Background
There are two main epithelial cell types in the bronchus – ciliated cells and goblet cells that secrete mucus. Within the submucosa there are also submucous glands present. The goblet cells have a glycoprotein which is acidic due to the presence of sialic sulphate groups and this secretion may vary with various diseases [
1]. SAα2,6Gal has been reported to be present on the apical surface of ciliated cells but there have been conflicting reports about SAα2,3Gal expression on cell types. In cultured epithelial cells SAα2,6Gal appears to be present on non-ciliated (goblet) cells while SAα2,3Gal is present on ciliated cells [
2]. On the contrary, others have reported that SAα2,3Gal expression is found in goblet cells [
3]. In addition, the patterns of glycosylation and the expression profile of SAα2,6 on cell surfaces may change during the course of developmental differentiation and following oncogenesis. For instance, SAα2,6Gal binding is weak during the glandular stage of lung development but increases as the lung matures [
4]. Furthermore, if cells are exposed to inflammation and tumour necrosis factor there may be qualitative changes in glycosylation and the glycosyltransferases that lead to sialylation [
5]. Recent publications, however, have indicated that both SAα2,3Gal and SAα2,6Gal may be present in the respiratory tract but with different distributions [
6,
7]. The presence or absence of these SA is important as human influenza A strains have been reported previously to preferentially attach to cells with SAα2,6Gal linkages and avian strains preferentially bind SAα2,3Gal [
8].
The affinity of the attachment of the HA to cell surface receptors is believed to be an important determinant in tissue tropism of the virus and constitutes part of the species barrier that keeps avian influenza viruses from readily infecting humans. Pigs contain a respiratory epithelium that has been reported to contain both "avian-virus" binding SAα2,3Gal and "human-virus" binding SAα2,6Gal linkages supposedly explaining why they can be infected with both human and avian influenza viruses [
9]. Therefore pigs have been regarded as a hypothetical "mixing" vessel where re-assortment of avian and human viruses can take place, potentially leading to the emergence of pandemic influenza [
9]. Given the presumed importance of the affinity of the influenza virus for its receptor, the distribution of SAα2,6Gal and SAα2,3Gal expression in the human respiratory tract is critically important for the understanding of influenza pathogenesis.
Some of the studies on the types of SA expressed on cell surfaces have been done on sialic acids extracted from cell membrane homogenates. However, for the understanding of influenza pathogenesis and pandemic emergence, it is important to have methods that can define the profiles of SAα2,6Gal and SAα2,3Gal in histological tissue sections in situ. The SAα2,6Gal and SAα2,3Gal expression on histological specimens in situ can be done using fluoresceinated lectins and by histochemistry. Since the early 1990's many histology laboratories have used unmasking or retrieval techniques to enhance immunohistochemical detection of antigens.
In this study we address three issues. The first was to investigate if antigen unmasking or retrieval would affect lectin-ligand expression in histological tissues. The second issue was to compare the findings obtained by using lectin fluorescence with cytochemistry for lectin-ligand analysis in the same tissue samples. Finally, we wanted to use optimized methods to re-evaluate the distribution of the SAα2,6Gal and SAα2,3Gal in human respiratory tissues and then correlate this with the reported presence or absence of influenza infection in different parts of the respiratory tract.
Methods
Biopsy samples were collected from the archived files of the Histopathology Department of Queen Mary Hospital, Pok Fu Lam, Hong Kong. Five surgically removed lungs from children with congenital cystic adenomatoid malformation (CCAM), seven non-neoplastic bronchial biopsies from patients investigated for possible malignancy, eight normal nasopharyngeal biopsies from patients with suspected nasopharyngeal carcinoma and eight lung biopsy samples from 20–40 week abortuses were also used. Seven biopsy tissues of CCAM, representing paediatric lung tissues of ages 1 month to 7 years were also retrieved from the files of the Department of Histopathology, Adelaide Women's and Children's Hospital, Adelaide, South Australia. All tissues had been fixed in 10% neutral buffered formalin, processed into paraffin and stored at room temperature. Tissue blocks from the lungs of five patients who had died of acute bacterial pneumonia were used as a non-influenza comparison. Intestinal tissue from four ducks kindly provided by the Agriculture, Fisheries and Conservation Department, Government of HKSAR, were used as a positive control for SAα2,3Gal binding. The research was approved by the Ethics Committee of the University of Hong Kong/Hospital Authority Western Cluster.
