Vascular E-selectin Expression Detected in Formalin-fixed, Paraffin-embedded Sections With an E-selectin Monoclonal Antibody Correlates With Ulcerative Colitis Activity

Abstract

It is widely accepted that E-selectin, an inducible endothelial cell adhesion molecule, plays a critical role in the initial step of neutrophil recruitment to sites of acute inflammation. However, immunohistological analysis of E-selectin has been hampered by lack of E-selectin-specific monoclonal antibodies that can stain formalin-fixed, paraffin-embedded (FFPE) tissue sections. Here, we employed E-selectin•IgM (a soluble form of E-selectin) as immunogen, and then, after negative selection with L-selectin•IgM and P-selectin•IgM and screening of FFPE sections of both COS-1 cells overexpressing E-selectin and acute appendicitis tissues, we successfully generated an E-selectin-specific monoclonal antibody capable of staining FFPE tissue sections. We used this antibody, designated U12-12, to perform quantitative immunohistological analysis of 390 colonic mucosal biopsy specimens representing ulcerative colitis. We found that the higher the histological disease activity, the greater the number of vessels expressing E-selectin, an observation consistent with previous analyses of frozen tissue sections. Furthermore, in active ulcerative colitis, E-selectin-expressing vessels contained neutrophils attached to endothelial cells, presumably in the process of extravasation, which eventually could cause epithelial damage. These results overall indicate that U12-12 is effective for E-selectin immunohistochemistry in archived FFPE samples representing various human diseases:

Introduction

Alterations in the endothelial cell surface in response to pro-inflammatory cytokines are of great importance in regulating leukocyte recruitment and thereby to progression to inflammatory states. ¹ Such alterations include expression of endothelial cell adhesion molecules that interact with counter-receptors on leukocytes. Among them, E-selectin is particularly important for neutrophil recruitment to sites of acute inflammation. E-selectin is not expressed in resting endothelial cells but is transcriptionally induced by interleukin 1β and tumor necrosis factor α secreted from macrophages activated by inflammatory stimuli such as bacterial lipopolysaccharide.^2,3

E-selectin is a type 1 transmembrane protein consisting of an N-terminal calcium-dependent (C-type) lectin domain [also called the carbohydrate recognition domain (CRD)], an epidermal growth factor (EGF)-like domain, six sequential short consensus repeats (also called Sushi) domains, a transmembrane domain, and a short C-terminal cytoplasmic tail (Fig. 1). ⁴ Through its CRD, E-selectin selectively, but relatively weakly, binds to its carbohydrate ligand sialyl Lewis x (sLe^x), Siaα2→3Galβ1→4(Fucα1→3)GlcNAcβ1→R (in which Sia is sialic acid, Gal is galactose, Fuc is fucose, and GlcNAc is N-acetylglucosamine), expressed on neutrophils, which leads to “rolling,” the initial step of leukocyte recruitment. ⁵ Under pathologic conditions, E-selectin also binds to sialyl Lewis a (sLe^a, a structural isomer of sLe^x also known as CA19-9), Siaα2→3Galβ1→3(Fucα1→4)GlcNAcβ1→R, expressed on circulating tumor cells and can contribute to formation of hematogenous metastasis. ⁶ Taking advantage of this sLe^x/sLe^a-specific carbohydrate-binding property of E-selectin, we previously used an E-selectin•IgM chimera, which is a soluble form of E-selectin, to functionally probe sLe^x and/or sLe^a in formalin-fixed, paraffin-embedded (FFPE) tissue sections of various human diseases, including chronic Helicobacter pylori gastritis, ⁷ ulcerative colitis, ⁸ testicular seminoma, ⁹ bladder urothelial carcinoma, ¹⁰ papillary thyroid carcinoma, ¹¹ clear cell renal cell carcinoma, ¹² and most recently, angioimmunoblastic T-cell lymphoma. ¹³

Figure 1.

Schematic representation of the structure of E-selectin, L-selectin, and P-selectin. Each consists of five domains: (1) an N-terminal C-type lectin domain (also called the carbohydrate recognition domain [CRD]), (2) an epidermal growth factor (EGF)-like domain, (3) a set number of sequential short consensus repeats (SCRs, also called Sushi) domains, (4) a transmembrane (TM) domain, and (5) a C-terminal short cytoplasmic tail (CT). Note that both E-selectin and P-selectin are expressed on endothelial cells.

