Chemoenzymatic Tagging of Tn/TF/STF Antigens in Living Systems

Abstract

Truncated mucin-type O-glycans, such as Tn-associated antigens, are aberrantly expressed biomarkers of cancer, but remain challenging to target. Reactive antibodies to these antigens either lack high-affinity or are prone to antigen escape. Here, we have developed a robust chemoenzymatic strategy for the global labeling of Tn-associated antigens, i.e. Tn (GalNAcα-O-Ser/Thr), Thomsen-Friedenreich (Galβ1–3GalNAcα-O-Ser/Thr, TF) and STF (Neu5Acα2–3Galβ1–3GalNAcα-O-Ser/Thr, STF) antigens, in human whole blood with high efficiency and selectivity. This method relies on the use of the O-glycan sialyltransferase ST6GalNAc1 to transfer a sialic acid-functionalized adaptor to the GalNAc residue of these antigens. By tagging, the adaptor functionalized antigens can be easily targeted by customized strategies such as, but not limited to, chimeric antigen receptor T-Cells (CAR-T). We expect this tagging system to find broad applications in cancer diagnostics and targeting in combination with established strategies.

Graphical Abstract

Aberrant expression of truncated mucin-type O-glycans is a hallmark of cancer.^[1] These truncated O-glycans, including Tn (GalNAcα-O-Ser/Thr), Thomsen-Friedenreich (Galβ1–3GalNAcα-O-Ser/Thr, TF) and STF (Neu5Acα2–3Galβ1–3GalNAcα-O-Ser/Thr, STF) antigens, are found in many carcinomas such as gastric, colon, breast, lung, esophageal, prostate, endometrial, bladder and pancreatic cancers.^{[1c–f, 2]} Upregulation of these antigens correlates with adverse outcome and decreased patient survival, attracting increased exploration of their potential use as cancer diagnostic markers as well as targets for cancer therapy.^[3]

A few high-affinity monoclonal antibodies (mAbs) against truncated mucin-type O-linked glycopeptides have been developed for diagnostics and cancer immunotherapy, with 5E5 being a prominent example.^{[3b, 4]}. Compared to truncated mucin-type O-linked glycopeptides, Tn-associated glycan antigens are more attractive targets to overcome tumor escape. Antibodies targeting glycopeptides may lose activity due to loss of proteinogenic antigens caused by amino-acid mutagenesis on the peptide backbones as a common immune escape mechanism.^[3b] By contrast, strategies that tag or target tumor-associated glycan are less susceptible to such problems because aberrant glycan expression results from alternations in non-templated metabolic pathways that usually require concerted action of multiple enzymes including those with functional redundancy.^[3b] However, currently developed antibodies targeting Tn-associated glycan antigens (Tn/TF/STF) are low-affinity IgMs.^{[3b, 4a]} Alternatively, glycan-binding lectins,^[5] e.g. Morniga G (MorG), have been developed to target Tn-associated glycan epitopes and utilized to induce the killing of human leukemia cells when combined with a photochemical strategy.^[6]

Over the past decade, chemoenzymatic glycan editing has emerged as an attractive approach for the detection and modification of cell-surface glycans.^[7] Through the incorporation of unnatural sugar analogues bearing bioorthogonal groups or biophysical probes through specific glycosyltransferases, such strategies have found applications for the detection of TF and STF antigens individually.^[8] However, to our knowledge, no chemoenzymatic method has been exploited for tumor tagging or targeting of all three Tn-associated antigens (Tn/TF/STF). Here, we report a robust chemoenzymatic strategy for tagging all three Tn-associated glycan antigens in challenging conditions such as in human whole blood with high selectivity and efficiency. Furthermore, through tagging, these difficult-to-target cancer-associated glycan antigens are decorated with an adaptor molecule that is readily targeted by tailor-made strategies such as but not limited to Chimeric Antigen Receptor T-Cells (CAR-T).

Results and discussion

Design and characterization of a chemoenzymatic strategy for selectively tagging Tn-associated antigens.

