Purification and characterization of a novel restricted antigen expressed by normal and transformed human colonic epithelium.

A cell surface antigen that is expressed by normal and 95% of transformed colonic epithelium and is recognized by the monoclonal antibody A33 (Welt, S., Divgi, C. R., Real, F. X., Yeh, S. D., Garin-Chesa, P., Finstad, C. L., Sakamoto, J., Cohen, A., Sigurdson, E. R., Kemeny, N., Carswell, E. A., Oettgen, H. F., and Old, L. J. (1990) J. Clin. Oncol. 8, 1894-1906) has been purified to homogeneity from the human colonic carcinoma cell line LIM1215. The A33 protein was purified from Triton X-114 extracts of LIM1215 cells under nondenaturing conditions. These extracts were applied sequentially to Green-Sepharose HE-4BD, Mono-Q HR 10/10, Superose 12 HR 10/30, and micropreparative Brownlee Aquapore RP 300. The purification was monitored by biosensor analysis using surface plasmon resonance detection with a F(ab′)2 fragment of the humanized A33 monoclonal antibody immobilized on the sensor surface and Western blot analysis following SDS-polyacrylamide gel electrophoresis (PAGE) under nonreducing conditions using humanized A33 monoclonal antibody. The purified A33 antigen has a Mr on SDS-PAGE of 43,000 under nonreducing conditions. By contrast, the purified protein displayed a Mr of approximately 180,000 under native conditions on both size exclusion chromatography and native PAGE, possibly due to the formation of a homotetramer. N-terminal amino acid sequence analysis of the purified protein identified 34 amino acid residues of a unique sequence: ISVETPQDVLRASQGKSVTLPXTYHTSXXXREGLIQWD. A polyclonal antibody was raised against a synthetic peptide corresponding to residues 2-20 of this sequence. The antipeptide serum recognized the purified protein using Western blot analysis under both nonreducing (Mr 43,000) and reducing (Mr 49,000) conditions.

methods for effective treatment of metastatic colorectal cancer. Passive cancer immunotherapy, using monoclonal or genetically engineered antibodies to deliver radioisotopes to tumor cells, can complement the effect of conventional treatments (surgery, chemotherapy, and radiotherapy), and a large number of monoclonal antibodies have already been administered to cancer patients following this approach.
The monoclonal antibody (mAb) 1 A33 detects a cell surface antigen expressed by 95% of primary or metastatic colon cancer cells and normal colonic epithelium but not by most other normal tissues and tumor types (2,3). Because of its restricted pattern of expression, the A33 antigen can be classified as a tissue-specific antigen for both normal and transformed colon, rectal, and small intestinal epithelium. Some human colon cancer cell lines express high levels of the A33 antigen, binding up to 800,000 molecules/cell. 2 The A33 antigen is not secreted or shed, and cell-bound radiolabeled A33 mAb is rapidly internalized into antigen-positive cells. Phase 1 quantitative dosimetry and human biodistribution studies demonstrated the tumor targeting capabilities of A33 mAb (2). Phase I and II radioimmunotherapy studies with 131 I-and 125 I-labeled A33 mAb further demonstrated the localization capabilities of this antibody, and, in addition, some antitumor effects were observed in these heavily pretreated patients (3,4). These studies have also shown that, although the isotope is rapidly cleared from the normal colon (5-6 days), it is retained for long periods (up to 6 weeks) in the tumor lesions, providing a rationale for radioimmunotherapy studies. The antibody has now been humanized (5), and studies are commencing using this reagent in multiple phase I trials.
However, despite extensive immunochemical, immunohistochemical, and clinical studies, the antigen for the A33 mAb has not previously been identified. Our initial attempts to purify the A33 mAb antigen by immunochromatography were hindered by low yields and leakage of the antibody from the affinity matrix. Furthermore, nonspecific interaction between the Fc domain of the murine A33 mAb and actin was observed. Therefore, a multidimensional chromatographic protocol was developed for the purification of the protein from the human colonic carcinoma cell line LIM1215 (6), which expresses significant levels of A33 antigen, as demonstrated by both immu-nocytochemistry and flow cytometry analysis.
