P16 Loss and Mitotic Activity Predict Poor Survival in Patients with Peritoneal Malignant Mesothelioma

Malignant mesothelioma is an aggressive neoplastic proliferation derived from cells lining the serosal membranes. Histologic patterns of malignant mesothelioma can be diverse; broadly, however, tumor growth patterns are described as epithelial or sarcomatous. Mixtures of the two patterns are described as biphasic. Of the ∼2,500 new cases of mesothelioma expected in the United States per year, 25% of those cases will arise in the peritoneum (1). Peritoneal mesothelioma has a historical median survival of <1 year (2). However, a significant proportion of patients survive for extended periods, with intensive therapy (1, 3–5).

The current therapy for peritoneal malignant mesothelioma consists of a multimodality approach that incorporates tumor debulking, i.p. chemotherapy, and radiation therapy. A second operation done (“2nd look”) after therapy can determine extent of residual tumor. Using this approach, long-term survival can be achieved, with a reported median survival of 92 months (5, 6).

For patients treated with multimodality regimens, several clinical variables have been shown associated with long-term survival. These include age, gender, histologic subtype (epithelial versus biphasic/sarcomatous), extent of debulking (cytoreduction score), invasiveness, metastatic status, incidental diagnosis, and second-look operation (4, 5, 7, 8). With the exception of second-look operation, these variables can be assessed using a multidisciplinary approach by oncologist, surgeon, and pathologist at the time of initial surgery.

Pathologic variables that identify prognostically distinct subsets have been studied in malignant pleural and peritoneal mesothelioma. Prior pathologic series have identified localized mesothelioma (9) and well-differentiated papillary mesothelioma (10) as mesothelioma subtypes associated with favorable prognosis. Once the analysis is confined to diffuse malignant mesothelioma, traditional pathologic variables of nuclear grade, subtype of epithelial histology, mitotic activity, and necrosis (3, 9) have not shown to be of significant predictive value of survival.

It is reported that tumors with a malignant spindle cell component (biphasic/mixed or sarcomatous histologic subtypes) tend to be locally aggressive, bulky, and associated with poor survival. Because this histologic subtype has a higher frequency of p16 loss than epithelial subtype (11), we hypothesized that p16 loss may represent a marker of poor prognosis in all malignant mesotheliomas.

Previously reported mesothelioma clinicopathologic predictive outcome variables do not identify all patients in which tumor progresses despite therapy. As new agents become available, the identification of patient subgroups that are most likely to benefit from aggressive initial therapy may enhance treatment allocation and lead to further progress in treatment of this disease. We report our institutional experience with peritoneal malignant mesothelioma, focusing on histologic variables and p16 immunoreactivity and their association with survival.

We identified 56 patients with peritoneal mesothelioma that were diagnosed and treated at Columbia University Medical Center during the period January 1995 to August 2003. These patients were evaluated for inclusion in a protocol to receive trimodal therapy (tumor debulking, i.p. chemotherapy, and radiotherapy). Patients received trimodal therapy, systemic chemotherapy, or no therapy. The treatment regimen was determined in part by ability to place i.p. catheters, ability to successfully debulk tumor, and patient health status. Second-look operations were done after completion of the i.p. phase of chemotherapy. All patients in the trimodal subgroup were potential candidates for second-look operation; poor patient health status after chemotherapy precluded second-look operations in 12 patients. Patient follow-up and survival time were recorded up to December 2003. The study was approved by the Columbia University Institutional Review Board.

Second-look operation data and cytoreduction score data were reviewed using detailed surgical reports and pathologic gross descriptions. Tumor debulking was recorded using the cytoreduction scoring system (5) in which: score 0, no residual tumor; score 1, <0.25 cm greatest tumor nodule; score 2, 0.25 to 2.5 cm greatest tumor nodule; and score 3, >2.5 cm or confluent.