We used lectin histochemistry and fluorescence which is the standard method of detection of the SA linkages [
10]. Lectin analysis was performed using the fluorescein labelled lectins Sambucus nigra agglutinin (SNA) which primarily detects 6-linked sialic acids and Maackia amurensis agglutinin (MAA) which primarily identifies 3-linked sialic acids. Both fluorescein isothiocyanate-labelled (FITC) and tetramethyl rhodamine isothiocyanate-labelled (TRITC) were used as fluorochromes and purchased from EY Laboratories (San Mateo, California). Additional FITC conjugated MAA1 was purchased from Vector Laboratories (Burlingame, CA). When peroxidase or biotin conjugation was used the conjugates were purchased from EY Laboratories (SNA and MAA) and Vector (MAA1 and MAA2). Digoxigenin labelled SNA and MAA was purchased from Roche as part of the Dig-Glycan Detection Kit.
For the initial trial of optimization for oligosaccharide ligand retrieval methods one lung block from a case of CCAM was used. The tissues were sectioned at 5 μm and deparaffinized. Control sections had no retrieval. Two different methods were used for retrieval: microwave and enzyme digestion. For microwaving, an Energy Beam Sciences microwave processor was used together with 2 types of buffer. 0.1 M EDTA and 10 mM citrate buffer was used and the sections were microwaved for 10,15, 20 and 25 minutes at 95°C. Two enzyme digestion methods were used: trypsin and pronase, and for both these methods sections were incubated for 15 mins at 37°C.
Single fluorescent studies were performed as follows. The sections were microwaved in 95°C citrate buffer pH 6.0 for 15 minutes, washed with 0.05 M Tris Buffer Saline (TBS) pH 7.6 and then incubated with either 1/100 FITC conjugated SNA (EY Laboratories), or 1/100 FITC conjugated MAA (EY Laboratories) for 1 hour at room temperature in the dark. Double fluorescent studies were performed according to the method of Mason et al. [
11]. Briefly, the sections were microwaved in 95°C citrate buffer pH 6.0 for 15 minutes, washed with 0.05 M Tris Buffer Saline (TBS) pH 7.6 and then incubated with 1/100 TRITC conjugated SNA (EY Laboratories) and 1/100 FITC conjugated MAA(EY Laboratories) for 1 hour at room temperature in the dark. The sections were washed with TBS 3 times for 5 minutes each and the nuclei stained with 5 μg/ml DAPI for 4 minutes followed by three washes with TBS of 5 minutes each and mounting with DAKO fluorescent mount (Dako, Glostrup, Denmark). Fluorescent examination was with a Nikon Eclipse microscope with SPOT Pursuit Camera (Sterling Heights, MI) and Image-Pro Plus software (MediaCybernetics, MD) was used.
Lectin horseradish peroxidase (HRP) detection: the sections were microwaved in 10 mM citrate buffer pH 6.0 for 15 min, blocked with 3% H2O2 in TBS for 12 min and after washing with TBS 3 times, 5 minutes each were then incubated with 1/50 HRP conjugated SNA (EY Laboratories) and 1/50 HRP conjugated MAA (EY Laboratories) at room temperature for 1 hour respectively. After 3 further washes in TBS the sections were developed with AEC substrate kit (Vector Laboratories) at RT for 30 minutes followed by counterstaining with Mayer's haematoxylin and mounting with DAKO aqueous mount (Dako, Glostrup, Denmark).
