Cancer-specific glycosylation of CD13 impacts its detection and activity in preclinical cancer tissues

Summary Harnessing the differences between cancer and non-cancer tissues presents new opportunities for selective targeting by anti-cancer drugs. CD13, a heavily glycosylated protein, is one example with significant unmet clinical potential in cancer drug discovery. Despite its high expression and activity in cancers, CD13 is also expressed in many normal tissues. Here, we report differential tissue glycosylation of CD13 across tissues and demonstrate for the first time that the nature and pattern of glycosylation of CD13 in preclinical cancer tissues are distinct compared to normal tissues. We identify cancer-specific O-glycosylation of CD13, which selectively blocks its detection in cancer models but not in normal tissues. In addition, the metabolism activity of cancer-expressed CD13 was observed to be critically dependent on its unique glycosylation. Thus, our data demonstrate the existence of discrete cancer-specific CD13 glycoforms and propose cancer-specific CD13 glycoforms as a clinically useful target for effective cancer-targeted therapy.


INTRODUCTION
2][3] Consequently, a significant effort has been focused on identifying novel targets that deliver drugs effectively with high specificity to cancer cells, while being less toxic to normal cells.One of such targets is CD13, [4][5][6] though it has thus far failed to fulfill its clinical potential due to the current lack of knowledge of its tissue-specific expression. 7,8D13 (Aminopeptidase N, APN) is a heavily glycosylated protein, whose expression and multiplicity of function have strongly been implicated in the hallmarks of cancer, particularly tumor angiogenesis and metastasis. 8,9][12] However, despite the well-reported evidence of CD13 over-expression and activity in cancer progression, the ubiquitous nature of its expression in normal tissues as generally reported remains a limitation to the full potential clinical utility of CD13 as a target for cancer therapeutics and drug discovery being exploited. 9D13 is regarded as a ubiquitously expressed glycoprotein, yet its multiple functions (often termed ''moonlighting activities''), as a peptidase, receptor, or signaling molecule, are well established as tissue dependent. 5,8,13,14It remains poorly understood as to how the same protein can function differently in cancers compared to normal tissues.][17] Glycosylation is a common posttranslational modification of proteins and is known to contribute significantly to the generation of different forms of a protein with altered localization and function. 18,19The glycosylation of CD13 accounts for about 20%-40% of its observed molecular weight (240-150 kDa), with several distinct glycoforms of CD13 that differ in size reported to be present in a single cell due to variations in glycosylation. 14,15Unfortunately, the implications of this differential glycosylation on CD13 expression and function, particularly in a tissuespecific manner, remain unestablished.Altered protein glycosylation is a common feature of cancer cells, which leads to modulation of structure and conformation, structure and functional activity of cancer-associated glycoprotein as compared to their normal healthy tissue counterparts. 20,21We therefore hypothesized that the glycosylation of CD13 is tissue specific, with cancer tissues-expressed CD13 differentially glycosylated as compared to normal tissues.The identification of cancer-specific glycoforms presents a novel targeted therapeutic

Validation of available CD13 antibodies with different epitope-binding CD13
The expression of CD13 across various tissues both in normal and cancers has been widely reported, though results are contradictory and highly dependent on the antibody and detection technique used.Hence, we first validated the reactivity of the antibodies used in this study to CD13 expression.Three commercially available CD13 antibodies, which recognize specific epitopes within the N-terminal [mAb 1-250], mid-region [mAb 400-500], and C-terminal [mAb 687-967] of the CD13 protein, (Figure 1A), in addition to 3D8 were used in this study.The 3D8 epitope is unknown, though it remains a highly published anti-CD13 antibody 16,22,23 The reactivity of the antibodies was tested using a CD13-CRISPR knockout THP-1 cell line.Anti-CD13 antibodies, mAb 1-250, mAb 400-500, and mAb 687-967, demonstrated reactivity to CD13 expression ($160 kDa) in CD13 wild-type THP-1 cells, but not in CD13 knockout THP-1 cells.Interestingly, 3D8 showed no reactivity to CD13 expression in THP-1 wild-type cells, despite CD13 expression in these cells being detected by the other antibodies (Figure 1B), and mRNA expression being confirmed (Figure S1).

