Colonic epithelial cell lines contain little or no GM1 ganglioside

Our inability to detect CTB crosslinking to GM1 in either colonic epithelial cell line stimulated us to evaluate the ganglioside content of these cells. Using Soxhelt extraction, lipids were extracted from T84, Colo205, and Jurkat cells cultured with either vehicle or the glycosphingolipid inhibitor NB-DGJ. Alkali-labile phospholipids were removed by mild alkaline hydrolysis, and some non-polar compounds were removed by silicic acid chromatography. When analyzed by thin-layer chromatography and resorcinol staining (Figure 3A), the fraction from untreated Jurkat cells yielded several bands migrating with mobilities similar to and slower than the GM3 ganglioside (Figure 3A, lane 5). In contrast, in the fraction from Jurkat cells cultured with NB-DGJ, only the most slow-migrating band was present (Figure 3A, lane 6). The fraction from T84 cells produced two bands, one migrating between GM3 and GM1 gangliosides, and one migrating similarly to GM1 (Figure 3A, lane 1). Finally, the fraction from Colo205 cells was only faintly stained (Figure 3A, lane 3).

Open in a separate window

Figure 3.

HP-TLC analysis of glycosphingolipids from T84, Colo205, and Jurkat cells.

Partially purified glycosphingolipid fractions isolated from T84, Colo205 and Jurkat cells cultured with either vehicle or the glycosphingolipid inhibitor NB-DGJ were separated on aluminum-backed silica gel plates using chloroform/methanol/water (60:35:8, by volume) as solvent and stained with resorcinol (A). Chromatograms with separated glycosphingolipids were incubated with ¹²⁵I-labeled CTB (B) or lectin from Erythrina cristagalli (C), followed by autoradiography for 12 hr. In (C), the asterisk (*) highlights the putative neolactotetraosylceramide band. Additional detail about samples analyzed is provided in the methods section.

DOI: http://dx.doi.org/10.7554/eLife.09545.008

To further identify these species, the partially purified glycosphingolipid fractions were probed for binding to ¹²⁵I-labeled CTB, also in thin-layer chromatography format. CTB bound strongly to the fraction isolated from untreated Jurkat cells (Figure 3B, lane 5), and also more weakly to the fraction isolated from Jurkat cells cultured with NB-DGJ (Figure 3B, lane 6). The CTB-binding material from Jurkat cells co-migrated with GM1, and appeared as a doublet, likely corresponding to GM1 species with different ceramide components. No binding of ¹²⁵I-labeled CTB to crude glycosphingolipid fractions from T84 or Colo205 cells was observed (Figure 3B, lanes 1 and 3), not even when high concentrations of the glycosphingolipid fractions and high concentrations of ¹²⁵I-CTB were used. Based on the amount of cells analyzed and the sensitivity of detection, we estimate that T84 and Colo205 cells contain no more than 5 ng of GM1 per million cells.

To evaluate the effect of NB-DGJ on glycosphingolipid production, binding of the Galβ4GlcNAc-binding lectin from E. cristagalli (Teneberg et al., 1994) to the partially purified glycosphingolipid fractions was tested. For the crude glycosphingolipid fraction from T84 cells, binding in the tetraglycosylceramide region, most likely to neolactotetraosylceramide (Galβ4GlcNAcβ3Galβ4Glcβ1Cer), was apparent (Figure 3C, lane 1), but this binding was not visible in the fraction from T84 cells cultured with NB-DGJ (Figure 3C, lane 2). We conclude that NB-DGJ effectively inhibits production of glucosylceramide glycolipids in T84 cells. Thus, if a low, undetectable amount of GM1 is present in the intestinal epithelial cell lines, it is reasonable to assume that this level is further reduced by culturing the cells with NB-DGJ.

Glycoproteins are the dominant CTB-binding molecules on the surface of human colonic epithelial cell lines

With the insight that CTB can recognize glycoproteins displayed on the colonic epithelial cells, we next assessed the relative contributions of different glycoconjugates to overall CTB binding in different cell types. First we examined Jurkat cells, where we had observed multiple CTB crosslinked species. Jurkat cells were cultured with inhibitors of glycosylation to prevent production of specific classes of glycoconjugates, then CTB binding was measured by flow cytometry (Figure 4A). The ganglioside biosynthesis inhibitor NB-DGJ reduced GM1 production in Jurkat cells (Figure 3B) and also resulted in a decrease in CTB binding (Figure 4A). In contrast, culturing Jurkat cells with benzyl-α-GalNAc or kifunensine had no significant effect on CTB binding (Figure 4A). The inability of benzyl-α-GalNAc to affect CTB binding to Jurkat cells is consistent with the known O-linked glycosylation defect in these cells, caused by a frame-shift mutation in the gene encoding chaperone C1GALT1C1 (Figure 4—figure supplement 1A) (Ju and Cummings, 2002). Jurkat cells do produce N-linked glycans, as evidenced by the enhancement of ConA binding to kifunensine-treated cells (Figure 4—figure supplement 1B), but since kifunensine treatment does not affect CTB binding to Jurkat cells (Figure 4A), N-linked glycans do not appear to be major contributors to CTB binding. Thus, the inhibition studies indicate that gangliosides are the dominant binding partners for CTB on Jurkat cells.

Open in a separate window

Figure 4.

CTB binding to human colonic epithelial cell lines depends on protein glycosylation.

(A) Jurkat cells were cultured with inhibitors of glycosylation. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from two independent experiments. (B) T84 cells were cultured with inhibitors of glycosylation, then incubated with increasing concentrations of CTB. Binding of CTB was measured by ELISA. Data presented are the mean values for duplicate samples with error bars indicating the standard deviation. A replicate experiment yielded similar results. (C) T84 cells were cultured with inhibitors of glycosylation. Binding of Alexa Fluor 647-CTB was measured by fluorescence microscopy. (D) Colo205 cells were cultured with inhibitors of glycosylation. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from two independent experiments. (E) Binding of CTB to wild-type or C1GALT1C1 KO Colo205 cells was measured by flow cytometry. Data shown are a single representative trial from three independent experiments.

DOI: http://dx.doi.org/10.7554/eLife.09545.009

Figure 4—figure supplement 1.

Open in a separate window

Characterization of protein glycosylation in Jurkat T cells.