The A33 antigen was extracted from LIM1215 cells by Triton X-114 phase partitioning (7) and purified by sequential use of ligand dye (Green-Sepharose HE-4BD), anion exchange (Mono-Q HR 10/10), size exclusion (Superose 12 HR 10/30), and micropreparative reversed-phase (RP) HPLC (Brownlee Aquapore RP 300; 30 ϫ 2.1-and 100 ϫ 1-mm internal diameter (ID)). The chromatographic purification was monitored by biosensor analysis (8) using surface plasmon resonance detection (9) with a F(abЈ) 2 fragment of the humanized A33 mAb immobilized onto the sensor surface and Western blot analysis using the A33 mAb under nonreducing conditions. N-terminal amino acid sequence analysis of the A33 antigen identified 34 amino acid residues of unique sequence, giving evidence of expression of a novel membrane protein expressed by normal and transformed human colonic epithelium.

MATERIALS AND METHODS
Triton X-100, Triton X-114, CHAPS, Tris, sodium chloride, sodium dihydrogen orthophosphate, glycine, leupeptin, aprotinin, phenylmethylsulfonyl fluoride, HEPES, EDTA, and actin were purchased from Sigma. Polyvinylidene difluoride membranes, Affi-Gel 10, and goat anti-mouse and anti-human horseradish peroxidase-conjugated antibodies were purchased from Bio-Rad. Prestained molecular weight standards were purchased from Bio-Rad and Novex (San Diego, CA). The fluorescein-conjugated anti-murine IgG was purchased from Silenus (Melbourne, Australia). Tween 20, trifluoroacetic acid, and a Micro BCA protein reagent kit were purchased from Pierce. An enhanced chemiluminescence Western blotting detection system was purchased from Amersham Corp. Biosensor reagents were from Pharmacia Biotech Inc.

Cell Culture Techniques
The LIM1215 colonic carcinoma cell line was grown in RPMI 1640 medium containing 10% fetal calf serum. Confluent cells (2.2 ϫ 10 5 /cm 2 ) were passaged using trypsin-Versene solution. Cells were seeded 1/10 into tissue culture dishes (150 ϫ 20 mm; Nuclon, Roskilde, Denmark) containing 25 ml of RPMI 1640 medium supplemented with 10% fetal calf serum, 1 g/ml hydrocortisone, 0.025 unit/ml insulin, and 10.82 g/ml ␣-thioglycerol. Dishes were incubated at 37°C in an atmosphere of 5% CO 2 for 5 days. After removing the media, cells were washed with phosphate-buffered saline (PBS) before being removed from the surface using a cell scraper (Costar Corp., Cambridge, MA). Cells were washed in PBS and resuspended at 10 9 cells/ml.

Flow Cytometric Analysis
A33 antigen expression on the surface of the LIM1215 cell line was analyzed by flow cytometry following standard techniques (10). The Hep-2 epidermoid carcinoma cell line (11) was used as a negative control. The cells were washed and resuspended at 5 ϫ 10 6 cells/ml in 500 l of PBS containing 5 mM EDTA and 5% fetal calf serum. The cells were incubated with 5 g of murine A33 mAb for 30 min at 4°C. After washing with buffer, the cell-antibody complex was incubated with fluorescein-conjugated anti-murine IgG (1/50 dilution). The negative control was performed by staining the cells with an isotypically matched nonrelated antibody (5 g), followed by fluorescein-conjugated anti-murine IgG alone. Flow cytometry was performed using a FACScan flow cytometer (Becton Dickinson, San Jose, CA).

