Phosphohistidine signaling promotes FAK-RB1 interaction and growth factor-independent proliferation of esophageal squamous cell carcinoma

Current clinical therapies targeting receptor tyrosine kinases including focal adhesion kinase (FAK) have had limited or no effect on esophageal squamous cell carcinoma (ESCC). Unlike esophageal adenocarcinomas, ESCC acquire glucose in excess of their anabolic need. We recently reported that glucose-induced growth factor-independent proliferation requires the phosphorylation of FAKHis58. Here, we confirm His58 phosphorylation in FAK immunoprecipitates of glucose-stimulated, serum-starved ESCC cells using antibodies specific for 3-phosphohistidine and mass spectrometry. We also confirm a role for the histidine kinase, NME1, in glucose-induced FAKpoHis58 and ESCC cell proliferation, correlating with increased levels of NME1 in ESCC tumors versus normal esophageal tissues. Unbiased screening identified glucose-induced retinoblastoma transcriptional corepressor 1 (RB1) binding to FAK, mediated through a “LxCxE” RB1-binding motif in FAK’s FERM domain. Importantly, in the absence of growth factors, glucose increased FAK scaffolding of RB1 in the cytoplasm, correlating with increased ESCC G1→S phase transition. Our data strongly suggest that this glucose-mediated mitogenic pathway is novel and represents a unique targetable opportunity in ESCC.


INTRODUCTION
Clonal expansion of cells resistant to growth factor inhibitors plays a critical role in tumor progression while on molecular targeted therapy [1][2][3][4][5]. However, the metabolic factors that allow tumor cells to escape dependence on growth factor for proliferation are unclear. Loss of regulation of glucose (Glc) uptake is a recognized feature of cancer metabolism [6]. In fact, positron emission tomography-based imaging utilizing 18 F-fluorodeoxyglucose uptake (FDG-PET) is routinely utilized for staging and assessing response to treatment in many patients including those with esophageal cancer [7][8][9][10]. FDG-PET activity varies amongst tumors likely dependent on glycolytic activity and proliferative rates [7]. We and others have long noted exceptionally high FDG-PET activity in tumors from esophageal squamous cell carcinoma (ESCC) vs. esophageal adenocarcinoma (EAC) patients [11][12][13][14]. Furthermore, some tumor types such as non-small cell squamous cell lung cancer, but not non-small cell adenocarcinomas of the lung, have distinct metabolic phenotypes vulnerable to inhibitors of glycolysis [15]. However, it remains unclear how Glc serves as mitogenic driver of these cancers and whether targeting these pathways might represent a novel therapeutic option in these cancers.
An increasing number of studies suggest that Glc may promote cancer by participating in alternate glycolytic pathways. For example, Vander Heiden et al. showed that in rapidly dividing tumor cells, decreased pyruvate kinase activity associated with the M2 isoform of pyruvate kinase (PKM2) induces phosphate transfer from phosphoenolpyruvate to the His11 residue on the glycolytic enzyme phosphoglycerate mutase (PGAM1) [16]. This in turn facilitated the high rates of anabolic metabolism by decoupling ATP production from glycolysis found in highly proliferating tumor cells. High levels of Glc have shown to increase the levels of another glycolytic enzyme, 6-phosphofructo-2-kinase/fructose-2,6bisphosphatase 3 (PFKFB3), leading to metabolism reprograming and gastric cancer progression [17]. Phosphohistidine signaling contributes to Glc uptake and metabolism in prokaryotes, eukaryotes and plants [16,18,19]. Protein-histidine kinases including the NME (non-metastatic) family members also known as NDPKs (nucleoside diphosphate kinases) catalyze the phosphate transfer to histidine residues to form phosphohistidine [20]. In addition, histidine phosphatases such as LHPP is associated with deregulation of phosphohistidine signaling in tumorigenesis [21].
We reported that Glc can function as a mitogen in ESCC but not in EAC, and that it induced the phosphorylation of FAK-His58 [22,23]. Moreover, we showed that expression of FAK H58A in ESCC abrogated Glc-induced proliferation concomitant with decreases in the activation levels of AMPK, 70S6K and AKT [22]. Expression of the phosphomimicking mutant, FAK H58E , was sufficient to induce DNA synthesis in serum-starved ESCC. This effect was not dependent on the autophosphorylation site on FAK, Y397, whose phosphorylation by growth factors induces intrinsic tyrosine kinase activity and mitogenic cell signaling [22,[24][25][26]. Y397 autophosphorylation normally associated with growth factor/integrin-stimulated FAK activity. We reported that the overexpression of FAK Y397F did not attenuate either Glc-increased FAK pHis levels or Glc-induced proliferation in KYSE70 cells [22], indicating that FAK pHis58 promoted Glc-stimulated proliferation independent of pY397-mediated FAK kinase activity. The stable overexpression of FAK H58A in ESCC cells attenuated Glc-induced AMPK, pP70S6K and pAKT levels [22], signaling mediators induced downstream of Src, PI3K and Grb2/SOS activated by other FAK scaffolding site modifications. This suggests that Glc-induced FAK pHis58 might activate these pathways in a growth factor-and FAKpY397independent manner. If so, then there is a likelihood that FAK pHis58 may mediate the action of these pathways by scaffolding with different, heretofore, not described binding partners.
To define the scaffolding partners for FAK in Glc-induced cell cycle progression, we compared by nano-LC-MS/MS the binding partners from FAK immunoprecipitates (IP) from ESCC lysates treated with Glc or vehicle. We show that RB1 associates with FAK, but not FAK H58A , only after Glc treatment, and this is likely facilitated by a "LxCxE" RB1-binding motif [27][28][29] found in FAK FERM domain. We show evidence that FAK-RB1 interaction enables CDK4/6-independent CDK2 activation and cell cycle transit [30]. Our studies define a novel mechanism through which Glc stimulates growth factor-independent proliferation by enabling His58-phosphorylated FAK to sequester RB1 to the cytoplasm.