We considered core 1/core 2 O-glycan modifying enzymes as candidates for the development of chemoenzymatic strategies to tag Tn-associated antigens in living cells. In the human O-glycan biosynthetic pathway, only the ST6GalNAc family of sialyltransferases is known to modify Tn-associated antigens (Tn, TF, STF) at the GalNAc residue with an α(2, 6)-linked sialic acid.^[1d] Six ST6GalNAcs have been identified to date,^[9] of which ST6GalNAc1 and ST6GalNAc2 generate α(2, 6)-linked sialyl-Tn antigen (Neu5Acα2–6GalNAcα-O-Ser/Thr), sialyl-6TF antigen (Galβ1–3(Neu5Ac2-6)GalNAc-O-Ser/Thr), and disialyl-TF antigen (Neu5Ac2-3Galβ1–3(Neu5Ac2-6)GalNAc-O-Ser/Thr) from Tn antigen, TF antigen, and sialyl-TF antigen, respectively,^[10] whereas ST6GalNAc3 and ST6GalNAc4 synthesize disialyl-TF antigen from sialyl-TF antigen.^[11] Based on this knowledge, we decided to evaluate the feasibility of using ST6GalNAc 1–4 to label our target glycan epitopes.

In order to label Tn, TF and STF antigens for their detection, the modifying enzymes must possess a relaxed donor substrate specificity to install biophysical probes for downstream applications. Therefore, as a first step in our tool development, we sought to determine whether recombinant ST6GalNAc 1–4 could transfer an unnatural Neu5Ac analogue bearing biotin (CMP-Sia-biotin) directly to Tn, TF, and STF antigens on the cell surface (Figure 1A and Figure S1). Jurkat cells, known to express high levels of Tn-associated antigens,^[12] were treated with CMP-Sia-Biotin in the presence or absence of ST6GalNAc1-4, followed by staining with APC-conjugated streptavidin (SA-APC) and flow cytometry analysis. As shown in Figure 1B, these four glycotransferases showed different efficiency of transferring CMP-Sia-Biotin, among which ST6GalNAc1 exhibited the highest activity. Specifically, the labeling efficiency of ST6GalNAc1, as quantified by APC-associated fluorescence, was approximately 70, 191 and 4-fold higher than that of ST6GalNAc2, ST6GalNAc3, and ST6GalNAc4, respectively, under the same reaction conditions (Figure 1B). In this experiment, HEK293T cells, which express low levels of Tn-associated antigens, were used as a negative control,^{[4c, 13]} and showed a significantly weaker labeling signal compared to that of Jurkat cells for each of the ST6GalNAc-treated groups (Figure 1B). Accordingly, the labeling signal generated on HEK293T cells was defined as the background signal, and the ratio of Jurkat/HEK293T relative labeling intensity was used to assess the labeling specificity. Using these criteria, we found that ST6GalNAc1 produced the highest labeling specificity compared to the other three enzymes (Figure S2). ST6GalNAc3 and ST6GalNAc4 recognized only STF, the relative low labeling efficiency may be due to the relatively low abundance of TF antigens on the cell surface and/or limited tolerance to the CMP-Sia-Biotin donor. While both ST6GalNAc1 and ST6GalNAc2 recognize Tn, TF, and STF antigens, the relative low labeling efficiency of ST6GalNAc 2 may be due to its limited tolerance to the CMP-Sia-Biotin donor. Accordingly, ST6GalNAc1 was chosen as the tagging enzyme for method optimization. We performed labeling in the presence of different concentrations of ST6GalNAc1 and found that the labeling reached saturation at 0.1 mg/ml ST6GalNAc1 (Figure S3). In parallel, we evaluated the donor substrate (CMP-Sia-Biotin) concentrations ranging from 10 nM to 1 mM in the ST6GalNAc1-mediated labeling (Figure S4A). Remarkably, as low as 10 nM CMP-Sia-Biotin could afford detectable signal (Figure S4B), and ~ 90% maximum labeling was reached when 200 μM CMP-Sia-Biotin was used (Figure S4A). This concentration was thus chosen as the optimized condition for subsequent studies.

Figure 1.

