Intratumoral lymphocyte density in serous ovarian carcinoma is superior to ERCC1 expression for predicting response to platinum-based therapy
- First Online:
- Cite this article as:
- Bösmüller, H., Haitchi-Petnehazy, S., Webersinke, G. et al. Virchows Arch (2011) 459: 183. doi:10.1007/s00428-011-1110-1
- 207 Views
Intratumoral immune cells and ERCC1 expression are likely to play a role in the response of ovarian carcinoma to chemotherapy, but their impact on therapy outcome is still unclear. Therefore, 41 cases of optimally resected high grade serous ovarian carcinomas were examined retrospectively for stromal and intraepithelial lymphocyte populations and ERCC1 status in relation to response to platinum-based therapy. Based on RECIST criteria, 27 patients were classified as responsive and 14 as therapy resistant, respectively. Using immunohistochemistry for CD3, CD8, CD4, TIA1, MUM1 and FOX P3 on representative tumor sections, we quantitatively evaluated the intratumoral density of lymphocyte subpopulations. In addition, ERCC1 protein and mRNA expression were determined by immunohistochemistry using the Steffensen score and quantitative RT-PCR, respectively. Furthermore, ERCC1 SNP’s C8092A and codon 118 were analysed. Response to chemotherapy was significantly associated with higher numbers of stromal CD3+ (mean 21.33 lymphocytes/HPF versus 8.21 lymphocytes/HPF, p = 0.002) and CD8+ lymphocytes (mean 9.22 lymphocytes/HPF versus 4.57 lymphocytes/HPF, p = 0.013). Counts of intraepithelial CD3+ and CD8+ lymphocytes, stromal and intraepithelial FOXP3+ and TIA1+ cells, CD4+ lymphocytes, and MUM1+ plasma cells did not reach statistical significance. Neither ERCC1 protein expression (p = 0.232) nor SNPs codon 118 and C8092A of the ERCC1 gene (p = 0.269 and p = 0.543) showed an association with therapy response. The same was true for ERCC1 mRNA levels (p = 0.896), probably due to intratumoral lymphocyte contamination. In conclusion, the density of CD3+ and CD8+ T-cells in tumor stroma proved to be a significant predictor for response to platinum-based therapy, whereas examination of ERCC1 failed to identify therapy-responsive patients.
KeywordsSerous ovarian adenocarcinomaERCC1 immunohistochemistryERCC1 mRNA expressionERCC1 SNPIntraepithelial lymphocytesStromal lymphocytes
Excision repair cross-complementing rodent repair deficiency complementation group 1
High power field
Non-small cell lung cancer
Single nucleotide polymorphism
Serous ovarian cancer
TATA box-binding protein
Platinum-based chemotherapy is the standard of care for ovarian carcinoma of tumor stages FIGO Ic to FIGO IV. Major side effects such as myelotoxicity are observed in about 20% of patients, and approximately 20% of serous ovarian carcinomas show primary resistance to Platinum-containing regimes. Therefore, there is particular interest in finding markers predictive of chemotherapy response. However, currently proposed biomarkers still lack predictive power in order to be used for pretherapeutic decision making. In recent years, both intra- and peritumoral immune cell infiltration and expression of proteins involved in the cellular response to chemotherapeutic agents have come into focus as potential predictors of chemotherapy response.
Among the specific markers investigated for their predictive value, the significance of ERCC1 (Excision repair cross-complementing rodent repair deficiency group 1) in ovarian carcinoma has received significant attention, analogous to non-small cell lung cancer [1–9]. Strongly positive ERCC1 immunohistochemistry (IHC) and a high level of ERCC1 mRNA are regarded as an indication for efficient repair of tumor DNA damaged by cytostatic agents, resulting in therapy resistance [10, 11]. A weak expression on mRNA or protein level is thought to indicate a higher likelihood of response to platinum-based chemotherapy [12–14]. In addition, the variant form of the single nucleotide polymorphism (SNP) of the ERCC1 gene in codon 118 was found to be associated with a worse outcome in ovarian carcinoma [15, 16].
