ELSEVIER Biochimica et Biophysica Acta 1312 (1996) 137-144

BB Biochi~ic~a et Biophysica A~ta

Biphasic modulation of cell growth by recombinant human galectin-1

Linda Adams, G. Kenneth Scott *, Cristina S. Weinberg School of Biological Sciences, University of Auckland, Private Bag 92019, Auckland, New Zealand

Received 17 November 1995; revised 20 February 1996; accepted 21 February 1996

Abstract

Human soluble galactose-binding lectin (galectin-1) has been expressed as an Escherichia coli fusion protein, following the amplification by polymerase chain reaction of cDNA prepared from a human osteosarcoma cell line. The fusion protein is a functional /3-galactoside-binding lectin, as is the recombinant galectin when purified from the cleaved fusion protein. The recombinant galectin has a biphasic effect on cell proliferation. Unlike the fusion protein, it functions as a human cell growth inhibitor, confirming earlier findings with natural human galectin-1, though it is less effective than the natural galectin. This reaction is not significantly inhibited by lactose, and is thus largely independent of the /3-galactoside-binding site. At lower concentrations, recombinant galectin-I is mitogenic, this activity being susceptible to inhibition by lactose, and thus attributable to the /3-galactoside-binding ability of the protein. Some tumour cells are susceptible to the growth-inhibitory effect, and the galectin-1 gene is expressed in both normal and tumour cells.

Keywords: Galectin-1; Lectin; Fusion protein; Cell growth control; Proteinase

I. Introduction

There is evidence for the existence of endogenous growth-related proteinases [1,2]. One such enzyme, the so-called growth-related proteinase (GRP), has been sub- jected to some functional characterisation. Antibodies to this enzyme inhibited the growth of many human cell types in culture, as did specific macromolecular proteinase inhibitors [3-6]. Investigations into the effects of GRP inhibition on intracellular signalling pathways have re- cently been summarised, but no firm conclusions as to a site of action can yet be drawn [7]. Though there were early indications that thrombin might mimic the GRP [3], it now seems clear that the GRP does not interact with the thrombin receptor [7,8].

A report of a small, negative growth-regulatory protein produced by human fibroblasts showed that its effect could apparently be mimicked by an exogenous proteinase in- hibitor [9]. We considered it possible that the role of the proteinase inhibitor was to protect rather than to mimic the growth regulator. Another endogenous negative growth regulator, first identified in mouse fibroblasts [10], was extensively characterised and shown to be a /3- galactoside-binding protein (mGBP) [11]. The lability of

* Corresponding author: Fax: +64 9 3737452.

0167-4889/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved PII S0167-4889(96)00031-6

the mGBP suggested the possibility that it might be a substrate for a proteinase such as the mouse homologue of the GRP. Using human cells, we have obtained some circumstantial evidence to support this hypothesis. A /3- galactoside-binding protein was isolated from human fi- broblast conditioned medium by affinity chromatography. Yields of the protein were very low, but were increased in cultures treated with GRP inhibitors. The purified protein acted as a cellular growth inhibitor [12].

In order to extend this work, it was necessary to have access to a recombinant form of the GBP. The human homologue of mGBP has been cloned on several occa- sions, and we chose the human hepatoma cDNA sequence reported by Abbott and Feizi [13] as the starting point for the expression of recombinant hGBP. It should be noted that the animal /3-galactoside-binding soluble lectins, of which GBP is one, have had a diverse and confusing nomenclature. A recent, well-supported proposal suggests that these lectins be referred to as 'galectins' [14]. The GBP protein is designated 'galect in- l ' , and we shall use this name henceforth.

Galectins, like other lectins, act as mitogens for a variety of cells, including spleen cells [15], lymphocytes [16] and vascular cells [17]. There is evidence in each case for inhibition of the mitogenic effect at supra-optimal galectin concentrations. It is possible that this inhibitory

138 L. Adams et aL / Biochimica et Biophysica Acta 1312 (1996) 137-144

phase may correspond to the specific growth-inhibitory activity more recently reported [11,12]. Further to this mitogenic activity, growth stimulation was also observed when bovine corneal endothelial cells were treated with human galectin [12]. Growth-stimulatory activity in a po- tentially therapeutic growth inhibitor is obviously worthy of investigation, irrespective of the intrinsic interest in a biphasic growth-modulatory protein acting on human cells.

We describe the production, amplification and cloning of human galectin-I cDNA, its expression as a bacterial fusion protein, and the effect of the purified recombinant protein on the growth of human cells in culture. The production of a galectin-glutathione-S-transferase fusion protein was of particular value in this study, as it retained lectin function, but was not a growth inhibitor, whereas the separated recombinant galectin displayed both activities.

2. Materials and methods

2.1. Cell culture

The MPS fibroblast cell strain is a normal human diploid fibroblast strain previously used by us in many experiments on growth-related proteolysis [5]. Tumour cell lines used included HELA and HEP 2 (human carcinomas), and U 2 0 S and MG 63 (human osteosarcomas), also previously used in such studies [6,18].

