Research Article

Protein crystallization promotes type 2 immunity and is reversible by antibody treatment

See allHide authors and affiliations

Science  24 May 2019:
Vol. 364, Issue 6442, eaaw4295
DOI: 10.1126/science.aaw4295

A crystal-clear ingredient for allergy?

Charcot-Leyden crystals (CLCs) are formed from the eosinophil granule protein galectin-10 (Gal10) and found in severe eosinophil-associated diseases like asthma and chronic rhinosinusitis. Whether CLCs actively contribute to disease pathogenesis is unknown. Persson et al. found that lab-grown Gal10 crystals are biosimilar to CLCs (see the Perspective by Allen and Sutherland). When given to mice, the crystals acted as a type 2 adjuvant, mimicking many of the features of human asthma. In contrast, a Gal10 mutein unable to crystallize had no effect. Antibodies against epitopes crucial for Gal10 autocrystallization could dissolve both in vitro–generated Gal10 crystals and patient-derived CLCs. Furthermore, these anti-Gal10 antibodies reversed the effects of Gal10 crystals in a humanized mouse model of asthma, suggesting a potential therapeutic approach for crystallopathies more broadly.

Science, this issue p. eaaw4295; see also p. 738

Structured Abstract

INTRODUCTION

Spontaneous protein crystallization rarely occurs in vivo. When it does, crystals are generated, which sustain long-term protein storage or enable the slow release of proteins. In 1853, Charcot reported extracellular bipyramidal crystals in the airways of asthmatics, an observation also made by Leyden in 1872. Charcot-Leyden crystals (CLCs) have since been described mostly in eosinophil-rich inflammatory lesions. They have become a hallmark of eosinophil death and can persist in tissues for months. CLCs are composed of galectin-10 (Gal10), one of the most abundant proteins in human eosinophils. Recent studies suggest that Gal10 is released from the cytoplasm of activated eosinophils. However, whether Gal10 can have a functional role in airway disease and type 2 immunity in vivo after a phase transition to a crystalline state is unknown.

RATIONALE

To test the hypothesis that CLCs stimulate immunity in the lung, we produced recombinant Gal10 crystals that were structurally and biochemically similar to CLCs obtained from patients with rhinosinusitis and asthma. Additionally, we engineered Gal10 muteins that selectively lost the ability to crystallize. Using these tools, we studied immune responses in mouse models of asthma. To complement these experiments in mice, we studied Gal10 expression in human samples and developed antibodies that bind and dissolve CLCs.

RESULTS

CLCs were abundantly present in the airways of chronic rhinosinusitis patients and correlated with the degree of eosinophil extracellular trap formation. Biosimilar crystalline Gal10 injected into the airways of naïve mice induced an innate immune response, rich in neutrophils and monocytes, and led to the uptake of crystals by dendritic cells (DCs). Soluble Gal10 muteins carrying a mutation of Tyr69 to glutamic acid were unable to crystallize and were immunologically inert. Simultaneous injection of CLCs with innocuous ovalbumin (OVA) resulted in DC uptake and T helper type 2 cell priming, together with airway eosinophilia and immunoglobulin G1 (IgG1) responses. Mechanistically, these effects were accompanied by NLRP3 inflammasome activation and interleukin-1β (IL-1β) release. However, the observed response to CLCs in vivo could occur independently of the NLRP3 inflammasome. In an effort to develop new therapeutic opportunities against this type of crystallopathy, we generated antibodies against crystalline Gal10. The epicenter of each crystal-dissolving antibody-binding epitope on Gal10 was situated at Tyr69, a residue we had identified as a critical crystal-packing hotspot. These antibodies rapidly dissolved preexisting CLCs in vitro and in the native mucus environment of patients. Crystal-dissolving antibodies suppressed airway inflammation, goblet-cell metaplasia, bronchial hyperreactivity, and IgE synthesis induced by CLC and house dust mite inhalation in a humanized mouse model.

CONCLUSION

Our results demonstrate that CLCs are more than just markers of eosinophilic inflammation. Rather, Gal10 is released by activated eosinophils and undergoes a phase transition to a crystalline state that actively promotes key features of asthma. Antibodies can rapidly dissolve CLCs abundantly present in the native mucus of patients and resolve key features of CLC crystallopathy in a preclinical model. Although protein crystallization is a rare event, we establish Charcot-Leyden crystallopathy as a druggable trait in patients with airway disease and provide a rationale for how antibodies can dissolve protein crystals.

Dissolving CLCs with antibodies.

CLCs that are abundant in airways in type 2 immunity can be dissolved by using antibodies that target key residues of the crystal-packing interface on Gal10. This strategy leads to the resolution of key asthma-like features in mice. mAb, monoclonal antibody.

Abstract

Although spontaneous protein crystallization is a rare event in vivo, Charcot-Leyden crystals (CLCs) consisting of galectin-10 (Gal10) protein are frequently observed in eosinophilic diseases, such as asthma. We found that CLCs derived from patients showed crystal packing and Gal10 structure identical to those of Gal10 crystals grown in vitro. When administered to the airways, crystalline Gal10 stimulated innate and adaptive immunity and acted as a type 2 adjuvant. By contrast, a soluble Gal10 mutein was inert. Antibodies directed against key epitopes of the CLC crystallization interface dissolved preexisting CLCs in patient-derived mucus within hours and reversed crystal-driven inflammation, goblet-cell metaplasia, immunoglobulin E (IgE) synthesis, and bronchial hyperreactivity (BHR) in a humanized mouse model of asthma. Thus, protein crystals may promote hallmark features of asthma and are targetable by crystal-dissolving antibodies.

Asthma is a chronic inflammatory disease characterized by the influx of type 2 immune cells, such as eosinophils, T helper type 2 cells, mast cells, and basophils, into the airways. These cells contribute to bronchial hyperreactivity (BHR) and mucus overproduction, leading to airway obstruction (1, 2). In 1853, Charcot and Robin, and then Leyden in 1872, described extracellular deposits of morphologically diverse crystals in inflamed tissues of patients with asthma (3, 4). Similar crystals have been observed in chronic rhinosinusitis, helminth infections, and cancer concurrent with tissue eosinophilia (58). Thus, Charcot-Leyden crystals (CLCs) serve as markers of eosinophilic inflammation. The precise cellular source and biochemical nature of CLCs had long remained enigmatic. These crystals were initially proposed to be made of an inorganic substance. Although spontaneous protein crystallization is a rare event, later studies revealed the proteinaceous nature of CLCs (9). Early biochemical work proposed that CLCs were made from a cytoplasmic protein with lysophospholipase activity (10), but the actual constituent was eventually found to be galectin-10 (Gal10), one of the most abundant yet least understood proteins found in the cytoplasm of eosinophils and basophils (11). Furthermore, CLCs rich in Gal10 were shown to form spontaneously upon hypotonic lysis of eosinophil cell lines or purified eosinophils obtained from inflamed airways (1214). CLCs can form in hypoxic eosinophils and in cells that have undergone NADPH (reduced nicotinamide adenine dinucleotide phosphate)–dependent eosinophil extracellular trap cell death (EETosis) (15, 16).

For >100 years, eosinophil-derived CLCs have been described anecdotally in mucus plugs obstructing the airways of patients with severe asthma and in the tenacious eosinophilic mucin of patients with chronic rhinosinusitis with nasal polyps (CRSwNP) (7, 17, 18). However, a direct role for CLCs or Gal10 in eosinophilic airway disease or mucus production has not been established. Establishing this link would be important because eosinophils are drug targets for the treatment of asthma and CRSwNP. How exactly eosinophils contribute to inflammation and chronic airway disease has remained unclear (19). Many of the abundant proteins contained within eosinophil granules, such as eosinophil cationic protein, eosinophil peroxidase, and major basic protein, can induce damage to lung structural cells. However, these proteins are still viewed mainly as biomarkers of inflammation and disease rather than drug targets for tissue eosinophilia (19, 20). As many inorganic and organic crystals have the capacity to trigger innate inflammation (21, 22), we hypothesized that eosinophil-derived CLCs may directly contribute to disease.

Abundance of CLCs in type 2 immunity

To evaluate the presence of CLCs in clinically relevant samples, we first performed immunofluorescence microscopy analysis of Gal10 in mucus and tissue biopsy samples from patients with severe CRSwNP who had undergone sinus surgery. Ten out of 15 CRSwNP patients had tenacious eosinophilic mucin (7, 15), which had abundant Gal10-immunoreactive bipyramidal crystalline structures (Fig. 1A). Gal10-immunoreactive CLCs were also found in resected nasal mucosa tissues from 11 out of 15 patients, in close proximity to areas of epithelial damage (Fig. 1B) and eosinophil degranulation (as characterized by the presence of extracellular DNA traps) (Fig. 1C). The number of CLCs found in biopsy samples strongly correlated with the presence of EETosis (coefficient of determination R2 = 0.6542; P = 0.0001) and with levels of the eosinophil growth factor interleukin-5 (IL-5) (R 2= 0.2988; P = 0.0285) (Fig. 1, D and E). Furthermore, the in vitro formation of CLCs from purified eosinophils occurred within minutes after activation by phorbol 12-myristate 13-acetate (PMA). This was accompanied by the extrusion of DNA into the extracellular space and the release of extracellular granules immunoreactive for Gal10 (Fig. 1F), as recently reported (15). Notably, immunoreactive CLCs were observed only occasionally in induced sputum samples from patients with severe asthma (see table S1 for clinical characteristics), most likely because sputum processing entails the use of reducing agents. However, when patient sputum samples were examined by an enzyme-linked immunosorbent assay (ELISA), immunoreactive Gal10 was detected in 8 out of 11 samples. By contrast, none of the control patient samples were positive for Gal10 (Fig. 1G). The concentration of soluble Gal10 correlated with the percentage of eosinophils in the induced sputum (Fig. 1H). Thus, the presence of CLCs or Gal10 is a biomarker for airway eosinophilia, as reported previously (15, 23).

Fig. 1 Gal10+ CLCs are present in chronic rhinosinusitis and asthma patients.

(A) Immunofluorescence staining with an anti-Gal10 antibody (green) or isotype (green) of tenacious eosinophilic mucin obtained from a CRSwNP patient. Scale bar, 7.5 μm. Data are representative of 10 patient samples, with one mucus sample per patient. (B) Immunostaining for Gal10 (green) on a biopsy sample taken from a patient with severe CRSwNP. Nuclei are labeled blue with DAPI. Scale bars, 25 μm. Data are representative of 16 patients; for each patient, three different tissue pieces and 10 randomly selected power fields (63× magnification) per tissue piece were analyzed. (C) Gal10-DAPI staining of tissue from CRSwNP patients shows a typical subepithelial EETosis lesion containing Gal10 associated with DNA, CLCs (*), and extracellular vesicle formation (arrows). Scale bar, 10 μm. Data are representative of 16 patients. (D and E) For each patient, three different tissue pieces were analyzed. The average number of crystals and the percentage of EETosis per high-power field (hpf) for 10 randomly selected power fields (63× magnification) were determined for each patient. Each patient was concluded to have no crystals, fewer than 10 crystals per power field (pf), or >10 crystals per power field on the basis of the average number of crystals. The data were analyzed by using pairwise comparisons in the Kruskal-Wallis H test with post hoc Dunn’s test accordingly to correlate with the levels of IL-5. Error bars represent the range from lowest to highest values. ns, not significant; *P < 0.05. (F) Peripheral eosinophils were stimulated with PMA for 15, 20, 25, and 30 min. At 15 min, a sporadic crystal (*) was observed. After 20 min, more extensive eosinophil extracellular traps (EETs) (arrowheads) containing Gal10 were observed. In addition, extensive extracellular vesicle formation (arrows) and a sporadic crystal were observed. After 25 min, most eosinophils adopted a typical EETosis morphology and almost no intact bilobed eosinophils remained. Gal10 immunoreactivity was abundantly present in close association with DNA, and extensive extracellular vesicle formation and more prominent crystals were observed. After 30 min, very prominent CLCs were observed and DNA appeared to disintegrate. Pictures are representative of results obtained from four donors, with three technical replicates and three fields per replicate. Scale bars, 10 μm. (G) Levels of soluble Gal10 measured in the sputum supernatants from 11 severe asthmatics and three healthy controls. Patient characteristics are listed in table S1. Data were analyzed by the Mann-Whitney U test. Error bars indicate the mean ± SD. (H) The percentage of eosinophils in the sputum cell pellet, plotted against the levels of soluble Gal10 across patients and controls. CI, confidence interval.

