To test if GM1 glycosphingolipids can be harnessed for biologic drug delivery, we first developed a non-degradable all D-isomer reporter peptide for structure-function studies on the ceramide domain. The reporter peptide was designed to contain two functional groups, a biotin for high-affinity streptavidin-enrichment, and an alkyne reactive group for chemical ligation to fluorophore molecules and quantitative detection. A C-terminal reactive aminooxy was used for coupling the reporter peptide to the oligosaccharide domain of the different GM1 species (Figure 1A and Figure 1—figure supplement 1A) (te Welscher et al., 2014). The functional groups on the reporter peptide, that is biotin, alkyne, fluorophore and combinations of, were tested to verify the absence of confounding effects on GM1 trafficking (Figure 1—figure supplement 1B). This was assessed by confocal microscopy for endosome sorting and transcytosis, using fluorescent cholera toxin B-subunit to label the GM1-peptide fusion molecules (Figure 1—figure supplement 1C). In all cases, the peptide-coupled GM1 species containing cis-unsaturated or short fatty acid ceramide domains sorted into small cytoplasmic vesicles and basolateral membranes consistent with the recycling and transcytotic pathways, whereas the peptide-coupled GM1 species containing saturated long fatty acid ceramide domains did not; they were sorted into larger cytoplasmic puncta consistent with the late endosome/lysosome instead (Figure 1—figure supplement 1C). Both events were blocked at 4°C consistent with uptake by endocytosis. These results are consistent with our previous studies (Chinnapen et al., 2012; te Welscher et al., 2014) and validate the reporter construct.

Figure 1 with 2 supplements see all

Download asset Open asset

All glycosphingolipid-peptide fusion molecules subsequently prepared were coupled to Alexa Fluor-488 (AF488), purified by HPLC, and structures confirmed by mass spectrometry (Figure 1—figure supplement 1A and Material and methods). When tested by pulse-chase in MDCK cells, the peptide-GM1 fusion molecules were internalized and sorted as predicted (Chinnapen et al., 2012). The GM1 species containing long saturated fatty acids (C16:0-GM1) were localized to intracellular puncta consistent with sorting to the lysosome (Figure 1B, bottom panels), and the GM1 species containing short fatty acids were sorted into the recycling and transcytotic pathways as evidenced by localization to apical and basolateral plasma membranes and small intracellular vesicles (Figure 1B middle panels). This interpretation was confirmed using lysotracker to mark the lysosome and the transferrin receptor to mark the recycling endosome (Figure 1—figure supplement 1D). The peptide alone did not bind or enter cell monolayers (Figure 1B, top panels). Because we use GM1 originally purified from bovine brain to synthesize the different GM1 species, the end products comprise two isoforms of the long chain base: one containing a sphingosine chain of C18:1 and the other of C20:1. For the GM1 species containing C12:0 fatty acids, the two sphingosine-isoforms were purified and found to track identically in transcytosis (Figure 1B, middle two panels). Thus, it is the structure of the fatty acid that dominates in the sorting reactions (Chinnapen et al., 2012).

To test structure-function of the ceramide fatty acid chain, we developed a quantitative assay for transcytosis (Figure 1—figure supplement 2A–B and Material and methods). The assay is sensitive to picomolar concentrations and linear over a large 6-log dynamic range (Figure 1—figure supplement 2B). Different GM1-peptide fusions (0.1 μM) were applied to apical reservoirs of polarized epithelial cell monolayers and transport to basolateral reservoirs analyzed after 3 hr by streptavidin-capture and quantitative fluorometric read (Figure 1C, Figure 1—figure supplement 2C). Defatted bovine serum albumin (1% w/v) was added to the basolateral reservoir to amplify release of the lipid-peptide fusion molecules from membrane to solution after transcytosis. In all studies, conditions for equal loading of the different GM1-peptide fusion molecules were determined by quantitative fluorescence measurement of washed cells treated with trypsin to remove adherent glycosphingolipids not incorporated into the membrane bilayer (Figure 1—figure supplement 2D).

