The current standard treatment for patients with epithelial ovarian cancer includes cytoreductive surgery followed by paclitaxel and platinum-based chemotherapy. This multimodality approach results in a complete clinical remission in nearly 80% of patients (McGuire et al.,1996) However, the majority of patients will relapse within a period of 2 years, indicating the presence of persistent small volume or microscopic disease not eradicated by primary therapy. Patients often have disease responsive to additional chemotherapy at the time of first relapse and will reenter clinical remission of short duration (Markman et al.,1997). Much of the current research in ovarian cancer is directed towards prolonging the duration of remission following primary treatment.
Among therapies that are being explored are those that use immunological approaches, including the use of monoclonal antibodies (MAbs) directed towards ovarian cancer antigens. An alternative approach to immunotherapy is the use of vaccines capable of directly stimulating an immune response directed towards antigens expressed on cancer cells (Livingston et al.,1997). Identifying carbohydrate epitopes that are overexpressed in ovarian cancer cells has opened up the possibility of using these structures as vaccines. Among possible targets are Lewisy (Ley), globo H, STn, Tn, and T-F, which are expressed on the glycolipids, glycoproteins, and mucins synthesized by ovarian cancer cells (Federici et al.,1999; Zhang et al.,1997). Some of these specificities are carried on the MUC-1 mucin, which serves as a target antigen in its own right. To some extent, the pattern of expression of these antigens varies according to the pathological subtype of the tumor (Federici et al.,1999). For instance, although Ley is highly expressed on the majority of serous and endometrioid carcinomas, it is poorly expressed on mucinous tumors. Correspondingly, STn and Tn epitopes are well expressed on mucinous tumors but poorly synthesized by serous tumors (Federici et al.,1999). Some normal tissues, e.g., epithelial cells and their secretions in the esophagus, stomach, proximal small intestine, and some acinar cells of the pancreas, also express Ley (Zhang et al.,1997; Hellström et al.,1990).
Advances in the chemical synthesis of oligosaccharides (Danishefsky and Bilodeau, 1996; Danishefsky et al.,1995) have encouraged the exploration of carbohydrate-based vaccines for the therapy of cancer. At MSKCC, trials have been initiated using GM2-KLH in melanoma patients (Livingston et al.,1994) and globo H-keyhole limpet hemocyanin (KLH) in prostate and breast cancer (Slovin et al.,1999). Miles et al. (1996) have pioneered the use of an sTn conjugate as a vaccine. On the basis of the high expression of Ley in ovarian cancer (Federici et al.,1999) and the results of a preclinical study in mice demonstrating the superior immunogenicity of a Ley oligosaccharide-KLH conjugate (Kudryashov et al.,1997), we initiated a trial of this vaccine in patients with ovarian cancer.
In this phase I study, we present the initial evaluation of a Ley vaccine based on a synthetic Ley pentasaccharide coupled to KLH carrier protein together with the QS-21 immunological adjuvant. Groups of patients with ovarian cancer who had either persistent or recurrent disease following primary therapy and were in complete clinical remission following additional chemotherapy were immunized with 4 different vaccine doses, and the safety of the vaccine and resulting antibody levels were assessed.
MATERIAL AND METHODS
Patient population and vaccination protocol
Between 10/27/97 and 7/21/98, 25 patients who had histologically documented epithelial ovarian, fallopian tube, or peritoneal cancer of any stage or grade at diagnosis were enrolled in this phase I study (Table I). Eligible patients were those who had either persistent or recurrent disease following primary therapy and were in complete clinical remission following additional chemotherapy. Complete remission was defined as a normal physical examination, normal serum CA-125 level, and normal computed tomography scan. The protocol also required normal hematologic, renal, and hepatic function. Baseline serum amylase was obtained. Stool guaiac had to be negative for occult blood. The study was carried out under an FDA-approved IND, and an Institutional Review Board approved protocol. Informed consent was obtained. Each of 4 groups (6 patients each) received s.c. administration of Ley-KLH conjugate at 3, 10, 30, or 60 μg (carbohydrate content) with QS-21 adjuvant (100 μg; Aquila, Worcester, MA) at 0, 1, 2, 6, and 18 weeks.