Comparison of the tagging efficiency of ST6GalNAc1-4 in transferring CMP-Sia-Biotin to the cell surfaces. (A) The reaction scheme of tagging process mediated by ST6GalNAc enzymes. (B) Flow cytometry analysis of the tagging efficiency across ST6GalNAc 1–4 on living cell surfaces respectively. Jurkat or HEK293T cells were treated with 100 μM CMP-Sia-Biotin in the absence (control group) or presence (experimental group) of 0.04 mg/ml different ST6GalNAc enzymes at 37 °C for 30 mins, respectively, followed by staining with SA-APC and analyzing by flow cytometry. Relative MFI (relative mean of fluorescent intensity) = labeling intensity of experimental group/labeling intensity of control groups. Mean ± SD (error bars).

Besides flow cytometry analysis of the chemoenzymatic labeling efficiency, Tn/TF/STF attached proteins on the Jurkat cell surface were biotinylated by ST6GalNAc1/CMP-Sia-Biotin and analyzed by Western blot. In this ST6GalNAc1/CMP-Sia-Biotin system, strong signals appeared only when adding both ST6GalNAc1 and CMP-Sia-Biotin, while no signal was detected in the control group without ST6GalNAc1 (Figure S5). HEK293T cells were treated under this “tagging” system as a negative control and showed much weaker labeling compared to that of Jurkat (Figure S5). Meanwhile, the labeling intensity mediated by ST6GalNAc1/CMP-Sia-X was resistant to the PNGaseF treatment, indicating this chemoenzymatic reaction was specifically on O-linked glycan (Figure S5). In an additional competition experiment, pre-labeling with the natural substrate CMP-Neu5Ac (CMP-Sia) dramatically inhibited the labeling efficiency of CMP-Sia-Biotin in a dose depended manner, indicating the unnatural analogue CMP-Sia-Biotin shares the same acceptors with the natural substrate (Figure S6). Taken together, these results demonstrated the remarkable efficiency as well as high selectivity of our ST6GalNAc1/CMP-Sia-X strategy for tagging Tn-associated antigens.

Tracking the expression level of Tn/TF/STF antigens in different cell lines using the ST6GalNAc1/CMP-Sia-Biotin tagging strategy. Next, using the ST6GalNAc1/CMP-Sia-Biotin tagging approach, we analyzed the expression levels of Tn/TF/STF antigens across different human cell lines and primary cells, including breast cancer cells (MCF7, MDA-MB-231), melanoma cells (Me290), leukemia cells (Jurkat), ovarian cancer cells (Hela), endometrial cancer cells (ECC1), HEK293T cells and human peripheral blood mononuclear cells (hPBMCs). Most of the cancer cells we examined were robustly labeled by this approach, with Jurkat cells showing the highest labeling (Figure 2). By contrast, normal cells such as HEK293T and hPBMCs showed much lower levels of labeling, suggesting that they express low Tn/TF/STF antigens (Figure 2). This result suggests that our strategy can be used to differentiate cancer cells from normal cells based on their Tn/TF/STF expression.

Figure 2.

Tracking the expression level of Tn/TF/STF antigens across different cells using ST6GalNAc1/CMP-Sia-Biotin tagging strategy. Different cells were treated with 200 μM CMP-Sia-Biotin in the presence of 0.1 mg/ml ST6GalNAc1 at 37 °C for 30 mins respectively, followed by staining with SA-APC and subjected to flow cytometry analysis. MFI (mean fluorescent intensity). Mean ± SD (error bars).

Selective tagging of cancer cells in human blood.