The conditions for maintenance and culture of cells have been previously described, as have assays for inhibi- tion of cell proliferation in microwell culture plates, using an ELISA spectrophotometer to measure fixed and stained cells [19], and the assay of [3H]thymidine incorporation into DNA [20]. In cell proliferation assays, recombinant galectin-1, the GST-galectin fusion protein, or GST itself were added to normal culture media at known concentra- tions. In some experiments, lactose (100 mM) was also added; control experiments established that this did not affect cell viability or proliferation. Thymidine incorpora- tion was measured in confluent, quiescent well-plate cul- tures, to which proteins and lactose were added as appro- priate. Cell viability was tested by Trypan Blue exclusion in confluent fibroblast monolayer cultures, using a 0.04% dye solution in sodium phosphate-buffered isotonic saline (PBS).

2.2. Cloning and expression of the galectin-1 gene

Unless otherwise noted, DNA-modifying enzymes were purchased from Life Technologies and all other reagents were from Sigma. The plasmid expression vector pGEX 4T-1 [21] and E. coli DH 5c~ cells were donated by Dr. David Christie. We used protocols described by Sambrook et al. [22] for techniques such as DNA restriction and ligation, bacterial transformation, and agarose gel elec- trophoresis.

Isolation of cytoplasmic RNA from mammalian cells was done initially using confluent cultures of human U2-OS osteosarcoma cells by a rapid method [23]. The extracted RNA was used as a template for a reverse transcription reaction performed using a RiboClone cDNA Synthesis System with AMV reverse transcriptase (Promega). The resulting cDNA was used as a template for PCR reactions with two pairs of convergent primers. To design these oligonucleotides, we referred to the sequence of the cDNA of human /3-galactoside-binding lectin from Hep G2 cells [16]. For initial experiments, the primers were 5'-CTGGA- GATCTTCATGGCTI'GTGGTCTG-3' and 5'-TGGCAGA- TCTCAGTCAAAGGCCACACA-3', and resulted in the production of a DNA fragment of approximately 450-bp in an agarose gel. The fragment was cloned into Bluescript, and sequenced by a modified Sanger technique using a 373A Sequencer (ABI) in the Sequencing Unit of the School of Biological Sciences, confirming the reported sequence. With a view to bacterial expression of the galectin, new primers were designed, introducing a Barn HI site in the GBP3 primer and a Sal I site in the GBP4 primer to permit cloning the amplified product into the pGEX vector. The GBP3 and GBP4 primers were, respec- tively, 5'-CGTGGATCCATGGCTTGTGGTCTGGTC-3' and 5'-ACGCGTCGACTCAGTCAAAGGCCACACA-3'. Conditions for PCR were 500 mM KC1, 1.5 mM MgC12, 100 mM Tris-HC1, pH 8.8, 250 mM dNTPs (Pharmacia), 20 pmol of each primer, 1.25 U Ampli Taq DNA poly- merase (Perkin Elmer) and 30 cycles of amplification (denaturing at 95°C for 1 rain, annealing at 55°C for 90 s and polymerizing at 72°C for 90 s). The amplified product was run on a 1.5% agarose horizontal gel in TAE buffer, cut out of the gel and purified using a Geneclean II kit (Bresatech). The product was restricted with Barn HI and Sal I restriction enzymes and ligated using T4 DNA ligase into the pGEX 4T-1 vector restricted with the same en- zymes, then transformed into competent E. coli DH 5a cells.

Transformants were screened for the presence of the insert by growing l0 ml overnight cultures from several transformant colonies and using a Wizard miniprep DNA purification system (Promega) to isolate plasmid DNA which was later Barn HI/Sal I restricted and run on a 1.5% agarose gel. The clones having the 450 bp insert were grown in 500 ml overnight cultures and the plasmid purified by equilibrium centrifugation in CsCl-ethidium bromide gradients [22] and then sequenced, as previously described.

2.3. Production of recombinant galectin-1

One of the several transformants having the insert was then innoculated in 3 ml LB/ampicillin medium and grown at 37°C in a shaking incubator until the absorbance at 600 nm reached 0.8. At this stage the cultures were induced for 3 h with 1 mM IPTG and then microfuged for

L. Adams et al. / Biochimica et Biophysica Acta 1312 (1996) 137-144 139

10 min at 5000 rpm. The cell pellets were boiled in gel loading buffer and applied to a 10% SDS polyacrylamide gel, to test for expression of the fusion protein. For large-scale expression, 1 litre cultures were grown and induced under the same conditions. The cultures were centrifuged in a Sorvall GSA rotor at 5500 rpm for 10 rain to collect cells and the pellet was resuspended in 10 ml of ice-cold sodium phosphate-buffered isotonic saline (PBS) containing 0.1 mM PMSF, to prevent proteolysis. The cells were lysed two times in a French Cell Press at 1400 psi and Triton X-100 was added to a concentration of 1%. The cell lysate was then centrifuged in a Sorvall SS-34 rotor at 10000 rpm for 15 min to remove insoluble material and intact cells. The supernatant was collected and passed through a glutathione-agarose affinity column (20 ml). The column was washed with 3 column volumes of PBS. The fusion protein was eluted with 50 mM Tris-HC1 buffer (pH 8.0), containing 5 mM reduced glutathione. The samples containing the fusion protein were pooled and dialysed overnight against thrombin cleavage buffer (50 mM Tris- HCI, pH 7.5, 150 mM NaCI and 2.5 mM CaCI2).