Structure of ex vivo CLCs and engineering of crystallization-deficient Gal10

To perform mechanistic studies of CLCs, we deemed it necessary to either isolate CLCs from their native source or find a way to produce biosimilar crystals. Previous biochemical and structural characterization of Gal10 relied on the autocrystallization behavior of protein-rich lysates from primary human blood eosinophils or leukemic cell lines, which led to the copurification of contaminating proteins such as lysophospholipase (12, 13, 2426). We harvested individual CLCs from the tenacious mucus of a CRSwNP patient to enable crystallographic studies using microfocus synchrotron x-rays (Fig. 2A). One such ex vivo patient-derived CLC measuring ~1 μm by 5 μm by 30 μm yielded high-quality x-ray diffraction data to 2.2-Å resolution (Fig. 2A and table S2). In parallel, we produced dimeric recombinant Gal10 as a hexahistidine-tagged protein in Escherichia coli (Fig. 2B), and the protein was soluble to at least 30 mg/ml. The protein preparation autocrystallized in phosphate-buffered saline (PBS) upon the proteolytic removal of the hexahistidine tag (Fig. 2, C and D). The crystals thus obtained enabled the structural determination of recombinant Gal10 to 1.3-Å resolution (table S2). Notably, patient-isolated CLCs, recombinant Gal10 crystals, and the published crystal structure of Gal10 obtained by lysis of the eosinophilic leukemia cell line AML14.3D10 and crystallization in vitro [Protein Data Bank (PDB) code 1LCL] (13, 27) were crystallographically equivalent, sharing the same hexagonal space group (P6522) and unit-cell parameters and featuring dimeric Gal10 wherein the Gal10 dimer was generated by crystallographic symmetry (Fig. 2, E and F). In all three crystal structures, the putative carbohydrate-binding sites were located at the dimer interface. By determining the crystal structure of Gal10 in complex with ribose, these binding sites were found to be fully solvent accessible (Fig. 2G). Analysis of Gal10 packing in this crystal form identified residues Cys29 and Cys57—pointing to the solvent channels of the crystal—as candidates for 5-iodoacetamidofluorescein (5-IAF) labeling. Thus, we were able to readily generate fluorescent Gal10 crystals (Fig. 2H), which closely resembled the various crystal morphologies originally described by Charcot and Leyden (Fig. 2I) (4).

Fig. 2 Recombinant Gal10 crystals are structurally similar to in vivo CLCs.

(A) Harvest of a CLC from tenacious eosinophilic mucus obtained from a CRSwNP patient and representative x-ray diffraction pattern determined by using microfocus synchrotron x-rays. Scale bar, 50 μm. d, Bragg planes d-spacing (Å). (B) SEC elution profile of recombinant N-terminally His-tagged Gal10 plotted as the light-scattering (LS) intensity at 90° (left vertical axis) as a function of the elution volume (V). The molar mass as determined by MALLS (right vertical axis) is reported as the number-average molar mass across the elution peak ± SD. n = 1 sample injected with His-tagged Gal10. (C) Coomassie-stained reducing SDS-PAGE gel for purified His-tagged Gal10 before and after TEV protease treatment. M, molecular mass. (D) The proteolytic removal of the N-terminal His tag of recombinant human Gal10 by TEV protease induces the spontaneous formation of crystals with CLC morphology. Scale bar, 50 μm. Spontaneous crystal formation by TEV-cleaved recombinant Gal10 was highly reproducible across different Gal10 protein batches. (E) Structural superposition of the x-ray structures determined from a patient-derived CLC (yellow), a recombinant Gal10 CLC (purple), and a cell line AML14.3D10 CLC (green). The low root mean square deviation (RMSD) shows similarity. C and N indicate the C and N termini. (F) The crystal packing and symmetry of an ex vivo CLC reveal the arrangement of Gal10 dimers along the c axis of the crystal unit cell. This axis is parallel to the long axis of the CLC bipyramids. (G) The binding of ribose to recombinant Gal10 crystals reveals putative carbohydrate-binding pockets at the Gal10 dimer interface. (H) Crystal formation after TEV treatment of Gal10 labeled with 5-IAF, demonstrating various crystal shapes within a single preparation. Scale bars, 50 μm. (I) CLC morphologies as described in original drawings by Jean-Martin Charcot, taken from (44). The lower part of this illustration shows a mucus plug obtained from a severe asthmatic, containing numerous CLCs.

The structural information at high resolution (Fig. 3A) revealed two potential crystal-packing interfaces (C1 and C2) in CLCs. Subsequently, several variants with single-point mutations in the C1 and C2 regions were engineered and screened for their potential to abrogate the autocrystallization of Gal10. For example, a Tyr69→Glu mutation that probed the C1 crystal-packing interface yielded a Gal10 variant that was resistant to autocrystallization and was soluble up to concentrations of 20 mg/ml in PBS (Fig. 3B). Wild-type (WT) Gal10 and the crystallization-deficient Tyr69→Glu mutant (referred to as Gal10mut hereafter) were biochemically and structurally equivalent, as they shared a constitutive dimeric assembly (Fig. 3, C to E).

Thus, it was possible to use autocrystallizing recombinant Gal10 to produce large quantities of CLCs that were biosimilar to patient-derived CLCs. Furthermore, we could engineer soluble and crystallization-deficient Gal10 muteins to serve as appropriate controls to investigate CLC biology.

Fig. 3 Design and characterization of nonautocrystallizing Gal10 variants.

(A) Overview and details of the two types of CLC crystal-packing interfaces. Gal10 molecules are shown as blue and yellow cartoons and surfaces. Amino acid residues probed by site-directed mutagenesis in C1 and C2 are shown in stick representation. (B) Comparison of the autocrystallization behaviors of a panel of mutant variants of Gal10 by light microscopy. Results are representative of two replicate experiments (n = 2) with identical outcomes. Scale bars, 500 μm. (C) SEC elution profile of the TEV-treated Gal10mut variant (Tyr69→Glu) plotted as the light-scattering (LS) intensity at 90° (left y axis) as a function of the elution volume (V). The molar mass as determined by MALLS (right y axis) is reported as the number-average molar mass across the elution peak ± SD. n = 1 sample injected with Gal10mut. (D) Cartoon representation of the determined crystal structure of the Gal10mut dimer in space group P21 with the protomers colored blue and yellow. The crystallographic dimer for WT Gal10 in space group P6522 is superposed as a gray cartoon with an overall RMSD of 0.37 Å for 265 aligned C-α atoms. (E) Theoretical scattering profile for dimeric Gal10mut calculated from the determined x-ray structure (red) fitted against the experimental scattering data (black). The lower inset plot is the error-weighted residual difference plot, Δ/σ. n = 1 sample analyzed by SAXS for Gal10mut.

CLCs promote innate and adaptive immunity

Although CLCs have been observed in patients with asthma or sinusitis for more than a century and are a sign of intense eosinophil activation (15), their possible role in causing airway inflammation or their presence as circumstantial bystanders has not been resolved. We next addressed whether CLCs could have proinflammatory effects in the lungs. Naïve C57BL/6 mice received an intratracheal injection of 100 μg of endotoxin-low and protease-free crystalline Gal10 or control soluble Gal10mut. After 6 and 24 hours, there were prominent influxes of neutrophils and monocytes, respectively, into the airways of mice receiving Gal10 crystals but not those receiving soluble Gal10mut or vehicle (Fig. 4A). This influx was accompanied by the production of large quantities of IL-6 and tumor necrosis factor–α (TNF-α) in bronchoalveolar lavage (BAL) fluid (Fig. 4B) and of IL-1β and the monocyte chemoattractant CCL2 in lung tissue (Fig. 4C) in the crystal-treated group. Twenty-four hours after the injection of fluorescent CLCs into lungs, intact extracellular crystals could be observed, together with endocytosed fluorescent material in Siglec F–expressing (Siglec F+) alveolar macrophages (Fig. 4D). Fluorescent material could be also seen within an intracellular compartment of CD11c+ MHCII (major histocompatibility complex class II)+ dendritic cells (DCs) in the draining mediastinal lymph nodes (mLNs) (Fig. 4E). DCs patrolling the lung and migrating to the nodes can be divided into two subsets (28). Further flow cytometric analysis of mLN migratory MHCIIhi DCs revealed that both Xcr1+ conventional DC subset 1 (cDC1) and Xcr1 CD172+ cDC2 carried fluorescent material to the draining mLNs on days 1 to 3 postinjection (Fig. 4, F and G). Thus, Gal10 induces an innate immune response only in its crystalline state, and crystals are taken up by antigen-presenting cells.

Fig. 4 Gal10 crystals activate innate and adaptive immune responses.

(A) Neutrophil (left) and monocyte (right) influx in the lungs was measured 6 and 24 hours after the injection of Gal10 crystals or soluble Gal10mut. Data are shown as mean ± SD. Results are pooled from two independent experiments with n = 6 mice per group. Two-way ANOVA: ns (not significant), P ≥ 0.05; ****P < 0.0001. (B) Concentration of cytokines in BAL fluid. Results are pooled from two independent experiments with n = 5 mice per group. Two-way ANOVA: ns, P ≥ 0.05; ****P < 0.0001. (C) Concentration of cytokines per milligram of protein of dispersed lung tissue. Data are shown as mean ± SD. Results are pooled from two independent experiments with n = 5 mice per group. Two-way ANOVA: ns, P ≥ 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. (D) IAF-labeled crystals 24 hours after intratracheal injection and uptake of fluorescent material by Siglec F+ (red) alveolar macrophages. (E) IAF-Gal10 crystal–derived proteins inside MHCII+ (red) DCs in CD11c-enriched cell fractions of mLNs. Scale bars for (D) and (E), 10 μm. (F) Representative dot plots of mLN cells gated for cDCs (Lin CD64 CD11c+ MHCIIhi) 24 hours after IAF-Gal10 injection. Cells are divided into XCR1+ cDC1 and CD172+ XCR1 cDC2. Numbers within plots are percentages of cells in each quadrant. Fluo Gal10 cryst, fluorescently labeled Gal10 crystals. (G) Percentages of fluorescent (Fluo+) cDC subsets (D1 and D3) at 24 and 72 hours after the injection of IAF-Gal10 crystals. Data are shown as mean ± SD. Results are pooled from two independent experiments with n = 3 to 6 mice per group. Three-way unbalanced ANOVA: *P < 0.05; **P < 0.01; ***P < 0.001. (H and I) Proliferation of OT-II T cells in the mLNs 72 hours after intratracheal administration of OVA alone, OVA plus Gal10 crystals, or OVA plus soluble Gal10mut. (H) Representative histograms of the division profile of the adoptively transferred OT-II T cells (CD4+ CD45.1+ Vα2+). (I) The numbers of OT-II T cells in the mLNs. Data are shown as mean ± SD. Results are from one experiment with n = 5 mice per group. One-way ANOVA: ns, P ≥ 0.05; **P < 0.01; ****P < 0.0001. (J to L) WT C57BL/6 mice were sensitized intratracheally with OVA alone, OVA plus Gal10 crystals, or OVA plus soluble Gal10mut on days 0 and 1 and challenged intranasally with OVA on days 11 to 13. (J) Flow cytometric analysis of the eosinophil (left) and neutrophil (right) influx in the BAL fluid. One-way ANOVA: ns, P ≥ 0.05; *P < 0.05; ***P < 0.001; ****P < 0.0001. (K and L) Serum OVA- and Gal10-specific IgG1 antibodies were measured by ELISA. One-way ANOVA: ns, P ≥ 0.05; ****P < 0.0001. Data are shown as mean ± SD. Results are pooled from two independent experiments with n = 6 to 8 mice per group. αOVA and αGal10, anti-OVA and anti-Gal10; A450, absorbance at 450 nm.

