The Development of Biologic End Points in Patients Treated with Differentiation Agents: An Experience of Retinoids in Prostate Cancer¹

Interest in the retinoids as therapeutic agents for prostate cancer stems from epidemiological studies suggesting that the consumption of vegetable-derived vitamin A protected against the disease (1, 2), in vitro studies showing that ATRA³ and cRA could inhibit cell proliferation and promote cellular differentiation and in vivo results with fenretinide, suggesting a role in both prevention and progression in the Dunning rat prostate cancer model (3). Clinical trials have shown mixed results (4, 5), but information is limited on the in vivo effects of these compounds on the human disease in relation to the clinical outcomes.

Previously, we proposed a methodology for evaluation of new agents in prostate cancer using posttherapy changes in PSA (6). This was required because the patterns of spread of the disease preclude the use of more traditional end points of efficacy. This method restricted entry to patients that had progressive disease, despite castrate levels of testosterone and tumor progression after the withdrawal of the antiandrogens and steroid hormones (6, 7). Outcomes are reported on the basis of changes in PSA, changes in bone, and changes in soft tissue independently (6). For differentiating agents in particular, it was recognized that a unique end point would be required because these compounds may initially increase PSA levels,which might be misinterpreted as “treatment failure.” As such, a pathological end point using accessible tumor tissue was used in these studies. Posttherapy PSA changes, increases or decreases, were not used to determine response or to continue treatment in the absence of other parameters.

ATRA was the first agent studied. The results showed that the compound induced its own degradation within 48–72 h, which may have contributed to the lack of clinical activity seen (8, 9). In the second trial, cRA was combined with IFN, based on an improved pharmacokinetic profile (10), the antiproliferative effects in vitro, and the additive effects of IFN. (11) Patients with both AI and AD tumors were studied, and the disease was evaluated by pre- and posttherapy biopsy for histology,proliferative index, and apoptotic changes. To assess for differentiation effect, PSMA expression was also assessed.

Entry required histologically confirmed prostate cancer that was progressing, despite castrate levels of testosterone (<30 ng/ml), and a life expectancy of >3 months. Disease progression was defined as:(a) a rising PSA of >50% from baseline on three successive occasions; (b) new metastatic lesions on bone scan;(c) a >25% increase in a bidimensionally measurable tumor mass; or (d) new symptoms related to tumor growth. Patients could have multiple previous hormonal manipulations and one previous chemotherapy. All patients had a KPS >60%, WBC >4000 cells/mm³, granulocytes >1500 cells/mm³, platelet count >150,000 cells/mm³, and normal coagulation parameters(prothrombin time, partial prothromboplastin time, and thrombin time). A serum creatinine <2.0 mg/dl and a baseline serum glutamic-oxaloacetic transaminase <2 times the institutional normal were also required.

Patients must have been at least 3 weeks from major surgery and at least 4 weeks from any prior radiation. Individuals were excluded if they had New York Heart Association class III or IV heart disease,history of bleeding disorders, peripheral neuropathy, central nervous system disease, active infection, or the presence of brain metastases. Patients who had not undergone a surgical orchiectomy were continued on medical therapies. These protocols were approved by the Institutional Review Board and written informed consent was required in all patients.

The pretreatment evaluation included a complete history and physical examination with a baseline KPS. Laboratory studies included an automated blood and platelet count, serum electrolytes, 12-channel screening profile (alkaline phosphatase, lactate dehydrogenase,aspartate transglutaminase, blood urea nitrogen, creatinine, calcium,phosphorus, uric acid, total protein, albumin, and total bilirubin),cholesterol, triglycerides, serum acid phosphatase (ACP), and PSA. Imaging studies included an abdominal and pelvic CT scan, bone scan,and chest radiograph or magnetic resonance imaging when clinically indicated.

In posttreatment, the automated blood and platelet count was performed weekly for the first 4 weeks and then every 2 weeks. A screening profile, electrolytes, creatinine, cholesterol, triglycerides, ACP, and PSA were repeated biweekly for the first 8 weeks and at a minimum of monthly thereafter. Measurable disease was evaluated at 8-week intervals, whereas radionuclide bone scans were generally repeated at 12-week intervals in the absence of other criteria of disease progression. All drug-related toxicities were recorded and graded according to the NCI Common Toxicity Criteria (version 1).