Extraction of A33 Antigen from LIM1215 Cells with
Triton X-100 or Triton X-114 LIM1215 colonic cells (1.6 ϫ 10 9 cells) were harvested, washed in PBS, and solubilized (10 8 cells/ml) for 30 min at 4°C with either 0.3% (v/v) Triton X-100 or 1% (v/v) Triton X-114 in 15 mM Tris-HCl, pH 7.4, containing 1 mM phenylmethylsulfonyl fluoride, 1 mM pepstatin, 0.1 mM leupeptin, and 0.01 unit/ml aprotinin. The resulting extracts were centrifuged twice at 4°C for 20 min at 14,000 ϫ g. The Triton X-100extracted supernatant was taken directly for Green-Sepharose HE-4BD chromatography (see below). The Triton X-114-extracted supernatant was layered over 6% sucrose in 15 mM Tris-HCl, pH 7.4, with 0.06% (v/v) Triton X-114 and containing the protease inhibitors listed above. The tubes were incubated at 37°C for 30 min and then centrifuged at 25°C for 15 min at 5,000 ϫ g. The detergent phase was collected for chromatographic purification.

Chromatographic Purification
Green-Sepharose Chromatography-Triton X-100 extract or the Triton X-114 detergent phase was diluted to a final concentration of 0.1% detergent and loaded at 4°C onto a Green-Sepharose HE-4BD column (100 ϫ 10 mm ID) connected to a fast protein liquid chromatography system (Pharmacia Biotech). The column was equilibrated with 10 mM Tris-HCl, pH 7.4, containing 0.1% (w/v) CHAPS. Bound proteins, including actin, were eluted stepwise with 1 M NaCl. The breakthrough contained the A33 antigen and was collected for anion exchange HPLC.
Anion Exchange HPLC-The Green-Sepharose breakthrough was injected at 4°C onto a Mono-Q HR 10/10 column previously equilibrated in 10 mM Tris-HCl, pH 7.4, containing 0.1% (w/v) CHAPS. The proteins were eluted from the column using a linear 0 -1 M NaCl gradient generated over 90 min at a flow rate of 1 ml/min. Fractions (1 ml) were collected automatically (FRAC 100, Pharmacia Biotech). Proteins were detected by absorbance at 280 nm. The A33 antigen in eluant fractions was detected using both biosensor analysis and Western blot analysis under nonreducing conditions, as described below.
Size Exclusion HPLC-The active fractions eluted from the Mono-Q column (10 ml) were concentrated 10-fold using a SpeedVac concentrator (Savant Instruments Inc., Farmingdale, NY), dialyzed against PBS containing 0.05% (w/v) CHAPS, and loaded at 4°C onto a Superose 12 HR 10/30 column. Proteins were eluted with PBS containing 0.05% (w/v) CHAPS at a flow rate of 0.5 ml/min. Fractions (0.5 ml) were collected. Proteins were detected at 280 nm. The A33 antigen was monitored by both Western blot and biosensor analysis.
Reversed-Phase HPLC Chromatography-Superose 12 active fractions (2.5 ml) were loaded at a flow rate of 1 ml/min by multiple 1-ml injections onto a Brownlee Aquapore RP 300 micropreparative RP-HPLC column (30 ϫ 2.1 mm ID) equilibrated with the primary solvent, 0.15% (v/v) trifluoroacetic acid in water. The proteins were eluted with a linear 60-min gradient to 60% aqueous n-propyl alcohol/0.125% (v/v) trifluoroacetic acid at a flow rate of 100 l/min, The column temperature was 45°C. Protein detection was performed at 215 nm. The A33 antigen was detected using both biosensor and Western blot analysis. Fractions containing the A33 antigen were repurified and further concentrated using a Brownlee Aquapore RP 300 micropreparative RP-HPLC column (100 ϫ 1 mm ID) prior to N-terminal sequence analysis (12), using the gradient conditions described above at a flow rate of 50 l/min. Eluant fractions were recovered manually.