Immunohistological analysis of the E-selectin molecule itself, however, has been hampered by lack of E-selectin-specific monoclonal antibodies capable of staining FFPE tissue sections; thus researchers had no choice but to use frozen tissue sections. In the case of ulcerative colitis, immunohistological analysis of E-selectin has been conducted with a limited number of frozen tissue sections using monoclonal antibodies such as BB11 (mouse IgG_2b),^14,15 BBIG-E6 (mouse IgG₁; R&D Systems, Minneapolis, MN), ¹⁶ and 1.2B6 (mouse IgG₁; Bio-Rad, Hercules, CA). ¹⁷ Unfortunately, BBIG-E6 and 1.2B6 were subsequently found to cross-react with P-selectin. ⁴ Another monoclonal antibody 16G4 (mouse IgG₁; Leica Biosystems, Nussloch, Germany) is reportedly usable on FFPE tissue sections; however, it was also subsequently reported by the vendor to cross-react with P-selectin. ³ Because both E- and P-selectin are expressed on vascular endothelial cells (see Fig. 1), it is currently not possible to accurately assess E-selectin expression using these antibodies. Therefore, to conduct immunohistological analysis of E-selectin using archived FFPE specimens of human disease, a true E-selectin-specific monoclonal antibody capable of staining FFPE tissue sections is required. FFPE tissue sections are much more advantageous than frozen tissue sections, not only in terms of availability but also in terms of preservation of morphology.

Here, employing our above-described E-selectin•IgM chimeric protein as an immunogen, we performed negative selection with L-selectin•IgM and P-selectin•IgM and then screened using FFPE sections of either COS-1 cells overexpressing E-selectin or those of acute appendicitis tissues. That analysis resulted in an E-selectin-specific monoclonal antibody, designated U12-12, capable of staining FFPE tissue sections. Then, using U12-12 to perform quantitative immunohistological analysis of 390 colonic mucosal biopsy specimens of ulcerative colitis, we observed that the higher the histological disease activity, the greater the number of vessels expressing E-selectin, consistent with previous reports using frozen tissue sections.^15,16 Thus, U12-12 can be used for immunohistological analysis of E-selectin in various human diseases employing archived FFPE sections.

Materials and Methods

Construction of Selectin Expression Vectors

DNA fragments encoding human E-selectin (amino acid residues 1-610) ¹⁸ were PCR-amplified using PrimeSTAR Max DNA polymerase (Takara Bio; Kusatsu, Japan) with the oligonucleotides 5’-TCgCTAgCTCTTgAAgTCATgATTgCTTCACA-3’ (5’-primer; NheI site underlined) and 5’-TgCTCgAgTTAAAggATgTAAgAAggCTTTTg-3’ (3’-primer; XhoI site underlined). Human lung first-strand cDNA (Clontech Laboratories; Mountain View, CA) served as template. After double NheI/XhoI digestion, DNA fragments of interest were subcloned into corresponding sites of pcDNA3.1/Hygro (Thermo Fisher Scientific; Waltham, MA), resulting in pcDNA3.1/Hygro-E-selectin.

DNA fragments encoding human P-selectin ¹⁹ were also PCR-amplified with the oligonucleotides 5’-TACCgAgCTCggATCCCACAgAggAgATggCCAA-3’ (5’-primer; BamHI site underlined) and 5’-gATATCTgCAgAATTCTCATTTATCgTCgTCATCCTTATAATCAggACTCgggTCAAATgCAgC-3’ (3’-primer; EcoRI site and FLAG sequence underlined). PCR products were subcloned into pcDNA3.1 using In-Fusion Snap Assembly Master Mix (Takara Bio) according to the manufacturer’s protocol, resulting in pcDNA3.1-P-selectin•FLAG.

