Long-Term Outcome and MGMT as a Predictive Marker in 24 Patients With Atypical Pituitary Adenomas and Pituitary Carcinomas Given Treatment With Temozolomide

Context/Objective:

Locally aggressive pituitary tumors (LAPT) and pituitary carcinomas respond poorly to conventional therapy and cytotoxic drugs. Temozolomide (TMZ) is an oral alkylating drug with good tolerability, approved for treatment of malignant gliomas. The experience of its use in pituitary tumors is limited.

Design and Setting:

We report on 24 patients with aggressive pituitary tumors (16 LAPTs, 8 carcinomas) treated with TMZ for a median of 6 months (range 1–23). Follow-up ranged from 4 to 91 months with a median of 32.5 months. 19/24 tumors were hormone secreting (PRL 9, ACTH 4, GH 4, GH/PRL 2). Ki-67 was 2–50% in LAPTs, and 5–80% in carcinomas.

Main Outcome:

Response to TMZ and the association with tumor expression of O6-methylguanine DNA methyltransferase (MGMT), MLH1, MSH2, and MSH6, examined by immunohistochemistry.

Results:

Complete tumor regression occurred in two carcinomas and persisted at follow-up after 48 and 91 months, respectively. Partial regress of tumor mass ranging from 35% to 80% occurred in 5 LAPTs and 2 carcinomas. Another patient with LAPT had a 71% decrease in prolactin levels without change in tumor volume. Three LAPTs could not be evaluated. Median MGMT staining was 9% (5–20%) in responders vs 93% (50–100%) in nonresponders. Loss of MSH2 and MSH 6 was observed in a single patient who had a rapid development of resistance to TMZ.

Conclusions:

This study shows that TMZ is a valuable treatment option for patients with uncontrolled pituitary tumors. The data suggest that tumoral MGMT staining below 50% is associated with a high likelihood of treatment response.

About 15% of pituitary adenomas are classified as “atypical” based on certain histopathological features: a Ki-67 labeling index above 3%, increased number of mitoses, and extensive p53 nuclear staining (1). Compared to ordinary benign adenomas, the atypical tumors are, in general, locally aggressive and frequently not controlled by multiple surgery, radiotherapy, and/or medical treatment. Aggressively growing pituitary adenomas do not always exhibit high proliferation indices, ie, are not classified as atypical on histopathological grounds.

Carcinomas represent about 0.2% of pituitary tumors (1). Their diagnosis is based on the identification of metastases (2) which can be systemic and/or cerebrospinal. Standard chemotherapy regimens typically have limited efficacy and the prognosis is poor. Two out of 3 patients die within 12 months after diagnosis (3, 4).

The first case reports with successful treatment of aggressive pituitary tumors with temozolomide (TMZ) were published in 2006 (5–7). TMZ is a lipophilic alkylating agent approved for treatment of glioblastoma multiforme. It has also been used for treatment of advanced neuroendocrine tumors (8) and malignant melanomas (9). The drug has a near 100% bioavailability through oral administration (10, 11), and crosses the blood-brain barrier easily. TMZ alkylates guanine at the O6 position, the O6-methylguanine base is recognized by the DNA mismatch repair system (MMR), which results in futile cycles, DNA strand breaks, and eventually cell death (11, 12). This effect is counteracted by removal of the methyl group by the DNA repair protein O6-methylguanine DNA methyltransferase (MGMT), thereby restoring the nucleotide to its normal form. In malignant gliomas, a low tumor MGMT expression is correlated with a better treatment response (13, 14). Independently of the MGMT status, inactivating mutations of the DNA mismatch repair protein MSH6 contribute to resistance to TMZ (15, 16).

In pituitary tumors, the role of MGMT as a predictive marker for response to TMZ has been debated. Some studies have shown no correlation (17–19), while others suggest MGMT to be valuable (20–22). Lack of distinct criteria regarding MGMT staining hampers the comparison between these studies. Semiquantitative measures for assessment have mostly been used. Results have been expressed as low, intermediate, and high staining, with no consensus on the cut-off values for the categories. Heterogeneity of MGMT distribution within the tumors may also occur (17). In one recent study, the immunostaining pattern of MSH 6 but not MGMT was found to correlate with clinical outcome (19).

