Low-molecular weight forms of cyclin E differentiate ovarian carcinoma from cells of mesothelial origin and are associated with poor survival in ovarian carcinoma

Authors

  • Ben Davidson MD, PhD,

    Corresponding author
    1. Pathology Clinic, Radiumhospitalet-Rikshospitalet Medical Center, Oslo, Norway
    2. Faculty Division Radiumhospitalet, the Medical Faculty, University of Oslo, Oslo, Norway
    • Department of Pathology, Radiumhospitalet-Rikshospitalet Medical Center, Montebello N-0310 Oslo, Norway
    Search for more papers by this author
    • Fax: (011) 47 22508554

  • Martina Skrede,

    1. Pathology Clinic, Radiumhospitalet-Rikshospitalet Medical Center, Oslo, Norway
    Search for more papers by this author
  • Ilvars Silins MD, PhD,

    1. Department of Gynecology, Radiumhospitalet-Rikshospitalet Medical Center, Oslo, Norway
    Search for more papers by this author
  • Ie-Ming Shih MD, PhD,

    1. Department of Pathology, Johns Hopkins Medical Institutions, Baltimore, Maryland
    2. Department of Gynecology and Oncology, Johns Hopkins Medical Institutions, Baltimore, Maryland
    Search for more papers by this author
  • Claes G. Trope MD, PhD,

    1. Faculty Division Radiumhospitalet, the Medical Faculty, University of Oslo, Oslo, Norway
    2. Department of Gynecology, Radiumhospitalet-Rikshospitalet Medical Center, Oslo, Norway
    Search for more papers by this author
  • Vivi Ann Flørenes PhD

    1. Pathology Clinic, Radiumhospitalet-Rikshospitalet Medical Center, Oslo, Norway
    Search for more papers by this author

Abstract

BACKGROUND.

The authors recently reported on the role of cyclin E in differentiating ovarian/primary peritoneal carcinoma from malignant peritoneal mesothelioma using gene expression arrays. In the current study, they analyzed the expression of low-molecular weight (LMW) forms of cyclin E in ovarian carcinoma, malignant mesothelioma, and benign reactive effusions.

METHODS.

Cyclin E protein expression was analyzed in 98 effusions (72 ovarian carcinomas, 14 malignant mesotheliomas, and 12 reactive specimens) using immunoblotting. Sixty-two ovarian carcinoma effusions were studied further for cyclin E expression using immunohistochemistry. The correlations between cyclin E expression in ovarian carcinoma and clinical parameters, including chemotherapy response, were analyzed.

RESULTS.

LMW forms of cyclin E were identified in 54 of 72 ovarian carcinoma effusions (75%) compared with 1 of 14 malignant mesothelioma effusions (7%) and 1 of 12 reactive effusions (8%) (P < .001). Their presence in ovarian carcinoma was associated with a higher percentage of cyclin E-positive cells (P = .001) and increased staining intensity (P < .001) using immunohistochemistry. The presence of LMW forms of cyclin E was correlated with shorter overall survival (P = .021) and progression-free survival (P = .020). The presence of a higher percentage of cyclin E-positive cells using immunohistochemistry was correlated with shorter progression-free survival (P = .026). No association with chemotherapy response was observed.

CONCLUSIONS.

LMW forms of cyclin E differentiated ovarian carcinoma from benign and malignant mesothelial cells and were associated with increased protein expression using immunohistochemistry. The expression of LMW cyclin E forms was not associated with chemotherapy response, although it may be a marker of aggressive disease in patients with metastatic ovarian carcinoma. Cancer 2007. © 2007 American Cancer Society.

One of the main survival advantages of cancer cells compared with their normal counterparts lies in their loss of replication control and deregulation of the cell cycle. The exit of cells from the quiescent state and progression along the cell cycle are regulated at 3 checkpoints—the G0→G1, G1→S, and G2→M transitions. These events are mediated by members of the cyclin and cyclin-dependent kinase (Cdk) families. Cyclins, the regulatory unit, bind Cdk, the catalytic component, in a specific manner.1, 2 The cyclin E-Cdk2 complex mediates the G1→S transition through phosphorylation and, thereby, inactivates the retinoblastoma protein (Rb) by releasing the E2F transcription factor. Among the additional proteins that are regulated by the complex is p27kip1, a cyclin-Cdk inhibitor, the transcription cofactor CBP/p300, and NPAT/p220, a transcription factor that controls histone gene transcription.1