CD13 expression in tissues varies depending on the antibody epitopes
Next the expression of CD13 in homogenized HT1080, MCF-7, and MDA-MB-231 cell line-derived xenograft (CDX) tissues, and normal mouse tissues (liver and kidney) was assessed ex vivo using the four anti-CD13 antibodies.We observed an inconsistent tissue expression profile of CD13 across the antibodies.All antibodies including 3D8 (which failed to detect CD13 in THP-1 wild-type cells) detected CD13 of similar size ($160 kDa) in HT1080 CDX .CD13 expression in liver ($150 kDa) and kidney ($160 kDa) tissue homogenates was detected by mAb 1-250, 400-500, and 687-987, but not 3D8.Interestingly, the expression of CD13 ($165 kDa) in breast cancer CDX (MCF-7 CDX & MDA-MB-231 CDX ) was faintly detected only by mAb 400-500 (Figures 1C and 1E).However, contrary to the observed antibody-dependent CD13 detections in cancer CDX tissues and normal tissues at the protein level, CD13 mRNA was detected in all examined tissues (Figure 1F).
We also evaluated the expression of CD13 in cancer cells and compared this to the corresponding xenograft tissue.Consistent with the lack of CD13 mRNA expression in MCF-7 and MDA-MB-231 cells (Figure 1F), none of the examined antibodies significantly detected CD13 protein in these cells (Figure 1D), although it is worth stating that mAb 687-967 and 3D8 detected very faint CD13 expression in MDA-MB-231 cells but not in MDA-MB-231 CDX .A significantly high fold increase both at mRNA level, >2000-fold (Figure 1G), and protein level, >60-fold, (detected by only mAb 400-500, Figure S2), was observed with MCF-7 CDX and MDA-MB-231 CDX homogenate tissues compared to their respective cancer cells.With the exception of mAb 1-250, which demonstrated about 2-fold relative CD13 protein expression in HT1080 CDX compared to the detection of CD13 in HT1080 cells, no significant change in CD13 protein expression was observed in HT1080 cells and respective CDX by mAb 400-500, mAb 687-967, and 3D8 (Figure S2), despite an approx.1,000-fold increase in mRNA expression.(Figure 1G).

Glycosylation of CD13 masks detection of tissue-specific glycoforms
The observed inconsistent CD13 detection by the different epitope-binding antibodies in the examined tissues is indicative of tissue-specific posttranslational modifications that generate tissue-specific forms of CD13 with differential reactivity to the antibodies.To test this hypothesis, we investigated the direct influence of glycosylation on the reactivity of the antibodies.First, we examined the effect of N-glycosylation on the reactivity of the antibodies.Peptide:N-glycosidase F (PNGase-F) is an endo-glycosidic enzyme that efficiently removes all N-glycans from a glycoprotein. 24Tissue samples were treated with PNGase-F before analysis by western blot.
Treatment with N-glycosidase significantly shifted CD13 band from $150-160 kDa to $115-120 kDa in all tissues with respective detecting antibodies, demonstrating the heavy N-glycosylated nature of CD13 (Figure 2A).Interestingly, the reactivity of the antibodies was observed to be significantly altered in a tissue-dependent manner after the removal of N-glycans.Enhanced reactivity of mAb 1-250 and mAb 687-967 to CD13 was observed only in N-glycosidase-treated HT1080 CDX and kidney tissues (Figures 2A, 2B, and 2D).Reactivity of mAb 400-500 was significantly enhanced in only N-glycosidase-treated MCF-7 CDX and MDA-MB-231 CDX (Figures 2A and 2C).In contrast, the reactivity of 3D8 to CD13 was observed to be significantly diminished in N-glycosidase-treated HT1080 CDX (Figures 2A and 2E).Meanwhile, the reactivity of antibodies in liver tissue was observed to be seemingly unaffected by PNGase-F treatment, although CD13 detection by mAb 1-250 and mAb 687-967 revealed a subset of CD13 glycoforms apparently present only in liver, which remain unaffected by N-glycosidase activity (Figure 2A).The observed differential reactivity of the antibodies to the same protein in different tissues after the removal of N-glycans (Figure 2F) is indicative of tissue-differential N-glycosylation of CD13 around the antibody recognition epitopes.We next evaluated the effect of O-glycosylation on the reactivity of the antibodies.O-glycosidase releases the Core 1 and Core 3 disaccharides (Galb1,3GalNAc and GlcNAcb1,3GalNAc) from O-glycans, which are linked to serine or threonine residues. 25,26However, the presence a terminal sialic acid residue is known to hinder O-glycosidase from functioning.Hence, for efficient hydrolysis of O-glycans, a combination of neuraminidase and O-glycosidase is required. 26The removal of O-glycans greatly enhanced the reactivity of anti-CD13 antibodies, particularly N-and C-terminal epitope-binding antibodies mAb 1-250 ($30-to 50-fold) and mAb 687-967 ($12-to 18-fold), to CD13 in MCF-7 CDX and MDA-MB-231 CDX tissues despite their inability to detect the glycosylated form of CD13 (Figures 3A, 3B, and 3D).
However, in normal tissues the removal of O-glycans either did not affect the reactivity of anti-CD13 antibodies or diminished reactivity (Figures 3A-3D).The removal of O-glycans was observed to selectively enhance the detection of CD13 in breast cancer CDX tissue by the anti-CD13 antibodies, particularly those with epitope-binding domains at the terminal ends of the CD13 protein (i.e., mAb 1-250, mAb 687-697, Figure 3E).Together, these data suggest tissue-specific glycosylation profiles that pose differential structural hindrance to CD13 antibody binding to CD13 in various tissues.These findings explain the contradictory and inconsistent reported tissue expression of CD13 in the literature.This suggests CD13 is likely to exist as different glycoforms in a tissue-specific manner.