Biosensor Analysis
Cell extracts and chromatographic fractions were monitored using an instrumental optical biosensor (BIAcore, Pharmacia Biosensor), with an F(abЈ) 2 fragment of humanized A33 mAb immobilized onto the biosensor surface using N-hydroxysuccinimide and ethyl-NЈ-dimethylaminopropyl-carbodiimide at a flow rate of 4 l/min as described previously (13). Antigen binding to the F(abЈ) 2 fragment is detected by surface plasmon resonance, which measures small changes in refractive index at, or near, the gold sensor surface (8,9). Prior to biosensor assay, cell extracts or aliquots of chromatographic fractions were diluted to 100 l final volume in BIAcore buffer (10 mM HEPES, pH 7.4, containing 3.4 mM EDTA, 0.15 mM NaCl, and 0.005% Tween 20). Samples (30 l) were injected over the sensor surface at a flow rate of 5 l/min. Following completion of the injection phase, dissociation was monitored in BIAcore buffer at the same flow rate for 360 s. Residual bound antigen was eluted, and the surface regenerated between injections using 40 l of 10 mM NaOH. This treatment did not denature the protein immobilized onto the sensor surface, as shown by equivalent signals on reinjection of a sample containing the A33 antigen.

Western Blot Analysis
Electrophoresis and Western blot analysis were performed on precast Phastgels using a Phastsystem separation and control unit (Pharmacia Biotech). Cell extracts and chromatographic fractions were electrophoresed under nonreducing conditions (14) on 8 -25% SDS-PAGE Phastgels or 8 -25% native Phastgels and transferred onto polyvinylidene difluoride membranes and incubated with murine or humanized A33 mAb. RP-HPLC-purified A33 antigen was also analyzed by Western blot under nonreducing and reducing conditions using polyclonal anti-N-terminal peptide antibodies. IgG binding was probed with horseradish peroxidase-labeled goat anti-mouse IgG, goat anti-human IgG, or goat anti-rabbit IgG and detected by enhanced chemiluminescence.

N-terminal Amino Acid Sequence Analysis
N-terminal amino acid sequence analysis of purified A33 protein was performed on a Hewlett-Packard G1005A protein sequencer operated with the routine 3.0 sequencer program described previously (14).

Polyclonal Antipeptide Antibody Production and Purification
Antibodies were generated in New Zealand White rabbits using as immunogen a chemically synthesized peptide corresponding to part of the N-terminal sequence of the A33 antigen described herein (amino acid residues 2-20) conjugated to keyhole limpet hemeocyanin (Quality Controlled Biochemicals Inc., Hopkinton, MA) in complete Freund's adjuvant/incomplete Freund's adjuvant. IgG was purified from immune sera by protein A affinity chromatography. Purified IgG was analyzed for reactivity with LIM1215 cell lysates and purified A33 antigen by SDS-PAGE and Western blot analysis.

Protein Quantitation
Protein concentration was determined using the bicinchoninic acid protein assay (16) or from the ratio of A 280 /A 260 (17).

RESULTS AND DISCUSSION
The colonic carcinoma cell line LIM1215 6 was shown by flow cytometry (Fig. 1) and immunocytochemistry and rosetting assays (data not shown) to strongly express the membrane antigen of the A33 mAb. Western blot analysis of a 0.3% Triton X-100 extract of LIM1215 cells under nonreducing conditions using both murine and humanized A33 mAb showed a major band with a M r of 43,000 ( Fig. 2A, lane 1). A M r 41,000 band was also recognized by Western blot analysis, as well as two other minor lower molecular weight bands (Fig. 2A, lane 1). Western blot analysis of a 0.3% Triton X-100 extract of Hep-2 cells, which were negative for A33 antigen expression by flow cytometry (Fig. 1), failed to demonstrate the M r 43,000 band but detected both the M r 41,000 and lower molecular weight bands (results not shown), suggesting that only the M r 43,000 protein was specifically recognized by the A33 mAb. The A33 mAb recognized a conformationally sensitive epitope, as evidenced by nonreactivity in Western blot analysis under reducing conditions. 3 A combined immunopurification and biosensor approach was initially designed in an attempt to purify the A33 ligand, using Triton X-100 extracts of LIM1215 cells as the starting material. The biosensor technology monitors protein-protein interactions in real time, using an optical detection principle based on surface plasmon resonance (9). In these initial experiments the murine A33 mAb was immobilized onto the surface of the sensor chip, whereas antigen-containing fractions were injected in a continuous flow over the surface. The surface plasmon resonance response reflects a change in refractive index, and hence a change in mass, at the detector surface as the antigen binds or dissociates from the immobilized antibody. For the immunochromatographic step, the murine A33 mAb was conjugated to Affi-Gel 10 (N-hydroxysuccinimide esters of a derivatized cross-linked agarose). By using this matrix, similar immobilization chemistry was used for both the biosensor analysis and affinity chromatography step, allowing the appropriate dissociation conditions for immunochromatography (10 mM NaOH) to be rapidly determined using biosensor analysis (18,19).