The levels of cyclin E (also termed cyclin E1), a 50-kilodalton (kD) nuclear protein, are subject to a variety of molecular control mechanisms, including positive regulation by Rb inactivation and increased E2F activity, regulation by several transcription factors, alternative splicing, and degradation through proteolysis.1 A second member of the family, cyclin E2, has 47% homology with cyclin E1 and has closely similar activity and regulation.1 Cyclin E expression is deregulated in a variety of epithelial and nonepithelial cancers through amplification of the cyclin E gene itself or by mutations in other components of related pathways.1

Ovarian carcinoma (OC) is the most lethal gynecologic malignancy in the Western world.3 OC and the closely related and morphologically indistinguishable primary peritoneal carcinoma (PPC), or tubal serous carcinoma, presents at advanced stage in 70% of patients, manifested most commonly as diffuse involvement of the peritoneal surface with the formation of solid lesions and accumulation of tumor cells in ascitic fluid. Platinum compounds and taxanes are the mainstay of chemotherapy for patients with OC/PPC. Although the 5-year survival rate for all stages improved from 39% during 1975 through 1979 to 43% during 1990 through 1999, it remains disappointingly low,4 largely because most patients present with metastatic disease5 and because of primary or acquired drug resistance.6

Cyclin E amplification has been reported in OC7–9 and is the most common gene amplification in serous carcinoma of the ovary,10 suggesting that deregulation of this molecule is involved in the development and progression of OC. However, to our knowledge, data are inconclusive regarding the clinical significance of cyclin E protein expression in OC. Three studies illustrated a significant correlation between higher cyclin E expression and poor progression-free survival (PFS) and/or overall survival (OS) in patients with OC,9, 11, 12 whereas 4 other studies did not report such a correlation.13–16

Porter et al. reported that low-molecular weight (LMW) forms of cyclin E are produced in breast carcinoma cell lines through the action of the enzyme elastase and that these forms have greater biologic activity than the full cyclin E molecule.17 Later, it was demonstrated that the presence of these forms was correlated with poor survival in clinical breast carcinoma.18 LMW cyclin E forms subsequently were identified in 17 of 25 primary OCs and were associated with increased phosphorylation activity of histone H1.19 Overexpression of LMW forms in MDAH-2774 cells was associated with increased cyclin E kinase activity, increased S-phase fraction and decreased doubling time, and greater sensitivity to cisplatin.19 To our knowledge, the prognostic role of LMW cyclin E forms in OC has not been studied.

We recently reported on the significantly higher level of cyclin E1 gene expression in OC/PPC effusions compared with diffuse malignant peritoneal mesothelioma (DMPM).20 DMPM is a native cancer of the serosal cavities that has the same clinical presentation as OC/PPC and is closely related to OC/PPC in terms of histogenesis, morphology, and the expression of diagnostic markers, including differentiation markers.21, 22 Immunohistochemical analysis of cyclin E expression in solid tumors and effusions from both tumor types confirmed this difference at the protein level.20 In the current study, we studied the presence of LMW cyclin E forms in OC/PPC, malignant mesothelioma (MM), and benign effusions that contained reactive mesothelial cells (RM) using immunoblotting. Furthermore, we analyzed the association between the presence and prognostic role of cyclin E LMW forms and chemotherapy response in OC/PPC effusions. Our data show that the presence of cyclin E LMW forms differentiate OC/PPC from both MM and RM cells and is associated with poor survival in patients with OC/PPC.

MATERIALS AND METHODS

Patients and Materials

The clinical material consisted of 98 effusions, all of which were submitted for routine diagnostic purposes to the Pathology Clinic at Rikshospitalet-Radiumhospitalet Medical Center during 1998 through 2004. The Regional Committee for Medical Research Ethics in Norway approved the study.

Seventy-two fresh, nonfixed, malignant peritoneal (n = 55 samples) and pleural (n = 17 samples) adenocarcinoma effusions were obtained from 63 patients (52 patients with OC, 3 patients with tubal serous carcinoma, and 8 patients with primary serous peritoneal carcinoma). Because of their closely linked histogenesis and phenotype, all of these tumors are referred to in the remainder of this report as OC. OC specimens and clinical data were obtained from the Department of Gynecology, Rikshospitalet-Radiumhospitalet Medical Center (Table 1). The agents used in first-line and second-line chemotherapy are detailed in Table 2. The International Federation of Gynecology and Obstetrics (FIGO) classification was used for both grading and staging.

