Animal studies were performed in α1,3-galactosyltransferase knockout (GalT-KO) mice. All animal procedures were supervised and approved by Bellvitge Biomedical Research Institute (IDIBELL) ethics committee for animal experimentation and the Catalonia Government (DMA 3225). The care and handling of the animals conformed to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication 85-23, revised 1996) and the European Agreement of Vertebrate Animal Protection for Experimental Use (86/609). Mice were maintained in the AAALAC-certified animal facility of IDIBELL under controlled temperature (21 ± 1°C), humidity (55 ± 5%), and cycles of light/dark (12/12 h), and food and water were given ad libitum. We used 16-week-old GalT-KO mice for all procedures, and each experimental group contained an equal number of male and female mice. Mouse blood collection was performed without anesthesia by submandibular bleeding. Serum was collected by mild centrifugation and stored at -20°C until further analysis.

The euthanasia procedure was established following the European Directive on protecting animals used for scientific purposes (2010/63/EU) and was performed in a CO₂ chamber. Animal death was never considered an endpoint criterion. Instead, this was determined through a protocol of animal observation with the corresponding corrective measures (Supplementary Table S1).

GAS914 (Novartis Pharma AG, Basel, Switzerland) is a poly-L-lysine backbone with an average degree of polymerization of 1,000 L-lysines and with 23-28% of lysines derivatized with αGal (27).

The quantification of anti-αGal antibodies by ELISA followed the general protocol previously described (15).

Glycochips (microchips printed with 577 different carbohydrate structures) were printed by Semiotik LLC (Moscow, Russia) using a collection of amine-functionalized glycans and bacterial polysaccharides covalently coupled to N-hydroxysuccinimide-derivatized glass slides (slide H, Schott-Nexterion, Mainz, Germany). The step-by-step protocol was deposited in Protocol Exchange (28). All data analysis was performed with the ScanArray^® Express Microarray Analysis System (PerkinElmer, Waltham, MA, USA). The binding results were expressed in relative fluorescence units (RFU) as median ± median absolute deviation (MAD). Interactive exploration of multidimensional data (heat mapping and clustering analysis) was performed with the Hierarchical Clustering Explorer application developed by the University of Maryland, MD, USA (http://www.cs.umd.edu/hcil/hce).

The Cecal ligation and puncture (CLP) procedure was performed as described elsewhere (29). GalT-KO mice were deeply anesthetized with 4% isoflurane (Esteve Veterinaria, Barcelona, Spain), and during the procedure, anesthesia was maintained at 2.5% isoflurane. Under sterile conditions, a 1-2 cm midline incision was made, and the cecum was exteriorized and ligated (4-0 Safil^® Violet, B. Braun, Melsungen, Hessen, Germany) distally to the ileocecal valve. To generate moderate-grade sepsis, defined as ~ 15 - 50% mortality during the acute phase of sepsis, 25% of the cecum (~ 0.5 cm) was ligated and punctured twice with a 30-gauge needle. The abdominal wall incision was closed, and recovery was facilitated by keeping the animal on a thermal blanket. Buprenorphine (0.05 mg/kg) and meloxicam (1 mg/kg) were administered subcutaneously (sc) as analgesics before and after the CLP procedure. The animals were returned to their cages one hour after surgery, where water and food were provided ad libitum and monitored twice a day for 15 days, to asses both early and chronic sepsis mortality (30). Animal recovery was also favored by sc of 1 mL dextrose solution administration for up to 7 days. Access to water and food was facilitated two hours after surgery when animals were placed in their corresponding cages. Body weight, mobility, food intake, cutaneous features, and respiratory frequency of animals were monitored twice a day over 15 days, providing a score that led to different corrective measures depending on animal status (Table S1) (31).

GalT-KO mice treated with 10 mg/kg of GAS914 or PBS sc on days -1, -3, and -5 before CLP (day 0) and scores ≥ 9 were euthanized at 12 h after the procedure (Table S1). Animals with scores < 9 were euthanized at 24 h after the CLP. To characterize bacteremia in peripheral blood, 100 µL of whole blood was seeded on a Petri dish with Blood Agar (Sigma-Aldrich, St. Louis, MO, USA) and incubated at 37°C for 24 h. Subsequently, bacteria were identified by a combination of different methods: Gram staining, plate culture with McConkey Agar differential medium, and biochemical tests (catalase and coagulase).

