- © 2008 by American Society of Clinical Oncology
Molecular Profiling of Carcinoma of Unknown Primary and Correlation With Clinical Evaluation
- Gauri R. Varadhachary,
- Dmitri Talantov,
- Martin N. Raber,
- Christina Meng,
- Kenneth R. Hess,
- Tim Jatkoe,
- Renato Lenzi,
- David R. Spigel,
- Yixin Wang,
- F. Anthony Greco,
- James L. Abbruzzese and
- John D. Hainsworth
- From The University of Texas M. D. Anderson Cancer Center, Department of Gastrointestinal Medical Oncology, Houston TX; Sarah Cannon Research Institute and Tennessee Oncology, Nashville TN; and Veridex, La Jolla, CA
- Corresponding author: John D. Hainsworth, MD, Sarah Cannon Research Institute, 250 25th Ave North, Suite 110, Nashville, TN 37203; e-mail: jhainsworth{at}tnonc.com
Abstract
Purpose To evaluate the feasibility of a 10-gene reverse transcriptase polymerase chain reaction assay to identify the tissue of origin in patients with carcinoma of unknown primary (CUP) site.
Patients and Methods Diagnostic biopsy formalin-fixed, paraffin-embedded (FFPE) specimens from 120 patients with CUP were collected retrospectively from Sarah Cannon Research Institute, Nashville, TN, and prospectively from The University of Texas M. D. Anderson Cancer Center, Houston, TX. Tissue of origin assignments by the assay were correlated with clinical and pathologic features and with response to therapy.
Results The assay was successfully performed in 104 patients (87%), and a tissue of origin was assigned in 63 patients (61%). In the remaining 41 patients (39%), the molecular profiles were not specific for the six tumor types detectable by this assay. The tissues of origin most commonly identified were lung, pancreas, and colon; most of these patients had clinical and pathologic features consistent with these diagnoses. Patients with lung and pancreas profiles had poor response to treatment. Patients with colon cancer profiles had better response to colon cancer–specific therapies than they did to empiric CUP therapy with taxane/platinum regimens. Patients with ovarian cancer profiles were atypical, with widespread visceral metastases and a paucity of overt peritoneal involvement.
Conclusion This gene expression profiling assay was feasible using FFPE biopsy specimens and identified a putative tissue of origin in 61% of patients with CUP. In most patients, the assigned tissue of origin was compatible with clinicopathologic features and response to treatment. Prospective studies in which assay results are used to direct therapy are indicated.
INTRODUCTION
Carcinoma of unknown primary (CUP) accounts for approximately 2% to 5% of all cancer diagnoses. Empiric treatment decisions are necessary in the absence of a specific diagnosis, and clinicians run the risk of inadvertently selecting a suboptimal treatment for an individual patient. Identification of the tissue of origin would facilitate the choice of therapy for patients with CUP; therefore, improved diagnostic methods are urgently needed.
The recent development of molecular technology that allows gene expression profiling of tumors provides an opportunity for improved diagnosis of patients with CUP. Specific gene expression profiles based on the tissue of origin have been identified for many tumor types.1-4 Quantitative reverse transcriptase polymerase chain reaction (RT-PCR) is a technique that allows gene expression profiling of formalin-fixed, paraffin-embedded (FFPE) tissue specimens and is therefore a potential diagnostic tool of broad applicability.5-7
A molecular assay (CUP assay) was developed (Veridex, La Jolla, CA) that evaluates the expression of 10 tissue type–specific gene markers using quantitative RT-PCR technology and is designed to detect tumors originating from six specific sites: lung, breast, colon, ovary, pancreas, and prostate.8 The accuracy of this assay was validated in a set of 260 FFPE tumor specimens obtained from metastatic sites in patients with cancers of known primary site. The assay correctly identified the primary site in 204 (78%) of 260 test examples. An incorrect diagnosis was made in 46 samples (18%), whereas no diagnosis was established in 10 samples.8
In this study, we performed the CUP assay on FFPE biopsy specimens from a large number of patients with CUP. The specific aims of this study were to assess the feasibility of the CUP assay in yielding diagnoses from FFPE biopsy specimens and to correlate assay results with clinical characteristics, pathologic features, and response to treatment in patients with CUP.
