Profiling Natural Serum Antibodies of Nonhuman Primates with a Carbohydrate Antigen Microarray

Abstract

Engineering of α-Galactosyltransferase gene-knockout pigs circumvented hyperacute rejection of pig organs after xenotransplantation in nonhuman primates. Overcoming this hurdle revealed the importance of non-α-Gal carbohydrate antigens in the immunobiology of acute humoral xenograft rejection. This study analyzed serum from seven naïve cynomolgus monkeys (blood type O/B/AB = 3/2/2) for the intensity of natural IgM and IgG signals using carbohydrate antigen microarray, which included historically reported α-Gal and non-α-Gal carbohydrate antigens with various modifications. The median (range) of IgM and IgG signals were 12.71 (7.23–16.38) and 9.05 (7.23–15.90), respectively. The highest IgM and IgG signals with narrowest distribution were from mono- and disaccharides, followed by modified structures. Natural anti-α-Gal antibody signals were medium to high in IgM (11.2–15.9) and medium in IgG (8.5–11.6) spectra, and was highest with Lac core structure (Galα1–3Galβ1–4Glc, iGb3) and lowest with LacNAc core structure (Galα1–3Galβ1–4GlcNAc). Similar signal intensities (up to 15.8 in IgM and up to 11.8 in IgG) were observed for historically detected natural non-α-Gal antigens, which included Tn antigen, T antigen, GM2 glycolipid, and Sd^a antigen. The hierarchical clustering analysis revealed the presence of clusters of anti-A antibodies and was capable of distinguishing between the blood group B and AB nonhuman primates. The results presented here provide the most comprehensive evaluation of natural antibodies present in cynomolgus monkeys.

INTRODUCTION

Organ transplantation is the treatment of choice for many patients with terminal or irreversible organ failure, but shortage of human donors demands the exploration of alternative sources of grafts. Cross-species xenotransplantation is a potential and promising answer to this challenge, and the pig is considered the suitable source of potential xenografts because of their similarities in physiology and size with humans, their high reproductive capacity, the low risk of disease transmission with designated pathogen-free donors, feasibility of genetic immunomodulation, and lower ethical concern compared with use of nonhuman primates (NHPs) as donors [1,2]. Early attempts at preventing xenograft rejection in preclinical studies in NHPs focused on overcoming hyper-acute rejection (HAR) mediated by natural antibodies directed toward grafts [3]. The major xenogeneic antigens are carbohydrate antigens on the cell surface, and the discovery of the galactose-α−1,3-galactose (Galα1–3Gal, α-Gal) epitope and the engineering of α1,3-galactosyltransferase gene-knockout (GGTA1-KO) pigs succeeded in eliminating HAR in xenotransplantation [4,5]. However, prolonged survival of xenografts is still impeded by thrombotic microangiopathy and consumptive coagulopathy due to the persistence of natural antibodies, which prevented xenotransplantation from progressing to a clinical reality [4,5]. A number of carbohydrate antigens have been suggested as candidate non-α-Gal antigens, which include N-glycolylneuraminic acid (Neu5Gc) and Sd^a antigen [6,7]. Multiple gene-knockout pigs that eliminate these xenoantigens have been created, but additional xenoantigens exist that react with elicited antibodies [8–10]. Identifying further non-α-Gal antigens still play an important role in all xenograft rejection and will be essential for developing strategies for establishing antigen-specific tolerance.

Two general approaches have been used to identify potential non-α-Gal antigens; biological studies profiling antibodies in human or NHP serum that are reactive to cells or tissues from genetically modulated pigs [8,9], and biochemical studies comparing the carbohydrate/protein profile of these pigs to human or NHP tissues [10,11]. However, due to cell culture conditions and contaminations, cellular assays have the potential risk of false positives and misinterpretation of the specificity of anti-carbohydrate antibodies [12]. The carbohydrate microarray (glycan microarray) has recently emerged as a novel tool for antibody studies of carbohydrates and glycoproteins [13–15]. Various forms of ‘arrays,’ in which glycoconjugates and carbohydrates are attached to solid supports, enable simultaneous assessment of many carbohydrate-protein interactions [13]. The understanding of natural antibody specificity toward various carbohydrate antigens provides deep insights into pre-transplantation crossmatch study results and antibody-mediated rejection, and may specify gene engineering targets and screening of potential recipients. In this study, we sought to use a high resolution carbohydrate array to perform a detailed analysis of natural anti-carbohydrate antibodies in NHPs.

MATERIALS AND METHODS

Animals

Seven naïve cynomolgus macaques (Macaca fascicularis), 6 males and 1 nulligravida female, of Mauritian origin were evaluated as recipients of islet xenografts in a preclinical study. All animals were purpose-bred and purchased from institutionally approved commercial vendors. They were imported to the U.S. at a median age of 1.4 years (range 1.1 to 2.0). All animals were quarantined via the same import facility following U.S. Centers for Disease Control guidelines. They underwent physical exam, received dental care, viral screening, tuberculosis testing, and fecal culture/float during quarantine. Animals were vaccinated for measles in the origin colony, their positive antibody status was verified. For samples used in this analysis, animals ranged in age between 3.6 to 5.2 years and had a median body weight (BW) of 4.8 kg (range 3.3 to 5.2 kg). ABO blood typing was performed by polymerase chain reaction and restriction fragment length polymorphism at Zoologix (Chatsworth, CA), and the distribution of the cohort was O (n = 3), B (n = 2), and AB (n = 2). They were housed in pairs or small groups of the same sex. They had free access to water and were fed biscuits (2055 Teklad Global 25% Protein Primate Diet or 7195 Teklad Hi-Fiber Primate Diet, Envigo, Madison, WI) based on BW. Their diet was enriched liberally with fresh fruits, vegetables, grains, beans, nuts and a multivitamin. The NHPs participated in an environmental enrichment program that included social play, toys, music, and regularly scheduled access to large exercise and swimming areas. To facilitate venous access, a totally implantable port and catheter was placed and managed using a previously described technique [16–18]. NHPs were trained to cooperate in medical procedures including blood collection [19]. Serum samples were collected via the venous access port into sterile additive-free collection tubes, allowed to clot for 30 min at room temperature, centrifuged with 2,000 x g for 10 minutes, stored at −80 °C, and shipped to Chemical Biology Laboratory, Center for Cancer Research, National Cancer Institute (Frederick, MD) for further analysis. All animal procedures were approved by the University of Minnesota Institutional Animal Care and Use Committee, and conducted in compliance with the Animal Welfare Act and adhere to principles stated in the Guide for Care and Use of Laboratory Animals.