Thrombin was added to the fusion protein sample (0.2, 0.6 or 1% w / w fusion protein) and the mixture was incubated for 90 min at 25°C. The mixture was added to the glutathione-agarose affinity column and the released protein was recovered by washing the column with 1 column volume of 50 mM Tris-HCl buffer (pH 7.5), containing 150 mM NaCI. The released protein was con- centrated using a Microsep centrifugal concentrator with a cut-off of 1 kDa, assayed using the Bradford protein assay [24], and was run on a 15% SDS-polyacrylamide gel for a purity check. For further purification, an asialofetuin- agarose affinity column (10 ml) was used for galectin fractionation, using conditions for absorption and elution as previously described [12]. For some control cell culture experiments, it was necessary to add unmodified GST, which was prepared essentially as described above, from cells transfected with the unmodified pGEX-4T 1 plasmid.

N-terminal amino acid sequence analysis was carded out by Ms. Catriona Knight, of the Auckland Protein Sequencing Laboratory, on samples electroblotted from SDS-polyacrylamide gels on to polyvinylidene difluoride membrane. Protein-staining bands in the membrane were cut out, and used directly for sequence analysis in a gas-phase protein sequencer (Applied Biosystems model 470A).

Erythrocyte agglutination by the recombinant lectin was tested using freshly-trypsinised rabbit erythrocytes [25,26]. Agglutination experiments were carried out with serial threefold dilutions of the purified fusion protein in ceramic tiles. Each well contained 100 /zl of 1% bovine serum albumin in 0.15 M NaC1, 100/zl of 0.15 M NaC1 and 100 /~1 of erythrocyte suspension, and 100 /xl protein solution (1 mg. m1-1 in phosphate-buffered saline) was used to start the serial dilution. Iqaemagglutination experiments were carried out in the presence and absence of /3-mer-

A B a b c a b c

Fig. 1. SDS-PAGE of GST, galectin-1 and GST-galectin fusion proteins. A. 10% SDS gel. Lane a: Mol. wt. standards (from the top of the gel; 205, 116, 97, 67, 45 and 29 kDa). Lane b: Purified GST. Lane c: Total bacterial protein; putative fusion protein arrowed. B. 15% SDS gel. Lane a: Mol. wt. standards (from the top of the gel; 67, 45, 36, 29, 24, 20 and 14 kDa). Lane b: Cleaved fusion protein (0.2% thrombin), Lane c: Fusion protein purified by glutathione-agarose chromatography.

captoethanol (10 mM), and of lactose (100 mM), and also with the cleaved, recombinant galectin. Visual identifica- tion of the endpoint of the titration of agglutination was confirmed by microscopic detection of agglutinated ery- throcytes.

3. Results

Agarose gel electrophoresis of the cDNA PCR-ampli- fled from U 2 0 S cells with the GBP3 and GBP4 primers yielded a fragment of approximately 450 base-pairs, and a similar fragment was seen with similarly-processed cDNA from normal human fibroblasts, and from other human tumour cells (MG 63, HELA). In some cases, an additional 250 base-pair fragment was also seen; this has not been further characterised. Nucleotide sequencing of the U 2 0 S 450 base-pair cDNA fragment restricted with BamHI/SalI, ligated into p-GEX 4T and cloned, con- firmed the nucleotide sequence previously reported for human galectin-1 [13]. We were unable to detect a corre- sponding cDNA from HEP 2 cells treated in the same way (data not shown).

Growth of small scale cultures of one of these clones confirmed the presence of a new protein of apparent molecular weight about 36-38 kDa in bacterial lysates (Fig. 1A, lane c). This corresponded to a putative fusion protein, and compares with the 29 kDa band which corre- sponds to GST (lane b). The fusion protein could be readily purified from large-scale bacterial culture lysates by affinity chromatography on glutathione-agarose. Elution profiles (not shown) displayed a large peak of protein passing unretarded through the chromatography column, and a much smaller peak initially retained on the column, and eluted by 5 mM glutathione. A 37 kDa protein was the major component in this glutathione-eluted fraction (Fig. 1B, lane c), though there was also some high-molecular- weight protein present in some preparations. Thrombin

140 L. Adams et al. / Biochimica et Biophysica Acta 1312 (1996) 137-144

A B a b c d e a b c d

Fig. 2. SDS-PAGE of galectin-1 purified from cleaved fusion protein by asialofetuin-agarose affinity chromatography. A. 15% SDS gel Lanes a,b: Lactose-eluted galectin-1 (0.6% thrombin), at different loadings. Lanes c,d: Unretarded fraction, at different loadings. Lane e: Mol. wt. markers (from top of gel; 67, 45, 36, 29, 24, 20 and 14 kDa). B. 15% SDS gel Rechromatography of unretarded fraction from (A). Lane a: Unretarded fraction. Lane b: Fusion protein prior to cleavage and chromatography. Lane c: Lactose-eluted galectin-1. Lane d: Mol. wt. markers (from top of gel; 67, 45, 36, 29, 24, 20, 14 kDa).

treatment (0.2% thrombin) of this purified fusion protein readily yielded additional bands of about 14 kDa and 29 kDa (Fig. 1B, lane b). It was sometimes difficult to get total conversion of the 37 kDa band, and some experi- ments were done with 0.6% or 1% thrombin, in an attempt to improve cleavage of the fusion protein. Though success- ful in this aim, proteolysis of the galectin could readily result (see below). Attempts to separate the putative 14 kDa galectin-1 from the other cleavage products by glu- tathione-agarose chromatography were not very successful, with all three protein components (fusion protein, galectin and GST) present in all fractions (not shown). Formation of a dimer of the 14 kDa galectin may have contributed to the staining intensity of the 29 kDa band (see below).