(A) The C1GALT1C1 gene was amplified from genomic DNA of T84 and Jurkat cells by PCR, then sequenced. Jurkat C1GALT1C1 was confirmed to contain a frame shift mutation that results in the expression of a truncated form of the protein (Ju and Cummings, 2002). (B) Jurkat T cells were cultured with inhibitors of glycosylation, then ConA binding was analyzed by flow cytometry. ConA binds high mannose structures that are produced when N-linked maturation is blocked.

DOI: http://dx.doi.org/10.7554/eLife.09545.010

Figure 4—figure supplement 2.

Open in a separate window

Representative fluorescence microscopy images of Alexa Fluor 647-CTB binding to T84 cells cultured with glycosylation inhibitors.

T84 cells were cultured with inhibitors of glycosylation. Binding of Alexa Fluor 647-CTB and DAPI staining were measured by fluorescence microscopy. Quantification of imaging data is presented in Figure 4C.

DOI: http://dx.doi.org/10.7554/eLife.09545.011

Next, we turned attention to the colonic epithelial cell lines. Because of difficulties associated with performing flow cytometry on the highly adherent T84 cells, CTB binding to the surface of T84 cells was measured by other methods. In the first approach, binding of biotin-labeled CTB to T84 monolayers was measured by an ELISA method (Figure 4B). In the second approach, binding of Alexa Fluor 647-labeled CTB to clusters of cells was quantified by fluorescence microscopy (Figure 4C and Figure 4—figure supplement 2). In the fluorescence microscopy approach, fluorescence was observed primarily to the outer surface of cell clusters, suggesting that Alexa Fluor 647-labeled CTB has limited access to the interior of the cell clusters, or that the CTB ligand is not displayed on the interior surface. Despite the differences in format, the results obtained by the ELISA and fluorescence microscopy methods were in agreement. NB-DGJ, the inhibitor of ganglioside biosynthesis, had no effect on CTB binding. Kifunensine, an inhibitor of N-linked  glycan maturation, resulted in only small effects on CTB binding. In contrast, culturing cells with benzyl-α-GalNAc, which blocks elaboration of O-linked  glycans, resulted in reductions in CTB binding in both assays. We conclude the glycoproteins are the dominant CTB binding partners in T84 cells and that gangliosides do not make a large contribution to CTB binding to T84 cells.

To measure CTB binding to Colo205 cells, we used flow cytometry. NB-DGJ treatment did not affect CTB binding to Colo205 cells (Figure 4D). However, in contrast to the results observed for T84 cells, benzyl-α-GalNAc did not cause a reduction in CTB binding to Colo205 cells (Figure 4D), nor did Colo205 cells lacking C1GALT1C1 show reduced CTB binding (Figure 4E). Instead, Colo205 cells cultured with the N-linked inhibitor kifunensine exhibited enhanced CTB binding (Figure 4D), consistent with a small increase in SiaDAz crosslinking that kifunensine causes in these cells (Figure 2C). We considered the possibility that CTB binds to multiple classes of glycoconjugates in Colo205 cells. We cultured Colo205 cells with pairs of inhibitors and measured CTB binding. The only case where we observed decreased CTB binding was when the two protein glycosylation inhibitors – kifunensine and benzyl-α-GalNAc – were used together (Figure 4D). In no case did culturing cells with NB-DGJ result in decreased CTB binding. Based on these data, we propose that CTB binds primarily to glycoproteins on Colo205 cells, with contributions from both N-linked and O-linked glycans. Gangliosides do not make a large contribution to CTB binding to Colo205 cells.

CTB interacts directly with glycoproteins, including CEACAM5

To demonstrate conclusively that CTB binds glycoproteins, we isolated crosslinked CTB complexes and used mass spectrometry to identify one of the crosslinked glycoproteins. T84 cells were cultured with Ac₄ManNDAz, and then incubated with biotin-CTB. After UV irradiation, the cells were lysed and the membrane fraction isolated. Biotinylated, crosslinked complexes were purified on streptavidin-agarose and subjected to trypsin digest, followed by LC-MS/MS analysis. Alternatively, purified crosslinked material was loaded onto an SDS-PAGE gel. Then the CTB-glycoprotein crosslinked material was excised, trypsin digest was performed, and the released peptides were analyzed by LC-MS/MS. We focused attention on proteins that were detected with a spectral count higher than 3 in either crosslinked sample, and not detected in the corresponding control sample (Table 1).

CEACAM5 (also known as CEA) and CD44 were among the top hits from the proteomics analysis and are both known to be highly glycosylated. We tested whether either of these proteins were present in the CTB-glycoprotein crosslinked complex. Cells were cultured with Ac₄ManNDAz, then incubated with CTB. After UV irradiation, the cells were lysed. Immunoprecipitation of CEACAM5 resulted in purification of the CTB crosslinked complex (Figure 5A). Similarly, immunoprecipitation performed with anti-CD44 resulted in purification of the CTB crosslinked complex (Figure 5B). We also performed immunoprecipitation with anti-CTB, which resulted in efficient purification of CEACAM5, but only weak purification of CD44 (Figure 5C). Notably, anti-CTB immunoprecipitated CEACAM5 even when the Ac₄ManNDAz crosslinker was omitted, suggesting that the CTB-CEACAM5 interaction is relatively strong (Figure 5C,D). Furthermore, the material purified by anti-CTB and recognized by anti-CEACAM5 had a slightly higher apparent molecular weight when the crosslinker was included than when it was omitted (Figure 5C). Thus, we observe direct crosslinking between CTB and CEACAM5. Additionally, CD44 co-purifies with the CTB-glycoprotein crosslinked complex, but is not directly crosslinked to CTB.

Open in a separate window

Figure 5.

Protein glycosylation is required for a CTB-glycoprotein interaction.

(A) T84 cells were cultured with Ac₄ManNDAz, incubated with CTB, and UV irradiated. Immunopurification was performed with control IgG or anti-CEACAM5. 7.5% SDS-PAGE immunoblots were performed with anti-CEACAM5 and anti-CTB. (B) T84 cells were cultured with Ac₄ManNDAz, incubated with CTB, and UV irradiated. Immunopurification was performed with control IgG or anti-CD44. 7.5% SDS-PAGE immunoblots were performed with anti-CD44 and anti-CTB. (C) T84 cells were cultured with Ac₄ManNDAz, incubated with CTB, and UV irradiated. Immunopurification was performed with control IgG or anti-CTB. 6% SDS-PAGE immunoblots were performed with anti-CTB, anti-CEACAM5, and anti-CD44. (D) T84 cells were cultured with inhibitors of glycosylation, then incubated with or without CTB. Immunopurification was performed with control IgG or anti-CTB. 6% SDS-PAGE immunoblots were performed with anti-CEACAM5. (E) T84 cells were cultured with Ac₄ManNDAz and with or without kifunensine, incubated with CTB, and UV irradiated. Immunopurification was performed with control IgG or anti-CEACAM5. 7.5% SDS-PAGE immunoblots were performed with anti-CEACAM5 and anti-CTB.