PATIENTS AND METHODS
Patient Selection
Two cohorts of CUP patients were included in this study. A retrospective patient cohort included tissue specimens from 78 patients with CUP who were diagnosed and treated at the Sarah Cannon Cancer Center, Nashville, TN, or the Minnie Pearl Cancer Research Network between April 1998 and October 2005. Seventy-two of the 78 patients included in this cohort were treated in one of five sequential clinical trials of empiric chemotherapy. Four of these trials evaluated various taxane and platinum combinations;9-12 the fifth and ongoing trial is a randomized comparison of paclitaxel/carboplatin/etoposide versus gemcitabine/irinotecan.
The prospective patient cohort included tissue specimens from 42 consenting patients with CUP who were referred to The University of Texas M. D. Anderson Cancer Center (Houston, TX) for comprehensive diagnostic work-up and treatment between March 2004 and June 2007. Serial radiographic, pathologic, and treatment data were available for this cohort. Treatment was tailored to the individual patient, based on clinical features, radiographic data, and immunohistochemistry (IHC) as per the institution's standard approach to CUP patients.
The definition of CUP was consistent in both cohorts and included patients with no detectable primary site after a diagnostic evaluation that included complete history; physical examination; blood counts; chemistry profile; computed tomography scans of the chest, abdomen, and pelvis; and appropriate targeted evaluation of any specific signs or symptoms. Patients with only fine-needle aspiration biopsy specimens were excluded. Study of both patient cohorts was approved by local institutional review boards.
Sample Collection and Processing
The FFPE biopsy tissue blocks were collected and sent to Veridex, where the CUP assay was performed. All samples were tested without the knowledge of either the clinical characteristics or the subsequent response to treatment, except for the sex of the patient and the site of biopsy.
Assay Procedure
The assay consists of 10 gene markers specific for lung (HUMPB, TTF-1, DSG3), pancreas (PSCA, F5), colon (CDH17), breast (MGB, PDEF), prostate (prostate-specific antigen), and ovarian (WT-1) primary tissues.8 These markers were selected from a large number of candidates based on RT-PCR results from tumors of known origin. Interpretation of the assay included an algorithm for excluding the gene expression patterns of normal cells included in the biopsy from the site of metastatic involvement.8
The assay was performed on RNA isolated from six to nine 10-μm FFPE tissue sections. Before selecting the tissue sections for assay, a hematoxylin/eosin stain was performed on an adjacent section to ensure the specimen contained more than 10% tumor cells. Assay results were reported as a series of probabilities for each of the six primary cancer types included in the assay and as “other” if the molecular profile did not fit any of six tissue types. Tissue of origin was assigned based on the highest probability reported.
Analysis of Clinical and Pathologic Information
Complete clinical and pathologic information was collected, including patient demographics, metastatic pattern of spread, results of clinical and laboratory tests, imaging data, results of pathologic evaluation (standard hematoxylin/eosin and IHC tests), chemotherapy administered, response to therapy, and overall survival. Overall survival was calculated as the time from the beginning of treatment until the date of death.
RESULTS
Assay Data and Patient Characteristics
The CUP assay was successfully performed on specimens from 104 (87%) of 120 patients. In 16 specimens, the assay could not be performed because of a low RNA yield (n = 3) or poor RNA preservation (n = 13).
The characteristics of the 104 patients in whom the assay was successfully completed are included in Table 1. The retrospective and prospective cohorts shared typical CUP characteristics: median age (56 years and 57 years, respectively), predominant histopathologic diagnosis of either adenocarcinoma or poorly differentiated adenocarcinoma (69% and 70%, respectively), and two or more metastatic sites (80% and 70%, respectively). The response rates to first-line treatment in the retrospective and prospective cohorts were 28% and 34%, respectively. The median survival for all patients in the retrospective cohort was 8.5 months; 2-year survival was 19%.
Patient Characteristics (N = 104)
Tissue of Origin Prediction Results
Assay results are listed in Table 2. A specific tissue of origin assignment was made in 63 patients with CUP (61%), with the remaining 41 patients (39%) classified as “other.” The most common tissues of origin identified were lung (22%) and colon (33%) in the retrospective and prospective populations, respectively. Other tissue of origin profiles identified in both cohorts included pancreas and ovarian (Table 2). Six patients had biopsies from more than one metastatic site, and the assay results are summarized in Figure 1. In four patients, the assay results were identical for all biopsy sites, confirming internal consistency.
Tissue of origin diagnoses in six patients with multiple specimens assayed. Assay results from all biopsy sites were identical in four patients (A through D). In patient E, eight of nine biopsies were diagnosed as “other”; one of nine was diagnosed as ovarian. In patient F, three biopsies were ovarian and two were “other.” In both patients, ovarian was the second highest probability in specimens diagnosed as “other.” LN, lymph node.
