(Received for publication, September 24, 1996, and in revised form, December 9, 1996)
From the Medizinische Klinik und Poliklinik, Innere Medizin I, Universitätskliniken des Saarlandes, D-66421 Homburg, Federal Republic of Germany
Using autologous serum for the immunoscreening of a cDNA expression library derived from tissue involved by Hodgkin's disease, a new 36-kDa protein with the characteristics of galectins (S-type lectins) was detected. Sequence analysis of the cDNA clone HOM-HD-21 revealed two homologous motifs known as lectin domains with galactoside binding capacity. The two domains are linked by a stretch of about 30 amino acid residues and share a sequence homology of 39%. While the N-terminal lectin domain shows merely moderate homologies with known galectins, the C-terminal lectin domain is highly homologous to rat galectin-5 with an amino acid sequence identity of 70%. We ruled out mutations of the tumor-derived transcript by sequence comparison with the respective cDNA cloned from normal peripheral blood leukocytes. Recombinant protein expressed in Chinese hamster ovary cells was purified from lysates by lactose and galactose affinity chromatography, proving the galactoside binding capacity of this new galectin. Northern blot analysis revealed an expression spectrum restricted to peripheral blood leukocytes and lymphatic tissues. In accordance with the nomenclature of known galectins, we suggest to designate this novel galactoside binding protein galectin-9.
Galectins, formerly known as S-type lectins or S-Lac lectins (1,
2), are a growing family of soluble animal -galactoside-binding proteins. Members of the galectin family are defined by two
characteristic features: affinity to
-galactosides and a specific
sequence motif called lectin domain. Based on the number of lectin
domains two groups of galectins are distinguished. The majority of
galectins, including galectin-1, -2, -3, -5, and -7 (3-9), have a
single lectin domain. The second group of galectins, which includes rat galectin-4 (10), rat galectin-8 (11), and the 32-kDa
galactoside-binding protein of Caenorhabditis elegans (12)
is characterized by two tandemly arranged lectin domains connected by a
linker peptide. Although none of the galectins contains a typical
secretion signal, several galectins are externalized by nonclassical
secretory mechanisms (13, 14) and play a role in modulating
cell-to-cell or cell-to-matrix interactions. Galectins are involved in
a number of different cellular events, including physiological (15, 16)
and malignant cell adhesion (17, 18), activation and proliferation of
immune cells (19, 20), as well as induction (21) and inhibition (22) of
programmed cell death.
Recently, we established SEREX1
(serological identification of antigens by recombinant expression
cloning), a novel approach for the molecular definition of human tumor
antigens using autologous serum from tumor patients. For SEREX,
tumor-derived -phage expression libraries are screened for
reactivity with high-titered IgG antibodies present in the autologous
serum of the analyzed patient. By applying this method to several human
neoplasms we identified numerous new tumor antigens (23, 24). In a
cDNA library derived from the Hodgkin's disease-involved spleen of
a 28-year-old female we detected four different antigens. The molecular
and biochemical analysis of one of these antigens, which was encoded by
the clone HOM-HD-21, revealed a novel human lectin, which shares the
structural and functional properties of galectins. In conformity with
the definition and the nomenclature of other known galectins (galectin 1-8) we propose to name this novel galactoside-binding protein galectin-9.
The study had been approved by the local
ethical review board ("Ethikkommission der Ärztekammer des
Saarlandes", Saarbrücken, Germany). Recombinant DNA work was
done with the official permission and according to the rules of the
state government of Saarland. Sera and tumor tissues were obtained
during routine diagnostic or therapeutic procedures and stored at
80 °C until use. Normal tissues were collected from autopsies of
tumor-free patients.
The
construction of the Hodgkin's-derived cDNA expression library has
been described elsewhere (23). In brief, a oligo(dT)-primed cDNA
expression library resulting in 3.2 × 106 primary
clones was established by directional cloning of the cDNA derived
from the infiltrated spleen of a 28-year-old woman with a nodular
sclerosis type of Hodgkin's disease into the EcoRI- and
XhoI-digested ZAPII phage (Stratagene, Heidelberg,
Germany).