RESULTS
Glucose promotes G1 to S/G2 transition and DNA replication in the absence of growth factors To establish the role of Glc as a sufficient mitogenic factor, we incubated ESCC and EAC cells under conditions capable of inducing quiescence: serum-reduced (5% FBS) and Glc-free medium to deplete Glc for 8 h and then in serum-free and Glc-free medium for 4 h. We then tested whether addition of Glc in serum-free media (for 3 h) could stimulate S-phase entry, as gauged by Click-it EdU flow cytometry (see "Methods"). Glc increased EdU-positivity in three ESCC cell lines, KSYE70, KSYE520 and TE10, but not in three EAC cell lines, FLO1, SK, and KYAE1 (Fig. 1A), supporting the notion that the ability of Glc to selectively increase ESCC cell numbers [23] results from increased G1→S transition.
We further analyzed which ESCC cell cycle phases were affected by Glc using PI staining and flow cytometry. Glc mimicked serumderived growth factors by increasing S-and G2/M-phase levels, and decreasing G1-phase levels (Fig. 1B), and by increasing the ratio of cells in either S or G2/M over those in G1 (Fig. 1C). Furthermore, Glc increased the number of ESCC (TE10), but not EAC (FLO1), cells transitioning to S-or G2/M-phases over time (Fig.  1D). These studies demonstrate that Glc can act as a mitogenic factor to stimulate the cell cycle progression and DNA replication of ESCC cells in the absence of growth factors.
Histidine phosphorylation of FAK contributes to Glcstimulated proliferation in the absence of serum-derived growth factors Activation of receptors by growth factors involves phosphorylation of tyrosine, serine, and/or threonine residues [31][32][33]. Under growth factor-deficient conditions, Glc can activate phosphohistidine signaling to promote growth factor-independent proliferation of ESCC cells [22]. Indeed, Glc increased the total relative protein phosphohistidine levels in ESCC (KYSE70 and KYSE180), but not in EAC (SK and FLO1), with the most induced band corresponding to a~125 kDa protein ( Fig. 2A). Nano-LC-MS analysis of purified recombinant FAK or FAK immunoprecipitated from Glc-stimulated ESCC cells identified increased His58 phosphorylation (Fig. 2B, C). ) and EAC (FLO1, SK and KYAE1) were incubated in serum-reduced (5% FBS) and glucose-free medium to deplete glucose for 8 h and then in serum-free and glucosefree medium for 4 h. After depletion of glucose and serum, the cells were kept in FBS-free medium without (−) or with (+) Glc in the presence of EdU for 3 h. Clik-it EdU flow cytometry kit was used for the assessments of newly synthesized DNA. B Assessments of Glc-induced changes of cell cycle phases using flow cytometry on LSR II. TE10 cells were incubated in serum-reduced (5% FBS) and glucose-free medium to deplete glucose for 8 h and then in serum-free and glucose-free medium for 4 h. After depletion of glucose and serum, the cells were incubated in the medium with or without FBS/glucose (Glc) stimulation for 1.5 h. TE10 cell were trypsinized, fixed in cold 70% ethanol, permeabilized/RNase treated, and stained with PI. ModFit LT program was utilized to analyze G1, S and G2/M phases of the treated cells. C Ratios of (S + G2/M) to G1 in TE10. D Time course assessments of Glc-induced G1 to S/G2/M phase transition. ESCC cells (TE10) and EAC (FLO1) were incubated in 5% FBS medium without Glc for 12 h, and then −/+ Glc stimulation for 0, 40, 80, 120 min. For all panels: *p < 0.05 **p < 0.01; and ***p < 0.001 ****p < 0.0001 vs -Glc.
Greater than 90% of the FAK coding sequence was identified in the FAK immunoprecipitated tryptic peptide fragments (Fig. 2D). These results demonstrate that Glc alone can stimulate histidine phosphorylation on FAK in ESCC cells.
To define the role of FAK pHis signaling in Glc-induced DNA replication, we compared relative FAK pHis levels with levels of newly synthesized DNA as marked by BrdU incorporation. Glc increased relative levels of FAK pHis but not pY397-FAK starting at 5 min (Fig. 2E upper panel), earlier than the first evidence Glcinduced DNA replication, which starts at 30 min (Fig. 2E, lower panel). In addition, the ability of Glc to increase FAK pHis in ESCC, but not in EAC (Fig. 2F)   incorporation (Fig. 2G). This suggests that Glc-induced FAK pHis signaling correlates with DNA replication in the absence of growth factor-stimulated pY397-FAK.

