Cancer of unknown primary: progress in the search for improved and rapid diagnosis leading toward superior patient outcomes

This paper explores the enigma of cancer of unknown primary (CUP) in relation to rapidly improving molecular diagnostic approaches. It is based on the first global collaboration meeting on improving research and clinical outcomes in CUP organized by the CUP Foundation. We review the difficulties of classifying this widely heterogeneous disease and the available diagnostic and pathological evaluative techniques, focusing on molecular profiling. Retrospective studies in CUP patients are shown to provide indirect validation of the accuracy of several platforms of gene expression profiling assays that may identify CUP subsets that respond favorably to active chemotherapy regimens. This review concludes that the recent major improvements in pathologic and molecular diagnostics, coupled with new improved therapies for several specific advanced solid tumors, need to be harmonized with more evidence from clinical-translational trials. All patients with CUP could thus be appropriately managed without the constant uncertainty that has previously severely hampered patient care and optimal outcomes. The longer-term objective is to understand the biology of highly metastatic disease, leading to the development of future global therapeutic programs. Current clinical studies, such as CUP-ONE, will address some of these issues.

HOXB13 May Predict Response to Neoadjuvant Letrozole in Patients with Estrogen Receptor-Positive Breast Cancer

Background. Third generation aromatase inhibitors (AIs) have been shown to be superior to tamoxifen in multiple randomized clinical trials and become the new standard of treatment for postmenopausal women with early or advanced endocrine-sensitive breast cancer. It is thus important to identify biomarkers predicting resistance to AIs so that alternative therapeutic options can be employed. A two gene expression index, HOXB13:IL17BR, and HOXB13 in particular, has been shown to be a negative predictor of clinical benefit from tamoxifen. It is unclear whether HOXB13:IL17BR or HOXB13 is also predictive of AI therapy.Methods. We performed a retrospective analysis of a phase II randomized neoadjuvant study of everolimus+letrozole vs. placebo+letrozole in patients with estrogen receptor-positive breast cancer. Total RNA was extracted from frozen tissue biopsies taken either prior to treatment or after 15 days of treatment. HOXB13:IL13BR and a molecular grade index (MGI) were measured in 107 pre-treatment samples and 111 samples after 15 days of treatment. Antiproliferative response as defined by a ln[Ki767] <1 at day 15 in the samples analyzed was correlated with these gene expression indices.Results. In the placebo+letrozole treatment arm, the antiproliferative response rate was 30% (14/47). Using a pre-specified cutpoint of 0.06, the response rate was 40% for low HOXB13:IL17BR patients, compared to 12% for high HOXB13:IL17BR patients (p=0.05). In the everolimus+letrozole arm, the antiproliferative response rate in the samples analyzed was 55% (33/60), and the response rates for low and high HOXB13:Il17BR patients were 70% and 37% (p=0.02), respectively. When examined separately, high HOXB13 expression was significantly associated with nonresponse to therapy (everolimus+letrozole arm, p=0.02; placebo+letrozle arm, p=0.03), but IL17BR was not (everolimus+letrozole arm, p=0.6; placebo+letrozle arm, p=0.88). As expected, we observed a significant correlation between MGI and Ki67 staining (Spearman r=0.67, p<0.0001).Conclusions. High HOXB13:IL17BR index and high HOXB13 expression may identify a possible subset of patients with estrogen receptor-positive tumor that could be less responsive to aromatase inhibitors such as letrozole. Furthermore, use of MGI may enable rapid assessment of antiproliferative response by fine needle aspiration.

Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 2120.

Unknown primary cancer (UPC): Accuracy of tissue of origin prediction by molecular profiling

11070

Background: Molecular profiling may be useful to identify the primary site and direct therapy for patients (pts) with UPC. Since most UPC pts never have a primary site identified, the accuracy of molecular profiling diagnoses are difficult to verify. We identified a group of UPC pts who had a primary site subsequently identified during their clinical course, and performed a 92-gene real time polymerase chain reaction (RT-PCR) assay (Arch Pathol Lab Med 130:465, 2006) on tissue from the initial diagnostic biopsy. We then compared the RT-PCR diagnosis with the subsequent clinical diagnosis. Methods: 38 of 501 UPC pts (7%) seen between 2000 and 2008 had their primary tumor subsequently identified during life. 24 of the 38 pts had tissue biopsies (excluding FNA cytology) and are the subject of this study. The RT-PCR assay was performed on unstained slides from the formalin-fixed, paraffin-embedded (FFPE) initial diagnostic biopsy, and the assay predictions were compared to the actual primary sites (found later). No clinical or pathologic data (other than sex, biopsy site, and 1 H&E stained slide) were used in the prediction of the primary site. Results: 16 of 24 assays were successful (8 had no tumor or RNA in the material). 11 of 16 predictions of the site of origin (68%) were correct, corresponding to the actual primary sites found 3–58 months (median 8.5 months) after the initial diagnosis of UPC. Primary sites correctly identified included breast 2, ovary/peritoneal 4, NSCLC 1, colorectal 2, gastric 1, melanoma 1. 3 predictions were inaccurate (colorectal, testicular, sarcoma) in patients with gastroesophageal, pancreas and NSCLC, respectively. 2 assays were unclassifiable. Conclusions: RT-PCR performed on FFPE initial diagnostic tissue was accurate in predicting the primary site of origin in 11 of 16 pts with UPC who eventually had their primary site identified clinically. These data provide a direct validation of the reliability of this RT-PCR assay in predicting the primary site in pts with UPC. When used in concert with clinical features and IHC stains, molecular profiling may provide the basis for more successful site-directed therapy for many of these pts. Prospective studies of RT-PCR in UPC are ongoing.