Carbohydrate Microarray Fabrication and Binding Assay

Carbohydrate microarray fabrication was performed as previously described [14,15], and variability of the microarray assay has been evaluated in detail previously [14,20,21]. Briefly, serum samples were profiled on a carbohydrate array (version A411) containing 408 components including glycopeptides, glycoproteins, N-linked carbohydrates, O-linked carbohydrates, glycolipid carbohydrates, and some controls (Table S1) [15]. Carbohydrates and glycopeptides were conjugated to albumin to produce neoglycoproteins as reported previously [22], and the neoglycoproteins were printed on the array. As a control, we print unmodified bovine serum albumin (BSA) and human serum albumin (HSA) as array components to detect any binding to these proteins that has not been fully inhibited. The number after the abbreviation indicates the average number of carbohydrates per molecule of albumin (e.g. GM3 – 12 has an average of 12 molecules of GM3 per molecule of albumin). Each of the components was printed in duplicate to produce a full copy of the array, and each slide contained 16 replicate copies of the array. Prior to the assay, slides were fitted with a 16 well module/gasket such that 16 independent array experiments could be carried out on each slide. A pooled human serum sample was assayed in one well of every slide and used as a reference. Each macaque sample was assayed in two separate wells for IgG and two separate wells for IgM. All serum samples were diluted 1:50 in 3% BSA (w/v, Sigma-Aldrich, St. Louis, MO), 1% HSA (Sigma-Aldrich, St. Louis, MO) in PBST [PBS with 0.05% (v/v) Tween 20 (Sigma-Aldrich, St. Louis, MO)], and 100 μl of each sample was added to arrays. Bound macaque antibodies were detected by incubating with tetramethylrhodamine isothiocyanate (TRITC) goat anti-monkey IgG (Brookwood Biomedial, Birmingham, AL) or TRITC anti-monkey IgM (Brookwood Miomedical, Birmingham, AL) at a dilution of 1:250 in 1% BSA, 3% HSA in PBST at 37 °C with gentle shaking at 100 rpm for two hours. The anti-IgG reagent detects all IgG subclasses, and signals for IgM and IgG are roughly comparable on micrograms/mL level. For the human serum reference sample, bound antibodies were detected with DyLight 549 goat anti-human IgG (Jackson ImmmunoResearch, West Groove, PA), and DyLight 649 goat anti-human IgM (Jackson ImmmunoResearch, West Groove, PA). After incubation with the secondary antibodies, slides were washed seven times with PBS and dried by centrifugation (1000 rpm, 200 g) prior to scanning. All clinical/demographic information was blinded during the profiling of serum samples.

Image Processing

Slides were scanned at 5 μm resolution with an InnoScan 1100 AL fluorescence scanner (Innopsys, Carbonne, France). Image analysis was carried out with Genepix Pro 6.0 analysis software (Molecular Devices Corporation, Union City, CA) as previously described [14,15]. Spots were defined as circular features with a diameter of 90 μm. Local background subtraction (median background) was performed and the background-subtracted median pixel intensity feature was used. To minimize the impact of noise on our comparisons, spots with intensity lower than 150 (1/2 the typical background signal when analyzing IgM and IgG at 1:50) were considered too low to be measured accurately and were set to 150. The average of duplicate spots was calculated to obtain a normalized value to the reference samples. A log-transformed (base 2) was applied for each slide, and the final data value was obtained from the normalized average of data from both wells for a given macaque sample. Full microarray data can be found in the Supplemental Table S1.

Data Analysis

We calculated the sample size based on the previous study by Oyelaran et al., which used the same array system as our study [14]. We estimated that the optimum number of primates included in this study to be seven allowing us to detect a difference of four log-transformed signals (which was equal to the signal difference between α-Gal and lowest 25% carbohydrates in the Oyelaran’s study), when α = 0.05, standard deviation (SD) = 1.25, and power = 0.9. Anti-carbohydrate antibody levels were summarized as median (SD), and were compared using the Mann-Whitney U test. A hierarchical clustering analysis was performed to group associated carbohydrate fractions using ClustVis (https://biit.cs.ut.ee/clustvis/) [23]. All fractions are clustered using correlation distance and average linkage. Analyses of Variance were performed to identify array components that correlated with blood type, and to determine P values. A two-sided P value less than 0.05 was considered statistically significant. All analyses were performed with JMP Pro 14.0 (SAS Institute, Inc., Cary, NC).

RESULTS

Anti-carbohydrate antibody profiles of pre-transplant monkeys

The distribution of 50 highest antibody signals for each array component (n = 408) is shown in Figure 1A (IgM) and Figure 1B (IgG), and the details of signal intensities are shown in Table S1. Each monkey had high antibody signals to many different carbohydrate antigens both in IgM and IgG repertoires. The median (range) of IgM and IgG signals were 12.71 (7.23–16.38) and 9.05 (7.23–15.90), respectively. The carbohydrate antigens with the highest levels of signals and narrowest distribution across all individuals in IgM antibodies were monosaccharides, disaccharides, and Tn antigens [α-linked N-acetylgalactosamine (GalNAc) to serine or threonine]. O-linked Glucose, Xylose, and N-acetylglucosamine (GlcNAc) also showed high signals. In the IgG repertoire, the highest and most consistent signals across all subjects were to the α-linked O-GlcNAc glycopeptide Ac-Ser-(GlcNAcα)Ser-Ser-Gly, followed by monosaccharides and disaccharides. O-linked Glucose and Xylose, as well as Tn antigens also exhibited high signals.