Table 1. Clinicopathologic Data of the Ovarian Carcinoma Cohort (63 Patients)
ParameterNo. of patients
  • FIGO indicates International Federation of Gynecology and Obstetrics; NA, not available.

  • *

    Including 3 patients with clear cell carcinoma.

  • Six effusions were from inoperable patients in whom the biopsy specimens were too small for grading, and 1 patient had no record.

  • Six patients were inoperable, and 4 patients had no records.

  • §

    Including mixed epithelial tumors and carcinomas not otherwise specified.

  • Four inoperable patients, where biopsy had too little tissue to establish histologic type.

  • For 72 effusions.

Mean age (range), y61 (34–82)
FIGO stage
 II2
 III32
 IV29
Grade
 16
 215
 3*35
 NA7
Residual disease, cm
 ≤122
 >131
 NA10
Histology
 Serous50
 Mucinous1
 Clear cell3
 Endometrioid1
 Undifferentiated2
 Other§2
 NA4
Effusion site
 Peritoneal55
 Pleural17
Table 2. First-line and Second-line Chemotherapy Regimens for Patients With Ovarian Carcinoma
AgentFirst-line (n = 62 patients)*Second-line (n = 58 patients)
  • TEC indicates combined paclitaxel, epirubicin, and carboplatin.

  • *

    One patient had no record.

  • One patient had no record, and 4 patients did not receive second-line chemotherapy.

Carboplatin and paclitaxel385
Carboplatin719
Paclitaxel25
Weekly paclitaxel012
Gemcitabine01
TEC90
Nolvadex17
Mithoxantrone01
Other58

Fourteen MM effusions (10 pleural, 4 peritoneal) were from patients who were diagnosed with epithelioid or biphasic MM in biopsy specimens. Twelve RM specimens (10 pleural and 2 peritoneal) were obtained from patients with different underlying cancers or clinical suspicion of malignancy.

Effusions were centrifuged within minutes after aspiration from patients, aliquoted, and frozen in RPMI medium supplemented with 20% fetal calf serum and 20% dimethyl sulfoxide. The diagnoses of all specimens were validated by cytopathologists using morphologic criteria and immunohistochemistry (ICH) with a broad antibody panel that recognizes carcinoma or mesothelial cells.23 Based on these analyses, only specimens with >50% target cells were selected for the current study.

Immunoblotting

Cells were lysed in ice-cold NP-40 lysis buffer (1% NP-40; 10% glycerol; 20 mM Tris-HCl [pH 7.5]; 137 mM NaCl; 100 mM sodium vanadate; 1 mM phenylmethyl sulphonyl fluoride; 0.02 mg/mL each of aprotinin, leupeptin, and pepstatin; and 10 μL/mL each of phosphatase inhibitor cocktails I and II). All protease inhibitors were purchased from Sigma Aldrich (St. Louis, Mo). Lysates were sonicated followed by centrifugation to collect the lysate supernatant fluid. Protein quantitation was performed by Bradford analysis, and 25 mg of protein per lane were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Proteins were transferred to polyvinylidine fluoride immobilon membranes (Millipore, Bedford, Mass) and subsequently hybridized to cyclin E1 (HE12) antibody, as recommended by the manufacturer (Santa Cruz Biotechnology, Santa Cruz, Calif).

ICH

Protein expression of cyclin E was analyzed in 62 of the 72 above-described OC effusions from 54 patients. IHC was performed using the EnVision + peroxidase system (Dako, Glostrup, Denmark). The cyclin E monoclonal antibody was purchased from Novocastra (Newcastle upon Tyne, U.K.) and was used at 1:100 dilution with antigen retrieval in ethylenediamine tetracetic acid buffer. Appropriate positive and negative controls were used.

Staining extent was scored by an experienced cytopathologist (B.D.) using a scale of from 0 to 4 corresponding to the percentage of immunoreactive tumor cells (0%, 1–5%, 6–25%, 26–75%, and 76–100%, respectively). Staining intensity was scored using a scale from 0 to 2 corresponding to absent, weak, or strong expression, respectively. No specimen contained <100 tumor cells.