Modifications of the previously described technique (23, 24) were implemented to circumvent the detection of false positive clones encoded by IgG-heavy chain transcripts derived from the numerous B-lymphocytes present in the spleen. After transfection for primary screening, plaques were transferred onto nitrocellulose membranes. After blocking with 5% (w/v) low-fat milk-TBS, nitrocellulose membranes were preincubated with an alkaline phosphatase-conjugated antibody specific for human IgG for 1 h. Reactive clones representing expressed IgG heavy chains were visualized by staining with alkaline phosphatase substrate and marked with a pencil. These prestained membranes were washed extensively with TBS, incubated with autologous patient's serum (1:1000), and the normal immunoscreening procedure was performed. Those plaques which appeared positive but had no pencil marks were considered as positives and subjected to retests and monoclonalization.
Sequence Analysis of Identified AntigensReactive clones were subcloned to monoclonality and submitted to in vivo excision (25) of pBluescript phagemids (26). Sequencing of cDNA inserts was carried out using a Sequenase 2.0 kit (U. S. Biochemical Corp., Bad Homburg, Germany) with vector-specific reverse and universal M13 primers according to the manufacturer's instructions. Specific internal oligonucleotides were designed as the sequencing progressed. Sequence alignments were performed with DNASIS (Pharmacia Biotech Inc.) and BLAST (27) softwares on EMBL (28) and GenBankTM (29) (Release 27.11.96) and PROSITE (30) data bases.
Northern and Southern Blot AnalysisNorthern blots were
performed with RNA extracted from tumors and normal tissues using
guanidium thiocyanate as a chaotropic agent (31). RNA integrity was
checked by electrophoresis in formalin/MOPS gels. Gels containing 10 µg of RNA/lane were blotted onto nylon membranes (Hybond N, Amersham
Corp.). After prehybridization the membranes were incubated overnight
at 42 °C in hybridization solution (50% formamide, 6 × SSC,
5 × Denhardt's, 0.2% SDS) with a 32P-labeled
specific full-length probe. The membranes were washed at progressively
higher stringency, with the final wash in 1 × SSC and 0.2% SDS
at 65 °C for 20 min. Autoradiography was conducted at 70 °C for
2 days using Kodak X-Omat-AR film and intensifying screen. After
exposure the filters were stripped and rehybridized with glyceraldehyde
3-phosphate dehydrogenase to prove RNA integrity. Densitometry of
autoradiographs was performed to compare the expression ratios of
different tissues using a XRS scanner and whole band analyzer software
(BioImage, Ann Arbor, MI). For Southern blot analysis 4 µg of human
DNA digested with EcoRI, HindIII, or
BamHI, respectively, was blotted and hybridized with a probe
specific for the C-terminal lectin domain.
Total RNA was isolated from peripheral blood leukocytes obtained from buffy coats by density gradient centrifugation (Ficoll-paque, Pharmacia, Freiburg, Germany). First strand cDNA was synthesized from 10 µg of total RNA with a (dT)18-oligonucleotide and Superscript reverse transcriptase (Life Technologies, Inc., Eggenstein, Germany). For amplification of human galectin-9 cDNA, transcript-specific oligonucleotides comprising the entire open reading frame were used. PCR was performed for 35 cycles with an annealing temperature of 68 °C under standard conditions with Tth Polymerase (Goldstar, Eurogentec). The resulting 978-bp product was cloned into TA-cloning vector and sequenced.