A major protein of M r 43,000 was immunopurified and found to be N-terminally blocked on amino acid microsequence analysis. Internal sequence analysis following "in gel" tryptic digestion (20) identified this protein as actin. Further biosensor studies confirmed this in vitro interaction between actin and immobilized A33 mAb. When a preparation of actin was injected over A33 mAb (Fig. 3) or nonrelated monoclonal antibodies of the same immunoglobulin subclass (IgG2a) immobilized onto the biosensor surface, a strong positive binding was observed. By contrast, when A33 F(abЈ) 2 fragments were immobilized, actin binding was no longer observed, suggesting that actin was binding to the Fc domain of the antibody (Fig. 3).
We therefore developed a multidimensional chromatographic protocol for the purification of the A33 mAb ligand. To avoid actin interaction during the screening of the chromatographic fractions, A33 F(abЈ) 2 fragments were immobilized onto the biosensor surface. Active fractions detected by biosensor analysis were confirmed positive by Western analysis following SDS-PAGE under nonreducing conditions.
The focus of the initial purification steps was to remove actin, which had been shown to be present in the cell extracts in significant quantities. Studies with Triton X-100 LIM1215 cell lysates showed that actin could be selectively separated from the A33 antigen by Green-Sepharose HE-4BD ligand dye chromatography. The A33 ligand did not bind to the Green-Sepharose matrix (Fig. 2A, lane 2), whereas actin was retained and could be eluted with 1 M NaCl (Fig. 2A, lane 3). Actin, purified from rabbit muscle, was weakly recognized on Western analysis with A33 mAb under nonreducing (a minor band of M r 41,000; Fig. 2A, lane 4) and reducing conditions (data not shown), confirming the cross-reactivity observed using the biosensor (Fig. 3).
Although the Green-Sepharose chromatography was effective for removal of actin, Triton X-114 extraction and phase separation was found to be preferable to Triton X-100 to obtain a detergent phase enriched in hydrophobic membrane proteins (Fig. 2B). The A33 antigen was detected in the Triton X-114 extract (Fig. 2B, lane 1) and in the detergent-rich phase following phase extraction (Fig. 2B, lane 3) but not in the aqueous phase (Fig. 2B, lane 2). The presence of the A33 antigen in the detergent-rich phase suggested that either the A33 antigen is an integral membrane protein or a glycosylphosphatidylinositol-anchored protein (21). The actin band of M r 41,000 and the other nonspecific low molecular weight components observed with Triton X-100 ( Fig. 2A, lane 1) were no longer detected using Triton X-114 extraction. The combination of Triton X-114 phase separation followed by Green Sepharose ligand dye chromatography to remove all traces of actin was therefore used as the primary purification step for the A33 antigen.
Injection of Triton X-114 extracts onto humanized A33 mAb immobilized to various chromatographic matrices (Affi-Gel 10, CNBr-activated Sepharose, and N-hydroxysuccinimide-activated Sepharose) resulted in the recovery of only very low amounts of a M r 43,000 protein. Furthermore, leakage of significant quantities of antibody from the supports (22) hampered the practical use of these methods to purify and identify the antigen. This, coupled with the fact that specific detection (biosensor or Western blot) was also based solely on affinity methods, led us to exclude an affinity step in the purification protocol.