Transcytosis for the different GM1-peptide fusion molecules was quantified as an apparent permeability coefficient (PAPP; cm/s) and compared against both the unfused reporter peptide (labeled peptide) or untreated monolayers as negative controls (Figure 1C, Figure 1—figure supplement 2C). When tested on human intestinal T84 cell monolayers, we found approximately a 10-fold increase in transepithelial transport (PAPP) for the GM1 ceramide species containing C6:0, C4:0, C2:0, fatty acids or lyso-GM1 as compared to controls. Introduction of an unsaturated cis-double bond to the short chain ceramide fatty acids (C12:1 and C6:1) had no effect on transcytosis in comparison to the saturated species (C12:0 and C6:0) (Figure 1C). This result is in contrast to the dramatic effect the cis-double bond induces in trafficking of the long fatty-acid chain GM1 glycosphingolipids (Chinnapen et al., 2012; Saslowsky et al., 2013). Transepithelial transport was dose-dependent for the C6:0-GM1-peptide fusion (grey bars) and greatly exceeded transport of the unconjugated reporter peptide (white bars) over a wide range of concentrations (Figure 1D). Mixing experiments using unconjugated GM1 and reporter peptide as individual molecules confirmed that transcellular transport of the peptide cargo was dependent on fusion to the GM1 glycosphingolipid (Figure 1E). Neither the unfused reporter peptide nor the GM1-peptide fusion had any detectable confounding effects on cell viability as determined by measurement of metabolic activity (MTT assay), or monolayer integrity and tight junction function assessed as trans-epithelial resistance (TEER) or dextran flux (Figure 1—figure supplement 2E–G).

Several approaches were used to confirm that the mechanism of cargo transport across epithelial cell monolayers was by transcytosis and not by paracellular leak. First, we tested for transport across epithelial monolayers at 4°C. Such low temperature effectively stops all forms of membrane dynamics including transcytosis, but has minimal effects on paracellular solute diffusion. We found no detectable transport of GM1-peptide fusions across T84 cell monolayers at 4°C, consistent with transport via transcytosis (Figure 2A). The same results were obtained when transcytosis was measured by live cell confocal microscopy. In these experiments, the apical membranes of epithelial cell monolayers were incubated with the C6:0-GM1-peptide fusion at 10°C for 45 min to allow for incorporation of the GM1 ceramide into the apical membrane with minimal uptake into the cell by endocytosis (Figure 2B, x–z and y–z images). Monolayers were washed and then chased for 15 min at 37°C or kept at a restrictive temperature of 10°C. We found the C6:0-GM1-peptide fusion localized to basolateral membranes in cells chased at 37°C, but not at 10°C (Figure 2B; left and middle panels respectively). Only after breaking open tight junctions by removal of extracellular Ca²⁺ (EDTA treatment) did the GM1-peptide fusion molecule gain access to the basolateral membrane at 10°C (Figure 2B, right panel).

Figure 2

Download asset Open asset

In a third approach, we blocked endocytosis at physiologic temperature using the dynamin inhibitor Dyngo-4A. For the C6:0 and C12:0 GM1-peptide fusion molecules, transport into the basolateral reservoir was strongly inhibited by Dyngo-4A treatment, consistent with active transcellular transport by transcytosis (Figure 2C). In contrast, Dyngo-4A had no detectable effect on transport of the reporter peptide alone, as expected for diffusion of solutes by paracellular leak. Similar results were obtained using a genetic approach. The exocyst complex is necessary for efficient receptor-mediated transcytosis of immunoglobulins (Oztan et al., 2007; Nelms et al., 2017), and esiRNA knock-down of EXOC2 subunit caused the predicted 50% decrease in trans-epithelial transport of the C6:0-GM1 peptide fusion molecule (Figure 2D). In contrast, transport of the unfused reporter peptide was not affected by exocyst KD. Thus, fusion of a peptide cargo to certain GM1 species enables active transport of the peptide across the epithelial barrier by transcytosis.