Because DCs bridge innate and adaptive immunity, we asked whether Gal10 crystals may boost adaptive immunity to co-administered model antigens. When the innocuous protein antigen ovalbumin (OVA) was injected intratracheally together with Gal10 crystals, adoptively transferred OVA-reactive CD4+ T cell receptor (TCR) MHCII-restricted transgenic T cells (OT-II T cells) showed increased proliferation 4 days later in draining mLNs (Fig. 4H). This led to the accumulation of higher numbers of OVA-reactive T cells in the mLNs (Fig. 4I), indicative of boosted cellular immunity. These effects were not seen when OVA was mixed with soluble Gal10mut. We next tested whether Gal10 crystals could also promote type 2 sensitization to inhaled harmless antigens, typically observed in asthmatics (1). Mice were immunized intratracheally with Gal10 crystals and OVA on days 0 and 1 and received OVA intranasal challenges on days 11 to 13. These mice exhibited airway inflammation characterized by an influx of eosinophils and neutrophils (Fig. 4J). Levels of type 2 cytokines in the supernatants of mLN restimulation cultures were below the limit of detection, yet mice mounted a robust OVA-specific immunoglobulin G1 (IgG1) response, a hallmark of type 2 immunity (Fig. 4K). These effects were not observed when OVA alone was used as the immunogen or when OVA was co-administered with soluble Gal10mut. Thus, the induction of type 2 immunity was specifically related to the crystalline state of Gal10 (Fig. 4, J and K).

Given that mice do not carry LGALS10 and therefore do not produce Gal10, we reasoned that Gal10 might be acting as a neoantigen, eliciting an antibody response directed to Gal10 epitopes that are displayed in a repetitive manner in the crystal lattice structure. Mice immunized intratracheally with Gal10 crystals, but not with soluble Gal10mut, in the absence of any adjuvant, mounted high titers of Gal10-specific IgG1 antibodies (Fig. 4L). The adjuvant activity of crystalline Gal10 on the humoral adaptive immune response to OVA or Gal10 was also seen when mice were injected intraperitoneally (fig. S1, a and b). Furthermore, this response was abolished in T cell–deficient Cd3e−/− mice. Thus, type 2 humoral immunity to crystalline Gal10 is T cell dependent (fig. S1c).

Inflammasome-dependent and -independent effects of CLCs

Many inorganic and organic crystals have the potential to elicit the secretion of IL-1β from inflammatory cells (21, 22) by triggering the NLRP3 (Nod-like receptor protein 3) inflammasome. This leads to ASC (apoptosis-associated specklike protein containing a caspase recruitment domain) adaptor recruitment and caspase 1 activation for pro–IL-1β processing, particularly when cells are first primed with lipopolysaccharide (LPS) (29, 30). A recent study suggested that CLCs can trigger the NLRP3 inflammasome in vitro and in vivo but did not report whether the NLRP3 inflammasome was required for pro-inflammatory effects (31). In an in vitro culture of bone marrow DCs, Gal10 crystals, but not soluble Gal10, induced low levels of IL-1β production in a caspase 1– and 11–, NLRP3-, and Toll-like receptor 4 (TLR4)–dependent manner, indicative of NLRP3 inflammasome activation. Notably, IL-1β levels were lower than those in the response induced by uric acid crystals (fig. S2a). However, there was no reduction in the crystal-induced cellular influx of neutrophils or monocytes when Gal10 crystals were injected into the airways of Casp1−/− Casp11−/−, Nlrp3−/−, or Tlr4−/− mice (fig. S2, b to d). In these mice, the type 2 humoral IgG1 immune response to Gal10 crystals was similarly unaffected (fig. S2, e and f). The finding that inflammation and immunity induced by Gal10 crystals were intact in Tlr4−/− mice also attests that residual endotoxin contamination could not be the reason for the observed biological effects induced by the crystals. Inflammasome activation can also lead to pyroptosis, and other crystals have been shown to cause necroptosis (22). However, after CLC administration, lactate dehydrogenase (LDH) was not released into the airways, indicating minimal necrotic cell death (fig. S2g). Additionally, innate inflammation was not reduced in Ripk3/ mice, which have defective necroptosis (fig. S2h). Thus, although CLCs can activate the NLRP3 inflammasome in vitro (31), this pathway is dispensable for CLCs to elicit inflammation and immunity in vivo. CCL2, IL-6, and TNF-α were also rapidly activated after crystal injection in vivo, pointing to other potential mechanisms by which these crystals, like others, can elicit inflammation and immunity, including complement and coagulation cascade activation (32).

Antibodies that dissolve CLCs

As the highly druggable NLRP3 pathway did not appear to be important in this form of protein crystallopathy, we sought to design an alternative strategy to interfere with Gal10 crystal biology. Antibodies from outbred llamas have highly functionally diverse variable (V) regions that are very similar to human V regions and can be screened by using phage display to design therapeutic antibodies with the SIMPLE Antibody platform (33). We immunized llamas three times with adjuvant-free Gal10 crystals, which led to a 4 log increase in Gal10-specific antibody titers compared with those in preimmune serum (fig. S3). After phage panning on immobilized Gal10 and periplasm collection, ELISA binding and surface plasmon resonance (SPR) demonstrated the successful generation of a high-affinity Gal10-reactive single-chain variable fragment (scFv), and these were engineered into conventional mouse and human IgGs (fig. S3). Because Gal10 autocrystallization is reliably reproducible in vitro, we used a crystallization robot to screen these antibodies for their potential to block Gal10 autocrystallization (fig. S4). We identified several that could abolish the propensity of Gal10 to crystallize. A natural extrapolation of these findings led us to investigate whether such antibodies would be able to solubilize already-formed Gal10 crystals. The addition of purified 1D11, 4E8, or 6F5 antibodies to preformed Gal10 crystals completely dissolved the crystals within 2 hours, an effect not seen with isotype controls (Fig. 5A). The rate of crystal dissolution was linear, albeit variable among the various antibody clones (Fig. 5B). The mode of crystal dissolution proceeded from the tips of the bipyramidal Gal10 crystals inward and along the long axis of the crystals (Fig. 5A and Movie 1).

Fig. 5 Characterization of anti-Gal10 crystal-dissolving antibodies.

(A) Time-lapse imaging after the addition of three different clones of Gal10 monoclonal antibodies to preformed, in vitro–grown Gal10 crystals. Scale bars, 50 μm. Snapshots in time are representative of three replicate experiments for each antibody. (B) Normalized total crystal surface area plotted as a function of time after the addition of Gal10 antibodies. The curves represent average values from three replicate experiments for each antibody. (C to E) X-ray structures for the Gal10-Fab complexes with Fab fragments derived from solubilizing anti-Gal10 antibodies. The insets magnify the interaction of the Fabs with Gal10 residue Tyr69. HC, heavy chain; LC, light chain. (F) Crystal-packing interface of Gal10 in CLCs, illustrating how residue Tyr69 mediates crystal-packing interactions by stacking against a symmetry-related Tyr69′ residue around a crystallographic twofold axis of symmetry perpendicular to the long c axis of the unit cell with 65 symmetry.

Movie 1. Dissolution of preformed Gal10 crystals by crystal-dissolving antibodies.

Time-lapse imaging following the addition of three different clones of Gal10 monoclonal antibodies to preformed, in vitro–grown Gal10 crystals. (Top left) Clone 4E8; (top right) clone 6F5; (bottom left) clone 1D11; (bottom right) isotype control. Scale bars, 50 μm.

To elucidate the mechanism of antibody-mediated crystal dissolution, we determined crystal structures of Fab fragments from clones 1D11, 4E8, and 6F5 in complex with Gal10 (Fig. 5, C to E). Notably, the epicenter of each antibody-binding epitope on Gal10 was situated at Tyr69, a residue we had identified as a critical crystal-packing hotspot (Figs. 3, A and B, and 5F). This residue was located at the tips of the Gal10 dimer, far from the putative carbohydrate-binding sites of Gal10 along the dimer interface. Antibody clones 1D11 and 6F5 appeared to mimic the same principles exploited by Tyr69 to mediate crystal-packing interactions in CLCs in vivo (i.e., a stacking interaction of tyrosine side chains and the concomitant accommodation of their hydroxyl groups via hydrogen bonding to an aspartic acid residue) (Fig. 5, C and D). In CLCs, Tyr69–Tyr69′ interactions via aromatic ring stacking spearheaded the packing of Gal10 along the long 65 screw axis of symmetry in the crystal. Furthermore, symmetry-related Tyr69 residues were concentrated within the lattice planes defined by the a and b axes of the unit cell and projected outward to the edges of the CLC lattice (Fig. 5F). Together, these features may explain why Tyr69 in crystalline Gal10 is such a strong immunogenic factor for the generation of crystal-dissolving antibodies and may provide a kinetic rationale for the observed mode of crystal dissolution. This may result from antibodies binding directly to Tyr69 residues that are exposed in the lattice planes at the tips of the CLC bipyramids and/or readily available in soluble Gal10 existing in equilibrium with crystalline Gal10.

Antibody treatment for CLC crystallopathy

To evaluate the therapeutic potential of these antibodies in humans, we next investigated whether they could also solubilize CLCs in tenacious eosinophilic mucus samples derived from patients with CRSwNP. Such CLCs had grown in vivo within chronically ill patients and were found in a native mucus environment. Incubating fresh allergic mucin containing numerous CLCs with Gal10 antibodies, we observed crystal dissolution with antibody clones 1D11, 4E8, and 6F5 but not with control isotype antibody (Movie 2 shows dissolution by the 1D11 clone). Encouraged by the potency of anti-CLC antibodies, we next sought to probe their therapeutic potential in vivo by resorting to a humanized mouse model of asthma. In this model, immunodeficient nonobese diabetic (NOD) Rag2−/− Il2rg−/− (NRG) mice received peripheral blood mononuclear cells (PBMCs) containing effector lymphocytes and mononuclear cells from donors allergic to house dust mites (HDMs), followed by HDM challenge exposures. This regimen results in the development of airway inflammation and human IgE synthesis (34). The use of immunodeficient mice allowed us to avoid any confounding effects of murine Gal10 IgG1 antibodies, which would inevitably be induced in this 28-day protocol. However, as human eosinophils are not contained in the PBMC fraction and it is difficult to adoptively transfer viable human eosinophils to mice, we still had to administer Gal10 crystals intratracheally at the time of HDM allergen exposure (fig. S5) to address our question. Mice were treated intratracheally with isotype antibodies or 1D11 crystal-dissolving antibodies every other day. After 28 days, isotype-treated mice receiving CLCs and HDMs had a markedly increased influx of inflammatory cells around the airways compared with that in mice receiving only HDMs, demonstrating that CLCs also had a strong pro-inflammatory effect in this more chronic mouse model (Fig. 6A, quantified in Fig. 6B). CLCs enhanced goblet-cell metaplasia and mucus production, as visualized by enhanced periodic acid–Schiff (PAS) staining of goblet cells (Fig. 6C) and substantiated by quantitative measurement of the mRNA of the goblet-cell mucin MUC5AC (Fig. 6D). The administration of CLCs with HDMs also boosted human IgE synthesis in this model, compared with that in the presence of HDMs alone (Fig. 6E). BHR is an essential feature of asthma. We therefore addressed the effect of CLC administration on the responsiveness of mechanically ventilated mice to the inhaled bronchoconstrictor methacholine. The addition of CLCs to HDMs boosted the degree of bronchoconstriction compared with that observed with HDMs alone (Fig. 6F). Mice in the HDM-CLC group did not survive doses of methacholine above 200 mg/ml. Treatment with the 1D11 antibody completely neutralized the inflammatory effect of CLCs, the ability of CLCs to enhance IgE synthesis, goblet-cell metaplasia, and BHR (Fig. 6). Thus, we have generated antibodies able to dissolve human CLCs in native mucus and to inhibit key features of airway CLC crystallopathy in a humanized mouse model.

Movie 2. Dissolution of CLCs in tenacious mucus obtained from a patient with severe CRSwNP.

Time-lapse imaging following the addition of monoclonal antibody 1D11 to mucus samples obtained from a CRSwNP patient undergoing sinus surgery. Arrows point to CLCs undergoing dissolution. Scale bar, 50 μm.

Fig. 6 Therapeutic potential of Gal10 antibodies.