Construction of expression vectors encoding three human selectins fused to human IgM, namely pcDNA1.1-E-selectin•IgM, pcDNA1.1-L-selectin•IgM, and pcDNA1.1-P-selectin•IgM, was described previously.^{7, 20} To construct an expression vector of human E-selectin fused to human IgG, DNA fragments encoding the Fc region of IgG were excised from pcDNA3-MAdCAM-1•IgG ²¹ by double BamHI/XbaI digestion and inserted into corresponding sites of pcDNA3.1/Hygro (pcDNA3.1/Hygro-IgG). DNA fragments encoding amino acid residues 1-300 of human E-selectin were PCR-amplified by replacing the 3’-primer above with the oligonucleotide 5’-gAggATCCTTACACgTTggCTTCTCgTTgTCC-3’ (BamHI site underlined). Amplicons were NheI/BamHI double-digested and subcloned into corresponding sites of pcDNA3.1/Hygro-IgG, resulting in pcDNA3.1/Hygro-E-selectin•IgG.

Expression vectors encoding a series of domain deletion mutants of E-selectin•IgM were constructed by inverse PCR using the KOD -Plus- Mutagenesis Kit (Toyobo; Osaka, Japan) with the following oligonucleotides: ΔCRD, 5’-ACAgCTgCCTgTACCAATACATCCTgCAgTgg-3’ (5’-primer) and 5’-ggCTCCACTCTCTTTAATgAgAAgCACCAAAgTgAgAg-3’ (3’-primer); ΔEGF, 5’-CAAATTgTgAACTgTACAgCCCTggAATCCCCTg-3’ (5’-primer) and 5’-gTAgCATAgggCAAgCTTCTTCTTgCTgCACCTCTC-3’ (3’-primer); ΔSushi 1, 5’-gTTgAgTgTgATgCTgTgACAAATCCAgCC-3’ (5’-primer) and 5’-CTCACACTTgAgTCCACTgAAgCCAgggTC-3’ (3’-primer); and ΔSushi 2, 5’-AAggATCCTgTgATTgCTgAgCTgCCTCCC-3’ (5’-primer) and 5’-CACATTgCAggCTggAATAggAgCACTCCATTC-3’ (3’-primer). To construct a soluble Sushi 2 domain fused to IgM, which does not contain the CRD, the EGF-like domain, or the Sushi 1 domain, the 5’-primer for ΔSushi 1 and the 3’-primer for ΔCRD described above were used.

Production of Anti-E-selectin Monoclonal Antibodies

COS-1 cells were transfected with pcDNA1.1-E-selectin•IgM using Lipofectamine Plus (Thermo Fisher Scientific), and 72 hr later, conditioned medium was recovered and concentrated ~50-fold using Amicon Ultra Centrifugal Filters (nominal molecular weight limit 100 kDa; Millipore, Billerica, MA). The protein solution was mixed with complete Freund’s adjuvant to create an emulsion used to inject into the footpad of both legs of 9-week-old female Wistar Kyoto rats, which had been anesthetized with isoflurane. Rats were immunized in the same manner two more times at 2-week intervals. One week after the third immunization, rats were sacrificed and lymphocytes were collected from inguinal and iliac lymph nodes. Lymphocytes were mixed with SP2/0 mouse myeloma cells and subjected to electrofusion using an ECFG21 Super Electro Cell Fusion Generator (Nepa Gene; Chiba, Japan), as described. ²² Fused cells were suspended in RPMI 1640 containing 8% fetal bovine serum, 10% BM Condimed H1 Hybridoma Cloning Supplement (Roche Diagnostics; Mannheim, Germany), and 55-µM 2-mercaptoethanol, and incubated in a CO₂ incubator at 37C overnight. After adding HAT Media Supplement (Sigma-Aldrich; St. Louis, MO), the cell suspension was dispensed into 96-well plates and cultured until surviving cells formed colonies. Culture media were screened by enzyme-linked immunosorbent assay (ELISA) using E-selectin•IgG as the coating antigen and horseradish peroxidase (HRP)-conjugated anti-rat IgG (Jackson ImmunoResearch; West Grove, PA) as the secondary antibody. Culture media strongly positive for the antigen were screened by another ELISA using a cocktail of L-selectin•IgM and P-selectin•IgM as the coating antigen. Culture media negative for these antigens were further screened by immunostaining of FFPE sections of COS-1 cells transiently expressing E-selectin. Hybridomas in the culture media that selectively stained the plasma membrane of COS-1 cells were cloned by limiting dilution, and the resulting hybridoma lines were expanded and cryopreserved. Among them, a line designated U12-12 was selected for subsequent analysis.