Statistical Analysis

Statistical analyses were conducted using the SPSS-PC software package (version 13; SPSS, Inc., Chicago, Ill. Probability (P) values <.05 were considered statistically significant. The association between cyclin E LMW form expression and specimen type (OC, MM, or RM) was analyzed using the 2-sided chi-square test, as was the association with clinicopathologic parameters and with cyclin E expression by IHC. Univariate survival analyses of OS and PFS were performed using the Kaplan-Meier method and the log-rank test. For these analyses, cyclin E IHC expression categories were grouped, scoring the extent of expression as focal (≤25% of cells) versus diffuse (>25% of cells) and the intensity of expression as absent/weak versus strong. For patients who had >1 effusion, expression in the first specimen was analyzed. Multivariate analyses for OS and PFS were performed using the Cox proportional hazards model.

RESULTS

LMW Cyclin E Forms Are Expressed Predominantly in OC

LMW cyclin E forms (bands of smaller size than the predicted 50-kD full protein) were observed in 54 of 72 OC effusions (75%) with immunoblotting. Of these, 20 effusions had weak expression, and 34 effusions had strong expression (Fig. 1). Analysis of MM specimens showed weak expression in 1 of 14 effusions (7%) and no expression in the remaining 13 effusions (Fig. 1). RM specimens demonstrated strong expression in 1 of 12 effusions (8%) and no expression in the remaining 11 effusions. The difference between OC, MM, and RM effusions was statistically significant (P < .001). These data suggest that the presence of LMW cyclin E forms is characteristic of OC but constitutes a rare event in cells of mesothelial origin.

Figure 1.

Immunoblotting. Analysis of 12 ovarian carcinoma (OC) effusions (top row) and 10 malignant mesothelioma (MM) effusions (lower row). Low-molecular weight (LMW) forms of cyclin E were found in 10 of 12 OC specimens (strong in Lanes 1–5, 8, 11, and 12; weak in Lanes 9 and 10). The effusions in Lanes 6 and 7 were negative. All MM effusions were negative except for the weakly expressing effusion in Lane 6. The positive control for cyclin E expression in the pilot study, the WM 1341B melanoma cell line treated with trichostatin A, also is shown (C). kD indicates kilodalton.

The Presence of LMW Cyclin E Forms Is Associated With Increased Cyclin E Expression by IHC

Analysis of cyclin E expression using IHC showed nuclear expression of this protein in 59 of 62 OC specimens (95%). Expression (as the percentage of immunostained cells) was scored as 1 in 11 effusions, scored as 2 in 21 effusions, scored as 3 in 18 effusions, and scored as 4 in 9 effusions. Staining intensity was weak in 19 effusions and strong in the remaining 40 positive specimens (Fig. 2A-D). The presence of LMW cyclin E forms in OC effusions was associated with a higher percentage of cyclin E-positive cells (P = .001) and increased staining intensity (P < .001) using IHC (Table 3). These data suggest that stronger and more diffuse cyclin E expression using IHC reflects the presence of LMW cyclin E forms.

Figure 2.

Immunohistochemistry analysis. (A and B) Two ovarian carcinoma (OC) effusions with strong expression of cyclin E in >75% of cells. (C and D) Two OC effusions with weak cyclin E expression in <25% of cells.

Table 3. Association Between the Presence of Low-molecular Weight Forms of Cyclin E Using Immunoblotting and Cyclin E Immunostaining in 62 Ovarian Carcinoma Effusions
VariableLMW cyclin E formsP
Absent (n = 13 patients)Present (n = 49 patients)
  1. LMW indicates low-molecular weight.

Cyclin E staining extent, % positive cells
 0%30 
 1–5%47 
 6–25%516 
 26–75%018 
 76–100%18.001
Cyclin E staining intensity
 Absent30 
 Weak712 
 Strong337<.001

The Presence of LMW Cyclin E Forms in OC Effusions Is Not Related to Chemotherapy Response

In vitro overexpression of LMW cyclin E forms was associated previously with higher sensitivity to cisplatin.17 Therefore, we analyzed the correlation between the presence of these LMW forms and chemotherapy status/response in our cohort. LMW form expression was comparable in prechemotherapy and postchemotherapy effusions (P = .488), with similar findings for previous treatment with platinum (P = .375) and with paclitaxel (P = .542). A weak association was observed between cyclin E staining extent using IHC and previous chemotherapy (P = .049; data not shown), and comparable results were observed for previous paclitaxel treatment (P = .03; data not shown).