The DNA of the E. coli isolated from blood cultures obtained from GalT-KO mice after the CLP procedure was stratified by pulsed-field gel electrophoresis (PFGE). Ten colony-forming units or more were randomly selected from blood agar plates, resuspended in 5 ml of tryptic soy broth (Becton Dickinson and Company, Franklin Lakes, NJ, USA), and incubated at 37°C overnight. Total DNA was extracted following standard procedures (32). DNA restriction of each isolate was carried out with 40 units of XbaI (New England BioLabs Inc., Beverly, MA, USA) in 40 µL of 1xNEB2 buffer for 6 h at 37°C. The chromosomal DNA fragments were separated in a 1% agarose gel (CertifiedTM Megabase Agarose, Bio-Rad, Hercules, California, USA) in 0.5X TBE mM separated from the values: TBE (890 mM Tris; 890 mM boric acid; 20 mM EDTA, pH8). The pulsed-field gel electrophoresis (PFGE) was carried out in a CHEF-DRIII apparatus (Bio-Rad, Hercules, California, USA) with 0.5X TBE as running buffer. The Lambda Ladder PFGE Marker (New England Biolabs Inc., Beverly, MA, USA) was used as molecular weight marker. The running conditions were: 6V/cm with initial pulses of 1 second, which increased until 30 seconds, for a total run of 18 hours. PFGE patterns were visually compared following the criteria previously described (33). The dendrogram showing the clustering of strains was generated from the analysis of PFGE profiles with the FINGERPRINTING II software, with 1% optimization and band position tolerance. The cut-off value to define the PFGE patterns was set at 80%.

Two different strains of Escherichia coli were used in the studies: the human pathogenic E. coli O86:B7, obtained from the American Type Culture Collection (ATCC^® 12701™, LGC Standards, Teddington, Middlesex, UK), and one representative strain isolated from GalT-KO mouse blood after CLP (E. coli A3). In addition, Pseudomonas aeruginosa (strain 21565) and Klebsiella pneumoniae (strain 35204) were isolated from sputum and human blood, respectively. Both strains belonged to the Microbiology Department of Bellvitge University Hospital collection, were characterized as multidrug-resistant bacteria and are representative of bacteria causing often infectious diseases in the hospital. The strains were cultured for 16 h in Nutrient Broth (NB) medium (Becton Dickinson, and Company, Franklin Lakes, NJ, USA) at 37°C and 200 rpm (New Brunswick Scientific, Edison, NJ, USA). In the case of solid cultures, NB was supplemented with 1.5 (w/v) agar and incubated under similar conditions. The overnight cultures were diluted (1:100) with fresh NB medium for strain conservation. Cells were grown under the same conditions until the exponential phase (between 2-3 hours), where cellular suspensions were supplemented with 15% (v/v) glycerol (Sigma-Aldrich, St. Louis, MO, USA) and stored at -80°C for up to 2 years.

Bacterial strains were incubated overnight in NB medium on an orbital shaker (New Brunswick Scientific, Edison, NJ, USA) at 200 rpm and 37°C. The bacterial suspension was diluted 100 times using a fresh medium the next day. Sterile baby rabbit complement (AbD Serotec, Kidlington, Oxfordshire, UK) was added to the bacterial suspension (2.5%, v/v). This mix was then incubated with the heat-inactivated (0.5 h at 56°C) mouse or human serum samples (5%, v/v) for 2-4 h at 200 rpm and 37°C. Cultures on 1.5% (w/v) agar NB plates were plated hourly (40 µL) at adequate dilutions (starting from 1:3125). After 18 h of incubation at 37°C, the resulting bacterial colonies were counted. The broth alone was used as the control for bacterial growth, and complement without serum was used for lysis via the alternative complement pathway (34). We calculated the bactericidal activity, comparing the number of bacteria in reaction mixtures containing the tested serum with the control.