FIGURE 1.

The distribution (log-transformed base 2) of 50 highest serum IgM (A) and IgG (B) natural antibody signals in seven naïve monkeys. Sequences and additional information for each array component can be found in the Supporting Information. The number after the abbreviation indicates the average number of glycans/glycopeptides per molecule of albumin. The circle represents the median of the 7 animals and the bars represent the standard deviation for the group.

To look at variation across samples and carbohydrates, the data were clustered by hierarchical clustering using correlation distance and average linkage. The heatmaps of the data allow evaluation of the relative signal from each monkey. Both in IgM and IgG repertoires, Tn antigens and blood group A antigens clustered together, respectively (Figure 2). Similarly, carbohydrates with the lacto-, ganglio-, or globo-series core structures partly clustered together both in IgM and IgG repertoires.

FIGURE 2.

Heat map of serum IgM (A) and IgG (B) natural antibody signals in seven naïve monkeys. Subjects are indicated in the rows and carbohydrates are indicated in the columns. Each rectangle represents the log-transformed (base 2) normalized signal.

Antibody signals to α-Gal carbohydrate and its analogs

Previous studies using flow cytometry confirmed that anti-α-Gal antibody in NHP is composed of primarily natural antibodies [24]. The anti-α-Gal response has recently been characterized polymorphic and the specificity of anti-α-Gal antibodies is not only determined by the α-linked galactose but also by the core chains carrying the α-Gal epitope [25]. For this reason, we compared the reactivity to different core chains carrying the α-Gal epitope both in IgM and IgG repertoires (Table 1) [26–28]. Signal toward α-Gal epitope that is β-linked to LacNAc type core structure (Galα1–3Galβ1–4GlcNAc, alphaGal) was lowest in the epitopes of α-Gal related analogs both in the IgM and IgG repertoires [11.2 (0.9) and 8.5 (0.5), respectively].

Table 1.

Comparison between different core chains carrying the α-Gal determinant

				IgM signal		IgG signal
a411 print #	Abbreviation	Description	Biological relevance	Median	SD	Median	SD
192	Galilli - 21	Gala1–3Galb1–4Glc-BSA	weak [23]	15.853	0.499	11.612	0.510
20	Bdi −23	Gala1–3Gal– BSA	NA	15.773	0.364	11.959	0.725
218	alphaGal-6-deoxy - 11	Gala1–3Galb1–4(6deoxy-GlcNAc)-HSA	NA	14.920	0.554	11.258	0.310
180	Gala3-type1 – 09	Gala1–3Galb1–3GlcNAc-BSA	strong [24]	14.868	0.632	10.992	0.581
290	alpha-Gal tetra - 17	Gala1–3Galb1–4GlcNAcb1–3Galb1-BSA	NA	14.276	0.776	10.282	0.561
76	Gal3– 07	Gala1–3Galb1–4Gala-BSA	NA	12.848	0.629	9.366	0.733
242	Bdi-g - 16	Gala1–3Galb– BSA	strong [25]	12.722	0.518	9.189	0.518
309	alpha-Gal tetra - 04	Gala1–3Galb1–4GlcNAcb1–3Galb1-BSA	NA	12.559	0.903	9.196	0.444
223	Bdi-g - 06	Gala1–3Galb– BSA	strong [25]	12.286	1.087	8.672	1.038
102	alphaGal- 08	Gala1–3Galb1–4GlcNAc-BSA	strong [24]	11.231	0.918	8.506	0.542

Biological relevance is stated high when it is reported that the antigen is expressed on wild type or gene modified porcine cells/tissues. Biological relevance is stated weak when it is reported that the antigen is not expressed on porcine cells/tissues. Numbers in brackets show reference. NA data not available

Antibody signals toward non-α-Gal pig antigens and its analogs

We have reviewed previous studies on non-α-Gal antigens in human and NHP serum; which included the Sd^a epitope [Neu5Acα2–3(GalNAcβ1–4)Gal] [8], Fucα1–3GalNAc epitope [11], Galβ1–3GalNAc epitope [11], uncapped LacNAc compunds [6], terminal blood group H type 2 antigen [6], P1 antigen (Galα1–4Galβ1–4GlcNAcβ1–3Galβ1–4Glc) [6], x₂ antigen (GalNAcβ1–3Galβ1–4GlcNAcβ1–3Galβ1–4Glc) [6], and the four N-linked carbohydrates described as ion 1595.7, 1851.7, 2245.8, and 2982.2 on matrix-assisted laser desorption/ionization time-of-flight mass spectrometry [10]. These antigens in our array set were listed in Table 2 along with reported biological relevance [6,8,10,11,29–39]. We excluded Tn antigens from this analysis because we already observed high signals for these antigens in Figure 1. The non-α-Gal antigens with the three highest signal and narrow distribution were T antigens [α-linked Galβ1–3GalNAc to serine or threonine; 15.1 (0.4) to 15.8 (0.2) in IgM, and 11.2 (0.6) to 11.8 (0.3) in IgG]. Our carbohydrate microarray revealed that the Sd^a antigen (CT/Sda-Sp-13) signal was 11.59 in IgM, and 9.05 in IgG, which were as high as alphaGal (11.2 in IgM and 8.5 in IgG). Globo-series glycolipids stage-specific embryonic antigens-3 (SSEA-3) and SSEA-4 had higher and broader distribution especially in IgM than Sd^a antigen [10.5 (2.1) to 12.7 (1.6), and 12.4 (2.1) to 13.5 (2.3), respectively]. Ganglio-series glycolipids, e.g., GM2 and GT2, also showed higher signals with broad distribution than Sd^a antigen. The P1 antigen and blood group H type 2 antigens were 13.2 (1.0) and 10.9 (0.7) to 12.7 (1.3) in IgM, respectively, but the signal intensity toward high-mannose carbohydrates were negative. The comparison of ganglio- (Table 3) [32,36–38,40–42] and globo-series (Table 4) [34,35,43–45] glycolipid signal intensity in pre-transplant monkeys suggested that certain carbohydrates in upstream of the cascades (GA1, GA2, GM2, GD3, or Gb3) showed higher signals than others (Figure 3). In the hierarchical clustering (Figure 2), GM2 was clustered with GD3 in IgM, and with GD3 and GD2 in IgG. Likewise, SSEA-3 was clustered with SSEA-4 and Gb3 was clustered with Gb4 in IgG.