All patients on the ATRA study and patients with assessable tumors in the cRA\/IFN study had pre- and posttherapy tumor biopsies obtained. Biopsies were obtained within 4 weeks of starting the therapy and were repeated after 4 weeks of therapy. The same lesions were biopsied when possible. Pre- and posttherapy tumor biopsies were evaluated for histological changes on H&E stains, for apoptosis using a modified terminal deoxynucleotidyl transferase-mediated nick end labeling method as described by Gavrieli et al. (12), and for proliferation index using anti-Ki67 monoclonal antibody MIBI (Immunototech; 1:50 dilution). Ki67 proliferative index was considered high when ≥20% of tumor cells displayed a positive MIBI nuclear immunoreactivity pattern.

Two different antibodies against the intracellular domain of PSMA, PM2(Hybritech), and 7E11(Cytogen Corp., Princeton, NJ) were used to detect expression of this antigen. MIgS-KpI, a mouse monoclonal antibody of the same subclass as the primary antibodies listed above, was used as negative control at similar working dilutions. Sections were subsequently immersed in boiling 0.01% citric acid (pH 6.0) for 15 min to enhance antigen retrieval, allowed to cool, and incubated with primary antibodies overnight at 4°C. Biotinylated horse antimouse IgG antibodies were applied for 1 h (Vector Laboratories, Burlingame,CA; 1:500 dilution), followed by streptavidin peroxidase complexes for 1 h (Vector Laboratories; 1:500 dilution). Diaminobenzidine was used as the nuclear counterstain. For each case, the percentage of invasive tumor cells expressing cytoplasmic immunoreactivity was estimated in a continuum from 1 to 100%. Intensity was recorded using a scale of 0–3 (0, undetectable staining; 3, highest staining).

All pathological specimens were reviewed by one pathologist (V. R.),who was blinded as to the pre- and posttherapy status of the specimen and clinical outcomes. All pathological slides were prepared at the same time to avoid intraexperimental variation of the results and so that both positive and negative controls were examined simultaneously.

For the administration of ATRA, the drug was supplied by the NCI and consisted of soft gelatin-filled capsules, each containing 10 mg of ATRA. Other ingredients included butylated hydroxyanisole, disodium edate, refined soybean oil, and a wax mixture consisting of purified beeswax, hydrogenated soybean oil flakes, and hydrogenated vegetable oil. Patients were started on 50 mg/m² ATRA p.o. every 8 h and instructed to take it with food. The dose was rounded to the nearest 10 mg.

The dose of ATRA was escalated by 25% if the patient did not experience a greater than grade 2 toxicity after 4 weeks of therapy. If a patient developed a grade 3 or 4 toxicity, the therapy was held until the adverse event resolved. Therapy was continued until progression of disease or dose-limiting toxicity.

On day 1 of therapy, 10 ml of heparinized blood were drawn at 0.5, 1,2, 3, 4, 6, 8, 12, and 24 h. Urine was also collected in three aliquots from 0 to 6, 6 to 12, and 12 to 24 h. These pharmacokinetic tests were repeated after 4 weeks of treatment to assess changes over time. After the initial pharmacokinetic parameters were evaluated for the first five patients, there was a marked drop in ATRA in the serum at day 28, and subsequently patients had pharmacology studies performed on day 3 (heparinized blood was drawn at 0.5, 1, 2,and 3 h). We have reported previously the pharmacokinetics of ATRA, and these results will not be discussed further (8).

Recombinant human IFN (Roferon-A) and cRA (Accutane) were obtained from Hoffmann-LaRoche, Inc. (Nutley, NJ) through the Division of Cancer Treatment, NCI. IFN was provided in 3-million unit and 18-million unit vials, which were mixed with sterile water. cRA was available in 10-and 40-mg tablets. Patients were administered IFN starting at 3 million units s.c. daily, and the dose was escalated to 6 million units the second week and 9 million units the third week if patients did not experience a grade 2 or greater drug toxicity. Patients were premedicated with 650 mg of acetaminophen p.o. prior to the administration of IFN. The dose was held for grade 3 or 4 NCI toxicity,except for a hematological toxicity manifested by low hemoglobin. In case of a severe adverse reaction, the IFN was held until the toxicity returned to grade 0 or 1. Patients with a persistent grade 3 or 4 toxicity for greater than 4 weeks were removed from the study. Patients received 1 mg/kg of cRA in two divided doses daily. This was taken with food for maximal absorption. The treatment of cRA was interrupted in patients who experienced any grade 3 or 4 toxicity and reinitiated at a dose of 0.5 mg/kg p.o. in two daily doses once patients’ toxicity resolved (grade 0 or 1). If grade 3 or 4 toxicity persisted on the attenuated dose of cRA, the treatment was discontinued.