Histidine kinase NME1 promotes phosphorylation of FAK and DNA synthesis
We examined the role of the histidine kinases, NME1 or NME2, in Glc-induced FAK pHis , given that these are the only known mammalian histidine kinases, whose upregulated expression is associated with cancer [21]. Glc induced a recompartmentalization of NME1, moving from the nucleus to the cytoplasm in KYSE70 cells (Fig. 3A). In contrast, NME2 abundance fell below the level of detection (Fig. 3A). The function of NME1 in the nucleus is largely unclear, but cytosolic NME1 can contribute to tumor invasion, a process that involves FAK. Thus, NME1 may be the major kinase in Glc-induced FAK pHis . Consistent with a role in promoting Glcinduced pHis levels in ESCC, siRNA knockdown of NME1 decreased basal and Glc-induced FAK pHis levels ( Fig. 3B). NME1 immunoprecipitated (IP) from Glc-induced KYSE70 cells could phosphorylate purified FAK in vitro (Fig. 3C). Phosphorylation of FAK by NME1 had an optimal pH of 8.0 (Fig. 3D) and an optimal substrate concentration of~200 ng (per 50 µL reaction) (Fig. 3E), but was inhibited by acidic pH (Fig. 3F). Each 100 µL sample was acidified with 25 µL of 1 M HCl before the incubation, then neutralize with 25 µL of 1 M NaOH prior to phosphohistidine detection. In addition, the sample was heated at 95°C for 10-15 min to reverse histidine phosphorylation (Fig. 3F). In KYSE70 cells expressing the kinase-dead NME1 mutant (H118F) [34,35], Glc failed to induced FAK pHis (Fig. 3G). Similarly, Glc increased FAK pHis levels in KYSE70 cells stably expressing WT-FAK, but not in cells expressing FAK-H58A (Fig. 3H). Importantly, neither Glc nor the expression of FAK-H58A affected relative FAK pY397 levels. These observations suggest that Glc induces the cytosolic accumulation of NME1 in ESCC cells, where it facilitates FAK histidine phosphorylation. We previously showed that Glc fails to induce proliferation in ESCC cells stably expressing FAK H58A [22]. We addressed whether NME1 knockdown phenocopied this effect and whether FAK-His58 phosphorylation was required. Figure 3I shows that KYSE70 cells with WT-FAK exhibited Glc-induced DNA synthesis in cells transfected with control siRNA. In contrast knockdown of NME1 abrogated this effect. DNA synthesis could be rescued by the expression of the phosphomimetic, FAK-H58E, even in cells with knocked down NME1. The absolute requirement for NME1 for Glcinduced ESCC proliferation is shown in Fig. 3J, where we showed that the CRISPR-based knockout of NME1 prevented Glc-induced proliferation over 75 h of observation (Fig. 3J). Although no specific NME1/2 inhibitors have been described, our observations that NME1 promotes FAK pHis and DNA replication will support the notion to target NME1 as a means of controlling growth factorindependent ESCC growth.

