Galectin-4 Antimicrobial Activity Primarily Occurs Through its C-Terminal Domain

Abstract

Although immune tolerance evolved to reduce reactivity with self, it creates a gap in the adaptive immune response against microbes that decorate themselves in self-like antigens. This is particularly apparent with carbohydrate-based blood group antigens, wherein microbes can envelope themselves in blood group structures similar to human cells. In this study, we demonstrate that the innate immune lectin, galectin-4 (Gal-4), exhibits strain-specific binding and killing behavior towards microbes that display blood group–like antigens. Examination of binding preferences using a combination of microarrays populated with ABO(H) glycans and a variety of microbial strains, including those that express blood group–like antigens, demonstrated that Gal-4 binds mammalian and microbial antigens that have features of blood group and mammalian-like structures. Although Gal-4 was thought to exist as a monomer that achieves functional bivalency through its two linked carbohydrate recognition domains, our data demonstrate that Gal-4 forms dimers and that differences in the intrinsic ability of each domain to dimerize likely influences binding affinity. While each Gal-4 domain exhibited blood group–binding activity, the C-terminal domain (Gal-4C) exhibited dimeric properties, while the N-terminal domain (Gal-4N) failed to similarly display dimeric activity. Gal-4C not only exhibited the ability to dimerize but also possessed higher affinity toward ABO(H) blood group antigens and microbes expressing glycans with blood group–like features. Furthermore, when compared to Gal-4N, Gal-4C exhibited more potent antimicrobial activity. Even in the context of the full-length protein, where Gal-4N is functionally bivalent by virtue of Gal-4C dimerization, Gal-4C continued to display higher antimicrobial activity. These results demonstrate that Gal-4 exists as a dimer and exhibits its antimicrobial activity primarily through its C-terminal domain. In doing so, these data provide important insight into key features of Gal-4 responsible for its innate immune activity against molecular mimicry.

Graphical Abstract

Highlights

• •Gal-4 binds to blood group antigens on both mammalian and microbial cells.
• •Gal-4 dimerizes via its C-terminal domain, displaying potent antimicrobial activity.
• •Gal-4 uniquely targets pathogens that employ molecular mimicry.

In Brief

Host immune factors typically recognize determinants distinct from self. However, certain microbes employ molecular mimicry to evade immune detection. Galectin-4 (Gal-4) demonstrates the ability to bind and eliminate microbes that utilize molecular mimicry. This antimicrobial activity primarily occurs through its C-terminal domain, which not only engages microbial glycans but also facilitates protein dimerization. These results provide important insight into the binding specificity and quaternary structure of Gal-4 that likely contributes to its antimicrobial activity.

Host–microbe interactions represent a fundamental component of host immunity (1). To provide protection against an evolving antigenic landscape, host immune factors evolved the ability to engage distinct features on microbes not present on host cells, allowing host immune activity to be directed specifically toward microbes (2, 3, 4, 5, 6). While this strategy of self/non-self recognition allows host immunity to focus on potential pathogens (7, 8), the tailoring of immunity to discriminate host from microbe leaves a gap in host immunity toward microbes that are decorated with host-like molecules as a form of molecular mimicry (9). Indeed, several microbes possess the ability to envelope themselves in molecular features that mimic mammalian structures, posing a challenge for host immunity when utilizing strategies of self/non-self recognition to specifically engage and then eliminate microbes (10, 11, 12).

Gaps in adaptive immunity toward molecular mimicry are perhaps most apparent toward microbes that express variants of mammalian ABO(H) blood group antigens. In contrast to red blood cell (RBC)-induced alloantibody formation (13, 14, 15, 16, 17, 18, 19), antibodies against ABO(H) antigens are naturally occurring and form within the first few months of life (20). While conflicting data exist regarding the development of naturally occurring anti-ABO(H) antibodies (21, 22, 23, 24, 25), several studies suggest that microbes that express ABO(H)-like antigens may stimulate anti-ABO(H) antibody formation relevant to transfusion and transplantation (26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36). However, as tolerance mechanisms prevent the formation of anti-ABO(H) antibodies in ABO(H) blood group–positive individuals, how these individuals protect themselves against blood group–decorated microbes is incompletely understood. As adaptive immunity is clearly limited in its ability to generate anti-ABO(H) antibodies in blood group–positive individuals, host factors that are responsible for protecting an individual against microbes that utilize blood group–like molecular mimicry are likely confined to innate immunity. As ABO(H) antigens are carbohydrate structures, factors that contribute to this form of immunity would be predicted to possess carbohydrate-binding activity.

Several studies suggest that galectins, an ancient family of carbohydrate-binding proteins, may possess the ability to fill gaps in adaptive immunity against blood group antigens by specifically targeting microbes that utilize molecular mimicry (37, 38). However, the fine specificity and overall structural features, including the individual carbohydrate recognition domains (CRDs) of each galectin family member responsible for this activity, are only beginning to be understood. Not all galectins exhibit blood group–binding preference or possess the ability to alter microbial viability, suggesting that this activity may not be a uniform property of galectin family members (37, 38, 39, 40, 41). Prior studies also suggest that galectin quaternary structure may be important for galectin-mediated antimicrobial activity (38, 41, 42). For example, the C-terminal domain of Gal-3, which exists as a monomer, fails to exhibit antimicrobial activity despite its ability to bind to blood group–expressing microbes, while the full-length protein, which forms oligomers, exhibits antimicrobial activity (37). However, whether similar features of other galectins are likewise required for antimicrobial activity remains relatively unexplored.

Galectin-4 (Gal-4) is a tandem repeat galectin, which is a subset of galectins that possess two unique CRDs linked by a peptide. Gal-4 was first described along the gastrointestinal tract (43, 44, 45, 46), suggesting that it is uniquely poised to protect the host against blood group–positive microbes. However, while the two CRDs within Gal-4 are tethered with a linker peptide (47, 48), the overall binding specificity and relative contribution of each domain to its binding specificity and potential antimicrobial activity remains largely unknown. Although tandem repeat galectins have been thought to be monomeric proteins that possess functional bivalency by virtue of their two domain configuration (49, 50), our results demonstrate that Gal-4 exists as a dimer, with dimerization occurring through its C-terminal domain. While each domain can recognize blood group antigens, the C-terminal domain possesses a more potent ability to bind and kill microbes expressing blood group and related antigens. In contrast, the N-terminal domain fails to exhibit potent killing activity to the same strains even in the context of full-length dimeric Gal-4. These results provide new insights into the overall binding specificity and quaternary structure of Gal-4 that contributes to its ability to provide innate immunity against molecular mimicry.