The overall experience of TMZ in locally aggressive pituitary tumors and pituitary carcinomas is limited and mainly based on single case reports or small series of 2–13 patients, with no standardized response criteria (20, 23, 24). In the present analysis the aim was to study long-term outcome and predictive tumor factors in 24 patients with aggressive tumors treated at seven pituitary centers in Sweden, Denmark, Belgium, and The Netherlands. Eight of the patients had pituitary carcinomas with systemic and/or cerebrospinal metastases. Tumor expression of Ki-67 labeling index, p53, MGMT, and the three MMR proteins MLH1, MSH 2, MSH6 were examined by immunohistochemistry at a single laboratory, and their value as biomarkers of clinical response evaluated.

Patients and Methods

Patients

Twenty four patients, 16 with locally aggressive pituitary tumors (LAPTs), defined as macroadenomas, in which tumor growth was not controlled by surgery, radiotherapy, and medical treatments, and 8 with pituitary carcinomas, were treated with TMZ. Median patient age at diagnosis was 48 (range, 13–71) years. Nineteen tumors were hormone secreting (PRL 9, ACTH 4, GH 4, PRL/GH 2), see Tables 1 and 2. Sixteen of the 24 patients had visual field defects /visual impairment, and/or hydrocephalus due to extensive suprasellar tumor extensions.

At diagnosis 8 of the 16 LAPTs presented with an aggressive growth behavior. The other 8 tumors acquired an aggressive phenotype with loss of response to regular therapy after a latency period of 5–300 months (median 110), Table 1. The 8 patients with pituitary carcinomas had systemic and/or cerebrospinal metastases. The median time from diagnosis to metastases was 85 months (range, 14–266), with the shortest time for ACTH producing carcinomas, (range 14–61 months, n = 3, Table 2).

At the start of TMZ, 12 of the 16 patients with LAPTs had undergone several surgical procedures. Radiotherapy in 13 patients, and cytotoxic drugs in 3, had not achieved growth arrest. Tumors had been resistant, or only partially responsive to dopamine agonists and/or somatostatin analogs (Table 1).

On histopathological examination 12 of 16 LAPTs displayed atypical features, ie, a Ki-67 labeling index above 3%, and/or p53 staining, and/or abundant mitoses/nuclear pleomorphism (Table 1). One young man with a giant prolactinoma (pat no 5) had MEN-1.

Temozolomide treatment

TMZ was administered in a dose of 150–200 mg/m² per day for 5 days every 28 days. Median time on TMZ treatment was 6 months (range, 1–23). In 2 nonresponders (No. 17 and 18), capecitabine was added during the last 2 and 4 cycles, respectively (see Table 4 below). Radiotherapy was performed 2 and 6 months before the start of TMZ in patient 2 and 13, and at TMZ cycle 2 and 3 in patient 6 and 19. Three of the 24 patients (No. 9, 12, and 22) have previously been reported with a shorter duration of follow-up (25).

Response criteria

The treatment response data are based on 21 out of 24 patients; three were excluded from the outcome analysis. Two of these patients (No. 1 and 10) stopped treatment after 1 cycle due to side effects, and in one patient (No. 3) with gigantism and residual tumor mass postsurgery, GH/IGF-I concentrations at the start of TMZ + radiotherapy were not available. In all tumors a complete response was defined as a complete regression of total tumor mass, and normalization of hormone concentrations. A partial response was defined as decrease in tumor volume ≥ 30% at MRI, and /or a hormonal decrease ≥50% in LAPTs, and a decrease in tumor burden ≥30% in carcinomas.