Flow Cytometry of CHO Cells Stably Expressing E-selectin

Chinese hamster ovary (CHO) cells were transfected with pcDNA3.1/Hygro-E-selectin and selected in the presence of Hygromycin B (Thermo Fisher Scientific). Cells positively stained with 68-5H11 (mouse IgG₁; BD Biosciences, Franklin Lakes, NJ), an anti-human E-selectin monoclonal antibody that is useful in unfixed cells but reportedly not suitable to stain FFPE tissue sections, were cloned by limiting dilution, resulting in CHO/E-selectin. Cells were dissociated into monodispersed cells using phosphate-buffered saline (PBS) supplemented with 0.5-mM ethylenediaminetetraacetic acid, and incubated with 68-5H11 or U12-12 at 4C for 15 min. After washing, cells were incubated at 4C for 15 min with anti-mouse IgG or anti-rat IgG, each conjugated with Alexa Fluor 488 (Thermo Fisher Scientific). Stained cells were analyzed using FACSCanto II (BD Biosciences) with FlowJo software (Tree Star; Ashland, OR).

Western Blot Analysis

Western blot analysis was conducted essentially as described. ²³ Briefly, conditioned media containing E-selectin•IgM (and its domain-deficient mutants), L-selectin•IgM, or P-selectin•IgM were subjected to 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred onto polyvinylidene difluoride membrane (Millipore). After blocking with 0.3% skim milk in Tris-buffered saline (TBS) at 4C overnight, the membrane was incubated with either HRP-conjugated anti-human IgM (Jackson ImmunoResearch) or U12-12 for 60 min. In the latter case, after washing with TBS supplemented with 0.05% Tween 20, the membrane was incubated with HRP-conjugated anti-rat IgG. After washing, the membrane was developed with SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) and imaged on ImageQuant LAS 4000 (GE Healthcare, Chicago, IL).

Double Immunofluorescence

COS-1 cells grown on coverslips were transiently transfected with pcDNA3.1-P-selectin•FLAG, and 24 hr later fixed with neutralized 20% formalin (pH 7.4) for 15 min. After permeabilizing cell membranes with 1% Triton X-100 in PBS for 15 min, cells were incubated 30 min with anti-FLAG M2 monoclonal antibody (mouse IgG₁; Sigma-Aldrich), followed by incubation with Alexa Fluor 555-conjugated anti-mouse IgG (Thermo Fisher Scientific) for 15 min. After washing, cells were incubated with U12-12 for 30 min, followed by incubation with Alexa Fluor 488-conjugated anti-rat IgG supplemented with 4′,6-diamidino-2-phenylindole (DAPI) for 15 min. Cells were mounted with 50% glycerol in TBS and observed under a fluorescence microscope.

Human Tissue Samples

FFPE tissue blocks of surgical specimens of acute appendicitis (n=5) and those of biopsy specimens of ulcerative colitis (n=390) were retrieved from the pathology archives of University of Fukui Hospital and Shinshu University Hospital, respectively. In addition, frozen blocks of acute appendicitis tissue were also prepared. Analysis of human tissues was approved by the institutional ethics committees of both universities.

Immunohistochemistry

Immunohistochemical staining for E-selectin was carried out using an indirect method as described. ²⁴ Briefly, after deparaffinization and rehydration, endogenous peroxidase activity was blocked by immersing sections in absolute methanol containing 0.3% H₂O₂ for 30 min. Antigens were then retrieved by autoclaving sections in 10 mM citrate buffer (pH 6.0) at 105C for 15 min. After blocking 30 min with 1% bovine serum albumin in TBS, sections were incubated with U12-12 at 4C overnight, followed by incubation with HRP-conjugated anti-rat IgG for 30 min. The color reaction was developed using a Betazoid DAB Chromogen Kit (Biocare Medical; Pacheco, CA). Sections were briefly counterstained with hematoxylin. In the case of frozen tissues, sections were fixed for 15 min in fixative consisting of 80% ethanol, formalin, and glacial acetic acid at a ratio of 10:1:1, and antigen retrieval was omitted.