Experimental Procedures

Protein Expression and Purification

Expression plasmids carrying the genes for human Gal-4, Gal-4N, Gal-4C, Gal-4NM, Gal-4CM, and Gal-4DM were introduced into Escherichia coli BL21 (DE3) and expressed following established protocols (37, 38, 41, 51, 52). In brief, bacteria harboring the plasmids were grown in LB broth supplemented with 100 μg/ml ampicillin at 250 rpm and 37 °C. Once the bacteria reached the mid-logarithmic phase, protein expression was induced by adding 1 mM IPTG, followed by incubation at 16 °C for 20 h as previously outlined for recombinant protein production in general (53, 54). After protein induction, the bacterial cultures (6 L) were pelleted, and the resulting pellets were resuspended in 60 ml of bacterial lysis buffer. The lysis buffer consisted of PBS with 14 mM 2-mercaptoethanol (2-ME), 60 μl ribonuclease A, 60 μl DNase I, 60 μl lysozyme, and two protease inhibitor cocktail tablets (Roche). Sonication-cooling cycles were employed to lyse the bacteria, with 10 s of sonication followed by 10 s of rest at 4 °C. Centrifugation at 17,000 rpm and 4 °C for 1 h was carried out twice to remove bacterial cell debris. The supernatant was passed through a lactosyl-Sepharose column for affinity purification of recombinant Gal-4, Gal-4N, Gal-4C, Gal-4NM, and Gal-4CM (52). Target protein fractions were eluted using PBS with 14 mM 2-ME and 100 mM lactose. Prior to performing biological assays, 2-ME and lactose was removed using a PD-10 gel filtration column (Cytiva). Additionally, His-tagged Gal-4DM was purified using Ni Sepharose Excel affinity resin (Cytiva) according to the manufacturer’s guidelines. Protein was concentrated and buffer exchanged to PBS using Amicon Ultra-15 10 kDa cutoff Centrifugal Filter Units (Millipore). The protein was aliquoted and stored at −80 °C prior to use as outlined previously (42, 55, 56).

ABO(H) Glycan Microarray

The ABH arrays were used as described previously (57). Briefly, Gal-4, Gal-4N, Gal-4C, Gal-4NM, or Gal-4CM were labeled with Alexa Fluor 647 N-hydroxysuccinimide ester (Thermo Fisher Scientific, A20006), with lactose added (final concentration of 100 mM) to aid in stability during the labeling reaction. The mixture was incubated for 1 h at room temperature while protected from light, following the previously outlined protocol (58). To remove any unbound lactose and unconjugated Alexa Fluor 647, a PD-10 gel filtration column was utilized. To eliminate any potentially inactive protein, labeled Gal-4, Gal-4N, Gal-4C, Gal-4NM, or Gal-4CM underwent an additional purification step using a lactosyl-Sepharose column. Bound Gal-4, Gal-4N, Gal-4C, Gal-4NM, or Gal-4CM was eluted using PBS with 14 mM 2-ME and 100 mM lactose, with lactose subsequently removed using a PD-10 gel filtration column. The purified labeled Gal-4, Gal-4N, Gal-4C, Gal-4NM, or Gal-4CM samples were used for the microarray experiments. To assess galectin recognition of glycans on the printed glycan microarray, slides were exposed to Alexa Fluor 647–labeled Gal-4, Gal-4N, Gal-4C, Gal-4NM, or Gal-4CM in PBS with 0.05% Tween-20 and 1% bovine serum albumin. Incubation took place for 1 h at room temperature in a dark and humid chamber. Subsequently, the slide underwent a series of washes, was dried using microcentrifugation, and then examined by a microarray scanner (GenePix 4000 B, Molecular devices) to capture an image of the bound fluorescence (59). ImaGene software (GenePix Pro 7) was employed to determine integrated spot intensities (60). Apparent K_D values were calculated using GraphPad Prism version 9 (https://www.graphpad.com/features) (55, 56).

Bacteria Strains

We gratefully acknowledge Moon H. Nahm from the University of Alabama at Birmingham, for kindly providing the Streptococcus pneumoniae strains. The S. pneumoniae strains were cultivated either on trypticase soy agar II blood agar plates (Becton Dickson BBL) or in Todd-Hewitt yeast broth (Todd-Hewitt broth with 0.5% yeast extract) at 37 °C in a 5% CO₂ using an Anaero-Pouch-MicroAero bag (Mitsubishi Gas Chemical Company). Once colonies had formed, they were transferred to Todd-Hewitt yeast broth and incubated at 37 °C with 5% CO₂ until reaching the exponential growth phase. Subsequently, the cultures were isolated and prepared for incorporation into the microbial microarray (MMA). The clinically isolated strains Haemophilus influenzae GAP 514, 516, 521, and 530 were cultured on chocolate agar plates (Remel, R01301) at 37 °C in a 5% CO₂ using an AnaeroPouch-MicroAero bag. The E. coli O86 and E. coli O86-ΔWaaL strains were generously provided by Peng George Wang from Georgia State University. The Providencia alcalifaciens O5 and P. alcalifaciens O19 strains were kindly provided by Y. Knirel from the ND Zelinsky Institute of Organic Chemistry in Moscow, Russia. The Klebsiella pneumoniae O1 and K. pneumoniae O4 strains were kindly provided by C. Whitfield from the University of Guelph. Each bacterial strain was grown and maintained at 37 °C using LB culture medium from Fisher.

Microbial Microarrays

The MMAs were created as described earlier (42). Bacterial cells were fixed with 1% paraformaldehyde for 24 h at 4 °C and subsequently washed with PBS. The cells were then labeled with 5 μM SYTO 13 (S7575, Invitrogen) for 15 min at room temperature in the dark. Following labeling, the bacterial cells were washed with PBS and printed onto Nitrocellulose Film-Slides (Grace bio-labs, 705108). Proteins were directly labeled with Alexa Fluor 647 (Thermo Fisher Scientific, A20006) and diluted to different concentrations in PBST (PBS, 0.05% Tween-20) containing 1% bovine serum albumin. The slides were blocked with SuperG blocking buffer (Bio-labs) for 1 h at room temperature. Similarly, the slides were incubated with varying concentrations of each galectin, as indicated, for 1 h in a dark chamber at room temperature. Slides were washed by successive immersion in PBS containing 0.05% Tween 20 (4x), PBS (4x), and water (4x) and were dried by microarray slide spinner. Fluorescence images were acquired using a microarray scanner (GenePix 4000 B, Molecular devices). Integrated spot intensities were obtained using ImaGene software (GenePix Pro 7).