To evaluate the feasibility of our strategy for selectively tagging cancer cells, we first mixed HEK293T (pre-stained with CellTracker Orange CMRA), Jurkat cells (pre-stained with CellTracker Green CMFDA) and unlabeled hPBMC at a ratio of 1:1:1, followed by incubating with ST6GalNAc1 and CMP-Sia-Biotin (Figure 3A). As expected, Jurkat cells showed approximately 45-fold and 28-fold higher tagging signals compare to those of HEK293T and hPBMCs, respectively (Figure 3B, 3C). To further test our strategy in a more challenging condition mimicking the in vivo scenario, we doped Jurkat cells (pre-stained with CellTracker Green CMFDA) into fresh human blood, and subjected the mixture to the ST6GalNAc1-mediated labeling using CMP-Sia-Biotin as the donor substrate (Figure 3D). As shown in Figure 3, the dominant cell population (hCD45-) that mainly consists of red blood cells showed negligible labeling. hCD45+/Green CMFDA− cells that include all hematopoietic cells from human blood except for mature erythrocytes showed very weak labeling (Figure 3E, 3F). By contrast, Jurkat cells (hCD45+/Green CMFDA+) were robustly labeled, showing approximately 43-fold higher labeling signal than that of hCD45+/Green CMFDA− cells.

Figure 3.

Tagging of Tn/TF/STF antigens in a cell mixture (A), (B), (C) and in human whole blood (D), (E), (F). (A) Dot plots show different cell populations in a cell mixture of Jurkat (pre-stained with CellTracker Green CMFDA), HEK293T (pre-stained with CellTracker Orange CMRA) and unlabelled hPBMCs. The double positive population (Green CMFDA+/Orange CMRA+) are Jurkat or HEK293T cells that are non-specifically interacted with each other. (B) Flow cytometry analysis of the labeling intensity of different cells in (A). The cell mixture was treated with 200 μM CMP-Sia-Biotin in the presence (red curve) or absence (grey curve) of ST6GalNAc1, followed by staining with SA-APC and subjected to flow cytometry analysis. (C) Quantitative analysis of the labeling intensity in (B). Mean ± SD (error bars). (D) Dot plots show different cell populations in human whole blood doped with Jurkat cells (pre-stained with CellTracker Green CMFDA) by using anti-hCD45 (PE) antibody. (E) Flow cytometry analysis of the labeling intensity of different cells in (D). The whole blood was treated with 200 μM CMP-Sia-Biotin in the presence (red curve) or absence (grey curve) of ST6GalNAc1, followed by SA-APC and anti-hCD45 (PE) staining and subjected to flow cytometry analysis. (F) Quantitative analysis of the labeling intensity in (E). Mean ± SD (error bars). MFI (mean fluorescent intensity). Notably, the Jurkat cell population and hCD45+/Green CMFDA− cell population could be clearly distinguished from each other through the tagging channel (SA-APC channel) via flow cytometry analysis, demonstrating the feasibility of our strategy in discriminate cancer cells from normal cells under challenging conditions (Figure S7).To further verify the sensitivity of this tagging system, we doped pre-stained Jurkat cells into human blood at different ratio from 10³ cells/mL to 10⁶ cells/ml, followed by chemoenzymatic tagging of CMP-Sia-Biotin. We found that tagging signal of Jurkat cells were much higher than that of hCD45+/Green CMFDA− cells from human blood at each doped ratio (Figure 4). Even at a ratio of as low as 10³ cells/ml, the tagging signal on Jurkat cells was still approximately 18-fold higher than that of hCD45+/Green CMFDA− cells from human blood (Figure 4).

Design of a TAT-CAR-T strategy.

We envision that our ST6GalNAc1/CMP-Sia-X tagging strategy can be integrated with a subsequent targeting step to achieve the selective killing of cancer cells with upregulated Tn/TF/STF expression. Inspired by the work reported by Min Soo Kim and colleagues using a bifunctional small molecule consisting of folate conjugated FITC (folate-FITC) to redirect an anti-fluorescein isothiocyanate (FITC) CAR-T toward folate receptor (FR)-overexpressing cancer cells,^[14] we conceived the design shown in Figure 5A. This design consists of two steps. In the first step, a bioorthogonal handle—methyltetrazine (Tz) is installed on the surface of cancer cells through the ST6GalNAc1-mediated Tn/TF/STF tagging. In the second step, a molecule consisting of the complementary bioorthogonal reaction group trans-cyclooctene (TCO) conjugated to FITC (TCO-PEG₃-FITC) is employed to recruit anti-FITC CAR-T cells to the tagged cancer cells (Figure 5A). The inverse electron-demand Diels-Alder cycloaddition reactions between TCOs and tetrazine are among the fastest bioorthogonal reactions known to date and both reactive partners have strict orthogonality to biological systems.^[15]

Figure 5.