Table I. Demographics of Patients Entered into the Study
Total number of patients
Number of evaluable patients
Median age (range, in years)
Stage at diagnosis
Median number of prior regimens (range)
Physical examination including stool guaiac, complete blood cell counts, routine chemistries, hepatic function, and amylase were obtained every 2 weeks initially to assess for potential toxicity. Peripheral blood (20–30 cm3) was additionally drawn before vaccination, and at 2 weeks after the 3rd, 2 and 6 weeks after the 4th, and 2 and 12 weeks after the 5th vaccination. Sera prepared from the bloods was aliquoted and stored at −20°C for antibody response assays.
Patients received skin tests for delayed hypersensitivity (DTH) with Ley oligosaccharide (10 μg) i.d. at the 1st and 3rd immunizations and 2 weeks after the 5th.
Ovarian cancer cell line NIH:OVCAR-3 was obtained from the ATCC (Rockville, MD); melanoma cell line SK-MEL-28 was isolated at MSKCC. Ley-ceramide was synthesized as described previously (Kudryashov et al.,1997). Ley and Leb-expressing mucin (Tighe) was purified from an ovarian cyst fluid (Lloyd et al.,1966). Lea-polyacrylamide and Lex-polyacrylamide were purchased from Glycotech (Rockville, MD). Mouse anti-Ley MAb (3S193) has been described (Kitamura et al.,1994).
Ley pentasaccharide was synthesized as its allyl glycoside (Fig. 1) as described previously (Danishefsky et al.,1995, Danishefsky and Bilodeau, 1996). The oligosaccharide was coupled directly to KLH (Perimmune, Rockville, MD), using the reductive amination method, with a yield of 7%–10% based on sugar recovery. The sugar content of the conjugate was determined by high-performance anion exchange chromatography after acid hydrolysis as described (Lloyd and Savage, 1991), and the protein content was analyzed by the Bradford (1976) method. The conjugate contained 310 moles Ley/mole KLH. The final product was filtered through a 0.2 μm filter and vialed into aliquots containing 3, 10, 30, and 60 μg carbohydrate in 1 ml PBS containing 100 μg QS-21. Prior to its use as a vaccine, the sterility of the sample and its toxicity in mice and hamsters were tested. The immunogenicity of the construct was also tested in mice as described previously (Kudryashov et al.,1997).
Enzyme-linked immunosorbent assay (ELISA)
Reactivity of the sera with glycolipid and glycoprotein antigens were assayed by ELISA as described previously (Kudryashov et al.,1997). Two test antigens were used: 1. a synthetic Ley glycolipid (Ley pentasaccharide-ceramide); 2. a Ley-expressing mucin from an ovarian cyst fluid (Tighe; Lloyd et al.,1966); this mucin also carries Leb determinants. ELISA was carried out as follows: Antigen was coated onto the wells of 60-well microtiter plates by allowing a water solution (glycoproteins) or ethanol solution (glycolipids) to evaporate at 37°C. After blocking with 2% BSA in PBS, diluted antiserum (10 μl) was added and allowed to incubate for 1 hr at room temperature. Excess antibody was removed, and the plate washed 3 times with 0.5% BSA/PBS. Antigen-antibody complexes were then detected with one of two alkaline-phosphatase-conjugated anti-Ig reagents: 1. rabbit anti-mouse IgG (Zymed, S. San Francisco, CA.) or 2. rabbit anti-mouse IgM (Southern Biotechnology Associates, Birmingham, AL.) After the plate had been washed, p-nitrophenylphosphate was added and allowed to react for 15 and 60 min before reading in a plate reader at 405 nm.