Phosphatase PP2A modulates phosphorylation of FAK and proliferation
We addressed whether the ability of Glc to induce FAK-His58 phosphorylation might be controlled by the downregulation of protein phosphatases. Nano-LC-MS analysis of cell lysates derived from KYSE70 cells with or without Glc stimulation showed decreased levels of PPP2R2A and PPP2R1B, the subunits of phosphatase PP2A, in Glc-treated cells were lower than that in control cells without Glc stimulation, respectively (Fig. 4A). This observation was verified using Western blot analysis, which indicated that Glc decreased PPP2R2A (Fig. 4B). Indeed, the siRNA-mediated knockdown of PPP2R2A increased relative FAK pHis levels ( Fig. 4C), whereas the overexpression of PPP2R2A in KYSE70 cells decreased basal and Glc-induced FAK pHis levels (Fig. 4D). Next, we examined whether inhibition of PPP2R2A impacted FAK pHis levels. An ELISA assay was used in which FAK protein was captured from cell lysates by immobilized FAK antibody, followed by anti-pHis antibody to detect the levels of captured FAK pHis . The knockdown of PPP2R2A in KYSE70 cells produced high FAK pHis levels irrespective of Glc treatment, commensurate with levels induced by Glc in siControl cells (Fig. 4E). To determine whether Glc modulated PP2A phosphatase activity, we added equal aliquots of PP2A (as PPP2R2A IPs), isolated from lysates of KYSE70 with or without Glc stimulation, to captured FAK ELISAs, and after incubating at 37°C for 10 min, determined the fraction of FAK pHis with anti-pHis antibody. PP2A activity was greater in cells treated with vehicle (Fig. 4F), suggesting that Glc treatment reduced PP2A activity, thereby contributing to the overall increase in FAK pHis activity. Furthermore, PPP2R2A overexpression attenuated Glcinduced BrdU-DNA synthesis in KYSE70 cells (Fig. 4G). These observations suggest that phosphatase PP2A is involved in and controlled by Glc in the induction of FAK pHis and ESCC proliferation in the absence of serum-derived growth factors.
Glucose promotes growth factor-independent FAK function and cell cycle progression via FAK-RB1 interaction Our data suggest a novel mitogenic role for FAK in Glc-induced ESCC G1/S phase transition that is not dependent on the welldefined FAK-Y397 autophosphorylation normally associated with growth factor/integrin-stimulated FAK activity [24][25][26]. Specifically, we reported that the overexpression of FAK Y397F did not attenuate either Glc-increased FAK pHis levels or Glc-induced proliferation in KYSE70 cells [22]. FAK pHis58 may mediate action of cell cycle progression by scaffolding with different, heretofore, not described binding partners. Sequence analysis revealed a consensus LxCxE motif (a.a. 171-175) in the FERM domain of FAK that facilitates protein binding to RB1 [27][28][29]. Although FAK has been shown previously to bind p53 via the FAK FERM domain, FAK binding to RB1 has not been reported. We observed that Glc treatment of ESCC induced cytosolic redistribution of RB1 from nuclei (Fig. 5A). Glc induced perinuclear co-localization of RB1 with FAK as assessed by confocal fluorescence microscopy (Fig. 5B). We then employed a proximity ligation assay (PLA) to assess direct RB1/FAK interaction in cells. Glc increased the interaction of FAK and RB1 at perinuclear sites in ESCC cells (Fig. 5C). This interaction was abrogated by expression of FAK H58A , yet induced by the expression of FAK H58E (a putative phosphomimetic) even in the absence of Glc (Fig. 5D). Mutation of the RB1-binding site ( 171 LxCxE → 171 LxAxA) on FAK abrogated Glc-induced FAK-RB1 interaction (Fig. 5E). We then investigated the FAK-RB1 interaction by co-immunoprecipitation (co-IP) assays. Glc induced increased FAK protein in RB1 IPs (Fig. 5F) and more RB1 in FAK IPs (Fig. 5G), compared to cells receiving vehicle. Consistent with the notion that FAK-RB1 interaction is promoted by Glc-induced FAK pHis , FAK H58A showed no interaction with RB1 interaction in ESCC cells (Fig. 5H). IP/IB analysis of purified recombinant (r) FAK and rRB1 proteins verified a direct interaction of FAK and RB1 (Fig. 5I). Lastly, the mutants of FAK C173A,E175A with altered RB1-binding site, 171 Lx 173 Cx 175 E → 171 Lx 173 Ax 175 A were delivered to the cells. The expression of FAK C173A,E175A , which failed to interact with RB1 ( Fig. 5E), diminished Glc-induced cell proliferation (as gauged by S phase induction) in multiple ESCC cell lines (Fig. 5J). These novel findings strongly suggest that Glc induces ESCC proliferation through the scaffolding of RB1 by FAK in a pHis-dependent manner in the cytosol.