To explain how the very short chain fatty acids amplified transport across epithelial cell monolayers, we first measured the rate of transcytosis by pulse chase. Apical membranes of MDCK cell monolayers were loaded at 10°C with equal amounts of C12:0 and C6:0-GM1-peptide fusions, washed, and then incubated at 37°C and imaged by live-cell confocal microscopy over time. Transcytosis was measured as fluorescence at basolateral membranes. By this method, we found no detectable difference among the two GM1 species in the rate of transcytosis (Figure 3—figure supplement 1A). In both cases, basolateral membranes were fluorescent after a 10 min chase. At longer chase times, however, we observed a dramatic difference between the C6:0- and C12:0-peptide fusion molecules (Figure 3A). Monolayers loaded with the C12:0-GM1 peptide fusion stained brightly at both the apical (bottom left panel) and basolateral membranes (bottom right panel). In stark contrast, monolayers loaded with the C6:0-GM1 peptide fusion showed no fluorescence (middle panels). We interpreted this result as indicating a higher rate of release from cell membranes to the basolateral solution and emptying the cell of the peptide-GM1 fusion over time. To test this idea, we quantified the rate of release from cell membranes for the fluorescent GM1-peptide fusion molecules (Material and Methods). We studied the rate of GM1 release into DMEM media alone (Figure 3B), as well as into DMEM containing defatted bovine albumin (BSA) (Pagano, 1989) (Figure 3C). Results show a faster and more complete diffusion from membrane to solution for the C6:0-GM1 fusion molecule (Figure 3B and C; red curve) compared to the longer chain C12:0-GM1-peptide (blue curve). Faster and more complete release into solution was also observed for the C2:0-GM1-peptide fusion molecule (Figure 3B and C; purple curve). Thus, the greater efficiency for transepithelial transport by the short chain GM1 species is largely explained by their greater efficiency of diffusion from membrane to solution after transcytosis.

Figure 3 with 1 supplement see all

Download asset Open asset

Glycosphingolipids contain another major functional domain in addition to the ceramide, the extracellular oligosaccharide head group. These are structurally diverse and operate in a variety of biologic activities (Cantù et al., 2011). In all cases, however, the oligosaccharide head group acts to trap sphingolipids in the outer leaflet of cell membranes, thus rendering the lipids dependent on membrane trafficking for their distribution across the cell. To test if the effects of ceramide structure on transcytosis and membrane-release were specific to GM1 glycosphingolipids, or could be generalized to other glycosphingolipid species, we fused our reporter peptide to a GM3 ganglioside synthesized to contain ceramide domains with either C12:0 or C6:0 fatty acids. The oligosaccharide domain of GM3 differs from GM1 by the absence of two sugars and thus lacks the terminal galactose (and GalNAc) that functions strongly as a lectin-binding site in GM1. When tested for transcytosis, we unexpectedly found that the GM3-C12:0-peptide fusion molecule crosses epithelial monolayers far more efficiently that the closely related GM1-C12:0-peptide; and as efficiently as the GM1-C6:0 and C2:0 species (Figure 3D). Similarly, transepithelial transport for the GM3-C6:0-peptide was approximately 2-fold greater than that observed for the GM1-C6:0-peptide when compared directly (Figure 3—figure supplement 1B). Transport was strongly inhibited by pretreatment with the dynamin inhibitor Dyngo-4A, implicating active transcellular trafficking by transcytosis. In membrane-release studies, we found a higher rate of release to solution for the C12:0-GM3-peptide fusion (green curve) when compared to the GM1 fusion molecule (Figure 3B and C). Thus, the GM1 glycosphingolipid species appear to be retained in the membrane more tightly than the GM3 species containing the same ceramide domains. Because GM3 lacks a free terminal galactose, we hypothesized the GM1 lipids, which contain the terminal galactose, might be further tethered to the membrane by a form of lectin-binding at the cell surface. To test this idea, we studied the rate of membrane release for the C12:0-GM1 species in the presence or absence of 100 mM lactose (Glc-Gal disaccharide) as a competitive ligand (Figure 3E). These studies show enhanced release from the membrane in the presence of excess free lactose, but not excess mannitol, implicating interaction with a galactose-specific lectin membrane tether (Figure 3E). To confirm this idea, we studied the rate of membrane release for the C12:0-GM3 species that lacks the terminal disaccharide galactosyl N-acetyl galactosamine (Gal-GalNAc) contained in GM1 linked (1→4) to the central galactose. Here we find that lactose at high 100 mM concentrations competed both the GM1 and GM3 species off the membrane (not shown) but at lower doses (5 mM) lactose released only the GM1-fusion (Figure 3—figure supplement 1C–D). Likewise, Gal-GalNAc (5 mM) was effective at enhancing release of only the GM1-fusion molecules (Figure 3—figure supplement 1E–F). Thus, the oligosaccharide domain of the glycosphingolipids can also affect the efficiency of transport across epithelial barriers, we propose by interacting with lectin-like molecules at the cell surface. The results also strengthen the implications and general principles for how the oligosaccharide head group of native glycosphingolipids may affect sorting and retention in specific regions of the cell.

