Immunoengineering can overcome the glycocalyx armour of cancer cells

Abstract

Cancer cell glycocalyx is a major line of defence against immune surveillance. However, how specific physical properties of the glycocalyx are regulated on a molecular level, contribute to immune evasion and may be overcome through immunoengineering must be resolved. Here we report how cancer-associated mucins and their glycosylation contribute to the nanoscale material thickness of the glycocalyx and consequently modulate the functional interactions with cytotoxic immune cells. Natural-killer-cell-mediated cytotoxicity is inversely correlated with the glycocalyx thickness of the target cells. Changes in glycocalyx thickness of approximately 10 nm can alter the susceptibility to immune cell attack. Enhanced stimulation of natural killer and T cells through equipment with chimeric antigen receptors can improve the cytotoxicity against mucin-bearing target cells. Alternatively, cytotoxicity can be enhanced through engineering effector cells to display glycocalyx-editing enzymes, including mucinases and sialidases. Together, our results motivate the development of immunoengineering strategies that overcome the glycocalyx armour of cancer cells.

Extended Data Fig. 1:. Additional analysis of Muc1-mediated protection against NK cell cytotoxic function.

a,b, Representative confocal and brightfield images of two clonally expanded MCF10A cell lines expressing Muc1-GFP with 42 tandem repeats (TRs) under the control of a tetracycline inducible promoter at the indicated doxycycline induction level. Scale bar, 10 μm. c, Representative confocal and brightfield images of Muc1 expresssing B2 clonal cells labelled with anti-Human Muc1 Janelia Fluor 549; the doxycycline induction level is indicated. Scale bar, 10 μm. d, NK-92 cell-mediated cytotoxicity against the polyclonal Muc1-GFP and three Muc1-GFP expressing clonal cell lines (1E7, 2E4, and 2G9) expressing MCF10A line from which the clonal lines were isolated, and a clonal MCF10A cell line expressing the Muc1 construct without the GFP reporter (B2 clone); the doxycycline induction levels are indicated. Results are the mean ± s.d of n=3 experimental replicates. NK cell to target cell ratio is 5:1. e. Mean fluorescence intensity of cell-surface Muc1-GFP in 1E7 cells treated with the indicated concentration of StcE mucinase; cell-surface Muc1-GFP probed with Alexa Fluor 647 conjugated GFP nanobody. Results are shown for induction of Muc1-GFP at 100 ng/ml and 1000 ng/ml doxycycline. Results are the mean ± s.d. of at least 14 cells from one representative of 3 independent experiments. Statistical analysis by one-way ANOVA with Tukey’s post-hoc tests.

Extended Data Fig. 2:. Additional data supporting Siglec-independent protection by the mucin layer.

a, b, Surface expression levels of CD3 (a) and CD56 (b) in NK-92 cells and primary NK cells analyzed by flow cytometry, validating isolation of CD56+/CD3− NK cells. c, d, Surface levels of Siglec-7 (c) and Siglec-9 (d) on NK-92 and primary human NK cells evaluated by flow cytometry, demonstrating minimal to no detectable levels of the receptors on NK-92. e, Western blot analysis confirming UDP-N-acetylglucosamine 2-epimerase (GNE) knockout (KO) in 1E7 cells to ablate sialylation. Flow cytometry data for primary NK cells are representative results for one of n=3 human blood donors.

Extended Data Fig. 3:. Measurement of the nanoscale glycocalyx thickness with Ring Scanning Angle Interference Microscopy (Ring-SAIM).