Reactivities of the sera with tumor cells were tested by flow cytometry (FACS). Briefly, cell suspensions were prepared from Ley-positive OV-CAR-3 ovarian cancer cells and Ley-negative SK-MEL-28 melanoma cells by trypsinization. After washing in PBS, the cells were treated with the sera (1:20 dilution) or the anti-Ley mouse MAb (3S193); 5 μg/ml for 2 hr at room temperature. After washing, the cells were incubated with rabbit anti-human IgG (γ-chain specific) or anti-human IgM (μ-chain specific) conjugated with fluorescein (Southern Biotechnology Associates) for 2 hr. After washing in PBS, the cells were analyzed by fluorocytometry (EPICS-Profile II flow cytometer; Coulter, Hialeah, FL).
Complement-dependent cytotoxicity assay
Complement-dependent cytotoxicity (CDC) was assayed at a serum dilution of 1:10 with OVCAR-3 cells and human complement 1:4; (Sigma, St. Louis, MO) by chromium-release assay as previously described (Slovin et al.,1999). All assays were performed in triplicate. Controls included cells incubated only with culture medium, complement, antisera, or MAb 3S193. Spontaneous release was the chromium released by target cells incubated with complement alone. The maximum release was the amount of 51Cr released from target cells lysed with 1% Triton X-100. Percent cytolysis was calculated according to the formula: specific release (%) = (experimental release − spontaneous release)/(maximum release − spontaneous release) × 100.
Neutral glycolipids were isolated from cells and tissues as described previously (Furukawa et al.,1985). Glycolipids were separated on HPTLC plates (Merck ) in CHCl3:CH3OH:H2O solvent (60:35:8) and detected with either an orcinol-H2SO4 spray or by immunostaining with patient serum (1:20) using rabbit anti-mouse IgM-horseradish peroxidase (Sigma) as the second antibody as described in detail by Yin et al. (1996).
Twenty-five patients were enrolled, and one patient withdrew consent following one vaccination. Twenty-four patients were able to be evaluated for antibody response and toxicity assessment. The median age was 49 years (range 37–67 years) with the majority having stage III disease at diagnosis. Patients were heavily pretreated with a median number of 3 prior regimens (range 2–6). The patient demographics are summarized in Table I.
The vaccine administration was well tolerated. Dose escalation occurred according to schedule. No gastrointestinal, hematologic, renal, or hepatic toxicity related to vaccine administration occurred. Six patients had intermittent grade I leukopenia, and 2 patients had grade II anemia likely related to previous chemotherapy administration. One patient developed a stool guaiac positive for occult blood, and endoscopy showed inflamed internal hemorrhoids. One patient developed elevated liver transaminases and alkaline phosphatase; subsequent evaluation showed symptomatic gallstones. Two patients developed transient amylase elevation (1 patient grade I, 1 patient grade II), which were asymptomatic and resolved within 3 weeks. Non-hematologic toxicity was mild and is as outlined in Table II. Most patients reported self-limited malaise/myalgia for 24–48 hr after vaccination, and local skin irritation at the vaccination site resolved within 5–7 days in all patients.
Table II. Non-Hematologic Toxicity Observed in Immunized Patients
The median duration of complete clinical remission prior to study entry was 2 months (range 1–7 months). While not the endpoint of this trial, the time to biochemical or measurable disease progression was recorded. At a median of 18 months follow-up, 19 of 24 patients have either biochemical or measurable disease recurrence. The median time to progression from trial entry using biochemical or measurable criteria was 6 months (range 2–17 months). Five patients remain in complete clinical remission at 18 months of follow-up.