Phagocytosis assay was performed based on a protocol described previously (35). Briefly, RAW 264.7 mouse macrophages were incubated in Dulbecco’s Modified Eagle Medium (DMEM) with a 5% Fetal Bovine Serum (FBS) at 37°C, 5% CO₂, and 4.2 x 10⁵ cells were placed in 6-well cell culture plates. Meanwhile, E. coli O86:B7 was incubated overnight in NB medium on an orbital shaker (New Brunswick Scientific, Edison, NJ, USA) at 200 rpm and 37°C.

The next day, the bacterial suspension was diluted 100 times using a fresh medium and incubated in the same conditions until the exponential phase. Later, bacterial suspension was centrifuged (4,000g, 10 min, 22°C), and cell concentration was adjusted to 5 x 10⁸ CFU/mL. Then, bacteria were stained with 20µM of the SYTO9 Green marker (Thermo Fisher Scientific, Waltham, MA, USA), and 10⁷ stained bacteria were pre-opsonized with heat-inactivated human serum at 10% (v/v) in PBS, previously treated overnight at 4°C with GAS914 (100 μg/ml) by a 20 min incubation at RT in orbital shaking). Meanwhile, macrophages were incubated (20 min, in the dark, without agitation, at room temperature) with 2.5 μg/mL of APC-anti-mouse-F4/80 (Thermo Fisher Scientific, Waltham, MA, USA).

After a washing step, stained bacteria and macrophages were incubated (multiplicity of infection (MOI) of 10) at 37°C, 5% CO_2, for 30 minutes. After a washing step, the samples were fixed in 400 µL of paraformaldehyde (4% v/v), and cytofluorometric determinations were performed on a Gallios flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). Every single FACS determination recorded about 20,000 total events. The events characterized by normal forward scatter (FSC) and side scatter (SSC) parameters were included in subsequent analyses. Fluorescence was collected through the corresponding bandpass filters for each indicated surface cell marker. Data were analyzed using KALUZA software (Beckman Coulter, CA, USA).

E. coli strains were incubated overnight in NB medium at 37°C with shaking at 200 rpm. The next day, the bacterial suspensions were diluted 100 times using a fresh medium and grown to an optical density of 0.3 units at 600 nm. Cells were harvested at 4,000g for 10 min at 4°C (Sorvall, Thermo Fisher Scientific, Waltham, MA, USA) and washed twice with an equal volume of Hank’s Balanced Salt Solution (HBSS, Life Technologies, Carlsbad, CA, USA). Cells were resuspended in an equal volume of HBSS supplemented with 1.3 mM calcium and 0.8 mM magnesium, divided (150 µL) into 96-well plates (conic bottom), and harvested at 2,500g for 10 min at 4°C (Eppendorf, Hamburg, Germany). Supernatants were carefully removed by aspiration. Heat-inactivated serum was added to the bacterial suspension at 10% (v/v) in supplemented HBSS. The cells were homogenized, and, in the case of complement deposition assays, 10% (v/v) standard mouse or human complement (Sigma-Aldrich, St. Louis, MO, USA) was added to the cell suspension. Plates were incubated for 0.5 h at 25°C and centrifuged under the same conditions after adding 100 µL of HBSS for washing. Supernatants were discarded by aspiration. Cells were then incubated in darkness (0.5 h at 4°C) with the secondary antibody labeled with the corresponding fluorochrome: anti-mouse C3 and C4 (Cedarlane, Burlington, ON, Canada), anti-human C3 and C4 (MP Biomedicals, Santa Ana, CA, USA), anti-mouse IgM, IgG (Invitrogen, Carlsbad, CA, USA), and IgG subclasses (Abcam, Milton, Cambridge, UK), and anti-human IgM, IgG, IgA (Invitrogen, Carlsbad, CA, USA), and IgG subclasses (Sigma-Aldrich, St. Louis, MO, USA), following the manufacturer’s instructions. After adding HBSS (100 µL) for washing, plates were centrifuged under the same conditions, and supernatants were discarded by aspiration. Finally, the cells were fixed (1 h at 4°C) with 4% (w/v) paraformaldehyde in HBSS.

Cytofluorometric determinations were performed on a Gallios flow cytometer (Becton Dickinson and Company, Franklin Lakes,NJ, USA). Every single FACS determination recorded about 20, 000 total events. The events characterized by normal forward scatter (FSC) and side scatter (SSC) parameters were included in subsequent analyses. Fluorescence was collected through the corresponding bandpass filters for each indicated surface cell marker. Data were analyzed using KALUZA software (Beckman Coulter, CA, USA). Cells incubated with supplemented HBSS alone were used as the negative control, and cells incubated with supplemented HBSS and secondary antibodies were considered the experiment’s background.