NME1 expression and FAK-RB1 interaction correlate with clinical ESCC progression
It is known that Glc utilization in ESCC is higher than in EAC [11]. In addition, we demonstrated increased NME1 activity in response to Glc (Fig. 3). Data from The Cancer Genome Atlas (TCGA) revealed increased NME1 and FAK mRNA levels in ESCC tumors relative to those in normal tissues (Fig. 6A) [36,37]. Our IHC staining studies showed elevation of NME1 levels in ESCC compared to local normal epithelium (Fig. 6B). Using the FAK ELISA modified to detect RB1 interaction with the secondary antibody, we showed that increased FAK-RB1 interaction in lysates from flash-frozen ESCC tumor tissue versus noncancerous tissues microdissected from esophageal resections (Fig. 6C). Furthermore, PLA analysis of paraffin-embedded human tissue sections showed that increased FAK-RB1 signals in ESCC but not in normal tissues (Fig. 6D). Lastly, lysates from ESCC tissues showed relatively decreased levels of PPP2R2A and increased levels of FAK pHis , NME1, E2F and CDK2 by IB when compared to levels in normal esophageal tissue (Fig. 6E), which is consistent with the increased roles for E2F and CDK2 in ESCC proliferation [30].

DISCUSSION
Our studies clearly indicate a direct onco-proliferative function of Glc in ESCC. These tumors have diverse genetic abnormalities but may share a common metabolic weakness, namely that their growth factor independence is Glc-dependent. This new finding can help resolve fundamental dilemmas: what is the driving force of uncontrolled ESCC growth and why clinical therapies targeting receptor tyrosine kinases have had limited or no effect? Unlike esophageal adenocarcinomas, ESCC acquire excessive Glc and are addicted to Glc. Interestingly, our observations indicate that the Glc levels required to induce ESCC proliferation in vitro is roughly 160-fold lower than those found in normal blood [22]. This suggests that physiological levels of Glc are ample to promote ESCC proliferation even under receptor tyrosine kinase inhibition therapy (interruption of growth factor signaling), supporting that Glc functions in ESCC as a growth factor-like mitogen. Moreover, Glc-stimulated ESCC proliferation in eukaryotes resembles nutrient control of prokaryotic growth.
Our studies have a considerable amount of experimental innovation, including (i) a tumor system describing how Glc can act as a sole mitogenic driver, (ii) data showing that Glc induces cell cycle progression in ESCC (but not in EAC) by inducing FAK pHis58 irrespective of the requirement for FAK-Y397 autophosphorylation or kinase activity, (iii) evidence that Glc-induced FAK pHis58 and DNA replication are likely facilitated by the upregulation of NME1 activity, and (iv) correlation of Glcinduced cell cycle progression with an increase in FAK-RB1 complexes in the cytoplasm. Importantly, our data incorporate a heretofore undescribed role for a histidine phosphorylation pathway in controlling ESCC proliferation induced by Glc as a sole mitogen (Fig. 7). Our reported observations suggested that growth factorstimulated receptor tyrosine kinase pathways were unlikely to be involved in FAK-RB1 interaction-modulated ESCC cell proliferation since growth factors such as insulin did not increase, tyrosine kinase inhibitors did not attenuate, and kinase-dead mutants did not affect Glc-induced FAK modulation of ESCC proliferation [22]. Data from our lab and others indicate that the histidine phosphorylation of metabolic and proliferative regulators in mammalian cells plays a role in tumor progression [16,[38][39][40]. For example, Cantley and coworkers identified the pHis-containing protein, phosphoglycerate mutase 1, involved in an alternative glycolytic pathway in proliferating cancer cells [16]. We have reported that Glc-induced