In recent years and with refined techniques, the density and composition of the intratumoral lymphocyte population have been found to be of prognostic relevance in a wide variety of neoplasms. In ovarian carcinoma, most studies so far have investigated their impact on overall and disease-free survival, mainly pointing out the importance of intraepithelial CD8+ lymphocytes [17–24]. The correlation between tumor-associated lymphocytes and chemotherapy response has been the subject of only a very limited number of studies [25, 26].
Given the in part contradictory results concerning the role of ERCC1 and the scarcity of data concerning the impact of intratumoral lymphocytes on therapy response, the present study of 41 cases of high grade serous ovarian carcinomas uniformly treated by platinum-based chemotherapy aims at elucidating the potential of ERCC1 immunohistochemistry, mRNA expression and the codon 118 polymorphism on one hand, and intratumoral lymphocyte content on the other hand as predictive markers for ovarian carcinoma chemotherapy.
Material and methods
Forty-one women aged 40 to 79 years with high grade serous ovarian carcinoma (SOC) were included in the present study. All patients had undergone optimal surgical resection (no visible residual tumor) under standardized conditions, including hysterectomy, bilateral adnexectomy, omentectomy, debulking procedures, peritoneal biopsies, and peritoneal lavage for cytologic examination. Thirty-two patients had undergone pelvine as well as paraaortal lymphadenectomy. Nine patients without evidence for retroperitoneal involvement had not received lymphadenectomy based on preoperative risk evaluation. Patients with lymph node metastasis, extraperitoneal metastasis or visible residual tumor were not included in the study. Pathologically, 36 patients were diagnosed with SOC pT 3c, two with pT 2c and five with pT 1c exhibiting tumor masses in both ovaries and positive peritoneal fluid, but without fallopian tube infiltration.
Standardized Carboplatin-Taxane chemotherapy was administered every three weeks for six cycles in all patients. Platinum resistance was defined as a relapse during or within 6 months after the end of six cycles of chemotherapy, proven by CT and/or PET-CT. Cancer Antigen 12-5 (CA 12.5) was used as an additional parameter of recurrence. Platinum sensitivity was defined by a progression-free interval for at least 12 months after the end of six cycles of chemotherapy. No second look surgery or tumor sampling in cases of recurrence was done. By means of these criteria, 14 tumors were classified as platinum-resistant and 27 as platinum-sensitive. Patients with partial platinum resistance, defined by recurrences within 6 up to 12 months after the end of six cycles of chemotherapy were not included in the study.
Eleven patients aged 36 to 54 years with serous borderline tumors, including two of micropapillary type, were recruited as a control group. Eight patients had a single lesion, three women showed involvement of both ovaries, including the two micropapillary tumors. The surgery performed was identical to the procedure described above, but patients did neither undergo lymphadenectomy nor chemotherapy.
Immunohistochemisty and molecular analyses
All cases were reviewed by an experienced gynecopathologist (A.S.), and appropriate formalin-fixed and paraffin-embedded tumor blocks were selected. Matched, but tumor-free fallopian tubes or uninvolved ovarian stroma served as control tissues for IHC, mRNA quantitation was calibrated with selected fallopian tube tissue.
The following primary antibodies were used for immunohistochemistry: ERCC1 (1:50, monoclonal 8F1, DCS Innovative Diagnostics, Hamburg, Germany), CD3/pan-T-cell (1:100, rat monoclonal SP7, DCS), CD4/T-helper cells (1:100, rat monoclonal SP35; Cytomed Baden-Baden, Germany), CD8/cytotoxic T-cells (1:200, mouse monoclonal C8/144B, DAKO, Glostrup, Denmark), MUM1/plasma cells (1:200 MUM1p, DAKO), TIA1/cytotoxic T-cells (1:200, mouse monoclonal GMP12, Cytomed), FOX-P3/regulatory T-cells ( 1:150, mouse monoclonal, mA22509, ABCAM, MA USA). The tissue sections were pre-treated with EDTA-buffer solution (pH 8.6) at 95° for 36 min (ERCC1, CD3, CD8), 64 min (MUM1, TIA-1) and 76 min (CD4). The antibodies were incubated at 37° for 32 min (ERCC1, CD3, CD8, MUM1, TIA-1) and 42° for 40 min, immunohistochemical staining was based on the avidin–biotin method. One representative paraffin section was used per case and antibody.