Table 2.

Antibody signals to historically identified non-α-Gal carbohydrate antigens in pre-transplant NHPs

					IgM signal		IgG signal
a411 print #	Abbreviation	Description	Prototypic xenoantigen	Biological relevance	Median	SD	Median	SD
159	Ac-TF(Ser)-G - 24	Ac-(Galb1–3GalNAca)Ser-Gly-Hex-BSA	Galβ1–3GalNAc [11]	possible [29]	15.81	0.22	11.75	0.31
128	Ac-S-TF(Ser)-S-G - 28	Ac-Ser-(Galb1–3GalNAca)Ser-Ser-Gly-Hex-BSA	Galβ1–3GalNAc [11]	possible [29]	15.50	0.34	11.58	0.47
109	Ac-S-TF(Ser)-S-G - 16	Ac-Ser-(Galb1–3GalNAca)Ser-Ser-Gly-Hex-BSA	Galβ1–3GalNAc [11]	possible [29]	15.14	0.42	11.19	0.56
22	GA1di −11	Galb1–3GalNAcb – HSA	Galβ1–3GalNAc [11]	possible [30, 31]	14.63	0.35	10.88	0.91
312	GM2-Sp - 07	Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	strong [32]	14.09	1.65	10.18	1.32
274	6’Neu5Ac-LDN-Sp - 13	Neu5Aca2–6GalNAcb1–4GlcNAcb-Sp-BSA	Siaα2–6Gal/GalNAc [11]	weak [33]	13.70	0.70	9.55	1.17
96	GM2-Sp - 14	Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	strong [32]	13.61	1.65	9.40	0.99
241	SSEA-4-Sp - 12	Neu5Aca2–3Galb1–3GalNAcb1–3Gala1–4Galb1–4Glcb-Sp-BSA	Galβ1–3GalNAc [11]	strong [34]	13.50	2.32	9.52	1.73
57	GA1 – 20	Galb1–3GalNAcb1–4Galb1-BSA	Galβ1–3GalNAc [11]	possible [30, 31]	13.35	1.01	10.33	1.14
80	P1 – 09	Gala1–4Galb1–4GlcNAc-BSA	P1 [6]	strong [6]	13.23	1.02	9.08	1.07
118	GA1 – 06	Galb1–3GalNAcb1–4Galb1-BSA	Galβ1–3GalNAc [11]	possible [30, 31]	13.21	0.77	9.11	1.00
158	Ac-TF(Ser)-G - 04	Ac-(Galb1–3GalNAca)Ser-Gly-Hex-BSA	Galβ1–3GalNAc [11]	possible [29]	12.98	0.52	9.55	0.52
139	BG-H2– 12	Fuca1–2Galb1–4GlcNAcb1-linker-BSA	terminal H type2 [6]	strong [6]	12.68	1.25	8.82	0.95
305	Gb5/SSEA3 – 12	Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	Galβ1–3GalNAc [11]	possible [35]	12.65	1.61	8.89	1.10
288	GT2-Sp - 08	Neu5Aca2–8Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	strong [36]	12.47	1.05	9.01	1.48
246	BG-H2 – 16	Fuca1–2Galb1–4GlcNAcb-HSA	terminal H type2 [6]	strong [6]	12.44	1.29	8.07	0.79
231	SSEA-4-Sp - 05	Neu5Aca2–3Galb1–3GalNAcb1–3Gala1–4Galb1–4Glcb-Sp-BSA	Galβ1–3GalNAc [11]	strong [34]	12.41	2.14	8.75	1.35
123	Ac-S-TF(Ser)-S-G - 04	Ac-Ser-(Galb1–3GalNAca)Ser-Ser-Gly-Hex-BSA	Galβ1–3GalNAc [11]	possible [29]	12.35	1.01	8.21	0.83
82	GM1a - 29	Galb1–3GalNAcb1–4(Neu5Aca2–3)Galb1–4(Glc)HSA	Galβ1–3GalNAc [11]	strong [32]	12.24	1.01	9.05	0.57
336	GM2-Sp - 04	Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	strong [32]	12.14	1.59	8.37	1.03
330	GD2-Sp - 10	Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	possible [37]	12.09	1.70	10.26	1.69
280	GQ2-Sp - 06	Neu5Aca2–8Neu5Aca2–8Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	NA	12.06	0.80	8.69	1.41
247	CT/Sda-Sp - 13	Neu5Aca2–3[GalNAcb1–4]Galb1–4GlcNAcb-Sp-BSA	Sda epitope [8]	strong [8]	11.59	1.10	9.05	0.93
324	6’Neu5Ac-LDN-Sp - 05	Neu5Aca2–6GalNAcb1–4GlcNAcb-Sp-BSA	Siaα2–6Gal/GalNAc [11]	weak [33]	11.58	1.63	7.69	1.08
47	6’Neu5Ac-Lac	Neu5Aca2–6Galb1–4Glc-APD-HSA	Siaα2–6Gal/GalNAc [11]	NA	11.52	2.09	8.28	1.23
236	GD1a-Sp - 10	Neu5Aca2–3[Neu5Aca2–3Galb1–3GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	strong [38]	11.37	1.55	7.88	1.14