DOI: http://dx.doi.org/10.7554/eLife.09545.013

We performed additional experiments to characterize the role of protein glycosylation in the CTB-CEACAM5 interaction. We cultured cells with inhibitors of glycosylation and assessed the effect on the noncovalent CTB-CEACAM5 interaction (Figure 5D). Both benzyl-α-GalNAc and kifunensine treatment caused reductions in the apparent molecular weight of CEACAM5, suggesting that this protein bears both N-linked and O-linked glycans. Treatment with benzyl-α-GalNAc enhanced the CTB-CEACAM5 interaction, while treatment with kifunensine abrogated the interaction, and treatment with NB-DGJ had no effect. Inhibition of N-linked glycan maturation also blocked crosslinking between CTB and CEACAM5 (Figure 5E). Thus, protein glycosylation regulates the ability of CEACAM5 to interact with CTB.

Taken together, these data demonstrate a direct interaction between CTB and at least one glycoprotein, CEACAM5. Recognition of CEACAM5 by CTB depends on protein glycosylation, suggesting that the glycan, and not the protein, may provide the recognition determinant. While culturing cells with kifunensine abolished the CEACAM5-CTB interaction (Figure 5E), overall CTB crosslinking in T84 cells was only slightly affected by kifunensine treatment (Figure 2B). Thus, additional CTB-crosslinking proteins exist in T84 cells and remain to be identified.

Fucose is important for CTB binding to human intestinal epithelial cell lines

Because glycan identity, not protein identity, appeared to be essential to CTB binding, our next goal was to gain greater insight into glycan structures recognized by CTB. Epidemiological studies implicate variation among fucose-containing glycoconjugates in human susceptibility to cholera (Barua and Paguio, 1977; Swerdlow et al., 1994; Harris et al., 2005; Holmner et al., 2010). In addition, some glycan array binding experiments performed by the Consortium for Functional Glycomics (CFG) suggest that fucosylated glycans can be recognized by CTB, albeit with lower avidity than GM1 (Consortium for Functional Glycomics, 2010). Importantly, fucose is not present in GM1, which is recognized by CTB through interactions with its terminal Neu5Ac and galactose residues (Merritt et al., 1994). We therefore tested whether monosaccharides, including fucose, Neu5Ac, and galactose, could competitively inhibit CTB binding to different cell lines. We found that 100 mM Neu5Ac potently blocked CTB binding to Jurkat cells, while 100 mM galactose inhibited slightly (Figure 6A,B). Other monosaccharides had no effect. In particular, 100 mM fucose did not interfere with CTB binding to Jurkat cells (Figure 6B), even though binding of a fucose-recognizing lectin, Aleuria aurantia lectin (AAL), was completely inhibited by the same concentration of free fucose (Figure 6—figure supplement 1). These results are consistent with CTB binding to Jurkat cells via recognition of GM1.

Open in a separate window

Figure 6.

Fucose blocks binding of CTB to human colonic epithelial cell lines.

(A) Jurkat cells were incubated with 4 µg/mL of CTB in the presence of 100 mM of free sugar. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from three independent experiments. (B) The median fluorescence intensity (MFI) for the no sugar treatment sample presented in panel A was normalized to 100% bound. Data shown represent an average of three independent trials and their standard deviations. (C) T84 cells were incubated with 200 mM of free sugar and variable concentrations of CTB. Binding of CTB was measured by ELISA. Data presented are the mean values for duplicate samples with error bars indicating the standard deviation. A replicate experiment yielded similar results. (D) T84 cells were incubated with variable free sugar concentrations and 10 µg/mL of CTB. Binding of CTB was measured by ELISA. Data presented are the mean values for duplicate samples with error bars indicating the standard deviation. A replicate experiment yielded similar results. (E) T84 cells were incubated with 100 mM fucose or 100 mM glucose in the presence of Alexa Fluor 647-CTB. Binding of Alexa Fluor 647-CTB was measured by fluorescence microscopy. (F) Colo205 cells were incubated with 10 µg/mL of CTB in the presence of 100 mM of free sugar. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from three independent experiments. (G) The median fluorescence intensity (MFI) for the no sugar treatment sample presented in panel F was normalized to 100% bound. Data shown represent an average of three independent trials and their standard deviations. (H) Colo205 cells were incubated with variable free fucose concentrations and 10 µg/mL of CTB. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from two independent experiments.

DOI: http://dx.doi.org/10.7554/eLife.09545.014

Figure 6—figure supplement 1.

Open in a separate window

Effects of free sugars on lectin binding to Jurkat cells.

(A) Jurkat T cells were incubated with 100 mM fucose or Neu5Ac and 4 µg/mL of biotin-AAL. Binding of AAL was measured by flow cytometry. (B) Jurkat T cells were incubated with 100 mM fucose or Neu5Ac and 4 µg/mL of biotin-MAL-II. Binding of MAL-II was measured by flow cytometry.

DOI: http://dx.doi.org/10.7554/eLife.09545.015

Figure 6—figure supplement 2.

Open in a separate window

Representative fluorescence microscopy images showing effects of free sugars on binding of Alexa Fluor 647-CTB to T84 cells.

Binding of Alexa Fluor 647-CTB and DAPI staining in the presence of 100 mM fucose or 100 mM glucose were measured by fluorescence microscopy. Quantification of imaging data is presented in Figure 6E.

DOI: http://dx.doi.org/10.7554/eLife.09545.016

Figure 6—figure supplement 3.

Open in a separate window

Effects of free sugars on lectin binding to Colo205 cells.

(A) Colo205 cells were incubated with 100 mM fucose or Neu5Ac and 10 µg/mL of biotin-AAL. Binding of AAL was measured by flow cytometry. (B) Colo205 cells were incubated with 100 mM fucose or Neu5Ac and 10 µg/mL of biotin-MAL-II. Binding of MAL-II was measured by flow cytometry. (C) Colo205 cells were incubated with variable free fucose concentrations and 10 µg/mL of biotin-AAL. Binding of AAL was measured by flow cytometry.