To test for glycosphingolipid-mediated transport across the intestine in vivo, the unfused reporter peptide or the C4:0-GM1-peptide fusion molecule were intragastrically gavaged to mice at equal doses (0.5 nmol/kg) and absorption into the blood analyzed after 15 and 30 min using the streptavidin-capture assay. At both time points, we find evidence of absorption into the systemic circulation for the GM1-peptide fusion molecules (nearly 3% of the applied dose), but not for the unfused peptide (Figure 4A). The same results were obtained for the C12:0-GM3-peptide fusion molecule (Figure 4B). We also measured uptake into the liver, where at 1 hr after gastric gavage we find the glycosphingolipid-peptide fusion molecule, but not the unfused peptide (Figure 4C). Thus, fusion to the glycosphingolipids facilitated absorption of the peptide cargo across the intestine and into the two tissues we sampled, blood and liver. The reporter peptide on its own was not detectably absorbed.

Figure 4

Download asset Open asset

Figure 4—source data 1

Summary Data for Figure 4A.

Columns represent treatment groups. Groupings of rows represent data collected on the same day (independent experiments). Every number is fmols per 100uls blood calculated from standard curves per individual mouse (each measured is an average of two technical replicates). Raw data and data corrected for vehicle (background) are provided.

https://doi.org/10.7554/eLife.34469.010

Download elife-34469-fig4-data1-v2.xlsx

Figure 4—source data 2

Summary Data for Figure 4E.

Columns represent treatment groups. Groupings of rows represent data collected on the same day (independent experiments). Every number is fmols per 100uls blood calculated from standard curves per individual mouse (each measure is an average of two technical replicates). Raw data and data corrected for vehicle (background) are provided.

https://doi.org/10.7554/eLife.34469.011

Download elife-34469-fig4-data2-v2.xlsx

To test if these results can be generalized, we applied the C6:0- and C12:0-GM1-peptide fusions to the nasal epithelium, another tight epithelial mucosal surface. In this case, the C6:0-GM1-peptide (green labeling; Figure 4D) could be visualized by two-photon microscopy within the epithelial barrier in all regions of the nasal epithelium (Figure 4D), including in areas of pseudostratified (top right panels) and simple columnar epithelial tissues (bottom right panels). Uptake of the unfused peptide, applied at the same dose, was very rarely detected (left panels). Absorption to the systemic circulation for the GM1-peptide fusion molecules was confirmed biochemically by measuring content in the blood 15 min after nasal administration (Figure 4E). Here, we find approximately a 10-fold increase in blood levels for the GM1-peptide fusion molecules, compared to peptide alone (which is close to background). Unexpectedly, in the nasal epithelium, we find evidence for efficient absorption of the C12:0-GM1-peptide fusion molecules, similar to our results with the C12:0-GM3-peptide species in the intestine. The result suggests that different tissues may interact in different ways with the oligosaccharide domains of glycosphingolipids. In this case, the nasal epithelium may not bind the GM1 oligosaccharide, thus allowing for more efficient release from cell membranes into solution after transcytosis and systemic absorption.

Glucagon-like peptide-1 (GLP-1) and related peptides have become important drugs in the management of type two diabetes mellitus, by both promoting weight reduction and sensitizing glucose-stimulated insulin release (van Bloemendaal et al., 2014; Heppner and Perez-Tilve, 2015; Tran et al., 2017). A major factor limiting the clinical utility in many individuals is the fact that all currently available preparations must be delivered by subcutaneous injection. To test if the properties of glycosphingolipid trafficking could be applied to enable oral absorption of GLP-1, we coupled a long-half-life version of GLP-1 (Figure 5A) with C-terminal peptide linker (termed here GLP-1 for simplicity) to the C6:0-GM1 ceramide species as described (te Welscher et al., 2014). The bioactivity of the glycolipid-GLP-1 fusion molecule was quantitatively assessed using HEK cells expressing the hGLP-1 receptor and CRE (cAMP) luciferase reporter (te Welscher et al., 2014). As controls, the commercially available long-acting GLP-1 (Exendin-4), and the unfused GLP-1-peptide were assessed in parallel (Figure 5—figure supplement 1A). The fusion of C6:0-GM1 to GLP-1 caused some loss of function, but the molecule remained highly potent as an incretin hormone, closely comparable to that of the controls.