Production of His-tagged Recombinant Protein for ImmunizationFull-length human galectin-9 protein turned out to
be toxic and not expressable in Escherichia coli. A 245-bp
cDNA fragment coding for the first 81 N-terminal amino acids of
galectin-9 was amplified using the identified HOM-HD-21 cDNA as
template and the primers (5-GCC TTC AGC GGT TCC CAG GCT CCC TAC-3
)
and (5
-CCC AGC TTC CGT GCC TCG TGT TGG ACA-3
) as sense or antisense
oligonucleotides, respectively. The PCR with Pfu polymerase
(Stratagene) was conducted for 20 cycles with an annealing temperature
of 60 °C. The blunt-end PCR product was gel-purified using the QiaEx
II kit (Qiagen) and ligated in frame to SmaI-digested,
dephosphorylated, and gel-purified pQE32 vector (Qiagen), allowing for
the translation of a fusion protein bearing a 6-histidine tail at the N
terminus. To express the His-tagged recombinant protein, the construct
was transformed into E. coli SG13009 (pREP4) strain and
selected on kanamycin/ampicillin-containing plates. Several colonies
were picked, and the production of recombinant protein was induced in
small scale by adding 2 mM
isopropyl-
-D-thiogalactopyranoside to the culture medium
in order to check for protein expression. Small scale purification over
Ni-NTA-columns was performed for each clone. One clone coding for a
protein of the expected length was selected and verified by sequence
analysis. Subsequently, large scale induction of recombinant protein
was performed. Cells were harvested 5 h after induction with 2 mM isopropyl-
-D-thiogalactopyranoside. Lysis
was performed in buffer A (8 M urea, 100 mM
Na2PO4, 10 mM Tris-HCl, pH 8, 0,01% Triton X-100) overnight. Debris was spun down, and supernatant
was loaded onto preequilibrated Ni-NTA-resin. Washes were performed
with 2 volumes of buffer A, pH 8.0, and at least 10 volumes of buffer
A, pH 6.3. Elution was performed with 250 mM imidazol in
buffer A. The yield of affinity-purified His-tagged protein ranged from
15 to 40 mg/liter of bacterial culture. Protein quantification was done
by the Bradford procedure (32) following the recommendations of the
manufacturer (Bio-Rad).
Polyclonal rabbit antisera were obtained from a custom antibody service (Eurogentec, Brussels, Belgium). A total of four immunizations with 100 µg of purified His-tagged fusion protein per immunization were performed. Sera before and after immunization were tested by Western blot for reactivity with the fusion protein. Serum obtained after the final boost was submitted to affinity purification. Briefly, His-tagged protein in buffer A was loaded onto Ni-NTA-resin. Washes with sequentially decreasing molarities of urea in buffer A were performed to equilibrate the column with 100 mM Na2PO4, 10 mM Tris-HCl, pH 8, 0.01% Triton X-100, pH 7.4. After several washes polyclonal serum was eluted with 4 M MgCl2, dialyzed against distilled H20 for 1 h and against phosphate-buffered saline (PBS) overnight. Sensitivity and specificity of the affinity-purified rabbit serum was tested using purified His-tagged protein and CHO/HD21-lysate as positive controls and two unrelated His-tagged proteins (MAGE-1 and HOM-Mel 40) and CHO/pcDNA3 as negative controls. Affinity-purified serum was used for all experiments.
Western Blot AnalysisRabbit antiserum obtained after immunization with the N-terminal fragment of human galectin-9 was used to detect protein expression in cell lysates. Samples of 2 µg of recombinant protein and 20 µg of cell lysates, respectively, were mixed with 2 × SDS buffer (0.1 M Tris-HCl, pH 6.8, 0.2 M dithiothreitol, 4% SDS, 0.2% bromphenol blue, and 20% glycerol), electrophoresed in 12% SDS-PAGE and then blotted onto nylon membranes (Schleicher & Schüll) by semidry transfer (Bio-Rad). After blocking unspecific binding with 5% low-fat milk in TBS for 1 h, the membranes were incubated with 1:100 diluted anti-galectin-9 rabbit serum. The blots were then incubated for 1 h with alkaline phosphatase-conjugated mouse anti-rabbit IgG (Dianova). The membranes were consecutively incubated for 30 min with rabbit anti-mouse Ig (Dianova) as bridging antibody, with anti-alkaline phosphatase and with 0.25 mg/ml alkaline phosphatase. After each incubation step the membranes were washed extensively in TBS and 0.01% Tween 20. Visualization of positive reactions was performed by staining with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium according to the manufacturer's instructions (Bio-Rad).