Once the safety of this regimen was established, the protocol was amended to include patients with a rising PSA after surgery or radiation therapy that had not undergone castration. The same dose and schedule was used. After 16 weeks of therapy, if the patient did not have a partial response (>80% posttherapy decline in PSA), they were withdrawn from the study and offered androgen ablation or observation.

Patients were followed in the outpatient department at weekly intervals for the first 4 weeks and then biweekly afterward. At each visit, the history and physical were repeated, and adverse events were recorded using the NCI Common Toxicity Scale (version 1.0). Patients were considered evaluable for response if they had completed a minimum of 4 weeks of treatment. Outcomes were assessed independently using radionuclide bone scan for osseous lesions, CT scan or magnetic resonance imaging for measurable lesions, and posttherapy changes in PSA (Tandem-E; Hybritech; upper limit of normal range, 4.0 ng/ml). Posttherapy PSA changes were assessed as described previously (6, 13).

Thirty patients with AI prostate cancer were treated with ATRA (14 cases) or the combination of cRA\/IFN (16 cases), of whom 14 (100%) and 13 (83%) were evaluable for response. Three patients treated with cRA\/IFN were in the study for less than 4 weeks and not evaluable for response. Two patients had rapid progression of disease requiring alternative therapies, and one patient requested to be removed from the study because of severe fatigue with anorexia. All 30 patients were evaluable for toxicity.

Patient characteristics and baseline biochemical parameters are detailed in Tables 1 and 2. The median performance status was 80%; 12 of 30 (40%) had more than one prior hormonal manipulation,and the majority had measurable disease. Other baseline parameters are consistent with patients with advanced prostate cancer.

All patients tolerated the full dose of ATRA, of whom one (7%)patient required a dose reduction for fatigue. In the cRA\/IFN trial, the maximum dose of IFN was escalated to 9 million units in nine patients and 6 million units in two patients, whereas two patients were maintained at 3 million units. Four of 13 (30%) patients required a dose reduction of the IFN, and one patient required a dose reduction of the cRA because of toxicity. The median duration of treatment was 71 days, with a range of 30–351 days.

A dry scaling skin rash and cheilitis were the most common adverse events (Table 3). Both were treated with skin emoluments and did not require a change in systemic therapy. Significant weight loss and anorexia was seen in <10% of the patients; however, one patient discontinued treatment because of persistent fatigue and anorexia. Hematological toxicities were most common in patients treated with cRA\/IFN with mild leukopenia; anemia and thrombocytopenia were the most common abnormalities.

Two patients with mild urinary hesitancy prior to treatment had acute urinary outlet obstruction within 1 week of initiating cRA\/IFN. One patient had a transurethral resection of the prostate and then continued on therapy with cRA\/IFN. This patient after 8 weeks of continuous treatment showed regression of a metastatic lymph node.

Mild elevation of hepatic transaminases was common, along with elevation of serum triglycerides. Most triglycerides were <200 mg/dl(normal range, 50–160 mg/dl); however, one patient had an extreme elevation of the triglycerides to 3280 mg/dl. Minor sensory, motor,visual, and mood changes occurred during treatment with cRA\/IFN, but headaches were the most common neurological abnormality reported with ATRA. Only one patient treated with cRA\/IFN reported a similar headache. One patient treated with ATRA had a nonfatal pulmonary embolism, which was treated with anticoagulants without further complications.

Twenty-five patients had metastatic disease to the bone prior to therapy. No improvement or progression in the bone scan was documented in all cases after treatment. One patient treated with cRA\/IFN with measurable disease and no osseous disease at the beginning of treatment developed bone metastasis after 9 months of therapy.

One of 24 patients (4%; 95% confidence interval, 3.8–11.8%) with measurable disease treated with cRA/IFN showed a <50% decrease in a pelvic nodal mass, which was confirmed as necrotic on posttherapy tumor biopsy.