Figure 5 with 1 supplement see all

Download asset Open asset

Figure 5—source data 1

Source Data for Figure 5D.

Columns represent treatment groups. Groupings of rows represent data collected on the same day (independent experiments, exp 1–10). Every number is the total area under glucose tolerance curves per individual mice calculated using GraphPad PRISM 7.0 software. This AUC data (not corrected) was used to produce Figure 5D.

https://doi.org/10.7554/eLife.34469.014

Download elife-34469-fig5-data1-v2.xlsx

Figure 5—source data 2

Source Data for Figure 5E.

Columns represent treatment groups. Groupings of rows represent data collected on the same day (independent experiments). Every number is fmols per 100uls blood calculated from GLP-1 or C6:0-GM1-GLP-1 bioactivity/luciferase curves for each individual mouse.

https://doi.org/10.7554/eLife.34469.015

Download elife-34469-fig5-data2-v2.xlsx

Figure 5—source data 3

Source Data for Figure 5F.

Columns represent treatment groups. Groupings of rows represent data collected on the same day (independent experiments). Every number is fmols per 100uls blood calculated from standard curves per individual mouse (each measured is an average of two technical replicates (with variance well less than 10%).

https://doi.org/10.7554/eLife.34469.016

Download elife-34469-fig5-data3-v2.xlsx

The GLP-1 peptide cargo is 40 residues, approximately 4-fold greater in size compared to the reporter peptide. We first studied transport across intestinal T84 cell monolayers in vitro to test if GM1 glycosphingolipids could transport such a larger cargo. In these studies, GLP-1 transport was quantified by luciferase bioassay as previously described (te Welscher et al., 2014) (Figure 5B). Here, we find an even greater effect of the glycosphingolipids on transepithelial GLP-1 transport (20–100-fold above controls). This is explained by a much lower rate of paracellular leak for the larger sized 40-residue GLP-1 peptide. Such size-exclusion from tight junctions is a well-known determinant of paracellular solute diffusion across intact epithelial barriers.

To test for absorption and biologic incretin activity in vivo, we gastrically gavaged equal doses (10 nmol/kg) of the C6:0-GM1-GLP-1 fusion, the unfused GLP-1 peptide (GLP-1 oral), or vehicle into wild-type mice and measured effects on glucose metabolism by glucose tolerance test (Figure 5C). Here, we find a lower peak and more rapid return of blood glucose to normal levels in the animals gavaged the C6:0-GM1-GLP-1 fusion molecules compared to animals gavaged GLP-1 peptide (Figure 5C and D). The effect on glucose tolerance by gastrically administered C6:0-GM1-GLP-1 was similar to the effect achieved by the intraperitoneal injection of GLP-1 peptide alone, implicating an equally high level of bioavailability for the gastrically-delivered GM1-fusion molecule (Figure 5D).

We confirmed intestinal absorption of the C6:0-GM1-GLP-1 into the systemic circulation in two ways. First, we measured GLP-1 activity in blood samples by streptavidin capture and quantitative luciferase bioassay (Figure 5E). The results show absorption of the GM1-GLP-1 fusion molecule into the blood, but not for unfused GLP-1. In a second approach, we synthesized an all D-amino acid (non-degradable) isomer of GLP-1 coupled to AF488 to allow for direct quantitative measurement of the 40-residue isomer in the blood using the same streptavidin capture assay as described for our reporter peptide (Figure 5A). Again, we find evidence for absorption of the GLP-1 cargo when fused to the C6:0-GM1 transport vehicle, but no absorption for the unfused GLP-1 peptide (Figure 5F). These experiments were performed in two different laboratories, using two different animal facilities with the same results. In all assays (Figure 5C–5F), we find the efficiency of intestinal absorption enabled by fusion to C6:0-GM1 was again almost as efficient as for IP injection of the peptide alone, implicating a high level of bioavailability for the GM1-fusion molecules applied by gastric gavage.