a, Muc1-GFP size standards with biopolymer domains comprised of 0, 10, 21, and 42 tandem repeats (TRs). b, Western blot analysis of Muc1-GFP constructs stably expressed in MCF10A epithelial cells; arrows indicate fully glycosylated Muc1 with 10, 21, or 42 TRs; the primary antibody reacts with the Muc1 TRs and does not probe the 0TR Muc1. c, Flow cytometry analysis of cell-surface Muc1-GFP probed with Alexa Fluor 647 conjugated GFP nanobody. d, Examples of the pixelwise SAIM interferograms for Muc1-GFP constructs labelled with Alexa Fluor 647 conjugated GFP nanobody. e, Representative widefield image and glycocalyx thickness map of live MCF10A epithelial cells expressing the indicated Muc1 constructs labeled with Alexa Fluor 647 conjugated GFP nanobody; the Muc1 x42 mutant construct is comprised of 42 TRs of the mutated Muc1 tandem repeat – PDARPAPGATAPPAHGVTAA – which has three of the five serine/threonine O-glycosylation sites mutated to alanine. Scale bars, 10 μm. f, g, Quantification of glycocalyx thickness in cells expressing the indicated Muc1 constructs (f) and the Muc1-expressing B2 clonal line induced at the indicated doxycycline levels (b). Boxes and whiskers show the first and third quartiles (boxes), median, and range of the data. Each condition include a minimum of 19 cells from a representative experiment (n=3 independent experiments). Statistical analysis by one-way ANOVA with Tukey’s post-hoc tests. h, Log of the mean glycocalyx thickness versus log of the mucin biopolymer length from (f); best-fit line to a power-law model with slope = 0.16 and R²= 0.62. i, Log of the mean glycocalyx thickness in 1E7 cells expressing Muc1-GFP at varying doxycycline induction levels versus the log of the mean GFP nanobody signal, a proportionate measure of Muc1 surface density; plotted data from Fig. 2f and Fig. 1d; best-fit line to a power-law model has slope = 0.06 and R² = 0.79. j, Quantification of glycocalyx thickness in cells expressing the Muc1 x42 and the Muc1 x42 mutant construct. Boxes and whiskers show the first and third quartiles (boxes), median, and range of the data. Each condition includes a minimum of 23 cells from a representative experiment (n=3 independent experiments). Statistical analysis by two-tailed t tests. Unless otherwise indicated, doxycycline induction of mucin expression is at 1,000 ng/ml.

Extended Data Fig. 4:. Validation of glycoengineered cell lines.

a, Agarose gel electrophoresis of PCR-amplified and EcoRI-digested segment of C1GALT1 showing successful homozygous knockout (KO) in 1E7 cells (See 2A7 sub-clone of 1E7); the homology directed repair (HDR) template for CRISPR/Cas9 mediated C1GALT1 KO introduces a unique EcoRI site. b, Sangersequencing results of a PCR-amplified segment of C1GALT1 with insertion of EcoRI restriction site (GAATTC) and stop codon (TAA) in the 1E7 C1GALT1 KO clone, verifying homozygous KO. (Top: 1E7; bottom: 1E7 C1GALT1 KO) c, Fluorescence images of Glucosaminyl (N-Acetyl) Transferase 1 (GCNT1) in wild-type and GCNT1 overexpressing (GCNT1 OE) 1E7 cells with Muc1-GFP expression induced at 1,000 ng/ml of doxycycline; GCNT1 is localized in the Golgi. d, Relative composition of Core O-glycans in wild-type, 1E7 GCNT1 OE, and 1E7 GNE KO cells measured with the Cellular O-glycome Reporter (CORA) method e, Flow cytometry analysis of wild-type and glycoengineered 1E7 cells. Results for the parental MCF10A epithelial cells that do not express Muc1-GFP (-Muc1-GFP) are included as a control. f, Western blot showing the relative size of Muc1 in the wild-type and glycoengineered 1E7 cells. g, sWGA lectin blot after extended SDS-PAGE run time showing increased molecular weight of Muc1-GFP in 1E7 GCNT1 OE cells compared to wild-type 1E7 cells. h, Rate of proliferation of wild-type and glycoengineered 1E7 cells. Results are the mean of 2 independent experiment replicates. i, Flow cytometry analysis of 1E7 cells and their glycoengineered progeny induced at 0 ng/ml and 1,000 ng/ml of doxycycline.