[Table: see text]

Comparison of risk stratification by OncotypeDX versus HOXB13/IL17BR index and molecular grade index in 166 estrogen receptor positive patients from multiple institutions

Abstract #1072

Background
 We have previously reported two complementary prognostic markers in breast cancer for predicting risk of recurrence (Ma et al, Clin Cancer Res, 2008): a molecular grade index (MGI) that recapitulates grade 1 and 3 tumors and stratifies grade 2 tumors, and a gene expression index (HOXB13/IL17BR or H/I) that predicts clinical outcome in estrogen receptor positive (ER+) patients and endocrine benefit (Jerevell et al, Breast Cancer Res Treat, 2007). In combination, MGI and H/I define three risk groups: Low risk (low MGI), intermediate risk (low H/I, high MGI) and high risk (high H/I, high MGI). We conducted a study to compare risk stratification by MGI and H/I within samples for which a previous recurrence score (RS) by Oncotype Dx was reported.
 Materials and Methods
 RNA was extracted from formalin-fixed paraffin-embedded sections from all obtainable samples of patients (pts) from five institutions in which a RS was reported MGH, 77pts; Brigham, 48pts; IU, 19pts; UAB, 13pts; Scripps, 9pts. Real-time RT-PCR assays for MGI and H/I were completed and binary results of low and high for both were determined using pre-defined cutpoints. Risk stratification via MGI + H/I for each sample was compared to previously reported RS's (low, intermediate or high).
 Results
 For the 166 pts, 48% had a low, 45% intermediate and 7% high RS's. Overall, risk stratifications by RS versus MGI + H/I were significantly correlated (p < 0.001) with 81% of patients with low RS also having low risk (low MGI). Risk stratification by H/I + MGI reduced intermediate risk via RS by 2.5-fold (45% versus 18%).
 
 Discussion
 In ER+ pts, MGI + H/I produces a significantly smaller group of patients with intermediate risk by re-stratifying a large number of patients with intermediate RS into low and high risk groups. Risk stratification by H/I + MGI in ER+ pts is by measurement of two discrete parameters: endocrine responsiveness (H/I) and proliferative status (MGI). This may allow for more informed treatment decisions by the treating oncologist. Further studies with clinical outcome are required to establish the clinical significance of these different risk stratification methods.

Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 1072.

Gene profiling validation for cancer classification using known primary carcinoma samples

21130

Background: Cancer of Unknown Primary Site (CUP) accounts for 3–5% of all malignant neoplasms. With current diagnostic procedures, approximately 20% of CUP patients have a primary identified. Since formulation of prognosis and treatment selection depends largely on the primary type, diagnostic uncertainty unfavorably affects CUP management. Gene expression profiling has permitted the development of diagnostic classifiers. A clinically usable methodology needs to differentiate a sufficiently large number of primary types. This study focuses on the estimate of the accuracy of gene expression for classification of metastatic tumors in formalin fixed (FF) FNA and core/surgical biopsies. Methods: CupPrint® is a 1900 gene microarray optimized for FF samples which uses a database of 643 cancer profiles to provide a robust method for the discrimination of 51 tumor (sub) types. Metastasis samples of six known primary carcinoma types were studied (lung, pancreas, colorectal, breast, liver, ovary/primary peritoneal carcinoma). To examine the impact of the metastatic site on diagnostic accuracy different biopsy sites were studied, including lung, brain, lymph nodes, liver, peritoneum. Samples were deemed evaluable that contained an estimated >33% tumor cells. mRNA samples from the metastatic lesions were provided in a blinded fashion to the microarray laboratory. Results: For 39 evaluable samples studied so far, the CupPrint® prediction was compared with the pathologic diagnosis. Of the 39, in 34 the primary site of origin was correctly predicted (87% accuracy). In two of the incorrectly predicted samples the site of biopsy (brain and liver respectively) was reported as the primary site. Correlations of molecular profiles to clinical parameters are being explored. Conclusions: CupPrint® appears to be sufficiently accurate to be of potential clinical utility in the diagnosis of CUP.

No significant financial relationships to disclose.

High expression of stanniocalcin in differentiated brain neurons

Stanniocalcin (STC) is a glycoprotein hormone first found in fish, in which it regulates calcium homeostasis and protects against hypercalcemia. Human and mouse stc cDNA were recently cloned. We found a dramatically upregulated expression of STC during induced neural differentiation in a human neural crest-derived cell line, Paju. Immunohistochemical staining of sections from human and adult mouse brain revealed abundant presence of STC in the neurons with no activity in the glial cells. STC expression was not seen in immature brain neurons of fetal or newborn mice. Given that STC has been found to regulate calcium/phosphate metabolism in some mammalian epithelia, we suggest that STC may act as a regulator of calcium homeostasis in terminally differentiated brain neurons.

In vivo characterization of the proteasome regulator PA28

A proteasome regulator, termed PA28, has been shown to modulate peptidase activities of the proteasomes in vitro. Two different but homologous PA28 molecules, designated as PA28alpha and PA28beta, have been cloned. Both alpha and beta polypeptides of PA28 are found in PA28 complexes isolated from cells, indicating that both are constituents of functional PA28 complexes. Using antisera specific to PA28alpha, PA28beta, and epitope-tagged PA28 molecules, we show that expression of PA28alpha and PA28beta is coordinately induced by various cytokines in different cell lines and that PA28 subunits and proteasomes have almost identical half-lives. In addition, we show that PA28 complexes are associated with 20 S but not 26 S proteasomes in vivo. Moreover, we demonstrate that PA28 complex is a heterohexamer composed of both alpha and beta subunits with a stoichiometry of alpha3beta3 in an alternating order.