Notably, however, the C2:0-GM1-GLP-1 fusion molecule had no effect on glucose tolerance (Figure 5D) and was not detectably absorbed after gastric gavage (Figure 5—figure supplement 1B), even though this molecule was readily transported across epithelial monolayers in vitro (Figure 1C). This may be explained by lower affinity of the C2:0- (and lyso-) ceramide domains for incorporation into cell membranes, as inferred from membrane loading and release assays (Figure 3B and Figure 1—figure supplement 2D). The difference in biology (transcytosis in vitro versus absorption in vivo) becomes apparent only in vivo where the conditions for epithelial uptake and transport are not optimized as they are in vitro. Thus, although it seemed at first glance that further shortening of the fatty acid beyond C4:0 should amplify transepithelial transport and thus clinical utility, this was not the case and the result informs further development of the technology.

In summary, we find that fusion of therapeutic peptides to GM1 and GM3 glycosphingolipids with short fatty acids enables their active transport across tight epithelial barriers by transcytosis. In the case of the incretin hormone GLP-1, fusion to the lipid carriers allows for gastric (oral) absorption with high bioavailability and the expected effects on blood glucose, highlighting the potential use of this technology in clinical applications.

T84 cells and MDCK cells were both obtained from ATCC (American Type Culture Collection). T84 cells were cultured in a 1:1 mixture of Ham's F12 medium and Dulbecco's modified Eagle's medium with 2.5 mM L-glutamine, 95%; fetal bovine serum, 5%. MDCK cells were maintained in DMEM supplemented with 10% heat-inactivated FBS and penicillin and streptomycin, all obtained from Thermo Fisher Scientific. T84 and MDCK cells were routinely confirmed to be negative for mycoplasma using a PCR kit from MD Biosciences.

T84 or MDCK-II cells plated on 24-well Transwell inserts (polyester membranes, Costar) were washed and equilibrated in DMEM without serum containing defatted-BSA (df-BSA). Unconjugated reporter peptide or GM1-peptide fusions at 0.1 μM complexed to df-BSA in a 1:1 ratio were then added apically for 3 hr. An excess of BSA (1% wt volume) was added basolaterally to aid in extraction of lipid from cell membranes. After a 3 hr continuous incubation, 1 mL basolateral media was collected and incubated with 10 µL magnetic streptavidin sepharose beads overnight at 4°C, washed with TBS-Tween, and eluted in 95% formamide/10 mM EDTA/0.4 mg/mL biotin. For detection of the reporter peptide or GM1-peptide fusion, fluorescence was read using an Infinite M1000 plate reader (Tecan). For each biological replicate concentrations were calculated from standard curves for each compound.

The apparent permeability coefficient (PAPP, cm/s) is calculated across cell monolayers grown in transwells based on the appearance rate of lipid-peptide fusions in the basolateral compartment over time:

Papp (cm/s) = (VD / (A*MD))* (DMR/Dt)

Papp (cm/s)= (cm³ / (cm² * mol)) * (mol/s)

VD = apical (donor) volume (cm³)

MD = apical (donor) amount (mol)

A = membrane surface area (cm²) of apical (donor) chamber (i.e. transwell surface area)

DMR/Dt = the amount of compound (mol) transferred to the basolateral (receiver) compartment over time (s).

The reader is referred to Bio-Protocol (Garcia-Castillo et al., 2018).

WT C57/BL/6 mice (male 7–9 weeks old) were purchased from Jackson Laboratory (Maine USA) and acclimatized for one week. For intestinal absorption experiments, mice that were fasted overnight were lightly anesthetized with isoflurane and gastrically gavaged with a 0.5 nmol/kg dose in a 200 µL volume. Compounds were diluted in PBS containing df-BSA in a 1:1 ratio prior to administration to mice. For analysis of systemic absorption, blood samples were obtained using standard cardiac puncture procedures at 15 or 30 min after compound administration. 100 µL blood was diluted with 400 µL RIPA buffer and incubated with 10 µL streptavidin Sepharose overnight at 4°C, washed, and eluted in 95% formamide/10 mM EDTA/0.4 mg/mL biotin as in our in vitro assay.