NRG mice received PBMCs from a donor allergic to HDMs. The mice then received seven intratracheal injections of HDM extract with or without Gal10 crystals. The mice receiving Gal10 crystals were treated with the 1D11 crystal-dissolving antibody intratracheally or with the control isotype every other day throughout the experiment. All mice (n = 6 to 8 mice per group) were challenged with HDM extract 1 day before section and end point analysis at day 28. Ab, antibody. (A) H&E staining (200× magnification). Scale bars, 200 μm. Data are representative of four independent experiments. (B) Investigator-blinded quantitative image analysis of the number of inflammatory cells extending into a 500-μm-perimeter region from the basement membrane (bas. membr.), expressed per length of basement membrane. Kruskal-Wallis H test: *P < 0.05. Data are representative of two experiments. (C) PAS staining (top, 200× magnification; bottom, 1000× magnification) of lung sections. Scale bars, 200 μm (top) and 20 μm (bottom). Data are representative of two independent experiments. (D) Quantitative reverse transcription PCR for the goblet cell–associated mucin MUC5AC. Kruskal-Wallis H test: *P < 0.05; **P < 0.01. Data are representative of two experiments. (E) Serum concentration of human IgE measured by ELISA. Kruskal-Wallis H test: **P < 0.01; ***P < 0.001. Data are representative of two experiments. (F) Bronchoconstriction measured as dynamic airway resistance (Rrs) after the inhalation of increasing concentrations of the bronchoconstrictor methacholine (MCh). Data are representative of two experiments. Kruskal-Wallis H test: *P < 0.05.

Discussion

Collectively, these reported data advance several exciting concepts. First, these results show that CLCs are more than just markers of tissue eosinophilia, solving a century-long biomedical puzzle and adding another member to the growing list of crystallopathies (22). We note that CLC crystallopathy is marked by increased airway inflammation, mucus production, BHR, and facilitated type 2 immunity. We propose that CLC formation constitutes a key effector mechanism and may explain how degranulating eosinophils contribute to the onset of disease, either by initiating sensitization to inhaled antigens or by perpetuating or altering key features of asthma. Notably, CLCs induced neutrophilic inflammation in mice. Mixed neutrophilic and eosinophilic inflammation is often reported in the most severe asthma cases (35), and it will be interesting to study whether CLCs are a mechanism by which severe eosinophilic activation can coincide with airway neutrophilia (35). We also observed boosted humoral immunity in WT mice and enhanced human IgE synthesis in humanized mice. Recent genome-wide genetic and epigenetic association studies have found a strong association of total IgE levels with hypomethylation at the CLC or LGALS10 gene locus, suggesting that crystalline Gal10 may also be a trigger for IgE synthesis in human type 2 immune disease (36, 37).

To date, there is no animal model of asthma wherein CLCs form endogenously. Yet, the chitinase-like protein Ym1 in mice is encoded by the Chil3 gene and also readily forms protein crystals at sites of type 2 inflammation. Ym1 crystals have been termed pseudo-CLCs (3840). However, Ym1 is not produced by mouse eosinophils. Rather, it is highly induced in alternatively activated macrophages that are stimulated by the type 2 cytokine IL-4 or IL-13 (41, 42). Blocking antibodies directed against Ym1 also suppress neutrophilic inflammation in the lungs of helminth-infected mice, demonstrating a notable parallel to the responses induced by CLCs in humans (43). Therefore, it appears that protein crystallization is a more generalized feature of type 2 immunity, originating from distinct protein sources in different species. Protein crystallization may offer an old evolutionary advantage to the host (e.g., to enable resistance to chronic helminth infections or enforce key aspects of innate and adaptive type 2 immunity).

Second, both in patient material and in mice, CLCs tended to associate with mucus and promote goblet-cell metaplasia, indicating a potential self-amplifying effect on mucus production. Seven years after his first report of CLCs in 1853, Charcot reported the accumulation of CLCs in mucus plugs of asthmatics (44). Similarly, Ym1 crystals have also been reported within mucus in mouse models of asthma (41, 45). In the mucus of CRSwNP patients, we observed numerous sharp crystals, directed in all angles, which appeared to impinge on epithelial cells. Thus, the sharpness and random orientation of these crystals as well as their extension into the epithelial barrier layer may impede mucus-plug clearance. The preferential accumulation of CLCs in mucus may be related to the carbohydrate recognition domain in the Gal10 dimer interface (14). In sheep, eosinophils selectively express Gal14, not encoded in the human genome. Although Gal14 does not crystallize into CLCs, it can bind directly to airway mucins in a sheep HDM-driven model of asthma (46).

Third and most notably, our findings are directly translatable as a therapeutic concept for very common yet severe diseases. CLCs are typically found in the airways of patients with tenacious mucus, high tissue eosinophil counts, and high serum IgE concentrations. This combination is found in patients who have severe asthma, some forms of cystic fibrosis, allergic bronchopulmonary aspergillosis, sinus disease with nasal polyps, and allergic fungal sinusitis (7, 17, 18, 47). These diseases are often recurrent after medical and surgical treatment and are increasingly treated with biologicals targeting eosinophils (1, 48). Although these treatments reduce tissue eosinophilia and asthma exacerbation frequency, the mucus plugging and concomitant fixed airway obstruction often persist, and there is an urgent need for additional therapies targeting airway mucus plugging (49). Given the stability and long half-life of CLCs (44), it is conceivable that CLCs persist in mucus plugs even after eosinophilia has been resolved and sustain the airway neutrophilia that has been linked to mucus production (50). High-affinity therapeutic antibodies to Gal10 rapidly dissolve crystals in patient mucus samples and reduce CLC-induced mucus production, BHR, inflammation, and IgE synthesis. However, it remains to be seen how they will perform in humans with asthma or other airway diseases rich in eosinophils.

Lastly, the physicochemical transition of a protein from solution to a crystalline state will likely substantially affect the subsequent recognition and response by the immune system. We here provide clear evidence that this is the case, as the crystalline but not the soluble state of the neoantigen Gal10 is able to induce immunity in mice and llamas. A similar observation was reported for crystalline versus soluble human serum albumin immunization in rats (51). In tumor-bearing mice, IgG autoantibodies to Ym1 have also been described, although whether their presence is linked to the crystalline state of Ym1 remains to be determined (52). The determination of immunogenicity by the physicochemical state of a protein may also affect our understanding of human autoimmunity. A well-known autoantigen, insulin, transitions in a highly regulated way from a stored crystalline state in the pancreas to an active soluble form in circulation upon glucose feeding. However, the transfer of crystalline insulin to DCs and macrophages via the uptake of dense core granules from pancreatic β cells is an early event in type 1 diabetes, leading to the initiation of autoimmunity (53, 54). Thus, protein crystallopathy is more widespread than originally thought and may be targeted by therapeutics that dissolve or prevent crystal formation.

Materials and methods

CRSwNP and asthma patients

Nasal polyposis was diagnosed on the basis of symptoms, clinical examination, nasal endoscopy, and sinus computed tomography scanning according to the guidelines of the “European position paper on rhinosinusitis and nasal polyps 2012” (55). All patients refrained from using oral and/or topical corticosteroids for at least 4 weeks before surgery. The study and collection of samples were approved by the ethics committee of the Ghent University Hospital, and an informed consent was obtained from all patients prior to enrollment in the study.

A diagnosis of severe asthma was defined according to the Global Initiative for Asthma (GINA) guidelines. Each subject underwent spirometry before the beginning of the procedure. All patients gave informed consent, and the collection of samples was approved by the local ethical committee (AP-HM) and was part of the bronchial obstruction and asthma cohort [Cohorte Obstruction Bronchique et Asthme (COBRA)] research sponsored by the French National Institute of Health and Medical Research, INSERM (IDRGB 2008-A00284-51, Afssaps 2008-0113).

Mice

WT C57BL/6 mice (6 to 7 weeks old) were purchased from Janvier Labs (Saint-Berthevin, France). CD45.1 Rag2−/− OT-II and Tlr4−/− mice were bred at the animal house facility of the VIB-UGent Center for Inflammation Research. Casp1/11−/− (56), Nlrp3−/− (57), and Asc−/− (58) mice were initially obtained from R. Flavell (Yale University) and V. Dixit (Genentech) and housed at the VIB-UGent Center for Inflammation Research. Cd3e−/− mice were kindly provided by S. Goriely (ULB, Belgium). We did not use randomization to assign animals to experimental groups. Investigators were not blinded to group allocation during experiments. No animals were excluded from the analysis. All experiments were approved by the animal ethics committee at Ghent University.

Sputum collection

Airway mucosal tissue and sticky allergic mucin–type mucus were obtained from CRSwNP patients undergoing endoscopic sinus surgery and stored overnight at 4°C in RPMI-1640 (Sigma-Aldrich) containing antibiotics (50 IU/ml penicillin and 50 mg/ml streptomycin; Thermo Fisher Scientific) and 0.1% bovine serum albumin (BSA) (Sigma-Aldrich).

Sputum from asthmatics was induced by inhalation of a hypertonic saline solution (3%). Processing was performed according to European Respiratory Society (ERS) recommendations (59), and after collection, the volume of induced sputum was determined. A volume of 0.1% dithiothreitol (DTT) (Sigma-Aldrich) was added equal to four times the sputum weight. After homogenization, sputum samples were filtered (48-μm nylon cell strainer) and subsequently centrifuged at 800 × g for 10 min. The supernatant was aspirated and stocked at −80°C for later analysis. Cell pellets obtained were then resuspended in PBS and cytocentrifuged (Cytospin 2; Shandon Instruments). Cytospins were stained with Diff-Quick (Merz-Dade), and differential cell counts were expressed as a percentage of 400 cells.

Soluble Gal10 in sputum samples was quantified by ELISA. A flat-bottom 96-well plate (Greiner) was coated overnight at 4°C with 50 μl per well with a monoclonal rabbit anti–human Gal10 antibody (EPR11197) (Abcam) diluted 1:1000 in coating buffer (Thermo Fisher Scientific). After washing and blocking with assay diluent (Thermo Fisher Scientific) for 1 hour at room temperature (RT), 50 μl per well of samples, diluted 1:100 in assay diluent, was added in duplicate and incubated for 2 hours at RT. After washing, 50 μl per well of a polyclonal goat anti–human Gal10 antibody (R&D Systems) at 0.4 μg/ml in assay diluent was added for 45 min at RT, followed by a polyclonal biotinylated donkey anti–goat antibody (Jackson ImmunoResearch) at 1.9 μg/ml in assay diluent for 45 min at RT. After washing, 50 μl per well of streptavidin–horseradish peroxidase (HRP) reagent (1:250 diluted in assay diluent; BD Biosciences) was added for 30 min at RT. The plate was then washed, and 50 μl per well of 1× trimethylboron (TMB) substrate solution (Thermo Fisher Scientific) was added. The plate was incubated at RT in the dark. To stop the reaction, 25 μl per well of stop solution (2.5 N H2SO4) was added. Finally, the absorbance was read at 450 nm with a Perkin Elmer multilabel counter, and data were collected with Wallac 1420 Manager software.

Immunostaining for CLCs

Allergic mucin from CRSwNP patients was fixed with 4% paraformaldehyde (PFA) and embedded in paraffin. Tissue slides (5 μm) of the embedded mucin were incubated for 1 hour with 0.05% trypsin-EDTA (Thermo Fisher Scientific) at 37°C and for 1 hour with blocking buffer [7.5% BSA (Sigma-Aldrich) in PBS] at RT. Anti–human Gal10 antibody (EPR11197) (5.22 μg/ml; Abcam) was applied overnight at 4°C, and bound antibody was detected with a fluorescein isothiocyanate (FITC)–labeled goat anti–rabbit antibody (A11034) (5 μg/ml; Thermo Fisher Scientific). Polyclonal rabbit IgG isotype control (5.22 μg/ml; Abcam) served as the primary staining control. After washing, slides were mounted with Vectashield containing 4′,6-diamidino-2-phenylindole (DAPI) (1.5 mg/ml; Sigma-Aldrich) and analyzed with a confocal laser-scanning microscope (Leica TCS SPE with an ACS Apo 63×/1.30 oil cs lens, and running with LAS AF 2.6.0 software).