Lectin biotin detection: The sections were microwaved in 10 mM citrate buffer pH 6.0 at 95°C for 15 min then blocked with 3% H2O2 in TBS for 12 min and with avidin/biotin blocking kit (Vector). They were then incubated with 1/100 biotinylated MAA1 (or 1/100 biotinylated MAA2) (Vector) for either 1 hour at RT or 4°C overnight, blocked with 1% bovine serum albumin for 10 mins at RT, and then incubated with strep-ABC complex (Dako Cytomation, K-0377) diluted 1/100 for 30 mins. at room temperature. Development was performed using the AEC substrate kit (Vector) at room temperature for 10 minutes. The nuclei were counterstained with Mayer's hematoxylin and then the sections were dried and mounted with DAKO aqueous mount (Dako Cytomation). Duck intestine sections were used as controls with and without pre-treatment with SAα2,3 specific neuraminidase from Glyko to ensure that sialic acids were specifically targeted. Stain intensity was measured semi-quantitatively using duck MAA as a control. Similar stain intensity to the duck intestine was graded as strong (++) and a weaker pattern as +.
To determine lectin binding profiles we used data from the Consortium for Functional Glycomics (CFG) web site [
12] using glycan array data for MAA1(also known as MAL), SNA and a human H5N1 virus, (A/Vietnam/1203/04).
Discussion
Cells of the respiratory tract express a number of glycan containing conjugates on the cell surface, many of which terminate with
N-acetylneuraminic (sialic, SA) acids – a series of 9-carbon sugars. Influenza virus infection of humans involves binding of the virus haemagglutinin (HA) to these sialyoligosaccharides on the surface of cells of the respiratory tract. In addition, the virus neuraminidase (NA) cleaves the sialic acid on the host cell and is important in releasing newly formed virus from the cell after virus replication is completed so these virions can spread out in search of other cells to infect. Since respiratory mucus is also rich in SA, this provides a potential barrier to the spread of newly formed virions. By cleaving these SA, the influenza virus NA facilitates the virus spread through this mucus layer [
14]. Thus the balance between the affinity of binding between the virus and the cell receptor and the virus-releasing activity of the NA is critical to virus replication in a host species.
The crystal structures of HA shows that the terminal sialic acids bind in a groove at the top of the HA molecule [
15]. Previous investigations had indicated that avian viruses would preferentially bind SAα2,3Gal and human viruses SAα2,6Gal [
8]. The avian and human H5N1 viruses causing the "bird flu" outbreak in Hong Kong in 1997 had affinity for binding to SAα2,3Gal but the virus associated with human disease in Hong Kong in 2003 had affinity to bind to both avian-like SAα2,3Gal and human-like SAα2,6Gal [
16]. H5N1 disease in humans has been reported to be different from conventional human influenza viruses (H3N2 or H1N1) in that the lower (rather than upper) respiratory tract is the major target for virus replication [
7].
Using retrieval methods and selection of lectin conjugate we have demonstrated that the lectin binding to the SAα2,3Gal receptor for avian influenza viruses is more widely expressed in the respiratory tract of humans than previously documented [
7]. In particular, we demonstrated that unlike the previous reports that indicated certain type of cells had only one lectin binding profile, SAα2,6 and SAα2,3Gal was found in ciliated epithelium, goblet cells and submucous glands in the bronchus as well as pneumocytes of the alveoli, and SAα2,3 was also present in the metaplastic epithelium. In keeping with an earlier report [
4], we found that neonatal pneumocytes expressed mainly SAα2,3Gal and the neonatal bronchus was also primarily SAα2,3Gal expressing. The respiratory tract of young children showed mainly SAα2,3Gal with a lower level of SAα2,6Gal expression than adult tissues. This may, in part explain why children appear to be more susceptible to avian influenza H5N1 in the recent outbreaks in East Asia.
The difference between our studies and previous ones on sialic acid expression in respiratory epithelial cells can be partially explained by the methods used for lectin analysis as well as the type of lectin conjugate used. Antigen retrieval or unmasking did not become an established procedure in many laboratories until the mid 1990's and the earlier publications used paraffin embedded tissues without retrieval [
3,
8]. Later studies on pigs, primates and ducks also did not use retrieval procedures [
9,
17]. While the precise mechanism of retrieval still remains not precisely defined the general consensus is that heating or the use of enzymes serves to unmask antigenic sites that have become cross-linked through formalin fixation [
18]. The unmasking of epitopes appears to extend to carbohydrate moieties as well as proteins. The increased detection of SAα2,6Gal through unmasking has also been recently shown in the liver [
19].