Liver tissue was flash frozen in liquid nitrogen and ground with a chilled mortar and pestle on dry ice. After obtaining dry weight, samples were homogenized in 1 mL RIPA buffer, centrifuged, and supernatant incubated with 10 µL streptavidin Sepharose and bound molecules eluted with 95% formamide/10 mM EDTA/0.4 mg/mL biotin. Amount of compound accumulated in the liver was normalized per mg dry weight.

For intraperitoneal glucose tolerance tests, a 10 nmol/kg dose was used to gavage overnight-fasted WT C57/BL/6 mice (male 7–9 weeks old) with GM1-GLP-1 fusion molecules or unfused GLP-1. Glucose measurements following i.p administration of 2 mg/g glucose solution were obtained from tail vein blood applied directly to glucose strips as in (te Welscher et al., 2014).

MDCK-II cells were plated on 96-well plates the day prior to the experiment. Cells were washed with 10°C serum-free DMEM (no phenol red) and equilibrated with DMEM containing 0.1 μM df-BSA for 15 min. Cells were loaded for 45 min at 10°C with 0.1 μM GM1-peptide molecule with a molar ratio of 1:1 (lipid:df-BSA). After loading, cells were washed, warmed to 37°C degrees in DMEM (no phenol red) to allow for proper lipid incorporation, and incubated with 0.25% trypsin in HBSS to release adherent glycosphingolipids not incorporated into the membrane bilayer.

Cells were then incubated in DMEM alone or DMEM containing 1% df-BSA for 2 min, 15 min, or 1 hr. After the indicated time, media was collected and GM1-peptide molecules released into solution quantified using standards for each compound. Cells were subsequently lysed in RIPA buffer and the amount of cell-associated GM1-peptide remaining at that time point quantified using known standards. Amount of GM1-peptide released into the solution was calculated as a ratio of total lipid incorporated (i.e GM1-peptide in media + cell associated GM1-peptide).

Gangliosides of different fatty acid species were supplied by Prof. Sandro Sonnino (U. Milan, Italy). Peptides containing modified functional residues were custom synthesized by Novo Nordisk (DK). Synthesis of peptide-lipid conjugates was accomplished by a modified method previously published (te Welscher et al., 2014). In a typical 2 mL reaction, 2 mg (approximately 1300 nmoles depending on fatty acid) of ganglioside was oxidized with sodium periodate (13 μmoles) in oxidation buffer (100 mM sodium acetate pH 5.5, 150 mM NaCl) for 30 min on ice and protected from light. The reaction was quenched by addition of glycerol (5% final). The reaction was desalted by Bond Elut SepPak C18 cartridge (Agilent, MA) and methanol used to elute from the column was removed by Speed Vac concentration (Savant). The oxidized product was then reconstituted in 2 mL PBS pH 6.9 in the presence of 10% DMF and reacted with 2700 nmoles of aminooxy-containing peptide in the presence of 10 mM aniline (Dirksen and Dawson, 2008). The reaction was incubated for 20 hr at room temperature with mixing on a nutator, where the GM1-peptide fusion product formed normally resulted in a white precipitate. The precipitate was separated from the solution by centrifugation, then resuspended in 400 μL 50% isopropanol/water after brief sonication. PBS pH 6.9 was added (200 μL) along with 4.8 μmoles of sodium cyanoborohydride and incubated for 3 hr to reduce the oxime bond. Lipid-peptide conjugates were purified by semi-preparative HPLC, and confirmed by either MALDI-TOF (AB Voyager), or ESI LC-MS (Agilent, MA).