Tissue of Origin Prediction Results for Retrospective and Prospective Cohorts
Correlation of Clinical and Pathologic Features With Assay CUP Results
The validity of the CUP assay results cannot be directly ascertained because the primary site in almost every case remained unidentified clinically. However, by correlating the putative tissue of origin assignment with clinical and pathologic features, as well as with response to empiric chemotherapy, inferences were made regarding the potential clinical value of this diagnostic procedure. The best correlations were obtained in patients with a molecular profile consistent with colon, non–small-cell lung cancer (NSCLC), or pancreatic cancer. Results with ovarian cancer were more mixed, perhaps because of the limited number of patients with this profile.
Correlation of colon cancer RT-PCR profile with clinical and pathologic data.
Clinical features, IHC results, and response to treatment data for the 23 patients with colon cancer profiles are listed in Table 3. Nineteen of 23 patients had liver or peritoneal metastases, consistent with the metastatic pattern of colon cancer. Seventeen of 23 patients underwent colonoscopy with negative results. IHC test results were available for 17 of 23 patients and suggested colon as the tissue of origin in 11 patients (CK20+ and/or CDX2+, CK7- or focal positive). In one patient, a primary site (appendix) was diagnosed 2 years after his initial CUP presentation.
Clinicopathologic Characteristics of Patients With Gene Profile Diagnosis of Colon Cancer (n = 23)
Only two of the 12 patients with a colon cancer RT-PCR profile in the retrospective cohort had partial responses to treatment; most received empiric paclitaxel and carboplatin–based treatment, and none received specific colon cancer regimens. In contrast, several patients in the prospective cohort were treated with optimal colon cancer–specific therapies as their first-line or second-line therapy and showed sustained partial responses. Four patients received fluorouracil, leucovorin, and oxaliplatin therapy as their first-line treatment and had partial responses; two of these patients received irinotecan and cetuximab as second-line therapy and also had a partial response to this regimen. Four other patients from the prospective group were treated with paclitaxel and carboplatin (n = 3) or gemcitabine and cisplatin (n = 1), with only one partial response. Two of these four patients showed partial response to fluorouracil, leucovorin, and oxaliplatin as second-line therapy.
Correlation of NSCLC RT-PCR profile with clinical and pathologic data.
Nineteen patients (retrospective, n = 15; prospective, n = 4) had an NSCLC RT-PCR profile, including eight patients with a squamous cell lung cancer profile. Clinical features of these patients were generally consistent with the diagnosis of NSCLC. Thirteen of 19 patients had predominantly intrathoracic metastases, whereas six patients had other sites of involvement (intra-abdominal, n = 4; bone, n = 1; inguinal node, n = 1). IHC data were available for 14 patients and suggested lung as the tissue of origin in 10 patients (TTF1+, two patients; CK7+/CK20−, nine patients). Most patients in the retrospective cohort received treatment with a paclitaxel and carboplatin–based regimen (13 of these 15 patients received paclitaxel, carboplatin, and etoposide). Five (33%) of these patients responded to treatment, and the 9-month median survival was typical of patients with advanced NSCLC. Two patients from the prospective cohort responded to paclitaxel and carboplatin, whereas one patient had no response to oxaliplatin and capecitabine. One patient in the prospective cohort with an NSCLC RT-PCR profile from his bone biopsy was found at the time of repeat imaging to have a primary pancreatic cancer with liver and bone metastases.
Correlation of pancreatic cancer RT-PCR profile with clinical and pathologic data.
A total of 13 (retrospective, n = 11; prospective, n = 2) patients had a pancreatic cancer RT-PCR profile, and most of these had compatible clinical and pathologic features. Twelve of the 13 patients had metastatic disease within the abdomen, most often in the liver (seven patients), mesenteric or retroperitoneal lymph nodes (four patients), spleen (two patients), or peritoneal surface (three patients). IHC profile was compatible with pancreas origin (CK7+ or CK7+/CK20+) in six patients and inconclusive in two patients. Only two patients in the retrospective cohort had partial responses to treatment with gemcitabine and irinotecan. The other nine patients had either stable disease or progressive disease after treatment with gemcitabine and irinotecan (two patients) or a paclitaxel and carboplatin–based regimen (seven patients). The median survival was 7 months.
Correlation of ovarian cancer RT-PCR profile with clinical and pathologic data.