Two isoforms of MAA have been recognized for many years and their binding profiles have been characterized. Since MAA1 bound to non-SAα2,3 glycans it has not been used as extensively by some researchers as MAA2. For instance, Shinya et al have recently demonstrated SAα2,3 Gal expression only in the lung but not the bronchus or upper respiratory tract [
7]. Since they only used MAA2 lectin binding (Y. Kawaoka, personal communication) our MAA2 results are in accord with theirs. But our finding of MAA1 binding in the upper and lower respiratory tract does have implications for the possible distribution of receptors for avian influenza viruses including the currently circulating H5N1 viruses. The crucial question therefore is whether for the identification of susceptible binding sites for avian influenza viruses which are known to bind SAα2,3Gal, should researchers should just use MAA2 or should they also use MAA1? While both of these isoforms do identify the SAα2,3Gal ending they differ in their affinity for the inner fragments of the glycans, and these inner fragments are known to affect different virus binding [
20]. To answer this we used glycan array data performed by the CFG for a known H5N1 avian influenza virus (Viet04) and also the affinity data for MAA1 (also known as MAL) [
12]. As expected Viet 04 does have strong affinity for SAα2,3Gal terminated glycans and that MAL does identify SAα2,3Gal glycans but it also identifies non-SAα2,3 moieties and there is an overlap in the glycans that Viet04 and MAA1 have high affinity to [see Additional file
1]. Thus, there are H5N1 binding SAα2,3 terminated oligosaccharides that are detected with MAA1 (e.g. glycans 235–237), but these potential binding sites may not be detected if only MAA2 is used. We therefore believe that MAA1 should be used in conjunction with MAA2 in determining tissue distribution but do acknowledge that as it does detect non-SA terminated residues ancillary methods such as neuraminidase treatment of sections may be needed. Furthermore our findings that the MAA supplied by Roche and EY Laboratories primarily identifies MAA1 also indicates that these lectins should be used in conjunction with specific lectins from Vector labs to avoid misinterpretation of potential binding sites for SAα2,3 binding viruses. It is of note that our findings are applicable not only to influenza viruses but also to some parainfluenza viruses which also have strict SAα2,3Gal binding preference.
Because of this varied distribution of MAA1 and MAA2 throughout the respiratory tract, we hypothesised that this may shed new light on the distribution and binding of H5N1 viruses to the upper and lower respiratory tract. Using this information we therefore used ex-vivo cultures of the upper and lower respiratory tract and infected them with different H5N1, H1N1 and H3N2 viruses and found that contrary to previous suppositions, H5N1 viruses were able to replicate in the upper respiratory tract – a region which lacked MAA2 binding but had abundant MAA1 binding, thus indicating that the virus is perhaps binding to SAα2,3Galβ1,3/4GlcNac motifs or even to non-sialyated receptors [
21].
Our results also indicated that the recent findings of Matrosovich and colleagues who found SAα2,3Gal in ciliated cells and SAα2,6Gal in the goblet cells of tracheobronchial cells cultured
in vitro are only a partial representation of the true nature of sialic acid expression in the adult respiratory tract [
2]. Their findings in the
in vitro model with cultured tracheobronchial epithelium showed more similarity to the receptor profile seen in the respiratory tract of children. Therefore it is possible that the human tracheobronchial epithelial culture model is more representative of the respiratory tract of children rather than that of adults, and represents a developmentally earlier model of the human respiratory tract.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Dr J Nicholls and Professor J S M Peiris designed the experiments and were responsible for primary analysis of tissues. Dr A J Bourne provided input to the CCAM cases from Australia. Dr H Chen and Y Guan assisted in the interpretation of results. All authors have read and approved the final manuscript.