Assessment of Histological Activity of Ulcerative Colitis

Based on the grading system proposed by Geboes et al., ²⁵ histological disease activity of each ulcerative colitis biopsy specimen was assessed by a certified pathologist (AK) without knowledge of the patient’s clinical status. Specifically, grade 0 was defined as structural change only, grade 1 as chronic inflammatory cell infiltrates, grade 2A as lamina propria eosinophils, grade 2B as lamina propria neutrophils, grade 3 as neutrophils in the epithelium, grade 4 as crypt destruction, and grade 5 as erosion or ulceration. ²⁵

Quantification of Vessels

Immunostained slides were scanned with NanoZoomer RS (Hamamatsu Photonics; Hamamatsu, Japan) to obtain whole slide images. The number of U12-12-positive vessels in each biopsy specimen was determined using NDP.view2 Plus software (Hamamatsu Photonics).

Statistical Analysis

Differences in the number of U12-12-positive vessels among Geboes grades were statistically analyzed using one-way analysis of variance (ANOVA) with Tukey-Kramer’s honestly significant difference test, and the association between Geboes grades and U12-12 positivity was assessed with the chi-squared test using JMP 15.1.0 software (SAS Institute; Cary, NC). Values of p<0.05 were considered significant.

Results

U12-12 Specifically Recognizes E-selectin

To generate an anti-E-selectin monoclonal antibody, we collected lymphocytes from rats immunized with E-selectin•IgM and established hybridomas that we subjected to a rigorous screening protocol (see Materials and Methods). Using that antibody, which we designate U12-12, we first performed flow cytometry of unfixed CHO/E-selectin cells. As shown in Fig. 2A (lower panel), U12-12 readily bound to CHO/E-selectin cells, and signal intensity was stronger than that of the commercially available anti-human E-selectin monoclonal antibody 68-5H11 (Fig. 2A, upper panel). Moreover, U12-12 was also effective in immunofluorescence analysis of formalin-fixed CHO/E-selectin cells (Fig. 2B). We next used U12-12 to perform Western blot analysis of lysates of COS-1 cells transiently expressing E-selectin. Under reducing conditions, we detected a single band of molecular weight ~115 kDa (Fig. 3A, right lane), but observed no bands in mock-transfected COS-1 cells (Fig. 3A, left lane). These results overall indicate that U12-12 can be used to detect E-selectin on both unfixed and fixed cells, as well as on immunoblots of cell lysates under reducing conditions.

Figure 2.

Immunocytological detection of E-selectin expressed in Chinese hamster ovary (CHO) cells. (A) Flow cytometry analysis of CHO/E-selectin cells performed with 68-5H11 (upper filled histogram) and U12-12 (lower filled histogram) antibodies. Unfilled histograms represent negative controls established in the absence of primary antibody. X- and y-axes indicate fluorescence intensity and number of events, respectively. Note that cells expressing E-selectin are distributed normally, indicating that they are of a single clone. (B) Immunofluorescence of formalin-fixed CHO/E-selectin cells stained with the U12-12 antibody. Nuclei were stained with DAPI. Bar = 50 µm. Abbreviation: DAPI, 4′,6-diamidino-2-phenylindole.

Figure 3.

E-selectin-specific U12-12 binding. (A) Western blot analysis of lysates of COS-1 cells overexpressing human E-selectin using U12-12 (right lane). Lysates of mock-transfected COS-1 cells served as controls (left lane). (B) Western blot analysis of conditioned media of COS-1 cells transfected with cDNA encoding E-selectin•IgM, L-selectin•IgM, or P-selectin•IgM. Membranes were immunoblotted (IB’d) with anti-human IgM (left panel) or U12-12 (right panel).

As noted (see Materials and Methods), U12-12 was screened to react solely with the E-selectin moiety of E-selectin•IgM, one of the three selectin•IgM chimeras. To confirm that U12-12 recognizes only E-selectin, we carried out Western blot analysis of conditioned media from COS-1 cells transfected with cDNA encoding either E-selectin•IgM, L-selectin•IgM, or P-selectin•IgM. As shown in Fig. 3B (left panel), all three selectin•IgM chimeras were decorated with anti-human IgM. However, U12-12 bound solely to E-selectin•IgM, not to L-selectin•IgM or P-selectin•IgM (Fig. 3B, right panel), indicating that the screening performed to establish U12-12 antibody resulted in the intended product.