Pathologic diagnoses were confirmed by two pathologists' review (A.B. and H.H.). In addition to morphologic assessment, a standard immunohistochemistry panel (calretinin, cytokeratin 5/6, WT-1, CEA, LeuM1, B72.3, and Ber EP4) was done to establish the diagnosis. Two cases were excluded from further study after morphologic assessment. In one case, the diagnosis was a well-differentiated papillary mesothelioma, determined using the criteria of Butnor et al. (12) of papillary growth with single layer of low–nuclear grade cells. The second was a primary pleural mesothelioma with secondary involvement of the peritoneum. Both pathologists scored additional pathologic variables independently; discrepancies were reviewed to achieve a consensus. The following variables were evaluated pathologically: histologic subtype (epithelial, biphasic, and sarcomatous), percent sarcomatous histology, nuclear grade, mitotic activity, necrosis, extent of invasion/invasion score, lymphoid response, stromal reaction/desmoplasia, organ involvement, nodal involvement, and predominant epithelial growth pattern. A tumor was diagnosed as biphasic if any sarcomatoid component was identified within the tumor. For nuclear grade, a score of 1 to 3 was used: 1, round uniform nuclei; 2, irregular nuclear contours, small nucleoli; 3, pleomorphism and markedly irregular nuclear contours and prominent nucleoli. For mitotic activity, the highest mitotic count per 10 high-power fields of three separate counts was recorded. For invasion score, an estimate of the total area of invasive tumor as a proportion of the overall tissue section was made and graded as follows: tumors with <25% of invasive growth were scored 1, moderate (25-75%) invasive growth scored 2, and extensive invasive growth (>75%) scored 3. For lymphoid response and stromal reaction/desmoplasia, cases were scored as 1, low; 2, moderate; or 3, extensive.

Tissue microarrays. Tissue blocks from 51 mesotheliomas were obtained and four tissue cores from each tumor were used to construct three tissue microarrays, using a MTA-1 arrayer (Beecher Instruments, Sun Prairie, WI) as described previously (13). For biphasic tumors, cores were obtained from epithelial and sarcomatous areas.

P16 immunohistochemistry. Tissue microarrays and paraffin sections (two cases) were used for p16 immunohistochemistry. Two different p16 clones were used in independent experiments using different antigen retrieval methods. For the first experiment, mouse monoclonal antibody 6H12 (Novocastra Laboratories, Newcastle, United Kingdom, 1:40 dilution) with antigen retrieval with Trilogy (Cell Marque, Hot Springs, AK) with EDTA for 40 minutes in a steamer was used. Primary antibody was incubated at room temperature for 1 hour and stained using a DAKO autostainer. For the second experiment, mouse monoclonal antibody 16P07 (Labvision, Fremont, CA) was used at a concentration of 4 μg/mL for 1 hour at room temperature, after antigen retrieval in 10 mmol/L citrate buffer (pH 6.0) in a steamer at 99°C for 40 minutes. Slides were stained manually using standard avidin-biotin complex protocol (Vector Laboratories, Burlingame CA). p16 immunoreactivity was nuclear and cytoplasmic and scored for intensity and extent using the following method: 0, no staining; 1+, weak staining; 2+, strong staining for intensity and focal (<5%) versus multifocal/diffuse for extent. Cases were considered negative if 0 or 1+ focal; positive cases were 1+ multifocal/diffuse, and all 2+ cases.

Statistical analysis. For the comparisons of means, a t test for independent variables was used for continuous data and χ² was used for categorical data. For both the entire data set and the trimodal subgroup, univariate analysis of survival was done using Cox regression analysis. For multivariate analysis in the entire data set, variables were adjusted for histologic subtype (epithelial or biphasic) and treatment (trimodal or other/none) using Cox regression analysis. For multivariate analysis in the trimodal subgroup, variables were adjusted for histologic type only. For median survival data, Kaplan Meier survival analysis was done, using a log-rank statistical method. All statistical analyses were done using the SPSS for Windows version 11.5.0 (SPSS, Inc., Chicago, IL).

Of the 54 patients with the diagnosis of primary peritoneal malignant mesothelioma, 39 received trimodal therapy, 10 received no therapy, and 4 received primary systemic chemotherapy-[cisplatinum/gemcitabine (n = 2), doxorubicin (n = 1), carboplatinum/gemcitabine (n = 1)]. Treatment information was not available for one patient. Clinical variables are summarized in Table 1. Survival time was longer in patients receiving trimodal therapy, and the performance of second-look operation was associated with increased survival time. Cytoreduction score also differed between the two subgroups; successful cytoreduction has effect on selection for trimodal therapy.