The “Targeting after Tagging” (TAT) strategy followed by CAR-T directed specific cancer killing. (A) A general workflow for TAT strategy in combination with anti-FITC CAR-T targeting (TAT-CART). (B) Targeted killing of Jurkat cells (Tn/TF/STF high expression) using TAT-CART strategy. Jurkat cells were first treated with or without CMP-Sia-Tz in the presence or absence of ST6GalNAc1. CAR-T cells were then added to the labeled Jurkat cells at a ratio of 10:1 with different concentration of TCO-PEG₃-FITC (10 nM, 1 nM, 0.1 nM) respectively. %cytotoxicity was measured by the LDH release measurement assay. Mean ± SD (error bars). (C) Selective killing of Tn/TF/STF high expression cancer cell lines. Jurkat cells (Tn/TF/STF high expression) or HEK293T cells (Tn/TF/STF low expression) were first treated with or without CMP-Sia-Tz in the presence or absence of ST6GalNAc1. Then, CAR-T cells were added to Jurkat cells at a ratio of 10:1 with 1 nM TCO-PEG₃-FITC. %cytotoxicity was measured by the LDH release measurement assay. Mean ± SD (error bars). In all figures, *P < 0.05; **P < 0.01; ***P< 0.001.

To evaluate this two-step approach, we synthesized a C-5 Tz functionalized CMP-Neu5Ac analogue (CMP-Sia-Tz) and tested its activity for tagging Jurkat cells through the ST6GalNAc1-mediated chemoenzymatic glycan editing. The detection of Tz molecules transferred by ST6GalNAc1 onto the cell surface was achieved by reacting with TCO-PEG₃-FITC and staining with an anti-FITC antibody (Alexa Fluor 647). We found that ST6GalNAc1-mediated tagging was remarkably efficient, and the labeling intensity was approximately 2333-fold higher than that of the control group without adding ST6GalNAc1 (Figure S8). Significantly, the tagging was found to be highly selective towards Tn/TF/STF⁺ Jurkat cells rather than HEK293T cells with basal levels of Tn/TF/STF expression—the relative tagging signal on Jurkat cells was approximate 64-fold higher than that of HEK293T (Figure S8). Next, we quantified the decay of Tz molecules on the cell surface post ST6GalNAc1-mediated tagging. We found that sufficient amounts of Tz remained on Jurkat cells even after 46 h post tagging, which was approximately 133-fold higher than that of control group without the ST6GalNAc1 treatment (Figure S9). By contrast, at this time point the quantity of Tz molecules on the HEK293T cell surface dropped to near the same level as the control group without the ST6GalNAc1 treatment (Figure S9).

We anticipate that the Tz molecules installed on the cell surface would readily react with TCO-PEG₃-FITC to redirect anti-FITC CAR-T cells for the targeted Jurkat cell killing. To this end, we assessed the efficiency of TCO-PEG₃-FITC at different concentrations (10 nM, 1 nM, 0.1 nM) to redirect and tune the activity of anti-FITC CAR-T cells towards Jurkat cells that had been tagged with Tz. We observed that anti-FITC CAR-T cells showed significantly enhanced cytotoxicity towards Jurkat cells that were tagged through ST6GalNAc1/CMP-Sia-Tz compared to the untreated Jurkat cells or Jurkat cells treated with CMP-Sia-Tz only (Figure 5B). Importantly, the CAR-T cell-induced toxicity was controlled by TCO-PEG₃-FITC in a dose-dependent manner, indicating the tunable nature of this approach (Figure 5B). Notably, even at 1 nM of TCO-PEG₃-FITC, this system exhibits remarkable specific cytotoxicity towards the tagged Jurkat cells but low background killing towards the untagged counterparts (Figure 5B). Finally, we also assessed the CAR-T mediated killing of HEK293T cells that express basal levels of Tn/TF/STF in the presence of 1 nM TCO-PEG₃-FITC at the same condition. To our delight, whereas 50% Jurkat cells underwent lysis, under the same condition only 9% specific lysis of HEK293T cells was detected (Figure 5C). Notably, The background levels of toxicity toward normal cells (i.e. HEK293T) observed here is consistent with the reported background toxicity levels detected in other CAR-T systems.^{[4c, 16]}