Transfection and Expression of Human Galectin-9 in Chinese Hamster Ovary CellsThe HOM-HD-21 cDNA insert was excised from the pBluescript-SK phagemid using restriction digestion with EcoRI and XhoI and was ligated into the EcoRI and XhoI cut, gel-purified eucaryotic expression vector pcDNA3 (Invitrogen). The ligation product was transformed in E. coli TOP-10, and plasmids were purified on a silica gel matrix (Qiagen) after alkaline lysis of bacteria. 5 µg of pcDNA3 plasmid containing the ligated HOM-HD-21 fragment were used to transfect CHO cells. As a control CHO cells were transfected with pcDNA3 plasmid containing no insert. Liposome-mediated transfection was carried out in RPMI 1640 medium with 1% fetal calf serum using N-[1-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methyl sulfate (DOTAP) (Boehringer Mannheim) according to the manufacturer's instructions. Transfected cells were designated CHO/HD21 or CHO/pcDNA, respectively, and were selected by a 4-week culture in RPMI 1640 medium containing 10% fetal calf serum and 250 µg/ml G418 (Sigma). Expression of human galectin-9 was analyzed by Western blot using rabbit serum.
Galactoside Affinity Purification1 × 107 CHO/HD21 and CHO/pcDNA cells were lysed in a buffer containing 100 mM Tris-HCl, pH 8, 100 mM NaCl, 2 mM EDTA, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride (Sigma). The lysate was loaded on a lactosyl-Sepharose column (Sigma). Unbound proteins were removed by extensively washing first with buffer B containing 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, and subsequently with PBS. Galactoside-binding proteins which had bound to the column were specifically eluted with 150 mM lactose-containing PBS. Samples of all fractions were analyzed by SDS-PAGE and subsequent silver staining according to the manufacturer's instructions (SilverPage, Biometra).
Applying the modified immunoscreening as described under
"Materials and Methods" to 1.0 × 106 recombinant
plaques, 14 positive clones representing four different transcripts
were identified. A group of six clones, designated HOM-HD-21,
HOM-HD-202, HOM-HD-297, HOM-HD-303, HOM-HD-402, and HOM-HD-415
represented the same transcript. Restriction enzyme mapping and 5 and
3
sequencing revealed that these clones had inserts of different
length ranging from 1586 to 1634 bp. The clone with the longest insert
HOM-HD-21 (Fig. 1) had a 5
-untranslated region of 70 bp, an N-terminal methionine complying with the general features of a
translation initiation site (33), followed by a 969-bp open reading
frame coding for a putative protein with a predicted molecular mass of
35,918 Da and a long 3
-untranslated region of 595 bp. Sequences of all
analyzed clones were identical. Length variation turned out to be due
to incompleteness of the 5
ends resulting from premature stops of the
cDNA synthesis. The alignment of the predicted amino acid sequence
with data bases demonstrated high homologies exclusively with the
members of the galectin family and revealed two domains of 140-150
amino acids in head-to-tail orientation with a mutual homology of 39%
linked by a stretch of 30 amino acids. Both domains contained sequence motifs that are conserved in the carbohydrate binding region of galectins (Fig. 2a; Ref. 2). While the
N-terminal domain showed a moderate overall homology to other known
galectins, the C-terminal domain had a 70% amino acid homology to rat
galectin-5 (Fig. 2b). Considering conserved amino acid
exchanges the homology of this domain to rat galectin-5 raised to 93%.
As described for other galectins, there was apparently no typical
secretion signal in the peptide sequence of HOM-HD-21. Alignment of the
linker peptide attaching both lectin domains revealed no significant
homology to the linker peptides of other galectins. Three putative
glycosylation sites were found at Asn34, Asn79,
and Asn137.
Production of Polyclonal Rabbit Antisera by Immunization with Recombinant Human Galectin-9 His-tagged Protein
The procaryotic
expression of a 5-fragment of HOM-HD-21 cDNA as His-tagged fusion
protein yielded a product with an apparent molecular mass of 12 kDa,
consistent with the expected molecular mass (Fig. 3).
The product was purified by nickel-chelate affinity chromatography and
used for the production of rabbit immune sera. Serum collected before
and after immunization was tested in Western blot against the
recombinant protein used for immunization and two unrelated recombinant
His-tagged proteins. The immune serum had a specific reactivity with
galectin-9 protein (data not shown). For further analysis the serum was
affinity-purified using immobilized His-tagged antigen bound onto the
nickel-chelate column.