Sephadex G-75 chromatography also failed to separate the products of digestion of the fusion protein (data not shown), but affinity chromatography on asialofetuin- agarose was more successful. The elution profile (not shown) was typical of that observed for the purification of natural galectin from cell extracts [12], with a larger protein peak unretarded by the matrix, and a smaller protein peak retained and eluted with 100 mM lactose. All three bands were observed in the material bound to asialofetuin-agarose, and eluted with lactose, but the 29 kDa and 37 kDa components were relatively minor (Fig. 2A, lanes a and b). Re-chromatography of the unretarded fraction on asialofetuin-agarose under the same conditions resulted in a purer galectin-1 preparation (Fig. 2B, lane c). Amino acid sequence evidence, discussed below, indicates that most, if not all, of the 29 kDa material remaining in the retarded fraction after the first asialofetuin-agarose chromatography step was in fact identical in sequence to galectin- 1.

Two signals were clearly apparent when PTH amino acids were identified following automated sequential degradation of the 14 kDa galectin-1 band electroblotted from SDS-PAGE. The major signals (approx. 78%, on the

basis of yields of PTH-amino acids) corresponded to a sequence gly-ser-met-ala-(-)-gly-leu-val (the fifth residue was unassigned), and the minor signals (ca. 22%) to phe-asn-ala-his-gly-asp-ala-asn. The N-terminal sequence of human galectin-1 has been predicted from the cDNA to be met-ala-cys-gly-leu-val- [13], and the additional N- terminal gly-ser- is expected from the sequence of the p-GEX linker region [21]. Amino acid sequencing was also carried out on material electroblotted from 29 kDa bands, to check the identity of this material. When the sample used was from the unbound protein washed off an asialofe- tuin column prior to the 100 mM lactose elution (i.e. the material from the 29 kDa band in lane (c) or (d) of Fig. 2 A), the major signal ( > 85%) corresponded to met-(-)- pro-ile-leu-gly-tyr, with the second residue not assigned; the sequence of glutathione-S-transferase is met-ser-pro- ile-leu-gly-tyr [27]. When the 29 kDa material from lane (a) of Fig. 2 B was electroblotted and sequenced, the only detectable sequence was gly-ser-met-ala-(-)-gly-leu, indi- cating that this corresponded to galectin-1, presumably running in SDS-PAGE as a dimer.

Erythrocyte agglutination experiments were carried out, starting with a 1 mg • ml-1 solution of the fusion protein. Agglutination was clearly apparent in a series of threefold dilutions, down to a concentration of 1.3 /xg. m1-1, both in the presence and absense of fl-mercaptoethanol. The presence of lactose (100 mM) completely inhibited the agglutination reaction. The purified recombinant galectin was also an effective agglutinin at a concentration of 1.3 /xg • ml-~. In control experiments, GST caused agglutina- tion at a concentration of 1 mg • m l - l , but was ineffective at lower concentrations. This was judged to be due to non-specific binding of dimeric GST to the erythrocytes.

The growth-inhibitory effect of galectin-1 is demon- strated graphically by Fig. 3, for which HEP 2 cells were used. Control experiments were also carried out with the GST-galectin fusion protein, and with GST itself, and neither protein had any discernable effect upon cell growth at concentrations equivalent to those of galectin-1. It is thus clear that the effect of galectin-1 is not due to residual GST or fusion protein in the galectin-1 preparations. The same results were found in similar control experiments with MPS fibroblasts (not shown); the growth-inhibitory effect of galectin-1 on these cells is summarised by Fig. 4A. The fusion protein is not a growth inhibitor in the same concentration range, nor is unmodified GST (data not shown). It is clear that the inhibitory effect of galectin-1 is not substantially inhibited by 100 mM lactose. Quantitative data for the effect of galectin-1 on HEP 2 and other tumour cell lines is summarised in Fig. 4B. Inhibition of the growth of HEP2 and U 2 0 S osteosareoma cells by galectin-1 is clearly significant, whereas inhibition of growth of HELA cells is not.

In some experiments, the inhibitory effect of galectin-1 was clearly lower at 500/zg • ml - 1 than at 166 /zg. ml - i. This phenomenon has not as yet been investigated more

L. Adams et al. / Biochimica et Biophysica Acta 1312 (1996) 137-144 141

Fig. 3. Inhibition of human turnout cell growth by recombinant galectin-1. Sparse cultures of human HEP 2 cells were treated with the GST-galectin fusion protein (left-hand rows), with cleaved and purified recombinant galectin-1 (centre), or with recombinant GST (right). Duplicate wells were treated at each concentration, starting at 500 /zg. ml-* in the case of galectin 1 (topmost pairs of wells, arrowed), with descending tripling dilutions (166, 55, 18, 6 and 2 /~g .ml- I ) . The corresponding wells treated with the fusion protein and with GST were at 1500, 500, 166, 55, 18 and 6 / z g - m l -~ , in each case. The lowest pairs of wells were a control with normal medium in each case. The protein was dissolved in normal growth medium in each case, and 100 /~1 was added to replace the medium on the cells 24 h after plating. After 96 h, the cells were fixed with neutral formalin, and stained with Giemsa reagent.

fully because of limited availability of the recombinant galectin, but may represent some form of galectin receptor down-regulation.