Assessing the Effect of Gal-4 on Bacterial Viability

The bacteria were cultured until reaching the mid-logarithmic phase. Bacterial cells were then suspended with the specified concentrations of Gal-4 or each of its domains at 37 °C for 2 h (61). The presence of colony-forming units (CFUs) was determined using limited dilution analysis (61).

Flow Cytometry Analysis

In relation to the flow cytometry analysis for bacteria, the bacteria were cultured until they reached the mid-logarithmic phase, washed twice with cold PBS at 4 °C, and subsequently incubated with 0.1 μM Alexa Fluor 647–labeled proteins, with or without thiodigalactoside (final concentration 20 mM TDG), at 4 °C for 10 min as done previously (55). After incubation, the bacteria were washed twice with PBS and resuspended in PBS for flow cytometry analysis. As for the flow cytometry analysis of RBCs, the RBCs were resuspended and washed twice in PBS at 4 °C. Then, 1 μl packed RBCs was incubated with 0.1 μM Alexa Fluor 647–labeled protein at 4 °C for 20 min. Once the incubation was complete, the cells were washed twice and resuspended in PBS for flow cytometry analysis. All flow cytometry analyses were performed using the FACS Canto II flow cytometer (BD Biosciences) (62, 63, 64). The acquired data were processed using FlowJo version 10 (https://www.flowjo.com/solutions/flowjo/free-trial).

Agglutination Assay

Reagent RBCs (Bio-Rad) of each blood type indicated were dispensed into round-bottom 96-well plates. Serial dilutions of each galectin were mixed with the RBCs and left to agglutinate. The agglutination end point was determined as the last concentration at which agglutination occurred.

Chemical Cross-Linking Using BS³

Gal-4, Gal-4N, Gal-4C, Gal-4NM, Gal-4CM, or Gal-4DM (5 μM each) were incubated with a 50-fold molar excess of bis(sulfo-succinimidyl) suberate (BS³) as per the manufacturer's instructions for 30 min at room temperature. The reaction was quenched with 1 M Tris–HCl to neutralize any unreacted BS³. The proteins were then visualized by staining with Coomassie Blue on the SDS-PAGE (41).

Analytical Ultracentrifugation

Sedimentation velocity experiments were performed in a ProteomeLab XL-I analytical ultracentrifuge (Beckman) at 42,000 rpm and 10 ºC using an AN-50 Ti rotor. Double sector cells equipped with 1.2 cm charcoal-epon centerpieces (Beckman) and quartz windows were used. Samples were allowed to equilibrate for 2.5 h at 10 °C prior to starting the run. Galectin partial specific volume and buffer density and viscosity were calculated using SEDNTERP. Data was analyzed using the c(s) distribution and c(M) distribution models in Sedfit. Data was plotted using GUSSI and Prism.

CFG Glycan Microarray

Glycan binding on the consortium for functional genomics (CFG) glycan microarray by Gal-4 was accomplished as outlined previously (40, 55). Briefly, Alexa Fluor 647–labeled Gal-4, Gal-4N, or Gal-4C was incubated on the microarray at the indicated concentrations for 1 h at room temperature within a dark, humidified chamber. Post-incubation, slides were washed, and the fluorescence was measured using microarray scanner (GenePix 4000 B, Molecular devices). Spot intensities were then quantified using Imagene software (GenePix Pro 7).

Experimental Design and Statistical Rationale

In order to assess the reproducibility and quantitative accuracy of the study, each experiment shown was conducted at least three independent times. For the measurement of antimicrobial activity, statistical analysis was conducted using a one-way ANOVA in conjunction with Dunnett's test. ∗∗∗∗p < 0.0001.

Results

Gal-4 Recognition of ABO(H) Glycans

While numerous studies have examined the binding specificity of human galectins (40, 65, 66, 67, 68, 69, 70, 71), little is known regarding the binding specificity of Gal-4, one of the first tandem repeat galectins discovered (44, 45) (Supplemental Fig. S1), especially toward blood group antigens and microbial glycans. To begin defining the binding specificity of Gal-4, we cloned and expressed full-length Gal-4, along with its individual C-terminal (Gal-4C) and N-terminal (Gal-4N) CRDs. Given the possible impact of ABO(H) substructures on the binding specificity of galectins (37), we employed an array platform populated with distinct ABO(H) antigens that reside on six different core structures (referred to as Types I-VI) (57) (Fig. 1A). As prior studies have demonstrated that subtle differences in printing efficiency may influence perceived binding affinity following evaluation at a single concentration (72), we examined Gal-4, Gal-4N, and Gal-4C binding over a range of concentrations to provide relative dissociation constants (K_Ds) (Fig. 1, B and C). Gal-4 exhibited a strong affinity for most of the blood group antigens tested, with particularly high affinity for blood group A and B antigens and weaker binding to H antigens. Among ABO(H) blood group antigens, Gal-4, Gal-4N, and Gal-4C all exhibited a preference for B type V and VI when compared to the same B antigen present on type I and II structures. Gal-4N also showed a preference for A type V and VI but comparatively weaker binding to A type I and II antigens. Gal-4 and Gal-4C exhibited binding toward blood group A on type I through type IV structures, while even stronger binding was observed for type V and type VI blood group A. While Gal-4 bound to H structures (blood group O), neither domain exhibited a high level of binding toward these structures. Although other glycan microarrays do not commonly possess a similar number distinct to ABH glycan determinants, they are populated with additional non-ABH glycans that can provide an opportunity to determine the level of specificity of Gal-4 for ABH antigens. To accomplish this, we next turned to the CFG glycan microarray, which contains hundreds of distinct glycan structures, including ABO blood group antigens (Supplemental Figs. S2 and S3). Gal-4 demonstrated a strong affinity and specificity for blood group A and B antigens and weaker binding to H antigens. Gal-4N exhibited a preference for structures containing blood group B, while Gal-4C showed binding affinity toward glycans containing blood group A. Blood group A and B presentation on N-glycans appeared to uniquely enhance Gal-4N binding. (Supplemental Fig. S3). Detailed glycan structures are summarized in Supplemental Table S1. These results suggest that Gal-4 exhibits strong binding toward A and B blood group antigens, although ABH substructures can influence overall ABH recognition especially in the context of the isolated CRDs.

Fig. 1.

Galectin-4 engages distinct blood group antigens across a range of concentrations.A, schematic representation of the ABO(H) glycan structures employed in ABO(H) glycan microarray. B, binding isotherms acquired following incubation of Gal-4, Gal-4N, and Gal-4C for A-II, B-II, and H-II as shown at the concentrations. Error bars represent the mean ± SD. C, a heat map of the relative fluorescence units (RFU) and K_D values using blue and red colors, respectively. Darker blue reflects a higher RFU value at each concentration shown, while white donates lower RFU values. The heat map of K_D values ranges from darker red (indicating low K_D) to light red (indicating high K_D). Gal-4, galectin-4.