Bacterial strains were grown in NB medium for 16 h at 37°C (200 rpm). Cultures were adjusted to an optical density of 0.5 units at 600 nm with fresh medium, the cells were pelleted by centrifugation (10,600g for 10 min, Eppendorf, Hamburg, Germany), and the supernatants were discarded. Samples were prepared from whole-cell lysates treated with 0.5 mg/mL proteinase K (Sigma-Aldrich, St. Louis, MO, USA). Lipopolysaccharide (LPS) was extracted using the hot aqueous-phenol method previously described (36). Subsequently, the extract was separated by SDS-PAGE (13%) and directly stained using a standard silver stain protocol (Bio-Rad, Hercules, CA, USA). LPS from E. coli O111:B4 was used as a standard for long-chain O-antigen (Sigma-Aldrich, St. Louis, MO, USA).

Bacterial strains were incubated for 16 h in NB medium on a shaker at 200 rpm and 37°C. The next day, the bacterial suspensions were diluted 100 times using a fresh medium and grown to an optical density of 0.3 units at 600 nm. Cells were harvested (5,000g, 10 min at 4°C) and washed twice with an equal volume of HBSS. Next, cells were resuspended in an equal volume of HBSS, divided (150 µL) into 96-well plates (conical bottom), and centrifuged at 2,500g for 10 min at 4°C. Supernatants were carefully removed by aspiration. Cells were then incubated for 0.5 h at 4°C (in darkness) with 100 µL of lectin IB4-FITC in HBSS (1:100, EY Laboratories, San Mateo, CA, USA) for α-galactosyl expression (37). For specific αGal expression, anti-αGal IgM (Gal-13) monoclonal antibody (titer ~1:1000) and anti-αGal IgG (M86) monoclonal antibody (titer ~1:80) (kindly facilitated by Dr. Uri Galili) were used. After adding 100 µL of HBSS for washing, the plates were centrifuged again under the same conditions, and supernatants were removed by aspiration. Finally, the cells were fixed (1 h at 4°C) with 4% (w/v) paraformaldehyde in HBSS. Cytofluorometric determinations were performed on a Gallios flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). Every single FACS determination recorded about 20, 000 total events. The events characterized by normal forward scatter (FSC) and side scatter (SSC) parameters were included in subsequent analyses. Fluorescence was collected through the corresponding bandpass filters for each indicated surface cell marker. Data were analyzed using KALUZA software (Beckman Coulter, CA, USA). Cells incubated with HBSS alone were used as the negative control for the experiment.

All experiments were performed at least three times unless otherwise stated. All data were analyzed using GraphPad Prism statistics software (GraphPad Software Inc., San Diego, CA, USA). First, the Gaussian data distribution was checked by the D’Agostino-Pearson omnibus normality test (alpha = 0.05), and the homogeneity of variances was determined by the F test (alpha = 0.05). Statistical analyses were performed using paired or unpaired parametric t-tests. The Wilcoxon matched-pairs signed-rank and Mann-Whitney test (unpaired data analysis) were used as non-parametric tests when data did not follow a Gaussian distribution. Grubbs’ test (extreme studentized deviation) was used to determine significant outliers in the data (alpha = 0.05). The results are expressed as the mean ± standard deviation (SD) and, in the case of glycan arrays, as the median ± median absolute deviation (MAD). Survival analysis was done with the Mantel-Cox log-rank test. Differences were considered statistically significant when p < 0.05.

GalT-KO mice reproduce the natural concentration of anti-αGal antibodies observed in humans (38). All the mice in these studies exhibited high levels of anti-αGal antibodies without external immunization at 16 weeks (Figure 1A). The concentration of anti-αGal IgM and IgG were similar, and IgG3 was the predominant anti-αGal IgG subclass in most of the mouse serum samples (Figure 1A). No levels of anti-αGal IgA were detected in any animal. In addition, no anti-αGal antibodies were detected in wild-type (CBA) mice.