Table 1 Mitogenicity of recombinant galectin-1

Agonist Concn. ( /~g.ml - t )

[3 H]Thymidine incorporation

(relative to (relative to control) maximum)

(%)

none 10% foetal calf serum Galectin-1

Galectin- 1 ( + 100 mM lactose)

1.0 ( + 0 . l ) 4.8 (+0 .4)

0.2 1.6 ( + 0.2) 76 1.0 2.1 (5:0.2) 100 5.0 1.9 (+0 .1) 90 25.0 1.5 ( + 0 . l ) 7l 125 0.9 (5:0.1) 43 5.0 0.9 (+0 .1) 43

25.0 0.7 (-L- 0.1) 33

Incorporation of [3H]thymidine into a TCA-insoluble form was measured in confluent MPS human fibroblast cultures in 24-well culture plates. Incorporation (the mean of four experimental determinations, + SEM) is expressed as a ratio with respect to the corresponding mean in serum-free controls, or is normalised relative to the maximal incorporation seen with galectin-l. Foetal calf serum was used as a positive control.

The mitogenic effect of galectin- 1 on MPS fibroblasts is summarised in the Table 1. The maximal observed stimula- tion in thymidine incorporation is relatively modest (29%) compared to that seen with foetal calf serum. This activity is discernable at galectin concentrations at and above 0.2 /xg • m1-1, is maximal around 1 tzg • ml- t , and begins to

A B

2

120

100

80

60

40

2O

140

120

e~ 100

¢~ 80

6O

4 0

20 i i i i i

10 40 160 (>4(I 10 40 160 640

protein concentration (p.g.ml-*) protein conccntratioa Otg.ml -t)

Fig. 4. Summary of cell growth inhibition by recombinant galectin-1. Cell cultures were established and treated with recombinant galectin-1 or other proteins as outlined in the legend to Fig 3. Fixed and stained cells in microwells were measured with an ELISA spectrophotometer equipped with a 540-nm filter. Cell growth is expressed as the percentage of the staining intensity seen with corresponding normal controls. Each point represents the mean of at least four determinations (5: S.E.M.), but error bars are included in only one data set for clarity. A. • MPS human fibroblasts treated with galectin-1. [] MPS fibroblasts treated with GST-galectin fusion protein. • MPS fibroblasts treated with galectin-I and 100 mM lactose. B. • HEP 2 carcinoma cells treated with galectin-l. [] HEP 2 cells treated with fusion protein. © U2 OS osteosarcoma cells treated with galectin-1. • HELA carcinoma cells treated with galectin-1.

142 L. Adams et al. /Biochimica et Biophysica Acta 1312 (1996) 137-144

decline at higher concentrations. It is clearly inhibited by lactose.

The question of possible cytotoxic effects of galectin-1 has been addressed. In experiments on MPS fibroblast proliferation, some cell detachment from the substratum was observed in culture wells treated with galectin-1. When this reagent was added to newly confluent MPS cultures for periods of 48 h, at the same concentrations as used to demonstrate inhibition of cell growth, there was no observable cell detachment, and the viability of the cells was unaffected. Trypan Blue exclusion indicated viabilty of > 98% in both test and control culture wells.

4. Discussion

The recombinant protein is clearly a lectin, as demon- strated by the haemagglutination reaction. This reaction is inhibited by lactose, which is consistent with the known specificity of galectin-1. It is noteworthy that the GST- galectin fusion protein is also a functional lectin. The dependence of haemagglutination on thiol reduction, gen- erally considered characteristic of galectins [28], was not observed in these experiments.

Amino acid sequencing confirmed that the recombinant protein was indeed derived from the galectin-1 gene. Apart from the non-assigned residue at cycle 6, the sequence was identical to that predicted for galectin-I [13], with an N-terminal dipeptide extension derived from the p-GEX linker region. The sequencing experiments also provided some evidence for thrombin degradation of the recombi- nant protein during cleavage of the fusion protein; the minor sequence observed corresponds to the eight amino acid residues C-terminal to arginine-48 in the galectin-1 sequence. This minor component was inseparable from the intact recombinant galectin by SDS-PAGE, so it seems likely that the N-terminal peptide generated by thrombin cleavage at arg-48 remains attached to the rest of the protein, perhaps due to the intramolecular disulphide bonds in galectin-1 [29]. There was also some evidence for the survival in SDS-PAGE of galectin-1 dimers.

Recombinant galectin-1 can act as a growth inhibitor for human fibroblasts, and for some human tumour cells. The GST-galectin fusion protein does not share this prop- erty, but the haemagglutination titres observed with the fusion protein and with the purified recombinant galectin were quantitatively similar. Taken together, these obseva- tions are a strong indication that growth inhibition does not depend upon the carbohydrate binding site. The anomalous mobility on SDS electrophoresis may indicate unusual structural interactions within the fusion protein, which may disrupt the site responsible for cell growth inhibition, but not the carbohydrate-binding site. The mobility in SDS- PAGE of the fusion protein corresponded to a kDa of about 37, whereas a value of about 43-44 kDa would be

expected from the known molecular weights of the two components.