DOI: http://dx.doi.org/10.7554/eLife.09545.017

Competitive inhibition of CTB binding by monosaccharides was also assessed on colonic epithelial cell lines. T84 cells were incubated with increasing amounts of biotin-CTB in a buffer containing 200 mM free monosaccharide. Binding of biotin-CTB was measured by an ELISA method (Figure 6C). Fucose and Neu5Ac, but not other sugars, effectively prevented binding of all concentrations of CTB. Further, both fucose and Neu5Ac inhibited biotin-CTB binding to T84 cells in a concentration-dependent manner (Figure 6D). Free fucose also blocked binding of Alexa Fluor 647-CTB to T84 cells, as measured by fluorescence microscopy (Figure 6E and Figure 6—figure supplement 2). The ability of monosaccharides to inhibit binding of biotin-CTB to Colo205 cells was measured by flow cytometry (Figure 6F). Fucose was the most effective inhibitor, while galactose and Neu5Ac each showed moderate inhibition (Figure 6G). While as little as 10 mM fucose could interfere with biotin-CTB binding to Colo205 cells, glucose did not affect biotin-CTB binding even at concentrations as high as 200 mM (Figure 6H). Similarly, free fucose, but not free Neu5Ac or glucose, blocked binding of fucose-recognizing AAL to Colo205 cells ( Figure 6—figure supplement 3) . A theme emerging from the monosaccharide competition experiments was that fucose specifically inhibits CTB binding to colonic epithelial cell lines, but not to Jurkat cells. In addition, Neu5Ac and galactose each displayed some ability to block binding of CTB to multiple cell lines.

Based on the results of the monosaccharide competition experiments, we used lectins as blocking reagents to assess whether fucosylated and/or sialylated structures are recognized by CTB. T84 cells were incubated with lectin, then variable concentrations of biotin-CTB were added and biotin-CTB binding was measured by ELISA (Figure 7A). Treatment with Ulex europaeus agglutinin I (UEA-1) or Lotus tetragonolobus lectin (LTL), each of which recognize specific fucosylated structures, had no effect on biotin-CTB binding. We also examined AAL and Lens culinaris agglutinin (LCA), which both recognize α1-6-fucosylated structures, although AAL has a broader substrate scope and also binds fucose in other linkages (Kochibe and Furukawa, 1980; Matsumura et al., 2007; Yu et al., 2012). AAL was able to block biotin-CTB binding, but LCA was not. We titrated the amount of lectins used in this blocking assay, while holding the biotin-CTB concentration constant (Figure 7B). AAL, but not other lectins, blocked biotin-CTB binding to T84 cells in a concentration-dependent way. We used flow cytometry to test the ability of lectins to block CTB binding to Colo205 cells. Colo205 cells were first incubated with lectin, then CTB was added. Similar to what we observed for T84 cells, AAL effectively blocked CTB binding, while other lectins had no effect (Figure 7C). AAL blocked CTB binding to Colo205 cells in a concentration-dependent way (Figure 7D), and when free fucose was included during the AAL pre-incubation, AAL was no longer able to block CTB binding (Figure 7E). While AAL effectively blocked CTB binding to both T84 and Colo205 cells, this fucose-recognizing lectin was not able to block biotin-CTB binding to Jurkat T cells, even when used in excess over toxin concentration (Figure 7F). Taken together, results from lectin blocking experiments are consistent with the idea that CTB binds fucosylated structures on the surface of colonic epithelial cell lines, while fucosylated structures do not contribute significantly to CTB binding to Jurkat cells. In contrast, neither MAL-II, which recognizes α2-3-linked sialic acid, nor Sambucus nigra lectin (SNA), which recognizes α2-6-linked sialic acid, affected biotin-CTB binding to colonic epithelial cell lines (Figure 7A–C).

Open in a separate window

Figure 7.

Aleuria aurantia lectin (AAL) blocks CTB binding to human colonic epithelial cell lines.

(A) T84 cells were incubated with 40 µg/mL of the indicated lectin, then variable concentrations of CTB were added. Binding of CTB was measured by ELISA. (B) T84 cells were incubated with variable lectin concentrations, then 10 µg/mL of CTB was added. Binding of CTB was measured by ELISA. Data presented are the mean values for duplicate samples with error bars indicating the standard deviation. A replicate experiment yielded similar results. (C) Colo205 cells were incubated with 100 µg/mL of the indicated lectins, then 10 µg/mL of CTB was added. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from three independent experiments. (D) Colo205 cells were incubated with variable AAL lectin concentrations, then 10 µg/mL of CTB was added. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from two independent experiments. (E) Colo205 cells were incubated with 40 µg/mL of AAL lectin in the presence of 200 mM free sugar (fucose or glucose), then 10 µg/mL of CTB was added. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from two independent experiments. (F) Jurkat T cells were incubated with 40 µg/mL AAL lectin, then 4 µg/mL of CTB was added. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from two independent experiments.

DOI: http://dx.doi.org/10.7554/eLife.09545.018

We tested whether inhibition of fucosylation affected production of CTB-glycoprotein crosslinked complexes. T84 or Colo205 cells were cultured with Ac₄ManNDAz and increasing concentrations of peracetylated 2-fluorofucose (2F-Fuc), a metabolic inhibitor of fucosylation (Rillahan et al., 2012). As the 2F-Fuc concentration increased, reduced amounts of the CTB-glycoprotein crosslinked complex were observed (Figure 8A,B), indicating that fucosylation is required for SiaDAz-dependent CTB crosslinking to glycoproteins. Next, we assessed whether fucosylation was required for the direct interaction between CTB and CEACAM5. T84 cells were cultured with Ac₄ManNDAz and 2F-Fuc, then crosslinking to CTB was performed. CEACAM5 was isolated by immunoprecipitation and probed with anti-CTB to assess CTB-CEACAM5 crosslinking (Figure 8C). Less crosslinked CTB-CEACAM5 was observed when fucosylation was inhibited. One possible explanation for this result is that the CEACAM5-CTB interaction is mediated by a fucosylated glycan. Additionally, the decreased crosslinking may reflect reduced overall binding of CTB to the cell surface when fucosylation is inhibited.

Open in a separate window

Figure 8.

Inhibition of fucosylation reduces CTB crosslinking and binding to human colonic epithelial cell lines.