314	GD2-Sp - 04	Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	possible [37]	11.33	2.24	9.65	1.68
295	GT2-Sp - 03	Neu5Aca2–8Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	strong [36]	11.10	1.18	8.25	1.52
186	GTSSA-TF(Ser)-TGHATPLPVTD	BSA-PEG7-Gly-Thr-Ser-Ser-Ala-(Galb1–3GalNAca)Ser-Thr-Gly-His-Ala-Thr-Pro-Leu-Pro-Val-Thr-Asp	Galβ1–3GalNAc [11]	possible [29]	11.03	1.37	7.92	0.50
304	GQ2-Sp - 03	Neu5Aca2–8Neu5Aca2–8Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Sda epitope [8]	NA	11.02	1.37	7.85	1.54
138	BG-H2– 06	Fuca1–2Galb1–4GlcNAcb1-linker-BSA	terminal H type2 [6]	strong [6]	10.92	0.72	8.37	0.53
291	Gb5/SSEA3 – 04	Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	Galβ1–3GalNAc [11]	possible [35]	10.49	2.08	7.43	1.07
135	LNnT - 04	Galb1–4GlcNAcb1–3Galb1-BSA	uncapped LacNAc [6]	strong [6]	10.47	0.84	8.13	0.77
229	GD1a-Sp - 05	Neu5Aca2–3[Neu5Aca2–3Galb1–3GalNAcb1–4]Galb1–4Glcb-Sp-BSA	Galβ1–3GalNAc [11]	strong [32]	10.44	2.03	7.26	0.92
319	6’Neu5Ac-LacNAc-Sp - 11	Neu5Aca2–6Galb1–4GlcNAcb-Sp-BSA	Siaα2–6Gal/GalNAc [11]	NA	10.37	1.18	7.52	0.15
298	CT/Sda-Sp - 05	Neu5Aca2–3[GalNAcb1–4]Galb1–4GlcNAcb-Sp-BSA	Sda epitope [8]	strong [8]	10.31	1.33	7.68	0.50
162	NGA2 – 07	GlcNAcb1–2Mana1–6(GlcNAcb1–2Mana1–3)Manb1–4GlcNAc -BSA	ion 1851.7 [10]	strong [10]	10.14	0.60	7.47	0.24
187	GTSSAS-TF(Thr)-GHATPLPVTD	BSA-PEG7-Gly-Thr-Ser-Ser-Ala-Ser-(Galb1–3GalNAca)Thr-Gly-His-Ala-Thr-Pro-Leu-Pro-Val-Thr-Asp	Galβ1–3GalNAc [11]	possible [29]	10.05	0.89	7.99	0.39
261	6’Neu5Ac-LacNAc-Sp - 05	Neu5Aca2–6Galb1–4GlcNAcb-Sp-BSA	Siaα2–6Gal/GalNAc [11]	NA	9.64	0.86	7.38	0.19
311	LacNAc (dimeric)-Sp - 06	(Galb1–4GlcNAcb1–3)2b-Sp-BSA	uncapped LacNAc [6]	strong [6]	9.52	1.40	7.77	0.62
189	GTSSA-TF(Ser)-TF(Thr)-GHATPLPVTD	BSA-PEG7-Gly-Thr-Ser-Ser-Ala-(Galb1–3GalNAca)Ser-(Galb1–3GalNAca)Thr-Gly-His-Ala-Thr-Pro-Leu-Pro-Val-Thr-Asp	Galβ1–3GalNAc [11]	possible [29]	9.45	1.47	7.74	0.53
56	LSTc - 07	Neu5Aca2–6Galb1–3GlcNAcb1–3Galb1-BSA	Siaα2–6Gal/GalNAc [11]	possible [39]	9.41	1.23	7.30	0.15
160	NA2 – 08	Galb1–4GlcNAcb1–2Mana1–6[Galb1–4GlcNAcb1–2Mana1–3]Manb1–4GlcNAc -BSA	ion 2245.8 [10]	possible [10]	8.83	0.73	7.32	0.17
318	GD1b - 05	Neu5Aca2–8Siaa2–3(Galb1–3GalNAcb1–4)Galb1–4-BSA	Galβ1–3GalNAc [11]	possible [37]	8.27	1.41	7.80	0.88
188	GTSSASTGHA-TF(Thr)-PLPVTD	BSA-PEG7-Gly-Thr-Ser-Ser-Ala-Ser-Thr-Gly-His-Ala-(Galb1–3GalNAca)Thr-Pro-Leu-Pro-Val-Thr-Asp	Galβ1–3GalNAc [11]	possible [29]	7.71	0.81	7.23	0.13
230	6’Neu5Gc-LacNAc-Sp - 05	Neu5Gca2–6Galb1–4GlcNAcb-Sp-BSA	Siaα2–6Gal/GalNAc [11]	NA	7.54	0.42	7.23	0.00
86	Man5 – 05	Mana1–6(Mana1–3)Mana1–6(Mana1–3)Manb1–4GlcNAc-BSA	ion 1595.7 [10]	possible [10]	7.23	0.24	7.23	0.00

Biological relevance is stated high when it is reported that the antigen is expressed on wild type or gene modified porcine cells/tissues. Biological relevance is stated possible when it is reported that the antigen is expressed on mammalian cells/tissues. Biological relevance is stated weak when it is reported that the antigen is not expressed on porcine cells/tissues. Numbers in brackets show reference. NA data not available, NHP nonhuman primate

Table 3.