Extended Data Fig. 5:. Galectin interactions with Muc1 O-glycans.

a, Quantification of recombinant human galectin-3 (rhGal-3) binding to wild-type, C1GALT1 knockout (KO), and GCNT1 overexpressing (OE) 1E7 cells; data normalized to the Muc1 signal intensity and presented as the mean and ± s.d. of 20 cells from one representative of n=3 independent experiments. b, Same as in a, with 5,000 U/mL PNGase F treatment to selectively remove N-glycans prior to analysis. c, Quantification of endogenous cell-surface galectin-3 (Gal-3) and Muc1 before and after treatment with 200 nM StcE mucinase. Results are the mean ± s.d. of at least 11 cells from one representative of n=3 independent experiments. Statistical analysis by two-tailed t tests. d, Representative confocal images of Muc1-GFP and immuno-labelled Gal-3 in wild-type and LGALS3 overexpressing (LGALS3 OE) 1E7 cells. Scale bars, 10 μm. e, Quantification of endogenous Gal-3 on the 1E7 cell surface following treatment with the indicated concentration of galectin inhibitor, TD139; data normalized to the Muc1 signal intensity and presented as the mean ± s.d. of 10 cells from one representative of n=3 independent experiments. f, Representative confocal images of endogenous galectin-3 on the cell membrane in presence and absence of 50μM TD139. Scale bars, 10 μm. g, Quantification of recombinant human galectin-1 (rhGal-1) binding to wild-type and glycoengineered 1E7 cells; data normalized to the Muc1 signal intensity and presented as the mean ± s.d. of 20 cells from one representative of n=3 independent experiments. h, Same as in g with 5,000 U/mL PNGase F treatment to selectively remove N-glycans prior to analysis. i, FRAP analysis of Muc1-GFP in wild-type and 1E7 GCNT1 OE cells treated with control buffer, as well as 1E7 GCNT1 OE cells treated with 10 μM rhGal-1. Results are the mean ± s.d. of intensity profiles of n=11, 10, 9 cells, respectively. j,k, Quantification of mobile fraction (j) and half recovery time (k) from (i). In all experiments, Muc1-GFP expression was induced with 1,000 ng/ml of doxycycline. Unless otherwise indicated, statistical analysis by one-way ANOVA with Tukey’s post-hoc tests.

Extended Data Fig. 6:. Apoptosis is not inhibited by Muc1-GFP expression.

a, Quantification of 1E7 cell death resulting from treatment with the indicated concentrations of recombinant perforin for 30 minutes; Muc1-GFP expression induced at 1, 100, 1,000 ng/ml doxycycline. Results are the mean ± s.d. of n=3 technical replicates for one representative of three independent experiments. b, Quantification of cell viability (%) of uninduced (0 ng/ml) and doxycyclineinduced (1,000 ng/ml) 1E7 cells following treatment with indicated concentration of doxorubicin for 48 hours. Results are the mean ± s.d. of n=3 technical replicates.

Extended Data Fig. 7:. Validation of endogenous labeling of Granzyme B (GzmB) in NK-92 cells.

Using CRISPR/Cas9 and an appropriate homology directed repair template, the gene sequence for a flexible GGSGGSGGS linker and the pH-tolerant green fluorescent protein, Gamillus, was inserted immediately before the endogenous GZMB stop codon in NK-92 cells. a, Representative brightfield and fluorescence images of NK-92 GzmB-Gamillus cells stained with Lysotracker Deep Red. Scale bar, 10μm. b, Pearson’s coefficients of correlation between GzmB-Gamillus and Lysotracker. Results are the mean ± s.d. of 20 cells. c,d, NK-92 GzmB-Gamillus mediated cytotoxicity against 1E7 with Muc1-GFP expression induced at the indicated doxycycline concentration. Results are the mean ± s.d. of n=3 independent experimental replicates. NK cell to target cell ratio is 10:1. Statistical analysis by one-way ANOVA with Tukey’s post-hoc tests. e, Western blot showing the relative expression level of Granzyme B (GzmB) in the NK-92 cells and NK-92 GzmB-Gamillus knock-in (KI) cells. Note that the acidlabile linker between GzmB and Gamillus is expected to undergo cleavage in the low pH of lysosomes and cytolytic granules.