The U12-12 Epitope Is Located in Sushi 2 Domain of E-selectin

To determine which domain of E-selectin protein exhibits the epitope recognized by U12-12, we generated four domain-deficient E-selectin•IgM mutants: constructs lacking either the CRD (ΔCRD), the EGF-like domain (ΔEGF), the Sushi 1 domain (ΔSushi 1), or the Sushi 2 domain (ΔSushi 2) (Fig. 4A). We then expressed mutants individually in COS-1 cells and assessed U12-12 binding by Western blotting of E-selectin•IgM mutant proteins secreted into the culture media. As shown in Fig. 4B (upper left panel), as expected, all secreted E-selectin•IgM deletion mutants were positive for anti-human IgM. However, although secreted E-selectin•IgM proteins that lacked the CRD, the EGF-like domain, or the Sushi 1 domain were also U12-12-positive, we observed no U12-12 binding to the Sushi 2 domain deletion mutant (Fig. 4B, lower left panel). Furthermore, U12-12 indeed bound to recombinant soluble Sushi 2 domain without the CRD, the EGF-like domain, or Sushi 1 domain (Fig. 4B, lower right panel, far right lane). These results combined indicate that the U12-12 epitope is located in the E-selectin Sushi 2 domain.

Figure 4.

Epitope mapping by domain deletion. (A) Schematic representation of an E-selectin•IgM chimera (wild-type) and corresponding domain deletion mutants (ΔCRD, ΔEGF, ΔSushi 1, and ΔSushi 2). A recombinant E-selectin Sushi 2 domain fused to IgM (Sushi 2), which does not contain the CRD, the EGF-like domain, or the Sushi 1 domain, was also constructed. Each mutant includes the wild-type signal peptide (SP) (dotted square on far left). The relative length of each domain has been adjusted to match the wild-type form, but the Fc region of IgM has been truncated. (B) Western blot analysis of deletion mutants shown in (A) immunoblotted (IB’d) with anti-human IgM (upper panels) or U12-12 (lower panels). (C) Double immunofluorescence staining of COS-1 cells expressing P-selectin•FLAG with anti-FLAG (left panel; red) or U12-12 (middle panel; secondary antibody Alexa Fluor 488 is not detectable). Merged images are shown at right. Nuclei were stained with DAPI. Bar = 50 µm. Abbreviation: CRD, carbohydrate recognition domain; EGF, epidermal growth factor; DAPI, 4′,6-diamidino-2-phenylindole.

We note that, in the expression vector pcDNA1.1-P-selectin•IgM, the region of the P-selectin extracellular domain from the N-terminal CRD to the Sushi 2 domain is connected to the Fc region of IgM,^{7, 20} and the region from the Sushi 3 domain to the C-terminal cytoplasmic tail is not included (see Fig. 1). To eliminate the possibility that U12-12 cross-reacts with amino acid sequences corresponding to the latter, particularly to areas corresponding to the Sushi 3 to Sushi 9 domains of human P-selectin, we conducted immunofluorescence staining of COS-1 cells transiently transfected with cDNA encoding full-length FLAG-tagged P-selectin. As shown in Fig. 4C, cells expressing P-selectin•FLAG were not decorated with U12-12, confirming that U12-12 is specific to E-selectin and does not cross-react with P-selectin.

U12-12 Selectively Stains Vascular Endothelial Cells at Acute Inflammatory Sites

As noted, our initial screen for an anti-E selectin monoclonal antibody was performed using FFPE sections of COS-1 cells overexpressing E-selectin. However, this screen cannot exclude clones that cross-react with other endogenous human antigens. Hence, as a final screen, we performed immunohistochemical staining of FFPE tissue sections of acute appendicitis, a canonical human acute inflammatory disorder. As shown in Fig. 5 (upper panels), we observed prominent transmural neutrophilic infiltration characteristic of acute phlegmonous appendicitis, and U12-12 selectively stained endothelial cells lining vessels in and around such neutrophil infiltrates. Interestingly, neutrophils in the vascular lumen frequently appeared to be attached to the luminal face of U12-12-positive vessels (Fig. 5, arrows in right panels). It is noteworthy that U12-12 did not stain any components of appendiceal tissues other than vascular endothelial cells. We observed similar staining patterns when we used frozen tissue sections (Fig. 5, lower panels). These findings collectively indicate that U12-12 works in an appropriate manner on FFPE human tissue sections.

Figure 5.