Antibody signals to ganglio-series glycolipids in pre-transplant NHPs

				IgM signal		IgG signal
a411 print #	Abbreviation	Description	Biological relevance	Median	SD	Median	SD
124	GA2di - 37	GalNAcb1–4Galb - BSA	possible [40]	15.89	0.47	12.23	1.99
170	GA2di - 28	GalNAcb1–4Galb - BSA	possible [40]	15.87	0.50	11.97	2.02
22	GA1di - 11	Galb1–3GalNAcb – HSA	possible [41]	14.63	0.34	10.88	0.91
116	GA2di - 05	GalNAcb1–4Galb - BSA	possible [40]	14.19	0.97	11.37	1.69
312	GM2-Sp - 07	Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	strong [32]	14.09	1.65	10.18	1.32
301	GD3-Sp - 08	Neu5Aca2–8Neu5Aca2–3Galb1–4Glcb-Sp-BSA	strong [32]	14.07	1.84	10.09	1.44
96	GM2-Sp - 14	Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	strong [32]	13.61	1.65	9.40	0.99
271	GD3-Sp - 04	Neu5Aca2–8Neu5Aca2–3Galb1–4Glcb-Sp-BSA	strong [32]	13.36	1.21	9.02	0.86
57	GA1 – 20	Galb1–3GalNAcb1–4Galb1-BSA	possible [41]	13.35	1.01	10.33	1.14
118	GA1 – 06	Galb1–3GalNAcb1–4Galb1-BSA	possible [41]	13.21	0.77	9.11	1.00
282	GM3-Sp - 11	Neu5Aca2–3Galb1–4Glcb-Sp-BSA	strong [32]	12.65	0.51	9.02	0.96
288	GT2-Sp - 08	Neu5Aca2–8Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	strong [36]	12.47	1.05	9.01	1.48
82	GM1a - 29	Galb1–3GalNAcb1–4(Neu5Aca2–3)Galb1–4(Glc)HSA	strong [32]	12.24	1.01	9.05	0.57
336	GM2-Sp - 04	Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	strong [32]	12.14	1.59	8.37	1.03
46	GM3 – 12	Neu5Aca2–3Galb1–4Glc-APD-HSA	strong [32]	12.12	2.13	8.91	1.37
330	GD2-Sp - 10	Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	possible [37, 42]	12.09	1.70	10.26	1.69
250	GM3(Gc)-Sp - 14	Neu5Gca2–3Galb1–4Glcb-Sp-BSA	possible [32, 42]	12.03	2.08	7.92	0.76
325	GT3-Sp - 07	Neu5Aca2–8Neu5Aca2–8Neu5Aca2–3Galb1–4Glcb-Sp-BSA	possible [42]	11.57	1.51	8.12	0.69
236	GD1a-Sp - 10	Neu5Aca2–3[Neu5Aca2–3Galb1–3GalNAcb1–4]Galb1–4Glcb-Sp-BSA	strong [38]	11.37	1.55	7.88	1.14
314	GD2-Sp - 04	Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	possible [37, 42]	11.33	2.24	9.65	1.68
285	GM3-Sp - 04	Neu5Aca2–3Galb1–4Glcb-Sp-BSA	strong [32]	11.17	1.13	8.15	0.84
257	GT3-Sp - 03	Neu5Aca2–8Neu5Aca2–8Neu5Aca2–3Galb1–4Glcb-Sp-BSA	possible [42]	11.10	1.99	7.58	0.81
295	GT2-Sp - 03	Neu5Aca2–8Neu5Aca2–8Neu5Aca2–3[GalNAcb1–4]Galb1–4Glcb-Sp-BSA	strong [36]	11.10	1.18	8.25	1.52
229	GD1a-Sp - 05	Neu5Aca2–3[Neu5Aca2–3Galb1–3GalNAcb1–4]Galb1–4Glcb-Sp-BSA	strong [32]	10.44	2.03	7.26	0.92
322	GM3(Gc)-Sp - 05	Neu5Gca2–3Galb1–4Glcb-Sp-BSA	possible [32, 42]	9.78	1.04	7.23	0.21
318	GD1b - 05	Neu5Aca2–8Siaa2–3(Galb1–3GalNAcb1–4)Galb1–4-BSA	possible [37]	8.27	1.41	7.80	0.88

Biological relevance is stated high when it is reported that the antigen is expressed on wild type or gene modified porcine cells/tissues. Biological relevance is stated possible when it is reported that the antigen is expressed on mammalian cells/tissues. Numbers in brackets show reference. NA data not available, NHP nonhuman primate

Table 4.