Histologic subtype (epithelial versus biphasic) has previously been shown associated with prognosis. The patterns of epithelial histology in our series included papillary, tubular, solid, microcystic, and single cell infiltration. In some cases of solid growth, tumors had a “hepatoid” appearance. No association was detected between histologic pattern of the predominant epithelial component and survival, including comparisons between tubulopapillary and solid growth. Of the biphasic cases, the estimated percentage of spindle cell component ranged from 5% to 95%, with an average of 37%. No particular epithelial pattern was associated with the biphasic cases. Four cases had ≤10% sarcomatous component. The extent of sarcomatous component in the biphasic tumors was not associated with survival time (data not shown).

Descriptive statistics of pathologic variables are summarized in Table 2. Increased mitotic rate, necrosis, high invasion score, organ involvement, and negative p16 immunoreactivity were all detected more frequently in patients that did not receive trimodal therapy.

Statistical analysis of clinical and histopathologic variables with survival time identified several variables associated with prognosis (Table 3). In univariate analyses, clinical variables such as trimodal treatment and low cytoreduction score were associated with decrease in risk of death. Among pathologic variables, biphasic histologic subtype, increased mitotic rate, organ involvement, high invasion score, high nuclear grade, necrosis and p16 loss were all associated with increased risk of death.

Because trimodal treatment was associated with survival (Table 1) and was a potential significant confounder, treatment type was incorporated in our multivariate analysis model. Similarly, biphasic histologic subtype was incorporated into the multivariate model. After adjustment for both treatment and histologic subtype, no second-look operation, higher mitotic rate, mitotic rate >4, and p16 loss remained significantly associated with risk of death. Of interest, the association of histologic type with survival seemed dependent on p16 status, indicating a relationship between p16 status and histologic type. In fact, p16 immunoreactivity was absent in 13 of 14 biphasic tumors (Fig. 1).

Focusing on the trimodal treated group (n = 39), univariate analysis of clinical and histologic variables indicated that no second-look operation, biphasic histologic subtype, p16 loss, higher mitotic count, necrosis, stromal response, organ involvement, and nodal involvement were associated with increased risk of death. Our multivariate model, adjusting for histologic subtype as a known variable influencing prognosis, indicated that no second look and higher mitotic activity (including mitotic rate above 4) were associated with increased risk of death (Table 4).

The overall median survival for all patients in this study is 34 months. For the trimodal subgroup, the median survival is not yet reached, and the mean survival is 78 months. The effect of the variables determined by multivariate analysis to be significantly associated with median survival for all patients and for the trimodal subgroup, is summarized in Table 5. These data complement the Cox regression analysis, showing significant reductions in median survival for p16 loss and mitotic rate above 4 in both groups and for no second-look operation in the trimodal group.

Malignant peritoneal mesothelioma research has yielded significant progress in treatment approaches, stratification of patients, and development of effective therapy (1, 4, 7). This progress has been due to a combination of refinements in pathologic diagnosis and development of combined treatment approaches that take advantage of the unique growth pattern of peritoneal mesothelioma. The accuracy of diagnosis of mesothelioma has been improved by electron microscopy and the discovery of immunohistochemistry markers that distinguish mesothelioma from adenocarcinoma (14–16). As bulky abdominal tumor is linked with complications of intractable ascites and colon obstruction, elements of therapy are focused on reduction of tumor burden. In this regard, surgical debulking, i.p. chemotherapy and radiotherapy are used for cytoreduction. A second-look operation determines the success of the above therapy and provides an opportunity for further reduction of residual tumor burden (1). A significant proportion of patients, however, does not respond to therapy. This may be because of local invasion, tumor resistance to therapy causing progression of the disease before initiation or completion of trimodal therapy, or recurrent disease becoming intractable or metastatic.

Variables predictive of peritoneal mesothelioma outcome have been reported previously. These variables include initial tumor burden, biphasic/sarcomatous histology, success of cytoreduction, gender, and second-look operation (4, 5). Analyses of pathologic variables have not yielded definite conclusions, although patients with exclusively tubulopapillary histology have been reported to have improved survival (3, 17). Mitotic rate has been previously examined, with the observation that many mesotheliomas have relatively low mitotic rates. Our series supports the observation that malignant mesotheliomas have a low mitotic rate, with only an upper quartile of patients with mitotic rates above 4 in 10 high-power fields. Despite the overall low mitotic rate in our series, survival time was independently inversely associated with mitotic rates above 4 in 10 high-power fields in the entire patient set and in the subgroup of patients completing trimodal therapy.