Gal-4 Recognizes Microbes that Express Blood Group and Related Glycans

Given the preference of Gal-4 and its individual domains for ABO(H) antigens, we next sought to investigate the binding preferences of Gal-4 and its sub-domains towards microbes that express surface carbohydrates resembling blood group antigens. To accomplish this, we employed a recently developed MMA (Fig. 2), a microarray platform populated with intact microbes, which has been shown to more accurately predict the antimicrobial activity of galectins and various microbes in a high throughput manner (42). Gal-4 selectively engaged gram-positive S. pneumoniae serotypes 33F (Sp 33F) and 14 (Sp 14), as well as gram-negative E. coli O86 (E. coli O86), K. pneumoniae O1 (KPO1), and P. alcalifaciens O5 (PAO5) (73, 74) (Supplemental Table S2). To determine whether similar binding preferences were observed for Gal-4N and Gal-4C, we likewise examined each domain over a similar concentration range (Fig. 2). Analysis of Gal-4N and Gal-4C revealed that each sub-domain also exhibited binding affinity towards Sp 33F. However, beyond engagement of Sp 33F, the binding specificity of Gal-4N and Gal-4C differed. While Gal-4C bound Sp 14, Gal-4N failed to show similar binding. Gal-4C also displayed stronger affinity toward E. coli O86, which expresses a blood group B-like antigen, when compared to Gal-4N. Importantly, neither domain bound the WaaL mutant of E. coli O86, which fails to express the blood group B-like antigen (75, 76). Similar to binding observed toward Sp 14, Gal-4C also bound PAO5 and KPO1. In contrast, Gal-4N showed minimal binding towards these same microbes. Using the same strategy employed to examine the relative binding affinity obtained from binding isotherms toward ABO(H) antigens, we likewise examined overall binding affinity of Gal-4, Gal-4N, and Gal-4C toward each microbe represented on the MMA. While Gal-4, Gal-4N, and Gal-4C all possess relatively high affinity for Sp 33F, Gal-4C exhibited more similarity in binding preferences to Gal-4 than Gal-4N.

Fig. 2.

Galectin-4 engages specific microbial strains that exhibit self-like antigens on a microbial microarray.A, binding isotherms acquired following incubation of galectins with each represented microbes depicted on the MMA. Glycan structures corresponding to each respective microbe are presented. Error bars represent the mean ± SD. B, a heat map of the relative fluorescence units (RFU) and K_D values using blue and red colors following incubation with Gal-4, Gal-4N, or Gal-4C, respectively. Darker blue reflects a higher RFU value at each concentration shown, while white donates lower RFU values. The heat map of K_D values ranges from darker red (indicating low K_D) to light red (indicating high K_D). Gal-4, galectin-4; MMA, microbial microarray.

To determine whether microbial microarray results accurately predict actual interactions with intact microbes in solution, Gal-4, Gal-4N, and Gal-4C binding towards the same microbes on the MMA bound by galectins were examined using flow cytometry (Fig. 3). Consistent with microarray results, Gal-4 readily engaged PAO5, KPO1, and Sp 33F, while pre-incubation of each galectin with TDG, a non-metabolizable inhibitor of galectins, prevented these interactions, confirming that binding required carbohydrate recognition (Fig. 3). To assess the specificity of these interactions, we investigated Gal-4 binding towards related strains that do not express blood group–like or related antigens and were not recognized in the MMA format, including Sp 2, PAO19, and E. coli O86-ΔWaaL. Gal-4 displayed negligible binding to these strains, demonstrating that, like the MMA results, Gal-4 specifically engaged microbes that express blood group–like and related mammalian carbohydrate-like structures (Fig. 3 and Supplemental Table S2). To determine whether each individual domain of Gal-4 likewise exhibits specific interactions with microbes expressing blood group and related antigens, Gal-4N and Gal-4C were similarly examined. Gal-4N and Gal-4C both recognized Sp 33F, although Gal-4C exhibited stronger binding to Sp 33F. In contrast, only Gal-4C bound to PAO5 and E. coli O86. Inclusion of TDG inhibited Gal-4N or Gal-4C interactions, confirming that these interactions were likewise carbohydrate-dependent. These results demonstrate that Gal-4 exhibits strain-specific interactions with microbes expressing blood group–like and related antigens, with the C-terminal domain appearing to most closely approximate the binding observed by the full-length protein.

Fig. 3.

Gal-4 recognizes blood group antigen-like microbes. Flow cytometric analysis of Gal-4, Gal-4N, and Gal-4C binding to Sp 33F, Sp 2, PAO5, PAO19, Escherichia coli O86, and E. coli O86-ΔWaaL with or without inclusion of TDG as indicated. The outcome of incubation of each galectin alone is labeled in blue. Each respective galectin plus TDG is labeled in orange. Unstained control is labeled in gray. Gal-4, galectin-4; TDG, thiodigalactoside.

Gal-4 Exhibits Antimicrobial Activity toward Distinct Microbial Strains

Given the ability of Gal-4 and associated domains to interact with S. pneumoniae, P. alcalifaciens, and E. coli, we sought to determine whether Gal-4 and its domains possesses antimicrobial activity toward microbes decorated with blood group–like antigens. To this end, we examined the effects of Gal-4 and each domain on microbial viability. Following incubation of Gal-4 over a range of concentrations with Sp 33F, PAO5, and E. coli O86, significant changes in microbial viability was observed, as measured by CFUs determination (Fig. 4A). In contrast, similar incubation of Gal-4 with Sp 2, PAO19, and E. coli O86-ΔWaaL failed to induce any detectable change in microbial viability when evaluated in parallel. Similar to Gal-4, Gal-4C likewise exhibited antimicrobial activity toward Sp 33F, PAO5, and E. coli O86 strains, while failing to alter the viability of Sp 2, PAO19, and E. coli O86-ΔwaaL. Notably, Gal-4N exhibited minimal antimicrobial activity against Sp 33F at the highest concentration tested, while no significant Gal-4N antimicrobial activity was observed toward the other five microbial strains examined. Inclusion of TDG prevented galectin-induced antimicrobial activity, demonstrating that the antimicrobial activity observed was carbohydrate-dependent (Fig. 4B). These results demonstrate that Gal-4 exhibits strain-specific killing activity and that the similarities in binding specificity and antimicrobial activity between Gal-4 and Gal-4C strongly suggest that the innate immune activity of Gal-4 likely resides within its C-terminal domain.