Glc
FAK pHis signaling is essential for Glc-induced growth factorindependent proliferation of ESCC [22]. The studies on pHisspecific antibody pulldown coupled with Nano-LC-MS/MS analysis confirmed Glc-induced FAK pHis58 in ESCC. Our study is also the first to identify that Glc-induced FAK pHis58 binds RB1, that FAK likely encodes a LxCxE RB-binding motif in its FERM domain. This is similar to other oncoproteins with this motif that disassociate RB1-E2F complexes to stimulate quiescent cells into a proliferative state [41]. FAK pHis58 interaction with RB1 can prevent its binding to E2F increasing cytoplasmic sequestration of RB1, leading to Glcdependent G1→S ESCC cell cycle progression. FAK interaction with RB1 may enable CDK4/6-independent CDK2 activation and cell cycle transit [30], and more especially, that FAK interacts with the RB1 dysregulated form produced by Cyclin D1:CDK4/6 during E2Fmediated G1/S progression [42][43][44][45]. This provides a molecular foundation linking deregulation of pHis signaling with growth factor-independent ESCC cell cycle progression, and importantly, identifies several potential intervention points that could be exploited to therapeutically target ESCC. This previously undescribed pathway helps explain how some tumors, such as ESCC, derive positive growth potential from glycolysis. Therefore, the studies on characterization of NME1 modulation of FAK pHis -RB1 interaction is an extension of our reported findings and yielded significant mechanistic data that will advance our understanding of Glc-promoted ESCC progression. Of importance, it sets the stage for the development of pHis signaling inhibitors at several critical points, such as NME1 kinase activity and FAK/RB1 association. Glc promoting FAK pHis signaling fills the knowledge gap between excessive Glc metabolism via glycolysis (to trigger alternative phosphohistidine signaling) and ESCC growth. This novel pathway likely holds relevance for other tumor types. Current tyrosine kinase inhibitors are typically ATP-competitive compounds or inhibitors of scaffolding activity with signaling partners [46], and they are usually assessed for inhibition of growth factor-induced signaling and/or proliferation. However, our studies indicate that Glc-induced FAK pHis58 -dependent ESCC proliferation does not require FAK-Y397 phosphorylation, suggesting that efforts to target NME1 and FAK pHis58 -RB1 axes will be an innovative way to inhibit ESCC growth, which have particularly evolved growth factor-independent pathways. Our study represents one of the first showing how protein-histidine phosphorylation mediates cancer progression pathways through FAK/RB1 interactions.

MATERIALS AND METHODS Cell models of Glc-stimulated proliferation
An established model of Glc-stimulated proliferation was utilized as previously described [22,23,47]. Briefly, in the presence of reduced serum (5%), cells were incubated in Glc-free medium for 3-8 h to deplete Glc and then in serum-free and Glc-free medium for 1-4 h. After depletion of Glc and serum, the cells were kept in serum-free medium with or without Glc for 1-4 h. WST1 was used to monitor cell viability/proliferation. BrdU or EdU were used to label Glc-induced DNA synthesis. BrdU/EdU-DNA levels were analyzed using ELISA or fluorescence flow cytometry. CRISPR-Cas9 gene editing systems were used to interrupt the FAK gene. Standard molecular biological techniques were used to make WT/mutant FAK constructs. All mutations such as FAK H58A/E were verified by sequencing. Cells expressing WT/mutant (His58A/E) FAK were obtained using flow cytometry sorting, G418 selection and colony expansion. WT/H118F NME1, PPP2R2A, WT/H58E/H58A FAK, and WT/RB1 site-mutated FAK. All mutations such as FAK H58A/E were verified by sequencing. Cells expressing WT/mutant (His58A/E) FAK were obtained by using flow cytometry sorting, G418 selection, and colony expansion. NME1, PPP2R2A, and control siRNAs were purchased from Santo Cruz Biotechnology.