Serological and DTH responses
Antibody responses in the 24 immunized patients were initially evaluated by ELISA for reactivity with two natural forms of Ley antigen: Ley ceramide glycolipid and a Ley-expressing mucin. IgG and IgM responses were assayed separately by using Ig class-specific second antibodies in the assays. The results are summarized in Table III. Immunization with Ley-KLH elicited an antibody response in the majority of the patients (16/24). Most of the responses were rather low in titer, although 8 patients had titers in the 1:80–5,120 range. The majority of the responses to Ley were of the IgM class, and only 4 patients exhibited clear IgG responses. In contrast, the patients responded to KLH with strong IgG production (titers: >51,200), as well as IgM antibodies (titers: >51,200). To examine the specificity of the response, the reactivity of the sera was also tested by ELISA on a Lea/Lex-expressing mucin. Four of the sera (11,14, 17, and 20) contained IgM antibodies to this antigen. When the reactivity of these antisera was further tested on Lea-polyacrylamide and Lex-polyacrylamide, it was found that only one serum (20) was reactive and that it reacted with both the Lea and Lex test compounds (data not shown). It was concluded that serum No. 20 contained broadly reactive antibodies recognizing Lea and Lex as well as Ley and that the other sera reacted with an unknown epitope(s) present on the Lea/Lex mucin sample. A peculiarity of the antibody responses in almost all the responding patients was the transient nature of the antibody response. In general, the antibody response peaked after the third or fourth immunization and subsequently fell even in the face of repeated immunizations; two examples of this effect are shown in Figure 2. In terms of an optimal immunizing dose of antigen, immunization with either 10 or 30 μg of carbohydrate gave the most consistent antibody responses (Table III).
Table III. Reactivities of Patient's Sera with LeY-Ceramide and LeY-Expressing Mucin Assayed by ELISA
The antibody responses were also analyzed by FACS for cell surface reactivity with the Ley-positive cell line OVCAR-3 and for their ability to lyse the same cell line in the presence of human complement (Table IV). Eight of the sera showed clear reactivity with the cell lines in FACS analysis and 12 showed strong CDC. Two examples of the FACS analysis are shown in Figure 3. The degree of positivity by FACS analysis did not completely correlate with the reactivity of the sera with Ley-Cer and Ley-mucin, although the serum showing the highest reactivity by ELISA (No. 12) was also the serum having the greatest number of cells positive by FACS. A number of the sera (12/24) also exhibited the ability to lyse OVCAR-3 in the presence of human complement (CDC). The correlation between cell surface reactivity as determined by FACS and the ability to lyse cells was quite high, although some sera that showed CDC activity did not show reactivity by FACS (Table IV).
Table IV. Reactivity of Patient Sera with OVCAR-3 Ovarian Cancer Cell Line Evaluated by Flow Cytometry (FACS) and Complement-Dependent Cytotoxicity (CDC)1
Amount of carbohydrate in vaccine per immunization.
Positive cells were detected with a fluorescein-conjugated anti-human IgM-specific reagent. Fluorescence given by preimmune sera was gated to 7.0–8.0% positive cells for each specimen and reactivity of postimmunization sera (after 3 or 4 immunizations) were compared with this value. None of the sera showed positive fluorescence with SK-MEL-28 cells.
Complement-mediated cytotoxicity was determined by a 3-hr 51Cr-release assay in the presence of human complement.
Bold figures indicate sera showing substantial difference between pre- and postimmunization values.
With regard to the chemical nature of the antigen recognized by the antibodies in the patient sera, more sera were reactive with Ley-ceramide (16/24) than with Ley-expressing mucin (8/24) as detected by ELISA (Table III). Testing the ability of serum 12 by TLC-immunostaining to react with the neutral glycolipids from a number of cells lines showed that this serum reacted with synthetic Ley pentasaccharide-ceramide and with Ley pentasaccharide-ceramide in ovarian cancer cell lines OVCAR-3 and SK-OV-3 (Fig. 4a, lanes 2 and 3) and in 3 ovarian tumor specimens (Fig. 4a, lanes 4–6) but not in Ley-negative melanoma cell line SK-MEL-28 (Fig. 4a, lane 7). Slower-moving positive bands were also observed in the ovarian cancer cell lines and tumor specimens; these species are probably extended forms of Ley glycolipids, e.g., trifucosyl Ley. Comparison of the immunostained plate with the total glycolipid species present in the samples as detected with an orcinol-H2SO4 spray (Fig. 4b) shows the specificity of the antibody response for Ley as the serum did not react with the faster-moving, simpler glycolipids present in these samples. In contrast with this result with glycolipid samples, attempts to detect glycoproteins and mucins from OVCAR-3 cells by immunoprecipitation of [3H] glucosamine-labeled cell lysates with serum 12 were unsuccessful (data not shown).
No delayed type hypersensitivity responses to Ley i.d. skin tests were observed.