Glycoforms of CD13 are tissue specific
To confirm the possibility that CD13 may exist as a distinct glycoforms in cancers and normal tissues, specific glycan-binding lectins (Table 1) were used to capture CD13 in cancer CDX and normal tissues.Captured CD13 glycoforms were assessed by CD13 immunoblotting using anti-CD13 antibody mAb 400-500.Although few significant differences were observed between HT1080 CDX and kidney tissue lectin-captured CD13 populations, populations captured in breast cancer CDX tissue (MCF-7 CDX & MDA-MB-231 CDX ) were significantly distinct compared to normal tissues and HT1080 CDX suggesting tissue-specific heterogeneous populations (glycoforms) of CD13.Focusing on the captured sialoglycans [a2,3-(MAL-II) and a2,6-linked sialic acid (SNA)] populations, CD13 present in breast cancer CDX tissue was observed to present different sialylation, particularly a2,3 sialoglycans, compared to normal tissues.In breast cancer CDX tissue, CD13 was less detected in captured a2,6-sialoglycan population though relatively prominently detected in captured a2,6-sialoglycan population.This was contrary to observations in normal tissues (Figure 4A; Table 2).This is a further confirmation of the tissue-specific manner of CD13 glycosylation, but more importantly an indication of a distinctive sialylation pattern on CD13 in breast cancer CDX tissue tissues compared to normal healthy liver and kidney.This observation appears to correlate with an observed lack of expression of a2,3 sialidases (Neu-2 and Neu-4) in breast cancer CDX tissue (Figure S3A).As demonstrated in Figures 2 and 3, the glycosylation of CD13 around the antibody epitopes leads to diminished recognition, likely to be due to structural hindrance to antibody binding.We therefore investigated the position of a2,3-and a2,3-sialoglycans in relation the recognition epitope of anti-CD13 antibody (mAb 400-500).This was done by incubating the immunoblot of cancer CDX and normal tissues with SNA and MAL-II lectins before the detection of CD13 using mAb 400-500.The lectin binding to sialic acids on CD13 will hinder CD13 detection if it occurs close to the mAb antibody recognition epitopes.Lectin blotting was performed to confirm the glycan affinity of lectin used (Figure S3B).In the presence of MAL-II (a2,3-linked sialic acid) lectin binding, CD13 detection by anti-CD13 antibody (mAb 400-500) was significantly reduced in breast cancer CDX tissues compared to normal tissues.However, the SNA (a2,6-linked sialic acid) lectin binding was observed to have no effect on the reactivity of mAb 400-500 to CD13 in all tissues.(Figures 4B and 4C).