The detergent phase of the Triton X-114 LIM1215 cell extracts, after passage through Green-Sepharose HE-4BD, was loaded onto a Mono-Q anion exchange HPLC column. Triton X-114 was exchanged with the zwitterionic detergent CHAPS after absorption of the detergent-membrane complex to the chromatographic support. CHAPS was found to greatly improve resolution on both the anion exchange and the subsequent size exclusion chromatography. Application of a 0 -1 M

FIG. 3. Biosensor analysis of the interaction between actin and either A33 IgG or the A33 F(ab) 2 fragment.
A preparation of rabbit muscle actin (0.3 g) was injected at a flow rate of 5 l/min over a sensor surface onto which either murine A33 IgG (upper trace) or A33 F(abЈ) 2 (lower trace) had been immobilized. Protein-protein interactions were monitored by surface plasmon resonance. At the end of the injection pulse, a signal of 247 RU was observed due to actin binding to A33 IgG, whereas the signal corresponding to actin binding to A33 F(abЈ) 2 was only 4 RU (arrows) .   FIG. 4. Anion exchange HPLC of the A33 antigen. Proteins contained in the Green-Sepharose breakthrough fraction were loaded onto a Mono-Q HR 10/10 anion exchange column and eluted at a flow rate of 1 ml/min with a linear NaCl gradient as indicated (---). One-ml fractions were collected, and aliquots (20 l) of each fraction were taken for biosensor assay. The M r 43,000 antigen was detected by Western blot analysis under nonreducing conditions (inset) in the labeled fractions. NaCl gradient resulted in the elution of at least 15 major protein peaks from the anion exchange column (Fig. 4). A strong biosensor signal (Ͼ200 RU) was registered when aliquots of fractions 2-11 were injected over the biosensor surface (Fig. 4). The presence of the M r 43,000 A33 antigen was confirmed by Western blot analysis (Fig. 4, inset) in fractions 2-9, eluting between 0.45 and 0.55 M NaCl. The subsequent fractions, eluting between 55 and 68 min, gave a small residual biosensor signal (approximately 100 RU) but were negative by Western blot analysis. The most active Mono-Q fractions (fractions 3-7) were pooled, concentrated 10-fold using a Savant SpeedVac concentrator, and further purified by size exclusion chromatography on Superose 12 HR 10/30 (Fig. 5). Interestingly, the M r 43,000 A33 mAb antigen was detected by both Western blot (Fig. 5, inset A) and biosensor analysis (Fig. 5) in early eluting fractions corresponding, by correlation with the retention time of protein standards of known molecular weight, to an apparent M r of 160,000 -200,000. Western analysis of the pooled Superose 12 active fractions (fractions 2-5), using a 8 -25% native Phastgel confirmed this result by revealing a protein migrating under native conditions (in the absence of SDS) with an apparent M r (23) of 180,000 (Fig. 5, inset B).
RP-HPLC of the Superose 12 active fractions (Fig. 5, fractions 1-5) on a Brownlee RP 300 micropreparative column (100 ϫ 2.1 mm ID) using n-propyl alcohol as the organic modifier allowed further purification of the A33 antigen (Fig. 6A). The M r 43,000 A33 antigen was found by biosensor analysis (Fig. 6, A and B) to be associated with a single symmetrical peak eluting from the RP-HPLC column between 45 and 48 min, demonstrating the highly hydrophobic character of this protein. Aliquots (2 l) of the active fractions were analyzed by SDS-PAGE with silver staining (Fig. 6A, inset A). Whereas fraction 3 contained traces of a higher molecular weight component, fractions 4 and 5 were essentially homogeneous and revealed a M r 43,000 protein, corresponding to the band recognized by A33 mAb using Western blot analysis (Fig. 6A, inset  B). The overall yield, based on comparison of the UV absorption at 215 nm with that of an ovalbumin standard, was approximately 1 g.