With exception of the fluorescent peptide described in Extended Data Figure 1 that was done by maleimide linkage, the labeling of peptides with Alexa fluorophore was typically done via copper-mediated Click chemistry. 320 μM peptide-lipid fusions containing an N-terminal alkyne residue (propargylglycine) were reacted with equimolar concentrations of Alexa Fluor 488 -azide under the following conditions. 50 mM Tris-Cl, 5 mM copper (II) sulfate, 100 mM sodium ascorbate, 37 mM (Tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine, TBTA in DMSO/t-butanol 1:4) 1 mM (Tris(2-carboxyethyl) phosphine hydrochloride, TCEP – Sigma) and reacted for 16 hr at room temperature with mixing via nutator. Products were purified by HPLC and confirmed by mass spectrometry. Products were lyophilized and stored at −20°C. Compounds were resuspended in 33% DMF/water to make stock solutions for assays. m/z mass spectrometry values ± 3 Da were as follows: For GM1-C12:0 reporter conjugates with different functional groups on the peptide and d18:1 long chain base, alkyne = 2475.5 Da(1+); biotin = 2734.4 Da(1+); alkyne-biotin = 2829.4 (Da) (1+); Alexa Fluor 488 maleimide = 1552.2 Da and d20:1 = 1566.2 Da. For GM1-C16:0 species in this series, (2+) mass was observed at 1580.2 Da.

Most of the structure-function studies with GM1 fatty acid species lyso to C12:1, were detected with a 3 + charge. m/z values ± 5 Da for d18:1 and d20:1 sphingosine, respectively, were as follows: lyso = 1101.1 Da (3+) and 1110.5 Da (3+); C2:0 = 1115.1 Da (3+) and 1124.5 Da (3+); C4:0 = 1124.5 Da (3+) and 1133.8 Da (3+); C6:0 = 1134.1 Da (3+) and 1142.5 Da (3+); C6:1 = 1133.1 Da (3+) and 1143.5 Da (3+); C12:0 = 1161.8 Da (3+) and 1171.2 Da (3+); C12:1 = 11161.1 Da (3+) and 1170.5 Da (3+). Free peptide was observed as a single ion peak at 2102.8 Da. For GM3 molecular species conjugates, m/z was observed at: C6:0 = 1212.1 Da (3+) and 1026.1 Da (3+); C12:0 = 1040.1 Da (3+) and 1054.1 Da (3+).

To generate bioactive GLP1 fusion lipids, two peptides were joined together via a triazole linkage. Long half-life GLP1 sequences were synthesized containing isobutyrate residues substituted at key dipeptidyl peptidase-4 (DPP-4) cleavage sites, and a C-terminal azido-lysine (Figure 5 and Extended Data Figure 1). A Tobacco Etch Virus protease site (ENLYFQS) was originally designed into the sequence but was not used for the purposes of this paper. The peptide was joined to reporter peptide-lipid conjugates via N-terminal alkyne using Click chemistry as described above.

To synthesize the all-D GLP1-lipid fusions, peptides were made as a complete chain on solid phase, and contained aminooxy and biotin groups (Figure 5). Linkage to oxidized ganglioside was performed as stated above. m/z values ± 8 Da for the biologic GLP1-fusion were observed at: C2:0 = 1808.4 Da (4+); C6:0 = 1822.2 with free peptide seen as a 3 + charge at 2251.0 Da. For the all-D isomer version of GLP1, GLPD fused to GM1-C6:0, m/z was seen as a 3 + charge at 2251.0 Da and the free peptide as a 2 + charge at 2726.7 Da.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

This project was supported by a NIH F-32 DK111072-01 to MDG-C; DK084424, DK048106, DK104868, and an unrestricted innovator grant from Novo Nordisk to WIL; DK104868 to DC.; DK031036 to C RK.; HD000850 to SS.; T32 DK007477 to RJG; and PO1 AI112521 and RO1 AR068383 to Uv-A. We acknowledge the Harvard Digestive Disease Center DK034854 and the Harvard Medical School Center for Immune Imaging. We are grateful to Dr. Elisha Fielding for helping to develop animal tissue harvesting protocols and Kiara Blue for chemical synthesis of lipid-peptide fusions, as well as members of the Lencer laboratory for their helpful discussions. An innovator grant and collaboration with Novo Nordisk was awarded to WIL with no restrictions for initial studies on peptide reporter design (2012–2014). Novo Nordisk declined any and all claims to the technology in 2015.

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved procedures and protocols by the Boston Children's Hospital Institutional Animal Care and Use Committee (IAUC protocol #15-05-2945R).

© 2018, Garcia-Castillo et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.