DISCUSSION
We have demonstrated that although CD13 is heavily glycosylated in both cancers (CDX) and normal tissues (liver and kidney), the nature (identity and position) of this glycosylation differs across tissues and as such poses tissue-specific structural hindrance to the detection by CD13 antibodies depending on the epitope-binding recognition of the antibody in question.This explains the largely inaccurate and contradictory reports on the expression of CD13 in these tissues, using different monoclonal antibodies and detection techniques. 16,22,27eglycosylation of CD13 before antibody detection may provide a more accurate CD13 expression profile across tissues, particularly, when using CD13 antibodies with binding epitopes within the N and C termini of the CD13 protein.The inactivation of peptidic epitopes by glycosylation as observed here has recently been reported in detection of various glycoproteins in cancer, not least in the detection of cancer immunotherapy target PD-L1. 28,29These data highlight the profound impact of differential tissue glycosylation of CD13 on its detection by antibodies, particularly in cancers.
We have additionally demonstrated that the frequently utilized anti-CD13 antibody that has appeared in a large number of publications, 22,30 3D8, whose epitope-binding characteristics remain unknown, may likely require specific glycosylation on CD13 for its reactivity (Figure 2).Thus, considering the demonstrated tissue-differential glycosylation of CD13, this may explain why the reactivity of this antibody was observed only in HT1080 CDX tissues: perhaps the required glycosylation is absent in the other examined tissues.This is not a unique phenomenon to this CD13 antibody, as similar observations have been reported with other antibodies. 29,31Caution should be exercised when assessing reports of CD13 expression with this antibody, particularly when comparing CD13 expression across different tissues and conditions.
CD13 is an established regulator of tumor angiogenesis and is markedly induced in hypoxia and in xenograft tumor growth; CD13 expression is induced in cancer cells that do not endogenously express the protein under these conditions. 32This was the case in breast cancer cells and respective xenograft tissues (Figures 1A and 1E).Importantly, however, our data demonstrate that the glycosylation of CD13 induced and expressed under tumor microenvironmental conditions, particularly in breast cancer CDX tissue, is very distinct from the glycosylation observed in endogenously CD13-expressing tissues, such as HT1080 CDX and normal tissues (liver and kidney).This observation complements a key study by Curnis et al. which demonstrated the selective binding of an asparagine-glycine-arginine (NGR) peptide motif to CD13 expressed on tumor angiogenic vessels; it did not bind to CD13 expressed in normal tissues, including kidney. 5The structural basis for this selective ligand binding remains unknown.However, to explain the observation, the authors ruled out possible cancer-specific glycosylation, suggesting the possibility of a CD13 isoform with a unique conformation expressed exclusively on tumor angiogenic vessels.
On the contrary, we suggest cancer-selective O-and N-glycosylation at terminals and mid regions respectively of the CD13 protein in breast cancers.This glycosylation was demonstrated to selectively block antibody reactivity to CD13 in breast cancer in a way not observed in liver and kidney.In addition, the metabolism of a tumor-specific CD13 substrate was observed to be critically dependent on the unique CD13 glycosylation in tumors (Figures 4D and S5).This finding, along with the evidence for the distinct structural differences discussed, is the first evidence to indicate that the differential tissue-specific glycosylation of CD13 is important not only for its detection but also its activity.These data also perhaps explain the many reported tissue-specific functions of CD13 despite its ubiquitous expression. 8,33,34Glycosylation of proteins, particularly membrane proteins, is known to be critical for ligand-receptor interactions, stability, and activities of these proteins. 35,36Also, the terminal epitopes of glycoproteins have been shown to play a significant role in cell-cell interactions, cell-cell and cell-matrix adhesion, and to influence cancer metastasis. 37,38Thus, the differential glycosylation of CD13, particularly at both the N and C termini in examined tissues as demonstrated in this study may imply specific ligands and substrates with affinity for CD13 in various tissues.This is likely to account for its tissue-specific functions.
0][41][42] Using lectin capture, we have demonstrated CD13 as a sialo-glycoprotein with differential a2,3-sialylation in cancers.Although several studies have reported differential a2,3-sialylation in metastatic breast cancer, 40,43,44   including CD44 and mucins have been identified to be responsible for this observation. 40Our data suggest that cancer-specific sialylation of CD13 may significantly contribute to the hyper-sialylation observed in cancers.
The distinct aberrant cancer-associated glycans on CD13 as observed may represent an important cancer biomarker and target for anticancer-specific therapy.Glycoform-specific targeting of proteins in recent years has been demonstrated as an effective approach to limit off-target toxicities associated with current anti-cancer therapy, while increasing the antitumor specificity. 45,46Thus, the application of cancer-targeting strategies including antibody-drug conjugates (ADCs), 47 and peptide-drug conjugate (PDC), 48 which are selective to cancer-specific glycoforms of CD13, hold significant promise for cancer-specific therapy.
In conclusion, CD13 is a target with significant unfulfilled clinical potential in cancer drug discovery due its ubiquitous expression in normal tissues. 8,9This study has demonstrated for the first time that tumor microenvironment-induced and expressed CD13 in breast cancers CDX carry distinct cancer-associated glycans compared to normal tissues.In addition, evidence has been provided that these differences relate to the metabolism activity of the protein.Thus, our data suggest the existence of cancer-specific glycoforms of CD13, which are selectively present in cancers, but absent in normal tissues.This work therefore provides an important foundation for further investigations to identify and fully characterize the distinctive nature of cancer-associated glycans on CD13 particularly in clinical cancer tissues, which could lead to enhanced molecular tools, diagnostic biomarkers, as well as clinically useful targeted therapeutics.Targeting cancer-specific glycoforms as demonstrated in recent times with highly specific antibodies of cancer-specific LYPD3, 46 MUC1, 49 and CD43 50 glycoforms is a promising therapeutic strategy to selective cancer cells but not normal tissues.Hence, the characterization of the distinctive glycostructure and substrate specificity of cancer-specific CD13 glycoforms could enable the development of highly specific CD13 antibody and peptide substrate for potent ADCs and PDCs that target only cancer-expressed CD13 glycoforms. 47,48hese tools hold significant potential to a better understanding of cancer-specific functions of CD13 and, ultimately, its clinical usefulness as a target for effective precision targeted therapy.
mixture was incubated at 4 C overnight with rotation.Prewashed streptavidin-agarose (50 ml) was added and incubated was for another 4 h.The beads were washed after centrifugation and subsequently separated with 8% SDS-PAGE after boiling at 95 C for 5 min.The samples were transferred to nitrocellulose filter membranes and then the membranes were incubated with CD13/APN antibodies.