Seven patients had an ovarian cancer RT-PCR profile, including four patients in the retrospective and three patients in the prospective cohort (Table 4). Only one of these seven patients had predominant peritoneal involvement, and five of seven patients had one or more metastatic sites outside the abdomen. IHC profiles were available for five of these seven patients, including CK7+ (n = 6), estrogen receptor (ER) positive (n = 2), and WT-1+ (n = 1). Two of four patients had partial responses to first-line treatment with paclitaxel and carboplatin–based regimens. One patient with an isolated pleural effusion and ER-positive IHC staining was treated as having metastatic breast cancer and responded to doxorubicin and cyclophosphamide–based therapy.
Clinicopathologic Characteristics of Patients With Gene Profile Diagnosis of Ovarian Cancer (n = 7)
Forty-one (39%) of 104 patients had assay results that did not identify any of the six cancers included; these patients were classified as “other.” Clinical features in this group were diverse; common metastatic sites included liver, lung, and bone.
Correlation of CUP assay results with IHC results.
Because specific IHC evaluation was not mandated, there was variation in the studies performed. However, 37 patients had evaluations that included, at minimum, staining for CK7, CK20, and TTF-1. Additional studies commonly performed in these 37 patients were CDX2 (41%), carcinoembryonic antigen (41%), ER/progesterone receptor (38%), neuroendocrine stains (38%), AE1/AE3 (30%), and prostate-specific antigen (22%). On the basis of IHC results, specific diagnoses were made in four (11%) of these 37 patients (colorectal, n = 3; lung, n = 1); in the other 33 patients, the IHC results were not specific enough to suggest only one diagnosis. The CUP assay gave diagnoses in 23 (62%) of these 37 patients. All four of the patients given specific diagnoses by IHC staining had the same tissue of origin identified by the CUP assay.
DISCUSSION
DNA microarray and quantitative RT-PCR platforms have been used to develop accurate multiclass classifiers to help predict tissue of origin.3,8,13-15 In clinical medicine, RT-PCR assays are of greater applicability, because they can be performed using FFPE biopsy tissue. Several gene expression assays have been validated using samples from primary tumors or known metastatic lesions.8,13,14 Although a small number of patients with CUP have been tested,13 none of these assays has been evaluated in a large group of patients with CUP, nor have assay results previously been correlated with clinical features and treatment response.
We attempted to identify the putative tissue of origin in a large group of patients with CUP, using the Veridex molecular assay. The assay was technically feasible in 104 (87%) of 120 FFPE biopsy specimens submitted and provided specific diagnoses in 63 (61%) of 104 specimens tested. In 34 patients (33%), tissue available for assay was limited to small core-needle biopsies. Thus the assay seems to be feasible in patients with CUP using clinically available tissue. Furthermore, the CUP assay suggested a specific tissue of origin more often than did extensive IHC testing; in the subgroup of 37 patients who had extensive IHC testing, specific tissues of origin were suggested in four (11%) versus 23 (62%) by the CUP assay.
Validating a prediction of tissue of origin is challenging in patients with CUP, because a definitive primary cancer is rarely identified. An ideal molecular profile study is one that would be correlated with an eventual primary tumor identified, but such a study is difficult because the so-called latent primary detection rate is low. Therefore, indirect validation of the putative diagnosis is based on clinicopathologic features, response to treatment, and clinical disease course. Using these parameters, we were able to make an assessment regarding whether the CUP assay diagnoses were believable and clinically useful.
Fifty-five of the 63 patients assigned a tissue of origin by the CUP assay had results consistent with lung, pancreas, or colon cancer, whereas identification of primary ovarian (seven patients), breast (one patient), and prostate (zero patients) cancers was uncommon. The frequent identification of lung, pancreas, and colon as the tissues of origin (accounting for 53% of the entire group) is consistent with the results of previous autopsy studies. In the largest autopsy study reported, patients with lung, pancreas, and colon cancer comprised 48% of the total, whereas the diagnoses of ovary, breast, and prostate cancers were uncommon (8% of total).16,17
Most patients who had a lung or pancreatic cancer profile had clinical and pathologic features consistent with these diagnoses. Response to treatment was poor and survival was short, as would be expected in patients with advanced pancreatic or NSCLC. Patients in the retrospective cohort who had an NSCLC profile had a 33% response rate to taxane and platinum–based therapy, which is typical of advanced NSCLC. Likewise, only two of the 13 patients with a pancreas cancer profile had any response to empiric chemotherapy (both responded to gemcitabine-containing regimens).