For the semiquantitative evaluation of ERCC1 staining, we used the Steffensen score . Based on the staining intensity of the control tissue, the intensity of ERCC1 expression in the tumor tissue was graded from 0 to 3, with 0 as negative staining, grade 1 as weaker, grade 2 as equal and grade 3 stronger than control tissue. The score was obtained by multiplying this grade with a factor determined by the percentage of positive tumor cells (0–10% / 0.1; 10–50% / 0.5; 50–100% / 1). Scores > 1 were considered positive.
The quantification of lymphocytes was carried by counting the average number of cells per high power field (HPF = 400×) reviewing a total of ten HPF, additionally we documented the maximum range for each subset. To control for observation bias, we statistically evaluated both average and maximum counts for each category. The lymphocyte subsets of the tumor stroma were quantified at the invasive tumor front in the stains for CD3, CD8, CD4, TIA 1, MUM1 and FOXP3. In a second step, intraepithelial lymphocytes were counted.
mRNA and DNA for ERCC1 quantitation and SNP analysis, respectively, were extracted after macrodissection of hematoxylin–eosin stained slides. The extraction of total RNA from paraffin-embedded tumor tissue was performed as described by Specht et al. . Five hundred nanogram of RNA was transcribed with Superscript II (Invitrogen, Carlsbad, CA) according to the manufacturer. Quantitative PCR was performed on a LightCycler 480 (Roche Diagnostics, Mannheim, Germany) using previously published intron-spanning primers and probes for the quantification of ERCC1 and TATA box-binding protein (TBP), which served as a reference gene . To calculate the real-time PCR results we used the ΔΔCt-method, normalising ERCC1 against TBP in the first step (ΔCt). In the second step, these data were calibrated to the ratio of the mean value of the control tissue (ΔΔCt), defining mRNA of fallopian tubes as 100%. A relative expression of 140% and higher was considered as high expression in the tumor.
Genomic DNA of tumor tissue and control tissue was isolated by the use of Maxwell® 16 FFPE Custom Kit (Promega, WI, USA) after denaturation with proteinase K. The SNPs for Codon 118 (rs11615) and C8092A (rs3212986) were analysed with TaqMan SNP genotyping assays (Applied Biosystems, CA, USA), according to the recommendations of the manufacturer . Allelic discrimination was performed on a 7900HT Fast real-time PCR System using SDS 2.2.2 Software Version (Applied Biosystems).
Statistical analysis was performed using SSPS statistical software (Microsoft, CA, USA). Univariate analyses were carried out using the Chi-Squared statistic. P values less than 0.05 were considered significant.
ERCC1 protein and mRNA expression and SNP analysis
Correlation of chemotherapy response with ERCC1 protein and mRNA expression and ERCC1 SNPs at Codon 118 and C8092A
ERCC1 mRNA level
ERCC1 SNP codon 118
ERCC1 SNP C8092A
For the evaluation of ERCC1 mRNA, the ERCC1/TBP mRNA ratio of normal fallopian tube tissue was set at 100%. For tumor tissue, a ratio 140% or more was considered to represent high expression. In summary, 24 patients (59%) showed low to normal levels, 17 patients (41%) high ERCC1 mRNA levels. There was no correlation between ERCC1 mRNA expression and chemotherapy response, as 16 (60%) of 27 chemosensitive and six (42%) of 14 resistant tumors had low to normal mRNA levels (range 62–139%), whereas high mRNA levels (range 140–192%) were encountered in 11 (40%) responsive and eight (58%) unresponsive carcinomas (p = 0.896). There was a poor correlation between ERCC1 protein expression by immunohistochemistry and ERCC1 mRNA levels.
The 11 serous borderline tumors exhibited ERCC1 expression levels comparable to the control tissue. None of the cases showed high ERCC1 expression.
The analysis of ERCC1 single nucleotide polymorphisms (SNP) of codon 118 (Asn118Asn) and C8092A showed allele distributions similar to published data, but failed to show a significant association with therapy response. Details are shown in Table 1.