(A)T84 cells were cultured with Ac₄ManNDAz and increasing concentrations of 2F-Fuc, incubated with CTB, and UV irradiated. Lysates were analyzed by 7.5% SDS-PAGE immunoblot with anti-CTB antibody. (B)Colo205 cells were cultured with Ac₄ManNDAz and increasing concentrations of 2F-Fuc, incubated with CTB, and UV irradiated. Lysates were analyzed by 7.5% SDS-PAGE immunoblot with anti-CT antibody. (C) T84 cells were cultured with Ac₄ManNDAz and with or without 2F-Fuc, incubated with CTB, and UV irradiated. Immunopurification was performed with IgG or anti-CEACAM5. 7.5% SDS-PAGE immunoblots were performed with anti-CEACAM5 and anti-CTB. (D)T84 cells were cultured with 2F-Fuc, then incubated with increasing concentrations of CTB. Binding of CTB was measured by ELISA. Data presented are the mean values for duplicate samples with error bars indicating the standard deviation. A replicate experiment yielded similar results. (E) T84 cells were cultured with 2F-Fuc. Binding of Alexa Fluor 647-CTB was measured by fluorescence microscopy. (F) Colo205 cells were cultured with 2F-Fuc or 3F-NeuAc. Binding of CTB was measured by flow cytometry. Data shown are a single representative trial from three independent experiments.

DOI: http://dx.doi.org/10.7554/eLife.09545.019

Figure 8—figure supplement 1.

Open in a separate window

Culturing T84 cells with 2F-Fuc causes decreased cell surface fucosylation.

(A) Culturing T84 cells with 100 µM 2F-Fuc results in reduced biotin-UEA-I binding by flow cytometry. (B) Culturing T84 cells with 100 µM 3F-NeuAc does not result in reduced biotin-MAL-II binding by flow cytometry. (C) Culturing T84 cells with 100 µM 3F-NeuAc does not result in reduced biotin-SNA binding by flow cytometry.

DOI: http://dx.doi.org/10.7554/eLife.09545.020

Figure 8—figure supplement 2.

Open in a separate window

Representative fluorescence microscopy images of Alexa Fluor 647-CTB binding to T84 cells cultured with 2F-Fuc.

T84 cells were cultured with 2F-Fuc. Binding of Alexa Fluor 647-CTB and DAPI staining were measured by fluorescence microscopy. Quantification of imaging data is presented in Figure 8E.

DOI: http://dx.doi.org/10.7554/eLife.09545.021

Figure 8—figure supplement 3.

Open in a separate window

Culturing Colo205 cells with 2F-Fuc and 3F-NeuAc causes decreased cell surface fucosylation and sialylation, respectively.

(A) Culturing Colo205 cells with 200 µM 2F-Fuc results in reduced biotin-UEA-I binding by flow cytometry. (B) Culturing Colo205 cells with 200 µM 3F-NeuAc results in reduced biotin-MAL-II binding by flow cytometry. (C) Culturing Colo205 cells with 200 µM 3F-NeuAc results in reduced biotin-SNA binding by flow cytometry.

DOI: http://dx.doi.org/10.7554/eLife.09545.022

We tested whether inhibiting fucosylation affected overall binding of CTB to colonic epithelial cell lines. First, we confirmed the effectiveness of 2F-Fuc to reduce cell surface fucose in T84 cells by showing that it caused a reduction in binding of UEA-1 (Figure 8—figure supplement 1). Then, T84 cells were cultured with or without inhibitory concentrations of 2F-Fuc, followed by incubation with increasing concentrations of biotin-CTB. The amount of biotin-CTB bound to the cell surface was measured by an ELISA method (Figure 8D). Inhibition of fucosylation reduced the plateau value for biotin-CTB binding to T84 cells. We also used fluorescence microscopy to measure binding of Alexa Fluor 647-labeled CTB to T84 cells cultured with or without 2F-Fuc. Cells cultured with the fucosylation inhibitor displayed a reduction in Alexa Fluor 647-CTB binding (Figure 8E and Figure 8—figure supplement 2). Similarly, Colo205 cells were cultured with 2F-Fuc and CTB binding was measured by flow cytometry (Figure 8F). Inhibition of fucosylation reduced CTB binding to Colo205 cells, too. In contrast, Colo205 cells cultured with a sialylation inhibitor, 3F-NeuAc, (Rillahan et al., 2012) displayed enhanced binding to CTB, consistent with a small increase in binding of a fucose-recognizing lectin that also occurs with inhibition of sialylation (Figure 8F and Figure 8—figure supplement 3). We conclude that recognition of fucosylated structures is an important and general mechanism for CTB binding to colonic epithelial cells. The data regarding sialylation are less clear-cut and additional experiments will be required to determine whether sialic acids are components of the glycan motifs that CTB recognizes on colonic epithelial cells.

Fucosylation and protein glycosylation mediate CTB internalization and host cell intoxication

Having discovered that fucosylation and protein glycosylation are important factors mediating CTB binding on the surface of colonic epithelial cell lines, we next tested whether fucosylated glycoproteins were involved in the mechanism by which CT intoxicates host cells. We used an in-cell ELISA method to measure the effects of inhibitors of glycosylation on the uptake of biotin-CTB by T84 cells. Cells were first cultured with glycosylation inhibitors, as described above. To quantify internalized biotin-CTB, cells were incubated in the presence of biotin-CTB at 37°C for the indicated times, then unbound biotin-CTB was washed away and the cells were rapidly cooled to 4°C to arrest internalization. Remaining surface-bound biotin-CTB was blocked with avidin, and also removed by acid washing. After permeabilizing cells, internalized biotin-CTB was measured by ELISA (Figure 9A). Total surface bound biotin-CTB was simultaneously assessed by maintaining control cells at 4°C (no endocytosis) (Figure 9—figure supplement 1) and yielded values consistent with other measurements (Figures 4B,C, Figure 8D,E). The amount of biotin-CTB internalized by cells treated with NB-DGJ was indistinguishable from that observed for control cells (Figure 9A), suggesting that gangliosides do not mediate the majority of biotin-CTB internalization. Kifunensine caused a slight enhancement in internalization, suggesting that N-linked glycosylation of proteins might serve as an impediment to CTB internalization. Culturing cells with either benzyl-α-GalNAc or 2F-Fuc each nearly abolished internalization, implying that fucosylated and/or O-linked glycoproteins may play functional roles in biotin-CTB internalization.

Open in a separate window

Figure 9.

Fucosylation mediates CT internalization and intoxication in a human colonic epithelial cell line.