Vascular expression of E-selectin in acute appendicitis. FFPE (upper panels) and frozen (lower panels) tissue sections of acute appendicitis were stained with hematoxylin and eosin (HE; left panels) or immunostained with U12-12 (middle and right panels). Images at right (upper and lower rows) are respective enlarged views of the region adjacent to the asterisk in the corresponding middle panel, viewed using an oil immersion lens. Signals were visualized with DAB (brown) and tissues were counterstained with hematoxylin. In right panels, note that neutrophils in the vascular lumen appear to attach to U12-12-positive endothelial cells (arrows). Bar = 50 µm for the left and middle panels and 10 µm for the right panel. Abbreviation: FFPE, formalin-fixed, paraffin-embedded.

E-selectin Is Preferentially Expressed in Highly Active Ulcerative Colitis

Finally, we performed immunohistochemical staining of a total of 390 biopsy specimens of colonic mucosa with ulcerative colitis using the monoclonal antibody U12-12. As shown in Fig. 6 (upper panels), U12-12 preferentially stained lamina propria venules in ulcerative colitis with marked neutrophilic infiltration characteristic of erosion and/or ulceration, corresponding to a Geboes grade of 5. Intriguingly, however, in the absence of neutrophilic infiltration, E-selectin was not expressed on mucosal lamina propria venules, even in the presence of marked chronic inflammation characterized by dense lymphoplasmacytic infiltrates, corresponding to a Geboes grade of 1 (Fig. 6, lower panels).

Figure 6.

Induction of E-selectin on venular endothelial cells in active ulcerative colitis. FFPE tissue sections of ulcerative colitis in either active phase (Geboes grade 5; upper panels) or remission phase (Geboes grade 1; lower panels) were stained with hematoxylin and eosin (HE; left panels) or immunostained with U12-12 (middle and right panels). Images at right (upper and lower rows) are respective enlarged views of the region adjacent to the asterisk in the corresponding middle panel, viewed using an oil immersion lens. Signals were visualized with DAB (brown) and tissues were counterstained with hematoxylin. In the right upper panel, note that neutrophils in the vascular lumen seen in active ulcerative colitis (Geboes grade 5) appear to attach to U12-12-positive endothelial cells (arrows). Bar = 50 µm for left and middle panels and 10 µm for right panels. Abbreviation: FFPE, formalin-fixed, paraffin-embedded.

Next, to determine whether the number of U12-12-positive vessels differs among Geboes’ histological disease activity grades, we counted U12-12-positive venules in each of 390 ulcerative colitis biopsy specimens and performed a one-way ANOVA of the number of U12-12-positive vessels based on Geboes grade. As shown in Fig. 7, we found that the higher the Geboes grade, the greater the number of U12-12-positive vessels (p<0.001), and the vessel number was particularly high in Geboes grades 4 and 5. We then divided the 390 samples into U12-12-positive and U12-12-negative groups (for that categorization, a sample containing even one U12-12-positive vessel was considered U12-12-positive). We also categorized samples with Geboes grades of 4 or higher as the high Geboes grade group and those with grades of 3 or lower as the low Geboes grade group. We then performed a chi-squared test to determine a potential association between Geboes grade and U12-12 positivity. That analysis indicated that the U12-12 positivity rate was significantly greater in the high relative to low Geboes grade group (odds ratio: 21.467, 95% confidence interval: 10.768-42.797, p<0.001). When the receiver operating characteristic curve of the Geboes grade was drawn according to the number of U12-12-positive vessels, the area under the curve was 0.769, indicating moderate accuracy (sensitivity: 0.560, specificity: 0.974).

Figure 7.

Scatter plot representing one-way ANOVA of the number of U12-12-positive vessels (y-axis) based on Geboes grade (x-axis). In the case of multiple identical values, data points are superimposed, and outliers are omitted. The vertical span of each green diamond represents the 95% confidence interval (CI) for each group. The center line across each green diamond corresponds to the group mean. Abbreviation: ANOVA, analysis of variance.

Discussion

Here, we report generation of an anti-human E-selectin monoclonal antibody. This antibody, which we call U12-12, binds exclusively to E-selectin and does not cross-react with either L- or P-selectin. Importantly, U12-12 can be used for immunohistochemical staining of FFPE tissue sections. Therefore, employing FFPE tissue archives, U12-12 can be widely applied to immunohistological analysis of E-selectin expression in various human diseases.