Mice were injected subcutaneously (sc) with 1, 2.5, 5, and 10 mg/kg of GAS914 on days 0, 2, and 4 (for three total doses). The best results were obtained with 10 mg/kg GAS914 (Figure 1B), which resulted in a decrease of approximately 90% in the level of circulating anti-αGal antibodies at the end of the challenge. With this dose, on day 11, one week after the last administration of GAS914, the levels of anti-αGal IgM and IgG antibodies were still 50% lower than those before the treatment (Figure 1B). The decrease in circulating anti-αGal antibodies was readily achieved with the first dose and changed very little with the two additional injections. Moreover, GAS914 significantly reduced both IgM and IgG, including the circulating anti-αGal IgG1 and IgG3 subclasses in GalT-KO mice (Figure 1C). No change occurred in IgG2a and IgG2b. Additionally, we did not observe any other side effects in the animals induced by the GAS914 treatment (Table S1).

The impact of the treatment with GAS914 on the specificity of natural mouse anti-glycan antibodies was assessed in four GalT-KO mice using microchips printed with 577 different carbohydrate structures and bacterial antigens. The glycochip analysis showed that GAS914 treatment significantly diminished antibody reactivity in all the mice against αGal trisaccharide (#222), αGal tetrasaccharide (#373), and αGal pentasaccharide (#481) structures. The treatment also significantly decreased antibody binding in some animals to other αGal and α-galactosyl residues (Figure 1D).

The effect of removing anti-αGal antibodies in vivo was studied in GalT-KO mice submitted to CLP, which initially produces polymicrobial sepsis (2-12 h), shifting to predominant Gram-negative sepsis with coliform bacteria in the bloodstream at 24 h that causes, in most cases, animal death (39). Firstly, GalT-KO mice were pre-treated with GAS914 or PBS on days 0 and 2, after which CLP was produced on day 3. Then, animals were euthanized 12 or 24 h after CLP, depending on the welfare status (Table S1), to assess the presence of bacteria in the blood. Four animals of the PBS group were euthanized 12 h after CLP, and two of the PBS and 8 of GAS914 at 24 h. E. coli was isolated in 7 out of the 8 mice treated with GAS914 and 4 out of the 8 that received PBS. Two animals that received PBS died within 12 h after CLP, and Enterococcus faecium was the only one isolated from a third animal. We did not detect any bacteria in the blood of two mice (one from each experimental group). The DNA of the isolated E. coli strains was stratified by pulsed-field gel electrophoresis (PFGE), showing two different bacterial clusters (Figure 2A). One cluster included five E. coli strains isolated from GalT-KO mice treated with GAS914 and PBS, and the other cluster involved two isolated from PBS-treated mice. Only E. coli A3 was isolated from both GAS914 and PBS-treated mice.

Next, we evaluated the impact of GAS914 treatment on the mortality rate by introducing the therapy to the mice before or after the CLP procedure (Figure 2B). The results showed that the survival of GalT-KO mice that received two doses (days 0 and 2) of GAS914 before CLP (day 3) and on days 4, 6, 9, 11, 13, and 16 after CLP was significantly higher (83%) than that of control mice treated only with PBS (35%). By contrast, no benefit was observed in mice in which treatment with GAS914 was initiated 12 h after CLP (Figure 2B). The progression and severity of the infection were also scored by detailed observation of all animals (Table S1). Mice treated with GAS914 before and after CLP showed better welfare, and the disease severity was significantly lower than in the control group (Supplementary Figure S1).

We also measured in vitro GalT-KO mice serum bactericidal activity, and the antibody and complement deposition against the pathogenic mouse isolate E. coli A3 and E. coli O86:B7 (ATCC^® 12701^™), a bacterium with high α-galactosyl content (37, 40). The results showed an average 5.4-fold enhancement of the killing capacity of mice sera against E. coli A3 in animals receiving GAS914 compared to PBS treatment (Figure 3A). Furthermore, the increased serum bactericidal capacity with GAS914 treatment was associated with a significantly lower recognition of the bacterium by IgG1 and IgG3 subclasses in animals treated with GAS914 compared to those treated with PBS (Figure 3B). However, no changes were observed in total IgG, IgM, C3, C4, IgG2a, and IgG2b deposition. Interestingly, IgG1 showed substantially higher binding to E. coli A3 than the other IgG subclasses, and GAS914 reduced the reactivity of these antibodies by an average of 20-fold, despite the low concentration of this subclass among IgG anti-αGal antibodies (Figure 1A). In contrast, serum from GAS914-treated mice evidenced somewhat improvement in bactericidal activity against E. coli O86:B7 without changes in IgG subclasses deposition (Figure S2).