All 30 had a rising PSA at the initiation of the study. No patients normalized their PSA or had an 80% reduction (partial response) in the PSA from the baseline. No patients with ATRA had a significant PSA decline. One of 13 (8%; 95% confidence interval, 0.5–15.5%)patients treated with cRA\/IFN had a >50% reduction in the PSA from baseline for a 4-month duration. Fig. 1 shows one patient treated with cRA\/IFN that had a steady rise in the PSA with no evidence of progression in his retroperitoneal and osseous disease. After 4 months of continuous therapy, the PSA declined below his pretherapy baseline value. The patient continued cRA\/IFN for 9 months and then had evidence of measurable disease progression associated with a 50% rise in the PSA from his nadir value. After the discontinuation of cRA\/IFN, a >50% decline in PSA was documented, which was maintained for 4 months. There were no other concurrent interventions.

There were 21 (14 ATRA and 7 cRA/IFN) paired pre- and posttherapy tumor samples. Nine patients treated with cRA\/IFN did not have tumor assessable for biopsy, refused additional biopsy, or no tumor was found in the biopsy sample. Twenty pre- and posttreatment biopsies were from the same site. Only one patient had his pretreatment biopsy taken from metastatic paraortic lymph nodes and his posttreatment biopsy from a more accessible metastatic iliac bone.

No difference in histology, Ki67 index, or apoptosis were observed after ATRA. Of the seven biopsy sites evaluated after cRA\/IFN, one patient had pelvic lymph node biopsy and transurethral resection of the prostate pre- and posttherapy, four had prostate biopsy, one had retroperitoneal lymph nodes, and one had bone marrow biopsy. The two patients with biopsies of the lymph nodes showed tumor necrosis on the posttherapy specimen,with only one of two of the patients showing a parallel regression on CT scan. No other specimens demonstrated other gross histopathological changes. Only one case showed a decrease in the proliferation index in a posttherapy biopsy of the prostate when compared with the pretherapy sample. No evidence of induction of apoptosis was seen in the posttherapy specimens.

Nine of 21 of the paired samples (in 12 cases, the tissue blocks were exhausted) had additional pathological material available to assess PSMA expression based on the percentage of cells stained and intensity(Table 4). In posttherapy, there was an increased expression of PSMA,defined as both an increase in percentage of positive tumor cells as well as an increased intensity in immunoreactivity. This was seen in seven of nine cases (77.7%) using the PM2 antibody and in two of nine(22.2%) using the 7E11 antibody. Fig. 2demonstrates the changes in PSMA expression using the PM2 antibody in cases 7 and 9. We observed increased intensity, and the percentage of cells that showed expression for PSMA was greater with the PM2 antibody than the 7E11 antibody in all cases. There was a concordance of increased expression using the two antibodies in the two cases (nos. 2 and 9).

Four patients with rising PSA after radical prostatectomy or radiotherapy were treated with cRA\/IFN. Patient characteristics and baseline biochemical parameters are detailed in Tables 1 and 2. Adverse events were similar to the AI patients(Table 3).

After 16 weeks of therapy, none of the patients had progression in the bone, but one of the four patients had interval development of pelvic lymphadenopathy, suggesting metastatic disease. All of the patients had a rising PSA prior to therapy, and the PSA stabilized in all cases(Fig. 3). A 50% decline in posttherapy PSA for 2 months was documented in one case. Pre- and posttherapy levels of testosterone were evaluated, which demonstrated normal physiological levels of serum testosterone in all patients. Because none of the patients achieved >80% decline in PSA or normalized the PSA (partial response or complete response), they were withdrawn from the study as per protocol at the end of 4 months. Three patients elected to be treated with androgen ablation. One patient was observed without treatment and has maintained a stable PSA for >1 year (Fig. 3 D) without other evidence of disease progression.

The present study showed limited antitumor effects with ATRA and cRA\/IFN, whether assessed by changes in measurable disease or changes on bone scan. Posttherapy changes in PSA were not used as the primary end point, but 1 of 30 patients did show a >50%posttherapy decline. Posttherapy tumor biopsies documented tumor necrosis in 2 of 21 cases, and only one patient had a decrease in the proliferation index posttherapy. Similar data from Trump (4) reported no major responses in 17 patients with advanced prostate cancer treated with ATRA. This data would suggest that retinoids did not have clinical activity in this disease.