Fig. 4.

Gal-4 exhibits antimicrobial activity toward microbes that express blood group and related mammalian-like antigens.A, colony forming unit (CFU) enumeration of each respective microbes after incubation with the indicated concentrations of Gal-4, Gal-4N, or Gal-4C. B, quantification of viable bacteria after incubation with PBS alone, 5 μM of Gal-4, Gal-4N, or Gal-4C with or without TDG as indicated. Data are represented as mean values ±SD. Statistical analysis was performed using one-way ANOVA using Dunnett’s test. ∗∗∗∗p < 0.0001. Gal-4, galectin-4; TDG, thiodigalactoside.

Gal-4 and Gal-4C Have the Ability to Form Dimers

The inability of Gal-4N to similarly induce microbial death when compared to Gal-4 and Gal-4C, despite displaying some recognition of Sp 33F, is consistent with prior results demonstrating that the individual C-terminal CRD of galectin-3 (Gal-3C) can bind but likewise fails to kill blood group–expressing microbes (37). As Gal-3C exists as a monomer, it remained possible that the inability of Gal-4N to similarly kill microbes may simply reflect its monomeric state. However, as the full-length Gal-4 protein is thought to be a monomer that is functionally bivalent through each CRD (49, 50), whether individual domains of Gal-4 can dimerize remains relatively unexplored. To test this, we examined the monomeric and dimeric states of Gal-4. To accomplish this, we performed chemical cross-linking Gal-4, Gal-4N, and Gal-4C to examine possible dimerization using BS³ (41). Incubation of Gal-4N with or without BS³ failed to result in detectable Gal-4N dimers (Fig. 5A), suggesting that like Gal-3C, Gal-4N simply exists as a monomer. In contrast, incubation of Gal-4 and Gal-4C with BS³ led to significant trapping of dimeric Gal-4 and Gal-4C (Fig. 5A), suggesting that Gal-4 and Gal-4C possess the ability to dimerize. Analysis of Gal-4, Gal-4N, and Gal-4C using analytical ultracentrifugation demonstrated similar results with Gal-4 sedimenting as two species with weighted sedimentation coefficients (S_20,w) of 2.7 S and 3.9 S (77). The species at 2.7 S is consistent with monomeric Gal-4, while the species at 3.9 S aligns with dimeric Gal-4 (Supplemental Fig. S4A). Gal-4N sediments as a single species with S_20,w of 1.8 S, which represents its monomeric form (Supplemental Fig. S4B). In contrast, Gal-4C exists as a dimer (2.7 S) and a monomer (1.9 S), with a minor fraction forming a tetrameric structure at 4.8 S (Supplemental Fig. S4C). These data are consistent with findings from the BS³ cross-linking and suggest that Gal-4 can exist as a dimer, likely through its C-terminal domain.

Fig. 5.

Gal-4 and Gal-4C form dimers and agglutinate RBCs.A, coomassie blue staining of Gal-4, Gal-4N, or Gal-4C following incubation with the chemical cross-linker BS³ followed by SDS-PAGE analysis. B, flow cytometric analysis of Gal-4, Gal-4N, and Gal-4C binding to blood group A, B, or O RBCs with or without inclusion of TDG as indicated. The outcome of incubation of each galectin alone is labeled in blue. Each respective galectin plus TDG is labeled in orange. Unstained control is labeled in gray. C, cellular agglutination by Gal-4, Gal-4N, and Gal-4C. Each concentration shown represents the lowest tested concentration tested in a serial dilution at which cell agglutination was visible. BS³, bis(sulfo-succinimidyl) suberate; Gal-4, galectin-4; TDG, thiodigalactoside.

As RBC agglutination requires both binding and dimerization (40), we next sought to confirm the ability of Gal-4C to dimerize by testing whether Gal-4C alone can agglutinate RBCs. To examine this, we first examined galectin binding by flow cytometry analysis. Gal-4 displayed higher binding toward blood group A and B RBCs when compared to RBCs from blood group O individuals, consistent with the results observed on the ABH array. Similar to Gal-4, Gal-4C likewise exhibited higher binding to A and B RBCs when compared to blood group O RBCs. In contrast to Gal-4 and Gal-4C, very little binding by Gal-4N was observed toward any of the RBCs tested regardless of blood group status (Fig. 5B). To determine whether Gal-4, Gal-4N, or Gal-4C possess the ability to agglutinate RBCs, each galectin was incubated with A, B, or O RBCs across a range of concentrations. Gal-4 displayed a stronger tendency to agglutinate A-RBCs and B-RBCs in comparison to O-RBCs, which aligns with its binding preference on the microarrays array and the binding toward RBCs observed by flow cytometry. Gal-4C also induced RBC agglutination, showing a preferential hierarchy of blood group A > B > H, which likewise aligns with its binding to ABH glycan microarray and flow cytometric analysis. In contrast, Gal-4N failed to induce agglutination at any concentration tested (Fig. 5C). These results demonstrate that Gal-4 can dimerize and dimerization likely occurs through its C-terminal domain, where distinct interactions may occur between each Gal-4C domain to facilitate dimerization (Supplemental Fig. S5).

Gal-4CM, Representing Gal-4N within Full-Length Gal-4, Exhibits Higher Affinity Interactions with Blood Group Antigens

The ability of Gal-4 and Gal-4C domains to form dimers suggests that the reduced ability of the individual Gal-4N domain to display antimicrobial activity may simply reflect its monomeric state. As Gal-4C can dimerize, we next sought to determine whether Gal-4N may exhibit antimicrobial activity within the context of the full-length dimeric Gal-4 protein. To explore this, we generated a mutant of each CRD within the context of full-length Gal-4. This was accomplished by mutating critical residues (R240 in the C-terminal domain or R67 in the N-terminal domain) that are necessary for carbohydrate recognition. This resulted in a C-terminal domain mutant (R240H, Gal-4CM) and a corresponding N-terminal domain (R67H, Gal-4NM) mutant of Gal-4 (Fig. 6A). To determine whether Gal-4NM and Gal-4CM retain the ability to dimerize, we employed the same chemical cross-linking approach used to assess Gal-4, Gal-4C, and Gal-4N. Similar to full-length Gal-4, Gal-4NM and Gal-4CM retained the ability to dimerize as evidenced by successful trapping of each protein as a dimer following BS³ incubation (Fig. 6B).

Fig. 6.