In comparing the experiments on thymidine incorpora- tion and growth inhibition, it should be realised that inhibi- tion of thymidine incorporation is measured relative to baseline incorporation in serum-free medium in the ab- sence of mitogenic stimulus, whereas inhibition of cell growth is measured against a background of normal growth in complete medium. The mitogenic activity of galectin-1 is seen in a relatively low concentration range, and is dependent upon the fl-galactoside binding site, as it is inhibited by lactose. The growth-inhibitory property of the lectin is only apparent in a higher concentration range, and is largely unaffected by lactose. This offers further confir- mation that growth inhibition is not primarily due to the fl-galactoside binding site. Galectin-1 is apparently not cytotoxic.

Experiments with recombinant murine galectin-1 ex- pressed in a eukaryotic COS-I cell system suggest that it is not a functional lectin, by virtue of a endogenous glycan component which binds non-covalently to the fl-galacto- side binding site. In the absence of the glycan component, the protein is less effective as a growth inhibitor [30]. The physiological significance of the endogenous glycan com- ponent is unclear. Obviously, it is absent from bacterially- expressed galectins, but the intrinsic lectin activity which is the basis for identification of most natural mammalian isolates suggests that it may not always, or even usually, be present on these proteins. A novel and intriguing possi- bility is that the glycan is itself a modulator of galectin growth-modulatory activity, with the potential to minimise non-productive or mitogenic interactions, and to maximise growth-inhibitory actions.

The murine recombinant galectin-1 was considerably more effective as a growth inhibitor, but direct comparison is difficult because this protein was expressed in eukary- otic cells and tested on murine fibroblasts [11]. Our bacte- rially-expressed galectin-1 is apparently 5-10-fold less effective as a growth inhibitor than the natural human protein [12]. This difference is clearly not connected with the presence of a glycan similar to that reported by Wells and Mallucci [30], as both the natural and recombinant human proteins have been isolated on the basis of their affinity for fl-galactosides. The partial proteolysis of the galectin, already discussed above, may contribute to this effect, but there must also be some other more important cause which has not yet been identified. One possibility is that the residual linker amino acids from the fusion protein also reduce growth-inhibitory activity. Yields of natural galectin from conditioned cell culture medium were so low [12], that it was not practicable to make a more extensive comparison between it and the bacterial recombinant pro- tein. We cannot, on the basis of the evidence obtained with this recombinant form of galectin-1, accurately assess the physiological significance of the biphasic cellular re-

L. Adams et al. /Biochimica et Biophysica Acta 1312 (1996) 137-144 143

sponse, as the concentration response range may be differ- ent from that with the natural protein.

An autocrine growth factor from a transformed rat cell line has been claimed to be related to rat lung galectin-1 [31], and overexpression of rat galectin-1 cDNA in mouse fibroblasts resulted in transformation. There was a strong clonal correlation between transformation and the appear- ance of agglutinating activity in the culture medium. /3- galactosides which inhibited the agglutination did not af- fect autocrine stimulation of DNA synthesis [32]. The mitogenic mechanism was thus apparently not due to lectin activity, as has been the case in most studies on galectins so far.

The concept of a bioactive peptide with both growth- stimulatory and growth-inhibitory activities is not new. It has been well-established that transforming growth factor /3 (TGF/3) has a biphasic effect on some cell types, and a model, combining this concept with the chemotactic prop- erty of TGF/3, has been proposed to account for sequential cell migration, proliferation and differentiation in response to this growth factor [33]. Using natural galectin-1, we have previously observed a growth-stimulatory effect on bovine corneal endothelial cells [12], suggesting that these cells have a different concentration response range for galectin, and responded positively at concentrations which caused a negative effect in the other human cells.

Changes in the cellular levels of galectin-1 and anti- genically-related proteins have been linked to mitogenic stimulation and tumour transformation in human leuko- cytes [34]. A recent report confirms that the anti-tumour action of sodium butyrate is linked to the induction of galectin-1 gene expression in human coloncarcinoma cells, and the subsequent re-differentiation of the cells [35]. It may thus be significant that we were unable to detect galectin-1 gene expression in HEP 2 cells. Since these cells still respond to the growth-inhibitory action of galectin-1, their apparent failure to synthesise the lectin may contribute to their transformation. However, in sev- eral other cell types, galectin-I expression is increased following tumour transformation. The concomitant eleva- tion of expression of glycoproteins able to bind to galectin-1 may be significant in this context [34-37].

In our earlier experiments with natural human galectin- 1, we observed inhibition of the growth of human fibroblasts and of U2 OS osteosarcoma cells, but not of HELA carcinoma cells. We attributed the apparent failure of galectin-1 to inhibit the growth of HELA cells as circum- stantial support for the hypothesis that continuous galectin degradation by the GRP occurs in normal cells, and is necessary for cell growth [12], since inhibition of the GRP does not inhibit HELA cell growth [4,6]. Growth inhibition by proteinase inhibitors is also seen with HEP2 cells [6], which nevertheless are subject to growth inhibition by galectin-1. The relationship between GRP action and the role of galectin-1 is cle~.rly not as simple as we have previously postulated.