As we have noted previously (19), one of the advantages of using the biosensor to monitor chromatographic purification is that complimentary kinetic data describing the specific interaction can also be obtained simultaneously. Using nonlinear least squares regression analysis (24,25) of the biosensor curves shown in Fig. 6B, the apparent K D for the interaction between the purified A33 antigen and the immobilized A33 F(abЈ) 2 fragment was found to be in the low nanomolar range. The purified protein was subjected to N-terminal amino acid sequence analysis in which, in an initial experiment, a unique sequence of 33 amino acids was obtained: XSVETPQDVL-RASQGKSVTLPXTYHTSXXXREGLIQWD. The initial yield was approximately 10 pmol. This is the anticipated yield (50%) for the quantity loaded onto the sequencer (0.8 g after aliquots had been taken for Western blot and biosensor analysis (see Table I) and indicates that no N-terminally blocked proteins were present. In a subsequent experiment, in which purified protein from two preparations was loaded onto the sequencer, the initial yield was 20 pmol. In this experiment isoleucine was identified as the N-terminal residue. A sequence similarity search of the available protein and nucleotide data bases failed to reveal any significant sequence identity. In particular, we noted that the sequence obtained for the A33 antigen showed no significant similarity with the epithelial glycoprotein 40, FIG. 6. Micropreparative RP-HPLC purification of Superose 12 active fractions. A, Superose 12 active fractions were separated by RP-HPLC using a Brownlee RP 300 column (30 ϫ 2.1 mm ID) using the chromatographic conditions described under "Materials and Methods." One-minute fractions (100 l) were collected during the 60-min elution gradient. Aliquots (2 l) of each fraction were analyzed by SDS-PAGE (8 -25% gel, silver stained; inset A) and Western blot analysis under nonreducing conditions (inset B). B, biosensor analysis of fractions from micropreparative RP-HPLC. Aliquots (20 l) of each fraction were concentrated using a SpeedVac concentrator and redissolved in 100 l of BIAcore buffer. 30-l aliquots were analyzed using the biosensor as described under "Materials and Methods." Activity was found in the fractions eluting between 46 and 48 min.  (Fig. 4, 1-10) were concentrated using a SpeedVac concentrator and applied to a Superose 12 HR 10/30 size exclusion column. The column was eluted at a flow rate of 0.5 ml/min with BIAcore buffer, and 0.5-ml fractions were collected. The elution positions of protein calibration standards (bovine serum albumin dimer, bovine serum albumin, and trypsin inhibitor) are indicated above the chromatographic trace. An aliquot (10 l) of each fraction was taken for biosensor analysis. The also known as the 17-1A antigen or epithelial cell adhesion molecule (26). This M r 41,000 protein is expressed by many carcinomas, and the 17-1A antibody has also been used in adjuvant immunotherapy for colorectal carcinoma (27).
Rechromatography of the purified protein on size exclusion chromatography or native PAGE followed by Western blot analysis indicated, as seen previously prior to the final RP-HPLC stage (Fig. 5), an apparent M r of approximately 180,000. A single protein peak was observed with the size exclusion chromatography. The recovery was essentially quantitative (based on peak area calculations), suggesting that no additional species had dissociated during the chromatography. In fact, even after treatment of the purified protein with 8 M urea, no change in elution position on size exclusion chromatography was observed. Furthermore, the A33 antigen did not appear to be complexed with another protein, since co-immunoprecipitation from [ 35 S]methionine-radiolabeled LIM1215 cell lysates by the A33 mAb yielded a single species with the characteristic apparent M r on SDS-PAGE of 43,000 under nonreducing conditions and 50,000 following reduction (data not shown). Taken together with the initial yields obtained during N-terminal sequence analysis, these observations suggest that the 180,000 form of the antigen observed under nondenaturing conditions could be a noncovalently associated homotetramer.
To further confirm the biological specificity of the N-terminal sequence, a polyclonal antibody was raised in rabbits against a synthetic peptide corresponding to residues 2-20 of the generated sequence. Purified IgG from immunoreactive sera recognized the same M r 43,000 RP-HPLC-purified protein as the A33 mAb under nonreducing conditions (Fig. 7, lane 1). The peptide antisera also recognized a single protein spot with an apparent M r of 49,000 under reducing conditions (Fig. 7, lane 2), in agreement with the mobility observed for the immunoprecipitated material obtained from [ 35 S]methionine radiolabeled LIM1215 cell lysates under reducing conditions (see above). The antipeptide antibody reactivity appeared to be stronger against the reduced form of the protein.