Gal-4CM and Gal-4NM form dimers, engage ABO(H) glycans, and agglutinate RBCs.A, a schematic representation of the Gal-4 and Gal-4 with mutations introduced into each carbohydrate recognition domain, where the R240H mutation was introduced into the C-terminal domain (Gal-4CM) and R67H mutation was introduced into the N-terminal domain (Gal-4NM). B, Gal-4, Gal-4CM, or Gal-4NM were incubated with the chemical cross-linker BS³, followed by SDS-PAGE and Coomassie blue staining. C, heatmaps depicted illustrate the RFU and K_D values for Gal-4CM and Gal-4NM across the indicated concentration range on the ABO(H) glycan microarray. Darker blue reflects a higher RFU value at each concentration shown, while white donates lower RFU values. The heat map of K_D values ranges from darker red (indicating low K_D) to light red (indicating high K_D). D, flow cytometric analysis of Gal-4CM and Gal-4NM binding to blood group A, B, or O RBCs with or without inclusion of TDG as indicated. E, agglutination of A, B, or O RBCs by Gal-4NM and Gal-4CM as indicated. Each concentration shown represents the lowest tested concentration tested in a serial dilution at which cell agglutination was visible. BS³, bis(sulfo-succinimidyl) suberate; Gal-4, galectin-4; TDG, thiodigalactoside.

To define whether the binding specificity of the C- and N-terminal domains differ within the context of the full-length protein, we next examined the binding specificity of Gal-4NM and Gal-4CM on the ABH glycan microarray. While Gal-4N displayed lower binding affinity toward ABH glycans as an individual domain (Fig. 1C), Gal-4CM (intact Gal-4N within full length Gal-4) exhibited higher affinity and broader binding activity against each ABH antigen (Fig. 6C). In contrast, a minimal increase in Gal-4C binding, when compared with Gal-4NM (intact Gal-4C within the full-length Gal-4), was observed. Similarly, while very little binding of Gal-4N was observed toward A, B, or O RBCs (Fig. 5B), the same domain exhibited higher binding toward each RBC population within the context of the full-length Gal-4CM protein (Fig. 6D). In contrast, no increase in binding was observed by Gal-4C when examined within Gal-4NM in parallel (Figs. 5B and 6D). To determine whether the N-terminal domain can induce RBC agglutination in the context of the full-length protein, we defined the relative ability of Gal-4NM and Gal-4CM to induce agglutination of A, B, and O RBCs. Unlike the inability of Gal-4N to agglutinate RBCs, Gal-4CM not only exhibited higher binding toward A and B RBCs (Fig. 6D) but possessed the ability to induce RBC agglutination, with increased agglutination of B RBCs observed when compared to A RBCs or O RBCs (Fig. 6E). In contrast, Gal-4NM exhibited a binding hierarchy of A RBCs > B RBCs > O RBCs (Fig. 6D), similar to that observed for Gal-4C alone, and possessed the ability to induce agglutination of A RBCs more readily than B RBCs, which were in turn were more sensitive to agglutination than O RBCs (Fig. 6E). These results suggest that in the context of the full-length protein, the N-domain exhibits significant increases in binding toward ABO(H) glycans in an array format and on the surface of RBCs that is likely a result of functional bivalency achieved in the context of the dimeric Gal-4 protein.

Gal-4CM Exhibits Enhanced Microbial Recognition Compared to Gal-4N

The enhanced ability of Gal-4CM to recognize ABO(H) glycans and A, B, and O blood group RBCs suggests that a similar increase in binding may be apparent toward blood group decorated and related microbes. To examine this, we evaluated Gal-4CM and Gal-4NM over a range of concentrations on the MMA. In contrast to the relatively narrow interactions observed toward each microbe by Gal-4N, Gal-4CM bound to Sp 33F, E. coli O86, and PAO5. This binding did not reflect a general increase in nonspecific interactions, as Gal-4CM failed to recognize Sp 2, E. coli O86-ΔWaaL, and PAO19. The N-terminal domain (Gal-4CM) not only exhibited broader binding toward the microbes represented on the array but also bound with a higher apparent affinity to each microbe. In contrast, no increase in the overall affinity or changes in binding specificity for the C domain were observed following analysis of Gal-4NM when evaluated in parallel (Fig. 7, A and B). To determine whether similar increases in the N-terminal domain within Gal-4CM were likewise observed toward microbes in solution, we evaluated Gal-4CM binding toward select bound and unbound microbes as observed on the MMA by flow cytometry. In contrast to the relative inability of Gal-4N to bind E. coli O86 and PAO5, Gal-4CM exhibited detectable binding to each strain, while inclusion of TDG inhibited these interactions, demonstrating that binding was dependent on carbohydrate recognition. Gal-4CM failed to engage Sp 2, PAO19, or E. coli O86-ΔWaaL, indicating that like the outcome observed on the MMA, these interactions were strain-specific. In contrast to Gal-4CM, Gal-4NM failed to exhibit a similar degree of change in binding when compared to Gal-4C alone (Fig. 7C). These results demonstrate that in the context of the full-length protein, Gal-4N displays higher affinity for mammalian and microbial blood group and related glycans.

Fig. 7.

Gal-4CM and Gal-4NM both recognize specific microbial strains expressing self-like antigens.A, binding isotherms obtained after incubation of Gal-4NM or Gal-4CM with microbes on the MMA. Error bars indicate the mean ± SD. B, a heat map of the relative fluorescence units (RFU) and K_D values using blue and red colors following incubation with Gal-4NM or Gal-4CM, respectively. Darker blue reflects a higher RFU value at each concentration shown, while white donates lower RFU values. The heat map of K_D values ranges from darker red (indicating low K_D) to light red (indicating high K_D). C, flow cytometric analysis of binding by Gal-4CM and Gal-4NM to Sp 33F, Sp 2, PAO5, PAO19, Escherichia coli O86, and E. coli O86-ΔWaaL with or without TDG as indicated. The outcome of incubation of each galectin alone is labeled in blue. Each respective galectin plus TDG is labeled in orange. Unstained control is labeled in gray. Gal-4, galectin-4; MMA, microbe microarray; TDG, thiodigalactoside.

Gal-4NM Exhibits Minimal Anti-Microbial Activity

Given the ability of Gal-4N to exhibit bivalent interactions and higher binding affinity toward distinct strains of microbes in the context of the full-length protein, these results suggest that Gal-4N may possess antimicrobial activity when present within full-length Gal-4. To test this, we examined the antimicrobial activity of Gal-4CM toward Sp 33F, PAO5, and E. coli O86. Despite displaying sub-micromolar affinity of Gal-4CM for Sp 33F, Gal-4NM failed to exhibit the same level of antimicrobial activity displayed by Gal-4 or Gal-4C. Similarly, while Gal-4CM bound PAO5 and E. coli O86, antimicrobial activity was only observed at the highest concentration tested toward PAO5; no significant changes in the viability of E. coli O86 were observed. In contrast, Gal-4NM retained the ability to induce microbial death in Sp 33F, PAO5, and E. coli O86. Neither Gal-4CM or Gal-4NM displayed any detectable antimicrobial activity toward Sp 2, PAO19, and E. coli O86-ΔWaaL and all killing activity observed was reversed by inclusion of TDG, demonstrating that Gal-4NM– and Gal-CM–induced changes in microbial viability was strain-specific and carbohydrate-dependent (Fig. 8).