The clinical responses to first-line chemotherapy for 62 patients was as follows: complete response, 31 patients; partial response, 9 patients; stable disease, 1 patient; disease progression, 11 patients; and other (allergic response or bone marrow toxicity), 10 patients. Responses to second-line chemotherapy were known for 57 of 58 patients who received it as follows: complete response, 17 patients; partial response, 1 patient; stable disease, 6 patients; disease progression, 29 patients; and other (allergic response or bone marrow toxicity), 4 patients. No association was identified between the presence of LMW forms of cyclin E using immunoblotting or cyclin E expression using IHC and response to first- or second-line chemotherapy (P > .3 for all; data not shown). In addition, no association was identified with other clinicopathologic parameters (histologic grade, FIGO stage, residual disease volume; data not shown). Separate analyses of prechemotherapy and postchemotherapy specimens or limitation of the analysis to patients who received platinum agents did not alter these results (data not shown). These data suggest that the presence of LMW forms of cyclin E in OC effusions is not a predictor of chemotherapy response.

The Presence of LMW Cyclin E Forms in OC Effusions Is Correlated With Poor Survival

Follow-up for the 63 patients who were diagnosed with OC effusions ranged from 1 month to 81 months (median, 34 months). At the last follow-up, 3 patients were alive with disease, and 60 patients had died of disease. In univariate survival analysis, the presence of LMW cyclin E forms was correlated with shorter OS (P = .021) (Fig. 3A) and shorter PFS (P = .020) (Fig. 3B). Similarly, the presence of higher percentages of cyclin E-positive cells, as determined by using IHC, was correlated with shorter PFS (P = .026) (Fig. 3C).

Figure 3.

Survival analysis. (A) This Kaplan-Meier survival curve illustrates the correlation between the presence of low-molecular weight (LMW) forms of cyclin E in effusions and overall survival (OS) in 63 patients with ovarian carcinoma (OC). Patients who had effusions with no LMW forms of cyclin E (n = 17 patients; solid line) had a mean OS of 45 months compared with 30 months for patients who had effusions with LMW forms (n = 46 patients; dashed line) (P = .021). (B) This Kaplan-Meier survival curve illustrates the correlation between the presence of LMW forms of cyclin E in effusions and progression-free survival (PFS) for 63 patients with OC. Patients who had effusions with no LMW forms of cyclin E (n = 17 patients; solid line) had a mean PFS of 14 months compared with 6 months for patients who had effusions with LMW forms (n = 46 patients; dashed line) (P = .020). The curve begins below 1.0 because some patients had a PFS of 0 months. (C) This Kaplan-Meier survival curve illustrates the correlation between the extent of cyclin E staining in effusions using immunohistochemistry and PFS for 54 patients with OC. Patients who had effusions with focal (≤25%) cyclin E expression (n = 31 patients; solid line) had a mean PFS of 10 months compared with 5 months for patients who had effusions with diffuse (>25%) expression (n = 23 patients; dashed line) (P = .026). The curve begins below 1.0 because some patients had a PFS of 0 months.

Response to first-line chemotherapy was the only clinical parameter that was correlated with survival in this cohort. A complete response to first-line chemotherapy was associated with longer OS (43 months vs 25 months; P = .002) and longer PFS (12 months vs 4 months; P = .001) compared with a partial response, stable disease, disease progression, and treatment failure caused by allergic response or bone marrow toxicity. No correlation was observed between OS or PFS and patient age (≤60 years vs >60 years), histologic grade, FIGO stage, residual disease volume, or response to second-line chemotherapy.

The parameters that were entered into the Cox analysis of OS were LMW form expression, cyclin E staining extent by IHC, and response to first-line chemotherapy. Chemotherapy response was the only independent prognosticator in this analysis (P = .046). Similar findings were observed for PFS (P = .026). These data suggest that cyclin E is a marker of poor OS and PFS in patients with metastatic OC, although it is not an independent factor when chemotherapy response is taken into account.