Assessments of Glc-induced DNA replication
BrdU ELISA and EdU coupling flow cytometry were utilized to analyze DNA synthesis as previously described [22,23]. Briefly, cells were incubated in Glc-free medium containing 5% FBS for 3-8 h, and then in FBS-free medium with or without Glc in the presence of BrdU for 1 h or EdU for 4 h. BrdU-DNA and EdU-DNA were determined using the Cell Proliferation ELISA (BrdU, colorimetric, Roche) and Click-iT® Plus EdU Pacific Blue™ Flow Cytometry Assay Kit (Life Science), respectively. Cells were trypsinized, mixed with trypan blue and counted on a TC-20 cell counter to examine the cell viability (Bio-Rad).

ELISA of pHis-FAK
The protocol for analysis of pHis-FAK (FAK pHis ) levels was modified since pHis is heat and acid labile. The wells were coated with an anti-pHis-N3 antibody [49]

Western blot analysis and immunoprecipitation (IP)
Cell lysates and SDS-PAGE gels were prepared as previously described [23]. Modified protocols of Western blot analysis for the detection of pHis protein levels were utilized. The modification includes omitting the step of sample heating, using alkaline buffer systems, and blocking members with dry milk.
IP was carried out as previously described [23]. Briefly, lysates derived from control or Glc-stimulated KYSE70 cells were pre-cleared using Agarose-IgG and then incubated with agarose-conjugated anti-FAK-AC antibody. After washing, the beads (immunoprecipitates) were directly loaded to a SDS-PAGE gel without heating to preserve pHis. After detection of FAK pHis using an anti-pHis antibody, the same membrane was stripped and then reprobed with the anti-FAK antibody.

Mass spectrometry
To overcome the labile nature of the pHis phosphoramidate moiety, we utilized the reported protocols that have been established for pHis detection [22,23]. Phosphorylated human FAK (PV3832, Invitrogen) or pHis antibody-immunoprecipitated pHis protein derived from ESCC cells was digested with trypsin as previously described [50][51][52]. The protein samples were processed using a surfactant-aided-precipitation/on-pellet digestion (SOD) procedure, which provides extensive cleanup to remove detergents and non-protein matrix components, deep protein denaturation (by both surfactants and precipitation) for rapid, efficient, and reproducible digestion, and thereby achieves reliable quantification of samples. Briefly, 100 µg protein was reduced with 10 mM DTT with incubation at 37°C for 30 min in Eppendorf Thermomixer (Eppendorf, Hauppauge, NY), and cysteine residues were alkylated with 20 mM iodoacetamide (IAM) at 37°C for 30 min in the dark. For protein precipitation, one volume of chilled acetone (-20°C) was gently added into each sample and mixed for 1 min to obtain a cloudy suspension. Then, another 8 volumes of chilled acetone were added to the mixture to precipitate proteins. The solution was vortexed until it became clear and stored at −20°C overnight to allow complete precipitation. Subsequently, samples were centrifuged at 20,000 × g for 30 min at 4°C to obtain a protein pellet. After removing the supernatant, 500 μl chilled acetone/water mixture (85/15, v/v %) was added to wash the pellet. Samples were centrifuged for 3-5 min, acetone/ water supernatant was discarded, and the sample was allowed to air-dry. The system was kept under pH=8.5 all the time to prevent potential acid catalysis of pHis degradation.
For protein digestion, the pellet was dissolved in 100 μl Tris (pH = 8.5) buffer and sonicated in a water bath at 37°C. Then, 80 μl Tris buffer was added to 20 μg enzyme powder (Sigma) on ice for activation. The digestion procedure composed 2 steps: (1) activated trypsin was added to the samples at a ratio of 1:40 (enzyme: substrate) and incubated for 6hrat 37°C in an Eppendorf Thermomixer; and (2) a second aliquot of trypsin solution with equal volume was added to the samples and incubated overnight. After centrifugation at 20,000 × g for 20 min at 4°C, 2/3 of the digestion solution was carefully transferred into a tube for LC-MS analysis.
The peptide fragments were subjected to HCD MS/MS analysis as previously described [53][54][55]. The nano-flow reverse phase LC included a Spark Endurance autosampler (Emmen, Holland) and an ultrahigh-pressure Dionex ultimate Nano-2D Ultra capillary/nano-LC system. Peptide separation employed a long nano-LC column (75-μm ID × 100 cm) with Pepmap 3 μm C18 particles. A large-ID trap (300 μm ID × 1 cm) was packed with Zorbax 5 μm C18 materials to allow large-capacity loading and removal of hydrophobic and hydrophilic matrix components. Mobile phase A was 10 mM ammonium acetate in 2% acetonitrile (pH = 8), and mobile phase B was 10 mM ammonium acetate, pH = 8 in 88% acetonitrile. A 4 μg peptide sample was loaded onto the trap with 1% B at 10 μl/min. After the trap was washed for 3 min, a 250 nl/min flow rate was used to back-flush the samples onto the nano-LC column for further separation. The column was enclosed in a heating sheath filled with heat-conductive silicone and warmed homogeneously at 40°C, which helps improve the chromatographic resolution and reproducibility. The following was the 2.5hrseparation gradient used on the column: 4% B for 15 min; 13-28 % B for 110 min; 28-44 % B for 5 min; 44-60 % B for 5 min; 60-97 % B for 1 min; 97% B for 17 min. The trap was turned offline at 45 min to flush hydrophobic components. No perceivable degradation of pHis was observed under this LC condition.
An Orbitrap Fusion Lumos Mass Spectrometer (Thermo Fisher Scientific, San Jose, CA) was employed for peptide identification and quantification. Data collection was operated in a 3-s cycle using the data-dependent topspeed mode. The MS1 survey scan (m/z 400-1500) was at a resolution of 120,000, with automated gain control (AGC) target of 500,000 and a maximum injection time of 50 ms. Precursors were fragmentized in HCD activation mode at a normalized collision energy of 35% and the dynamic exclusion was set with 45 s. Precursors were filtered by quadrupole using an isolation window of 1 Th. The MS2 spectra were collected at a resolution of 15,000 in the Orbitrap, with an AGC target of 50,000 and a maximum injection time of 50 ms.
The raw files (.raw) generated by LC-MS were matched to database with SEQUEST-HT searching engine embedded in Proteome Discoverer (v2.1, Thermo Scientific). The search parameters were set as follows: (1)