Fig. 8.

Antimicrobial activity of Gal-4NM and Gal-4CM toward strains of microbes.A, colony forming unit (CFU) enumeration of each respective microbe after incubation with the indicated concentrations of Gal-4NM or Gal-4CM. B, the CFUs remaining were determined after incubating Sp 33F, Sp 2, PAO5, PAO19, Escherichia coli O86, and E. coli O86-ΔWaaL with PBS alone, 5 μM of Gal-4NM or Gal-4CM with or without TDG as indicated. Data are represented as mean values ± SD. Statistical analysis was performed using one-way ANOVA using Dunnett’s test. ∗∗∗∗p < 0.0001. Gal-4, galectin-4; TDG, thiodigalactoside.

We next considered the possibility that the microbial killing activity of Gal-4NM may reflect retained carbohydrate-binding activity of the C domain despite the R67H mutation. To control for the possibility that each mutated domain within Gal-4NM or Gal-4CM retains carbohydrate activity and that the killing activity of Gal-4NM, although less potent, may reflect C domain activity, we generated a Gal-4 mutant that contains both mutations (R67H and R240H), referred to as the Gal-4 double mutant (Gal-4DM) (Fig. 9A). We first sought to determine whether the DM retains the ability to dimerize. Similar to Gal-4, Gal-4NM, and Gal-4CM, incubation of the Gal-4DM with the chemical cross-linker BS³ resulted in trapping of the dimeric galectin, strongly suggesting that the mutations do not prevent dimerization (Fig. 9B). To determine whether Gal-4DM possesses the ability to recognize glycan ligands, we next examined its ability to engage lactosyl-Sepharose, the common affinity support used to assess general galectin activity (78). Despite retaining its dimeric state, Gal-4DM did not bind lactosyl-Sepharose (Fig. 9C). Consistent with its lack of carbohydrate-binding activity, Gal-4DM likewise failed to bind and agglutinate A, B, or O RBCs (Fig. 9, D and E). Finally, to determine whether Gal-4DM possesses any antimicrobial activity, we examined microbial viability over a range of concentrations. In contrast to the ability of Gal-4NM and Gal-4CM to exhibit some degree of antimicrobial activity toward Sp 33F, PAO5, and E. coli O86, Gal-4DM was inactive against the same panel of microbes (Fig. 9, F and G).

Fig. 9.

The Gal-4 double mutation does not recognize RBCs or alter microbial viability.A, a schematic representation of the Gal-4 (R67H and R240H) double mutant (Gal-4DM) is shown. B, incubation of Gal-4DM with or without the chemical cross-linker BS³ followed by SDS-PAGE and Coomassie blue staining. C, the chromatograms of lactose affinity column analysis of Gal-4 and Gal-4DM. D, flow cytometric analysis of Gal-4DM binding to A, B, or O RBCs with or without TDG, as indicated. E, agglutination of A, B, or O RBCs by Gal-4DM as indicated. Each concentration shown represents the lowest tested concentration in a serial dilution at which cell agglutination was observed. F, colony forming units (CFUs) remaining after incubation of each microbe with the indicated concentrations of Gal-4DM. G, CFU enumeration of each respective microbe after incubation with PBS or 5 μM of Gal-4DM in the presence or absence of TDG as indicated. BS³, bis(sulfo-succinimidyl) suberate; Gal-4, galectin-4; TDG, thiodigalactoside.

Discussion

Through an integrated application of mammalian and microbial glycan array analysis, quaternary structural determination, and evaluation of antimicrobial activity, the findings of the present study demonstrate that the tandem repeat galectin, Gal-4, actually exists as a dimer and that dimerization occurs through its C-terminal domain. These data also suggest that while both domains may contribute to microbial recognition and participate in antimicrobial activity in the context of the full-length protein, the C-terminal domain appears to be the primary domain involved in its antimicrobial activity. In doing so, these data suggest that, in contrast to prior concepts regarding the overall configuration of tandem repeat galectins in general wherein each separate CRD is tethered through a linker peptide to produce a monomeric and functionally bivalent protein (79, 80), some tandem repeats can actually exist as a dimer, with functional consequences with respect to their antimicrobial activity.

The ability of Gal-4 to form dimers adds to growing evidence that tandem repeat galectins, like prototypical galectins, can dimerize. Previous studies demonstrated that like Gal-4, Gal-8 can also exist as a dimer (41). In contrast to Gal-4, Gal-8 dimerization occurs through its N-terminal as opposed to C-terminal domain (41). Gal-8 also differs from Gal-4 in that the binding specificity of each domain is very distinct, with the C-terminal domain engaging blood group antigens, while blood group modifications actually inhibit the ability of the N-terminal domain to recognize the underlying lactosamine structure (38, 41, 42, 66, 81). In contrast to Gal-8, Gal-4N and C-terminal domains display similar binding profiles, where both domains can engage blood group antigens. However, the fine specificities of each domain did differ. For example, while Gal-4N and Gal-4C exhibited blood group–binding activity, the C terminal domain of Gal-4 exhibited a slightly enhanced preference for blood group A when compared to blood group B, while the N-terminal domain instead displayed some binding preferences for blood group B. Despite these differences, each domain bound A and B glycans, in the context of mammalian glycan presentation. It is possible that native Gal-4 possesses distinct glycan-binding preferences from that observed for the recombinant protein. However, the ability of recombinant Gal-4 and each domain to be purified over lactosyl-Sepharose and therefore retain carbohydrate-binding activity strongly suggests that the native confirmation is preserved (51, 52, 82). Indeed, challenges associated with isolating and labeling sufficient native protein to conduct similar studies, especially from human tissues, have resulted in recombinant approaches being nearly uniformly used to study galectin–glycan interactions (51, 52, 81, 82, 83). These results add to growing evidence that galectins in general exhibit binding affinity for these polymorphic carbohydrate antigens (40). These data also add to growing insight into the glycan-binding preference of Gal-4 in general (84, 85, 86, 87, 88, 89, 90) and demonstrate that individual galectins can possess very distinct glycan preferences depending on the type of modification occurring on lactosamine or similar base glycan configurations.