Extended Data Fig. 8:. Supporting characterization for StcE NK-92 and Zip-NK-92 cells.

a, Representative confocal images of Muc1-GFP on 1E7 cells treated with control buffer or 100 nM StcE, StcE ΔX409, StcE ΔX409 fused to the EE leucine zipper (StcE-EE); Alexa Fluor 647 conjugated GFP nanobody is used to probe cell-surface Muc1-GFP constructs. Scale bar, 10 μm. b, Viability of NK-92 and Zip-NK-92 after treatment with indicated concentration of StcE-EE. Propidium iodide (20 μg/ml) was used to detect dead cell population by flow cytometry. c, Cytotoxicity mediated by Zip-NK-92 cells coupled with the indicated concentration of Sialidase-EE zipper fusion protein (Sialidase-EE) against B2 cells at the indicated doxycycline induction level; NK cell to target cell ratio is 5:1. Results are the mean ± s.d. of n=3 technical replicates for each doxycycline concentration.

Fig. 1:. Muc1-mediated protection against Natural Killer (NK) cells.

a, Doxycycline inducible system for controlled expression of Muc1 or Muc1-GFP possessing 42 tandem repeats (TRs) in MCF10A epithelial cells. b, Isolation of stable cell clones with titratable expression of Muc1-GFP (1E7 clone) or Muc1 (B2 clone). c, Fluorescence and brightfield images of 1E7 cells induced at the indicated concentration of doxycycline. Scale bar, 10 μm. d, Flow cytometry analysis of 1E7 cells showing titratable Muc1-GFP levels across the indicated concentrations of doxycycline inducer. e, NK-92 cell-mediated cytotoxicity against 1E7 cells treated with control buffer or 100 nM StcE mucinase for 1 hour prior to co-culture; Muc1-GFP induction is at the indicated doxycycline concentration. Results are the mean ± s.d. of n=3 independent experimental replicates. f, NK-92 cell-mediated cytotoxicity against an MCF10A cell line expressing a Muc1-GFP construct with the TRs deleted (Muc1 0TR) under the control of a tetracycline inducible promoter at the indicated doxycycline induction level. Results are the mean ± s.d of n=3 technical replicates. g, Primary human NK cell-mediated cytotoxicity against 1E7 cells and 1E7 GNE KO cells at the indicated doxycycline concentration. NK cell to target cell ratio is 5:1. Results are the mean ± s.d. of n=3 biological replicates corresponding to three distinct human blood cell donors. h, Primary equine PBMC-mediated cytotoxicity against 1E7 cells at the indicated concentration of doxycycline; NK cell to target cell ratio is 20:1. Results are the mean ± s.d. of n=3 biological replicates corresponding to unique equine blood donors. i, NK-92 cell-mediated cytotoxicity against ZR-75-1 breast cancer cells sorted into sub-populations with low, intermediate, and high endogenous Muc1 expression. NK cell to target cell ratio is 10:1. Results are the mean ± s.d. of n=3 experimental replicates. j, NK-92 cell-mediated cytotoxicity against ZR-75-1 treated with control buffer or 100 nM StcE prior to co-culture. NK cell to target cell ratio is 10:1. Results are the mean ± s.d. of n=3 independent experimental replicates. Statistical analysis by two-tailed t tests. In panels e, g, and h, statistical analysis by one-way ANOVA with Tukey’s post-hoc tests.