Purification of patient-derived CLCs

One gram of the allergic mucin was finely cut and thoroughly diluted in 10 ml of RPMI-1640 (Sigma-Aldrich) containing antibiotics (50 IU/ml penicillin and 50 mg/ml streptomycin; Thermo Fisher Scientific), 0.1% BSA (Sigma-Aldrich), and 1 mg/ml collagen type 2 (Worthington). The mixture was further homogenized using a GentleMACS Dissociator (Miltenyi Biotec) and subsequently incubated at 37°C for 45 min under continuous rotation. After centrifugation, the pellet was dissolved in 3 ml of PBS containing antibiotics, to which 6 ml of Ficoll-Paque (GE Healthcare) was added. After centrifugation at 250 × g and the removal of the supernatant and most of the Ficoll layer, the pellet was dissolved 1:10 in PBS supplemented with antibiotics. This precipitation process was repeated five additional times. The final crystal-containing pellet was resuspended in 0.2 ml of PBS supplemented with antibiotics.

Production of recombinant Gal10 crystals

A synthetic codon-optimized DNA sequence encoding human Gal10 (residues 1 to 142, Uniprot Q05315) was cloned into the NcoI/XhoI sites of the pET28a bacterial expression vector (Novagen, catalog no. 69864-3) with a His tag and two protease cleavage sites, enterokinase (DDDDK) and tobacco etch virus (TEV) protease (ENLYFQG), at the N terminus (MASTTHHHHHHDTDIPTTGGGSRPDDDDKENLYFQGHM) (single-letter abbreviations for the amino acid residues are as follows: A, Ala; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr). pET28a-Gal10 was transformed into BL21(DE3) cells by using kanamycin (25 μg/ml) as a selection marker. Expression cultures were grown at 28°C in Luria-Bertani medium containing kanamycin (25 μg/ml). The expression of Gal10 was induced when the optical density at 600 nm (OD600) of the culture was 0.6, by the addition of isopropyl-β-d-thiogalactopyranoside (ITPG) to a final concentration of 1 mM. The culture was then allowed to grow overnight. The bacteria were harvested by centrifugation (6000 × g for 20 min at 4°C), and the cellular pellet was stored at −80°C. The bacterial pellet was thawed and resuspended in lysis buffer (50 mM NaH2PO4 and 300 mM NaCl, pH 7.4). The cells were lysed by sonication on a Branson sonifier (total run time of 4 min with pulses of 30 s at 30% output interspersed with 30 s of downtime). Cell debris was removed by centrifugation at 4°C (20,000 × g for 30 min). The supernatant was clarified by filtration using a 0.22-μm bottle-top filter and loaded onto a Ni Sepharose column equilibrated with 50 mM NaH2PO4 and 300 mM NaCl, pH 7.4. The column was then washed with loading buffer supplemented with 20 mM imidazole and 0.1% EMPIGEN detergent, followed by washing with loading buffer supplemented with 20 mM imidazole. The protein was subsequently eluted using loading buffer supplemented with 50 and 500 mM imidazole. The 50 and 500 mM elution peaks were pooled, concentrated, and injected onto a HiLoad 16/600 Superdex 200-pg column using PBS (pH 7.4) as the running buffer. The elution fractions corresponding to Gal10 were pooled and stored at −80°C. Endotoxin levels were determined with an Endosafe-PTS system (Charles River) to be lower than 1 EU/mg of recombinant protein. The His-tagged Gal10 protein was soluble up to at least 30 mg/ml. To form recombinant Gal10 crystals, N-terminally His-tagged Gal10 (2 to 4.5 mg/ml) was incubated with in-house–produced TEV protease (60) at a protease:target protein ratio of 1:100 (w/w). The pRK793 plasmid encoding His-tagged TEV protease was a kind gift from D. Waugh (Addgene plasmid 8827). Following overnight digestion, the protein solution was agitated by inverting it five times, after which the solution turned cloudy in about 30 min due to autocrystallization. To remove the TEV protease and to obtain a concentrated crystal suspension, the crystal-containing suspension was spun down (600 × g for 5 min) and washed three times with sterile and endotoxin-free PBS before the crystals were resuspended in 15 ml of PBS. The total protein concentration of this solution was determined by solubilizing 20 μl of the crystal suspension in an equal volume of 6 M guanidinium hydrochloride and measuring the absorbance at 280 nm with a Thermo Scientific NanoDrop 2000. The calculated extinction coefficient for cleaved Gal10 was 19,940 cm−1 M−1.

Fluorescently labeled forms of Gal10 crystals were produced by adding a fluorescein group to Gal10. The thiol-reactive dye 5-IAF was solubilized in 100% dimethylformamide to a concentration of 100 mM. The pH of the Gal10 protein solution (~5 mg/ml) was adjusted to 8.5 by adding 100 mM tris (pH 8.5) [using a 1 M tris (pH 8.5) stock solution]. A 10-fold molar excess of 5-IAF to Gal10 (monomer) was added to the protein solution, and the labeling reaction mixture was kept at RT in the dark for 2 hours. For WT Gal10 carrying an N-terminal His tag, a molar extinction coefficient of 21,430 cm−1 M−1 was used. For the Gal10 Tyr69→Glu variant (without an N-terminal His tag), an extinction coefficient of 18,450 cm−1 M−1 was used. The excess of unreacted 5-IAF was then quenched by adding 5 mM DTT. The excess 5-IAF was then removed by running the sample on a 50-ml HiTrap desalting column (GE Healthcare) using PBS as the running buffer. 5-IAF–labeled Gal10 was then concentrated and injected onto a HiLoad 16/600 Superdex 200-pg column. Elution peak fractions were then pooled and stored at −80°C. Endotoxin levels were determined with an Endosafe-PTS system (Charles River) to be <1 EU/mg of recombinant protein. Crystal formation was induced by TEV protease as for nonfluorescent crystals.

Production of a crystallization-resistant Gal10 mutein

To produce a nonautocrystallizing Gal10 variant, Gal10 muteins carrying single mutations at residues involved in crystal-packing interactions were designed. The following Gal10 mutations were produced using QuikChange site-directed mutagenesis (Agilent): Tyr69→Glu, Lys99→Glu, Tyr139→Ala, Tyr139→Val, Tyr139→Gln, Glu10→Ala, and Glu10→Lys. The pET28a-Gal10 plasmid was used as a template. The expression and purification of Gal10 muteins were identical to those for WT Gal10. The autocrystallization propensity of Gal10 muteins (4 mg/ml) was then evaluated by incubating the purified proteins with TEV protease at a protease:target protein ratio of 1:100 (w/w). The formation of crystals was evaluated by observation under a Leica M1615C (with PLAN 0.8× LWD objective) light stereo zoom microscope equipped with a Leica IC80 HD camera. The Gal10 Tyr69→Glu variant (Gal10mut) was selected for in vivo studies. Following TEV-mediated removal of the N-terminal tag on Gal10mut, the N-terminal tag and His-tagged TEV were removed by running the digestion mixture on a 5-ml Ni Sepharose HisTrap column using PBS as the running buffer. The column flowthrough, containing Gal10mut, was then concentrated and injected onto a HiLoad 16/600 Superdex 200-pg column. The size exclusion chromatography (SEC) elution fractions corresponding to Gal10mut were pooled and stored at −80°C. Endotoxin levels were determined with an Endosafe-PTS system (Charles River) to be <1 EU/mg of recombinant protein.

SEC–multiangle laser light scattering (MALLS)

Protein samples (100 μl) were injected onto a Superdex 200 Increase 10/300 GL column (GE Healthcare), with PBS (pH 7.4) as the running buffer at 0.5 ml min−1, coupled to an online UV detector (Shimadzu), a multiangle light-scattering miniDAWN TREOS instrument (Wyatt), and an Optilab T-rEX refractometer (Wyatt) at 25°C. A refractive index increment (dn/dc) value of 0.185 ml g−1 was used for protein concentration and molecular mass determination. Data were analyzed using the ASTRA6 software (Wyatt). Correction for band broadening was applied using parameters derived from BSA injected under identical running conditions.

Crystal structure determination of ex vivo CLCs and recombinant Gal10

Single ex vivo CLCs were harvested from enriched patient mucus solution by using mounted cryoloops. Before being cryo-cooled in liquid nitrogen, crystals were protected by a brief soak in PBS supplemented with 35% (v/v) glycerol. Autocrystallized recombinant Gal10 crystals were cryoprotected in PBS supplemented with 35% (v/v) glycerol. Recombinant Gal10 crystals in complex with d-ribose were obtained by autocrystallization in PBS in the presence of 300 mM d-ribose, and crystals were cryoprotected with 25% (w/v) polyethylene glycol, molecular weight 400 (PEG-400). A crystal structure of Gal10mut was also obtained. For this, Gal10mut was concentrated to 6 to 7 mg/ml before crystallization experiments. Sitting-drop nanoliter-scale vapor diffusion crystallization experiments were performed at 293 K using a Mosquito crystallization robot (TTP Labtech) and commercially available sparse-matrix screens (Molecular Dimensions, Hampton Research). Gal10mut crystals appeared after 24 hours in condition D12 of the PEG/Ion screen (Hampton Research) [0.2 M ammonium citrate (pH 5.1), 20% PEG-3350]. Before being cryo-cooled in liquid nitrogen, Gal10mut crystals were cryoprotected by briefly soaking the crystals in mother liquor supplemented with 25% PEG-400. Diffraction experiments at 100 K were conducted on beamlines P14 of Petra III (DESY, Hamburg, Germany) and X06SA (PXI) of the Swiss Light Source (Paul Scherrer Institute, Villigen, Switzerland). All data were integrated and scaled using the XDS suite (61). Molecular replacement (MR) was performed with Phaser using search models based on the available structure of Gal10 (PDB 1LCL). Model (re)building was performed in COOT (62), and individual coordinate and atomic displacement parameter (ADP) refinement was performed in PHENIX (63) and autoBuster (64). Model and map validation tools in COOT and the PHENIX suite were used throughout the workflow to guide improvement and validate the quality of crystallographic models.

Small-angle x-ray scattering in solution (SAXS)

SAXS data were measured on the SWING beamline at Synchrotron SOLEIL (Gif-sur-Yvette, France). Fifty microliters of Gal10mut was injected onto an Agilent 4.6-mm by 300-mm Bio SEC-3 column with a 300-Å pore size and HBS (pH 7.4) as a running buffer at a flow speed of 0.3 ml min−1 at 15°C. X-ray–scattering data were collected in continuous-flow mode with an exposure time of 1 s per frame. Data were recorded within a momentum transfer range of 0.0066 Å−1 < q < 0.609 Å−1, with q = 4πsinθ/λ. Raw data were radially averaged and buffer-subtracted using Foxtrot v3.3.4 as developed at Synchrotron SOLEIL (Gif-sur-Yvette, France) and provided by Xenocs (Sassenage, France). Data quality was analyzed with Foxtrot by checking the stability of the radius of gyration over the length of the elution peak and by scaling all curves to the most intense scattering profile. The final scattering curve was obtained by averaging the unscaled, buffer-subtracted scattering profiles from frames 255 to 268, which corresponded to the top of the elution peak. Structural parameters were determined with the ATSAS suite version 2.8.3 (65). Molecular weight estimates were calculated using DATMW by methods based on the Porod volume (66), the volume of correlation (67), and the apparent volume (68). The theoretical SAXS profile for dimeric Gal10mut was calculated from the determined x-ray structure and fitted to the experimental data using the FoXS server (69). The error-weighted residual difference plot was calculated as Δ/σ = [Iexp(q) − cImod(q)]/σ(q) versus q (70). SAXS analysis revealed that the dimeric assembly obtained by x-ray crystallography corresponds to the in-solution structure.

Analysis of in vivo innate immune responses

WT, Nlrp3−/−, Casp1−/− Casp11−/−, and Tlr4−/− mice on a C57BL/6 background were anesthetized with isoflurane (2 liters/min, 2 to 3%; Abbott Laboratories) and then injected intratracheally with 100 μg of Gal10 crystals or control soluble Gal10mut (in 80 μl of PBS). After 6 and 24 hours, mice were euthanized by CO2 inhalation and lungs were collected. To obtain single-cell suspensions, lungs were first cut with a scissor and then digested at 37°C for 30 min in RPMI-1640 (Thermo Fisher Scientific) containing Liberase (20 μg/ml; Sigma-Aldrich) and deoxyribonuclease (DNase) I (0.01 U/μl; Sigma-Aldrich). The resultant lung cell suspensions were filtered through a 70-μm filter.