Tumor markers/histopathology

Three μm sections were cut from paraffin embedded tissue and stained by immunohistochemistry on a Ventana BenchMark Ultra instrument using Optiview, Ventana as a detection system. As primary antibodies, MGMT (Invitrogen clone MT23.2, 1:100), Ki67 (Ventana clone DE-r-11 RTU), p53 (Ventana cloneDO7, RTU), MSH2 (Ventana clone G219-1129, RTU), MSH6 (Epitomics1:50), and MLH1 (Novocastra clone ES05, 1:50) were used. Staining for MGMT, MLH 1, MSH2, and MSH6 were performed with tyramide signal amplification (TSA). CC1 (Ventana) 100°C 32–48 minutes was used for epitope retrieval.

Estimation of the MGMT staining % was performed by counting stained and total number of tumor nuclei in well preserved fields. The aim was for a minimum of 200 events. Positive staining of endothelial cells was used as internal control of tissue quality, and only fields with positive endothelium were assessed. Ki67 was estimated in hot spots. p53 was considered positive if more than 10% of the nuclei stained densely. The mismatch (MMR) repair proteins were considered positive if they stained positive in a well-conserved area. Negative staining was accepted when tumor cells were negative in areas with positive staining of endothelial and mesenchymal cells.

Statistical analyses

Data are presented as median and range from minimum to maximum. Continuous variables were compared between groups using Mann-Whitney U-test or between matched pairs using Wilcoxon matched pair test. Categorical variables were compared using double-sided Fischer exact test. P values less than 0.05 were considered significant. P levels of 0.001 are the lowest presented

Ethical approval

The study was approved by the Ethics Committee of Lund University, Sweden, Dnr 2014/347, and the Ethics Committee of Region Hovedstaden, Denmark, J.nr H-1–2014–101. The biomaterial from Belgian patients was kindly provided by the biolibrary of the Université Catholique de Louvain.

Results

Effects of TMZ

Eighteen of the 21 evaluable patients were given TMZ without a treatment-free interval. In 3 patients the drug was paused for 2, 5, and 10 months, respectively, according to local clinical practice, despite a remaining tumor mass. At the end of the first treatment period (median 6, range 2–23 months), 10 of 21 patients were responders (Table 3 and 4). In these patients the first effect on tumors was observed after a median period of 3 months (range, 1–6). Patient status at the last observation is shown in Table 3 and 4. Two patients had a complete regression and seven patients had a partial regression of tumor mass >30% in 3, >50% in 3, and >75% in 1. In the 16 patients with hormone secreting tumors, hormonal concentrations were decreased by 53–98%, and normalized in 3 patients. The changes were generally in concert with the changes in tumor volume, except in patient No. 6 in whom prolactin concentration was lowered by 71% without a change in tumor volume, Table 3. There was no detectable relationship between tumor subtype and effect of TMZ. A response was seen in 2/2 mixed GH-prolactin secreting tumors, 4/9 prolactinomas, 2/3 somatotropinomas, 1/3 corticotropinomas, and 1/4 nonfunctioning tumors.

Of the three patients in whom an effect of TMZ could not be evaluated, two are still alive (No. 1, 3). These patients were given radiotherapy and TMZ at the same time. In patient No. 1, the tumor remained stable at the last follow-up 17 months after TMZ discontinuation. In patient No. 3, complete regression remained at the 5 year follow-up. The third patient (No. 10) was treated with cabergoline after stopping TMZ. The tumor continued to grow and the patient died 20 months later.

MGMT immunostaining and response to TMZ

Most tumors, 17/21, displayed a homogeneous MGMT staining pattern (Figure 1 A and B). Three (No. 2, 6, and 17) had a heterogeneous pattern with areas of low and high contents (Figure 1C), and were not included in the statistical analysis since an absolute staining percentage could not be defined. In one patient with carcinoma there was not sufficient material to analyze MGMT. The tumor samples obtained at the last surgery preceding treatment with TMZ were used to study a relation between tumor markers and response to TMZ. Tissue from repeat surgeries was available in 12 patients; median time between the first and last surgery was 30 (3–84) months. Overall, MGMT was not significantly different between the first and last MGMT with the exception of two cases in which homogeneous staining of <10% and 20%, respectively, had changed to heterogeneous patterns ranging from 9% to 100%. The limited material in these cases did not allow firm conclusions regarding a true change.