Fig. 2:. Molecular determinants of the protective mucin layer.

a, Schematic illustration of Ring-SAIM; interference of direct and reflected laser light off a silicon mirror creates standing waves of excitation light that depend on the laser incidence angle, θ, and are used to probe the glycocalyx thickness, which is measured as the gap between the substrate and fluorescently labelled membrane. b, Optical configuration for Ring-SAIM showing azimuthally scanned excitation light at varying θ; AOTF, Acousto-optic tunable filter. c, Representative widefield image and glycocalyx thickness map of live 1E7 cells expressing Muc1-GFP at the indicated doxycycline induction level. Scale bars, 10 μm. d, Glycocalyx thickness in ZR-75-1 cells before and after treatment with 100 nM StcE. Boxes and whiskers show the first and third quartiles (boxes), median, and range of the data. Each condition includes a minimum of 6 cells from a representative experiment (n=3 independent experiments). Statistical analysis by two-tailed t test. e, Biosynthetic pathway for mucin-type O-glycans. f, Relative O-glycan distributions measured by the Cellular O-glycome Reporter Amplification (CORA) analysis in 1E7 wild-type, 1E7 GCNT1 overexpressing (OE), 1E7 C1GALT1 knockout (KO), and 1E7 GNE KO cells. g Glycocalyx thickness in the 1E7 cell panel at the indicated doxycycline concentration; some cellular treatments include 10 μM recombinant human galectin-1, 5 μM recombinant human galectin-3, or 10 μM TD139. Boxes and whiskers show the first and third quartiles (boxes), median, and range of the data. Each condition includes a minimum of 13 cells from a representative experiment (n=3 independent experiments). Statistical analysis by one-way ANOVA with Tukey’s post-hoc tests. h, NK-92 cell-mediated cytotoxicity against the wild-type and engineered 1E7 cells at the indicated doxycycline induction level. NK cell to target cell ratio is 5:1. Results are the mean ± s.d. of at least n=3 independent experimental replicates. Statistical analysis by one-way ANOVA with Tukey’s post-hoc tests. i, NK-92 cell-mediated cytotoxicity is inversely proportional to the measured glycocalyx thickness with Pearson correlation coefficient, r = −0.91; plotted data is from (g) and (h); dashed line indicates a linear fit to the data.

Fig. 3:. Enhanced receptor-mediated activation can overcome the mucin barrier.

a, Representative fluorescence images of endogenous GzmB-Gamillus in NK-92 cells following conjugation with Muc1-expressing B2 cells at minimal and maximal doxycycline induction of Muc1 expression; B2 cells also express cytosolic mScarlet-I as a red fluorescent marker. Scale bar, 10 μm. b, Polarity index for Gamillus-labelled cytolytic granules in NK-92 cell conjugates with B2 cells at the indicated doxycycline induction of Muc1 (0 and 1,000 ng/ml). Results are the mean ± s.d. of at least 38 cells over n=3 independent experimental replicates. Statistical analysis by two-tailed t test. c, Schematic showing structure of chimeric antigen receptor (CAR) targeting HER2 and CD19. d, Western blot analysis of HER2 in wild-type and HER2 overexpressing 1E7 cells (+ HER2 OE). e, NK-92 and HER2 CAR NK-92 cell-mediated cytotoxicity against 1E7 HER2 OE cells at the indicated concentration of doxycycline. f, HER2 CAR-T cell-mediated cytotoxicity against 1E7 HER2 OE cells at the indicated concentration of doxycycline. g, HER2 CAR T cell-mediated cytotoxicity against 1E7 HER2 OE cells at the indicated doxycycline induction level and target:effector ratio. h, Flow cytometry analysis of CD19 surface levels on 1E7 wild-type and CD19 overexpressing 1E7 cells (+ CD19 OE) cells at the indicated concentrations of doxycycline. i, NK-92 and CD19 CAR NK-92 cell-mediated cytotoxicity against 1E7 or CD19 OE 1E7 cells at the indicated concentration of doxycycline. n=3 independent experimental replicates. j, CD19 CAR-T cell-mediated cytotoxicity against 1E7 and 1E7 CD19 OE cells at the indicated concentration of doxycycline. k, CD19 CAR T cell-mediated cytotoxicity against 1E7 and 1E7 CD19 OE cells at the indicated doxycycline induction level and target:effector ratio. Unless otherwise indicated, effector cell to target cell ratio is 5:1. All results in e-g and i-k are shown as mean ± s.d. of n=3 independent experimental replicates. In panels e-f and i-j, statistical analysis by one-way ANOVA with Tukey’s post-hoc tests. HER2 CAR NK-92/T and CD19 CAR-T cells are second-generation CARs, while CD19 CAR NK-92 is a third-generation CAR.