For measurements of pro-inflammatory cytokines, mice were euthanized with an overdose of pentobarbital (KELA Laboratoria) and BAL was performed by injecting 1 ml of PBS containing 0.01 mM EDTA (Lonza) once. Subsequently, the BAL fluid was centrifuged (400 × g for 5 min at 4°C) and the supernatant was stored at −20°C. Lungs were snap frozen in liquid nitrogen and homogenized with a TissueLyser II from Qiagen in tissue lysis buffer [40 mM tris-HCL (pH 6.8), 275 mM NaCl, and 20% glycerol (Sigma-Aldrich)]. One tablet of PhosSTOP (Sigma-Aldrich) and one tablet of Complete ULTRA tablets, Mini, Easypack (Sigma-Aldrich) were added per 10 ml of tissue lysis buffer. After homogenization, 2% IGEPAL CA-630 (U.S. Biological) was added. Subsequently, the samples were rotated for 30 min and centrifuged (20,000 × g for 10 min at 4°C). Supernatants were stored at −20°C. The total protein concentration was measured with the NanoOrange protein quantitation kit from Thermo Fisher Scientific according to the manufacturer’s instructions. IL-1β, IL-6, TNF-α, and CCL2 secretion was evaluated using Ready-Set-Go ELISA kits (Thermo Fisher Scientific) according to the manufacturer’s instructions.

The Pierce LDH cytotoxicity assay kit (Thermo Fisher Scientific) was used to evaluate LDH according to the manufacturer’s instructions.

In vivo uptake of fluorescent crystals

WT C57BL/6 mice were injected intratracheally with 100 μg of IAF-Gal10 crystals. After 1 day and 3 days, mice were euthanized by CO2 inhalation, and lungs and mLNs were collected. To obtain single-cell suspensions, lungs and mLNs were first cut with scissors and then digested at 37°C for 30 and 15 min, respectively, in RPMI-1640 (Thermo Fisher Scientific) containing Liberase (20 μg/ml; Sigma-Aldrich) and DNase I (0.01 U/μl; Sigma-Aldrich). The obtained lung cell suspensions were filtered through a 70-μm filter and then pre-enriched for CD11c+ cells using CD11c magnetic beads (Miltenyi Biotec). Subsequently, cells were immunostained with phycoerythrin (PE)–conjugated anti–Siglec F (E50-2440) (BD Biosciences). mLN cells were also pre-enriched for CD11c+ cells using CD11c magnetic beads and subsequently stained with allophycocyanin (APC)–conjugated anti–I-A/I-E (M5/114.15.2) (Thermo Fisher Scientific) prior to confocal analysis.

Adaptive T cell immune response

CD4+ OT-II cells were purified from CD45.1 Rag2−/− OT-II donor mice with the MagniSort mouse CD4 T cell enrichment kit (Thermo Fisher Scientific) and labeled with cell proliferation dye eFluor 450 (0.01 mM; Thermo Fisher Scientific) for 20 min in the dark at RT. Subsequently, 1 × 106 of these labeled cells (in 100 μl of PBS) were injected intravenously into WT C57BL/6 mice. On day 0, mice were injected intratracheally with 10 μg of OVA (Worthington) or with 10 μg of OVA mixed with 100 μg of Gal10 crystals or 100 μg of control soluble Gal10mut (in 80 μl of PBS). On day 3, mice were sacrificed with an overdose of pentobarbital (KELA Laboratoria).

Adaptive humoral immune response

On day 0, WT C57BL/6 mice were sensitized intraperitoneally with 10 μg of OVA (Worthington), 10 μg of OVA mixed with 1 mg of alum (Thermo Fisher Scientific), or 1 mg of Gal10 crystals (in 500 μl of PBS). On day 14, mice were challenged intraperitoneally with 10 μg of OVA (in 500 μl of PBS). In separate experiments, WT, Casp1−/− Casp11 −/−, Nlrp3−/−, Asc−/−, Tlr4−/−, and Cd3e−/− mice were sensitized intraperitoneally with 0.5 mg of Gal10 crystals or 0.5 mg of control soluble Gal10mut (in 500 μl of PBS). On day 21, all mice were sacrificed with an overdose of pentobarbital (KELA Laboratoria). Blood was collected and centrifuged at 9600 × g for 10 min at RT, and the serum was stored at −20 °C.

Serum OVA-specific and Gal10-specific antibodies were quantified by ELISA. A flat-bottom 96-well plate (Greiner) was coated overnight at 4°C with 50 μl per well of 0.1 mg/ml OVA (Worthington) in 0.1 M sodium carbonate buffer (pH 9.5) or with 0.1 mg/ml His-tagged soluble Gal10 in PBS. After washing and blocking for 1 hour, 50 μl per well of the samples diluted in PBS was added in duplicate for 2.5 hours at RT. Then 50 μl per well of working detector [biotinylated anti–mouse IgG1 detecting antibody (0.5 μg/ml; BD Biosciences) plus streptavidin–HRP reagent (1:250; BD Biosciences), both diluted in blocking buffer] was added, followed by a 1-hour incubation at RT. The wells were then washed seven times with wash buffer, and 50 μl per well of 1× TMB substrate solution (Thermo Fisher Scientific) was added. The plate was incubated at RT in the dark. To stop the reaction, 25 μl per well of stop solution (2.5 N H2SO4) was added. Finally, the absorbance was read at 450 nm with a Perkin Elmer multilabel counter, and data were collected with Wallac 1420 Manager software.

Model of type 2 sensitization to OVA and crystals

On days 0 and 1, WT C57BL/6 mice were sensitized intratracheally with 100 μg of OVA (Worthington), 100 μg of OVA mixed with 100 μg of Gal10 crystals, or 100 μg of control soluble Gal10mut (in 80 μl of PBS). On days 11 to 13, mice were challenged daily intranasally with 20 μg of OVA (in 40 μl of PBS). On day 14, mice were sacrificed with an overdose of pentobarbital (KELA Laboratoria), and BAL fluid and blood were collected. BAL was performed by injecting 1 ml of PBS containing 0.01 mM EDTA (Lonza) three times. BAL fluid was centrifuged (400 × g for 5 min at 4°C), and cells were stained for flow cytometry.

Blood was centrifuged at 9600 × g for 10 min at RT, and the serum was stored at −20°C. Serum OVA- and Gal10-specific antibodies were quantified by ELISA as described above for the adaptive humoral immune response.

GM-CSF bone marrow cultures

The femurs and tibias of WT C57BL/6, Tlr4−/−, Nlrp3−/−, and Casp1/11−/− mice were isolated and collected in cold Hanks’ balanced salt solution (HBSS) (Thermo Fisher Scientific). The bones were incubated with 70% ethanol (EtOH), washed twice with cold HBSS, and crushed with a mortar and pestle in HBSS. The cells were then filtered through a 70-μm filter and depleted of red blood cells (RBCs) by RBC lysis buffer [0.15 M NH4Cl (Sigma-Aldrich), 1 mM KHCO3 (Sigma-Aldrich), and 0.1 mM Na2-EDTA (Sigma-Aldrich) in MilliQ H2O]. To obtain granulocyte-macrophage colony-stimulating factor (GM-CSF) DCs and macrophages, bone marrow cells were grown in culture at 2.5 × 105 cells/ml of culture medium [RPMI-1640 (Thermo Fisher Scientific), 5% fetal calf serum (FCS; Bodinco), GlutaMAX supplement (2 mM; Thermo Fisher Scientific), gentamicin (50 μg/ml; Thermo Fisher Scientific), 0.05 μM β-mercaptoethanol, and recombinant mouse GM-CSF (20 ng/ml; Sigma-Aldrich)]. On day 3, fresh culture medium was added, and on day 6, half of the medium was replaced. On day 8, cells were primed using either PBS or 10 ng/ml LPS (InvivoGen) for 16 hours and subsequently exposed to 150 μg/ml Gal10 crystals, 150 μg/ml control soluble Gal10mut, or 150 μg/ml uric acid crystals (30). After 6 and 24 hours, the culture supernatant was harvested and stored at −20°C. IL-1β production was measured using the Ready-Set-Go ELISA kit from Thermo Fisher Scientific, according to the manufacturer’s instructions. The absorbance was read at 450 nm with a Perkin Elmer multilabel counter, and data were collected with Wallac 1420 Manager software.

Flow cytometry

Prior to flow cytometry, all single-cell suspensions were depleted of RBCs by using RBC lysis buffer [0.15 M NH4Cl (Sigma-Aldrich), 1 mM KHCO3 (Sigma-Aldrich), and 0.1 mM Na2-EDTA (Sigma-Aldrich) in MilliQ H2O] produced in-house.

For the analysis of innate immune responses, lung single-cell suspensions were stained for flow cytometry using FITC-conjugated anti-CD3ε (145-2c11) (Tonbo Biosciences), FITC-conjugated anti-CD19 (1D3) (Tonbo Biosciences), FITC-conjugated anti-CD11c (HL3) (BD Biosciences), PE-conjugated anti–Siglec-F (E50-2440) (BD Biosciences), BD Horizon V450–conjugated anti-CD11b (M1/70) (BD Biosciences), Brilliant Violet 605–conjugated anti-CD45 (30-F11) (BD Biosciences), and Alexa Fluor 700–conjugated anti-Ly6G (1A8) (BD Biosciences).

To study DC subsets, mLN single-cell suspensions were stained using PerCP (peridinin-chlorophyll-protein complex)–eFluor 710–conjugated anti-CD172a (P84) (Thermo Fisher Scientific), PE-conjugated anti-XCR1 (ZET) (BioLegend), PE-Cy5–conjugated anti-CD3ε (145-2c11) (Thermo Fisher Scientific), PE-Cy5–conjugated anti-CD19 (1D3) (Thermo Fisher Scientific), PE-Cy7–conjugated anti-CD11c (N418) (Thermo Fisher Scientific), BD Horizon V450–conjugated anti-CD11b (M1/70) (BD Biosciences), Alexa Fluor 647–conjugated anti-CD64 (X54-5/7.1) (BD Biosciences), and APC–eFluor 780–conjugated anti–I-A/I-E (M5/114.15.2) (Thermo Fisher Scientific).

For T cell response analysis, single-cell suspensions of mLNs were stained using FITC-conjugated anti-Vα2 TCR (B20.1) (BD Biosciences), PerCP-Cy5.5–conjugated anti-CD69 (H1.2F3) (BD Biosciences), PE-conjugated anti-CD62L (MEL-14) (BD Biosciences), PE-Cy5–conjugated anti-CD3ε (145-2c11) (Thermo Fisher Scientific), PE-Cy7–conjugated CD44 (IM7) (Thermo Fisher Scientific), Brilliant Violet 605–conjugated anti-CD45.1 (A20) (BioLegend), APC-conjugated anti-CD4 (RM4-5) (Thermo Fisher Scientific), and Alexa Fluor 700–conjugated anti-CD45.2 (104) (Thermo Fisher Scientific).

BAL cells were stained with PE-conjugated anti–Siglec F (E50-2440) (BD Biosciences), PE-Cy5–conjugated anti-CD3ε (145-2c11) (Thermo Fisher Scientific), PE-Cy5–conjugated anti-CD19 (1D3) (Thermo Fisher Scientific), PE-Cy7–conjugated anti-CD11c (N418) (Thermo Fisher Scientific), BD Horizon V450–conjugated anti-CD11b (M1/70) (BD Biosciences), Alexa Fluor 700–conjugated anti-Ly6G (1A8) (BD Biosciences), and APC–eFluor 780–conjugated anti-MHCII (M5/114.15.2) (Thermo Fisher Scientific).