MGMT staining was lower in patients responding to treatment with a median of 9% [range: (5–20)] compared to the nonresponders [median 93% (50–100), P < .001, Figure 2]. One patient (No. 13) with a nonfunctioning tumor had a decrease in tumor size by nearly 30% despite a high MGMT staining. The Ki-67 level in this case was low, 2%. The patient had received radiotherapy 6 months prior to start of TMZ. In the three patients who had a heterogeneous MGMT staining (No. 2, 6, 17) the response varied. One patient (No. 2) had a locally aggressive somatotropinoma, which partially responded to TMZ. The tumor recurred 15 months after drug discontinuation and then rapidly progressed leading to death. In the second patient (No. 6) who had a locally aggressive prolactinoma, the prolactin levels decreased by 71% but the tumor volume did not change. In the third patient (No. 17) an initial >50% regress of liver metastases was followed by a progress while on treatment. Addition of capecitabine during the last 2 cycles had no effect.

In three patients (No. 5, 19, and 24) in whom TMZ was paused after an initial response, the tumors progressed during the pause. A second treatment period was accompanied by either none (No. 5 and 19) or a lesser effect (No. 24) (Tables 3 and 4).

Mismatch DNA repair proteins and response to TMZ

MMR proteins were analyzed in 19 patients in whom there was sufficient tumor material. MLH1 stained positive in all patients. MSH2 and MSH6 were positive in all but one who had an initial response but became resistant to TMZ within 6 months.

Ki-67 labeling index, p53 immunostaining, and response to TMZ

Ki-67 labeling index was lower in LAPTs compared to carcinomas 7.5% (2–50%) and 20% (5–80%), respectively (P = .038). The index increased over time in nine of 16 patients where tissue from repeat surgeries was available, median 5% (range 2–70) vs 10% (2–80), first and last surgery, respectively, P = .008.

The p53 immunostaining was positive in 3 of 15 patients with LAPTs, and in 3 of 7 carcinomas (p = NS). Sufficient material for the analysis was not available in 2 patients. There were no relations between the response to TMZ, and the Ki-67 indices and the p53 immunostaining patterns, respectively.

Adverse events

Adverse events were reported in 13/24 (54%) patients [hematological (n = 4), tiredness (n = 8), others (including increase in liver enzymes, urinary tract infection and extreme fatigue, hearing loss, phlegmone) (n = 5)]. Most events were mild. TMZ was discontinued in three cases due to hearing loss after the first cycle, urinary tract infection, and extreme fatigue after the first cycle, and elevated liver enzymes after four cycles, respectively. Thrombocytopenia occurred in four patients, and the dose was reduced in two of these.

Discussion

This study reports on the largest cohort of patients with aggressive pituitary tumors treated with TMZ. All patients who received TMZ at the seven participating centers were enrolled, thereby avoiding reporting bias. The patients were monitored over a long time; the median duration of follow-up after TMZ discontinuation in initially responding patients was 32.5 (4–91) months. Overall, about 50% of the patients responded to TMZ and an MGMT level below 50% in tumor tissues predicted a response to treatment. After drug discontinuation, the effect was sustained in 7 of 10 of the responding patients.

Previous experience with TMZ is mostly based on single case reports and/or literature reviews of such reports (5–7, 20–24), although a few limited series of cases have been published recently. TMZ has usually been given as monotherapy. One study of four cases used a combination with capecitabine (26), a drug which is converted to 5-fluorouracil in vivo, and acts as a pyrimidine antagonist. In total, literature search revealed 74 cases, of which 31 were carcinomas. The largest series have included 6, 7, 8, and 13 patients, respectively (17–19, 27). In these four studies, treatment and follow-up periods from start of TMZ have ranged from 2 to 24 months (median 10), and 2 to 34 months (median 11), respectively. Response criteria have not been defined uniformly but expressed as >50% size reduction (27), >80% reduction (17), >20% decrease of tumor size (18), and >30% decrease of the largest tumor diameter (19). By use of these different criteria in the four studies, respectively, a total of 12 of the 34 patients had tumor shrinkage during the time of observation, with the first effect seen within 4 months.