Galactoside Binding Activity of Recombinant Human Galectin-9 Protein Expressed in CHO Cells
To confirm the galactoside binding
activity of eucaryotically expressed galectin-9, CHO cells showing no
detectable endogenous galectin-9 expression in Northern blot and
Western blot were chosen. CHO/HD21 and CHO/pcDNA (negative control)
cells were passaged in selection medium containing G418. After 3 weeks
the expression of a specific 35-kDa protein was demonstrated in Western
blots using rabbit antiserum raised against the N-terminal 81 amino acids of human galectin-9 (Fig. 4). The protein is
comigrating with its putative counterpart derived from density gradient
separated peripheral blood leukocytes (data not shown). No additional
smaller bands were detected, implying that the two lectin domains of
the molecule are not cleaved by endogenous proteases. Extracts from CHO/HD-21 and from CHO/pcDNA cells were loaded onto columns with immobilized galactose and lactose. After washing, competitive elution
was performed with galactose and lactose, respectively. Silver staining
of the initial cell lysate and the eluted fraction (Fig.
5) demonstrated binding of the 35-kDa protein to
galactose and lactose. The specificity of the eluted protein was
demonstrated with the rabbit antiserum raised against the N-terminal
fragment of galectin-9 (data not shown). The binding to the galactose
column was apparently weaker than to the lactose column, resulting in a
higher portion of recombinant protein in the flow-through (data not
shown). This is consistent with the nearly 100-fold higher affinity of
galectins to lactose as compared with galactose (34). No binding of
proteins was immunodetected using lysates from CHO/pcDNA cells
transfected with vector lacking the human galectin-9 cDNA.
Human Galectin-9 from Hodgkin's-involved Tissue Is Not Mutated
To exclude mutations of the tumor-derived human galectin-9 transcript as a reason for immunogenicity in the autologous host, we cloned the respective cDNA from normal peripheral blood leukocytes by reverse transcription PCR. Sequencing of the entire open reading frame revealed no differences to the tumor-derived cDNA, ruling out mutations as the cause for the generation of the detected antibody response.
Expression Spectrum of Galectin-9The expression of human
mRNA was analyzed by Northern blot hybridization using 20 µg of
total RNA blotted onto nylon membranes. We detected a moderate
expression of a 1.7-kilobase transcript in peripheral blood leukocytes,
lymph nodes, and tonsils. No expression was detected by Northern blot
in several tissues, including breast, kidney, brain, skeletal muscle,
skin, testis, and stomach. A very weak expression signal detected in
some of the tissue samples derived from colon and lung was most likely
caused by resident leukocytes. We detected a high expression of
galectin-9 transcripts in the Hodgkin's-diseased spleen used for
library construction and in two lymph node samples derived from other
Hodgkin's patients. The densitometric determined expression level in
each of the Hodgkin's disease involved tissues was at least 10-fold
higher than the level of expression found in normal lymphatic tissues
(Fig. 6).
Southern blot analysis with probes specific for the C-terminal lectin
domain revealed hybridization of the probe with at least two distinct
bands for each DNA restriction digestion (Fig. 7).
Hodgkin's disease is a complex lymphoproliferative malignant disorder. The histological diagnosis is based on the presence of Hodgkin-Reed-Sternberg (H&RS) cells surrounded by a cellular infiltrate composed of reactive lymphocytes, plasma cells, histiocytes, neutrophils, eosinophils, and stroma cells (35). The mononucleated Hodgkin and the multinucleated giant H&RS cells usually represent less than 1% of the cellular population of an involved tissue, although they are presumed to be the neoplastic cell population. Specific interactions of H&RS cells with the surrounding lymphocytes mediated by characteristic profiles of cytokines and growth factors have been reported. Hodgkin's disease is often associated with impaired immune functions (36). Aberrations of the humoral immunity include the presence of immune complexes in sera of patients with Hodgkin's disease. Previous investigations revealed that these immune complexes contain antigens that are present in the cytoplasm of H&RS cells (37). However, the nature of these antigens has remained undefined. Using the newly established approach of SEREX, we succeeded in identifying four different antigens in a Hodgkin's-derived cDNA expression library (23, 24). One of these 9 antigens, initially named HOM-HD-21, turned out to be a novel human galectin. It has two homologous lectin domains separated by a linker peptide. This structure is similar to the previously described members of the galectin family consisting of rat galectin-4 (10), rat galectin-8 (11), and C. elegans galectin (12).