QUANTIFICATION AND STATISTICAL ANALYSIS
All results were expressed as mean G standard error of measurement (SEM) from at least 3 independent experiments.The statistical analysis was performed using GraphPad Prism 8 (GraphPad Software, San Diego, CA, USA).P < 0.05 was considered statistically significant.

Figure 1 .
Figure 1.Expression of CD13 in cancer cell, CDX, and normal tissues (A) Schematic diagram of CD13 structure showing N-glycosylation sites and epitopes on the protein where the anti-CD13 antibodies used bind.(B) Reactivity of anti-CD13 antibodies to CD13 validated in CD13-KO THP-1 cells.CD13 expression in human cancer CDX and normal tissues (liver and kidney) using different epitope-binding CD13 antibodies; (C) Western blot and (E) quantified relative expression of CD13 in cancer CDX as detected by the different

Figure 1 .
Figure 1.Continued anti-CD13 antibodies.Band intensity measured by Image Lab Software 6.1 Software 6.1 and normalized to a b-actin loading control (D) CD13 expression in cancer cells and respective xenografts using different epitope-binding CD13 antibodies.(F) CD13 mRNA expression in cancer cells, [C], and respective xenografts, [CDX], and normal tissues.(G) CD13 mRNA expression in cancer cells, [C] and respective xenografts [CDX], and normal tissues Data shown are the mean of 3 independent experiments GSEM.**p > 0.01 and ****p > 0.0001 (two-way ANOVA).

Figure 2 .
Figure 2. Differential effect of N-de-glycosylation on CD13 antibody reactivity in human cancer CDX and normal tissues (A) Western blot analyses and (B-E) quantified fold change of the reactivity of anti-CD13 antibodies on CD13 expression in human cancer CDX and normal tissue after N-de-glycosylation in mAb 1-250 (B), mAb 400-50 (C), mAb 687-967 (D), and 3D8 (E), respectively.Mean fold change in protein band intensity was measured using Image Lab Software 6.1 and normalized to a b-actin loading control (F).HeatMap of anti-CD13 antibody reactivity to CD13 in human cancer CDX and normal tissues after N-de-glycosylation.Data shown are the mean of 3 independent experiments GSEM.**p > 0.01, ***p > 0.001 and ****p > 0.0001 (two-way ANOVA).

Figure 3 .
Figure 3. Differential effect of O-deglycosylation on CD13 antibody reactivity in human cancer CDX and normal tissues (A) Western blot analyses and (B-D) quantified fold change of the reactivity of anti-CD13 antibodies on CD13 in human cancer CDX and normal tissue after O-deglycosylation in mAb 1-250 (B), mAb 400-50 (C), and mAb 687-967 (D), respectively.Mean fold change in protein band intensity was measured using Image Lab Software 6.1 and normalized to a b-actin loading control (E).HeatMap of anti-CD13 antibody reactivity to CD13 in human cancer CDX and normal tissues after O-glycosylation.Data shown are the mean of 3 independent experiments GSEM.**p > 0.01, ***p > 0.001 and ****p > 0.0001 (two-way ANOVA).

Figure 4 .
Figure 4. Human cancer CDX and normal tissue lysates were subjected to lectin affinity capture (A) Captured proteins were assessed by immunoblotting for CD13 using mAb 400-500 anti-CD13 antibody.Loading was assessed by Coomassie stain (Figure S4).(B and C) Effect of sialo-glycan lectin binding on CD13 detection using mAb 400-500 anti-CD13 antibody and in cancer CDX and normal tissues.(D) Significance of glycosylation on the metabolic half live (min) of CD13 substrate in MCF-7 CDX homogenate.Data shown are the mean of 3 independent experiments GSEM.***p > 0.001 and ****p > 0.0001 (two-way ANOVA).

Table 1 .
Lectins and their glycan-binding specificities