In the present study, treatment with TMZ in 5/28 day cycles was accompanied by significant tumor shrinkage in 9/21 patients, and a clinically significant decrease of prolactin concentration in one. The effect of TMZ was not dependent on tumor type. In total, 10/21 (48%) were considered responders according to the study response criteria defined above. Two of these patients were partial responders during the first treatment period. After a treatment pause the effect was less pronounced during the second treatment period. Three of the 16 with LAPTs suffered from hydrocephalus and/or markedly impaired vision, which disappeared or was improved during treatment. In carcinomas, complete regression of metastases occurred in two out of eight patients. This effect persisted at 48 and 91 months of follow-up after discontinuation of the drug. Complete remission has never been reported in pituitary carcinomas with the commonly used combination of cytotoxic drugs, 5-fluorouracil and lomustine. Tumors progressed while on treatment in 8 of 21 patients. Nine patients are deceased, 2 from an unrelated cause. Addition of capecitabine in two of the unresponsive patients both having ACTH-producing carcinomas, had no effect. This is in contrast to a report on successful combination of TMZ and capecitabine in 4 patients with refractory Cushing's disease (26). In the latter study, a combination treatment was given from the start, which might be advantageous. In an in vitro model, a synergistic effect was apparent only when 5-fluorouracil was initiated before TMZ (26). Another difference is that unlike the two patients in the present study, MGMT was low, <15% in the patients of Zacharia et al. (26). Whether the combination is effective in pituitary tumors with high MGMT content remains to be studied.

In gliomas, levels of MGMT have been determined mainly by promoter methylation analysis, and less often by immunohistochemistry or direct enzymatic assays. In pituitary tumors, immunohistochemistry has been the method of choice; methylation status has not been closely associated with MGMT protein immunoexpresssion. Technical factors and yet unknown mechanisms for regulation of protein levels may contribute to this finding (28). In our experience, traditional 2 and 3 layer IHC techniques, as used for instance in Refs 19 and 20, does not always provide a satisfactory reaction considering signal/noise. The TSA technique has an advantage of improving specificity by the use of lower antibody concentrations, which reduces nonspecific binding. We introduced a quantitative way of determining MGMT levels resulting in an absolute percentage of positive nuclei. By this procedure, we found an inverse relationship between MGMT and the effect of TMZ, such that an MGMT level above 50% would indicate a low likelihood of an antitumoral response to the drug. In one outlier (pat no 13), tumor shrinkage of nearly 30% took place in spite of 100% MGMT staining (and low Ki-67) in the tumor specimen removed at surgery. The tumor had continued to grow and received stereotactic radiotherapy 6 months prior to TMZ. Possibly, the exceptional tumor shrinkage could be explained by TMZ having a synergistic effect with radiotherapy, as described in glioblastomas (14). Notably, marked responses occurred also in the other three patients treated with radiotherapy within 6 months of the first cycle of TMZ.

An antitumoral effect of TMZ requires an intact MMR pathway. In patients with glioblastoma, a treatment escape while on TMZ has been associated with loss of function of the MMR protein MSH6 (15). Inactivating mutations of MSH6 have been proposed to play a key role in the development of TMZ resistance when MGMT expression is low (16, 29). In the present study, absence of both MSH 2 and MSH 6 immunoexpression was found in a single patient who rapidly developed resistance to the drug. p53 status has been reported to predict response to TMZ (30), but we did not find such a relation.

To date, TMZ has mostly been used as salvage treatment when other treatments have failed, but the drug could probably be considered at an earlier stage. One challenge is to identify possible aggressive/malignant tumors already at the first presentation. Of the 16 LAPTs 9 initially behaved as typical benign adenomas and later acquired a more aggressive growth behavior. Six prolactinomas concomitantly lost their sensitivity to dopaminergic drugs, a phenomenon previously described (31). Another prolactinoma acquired an ability to secrete GH in parallel with rapid tumor growth, and in two of the carcinomas, cosecretion/overt secretion of GH coincided with the appearance of metastases. One of these carcinomas initially secreted prolactin only, and the other was a silent somatotropinoma. In the literature, there are three case reports linking a gain of/shift in hormone secretion to an onset of a more aggressive behavior; a nonfunctioning adenoma transforming into a TSH and prolactin secreting carcinoma (32), an initially silent somatropinoma (33), and an initially benign microprolactinoma (34), both evolving into aggressive somatotropinomas. In one of these cases, the tumor had acquired a GNAS gene mutation (34). There are also rare observations of benign prolactinomas switching to benign somatropinomas during treatment with dopamine agonists (35). The collective findings, however, emphasize that a change in the secretory phenotype of a pituitary adenoma should alert the treating physician of a possible evolvement of an aggressive tumor.

