards. All patients signed a written informed consent form. Patients received no chemotherapy or radiotherapy before surgery. Fresh tissues were pathologically certified and kept at -80 • C until use.

Statistical analysis
Data were presented by mean ± standard error of the mean (SEM). Average expression levels were used to set cutoff values. Before unpaired Student's t test, data were checked for normality. For comparisons of two groups, Student's t test was used. For comparisons of multiple groups, oneway analysis of variance (ANOVA) was undertaken, using the Student-Newman-Keuls (S-N-K) post hoc test. Statistical significance of overlapping was determined using Fisher's exact test. For analyzing expression correlation, Pearson's correlation coefficient assay was used. Log-rank test with/without Bonferroni correction (for more than three groups) was employed to analyse difference in survival. All statistical tests were two-sided, with p values less than .05 considered statistically significant when false discovery rate (FDR) was adjusted.