To produce anti-E-selectin antibodies, we exploited E-selectin•IgM chimeric protein as the immunogen. We did this for two reasons. First, and most practically, we had already constructed an E-selectin•IgM expression vector ⁷ and this chimeric protein was readily available to us. Second, we wanted to develop antibodies that recognize the extracellular domain of E-selectin rather than its transmembrane or cytoplasmic domains, so the antibodies could be used for flow cytometry or immunofluorescence staining of cultured cells that had not been permeabilized. Indeed, we show that U12-12 can be used for flow cytometry and immunofluorescence staining of CHO/E-selectin cells without permeabilizing cells (see Fig. 2).

Due to the high homology of the amino acid sequence of the extracellular domain of E-selectin and P-selectin (67% in the CRD, 63% in the EGF-like domain, and 41% in Sushi domains), both polyclonal antibodies and also several monoclonal antibodies raised against E-selectin reportedly cross-react with P-selectin.^3,4 Indeed, BBIG-E6 and 1.2B6 antibodies reportedly recognize identical or overlapping amino acid sequences in the CRD of E-selectin and P-selectin. ⁴ Hence, in this study, after positive selection with E-selectin•IgG to exclude clones that recognize the Fc region of IgM, culture media of hybridoma cells were negatively selected with L-selectin•IgM and P-selectin•IgM to exclude clones that cross-react with either. This screening protocol allowed us to obtain a truly E-selectin-specific monoclonal antibody. Furthermore, we confirmed that U12-12 does not cross-react with P-selectin by immunofluorescence staining of COS-1 cells expressing full-length P-selectin tagged with FLAG.

In this study, our analysis of E-selectin•IgM domain deletion mutants indicated that the epitope recognized by U12-12 is located in the Sushi 2 domain. The amino acid identity of Sushi domains among the three human selectins is 35%, which is low relative to other motifs in the extracellular domain (57% for the CRD and 54% for the EGF-like domain). ⁴ Thus, it is anticipated that clones with epitopes in E-selectin Sushi domains would be preferentially selected following negative selection with a cocktail of L-selectin•IgM and P-selectin•IgM.

One might suspect that U12-12 could cross-react with other molecules with a Sushi 2 domain. Therefore, we conducted a homology search for other proteins with the Sushi 2 domain and found that P-selectin (45%) and L-selectin (38%) were the two most homologous molecules. Because U12-12 does not cross-react with either of these proteins (see Fig. 3B, right panel), it is reasonable to conclude that U12-12 does not cross-react with most proteins with a Sushi 2 domain.

In Western blot analysis of ΔEGF, we observed a doublet rather than a single band, a pattern not seen with other deletion mutants. Currently, the reason for this is unclear, but there are at least two possibilities. One is that in ΔEGF, the fused CRD and Sushi 1 domains may create a new cleavage site, and the other is that ΔEGF may be selectively N-glycosylated due to a conformational change in the mutant protein.

After examining 390 ulcerative colitis tissue samples, we found that E-selectin is preferentially induced in highly active (Geboes grades ≥4) ulcerative colitis. A Geboes grade of 4 or higher means crypt destruction or erosion/ulceration is present due to marked neutrophilic infiltration. These changes are usually accompanied by formation of granulation tissues rich in capillaries and venules. In these conditions, E-selectin-expressing venules contain neutrophils apparently attached to endothelial cells (see Fig. 6, upper panels), which we presume are in the process of extravasation. Because E-selectin plays a pivotal role in the “rolling” step of neutrophil recruitment, it is reasonable to assume that the number of E-selectin-expressing vessels positively correlates with the degree of neutrophil infiltration (which can be assessed by the Geboes grading system ²⁵ in ulcerative colitis).

Using a relatively small number of frozen sections of ulcerative colitis tissues, Cellier et al. ¹⁶ previously reported that vascular expression of E-selectin correlated positively with clinical activity and endoscopic severity, as well as histological activity score. In this study, we were unable to correlate vascular E-selectin expression with clinical activity or endoscopic severity due to limited availability of clinical information relevant to the ulcerative colitis samples. Further multi-institutional analyses of a larger number of samples with more complete clinical and endoscopic information are required to determine the precise impact of vascular E-selectin expression on disease activity in ulcerative colitis.