Pituitary carcinomas are rare (36) and generally appear at diagnosis as invasively growing macroadenomas with significant suprasellar extensions (37). In contrast to benign adenomas, where corticotropinomas are in the minority and somatotropinomas the third most prevalent tumor, the carcinomas are more frequently corticotropinomas (30–40% of cases) while only about 5% secrete GH (4, 38). Most carcinomas have a Ki-67 proliferation index ≥12%, a high number of mitoses and/or nuclear pleomorphic features, and a positive nuclear staining for p53 (39). The tumors disseminate systemically to lymph nodes, liver, bone, lung, or/and within the CNS to cerebral cortex, cerebellum, spinal cord, and leptomeninges. The time to metastases ranges from a few months to 18 years with a median of 5–7 years (4, 37). The three corticotropinomas in the present series had the shortest time to metastases. This could be a chance finding, as other studies have not found an association between tumor subtypes and propensity for early dissemination.

There is a need to identify premetastatic pituitary carcinomas that might benefit from early aggressive therapy (40). Upon examination of the last removed tissues before the start of TMZ, seven of the eight carcinomas in the present series had a Ki-67 index >10%, and three had a strong positive p53 immunostaining. However, the Ki-67 indices were generally lower in tissues from the first surgery, eg, before development of metastases, and based on these specimens, a carcinoma was suspected only in one case.

TMZ is an alkylating agent and MGMT repairs the DNA by transferring the alkyl group to a cysteine in its own active center, which thereby gets inactivated (“suicide enzyme”) (12). The drug has mostly been administered in cycles, 5/28 days, or continuously in a lower daily dose, a regimen considered to deplete MGMT stores more avidly. It is not known which regimen offers a better long-term outcome. Another question is when to discontinue treatment. We observed that a second treatment period after pausing the drug seemed to be less effective. Likewise, in a tumor with patchy aggregations of MGMT (9–100% of tumor cells) and a 50% decrease in tumor volume while on TMZ, a late recurrence of a rapidly growing tumor occurred after cessation of TMZ. It is conceivable that cells with a low MGMT content had been successfully targeted, and subsequently been replaced by MGMT positive, and thus TMZ resistant clones. Collectively, the data could speak against a routine use of structured treatment interruption (STI).

Limitations with this study are that the data collection is retrospective, and that there may be a risk of underreporting adverse events, in particular events not leading to dose reductions or cessation of treatment. The number of patients is by necessity relatively small, given the rarity of the condition. A strength is that all patients treated at our hospitals are included. The material is therefore likely representative of patients treated with TMZ today. Another strength is that the tumor specimens were analyzed at a single laboratory.

To summarize, this study extends previous observations on temozolomide as a useful drug in patients with aggressive pituitary tumors, and establishes a role of MGMT as a biomarker of response. Future prospective randomized studies on treatment regimens are needed. Synergistic effects with radiotherapy and cytotoxic drugs deserve to be addressed, as well as the potential value of MGMT inhibitors. At the present stage, given the risk benefit profile of TMZ, a 3-month trial seems indicated in any patient with an aggressively growing pituitary tumor where other options of treatment may be limited.

Acknowledgments

This work was supported by grants from Lund University, and the Arvid Nilsson Fund.

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• LAPTs

  locally aggressive pituitary tumors

 
• MMR

  mismatch repair system

 
• STI

  structured treatment interruption

 
• TSA

  tyramide signal amplification.

References