(A) Internalized biotin-CTB for vehicle-, NB-DGJ-, and kifunensine-treated cells and vehicle-, benzyl-α-GalNAc-, and 2F-Fuc-treated cells are displayed as a function of time. Cells were incubated with biotin-CTB at 37°C for the indicated amount of time to allow internalization to occur. The amount of internalized biotin-CTB was measured by ELISA. Data presented are the mean values for duplicate samples with error bars indicating the standard deviation. Replicate experiments yielded similar results. (B) T84 cells were cultured with inhibitors of glycosylation. Cells were exposed to variable concentrations of CT holotoxin for 1 hr. cAMP levels were measured by ELISA. The cAMP levels for experimental samples are reported relative to the corresponding vehicle-treated control cells. Data shown represent an average of four independent trials.

DOI: http://dx.doi.org/10.7554/eLife.09545.023

Figure 9—figure supplement 1.

Open in a separate window

Total surface-bound biotin-CTB for vehicle-, NB-DGJ-, and kifunensine-treated cells and vehicle-, benzyl-α-GalNAc-, and 2F-Fuc-treated cells.

Cells were incubated with biotin-CTB at 4°C. Surface-bound biotin-CTB was measured by ELISA.

DOI: http://dx.doi.org/10.7554/eLife.09545.024

Figure 9—figure supplement 2.

Open in a separate window

The inhibitors of glycosylation kifunensine and benzyl-α-GalNAc exhibit off-target effects on cAMP production.

(A) T84 cells were cultured with inhibitors of glycosylation. Cells were exposed to 10 μM forskolin for 1 hr at 37°C. cAMP levels were measured by ELISA. The cAMP levels are reported relative to control cells induced with forskolin. Data shown represent an average of two independent trials. (B) T84 cells were cultured with inhibitors of glycosylation. Cells were exposed to 1 μM VIP (vasoactive intestinal peptide) for 1 hr at 37°C. cAMP levels were measured by ELISA. The cAMP levels are reported relative to control cells induced with VIP. Data shown represent an average of two independent trials. (C) T84 cells were cultured with inhibitors of glycosylation. Cells were exposed to variable concentrations of CT holotoxin for 1 hr. cAMP levels were measured by ELISA. The cAMP levels for experimental samples are reported relative to the corresponding vehicle-treated control cells. Data shown represent an average of four independent trials.

DOI: http://dx.doi.org/10.7554/eLife.09545.025

Figure 9—figure supplement 3.

Open in a separate window

Brefeldin A blocks CT-induced cAMP accumulation.

T84 cells were cultured with or without the ganglioside inhibitor NB-DGJ. Cells were exposed to 1 μg/mL brefeldin A at 37°C for 30 min, then 100 nM of CT holotoxin for 1 hr. cAMP levels were measured by ELISA. The cAMP levels are reported relative to control cells induced with CT in the absence of brefeldin A treatment. Data shown represent an average of three independent trials.

DOI: http://dx.doi.org/10.7554/eLife.09545.026

Next, we examined the effects of inhibitors of glycosylation on CT-induced elevation of cAMP, a later step in host cell intoxication. T84 cells were cultured in monolayers in the presence of glycosylation inhibitors. First, we evaluated whether inhibitors of glycosylation caused off-target effects on adenylate cyclase activity. Indeed, both benzyl-α-GalNAc and kifunensine caused reductions in the amount of cAMP that accumulated in response to stimulation with forskolin (Figure 9—figure supplement 2A), and with vasoactive intestinal peptide (VIP; Figure 9—figure supplement 2B), and similarly reduced cAMP accumulation in response to CT (Figure 9—figure supplement 2C). While the effects of benzyl-α-GalNAc and kifunensine on adenylate cyclase activity make it difficult to interpret how these inhibitors modulate CT intoxication, interpreting the effects of NB-DGJ and 2F-Fuc is more straightforward since neither affected cAMP accumulation in response to forskolin or VIP (Figure 9—figure supplement 2A,B).

CT holotoxin was incubated with cells at 37°C for 1 hr, then cAMP accumulation was measured. 2F-Fuc treatment resulted in dramatically decreased accumulation of cAMP, while treatment with NB-DGJ had a more moderate effect (Figure 9B). Thus, host cell fucosylation is important for intoxication by CT. Since 2F-Fuc treatment also causes reduced CTB binding and cell entry, the most likely explanation for this result is that reduced fucosylation leads to less CT entering cells, and thereby reduces host cell intoxication. In contrast, NB-DGJ does not dramatically affect CTB binding or internalization, so the explanation for this result is less clear. One possibility is that there is a small amount of GM1 in the cells, which is capable of mediating host cell intoxication via pathway that operates in parallel to the fucosylated glycoproteins. A second possibility is that GM1 is not required for the initial steps of CT binding and internalization, but is required for later steps in intoxication. A final possibility is that GM1 is not essential, but other glucosylceramide glycolipids, which are also reduced due to NB-DGJ treatment, play roles in the trafficking and intoxication process. Notably, CT intoxication appeared to be sensitive to brefeldin A in both untreated cells and in cells that were cultured with NB-DGJ (Figure 9—figure supplement 3), implying that in both cases, CT intoxication of host cells occurs via retrograde transport through the secretory pathway, consistent with other studies (Lencer, 2003).