The present report provides some further evidence to suggest that galectin-1 is particularly susceptible to prote- olysis, but the possible effects of thrombin or E. coli proteinases may not equate with those of human cell- surface enzymes, and direct experimental investigation of galectin-I degradation by cultured cells is an obvious extension of this work. Sensitivity to trypsin and staphylo- coccal V8 proteinase degradation has been observed with recombinant bovine galectin-1 synthesised in vitro [38].

At first sight, the activity of galectin-1 as an inhibitor of tumour cell growth has obvious potential for application. However, galectin-1 has both positive and negative growth-regulatory properties, depending on circumstances. Production of recombinant galectin-1 in a eukaryotic ex- pression system may minimise or eliminate this difficulty, by the generation of galectin-glycan conjugates, as dis- cussed above, but the nature of the galectin-glycan interac- tion requires further study. An alternative to the use of such conjugates could result from site-directed mutagene- sis to inactivate the fl-galactoside binding site, and consid- erable progress has already been made towards this aim [38]. A recombinant version of galectin-1, lacking /3- galactoside binding activity but retaining its growth-inhibi- tory site, could be suitable for therapeutic use. Site-di- rected mutagenesis may also be of value in the creation of proteinase-resistant forms of this cell growth effector.

Acknowledgements

We wish to thank the Cancer Society of New Zealand and the University of Auckland Research Fund for finan- cial support, and the Auckland Medical Research Founda- tion for a scholarship (to C.W.). We also wish to thank Ms Catriona Knight for the protein sequencing and oligo- nucleotide synthesis, lain MacDonald for photographic assistance and several colleagues, in particular Drs David Christie and Ruth Mould, for advice and assistance in establishing the molecular biological techniques.

References

[1] Scott, G.K. (1987) Proteinases and eukaryotic cell growth. Comp. Biochem. Physiol. 87B, 1-10.

[2] Scott, G.K. (1992) Proteinases and proteinase inhibitors as modula- tors of animal cell growth. Comp. Biochem. Physiol. 103B, 785-793.

[3] Allen, R.J., Rattray, S. and Scott, G.K. (1981) Preliminary evidence that thrombin may mimic a naturally occurring proteinase in cul- tured cells. Biosci. Rep. 1, 881-884.

[4] Pitts, J.D. and Scott, G.K. (1983) Growth inhibition of normal, tumour and transformed cells by antibodies to a cell-surface pro- teinase. Biosci. Rep. 3, 47-51.

[5] Scott, G.K. and Seow, H.F. (1985) Further evidence for a cell surface proteinase essential to the growth of cultured fibroblasts. Exp. Cell Res. 158, 41-52.

[6] Scott, G.K. (1988) Inhibition of plasminogen activators and the growth of cultured human tumour cells. Int. J. Biochem. 20, 817- 822.

144 L. Adams et al. / Biochimica et Biophysica Acta 1312 (1996) 137-144

[7] Borich, S.M., Englebretsen, D., Harding, D.R.K. and Scott, G.K. (1994) The thrombin receptor and endogenous growth-related prote- olysis in human fibroblasts. Cell Biol. Intl. 18, 639-646.

[8] Scott, G.K., Seow, H.F. and Tse, C.A. (1989) Investigations into the nature of growth- related proteolysis in human fibroblasts. Biochim. biophys. Acta. 1010, 160-165.

[9] Strobel-Stevens, J.D. and Lacey, J.C. (1981) Further evidence for an inhibitor of proliferation elaborated by normal human fibroblasts in culture: partial characterisation of the inhibitor. J. cell. Physiol. 106, 201-207.

[10] Wells, V. and Mallucci, L. (1983) Properties of a cell growth inhibitor produced by mouse fibroblasts. J. cell. Physiol. 117, 148- 154.

[11] Wells, V. and Mallucci, L. (1991) Identification of an autocrine negative growth factor: mouse /3-galactoside-binding protein is a cytostatic factor and cell growth regulator. Cell 64, 91-97.

[12] Manilal, S.B., Scott, G.K. and Tse, C.A. (1993) Inhibition of an endogenous growth- related proteinase enhances the recovery of a negative growth regulator from cultured human cells. Cell Biol. Intl. 17, 317-323.

[13] Abbott, W.M. and Feizi, T. (1989) Evidence that the 14 kDa soluble /3-galactoside-binding lectin in man is encoded by a single gene. Biochem. J. 259, 291-294.

[14] Barondes, S.H., Castronovo, V., Cooper, D.N.W., Cummings, R.D., Drickamer, K., Feizi, T., Gitt, M.A., Hirabayashi, J., Hughes, C., Kasai, K., Leffler, H., Liu, F.T., Lotan, R., Mercurio, A.M., Mons- gny, M., Pillai, S., Poire, F., Raz, A., Rigby, P.W.J., Rini, J.M. and Wang, J.L. (1994) Galectins: a family of animal /3-galactoside-bind- ing lectins. Cell 76, 597-598.

[15] Lipsick, J.S., Beyer, E.C., Barondes, S.H. and Kaplan, N.O. (1980) Lectins from chicken tissues are mitogenic for thy-1 negative murine spleen cells. Biochem Biophys. Res Commun. 97, 56-61.

[16] Pitts, M.J. and Yang, D.C.H. (1981) Mitogenicity and binding properties of b-galactoside-binding lectin from chick-embryo kidney. Biochem. J. 195, 435-439.