P16 is a tumor suppressor, known to prevent progression through the G₁-S restriction point of the cell cycle by blocking the action of CDK4/6 (18). Germ line mutations of P16 were first described in familial melanoma (18), but subsequently, multiple tumor types have been shown to exhibit somatic p16 loss. Mechanisms of p16 loss include deletions, point mutation, and epigenetic silencing by promoter hypermethylation (for review, see ref. 19). It has been recognized that deletion of p16 is a frequent finding in malignant mesothelioma cell lines (11) and that p16 loss can be identified in tumor tissue from pleural and peritoneal mesothelioma, although at a lower frequency than in mesothelioma cell lines (11, 20, 21). In one study of 45 mesotheliomas, 33% had altered p16, with 22% showing homozygous deletion, 9% showing promoter hypermethylation, and 2% containing a point mutation (21). A high frequency of p16 loss has also been reported in tumor tissues from biphasic mesothelioma when compared with epithelial type (11). Restoration of p16 activity has been a target for therapeutic intervention in malignant mesothelioma. Mesothelioma cell lines treated with adenoviral vectors containing p16 underwent G₁ cell cycle arrest (22).

P16 loss in malignant mesothelioma has indirect therapeutic implications as well. In pleural mesothelioma, it has been shown that homozygous deletion of p16 is frequently associated with co–deletion of MTAP, which encodes methylthioadenosine phosphorylase (Mtap) and is similarly located on chromosome 9p21 (23). This enzyme is part of the salvage pathway that leads to production of adenine for synthesis of AMP. Loss of Mtap leads to cellular dependence on purine biosynthetic pathways for the production of AMP. In cells that are MTAP deficient, l-alanosine (inhibitor of de novo purine synthesis) therapy can block the cellular production of AMP with a resultant toxic effect. This has been shown in hematopoietic malignancies that lack MTAP activity (24, 25). Given the observation that p16 loss in mesothelioma is frequently due to homozygous deletion, the frequent codeletion (91%) of p16 and MTAP suggests that p16 deficient tumors may have greater sensitivity to l-alanosine therapy.

Our immunohistochemistry data confirm a frequent loss of p16 immunoreactivity in peritoneal mesotheliomas, with nearly universal loss in biphasic peritoneal mesothelioma. Based on immunohistochemistry, our rate of p16 loss is 48%, which is intermediate between that detected by Illei et al. (74%) and that reported by Hirao et al. (31%) in pleural mesothelioma. This may reflect differences between peritoneal and pleural tumor as well as differences in detection techniques.

In our series, immunohistochemistry showed p16 loss in both epithelial and sarcomatous components, even in cases where sarcomatous component was ≤10%. Also notable was the association of p16 loss and survival time, even after adjustment for treatment subgroup and histologic subtype. Within the trimodal subgroup, p16 status was also associated with survival; this association was not detected after adjustment for histologic type in the model. The reason for this difference between subgroups is unclear, but given the relatively small number of cases (N = 39) and small number of events (n = 14), it is possible that the multivariate analysis within the trimodal subgroup is underpowered to evaluate the separate effect of histologic type and p16 loss on survival.

Second-look operation in the trimodal subgroup is a variable that is independently associated with survival time in our series. This observation has been previously reported in peritoneal mesotheliomas (4, 7). The association between performance of a second-look operation and clinical outcome may indicate a therapeutic benefit of the procedure or may be a surrogate indicator of independent clinical variables such as patient performance status, initial cytoreduction, and response to chemotherapy. This is an important question that is beyond the scope of the present study.

Further molecular studies including gene expression profiling may assist in discovery of variables predictive of primary disease progression requiring early systemic therapy and may guide selection of agents. Multimodality therapy may need to be modified in patients within a potential poor prognosis subset.

In summary, the present series shows that biphasic histology, mitotic rate above 4 in 10 high-power fields, and p16 loss may predict a poor outcome subgroup in peritoneal mesothelioma patients.

We thank Shing Lee of the Department of Biostatistics for assistance in the statistical analysis.