Fig. 4:. Tethering StcE mucinase to the NK-92 cell surface enhances their cytolytic efficiency.

a, StcE NK-92 cells prepared through tethering of StcE mucinase to the NK-92 cell surface by the native StcE X409 lectin domain. b, Surface levels of StcE and X409-deleted StcE (StcE ΔX409) on NK-92 cells following mucinase incubation at the indicated concentrations. c, ZR-75-1 cell killing by StcE NK-92 cells prepared with the indicated StcE concentration. NK cell to target cell ratio is 10:1. Results are the mean ± s.d. of n=3 technical replicates. d, Killing of 1E7 cells by NK-92, StcE NK-92, NK-92 cells tethered with catalytically dead StcE E447D, and NK-92 cells treated with StcE ΔX409. NK cell to target cell ratio is 10:1. Results are the mean ± s.d. of n=3 independent experimental replicates. e, Brightfield and Muc1-GFP fluorescence images of 1E7 cells conjugated with NK-92 or StcE NK-92 cells; lower images show zoomed-in regions of the NK-1E7 cell interface. Scale bar, 10 μm. f, Mean Muc1-GFP intensity (dark lines) and standard deviations (lighter bounding areas) along line profiles across the contact interface between 1E7 cell and NK-92 cells (n=6) or StcE NK-92 cells (n=10); the intensity is normalized to the mean intensity outside of the contact interface; red dotted arrows in e show representative profile lines. g, Glycocalyx thickness before and after treatment with 100 nM StcE. Boxes and whiskers show the first and third quartiles (boxes), median, and range of the data. Each condition includes a minimum of 11 cells from a representative experiment (n=3 independent experiments). Statistical analysis by two-tailed t test. h, Cytotoxicity mediated by control and engineered NK-92 cells. Results are the mean ± s.d. of n=3 target cancer cell lines with individual data points indicating average cytotoxicity against a specific cell line (SKOV3, SKBR3, and 1E7 HER2 OE). NK cell to target cell ratio is 5:1 for 1E7 HER2 OE and 10:1 for SKOV3 and SKBR3. In d and h, statistical analysis by one-way ANOVA with Tukey’s post-hoc tests. For all experiments, Muc1-GFP expression in 1E7 and 1E7 HER2 OE cells was with 1,000 ng/ml doxycycline.

Fig. 5:. Zip-NK-92 cells for modular display of glycocalyx-editing enzymes.

a. Cartoon showing the design of Zip-NK-92 cells for modular surface display of glycocalyx-editing (GE) enzymes using leucine zippers. b, Flow cytometry analysis of sfGFP and ALFA-Tag on wild-type NK-92 and Zip-NK-92 cells expressing the membrane-anchored RR zipper. c, Binding isotherms for StcE-EE zipper fusion protein (StcE-EE) incubated with Zip-NK-92 cells (red) and wild-type StcE incubated with NK-92 cells (black). Representative normalized intensity plot from n=3 independent experimental replicates. d, Cytotoxicity mediated by Zip-NK-92 cells coupled with StcE-EE at the indicated concentration against Muc1-expressing B2 cells at the indicated doxycycline induction level. NK cell to target cell ratio is 5:1. Results are the mean ± s.d. of n=3 technical replicates for each doxycycline level. e, f, Cytotoxicity mediated by Zip-NK-92 cells coupled with the indicated concentration of StcE-EE (e) and Sialidase-EE (f) against target cancer cell lines. NK cell to target cell ratio is 10:1. Results are the mean ± s.d. of n=4 or 5 target cancer cell lines (Capan-2, SKBR3, KPL-1, T47D, ZR-75-1). Statistical analysis by two-tailed t test.