DISCUSSION

The microenvironment of effusions within the serosal cavities is unique. Stromal myofibroblasts and endothelial cells are replaced by resident mesothelial cells, and leukocyte populations also may differ. The hypoxic conditions and reduced nutrient supply dictate a need for multiple cell survival mechanisms. Whether these mechanisms are universal for all cancers at this anatomic site or are tumor type-specific currently is unknown.

In this study, we observed that the expression of LMW forms of cyclin E largely was limited to OC/PPC, with rare expression in both benign and malignant mesothelial cells. This finding is in agreement with our recent data showing higher cyclin E expression using gene expression arrays, quantitative reverse transcriptase-polymerase chain reaction analysis, and IHC, suggesting that the up-regulation of this molecule plays a major biologic role in OC/PPC, but not in MM or RM. The association between the presence of LMW cyclin E forms using immunoblotting and increased expression using IHC in OC/PPC suggests that both methods are useful for detecting overexpression of this protein.

Bedrosian et al. recently reported that transfected clones of MDAH-2774 ovarian carcinoma cells, which express a cyclin E LMW doublet, had increased S-phase fraction, decreased doubling time, increased colony-forming capacity, and higher sensitivity to cisplatin.19 In contrast, Rosen et al. recently analyzed cyclin E protein expression using IHC in a large series of primary OC and reported a significant association between higher expression of this protein and lack of response to chemotherapy.12 We analyzed the association between cyclin E LMW form expression and chemotherapy response in our OC/PPC cohort. No relation was observed between cyclin E expression by immunoblotting or IHC and response to first- or second-line chemotherapy. Our findings suggest that cyclin E LMW form expression in OC/PPC cells in effusions is not a marker of sensitivity to chemotherapy, a finding that may be explained by the obvious differences between in vitro conditions and clinical specimens. The lack of concordance between our data using IHC and those of Rosen et al. is in agreement with our previous observations of extensive genotypic and phenotypic differences between OC/PPC cells in effusions and their counterparts in primary carcinomas.24–26

Analyses of protein expression of p27kip1, which is a negative regulator of the cyclin E-Cdk2 complex, have demonstrated that the levels of this molecule are correlated with survival in both MM27, 28 and OC,29–31 although the association was observed with both long and short survival in both tumor types. In the largest study of OC, Rosen et al. recently demonstrated an association between higher cytoplasmic levels of this protein and a poor prognosis.29 These data suggest that cell cycle-related molecules may affect the clinical behavior of OC.

In the current study, our univariate survival analysis demonstrated a correlation between cyclin E LMW form expression and poor OS and PFS, and similar results were obtained for PFS using IHC. With the exception of response to first-line chemotherapy, none of the clinical parameters in our cohort demonstrated any correlation with survival. To our knowledge, the prognostic role of LMW cyclin E forms in OC has not been investigated previously. However, our findings are in agreement with the data reported by Keyomarsi et al. in breast carcinoma.18 The current IHC findings are in agreement with 3 reports of cyclin E expression in primary OC,9, 11, 12 but not with 4 other studies, in which this protein showed no association with survival.13–16 Despite our belief that data from primary and metastatic tumors should be analyzed separately, our findings provide support for the hypothesis that, in general, cyclin E expression is related to a more aggressive clinical course in OC.

In the current series, response to chemotherapy was found to be a stronger predictor of both OS and PFS than cyclin E expression, a difference that was reflected in the multivariate survival analysis. This finding underscores the central role of chemotherapy response in determining survival in patients with advanced-stage OC/PPC. The lack of significance between cyclin E expression and disease outcome in the Cox analysis does not concord with the 3 above-described studies9, 11, 12 in which cyclin E expression alone11, 12 or as part of a combined cyclin E-cdk2-p27kip1 score9 independently predicted poor survival. However, chemotherapy response was not entered into the multivariate analysis in those reports, which may explain the discrepancy.

In conclusion, this study presents the first evidence that OC/PPC differs from MM and RM in terms of the expression of LMW forms of cyclin E. This expression does not predict chemotherapy response in patients with OC/PPC, but is associated with a more aggressive clinical course.

Acknowledgements

We gratefully acknowledge the technical help of Ms. Ellen Hellesylt and Ms. Inger-Liv Nordli at the Pathology Clinic, Radiumhospitalet-Rikshospitalet Medical Center.

Ancillary