Antibody signals to globo-series glycolipids in pre-transplant NHPs

				IgM signal		IgG signal
a411 print #	Abbreviation	Description	Biological relevance	Median	SD	Median	SD
108	Forssman Di - 21	GalNAca1–3GalNAcb1-BSA	weak [43]	16.32	0.31	12.40	1.63
244	Forssman Tetra-BSA - 05	GalNAca1–3GalNAcb1–3Gala1–4Galb-BSA	weak [43]	15.96	0.44	12.01	0.85
125	Forssman Di - 31	GalNAca1–3GalNAcb1-BSA	weak [43]	15.87	0.52	12.29	1.61
119	Forssman Di - 04	GalNAca1–3GalNAcb1-BSA	weak [43]	15.85	0.56	12.00	1.27
289	Forssman Tetra-BSA - 13	GalNAca1–3GalNAcb1–3Gala1–4Galb-BSA	weak [43]	15.60	0.43	12.00	0.76
42	Gb3– 13	Gala1–4Galb1–4Glc-HSA	strong [44]	14.60	0.67	10.40	1.48
254	Gb4 tetra (P1 tetra)-Sp - 15	GalNAcb1–3Gala1–4Galb1–4GlcNAcb-Sp-BSA	NA	14.44	1.55	10.89	1.12
87	Gb4 tetra (P1 tetra)-Sp - 06	GalNAcb1–3Gala1–4Galb1–4GlcNAcb-Sp-BSA	NA	13.69	1.84	10.28	1.13
294	Globo A - 09	GalNAca1–3(Fuca1–2)Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	possible [45]	13.64	1.92	9.33	1.36
241	SSEA-4-Sp - 12	Neu5Aca2–3Galb1–3GalNAcb1–3Gala1–4Galb1–4Glcb-Sp-BSA	strong [34]	13.50	2.32	9.52	1.73
305	Gb5/SSEA3 – 12	Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	possible [35]	12.65	1.61	8.89	1.10
231	SSEA-4-Sp - 05	Neu5Aca2–3Galb1–3GalNAcb1–3Gala1–4Galb1–4Glcb-Sp-BSA	strong [34]	12.41	2.14	8.75	1.35
110	Globo H - 03	Fuca1–2Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	possible [45]	10.56	1.00	7.94	0.57
112	Globo B - 12	Gala1–3(Fuca1–2)Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	possible [45]	10.52	1.67	7.95	0.65
291	Gb5/SSEA3 – 04	Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	possible [35]	10.49	2.08	7.43	1.07
94	Globo A - 03	GalNAca1–3(Fuca1–2)Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	possible [45]	10.39	2.09	7.32	0.88
81	Gb4 – 09	GalNAcb1–3Gala1–4Galb1-BSA	strong [44]	10.34	1.25	8.02	0.45
293	Globo H - 10	Fuca1–2Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	possible [45]	10.31	2.11	7.49	1.26
106	Globo B - 05	Gala1–3(Fuca1–2)Galb1–3GalNAcb1–3Gala1–4Galb1-BSA	possible [45]	9.90	0.66	7.52	0.26

Biological relevance is stated high when it is reported that the antigen is expressed on wild type or gene modified porcine cells/tissues. Biological relevance is stated possible when it is reported that the antigen is expressed on mammalian cells/tissues. Biological relevance is stated weak when it is reported that the antigen is not expressed on porcine cells/tissues. Numbers in brackets show reference. NA data not available, NHP nonhuman primate

FIGURE 3.

Biosynthetic pathways of ganglio-series and globo-series glycolipids. Values listed below carbohydrates show IgM and IgG signals for approximate density of 10 per BSA/HSA. A4GALT1 alpha 1,4-galactosyltransferase, B3GALNT1 beta-1,3-N-acetylgalactosaminyltransferase 1, B3GALT5 beta-1,3-galactosyltransferase 5, B3GALT4 beta-1,3-galactosyltransferase 4, B4GALNT1 beta-1,4-N-acetylgalactosaminyltransferase 1, FUT2 fucosyltransferase 2, GBGT1 globoside alpha-1,3-N-acetylgalactosaminyltransferase 1, ST3GAL2 ST3 beta-galactoside alpha-2,3-sialyltransferase 2, ST3GAL5 ST3 beta-galactoside alpha-2,3-sialyltransferase 5, ST8SIA1 ST8 alpha-N-acetyl-neuraminide alpha-2,8-sialyltransferase 1, ST8SIA3 ST8 alpha-N-acetyl-neuraminide alpha-2,8-sialyltransferase 3

Differences in antibody signals based on blood type

Although a previous report showed that A, B, and O (H) blood type antigens are not expressed on the surface of red blood cells in NHPs [46], the hierarchical clustering in our study revealed the presence of clusters of anti-A antibodies (Figure 2). The analysis also distinguished between the blood group B NHPs with one blood group O NHP (13GP08) and the blood group AB NHPs with the other two blood group O NHPs. The former group showed significantly higher signals for 20 and 8 out of 26 blood group A antigens in IgM and IgG repertoires than the latter group, respectively. Analyses to elucidate association between each blood type and anti-carbohydrate antigen signals found 6 and 5 carbohydrates that reached significance at level of P < 0.01 in IgM and IgG repertoires, respectively (Figure S1), which we considered unremarkable from statistical view point.

DISCUSSION

Natural antibody directed toward α-Gal and non-α-Gal epitopes in NHPs are implicated in the pathobiology of xenograft rejection. The generation of multiple-gene knockout pigs that lack α-Gal or non-α-Gal epitopes have reduced the impact of antibody binding, but is insufficient in eliminating rejection in immunosuppressed recipients [47–49]. The goal of this study was to identify the capacity and specificity of antibodies within naïve NHP serum using a carbohydrate microarray system [13–15]. Using the highly sensitive methodology we were able to confirm the reactivity of common carbohydrate xenoantigens while elaborating on novel non-α-Gal epitopes.

Byrne et al. identified B4GALNT2 gene and Sd^a antigen as a non-α-Gal antigen by using a GGTA1-KO and CD46 transgenic pig expression library [8]. The Sd^a antigen was reported negative on red blood cells or peripheral blood mononuclear cells of human and NHPs [50]. The addition of the B4GALNT2 mutation to the GGTA1/CMAH-KO background significantly reduced the level of human serum immunity to pig cells [9]. Byrne et al. further discussed GM2 glycolipid as a xenoantigen because it shares the Sd^a epitope with Sd^a antigen [7]. We observed higher signal intensity against GM2 glycolipids as well as GM3 and GD3 glycolipids than Sd^a antigen in our microarray system using NHP serum (Table 3). The expression of these glycolipids has been reported on porcine heart valve cusps [32] and all mammalian tissue [51,52]. Antibodies against these antigens have been identified in several human diseases such as neurological disorders and cancers [53,54]. But a blockade of GM3 production with the mutations in ST3GAL5 transferase has been known for neuromotor retardation both in mice and humans [55], and the utility of mutating ST5GAL5 in xenotransplant donor pigs would not be feasible.