Despite exhibiting enhanced binding affinity toward ABO(H) and microbial blood group-like glycans in the context of the full-length proteins, the N-terminal domain failed to exhibit potent antimicrobial activity within Gal-4CM. These results stand in contrast to Gal-3, where reconstitution of the C-terminal domain within the context of the full-length protein not only facilitates oligomerization but also restores potent antimicrobial activity (37). While Gal-4N did exhibit some antimicrobial activity within the context of the full-length protein (Gal-4CM), it was significantly reduced when compared to Gal-4, suggesting that the primary domain responsible for antimicrobial activity is Gal-4C (Table 1). Consistent with this, Gal-4NM, in which the C-terminal domain is intact, displayed more potent antimicrobial activity than Gal-4CM. However, these findings do not rule out the possibility that in the context of the full-length protein, the N domain of Gal-4 contributes to microbial glycan recognition and/or antimicrobial activity. The relatively reduced antimicrobial activity of Gal-4NM when compared to Gal-4C alone may also reflect possible interference by the N-terminal domain of optimal Gal-4C oligomerization and/or engagement of secondary microbial-binding events that are required for optimal antimicrobial activity as an isolated domain. However, when both domains are present, this inhibition may be offset by the ability of the N-terminal domain to contribute to microbial engagement and/or antimicrobial activity. However, the relatively potent antimicrobial activity of Gal-4C alone demonstrates that the N-terminal domain is not required for Gal-4 to exhibit antimicrobial effects on target microbes. These findings suggest that in the context of the full-length protein, each domain likely participates in microbial binding and antimicrobial activity but that Gal-4C retains this activity in the absence of Gal-4N (Table 1).

Table 1.

Summary of Gal-4: Bacterial binding, killing, and protein dimerization

Protein	Structure	Bacterial binding	Bacterial killing	Protein dimerization
Gal-4		+++++	++++	Yes
Gal-4N		+	-	No
Gal-4CM (R240H)		++	+	Yes
Gal-4C		++++	+++	Yes
Gal-4NM (R67H)		+++	++	Yes
Gal-4DM (R67H and R240H)		-	-	Yes

The ability of Gal-4 to dimerize and more particularly Gal-4C to retain dimeric activity has broad implications on Gal-4 activity. As noted above, galectins have been thought to exert many of their biological activities through engagement and cross-linking of counter receptors through their dimeric activity (91, 92, 93, 94, 95). Galectins expressed along epithelial surfaces may not only engage host cell glycans, where they can modify the behavior of host cells (79, 96), but have also been shown to be secreted in an apical fashion (97, 98, 99), suggested that the protein may accumulate in the intestinal lumen. Several galectins are sensitive to oxidative inactivation as a mechanism to temporally and spatially regulate their activity, while other galectins, including chimeric Gal-3 and tandem repeat galectins, have been primarily thought to be regulated through proteolytic cleavage of the intervening sequences responsible for functional bivalency (79, 95, 100). Indeed, several proteases have been shown to cleave tandem repeat Gal-8 and Gal-9, rendering these proteins as isolated CRDs (101). While similar cleavage of Gal-4 may reduce key features of its biological activity, with respect to its antimicrobial properties, such an event would be less likely to similarly impact its function in innate immunity. However, as the N-terminal domain likely possess other activities, cleavage of Gal-4 may reflect an important ability to dissociate biological activities that require an intact N-terminal domain within the full-length protein without compromising the ability of Gal-4C to continue to provide innate immunity against molecular mimicry.

These results illustrate the utility of microarrays when seeking to establish the binding specificity of carbohydrate-binding proteins in general and galectins in particular (102). The ability to examine galectins over a range of concentrations on the MMA provides an opportunity to establish relative binding affinities that may not be appreciated following examination of each protein at a single concentration as done previously (103, 104). Indeed, in prior studies, glycan microarray analysis has primarily occurred at a single concentration where subtle differences in printing efficiency may impact the apparent binding preferences of a given glycan-binding protein (60, 105, 106). Similarly, when direct labeling techniques are employed to produce a fluorescently or otherwise labeled protein, variation in the labeling efficiency of a given galectin or glycan-binding protein may alter the perceived strength of the corresponding signal generated following analysis on an array or by flow cytometry, which can make it difficult to compare the relative affinity of different galectins when using a single or only a few concentrations. By examining the binding affinity over a range of concentrations, relative binding affinities can provide a more complete picture of the binding affinity of a given galectin toward an array of mammalian and microbial glycans. While the distinct glycan features on each microbe engaged by galectins remain to be defined, the ability of galectins to selectively engage individual strains of microbes that have previously been shown to express unique glycans with blood group–like features coupled with the preference of Gal-4 for blood group antigens on mammalian glycan microarrays suggests that Gal-4 may engage microbial mimics of blood group antigens. As microbes often configure their blood group expression as polymers with unique linkages, the distinct binding preferences observed by Gal-4 toward various A, B, and H antigens may actually reflect underlying preferences for blood group antigens expressed by microbes in which similar glycan presentation may be present. Consistent with this, while a variety of infectious diseases may have facilitated the selection of blood group polymorphisms within the human population (107, 108, 109, 110), it is unlikely that galectins evolved to recognize mammalian blood group antigens to directly alter the biology of the host as these structures are polymorphic and are not known to be functional ligands whereby galectins influence host cell behavior. Instead, these binding preferences suggest that the ability of galectins to engage microbes that utilize blood group and related forms of molecular mimicry likely reflects their ability to provide innate immunity against molecular mimicry (37, 38, 42, 55, 56, 73).

In summary, these results demonstrate that Gal-4 exists as a dimer, likely through dimeric interactions of its C-terminal domain. These data also demonstrate that despite possessing the ability to engage ABO(H) antigens and similar microbial glycans, changes in microbial viability is not the inevitable outcome of galectin engagement. The reduced ability of Gal-4N to kill microbes expressing blood group and related antigens suggests that in addition to binding a microbe, other features of antimicrobial galectins may be ultimately responsible for their antimicrobial activity. Future studies will be needed to further dissect other features of galectins that are necessary for their antimicrobial selectivity. Taken together, our results provide unique insights into the overall quaternary structure of Gal-4 and the key domains responsible for its binding specificity and antimicrobial activity.

Data Availability

All data presented in the main text or supplementary information of this work are accessible from the lead contact upon request. Requests for additional information, resources, and reagents should be directed to the lead contact, Sean R. Stowell (srstowell@bwh.harvard.edu).

Supplemental data

This article contains supplemental data.

Conflict of interest

The authors declare no competing interests.