Details of all clones and concentrations used are listed in table S3. For all flow experiments, the viability of cells was discriminated by using eBioscience fixable viability dye eFluor 506 (Thermo Fisher Scientific). Fc Block 2.4.G2 (4.3 μg/ml; Bioceros) was used to block nonspecific antibody binding. Cell surface markers were stained for 30 min at 4°C in the dark. 123count eBeads counting beads (Thermo Fisher Scientific) were added to each sample. Settings were calibrated using UltraComp eBead compensation beads (Thermo Fisher Scientific). Data were collected on an LSRFortessa (BD) and were analyzed with FlowJo software (Tree Star).

Generation of anti-Gal10 crystal-dissolving antibodies

Two llamas (Lama glama) were immunized intramuscularly with Gal10 crystals (1 mg per dose per llama) in the absence of adjuvant on days 0, 14, and 28. Five days after the last immunization, PBMCs were collected for RNA extraction. The antibody response to Gal10 was measured by ELISA. For this purpose, a MaxiSorp plate (Thermo Fisher Scientific) was coated with His-tagged Gal10 (100 μg/ml) and blocked with casein (PBS–1% casein) before the addition of serial dilutions of llama serum pre- and postimmunization. Next, llama IgG1 bound to coated Gal10 was detected with a mouse antibody specific for llama CH1 (domain 1 of the constant portion of the Ig heavy chain) (10D12) and anti–mouse IgG-HRP (DAMPO) (Jackson ImmunoResearch). Lastly, after the addition of TMB (CHEM LAB), the reaction was stopped with 0.5 M H2SO4 (CHEM LAB) and the absorbance was measured at 450 nm (Tecan Sunrise, Magellan software). Both immunized llamas showed a strong immune response against Gal10, even though only three injections were performed.

Total mRNA was purified from the PBMCs isolated from the blood of immunized llamas and reverse-transcribed with random hexamer primers to obtain cDNA. To construct heavy- and light-chain libraries, a two-step polymerase chain reaction (PCR) was performed, using primers listed in table S4. First, nontagged primers were used directly on the cDNA to amplify the VH-CH1, Vλ-Cλ, and Vκ-Cκ regions (where VH, Vλ, and Vκ are the variable regions of the Ig heavy, λ, and κ chains, respectively, and Cλ and Cκ are the constant portions of the λ and κ chains, respectively). The PCR products were then purified and used in a second PCR with the tagged scFv primers to amplify VH, Vλ, and Vκ, which were cloned separately into the phagemid vector to create the “lambda” and “kappa” llama scFv libraries. The scFv fusion protein consisted of the VH and VL (variable region of the Ig light chain) sequences coupled by a (Gly4Ser)3 linker for a size of ~25 kDa. Phages expressing specific Gal10 scFv fragments were enriched by three rounds of selection on immobilized Gal10. Briefly, His-tagged Gal10 was immobilized on MaxiSorp ELISA plates (0.02 to 10 μg/ml), the plates were blocked with PBS–1% casein, and then the scFv phage library (input) was added. Unbound phages were removed via multiple washing steps. Finally, the bound phages were eluted with trypsin, and E. coli infection was performed to amplify the selected phages. This process resulted in the enrichment of the phage population expressing scFv with high-affinity anti-Gal10. The first round of λ and κ library selection from both llamas resulted in a minor enrichment of specific anti-Gal10 phages. The second and third rounds of selection resulted in an enrichment of phages expressing scFv with a likely higher affinity for Gal10. Production of the scFv by single clones was induced by an overnight incubation with IPTG, inducing secretion in the periplasmic space. The next day, bacteria were lysed by two freeze-thaw cycles at −80° and −20°C. After centrifugation, the supernatant (periplasmic extract) was collected and transferred into separate 96-well plates to be tested for binding capacity [by ELISA and biolayer interferometry (BLI)]. Briefly, a MaxiSorp plate (Thermo Fisher Scientific) was coated with His-tagged Gal10 (1 μg/ml) and then blocked with casein before being incubated with the periplasmic extract (diluted 1:5 in PBS) containing the Myc-tagged scFv. Detection of the binders was carried out with an anti-Myc–HRP antibody (Bethyl). Absorbance was measured at 450 nm (reference at 620 nm) with a Tecan instrument. The selected clones were also screened by BLI using the Octet RED96. For this purpose, His-tagged Gal10 was diluted in kinetic buffer (0.1% BSA and 0.002% Tween 20 in PBS) and captured on anti–penta-His (HIS1K) tips (ForteBio) until a capturing level of 1 nm was reached (loading step, 20 s). Diluted periplasmic extracts were then applied for 120 s (association), followed by 120 s of dissociation. Tips were regenerated by three 5-s washing steps in 0.5 M H2SO4. In order to assess the binding specificity of the periplasmic extract for the sensor tips, reference anti–penta-His (HIS1K) tips were used. Kinetic parameters were determined by fitting the binding of the scFv (binding to reference tips was subtracted) on captured His-tagged Gal10 with a 1.1 binding model using the ForteBio data analysis 9.0 software.

Selected scFv clones that showed binding to Gal10 were sequenced. Based on the complementarity-determining regions (CDRs) of the VH and VL sequences, each clone was classified as belonging to a particular family. This process resulted in the identification of 65 VH families, 13 Vκ families, and 23 Vλ families. Twelve clones were selected for further characterization. The 12 scFv clones were recloned as scFv–human Fc fusion molecules. For this purpose, the DNA of each selected scFv clone was cloned into a vector containing the CH2–CH3 constant domains of human IgG1. In addition, some selected clones were recloned into a mouse IgG1 backbone for further characterization. For this purpose, the VH and the VL of each clone were PCR amplified by using specific primers, isolated by electrophoresis, purified, and digested with restriction enzymes (BsmBI). After digestion and cleanup, ligation of the DNA (for VH or VL) was performed into BsmBI-predigested vectors containing the constant domains of the mouse λ or κ light chain or of the mouse IgG1 heavy chain (CH1-CH2-CH3).

Production of anti-Gal10 antibodies

Human embryonic kidney (HEK) 293E cells were transfected with the plasmid DNA encoding the scFv–human Fc or mouse IgG1 clones via polyethylenimine (PEI) (Polysciences). A total of 10 different mouse IgG1 antibodies and 12 different scFv–human molecules were produced in mammalian cells. For mouse IgG1 antibody transfections, a ratio of 1 μg of heavy chain to 3 μg of light chains was used. After 6 days of production, anti-Gal10 molecules (scFv–human Fc or mouse IgG1) were purified from the cell supernatant by using the protein-A Sepharose beads (GE Healthcare). Finally, SDS–polyacrylamide gel electrophoresis (PAGE) analysis was carried out to assess the purity and the integrity of the anti-Gal10 molecules (~100 kDa for the scFv–human Fc and ~150 kDa for the murine IgG1).

The scFv–human Fc or mouse IgG1 antibodies were characterized by ELISA and SPR (Biacore T3000). For the ELISA, 96-well MaxiSorp plates (Thermo Fisher Scientific) were coated with 0.2 μg/ml of His-tagged Gal10. A serial dilution of the scFv–human Fc fusion molecules or mouse IgG1 antibodies (from 100 nM; dilution, 1:5; eight serial dilutions) was applied before detection by a peroxidase donkey anti–mouse IgG (DAMPO) (0.16 μg/ml; Jackson ImmunoResearch) or peroxidase goat anti–human IgG (0.16 μg/ml; Jackson ImmunoResearch) antibody. The Gal10 mouse IgG1 antibodies showed a relative binding capacity ranking from 3.22 up to 0.04 nM against coated Gal10, whereas the panel of scFv–human Fc fusion molecules showed a relative binding capacity between 0.48 and 0.02 nM on the coated target.

The kinetics for the binding of scFv–human Fc or mouse IgG1 antibodies to Gal10 were measured in HBS-EP buffer by using a Biacore T3000 instrument (GE Healthcare). For this purpose, a goat anti–mouse IgG Fc (Jackson ImmunoResearch) or goat anti–human IgG Fc (Jackson ImmunoResearch) was immobilized onto a CM5 sensor chip (8000 RU) by using an amine-coupling kit (GE Healthcare) according to the manufacturer’s instructions. The antibodies were then captured (150 to 250 RU), and a serial dilution of His-tagged Gal10 (twofold serial dilution beginning at 5 μg/ml) was applied. The surface was regenerated using 10 mM glycine, pH 1.5 (GE Healthcare). Data were analyzed using the BIAevaluation software (GE Healthcare, version 4.1.1) with a blank subtraction. Kinetic parameters were determined by fitting the sensorgrams with a mass transport effect (Fit Kinetics simultaneous ka/kd). From the panel of scFv–human Fc molecules, clones 6F05 and 2C07 clearly showed the best affinity (0.9 and 0.6 nM, respectively), with off rates of 1.17 × 10−4 and 2.59 × 10−4 s−1, respectively, whereas the rest of the panel showed a fast off rate (>17 × 10−4 s−1) with affinities in a subnanomolar up to nanomolar range. In opposition, the panel of mouse IgG1 showed an affinity in a nanomolar up to subnanomolar range, with an off rate ranging between 3.4 × 10−4 and 53 × 10−4 s−1.

Inhibition of CLC formation and time-lapse solubilization of CLCs

The potency of Gal10 antibodies to inhibit Gal10 autocrystallization and to solubilize recombinant CLCs was initially evaluated using a Mosquito crystallization robot (TTP Labtech). To evaluate the inhibition of crystal formation, 250 nl of soluble WT recombinant Gal10 cleaved with TEV protease in PBS at a concentration of 0.4 to 0.7 mg/ml was mixed with 100 nl of anti-Gal10 antibody or an irrelevant antibody. The concentrations of the different antibodies are listed in the legend for fig. S4. The protein mixture was equilibrated against 40 μl of 50% (v/v) PEG-3350 contained in the reservoir well of a 96-well crystallization plate. After overnight incubation, the presence or absence of CLCs was evaluated by using an SZX16 Olympus (with SDF PLAPO 1× PF objective) or Leica M1615C (with PLAN 0.8× LWD objective) light stereo zoom microscope. The latter microscope was equipped with a Leica IC80 HD camera. To evaluate the potency of antibodies to solubilize recombinant CLCs, crystals were formed overnight by equilibrating 250 nl of soluble WT recombinant Gal10 cleaved with TEV protease in PBS against 50% (w/v) PEG-3350, as described above. The next day, 100 nl of anti-Gal10 or control antibody was added and the solubilization of Gal10 crystals was observed over time and imaged by using the Leica stereomicroscope described above.

To better document and characterize the process of solubilization of CLCs by anti-Gal10 antibodies, time-lapse experiments were conducted on a spinning-disk confocal microscope (details below). To this end, 2.5 μl of autocrystallized CLC solution (at 0.7 mg/ml) in PBS was spotted and imaged either on a 15-well μ-Slide angiogenesis coverslip or in a μ-Slide two-well coculture chamber (81506 and 81806, respectively; both from iBidi). Crystal solubilization was induced just before imaging, by the addition of 2 μl of anti-Gal10 antibodies at a concentration of 7 mg/ml, at which point the chambers were sealed with vacuum grease and a #1 coverslip (Menzel-Gläser). Samples were imaged on an Axio Observer.Z1 (Zeiss) equipped with a CSU-X1 Yokogawa spinning-disk head (Yokogawa Corporation) and a Zeiss AxioCam Mrm (Zeiss), with either an EC Plan-Neofluar 10× (NA 0.30) or a Plan ApoChromat 20× dry objective (NA 0.80, DIC). Images were acquired at 5-min intervals for up to 3 hours. Data analysis and image reconstruction were performed with ImageJ (NIH).

Time-lapse experiments for the antibody-mediated solubilization of ex vivo CLCs in patient mucus were conducted as follows. Mucus (4 μl) containing CLCs was spotted and imaged on a 15-well μ-Slide angiogenesis coverslip as described above. To induce CLC solubilization, 4 μl of anti-Gal10 or control antibody at 7 mg/ml was added to the mucus. Images were acquired every 10 min for up to 16 hours, using an identical spinning-disk microscope setup as described above. Data analysis and image reconstruction were performed with ImageJ (NIH).