Conclusions

Chemoenzymatic glycan editing has recently attracted increasing attention as a versatile approach to label and modify target-cell glycans for therapeutic applications.^[17] For example, in the elegant work of Bertozzi and colleagues, antibody-sialidase conjugates were used to selectively remove sialic acid from tumor cell surfaces, resulting in attenuated signaling through inhibitory Siglec receptors and increased susceptibility of tumor cells to antibody-dependent cell-mediated cytotoxicity (ADCC).^[17a] Using H. pylori α1,3-fucosyltransferase, which is capable of transferring an antibody-conjugated fucose to the cell surface, our own team established an antibody-cell conjugate platform that was successfully employed to augment antitumor activities of NK-92 MI cells in xenograft models.^[17c] By directly applying two solubilized glycosyltransferases, B4GalT1 and ST6Gal1, to modify antibody N-glycans in vivo, Anthony and coworkers were able to convert endogenous IgGs into anti-inflammatory mediators to attenuate autoimmune symptoms.^[17b] Inspired by these precedents, we have here developed a robust chemoenzymatic strategy to selectively tag Tn-associated antigens in living systems with high efficiency. We expect that our strategy could potentially be applied to tag or enrich Tn/TF/STF⁺ cancer cells in human blood for diagnostic purposes. Furthermore, by redirecting CAR-T cells to target cancer cells with high expression of Tn-associated antigens (Tn/TF/STF) via ST6GalNAc1-mediated chemoenzymatic glycan editing, efficient and selective killing of cancer cells vs. normal cells with basal levels of Tn/TF/STF is realized.

In addition to Biotin and the bioorthogonal tag Tz, other functional molecules can be transferred to the cell surface due to the substrate promiscuity of ST6GalNAc1. As an example, we have confirmed that Sia modified with mouse IgG (CMP-Sia-mIgG) can be successfully transferred to the Jurkat cell surface by ST6GalNAc1 (Figure S10). Based on this observation, we expect that a wide variety of functional groups may be employed as the adaptor molecule X in CMP-Sia-X and be installed onto the cell surface for different applications. For example, a toxin can be introduced to induce cancer cell apoptosis.^[18] Similarly, antibody-recruiting molecules such as α−1,3-Gal may be good candidates for recruiting pre-existing antibodies in the circulation for cancer targeting.^[19] Significantly, CMP-Sia analogues are easy to conjugate with functional molecules and allow for a wide range of applications or can be combined with established targeting modalities.

Figure 4.

Tagging Jurkat cells that are doped into human whole blood at different ratios. Jurkat cells (pre-stained with CellTracker Green CMFDA) were doped into whole human blood at ratios of 10⁶ cells/ml, 10⁵ cells/ml, 10⁴ cells/ml and 10³ cells/ml respectively, followed by ST6GalNAcI/CMP-Sia-Biotin tagging, probed with SA-APC and subjected to flow cytometry analysis. Mean ± SD (error bars). MFI (mean fluorescent intensity).

Scheme 1. Design of a Chemoenzymatic strategy for selectively tagging of cancer cells with Tn/TF/STF antigens in living systems.

Cell surface Tn/TF/STF antigens were first tagged with an adaptor molecular X through ST6GalNAc1-mediated chemoenzymatic labeling, which could be further used for targeted killing via established therapeutic strategies, e.g. CAR-T.

Abbreviations

CAR-T

Chimeric Antigen Receptor T-Cells

SA-APC

APC conjugated-streptavidin

TAT

Targeting after tagging

TCO

trans-cyclooctene

FITC

fluorescein isothiocyanate

Tz

Methyltetrazine

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.