SiaDAz-mediated CTB crosslinking

For photocrosslinking of CTB to Jurkat T cells (a suspension cell line), 2 million cells were first seeded in 10 mL media into 10-cm tissue culture plates treated with either vehicle (evaporated ethanol) or 100 μM Ac₄ManNDAz. After culturing for 72 hr, cells were counted, re-suspended in fresh media to a concentration of 5 million cells/mL, transferred to two separate multiwell tissue culture plates (for –/+ UV), and CTB (Sigma) was added at a concentration of 2.5 μg/mL. The toxin was allowed to bind to the cell surface for 45 min at 4°C in the dark. The cells were then either kept at 4°C for an additional 45 min (for the – UV samples) or irradiated on an ice/water bath for 45 min (for + UV samples) at 365 nm (UVP, XX-20BLB lamp). The cells were then collected, washed with Dulbecco’s Phosphate Buffered Saline (DPBS), and lysed on ice for 30–60 min in RIPA buffer (50 mM TrisHCl, pH 8.0, 150 mM NaCl, 1% (v/v) NP-40, 0.5% (v/v) sodium deoxycholate, 0.1% (w/v) sodium dodecyl sulfate (SDS) and a protease inhibitor cocktail (Roche, Indianapolis, IN) (catalog no. 11836170001)). The lysate was pelleted at 20,817g for 10 min at 4°C to clear the insoluble debris, and the supernatant was retained for further immunoblot analysis. Protein content was quantified with a BCA assay kit (Thermo Scientific Pierce Protein Biology, Waltham, MA) against a BSA standard curve for normalization. For resolution on a high percentage (15%) Tris-glycine gel, 9 μg of lysate was denatured in 2X SDS loading dye (100 mM TrisHCl pH 6.8, 4% (w/v) SDS, 0.04% (w/v) bromophenol blue, 20% (v/v) glycerol, and 10% (v/v) 2-mercaptoethanol) for 5 min at 90°C. For resolution on a lower percentage (6%) Tris-glycine gel, 12 μg of lysate was denatured in 4X SDS loading dye (200 mM TrisHCl pH 6.8, 8% (v/v) SDS, 0.08% (w/v) bromophenol blue, 40% (v/v) glycerol, and 40 mM DTT) for 5 min at 90°C. The samples were separated by SDS-PAGE and transferred to a PVDF membrane (EMD Millipore, catalog no. IPVH00010). The blots were probed overnight at 4°C for anti-CTB (Abcam, 1:10,000 dilution) in 5% (w/v) non-fat milk in TBST. The blots were then probed with a goat anti-rabbit HRP conjugated secondary antibody (1:5000 dilution) for 1 hr at room temperature, and developed using the SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific, catalog no. 34080) and X-ray film. Membranes were stripped in mild stripping buffer (200 mM glycine, 0.1% (w/v) SDS, 1% (v/v) Tween-20, pH 2.2) for 45 min at 37°C before re-probing for the loading control anti-α-tubulin (Sigma, 1:10,000 dilution) for 1 hr at room temperature. The blots were then probed with a goat anti-mouse HRP conjugated secondary antibody (1:5000 dilution) for 1 hr at room temperature, and developed using the SuperSignal West Pico Chemiluminescent Substrate and X-ray film.

For photocrosslinking of CTB to human colonic epithelial cell lines (both adherent cell lines), either 250,000 T84 cells or 500,000 Colo205 cells were seeded in 2 mL of media into two separate 6-well tissue culture plates (for –/+ UV) that were treated with either vehicle (evaporated ethanol) or Ac₄ManNDAz to a final concentration of 100 μM. When used, glycosylation inhibitors were also added at the time of seeding to achieve these final concentrations: 50 µM NB-DGJ, 2 mM deoxymannojirimycin, 1 µg/mL kifunensine, 2 mM benzyl-α-GalNAc, and 5 – 200 μM 2F-Fuc. Experimental samples were compared to the appropriate vehicle-only control: water for NB-DGJ, deoxymannojirimycin, and kifunensine; DMSO for benzyl-α-GalNAc and 2F-Fuc. After culturing for 72 hr, the media in each well was replaced with 1 mL fresh media containing approximately 4.5 μg of CTB (Sigma). The toxin was allowed to bind to the cell surface for 45 min at 4°C in the dark. The cells were then either kept at 4°C for an additional 45 min (for – UV samples) or irradiated on an ice/water bath for 45 min (for + UV samples) at 365 nm. Cells were then washed with DPBS, collected into RIPA lysis buffer with a cell scraper, and incubated on ice for 30 – 60 min. The lysate was centrifuged at 21,000g for 10 min at 4°C to remove insoluble debris, and the supernatant was retained for separation on both a higher (15%) and lower (ranging from 6 – 7.5%) percentage gel, as described previously for Jurkat T cells. The samples were then transferred to a PVDF membrane, and the blots were probed overnight at 4°C for either anti-CTB (Abcam, 1:10,000 dilution) or anti-CT (Sigma, 1:10,000 dilution) as indicated in the figure labels. Membranes were stripped and re-probed for the loading control anti-α-tubulin as described previously. To ensure the glycosyltransferase inhibitors were effective in T84 cells, samples were probed for 1 hr at room temperature for either anti-LAMP-1 (BD Biosciences, 1:5000 dilution) or anti-CD44 (Cell Signaling Technology, 1:2500 dilution). Photocrosslinking of CTB to hCMEC/D3 or HBEC cell lines was carried out in an analogous manner, with the exception that 200,000 cells were seeded into 6-well tissue culture plates, and approximately 1.1 μg of CTB (Sigma) was added to each well.

Ganglioside analysis

CTB binding assays

Proteomics analysis of CTB crosslinked complex

Immunopurifications

For crosslinking for immunopurification (IP) experiments, 750,000 T84 cells were seeded into 6-cm tissue culture plates in 6 mL of media containing 100 μM Ac₄ManNDAz; if required, a glycosylation inhibitor was added at this time. After culturing for 72 hr, the cells were replenished with 1.5 mL of fresh media containing 4 μg/mL CTB (Sigma) (i.e., 6 μg per plate). The toxin was allowed to bind to the cell surface for 45 min at 4°C in the dark. The cells were then either kept at 4°C for an additional 45 min (for the – UV plates) or irradiated on an ice/water bath for 45 min (for + UV plates) at 365 nm. The cells were then washed with DPBS, collected into RIPA lysis buffer with a cell scraper, and incubated on ice for 30–60 min. The lysate was pelleted at 21,000g for 10 min at 4°C to clear the insoluble debris, and the supernatant was retained; protein content was quantified with a BCA assay kit (Pierce) using a BSA standard curve.

For IP with the anti-CEACAM5 and anti-CD44 antibodies, the crosslinked lysates were diluted to 1.5 mg/mL in RIPA buffer, and then 270 μg was added to a 1.5 mL microcentrifuge tube. A 1:50 dilution of normal mouse IgG (EMD Millipore, catalog no. NI03), anti-CEACAM5 antibody, or anti-CD44 antibody was added, and the samples were mixed by end-over-end rotation overnight at 4°C. The lysate/Ab mixture was then added to 10 μL of Protein G sepharose (Sigma, catalog no. P3296) and mixed by end-over-end rotation for 2.5 hr at 4°C. The beads were washed three times with 200 µL RIPA buffer, and eluted with 10 μL 2X SDS loading dye (100 mM TrisHCl pH 6.8, 4% (w/v) SDS, 0.04% (w/v) bromophenol blue, 20% (v/v) glycerol, and 20 mM DTT) for 5 min at 90°C. The supernatant was collected and loaded into a single well of a 7.5% Tris-glycine gel. Two sets of samples were analyzed at a time: the first set to ensure that the IP of the target protein was successful (i.e., probing for CEACAM5 or CD44), and a second set to probe for association of the target protein with CTB. Therefore, after separation and transfer to a PVDF membrane, the blots were probed overnight at 4°C for either CEACAM5 (1:5000 dilution), CD44 (1:2500 dilution), or CTB (1:10,000 dilution) in 5% (w/v) non-fat milk in TBST. The blots were then probed with a HRP conjugated secondary antibody (1:5000 dilution) for 1 hr at room temperature, and developed using the SuperSignal West Pico Chemiluminescent Substrate and X-ray film.