Crystal structure of Gal10-Fab fragments engaging Gal10 target

Four clones of interest (6F05, 1C09, 1D11, and 4E08) were reformatted as Fab fragments. As a first step, the VH and the VL of each clone were PCR amplified using specific primers, purified by electrophoresis, digested with restriction enzymes (BsmBI), and ligated in the predigested vectors containing the human constant domains: the human λ constant domain for the VL or the CH1 constant domain for the VH (including part of the hinge region). Each ligated product was transformed into Top10 bacteria by heat shock. Bacteria were then transferred to agarose plates with ampicillin (vector resistance gene). Four to eight colonies per clone (VH and VL) were picked, sequenced, and amplified in order to purify the DNA sequence (MidiPrep). The vector was transfected into HEK293E cells by using a ratio of 1 μg of heavy-chain DNA to 1 μg of light-chain DNA via PEI. After 10 days of production, human Fab fragments were purified using the Capture Select IgG-CH1 Sepharose beads. Finally, SDS-PAGE analysis was carried out to assess the conformation, purity, and integrity of the Fab molecules (55 kDa). The binding properties of the Fab format were confirmed by ELISA and BLI binding, as described above, with median effective concentrations (EC50 values) between 1.6 and 26.7 nM.

Recombinant His-tagged Gal10 at 1 mg/ml was digested with TEV protease at RT overnight using a TEV protease:Gal10 ratio of 1:100 (w:w). Next, purified Fab was added to digested Gal10 in a 1.25 molar excess. The protein mixture was then injected onto a HiLoad 16/600 Superdex 200-pg column running on HBS buffer to isolate the Gal10-Fab complex. Fractions corresponding to the Gal10-Fab complex were pooled and stored at −80°C until further use. Gal10-Fab complexes were concentrated to 6 to 7 mg/ml before crystallization experiments. Sitting-drop nanoliter-scale vapor diffusion crystallization experiments were performed at 293 K by using a Mosquito crystallization robot (TTP Labtech) and commercially available sparse-matrix screens (Molecular Dimensions, Hampton Research). Crystals of Gal10 complexed with Fab 1D11 grew overnight in condition B7 of the ProPlex screen (Molecular Dimensions) [0.2 M ammonium acetate, 0.1 M sodium acetate (pH 4.0), and 15% PEG-4000]. Gal10 complexed with Fab 6F5 crystallized within 24 hours in condition G7 of the BCS Eco screen (Molecular Dimensions) [0.04 M CaCl2, 0.04 M Na-formate, 0.1 M PIPES (pH 7.0), and 8% PEG Smear High]. Crystals of Gal10 in complex with Fab 4E8 appeared after 2 weeks in condition B7 of the PEG/Ion screen (Hampton Research) (0.2 M ammonium nitrate and 20% PEG-3350). Before cryo-cooling into liquid nitrogen, crystals of the Gal10:Fab complexes were cryoprotected by briefly soaking the crystals in mother liquor supplement with 25% PEG-400. Diffraction experiments at 100 K were conducted on beamlines Proxima 2A of the SOLEIL synchrotron (Gif-sur-Yvette, France) and ID23-2 of the ESRF (Grenoble, France). All data were integrated and scaled using the XDS suite (71). MR was performed with Phaser (72) by using search models based on the structure of Gal10 (PDB 1LCL) and a high-resolution mouse Fab structure (PDB 5X4G). Model (re)building was performed in COOT (62), and individual coordinate and ADP refinement was performed in PHENIX (63) and autoBuster (64). Model and map validation tools in COOT and the PHENIX suite were used throughout the workflow to guide improvement and validate the quality of crystallographic models.

Analysis of humanized SCID mice

On day 0, immunodeficient NRG mice were reconstituted by intraperitoneal injection of 3 × 106 PBMCs obtained from an asthmatic patient allergic to HDMs. On days 1 to 4 and 7 to 9, all mice were injected intratracheally with 20 μg of HDM extract (Greer) diluted in 50 μl of PBS. In experiments addressing the pro-inflammatory effects of Gal10 crystals, on days 1, 3, 7, and 9, NRG mice were treated according to the following regimens: regimen 1, 200 μg of isotype control antibodies intratracheally (diluted in 30 μl of PBS); regimen 2, 10 μg of recombinant Gal10 crystals (1 μl of the stock) plus 200 μg of isotype control antibody intratracheally (diluted in 30 μl of PBS); and regimen 3, 10 μg of recombinant Gal10 crystals (1 μl of the stock) plus 200 μg of 1D11 antibody intratracheally (diluted in 30 μl of PBS). From day 11 onward, mice received intratracheal injections of isotype antibody (200 μg) or 1D11 antibody (200 μg) three times per week until the mice were sacrificed. On day 27, all mice were challenged one final time intratracheally with 20 μg of HDM extract (Greer, batch no. 002525848) diluted in 80 μl of PBS. All mice were sacrificed on day 28.

The upper and lower lobes of the right lung were fixed in 4% PFA overnight before being embedded in paraffin for histology. The middle lobe of the right lung was embedded in OCT and frozen at −80°C until further use (for quantitative reverse transcription PCR and immunofluorescence). To detect human cells, single-cell suspensions from the left lungs of mice were incubated for 20 min at 4°C with APC‐conjugated anti–human CD45 (HI30) (BD Biosciences). Dead cells were stained using the Aqua Live/Dead fixable dead cell stain kit (BD Biosciences). After cells were washed in PBS, 123count eBeads counting beads (Thermo Fisher Scientific) were added to each sample. Data were collected on a BD LSRFortessa and were analyzed with FlowJo software (Tree Star). None of the antibodies used cross‐reacted with murine tissues.

Human IgE concentrations were measured in the serum of NRG mice by using a human IgE uncoated ELISA kit (Thermo Fischer Scientific) according to the manufacturer’s instructions.

The presence of inflammation was assessed on paraffin-embedded lung slices that were stained with hematoxylin and eosin (H&E). We employed an investigator-blinded morphometric measurement of the number of inflammatory cells around the airways, in a 500-μm-perimeter area taken from the basement membrane of each visible airway in the slide. QuPath software (made available online by Queen’s University Belfast) (73) was used to analyze the number of nuclei per area, which was expressed as the number of cells per 10 μm of basement membrane. Goblet-cell metaplasia was visualized by PAS staining on paraffin-embedded slices. It was quantified by using mucin MUC5AC mRNA levels on snap-frozen middle-lobe samples. For this, frozen lung tissue was collected in a 1.5-ml microcentrifuge tube and 1 ml of TriPure (Sigma-Aldrich) was added. Tissue was homogenized by using a tissue homogenizer. To extract RNA, 200 μl of chloroform was added to the tubes containing the homogenized lung. After an incubation period of 5 min, tubes were centrifuged at 12,000 × g for 15 min. The upper transparent phase was collected in a ribonuclease (RNase)–free microcentrifuge tube and was mixed with 500 μl of isopropanol and 1 μl of glycogen for 10 min. The tubes were centrifuged at 12,000 × g for 5 min. The supernatant was discarded, and the pellet containing the purified RNA was washed in 75% EtOH (centrifugation at 7500 × g for 5 min). The pellet was resuspended in 20 μl of RNase-free water. The tubes were placed for 10 min at 60°C. RNA concentrations for each sample were determined by using a Nanodrop instrument. One microgram of RNA was used to make cDNA using the Sensifast cDNA synthesis kit (Bioline). The leftover RNA was frozen at −80°C. The cDNA was diluted 10 times in water and frozen until further use. For real-time PCR, the following mastermix was used for each well of the PCR plate: 10 μl of Sensifast SYBR No-Rox mix, 4.75 μl of water, 5 μl of cDNA, 0.125 μl of forward primer, and 0.125 μl of reverse primer (taken from a 100 μM stock). Primers used were as follows: murine Muc5ac, forward (Fwd), 5′-CTCCGTCTTAGTCAATAACCACC-3′ (SEQ ID no. 144), and reverse (Rev), 5′-GTCAGGTTTTAGGTTGCTCAAGG-3′ (SEQ ID no. 145); murine Gapdh as a housekeeping gene, Fwd, 5′-ACAAAATGGTGAAGGTCGGTG-3′ (SEQ ID no. 146), and Rev, 5′-CGTTTCACCTCTAACAACGGT-3′ (SEQ ID no. 147).

Lung function measurements

Measurement of lung function was performed using the FlexiVent invasive measurement of dynamic resistance, as described (74). In brief, mice were anesthetized with urethane, paralyzed using d-tubocurarine, tracheotomized, and intubated with an 18G catheter, followed by mechanical ventilation by a Flexivent apparatus. Increasing concentrations of methacholine (0 to 200 μg/ml) were aerosolized via the catheter. Dynamic resistance (Rrs) was recorded after a standardized inhalation maneuver given every 10 s for 2 min after methacholine administration.

Statistical analysis

Standard statistical analyses were performed by using GraphPad Prism 7 with default parameters. The significance of differences between groups was evaluated by using the Kruskal-Wallis H test or one-, two-, or three-way analysis of variance (ANOVA). For pairwise comparisons on nonparametric data sets, the Mann-Whitney U test was used. For three-way ANOVA comparing groups with unequal numbers of data points, an unbalanced ANOVA was performed by using regression facilities in Genstat v19, followed by multiple-comparison tests using Fisher’s protected least significant difference (LSD) test. A P value of <0.05 was considered statistically significant and was represented by *, a P value of <0.01 by **, a P value of <0.001 by ***, and a P value of <0.0001 by ****.

Supplementary Materials

References and Notes

Acknowledgments: We acknowledge the help of J. Haustraete of the VIB Protein Synthesis Facility, G. Van Isterdael of the VIB Flow Cytometry Core Facility, and K. Lemeire of VIB. We thank S. Goriely for his help with experiments in Cd3e−/− mice and K. S. Ravichandran for critical reading of the manuscript. Funding: This work was supported by Fonds Wetenschappelijk Onderzoek Vlaanderen (FWO Flanders) under EOS project GOG2318N and grant FWO/OPR2017004401 (to H.H. and B.N.L.) and by a Concerted Research Action (GOA) grant from Ghent University (to B.N.L., H.H., and S.N.S.) and the VIB. E.K.P. was supported by a grant from the Swedish Research Council (2014-6852). K.V. was supported by a postdoctoral research fellowship from the FWO. I.H. was supported by a predoctoral grant from Ghent University (BOF/DOC2017005501). C.B. and E.G. were supported by grants from FWO Flanders (1515516N and FWO/PDO/108) and the Interuniversity Attraction Poles grant P7/30. B.N.L. is supported by an ERC Advanced grant. Author contributions: E.K.P. and I.H. performed in vivo and in vitro experiments; K.V. produced recombinant proteins, performed structural studies, and carried out antibody dissolution experiments; H.H. and H.A. performed in vivo experiments in mice; K.D. assisted with in vivo experiments; E.G. and D.G. performed studies on CLCs in patients; P.C. and C.B. recruited patients; J.-M.P., H.D.H., C.B., and M.S. produced and characterized the llama antibodies; K.H.G.V. and A.D. performed key experiments regarding protein production and crystal structure elucidation; A.G. performed imaging of crystals and time-lapse crystal dissolution; H.V.G. performed inflammasome experiments; and S.N.S. and B.N.L. conceived and supervised the study, collaborated with clinicians, analyzed data, and wrote the manuscript with contributions from all authors. Competing interests: The authors declare the following financial competing interests. J.-M.P., H.D.H, C.B., E.K.P., and M.S. are full-time employees of arGEN-X. Patents pertaining to the results presented in the paper have been filed under the Patent Cooperation Treaty (UK Intellectual Property Office application 1806099.6) with arGEN-X, VIB, and Ghent University as applicants. Data and materials availability: All reported structural coordinates and related crystallographic data have been deposited in the Protein Data Bank with access codes 6GKQ (ex vivo CLC), 6GKS (recombinant human Gal10), 6QRN (recombinant human Gal10-ribose), 6GKT (recombinant human Gal10mut), 6GLW (Gal10–Fab 1D11), 6GLX (Gal10–Fab 4E8), and 6GKU (Gal10–Fab 6F5). SAXS data and the corresponding model for recombinant Gal10mut have been deposited in the Small Angle Scattering Biological Data Bank with access code SASDF86. Reasonable requests for materials described in this paper will be possible under a material transfer agreement with VIB and arGEN-X.
View Abstract

Stay Connected to Science

Subjects

Navigate This Article