The yields obtained for the individual purification steps, calculated from the biosensor responses (13), are summarized in Table I. For this purpose, a calibration curve of response (RU) against dilution was constructed retrospectively using the purified antigen. The responses observed for all of the purification steps, with the exception of the final reversed-phase chromatography (Fig. 6B), were found to fall in the linear range of the calibration curve (0 -700 RU). For quantitation of the RP-HPLC fractions, in which higher responses were obtained, values were extrapolated to the linear fit.
The Triton X-114 phase extraction was a useful initial purification and enrichment step, since almost 70% of the total extracted protein partitioned into the aqueous phase, whereas there was good recovery (83%) of the A33 ligand activity in the corresponding detergent phase. Recovery from the Green-Sepharose ligand dye column was 92%, with a 50% increase in the specific activity. The recovery from the Mono-Q column (fractions 2-10) was only 39%, possibly due to a number of factors, including the hydrophobic nature of the molecule, the detergent exchange from the Triton X-114 to CHAPS that was performed at this stage, and the use of a semipreparative Mono-Q HR 10/10 column (the use of the smaller HR 5/5 column resulted in unacceptably high back pressures). Additionally, it should be noted that, due to the large number of chromatographic fractions and the reduction in protein concentration, the fractions were monitored by UV absorption (17) instead of the BCA reaction (16), which had been used for the quantification of preceding fractions. Despite the observed losses, a significant (5-fold) increase in specific activity was obtained. Recovery from the Superose 12 size exclusion HPLC was greater than 60%, with a further 2-fold purification. The purification factor achieved with the final RP-HPLC step (62.5fold) enabled the protein to be purified to homogeneity. One microgram of A33 antigen was purified from each preparation (1.6 ϫ 10 9 LIM1215 cells), with a cumulative purification factor  a Calculated as a percentage of RU in initial extract. b For the calculation of the specific activity (SA): (i) the protein concentration was measured as described; (ii) a 10-l aliquot from the chromatographic fraction was diluted with 90 l HBS buffer, and 30 l of this sample was injected over the biosensor surface; and (iii) the RU values obtained were correlated with the amount of protein injected over the biosensor.
c Determined by BCA reaction (16). d Monitored from UV absorption at A 280 /A 260 (17). e Monitored from UV absorption at 215 nm compared with an ovalbumin standard.
of approximately 3750 and an overall recovery of 6%. By reinjection of an aliquot of the purified protein back onto the Brownlee RP 300 column and monitoring both the relative peak recovery from the chromatographic trace at 215 nm and binding response on the biosensor, it was established that the protein recovery from the RP-HPLC column was approximately 45%, whereas the recovery of biosensor activity was only 23%, suggesting that there were both on-column protein losses and partial denaturation of the ligand during this chromatographic procedure. If this loss of reactivity is taken into account, an overall purification of approximately 15,000-fold could be calculated. Additionally, it should be noted that these purification factors are based on the protein content of the initial LIM1215 detergent extract and not the total protein content of the cells, which would have been considerably higher.
In conclusion, a method for the purification and characterization of a novel tissue-specific cell surface marker for normal and transformed human colonic epithelium has been described in this article. The biosensor analysis of the interactions between the A33 mAb and its antigen, together with Western blot analysis performed under nonreducing conditions, provided a rapid, specific, and sensitive method for monitoring the purification of a novel cell protein. The availability of pure A33 antigen will now allow us to perform structural and functional studies on this molecule, as well as detailed kinetic studies on the interaction with the A33 mAb and related molecules. Kinetic characterization of intact IgG, F(abЈ) 2 and FabЈ fragments, and corresponding antibody conjugates will be studied using the biosensor in conjunction with their use in clinical imaging and therapeutic trials. An understanding of the structure and function of this interesting new tissue-specific antigen for normal and transformed colonic epithelium could lead to the development of new and improved reagents and methods for the diagnosis and treatment of colon cancer.