Two lines of evidence demonstrate that HOM-HD-21 cDNA codes indeed for a novel galectin: first, its deduced amino acid sequence contains two domains with conserved motifs that are implicated in the carbohydrate binding of galectins; second, eucaryotically expressed recombinant protein is biologically active and possesses sugar binding activity. Based on the nomenclature (1) for the previously described galectins, we suggest to designate this new transcript galectin-9.
The two carbohydrate binding domains of galectin-9 share an amino acid sequence homology of 39% to each other. This is similar to the other previously defined two lectin domain galectins, rat galectin-4 and rat galectin-8, which also have a limited interdomain homology of 35 and 33%, respectively. This limited homology may indicate that the two carbohydrate binding domains may recognize different ligands. The linker peptide sequence of human galectin-9 demonstrates no significant homology to the linker peptides of the other galectins. The same holds true for the previously described two-carbohydrate domain galectins. This may indicate that the structural sequence requirements for the functionality of such a linker peptide are limited.
The C-terminal lectin domain has a high homology to rat galectin-5. Although the published sequence of rat galectin-5 cDNA indicates only one lectin domain (8), splice variants with two lectin domains resident at the same gene locus cannot be excluded, since the genomic clone for rat galectin-5 has not been published to date. In addition, a human analogue of rat galectin-5 has not been identified yet. On this background the C-terminal lectin domain of human galectin-9 might be encoded by the gene for the human counterpart of rat galectin-5. To test this hypothesis we performed a Southern blot hybridization with genomic DNA obtained from human tissue using the C-terminal lectin domain of human galectin-9 as a probe. The finding that besides a strong unique hybridization signal at least one additional signal with weaker intensity was detected (Fig. 7) does contradict this speculation. Furthermore, since the 70% homology of human galectin-9 protein to rat galectin-5 is significantly lower than the homology of other rat galectins to their human counterparts (90% for galectin-1, 81% for galectin-3), it appears more likely that the gene encoding human galectin-9 is not the human counterpart to rat galectin-5.
As reported recently, antibodies reactive with human galectin-9 were detected in about 50% of the sera derived from patients with Hodgkin's disease, but not in the sera of healthy individuals or patients suffering from other tumors (23), suggesting that the antibodies might have been generated as a tumor-specific response. Potential causes for antibody responses may be mutations, neoexpression of viral epitopes, overexpression, or reexpression of embryofetal proteins which are normally silenced in adult tissue. By sequencing we could rule out mutational alterations of the transcript derived from the Hodgkin's spleen cDNA library as initiators for the antibody response. Northern blot studies revealed a very limited expression pattern restricted to peripheral blood leukocytes and lymphatic tissues. Furthermore, a significantly higher expression of the galectin-9 mRNA in tissues involved by Hodgkin's disease was observed. This suggests that loss of tolerance and the development of a strong humoral immune response were initiated by overexpression of the galectin in the tumor tissue. The functional significance of the respective antibodies must be evaluated in future studies.
Although its function is not yet known with certainty, the restricted expression of the new galectin to lymphatic tissues also suggests that it might have an important role in the regulation of cellular interactions of the immune system. Such a role has also been demonstrated for other galectins, e.g. for galectin-1, which induces apoptosis in activated T-cells (21) and galectin-3, which is up-regulated in proliferating T-cells and can inhibit Fas-mediated apoptosis (22). As this new galectin is strongly overexpressed in Hodgkin's disease tissue, it is conceivable that it might participate in the interaction between the H&RS cells with their surrounding cells and might thus play a role in the pathogenesis of this elusive disease and/or its consistently associated immunodeficiency.
So far we have not been able to assign the cellular origin of the galectin-9 transcripts to defined subpopulations in the peripheral blood or in the Hodgkin's infiltrated tissues, since the polyclonal rabbit serum, which reacted with His-tagged protein in Western blot, did not work in immunocytology and immunohistology (data not shown). Preliminary studies using peripheral blood leukocytes separated by magnetic cell sorting indicate that the expression level of the transcript is similar in enriched populations of B-cells, T-cells, and macrophages (data not shown). The availability of monoclonal antibodies that function in immunocytology will enable us to define more precisely the subpopulation(s) with high human galectin-9 expression.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) Z49107[GenBank].