The levels of anti-αGal antibodies in the serum of 8 healthy blood donors displayed a higher concentration of anti-αGal IgG than IgM and IgA antibodies and the predominance of the IgG2 subtype (Figure S3). GAS914 was combined with human sera from healthy blood donors in vitro at a 100 µg/mL concentration, estimated as the exposition achieved with in vivo doses of 5 mg/kg in primates (27). Exposure to GAS914 significantly reduced the detection of anti-αGal IgM, IgG, and IgA antibodies from individual human sera, along with the IgG2 subclass (Figure S3).

Next, we investigated whether the blockade of anti-αGal antibodies from sera of human blood donors in vitro modifies the complement-mediated killing activity against human pathogenic Gram-negative bacteria isolates. These included human E. coli O86:B7 and multidrug-resistant (MDR) human Pseudomonas aeruginosa 21565 and Klebsiella pneumoniae 35204 isolates. Inhibition of anti-αGal antibodies with GAS914 in vitro increased an average of 2-fold the bactericidal activity of human sera against E. coli O86:B7, 3.5-fold against K. pneumoniae and 1.3-fold against P. aeruginosa 21565 (Figure 4A). Although E. coli strains are mainly killed in a bactericidal assay using antibodies and complement, they can also be killed by phagocytes after bacterial opsonization by antibodies (41). Therefore, we also tested the opsonophagocytic activity of macrophages against E. coli O86:B7. Blockade of anti-αGal antibodies with GAS914 in a pool with the 8 human sera was associated with augmenting bacterial phagocytosis by macrophages (Figure S4).

We subsequently tested the deposition of IgM, IgA, IgG, and IgG subclasses, C3, and C4, on the surface of E. coli O86:B7. Neutralizing anti-αGal antibodies with GAS914 from human sera was associated with a decreased binding of IgG, specifically IgG2 antibodies (Figure 4B). We also assessed the bactericidal activity, and antibody and complement deposition from the serum of healthy blood donors against the E. coli A3 strain (isolated from the blood of GalT-KO mice after CLP) after the in vitro inhibition of anti-αGal antibodies with GAS914. No changes occurred in the bactericidal capacity, antibody binding, or complement deposition in the serum from 8 healthy blood donors, regardless of whether anti-αGal antibodies were removed (Figure S5).

It has been suggested that the impairment of serum killing by anti-αGal and other antibodies depends on binding antibodies to the long-chain O-antigen of LPS (10, 13), composed of many repeating oligosaccharide units. To characterize the Gram-negative bacteria used in the studies, we quantified and separated the repeating oligosaccharide units of LPS. We also assessed α-galactosyl expression in bacteria by lectin IB4-FITC staining and the presence of αGal by monoclonal antibodies. E. coli O86:B7 and P. aeruginosa 21565 did not show repeating oligosaccharides bands in the LPS O-antigen chains, indicating that both bacteria have short O-antigens of LPS. In contrast, K. pneumoniae 35204 and mouse E. coli A3 displayed repeating oligosaccharides bands in the LPS, like the pattern of a standard LPS with a long-chain O-antigen isolated from E. coli O111:B4 (Sigma-Aldrich, St. Louis, MO, USA) (Figure 5A).

The bacteria showed a variable expression of αGal and α-galactosyl residues assessed by monoclonal antibodies and IB4 lectin, respectively (Figure 5B). K. pneumoniae 35204 displayed the highest expression of αGal, followed by P. aeruginosa 21565 and E. coli A3. The expression of α-galactosyl residues was also very high in K. pneumoniae 35204, followed by E. coli O86:B7 and P. aeruginosa 21565. In E. coli O86:B7, there was no evidence of αGal residues, indicating that most of the antigen expressed in this bacterium corresponds to the B-blood group moiety (Galα1-3(Fucα1-2)Galβ) (40). In contrast, E. coli A3 showed some αGal without any α-galactosyl expression.