Our results show that it is possible to induce an antibody response in humans by immunization with a synthetic Ley glycoconjugate vaccine and that the vaccine was well tolerated with no adverse effects related to autoimmunity. Although many of the antibody responses were rather modest, significant antibody titers were attained in a proportion of the patients. One patient (12) showed a very high response, and it is interesting that this individual had anti-Ley antibody levels as measured by ELISA (Table III), FACS (Fig. 3), and CDC (Table IV) prior to immunization. In this patient, the anti-Ley response may have been an anamnestic response. Even though a synthetic construct was used as the immunizing antigen, the resulting antibody was capable of reacting with natural form of the Ley (mucin and glycolipid) and with Ley-expressing tumor cells. The antibodies produced were mainly of the IgM class with only 3 patients exhibiting detectable levels of IgG antibodies. This is in contrast with our results from immunizing mice with the same conjugate and the same adjuvant where both IgM and IgG antibodies were observed (Kudryashov et al.,1997). It appears that the KLH carrier was not able to induce a sufficient helper T-cell response to aid in a class switch to IgG antibodies in the majority of patients. It is also possible that these heavily pretreated cancer patients may be relatively anergic, although their normal antibody responses to KLH does not support this interpretation. Another possibility is that the highly substituted KLH cannot efficiently serve as an inducer of T-cell help, although this was not our experience with other KLH conjugates (Livingston et al.,1994). An interesting characteristic of the antibody responses were their transient nature. Thus, the maximum response was observed after 3 or 4 immunizations in most responding patients. Subsequently, antibody levels fell even in the face of repeated immunizations. These findings remain unexplained.
As to the molecular form of Ley recognized by the patient antibodies, it appears that the antibodies reacted better with Ley glycolipids rather than with Ley-expressing mucins (Table III). Also, serum from patient 12 recognized Ley glycolipids in a number of tumor cells but it did not react with mucins or glycoproteins in the same cells even though the cell line is known to carry Ley determinants on both glycolipids and mucins (Yin et al.,1996). This result is consistent with the ability of the patient sera to lyse Ley-expressing tumor cells in the presence of complement, as we have found that glycolipid antigens are better cellular targets for complement-dependent cytotoxicity than are glycoproteins and mucins (data not shown).
Considering that Ley is expressed on a number of normal tissues in humans, including breast, intestine, and pancreas, it is significant that it is possible to break tolerance to this antigen. Similar results have also been obtained with other carbohydrate-conjugate vaccines, e.g., GM2-KLH (Livingston et al.,1994), STn-KLH (Miles et al.,1996), and globo H-KLH (Slovin et al.,1999). The mechanism for the breaking of B-cell tolerance to these epitopes may be related to the recognition by B cells of a self-determinant if the determinant is on a protein possessing foreign T-cell epitopes, in our case KLH. Despite the normal distribution of Lewisy antigen in multiple sites, no symptomatic autoimmune toxicity was seen in this trial. Two patients had a mild, clinically insignificant elevation in serum amylase that resolved. Previous trials of anti-Lewis-y antibody (BR96) coupled to doxorubicin resulted in gastrointestinal toxicity including nausea, vomiting, and gastritis (Tolcher et al.,1999). No gastrointestinal toxicity resulted from immunization with the Lewisy antigen. In a number of mouse therapy models to other antigens also it has been noted that either autoimmune disorders do not develop or are limited in scope (Speiser et al.,1997; Tempero et al.,1999; Weber et al.,1998).
Clinical outcome was not the endpoint of this phase 1 trial. Some patients have recurred, as would be expected in this population. Whether a single component vaccine, capable of inducing only an antibody response, would be expected to be effective is unknown. The greatest impediment to a successful outcome is probably the heterogeneity of antigen expression within tumor lesions, although the expression of Ley in ovarian tumors, as judged by immunohistology, is relatively uniform (Federici et al.,1999), particularly in comparison with other carbohydrate specificities.
We thank Ms. C. Bryant for expert secretarial assistance and Mr. C. Kandell for Figure 1.