The SSEA-3 and SSEA-4 also exhibited higher signal and wide distributions than Sd^a antigen in our dataset (Table 2). The glycolipid molecules SSEA-3 and SSEA-4 are known as embryonic stem cell markers that are identified in stem cells and cancer cells [56,57]. An expression of SSEA-4 was reported on porcine cardiac valves [34]. Zulli et al. reported that diseased human and rabbit blood vessels expressed for SSEA-4, and both SSEA-3 and SSEA-4, respectively, although normal blood vessels were negative for both SSEA-3 and SSEA-4 [35,58]. A SSEA-4 expression, but not SSEA-3 expression, was reported on monocytes and granulocytes of cynomolgus macaques [59]. It is unclear at this time how SSEA-3 or SSEA-4 is related to naïve NHP but further study is necessary.

The present study revealed higher signal intensity against some of the precursor carbohydrate antigens in the upstream of cascades compared to the carbohydrate antigens in downstream of cascades (Figure 3). It is suggested that genetic modification could interfere with normal carbohydrate modification cascade and result in increased synthesis of incomplete carbohydrate components and in accumulation of precursor or abnormal carbohydrate components [10,60]. Estrada et al. showed that NHP serum acquired higher humoral immunity against GGTA1/CMAH-KO pig cells than GGTA1-KO pig cells [61]. Miyagawa et al. compared carbohydrate profiles between domestic and GGTA1-KO pig cells using a lectin microarray, and revealed modest signal increase in Galβ1–3GalNAc (T antigen) component in GGTA1-KO pigs [11]. The T antigens are also known as markers of aggressiveness of malignant neoplasms, which is caused by aberrant or incomplete glycosylation in cancer cells [62]. Our data and these previous studies have led us to consider that not only the blockade of specific pathways but also their specific downstream modification of carbohydrates may be helpful in eliminating the antigenicity of the xenograft.

As the hierarchical clustering in our study suggested, the fine specificity against carbohydrate antigen is determined not only by terminal epitopes but also by core structures. Milland et al. reported that α-Gal-related natural antibodies recognize the Lac form core structure rather than the LacNAc form core structure [25]. We observed similar results in our study; Galilli (or iGb3), which has a core chain of Lac form, had higher signals compared to alphaGal, which has a core chain of LacNAc form although the densities of the two antigens were different (Table 1). But this configuration appears analogous to the relationship between Sd^a antigen and GM2 glycolipids with similar densities; Sd^a related natural antibody recognize the Lac form (GM2) rather than LacNAc form (Sd^a antigen) both in IgM and IgG repertoires (Table 2).

The highest and most consistent signal in our IgG repertoire was to the glycopeptide Ac-Ser-(GlcNAcα)Ser-Ser-Gly, which has also been observed in the previous microarray study on human natural antibodies [15]. This α-linked O-GlcNAc has been found in prokaryotes but not in human, NHPs, or pigs, and are less likely to be a target in pig-to-NHP xenograft rejection [63]. We also observed high IgM and IgG signals for sialylated LacdiNAc (LDN, Table 2), but LDN is also a rare structure in mammalian glycoproteins [33]. Forssman antigen is another target of rejection in humans and NHPs, but the accumulated evidence suggests that the pig is Forssman-negative and that this carbohydrate is not playing a role in the rejection of pig grafts in NHPs [43]. The iGb3 glycolipid is less likely to be a target of rejection, since iGb3 was not shown to be expressed in sufficient amount in GGTA1-KO fetal pig fibroblasts and mini-pig kidney cells to mediate cell destruction [26,44]. The P1 antigen is synthesized by A4GALT, which is a direct competitor of the GGTA1, suggesting that A4GALT may substitute for GGTA1 in the GGTA1-KO pigs [6]. However, the P1 antigen may not be responsible for rejection because a lack of induced antibody response to P1 antigen by baboon serum was reported after exposure to GGTA1-KO pigs [43].

Although cynomolgus macaques share the A and B blood group antigens with humans, it was believed that these antigens were found in bodily fluids and only weakly expressed on the surface of red blood cells [46]. Our analysis insisted that cynomolgus serum was sensitized for specific blood group antigens, however, our data could not predict the sensitization to group A/B antigens in the blood group O NHPs. These should be tested in adequately powered future studies.

We are aware that our research has some limitations. First, our array is a model system that mimics some key features of natural presentation, and it is not identical to the natural system. Although serum antibodies have fairly high selectivity [64] and the array provides a tool to detect different antibody populations present in serum, positive binding does not mean that it will always bind those glycans as presented on a cell surface. Furthermore, the scope of our current study did not include a comparison to sensitized sera after xenotransplantation and therefore we could not confirm or discover new antigens that lead to rejection. However, the values are normalized and can directly be compared to values from future studies using sensitized sera. Second, the signal intensity of our model can be affected not only by antibody concentration but also antigen affinity/avidity. The differences in density and linker in our assay can affect recognition. Third, our array did not cover all carbohydrate antigens and we may miss relevant antibodies that bind carbohydrates that were not present. These limitations highlight evidence of the difficulty of collecting data on natural system, which may be true to any models.

In conclusion, our analysis suggests that natural antibodies are directed towards a series of α-Gal and non-α-Gal antigens including Tn antigen, T antigen, and ganglio- and globo series glycolipids. The comprehensive library of our data has a potential to allow us to make further comparative analysis with previous and future studies. This high throughput discovery tool also identifies numerous antigens to which there are natural antibodies among a small cohort of Mauritius cynomolgus monkeys. While these results need to be confirmed among other NHP, preclinical studies using Mauritius cynomolgus monkeys would benefit from screening for antibody reactivity to minimize innate immune complications.

Supplementary Material

Supp FigS1

FIGURE S1. Blood type covariate association for carbohydrate antigens in IgM (A) and IgG (B) that reached significance at level of P < 0.01. Values are expressed as median and standard deviation.