Distribution of lymphocyte subsets
Lymphocyte subset counts in responders versus non-responders
Responder (27) mean (range)
Non-responder (14) mean (range)
p value average (maximum)
Correlation of chemotherapy response with density of tumour infiltrating lymphocyte subsets
PPV = 0.91
NPV = 0.61
PPV = 0.76
NPV = 0.45
PPV = 0.82
NPV = 0.55
PPV = 0.64
NPV = 0.33
The serous borderline tumors used as controls had low lymphocyte counts in general. Of note, we did not find any differences for stromal and intraepithelial lymphocytes in the stains for CD3 and CD8. (0–5 cells/HPF). CD4, MUM1 and TIA1A-positive cells were sparse, FOXP3+ lymphocytes were virtually absent.
In order to avoid ineffective and potentially toxic chemotherapy in women with advanced ovarian carcinoma, there is increasing interest in reliable predictive markers for platinum responsiveness. The intention of the present study was to compare the predictive values of ERCC1 expression both on mRNA and protein levels, the influence of ERCC1 SNPs and the density of intratumoral lymphocytes in a very homogenously treated patient collective of serous ovarian carcinoma. The study demonstrates that a high density of stromal T-cells is predictive of response to platinum-based chemotherapy, whereas ERCC1 expression and ERCC1 allelotypes failed to identify patients who would benefit from this therapy.
ERCC1 is a key enzyme in the nucleotide excision repair pathway of platinum-damaged cells [11, 13, 14] and therefore has been the subject of many studies using different approaches, including immunohistochemistry, mRNA quantification and SNP analyses from peripheral blood cells or tissue. The group studied most extensively for the impact of ERCC1 so far were patients with non-small cell lung cancer (NSCLC), though the studies differed substantially with regard to selection criteria such as adjuvant therapy versus palliative treatment or progressive disease [1–3]. The different methods used, mostly IHC and RT-PCR, and the heterogeneous patient collectives do not always allow comparison of results. The same heterogeneity also applies to ERCC1 studies of ovarian carcinoma. An association between low ERCC1 expression both by RT-PCR, as well as by immunohistochemistry and improved overall survival (OS) has been described for patients with progressive disease [11, 14, 28] whereas Steffensen found a better response to platinum-based chemotherapy for patients negative by IHC, but no influence on OS [8, 15]. Further investigations concentrated on mRNA (RT-PCR) and SNP analyses, pointing out a correlation between platinum sensitivity on one side and low mRNA levels and T/T and C/T alleles in codon 118 on the other side, respectively. Similarly, a relation was found between poor response and positive ERCC1 IHC and C/C variant in codon 118 [5, 7, 15, 16]. Krivak identified the polymorphism C8092A of ERCC1 as an independent marker for progression-free survival (PFS) , but failed to find clinical relevance for OS [6, 29]. Our results reflect this lack of a clear-cut predictive value of ERCC1 immunohistochemistry, mRNA expression levels and SNPs codon 118 and C8092A. The main limitation of immunohistochemistry lies in the poor reproducibility and the difficulties in quantification of immunostaining results despite the use of the standard antibody clone 8F1. Although absent or reduced staining seems to be an indication for therapy response also in our study, immunopositivity is nevertheless found in a significant fraction of responders, making ERCC1 staining unsuited for therapy planning. It needs to be demonstrated whether automated image analysis of ERCC1 immunostaining will produce better results [8, 30].
Results for the SNP in codon 118 showed a trend towards discrimination of chemoresistant and chemosensitive tumors, but failed to reach statistical significance in our small collective. Larger studies are needed to decide whether SNP in codon 118 may form part of a marker panel for predicting response to platinum-based therapy.