Because retinoic acid may confound the interpretation of posttherapy changes in PSA and early biological effects from the retinoic acid therapy may not be detected using standard radiographic and pathological techniques, we also assessed changes in tumor cell antigen expression during therapy. Posttherapy tumor specimens did show an increased expression of PSMA after therapy in seven of nine cases using the PSM2 antibody and a parallel observation in two of nine cases using the 7E11 antibody. This increased expression of PSMA was seen with ATRA and cRA\/IFN. Two different antibodies were used because PM2 and 7E11 bind to different intracellular PSMA domains with different affinities (14). As seen in other studies, the PM2 antibody consistently had a greater expression in intensity and percentage of cells stained than the 7E11 antibody (14). PSMA is a type II membrane glycoprotein of M_r ∼100,000 and is expressed in benign prostate secretory-acinar epithelium, prostatic intraepithelial neoplasia, and prostatic adenocarcinoma (15). The PSMA functions as a neurocarboxypeptidase and folate hydrolase and may be involved in neuroendocrine regulation of prostate growth and differentiation (15, 16). Increased PSMA expression is reported in high-grade localized tumors (16, 17); however, PSMA expression is decreased in metastatic disease (18). In this study, an increased expression of PSMA posttreatment suggests a biological effect from the therapy and may indicate a favorable differentiation effect on the tumor. This finding needs to be further confirmed in larger prospective trials.

Prior to the studies, it was a concern that retinoids might increase PSA secretion independent of cell growth. This was not generally observed; however, in one case, a 50% posttherapy decline in PSA was documented (Fig. 1) but only after 4 months of treatment and a transient rise in the PSA. The 4-week biopsy in this patient also showed no tumor necrosis but an increased expression of PSMA from the baseline (patient 8 in Table 4). After 9 months of treatment, this patient discontinued cRA\/IFN secondary to objective disease progression. Without further therapeutic intervention, the patient’s objective disease stabilized, and a decline in PSA >50% was observed. This is similar to what has been described in the “steroid withdraw syndrome” (7, 19).

Preclinical studies and our initial results suggested that patients might require prolonged exposure to the retinoids to see a clinical benefit. Often in patients with advanced PC, symptomatic progression occurs rapidly, not allowing adequate drug exposure. This was the case in our study with the majority of patients with advanced AI prostate cancer progressing in 2–3 months. This would indicate that patients with minimal tumor volume might be better candidates for this treatment approach. To test this hypothesis, hormone-naive patients that had a rising PSA after radical prostatectomy or radiation therapy were treated with cRA\/IFN. The PSA declined or stabilized in all cases (Fig. 3), but the treatment was associated with cumulative fatigue. Serum testosterone levels were not altered by cRA\/IFN treatment in our patients, but retinoic acid may inhibit 5α-reductase, decreasing the conversion of testosterone to dihydrotestosterone (20). This could effect PSA secretion, although the inhibition of PSA should reverse once the drugs are discontinued. One patient was observed without androgen ablation after cRA\/IFN therapy and has remained clinically stable without biochemical progression for >1 year (Fig. 3 D). Although the natural history of his cancer is unknown,this patient appears to have had a clinical benefit from the combined therapy, which cannot be contributed to the suppression of DHT. Similar studies in AD patients by Dipaola et al.(5) showed that 5 of 21 patients treated with cRA and lower doses of IFN had a decline in posttherapy PSA. Plasma TGF-βlevels increased with the cRA\/IFN and correlated with a decrease in the PSA (5). This also suggested an indirect biological effect from the therapy. The results of cRA\/IFN in patients with minimal disease are encouraging, but the cumulative fatigue from the therapy is a concern in this asymptomatic population and often deters patients from continuing therapy.

The clinical observation made from this series of investigations with retinoic acid has provided strategies for developing other differentiation therapies. The increased expression of PSMA on posttherapy biopsies suggested that there was a biological impact from the therapy that was not apparent while using conventional measures of outcome. In the future, the modulating effect of retinoic acid may be exploited to enhance PSMA expression on prostate cancer cells to increase the targeting of specific antibodies to the PSMA.

We thank Hybritech, Inc. for supplying the PM2 antibody for the study.