Proximity ligation assay (PLA)
Briefly, Duolink™ In Situ Detection PLA kit was used with anti-HA tag/anti-FAK antibody for FAK detection (mouse) and anti-RB1 (rabbit) antibody. Deparaffinized slides or fixed cells on a chamber slide were blocked, incubated with primary antibodies (mouse anti-HA-tag/FAK and rabbit anti-RB1 antibodies) and with PLA probes. After DNA ligation and amplification, slides were examined under a fluorescence confocal microscope using DAPI for nuclear staining. Controls such as FAK KO cells and primary antibody omission were used to verify the specificity of FAK-RB1 interaction.

Microscopy examination
ESCC cells were cultured on a 25 CM 2 flask. Live cell images were acquired using a Zeiss AX10 Observer microscope with a LD APIan 20X/0, 30 Ph 1 lens.

Human specimens
Biospecimens or research pathology services for this study were provided by the Pathology Resource Network, which is funded by the National Cancer Institute and is a Roswell Park Comprehensive Cancer Center Support Grant shared resource. Clinical Data Delivery and Honest Broker services for this study were provided by the Clinical Data Network, which is funded by the National Cancer Institute and is a Roswell Park Cancer Center Support Grant shared resource. All protocols were approved by the Roswell Park Comprehensive Cancer Center Institutional Review Board (IRB). Frozen tumor and non-tumor tissue from esophageal adenocarcinoma and squamous cell carcinoma patients were obtained (BDR 071316). Paraffin-embedded tissue samples of esophageal squamous cell carcinoma, esophageal adenocarcinoma and normal esophagus were obtained (BDR 114019). Informed consent was obtained from all subjects.

Statistical analyses
GraphPad Prism was used for statistical analysis. Student's t tests were used for single comparisons. For comparisons that involve multiple variables and observations, ANOVA was used. For in vitro and in vivo studies, the number of biological replicates was calculated using a statistical analysis for power determination. For all studies, we set an alpha value of 0.05, a power of 0.8, and a standard deviation of 0.25.

DATA AVAILABILITY
All data generated or analyzed during this study are included in this published article and its supplementary information files. Fig. 7 Glucose-induced NME1/PP2A-pHis-FAK-RB1 signaling and tumor growth. Glucose increases histidine kinase NME1 and decreases phosphatase PP2A, resulting in increased histidine phosphorylation of FAK. Histidine 58 phosphorylation of FAK interacts with and sequesters RB1, which promotes G1/S phase transition and tumor growth.