[17] Sanford, G.L. and Harris-Hooker, S. (1990) Stimulation of vascular cell proliferation by /3-galactoside specific lectins. FASEB J. 4, 2912-2918.

[18] Scott, G.K. and Tse, C.A. (1994) Changes in sensitivity of human tumour cells to growth inhibition by proteinase inhibitors. Cell Biol. Intl. 18, 89-93.

[19] Scott, G.K. and Tse, C.A. (1988) Modulation of cell proliferation by protein proteinase inhibitors; a new analytical approach. Biol. Chem. Hoppe-Seyler 369, Suppl.5, 131-135.

[20] Hamilton, I., Reynolds, G.W., Scott, G.K., Sharfe, N. and Tse, C.A. (1990) Effects of human and ovine pancreatic secretory trypsin inhibitors on the proliferation of normal human fibroblasts. Biol. Chem. Hoppe-Seyler, 371, 78-83.

[21] Guan, K.L. and Dixon, J.E. (1991) Eukaryotic proteins expressed in Escherichia coli: an improved thrombin cleavage and purification procedure of fusion proteins with glutathione-S-transferase. Anal. Biochem. 192, 262-267.

[22] Sambrook J., Fritsch F., Maniatis T. (1989) Molecular cloning; a laboratory manual, (2nd ed.) Cold Spring Harbor Laboratory Press.

[23] Gough, N.M. (1988) Rapid and quantitative preparation of cytoplas- mic RNA from small numbers of cells. Anal. Biochem. 173, 93-95.

[24] Bradford, M.M. (1976) A rapid and sensitive method for the quanti- tation of microgram quantities of protein utilising the principle of protein-dye binding. Anal. Biochem. 72, 248-252.

[25] Lis, H. and Sharon, N. (1972) Soy bean (Glycine max) agglutinin. Methods Enzymol. 28, 360-368

[26] Nowak, T.P., Kobiler, D., Roel, L.E., and Barondes, S.H (1977) Developmentally regulated lectin from embryonic chick muscle. J. Biol. Chem. 252, 6026-6030

[27] Smith, D.B., Davern, K.M., Board, P.G., Tiu, W.U., Garcia, E.G. and Mitchell, G.F. (1986) Mr 26, 000 antigen of Schistosoma japonicum recognised by resistant WEHI 129/J mice is a parasite glutathione S-transferase. Proc. Natl. Acad. Sci. U.S.A. 83, 8703- 8707.

[28] Barondes, S.H. (1984) Soluble lectins: a new class of extracellular proteins. Science 223, 1259-1264.

[29] Tracey, B.M., Feizi, T., Abbott, W.M., Carruthers, R.A, Green, B.N. and Lawson, A.M. (1992) Subunit molecular mass assignment of 14, 654 Da to the soluble fl-galactoside-binding lectin from bovine heart muscle and demonstration of intramolecular disulfide bonding asso- ciated with oxidative inactivation. J. Biol Chem. 267, 10342-10347.

[30] Wells, V. and Mallucci, L. (1992) Molecular expression of the negative growth factor marine fl-galactoside binding protein (mGBP). Biochim. Biophys. Acta 1121, 239-244.

[31] Yamaoka, K., Hirai, R., Tsugita, A. and Mitsui, H. (1984) The purification of an acid and heat labile transforming growth factor from an avian sarcoma virus-transformed rat cell line. J. Cell. Physiol. 119, 307-314.

[32] Yamaoka, K., Ohno, S., Kawasaki, H. and Suzuki, K. (1991) Overexpression of a /3- galactoside binding protein causes transfor- mation of BALB3T3 fibroblast cells. Biochem. Biophys. Res. Com- mun. 179, 272-279.

[33] Moses, H.L., Yang, E.Y. and Pietenpol, J.A. (1990) TGF/3 stimula- tion and inhibition of cell proliferation: new mechanistic insights. Cell 63, 245-247.

[34] Carding, S.R., Thorpe, S.J., Thorpe, R. and Feizi, T. (1985) Trans- formation and growth related changes in levels of nuclear and cytoplasmic proteins antigenically related to mammalian galacto- side-binding lectin. Biochem. Biophys. Res. Commun. 127,680-686.

[35] Ohannesian, D.W., Lotan, D. and Lotan, R. (1994) Concomitant increases in galectin-1 and its glycoconjugate ligands (carcinoembryonic antigen, lamp-1 and lamp-2) in cultured human colon carcinoma cells by sodium butyrate. Cancer Res. 54, 5992- 6000.

[36] Raz, A., Avivi, A., Pazerini, G. and Carmi, P. (1987) Cloning and expression of cDNA for two endogenous UV2237 fibrosarcoma lectin genes Exp. Cell Res. 173, 109-116.

[37] Chiariotti, L., Berlingieri, M.T., De Rosa, P., Battaglia, C., Berger, N., Bruni, C.B. and Fusco, A. (1992) Increased expression of the negative growth factor, galactoside-binding protein, gene in trans- formed thyroid cells and in human thyroid carcinomas. Oncogene 7, 2507-2511.

[38] Abbott, W.M. and Feizi, T. (1991) Soluble 14 kDa b-galactoside- specific bovine lectin. J. Biol. Chem. 266, 5552-5557.