For IP with the anti-CTB antibody, the crosslinked lysates were diluted to 1.5 mg/mL in RIPA buffer, and then 300 μg was added to a 1.5 mL microcentrifuge tube. Then, 0.5 μg of either ant-CTB antibody or normal rabbit IgG (EMD Millipore, catalog no. NI01) was added, and the samples were mixed by end-over-end rotation overnight at 4°C. The lysate/Ab mixture was then added to 10 μL of TrueBlot anti-rabbit Ig IP beads (Rockland Immunochemicals, Limerick, PA) (catalog no. 00-8800-25) and mixed by end-over-end rotation for 2.5 hr at 4°C. The beads were washed four times with 200 µL RIPA buffer, and eluted with 10 μL 2X SDS loading dye for 5 min at 90°C. The supernatant was collected and loaded into a single well of a 6% Tris-glycine gel. After separation and transfer to PVDF membranes, the blots were probed for CTB, CEACAM5, and CD44 as described above.

For co-IP of CTB and CEACAM5 within non-crosslinked lysates, 750,000 T84 cells were seeded in 5 mL of media into 6-cm dishes, and glycosylation inhibitors were added at this time. After culturing for 72 hr, the cells were replenished with 1.5 mL of fresh media containing 4 μg/mL CTB (Sigma) (i.e., 6 μg per plate) and kept at 4°C for 1.25 hr. The cells were then washed with DPBS, collected into RIPA lysis buffer with a cell scraper, and incubated on ice for 30–60 min. The lysate was pelleted at 21,000g for 10 min at 4°C to clear the insoluble debris, and the supernatant was retained; protein content was quantified with a BCA assay kit (Pierce) using a BSA standard curve for normalization. For IP with the anti-CTB antibody, the lysates were diluted to 1.4 mg/mL in RIPA buffer, and then 350 μg was added to a 1.5 mL microcentrifuge tube. Then, 0.5 μg of either anti-CTB antibody or normal rabbit IgG was added, and the samples were mixed by end-over-end rotation overnight at 4°C. The lysate/Ab mixture was then added to 10 μL of TrueBlot anti-rabbit Ig IP beads and mixed by end-over-end rotation for 2.5 hr at 4°C. The beads were then washed, eluted, loaded onto a 6% Tris-glycine gel for separation, and probed on a PVDF membrane for CEACAM5 as described previously for the anti-CTB IP of crosslinked samples.

Measurements of CTB uptake

CTB internalization was measured by the in-cell ELISA, described above, with the following adaptations. The biotin-CTB concentration was 1 µg/ml. During the experiment, control samples (to measure total surface bound biotin-CTB) were washed three times with ice-cold PBS, then kept on ice awaiting analysis. Experimental samples were warmed to 37°C for the indicated times to allow endocytic uptake, then endocytosis was halted by returning cells to ice and washing three times with cold PBS. Non-internalized biotinylated CTB was masked by successive treatment with 50 µg/ml of avidin (Sigma-Aldrich) for 1 hr on ice, followed by three 1-min cold acid washes (0.2 M acetic acid/0.2 M NaCl). Cells were fixed with 4% paraformaldehyde and further permeabilized with 0.1% (v/v) Triton X-100 in PBS. Cells were then incubated at room temperature for 1 hr in streptavidin-HRP conjugate diluted in Q-PBS (PBS supplemented with 0.01% (w/v) saponin, 2% (w/v) BSA, and 0.1% (w/v) lysine, pH 7.4). Reactive aldehydes and nonspecific binding sites were quenched with Q-PBS. HRP activity was measured as described above.

Measurement of cAMP

350,000 T84 cells were cultured for 8–10 d in the absence or presence of each glycosylation inhibitor or the corresponding vehicle control, to achieve the following final concentrations: 50 µM NB-DGJ, 1 µg/ml kifunensine, 2 mM benzyl-α-GalNAc, or 100 µM 2F-Fuc. Stock solutions of NB-DGJ and kifunensine were dissolved in water and benzyl-α-GalNAc and 2F-Fuc were dissolved in DMSO. Cells were cultured with inhibitors on 4.67-cm² Transwell inserts (Costar Laboratories, Cambridge, Mass). Monolayer integrity was evaluated by transepithelial electrical resistance (TEER) measurements, which consistently increased over the 8–10 d culturing period. Typical TEER values for cells treated with the water control, NB-DGJ, and 2F-Fuc were 800–900 Ω cm²; for cells treated with kifunensine were 700 Ω cm²; for cells treated with DMSO were 1400 Ω cm²; for cells treated with benzyl-α-GalNAc were 1800 Ω cm². Cells were incubated with forskolin (10 µM), VIP (1 µM), or cholera holotoxin (0, 1, or 100 nM) in culture media with 5% CO₂ at 37°C for 1 hr, followed 2 washes with cold PBS, then lysed and flash frozen in liquid N₂. Accumulated cAMP was measured using Direct Biotrak EIA kit (GE Healthcare), according manufacturer’s instructions and normalized by protein content (bicinchoninic acid assay, Pierce BCA protein assay kit, Pierce).

The authors would like to acknowledge the assistance of the UT Southwestern Live Cell Imaging Facility, a Shared Resource of the Harold C Simmons Cancer Center. We thank Lu Zhang, Andres Roig, Jerry Shay, and John Minna (UT Southwestern) for sharing HCEC and HBEC cells, and Babette Weksler (Weill Cornell Medical College), Pierre-Oliver Couraud (INSERM), and Ignacio Romero (The Open University) for sharing hCMEC/D3 cells. We thank Carlos Reis and Sandra Schmid (UT Southwestern) for help with the in-cell ELISA assay. We thank Jan Holmgren (University of Gothenburg), Kim Orth (UT Southwestern), Arun Radhakrishnan (UT Southwestern), and Michael Shiloh (UT Southwestern) for comments on the manuscript.