The lack of a predictive value of ERCC1 mRNA levels may seem surprising at first. Most studies on ERCC1 mRNA quantification relied on RT-PCR of frozen tissue, and it could be argued that technical factors play a role [10, 11]. However, we have previously shown the high reliability of mRNA quantification on formalin-fixed, paraffin-embedded and manually dissected tissue, using the protocol initially published by Specht et al, and employing intron-spanning primers to avoid DNA contamination [27, 31]. When comparing our ERCC1 expression data with the lymphocyte counts, we observed a notable correlation between stromal CD3 counts and mRNA levels in the chemosensitive subset (p = 0.009), indicating that a high number of contaminating lymphocytes correlates with higher ERCC1 levels and thus possibly obscures any correlation between low ERCC1 expression in tumor cells and chemoresponse. All 13 tumors with less than 18 lymphocytes/HPF had only low ERCC1 mRNA levels, whereas seven tumors with more than 26 lymphocytes/HPF revealed only high levels.
The dataset for the resistant subset was not as consistent, but four tumors with few lymphocytes showed markedly elevated levels, indicating that increased ERCC1 mRNA levels indeed derived from the neoplastic population. In conclusion, these results throw doubt on the practical value of ERCC1 mRNA quantification in ovarian carcinoma tissue without correction for lymphocyte density.
Tumor-associated lymphocytes are a common feature of serous ovarian cancer and have been investigated in many studies concerning their potential prognostic or predictive impact. Most published data have shown an improved overall survival for cases with increased intratumoral/intraepithelial CD8+ lymphocytes, [17–19, 21, 24], a finding which is also supported by the results of our study. The phenotype and function of this T-cell subset and its interactions with other lymphocyte subpopulations still remain unclear, but the microenvironment is thought to play a major role . CD4+ intraepithelial lymphocytes and regulatory T-cells (FOXP3+) are supposed to counterbalance the beneficial effect of a high CD8+ T-cell density. Only high CD8/CD4 and CD8/Treg ratios were associated with better prognosis .
Reports about the correlation between intratumoral CD3+ lymphocytes and overall survival are inconsistent. Whereas Sato and Clarke failed to find a correlation between OS and intraepithelial lymphocytes [18, 20], Milne and Tomsova demonstrated intraepithelial lymphocytes to be relevant for survival, when stromal and intraepithelial lymphocytes were evaluated separately [33, 34]. In a recent study Al-Attar et al. counted intraepithelial and stromal lymphocytes in ovarian cancer and calculated a ratio, emphasising that high numbers of intraepithelial CD3+ lymphocytes correlated with longer survival. High numbers of stromal CD3+ lymphocytes and a low intraepithelial/stromal ratio predicted worse prognosis .
These survival data have to be discerned from data correlating lymphocyte counts with response to chemotherapy, which has been investigated only in a handful of studies. Stumpf et al. pointed out the predictive value of intraepithelial CD8+ lymphocytes, whereas Raspollini et al. observed a correlation between response and tumor infiltrating CD3+ lymphocytes [25, 26].
A recurrent problem in comparing studies on lymphocyte densities in ovarian cancer is the lack of a stringent terminology. Not all investigators make a clear distinction between stromal and intraepithelial lymphocyte populations. In particular, the term “intraepithelial” is misguiding in some publications, since it is used as a generic term for tumor infiltrating lymphocytes.
In our study, intratumoral lymphocytes were differentiated based on their location and phenotype. Whereas intraepithelial CD3+ and CD8+ lymphocytes (IL) were seen throughout the tumor, stromal lymphocytes (SL) were increased at the invasion front. These distribution patterns could not be identified for TIA1 and FOXP3-positive cells, probably due to their low numbers. CD4+ lymphocytes and plasma cells were strictly restricted to the tumor stroma.
The number, function and relevance of FOXP3+ T-cells in ovarian cancer are still a matter of controversy. The presence of FOXP3+ T-cells was initially thought to be a feature of advanced serous ovarian cancer and unfavourable prognosis, but recent studies found the reverse [36, 37]. In our hands, the low numbers of FOXP3+ cells probably played a role in the lack of statistical significance with the approach used in this study.
In conclusion, only the density of stromal CD3+ and CD8+ lymphocytes proved to be significantly associated with response to platinum-based chemotherapy in our study. Quantification of ERCC1 mRNA is probably influenced by the amount of intratumoral lymphocytes and therefore does not faithfully reflect the ERCC1 levels in tumor cells, resulting in poor response prediction. Larger prospective studies will be required to confirm these findings.
Conflict of interests
The authors declare that they have no competing interests.