The 23 patients with a colon cancer profile (22% of the entire group) are worthy of further discussion. Clinical and pathologic features of these patients were compatible with colon origin, although no colon primary site was visualized in the 17 patients who had colonoscopy. An accurate tissue of origin assignment in this group of patients is of potential clinical importance, because standard regimens for advanced colon cancer differ substantially from empiric CUP regimens (usually taxane and platinum). Indeed, in the retrospective cohort where all patients received standard CUP regimens, only one (14%) of seven patients responded to a taxane and platinum regimen, and none of three patients responded to gemcitabine and irinotecan. Conversely, all the patients in the prospective cohort who received standard colon cancer regimens showed a clinical response. Given the small sample size, these early results should be regarded with cautious optimism. Further experience in identifying and treating this patient subgroup is critical, because confirmation of our findings would change clinical practice.
The tissues of origin in the 41 patients categorized as “other” by this CUP assay are unknown, but this is likely to be a diverse group. Some of these patients probably had tissues of origin other than the six targeted by this CUP assay. Others may have had one of the six primary cancers included, but with atypical gene expression patterns. It is conceivable that the atypical biologic behavior of CUP is correlated with (and perhaps mediated by) a group of unique genetic abnormalities, so that the molecular gene profile differs from those of recognized primary tumors.
Additional investigation may help further clarify the role of molecular profiling in the diagnosis of patients with CUP. First, the assay we evaluated measured the expression of a limited number of genes targeting common cancers. Expansion of the assay to detect molecular profiles of additional cancer types would likely result in specific diagnoses for some of the patients categorized as “other” in this study.13,14 Second, the utility of this and other molecular assays in identifying the tissue of origin from fine-needle aspiration tissue samples should be further investigated, because this is the only available biopsy material in many patients with CUP.
In summary, our study is the first to evaluate RT-PCR molecular profiling in a large group of patients with CUP. Our results show that this approach is feasible using FFPE biopsy specimens and identifies a putative tissue of origin in the majority of patients. In most patients, clinical and pathologic features are consistent with the diagnosis made by the CUP assay. In addition, our results suggest that tissue of origin identification by this method may be of clinical value in selecting treatment, particularly in patients with a colon cancer profile. As more effective cytotoxic and targeted therapies emerge for additional known cancers, accurate identification of CUP subtypes, facilitated by the emergence of robust molecular profiling tools, will become increasingly important to the appropriate care of patients with CUP. In addition, these assays may also become research tools providing further insight into the complex biology of CUP.
AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Employment or Leadership Position: Dmitri Talantov, Johnson & Johnson (C); Tim Jatkoe, Johnson & Johnson (C); Yixin Wang, Johnson & Johnson (C) Consultant or Advisory Role: James L. Abbruzzese, Pathwork Diagnostics (U), AviaraDx (U) Stock Ownership: Dmitri Talantov, Johnson & Johnson; Tim Jatkoe, Johnson & Johnson; Yixin Wang, Johnson & Johnson Honoraria: James L. Abbruzzese, Pathwork Diagnostics Research Funding: Gauri R. Varadhachary, Veridex LLC; James L. Abbruzzese, AviaraDx, Veridex LLC; John D. Hainsworth, Veridex Expert Testimony: None Other Remuneration: None
AUTHOR CONTRIBUTIONS
Conception and design: Gauri R. Varadhachary, Dmitri Talantov, Kenneth R. Hess, Yixin Wang, James L. Abbruzzese, John D. Hainsworth
Administrative support: Gauri R. Varadhachary, Christina Meng, James L. Abbruzzese
Provision of study materials or patients: Gauri R. Varadhachary, Martin N. Raber, Christina Meng, David R. Spigel, F. Anthony Greco, John D. Hainsworth
Collection and assembly of data: Gauri R. Varadhachary, Dmitri Talantov, Christina Meng, John D. Hainsworth
Data analysis and interpretation: Gauri R. Varadhachary, Dmitri Talantov, Martin N. Raber, Tim Jatkoe, Yixin Wang, James L. Abbruzzese, John D. Hainsworth
Manuscript writing: Gauri R. Varadhachary, Dmitri Talantov, John D. Hainsworth
Final approval of manuscript: Gauri R. Varadhachary, Dmitri Talantov, Martin N. Raber, Christina Meng, Kenneth R. Hess, Tim Jatkoe, Renato Lenzi, David R. Spigel, Yixin Wang, F. Anthony Greco, James L. Abbruzzese, John D. Hainsworth
Glossary Terms
- Gene expression profiling:
- Identifying the expression of a set of genes in a biologic sample (eg, blood, tissue) using microarray technology.
- Molecular profiling:
- With the advent of bioinformatics, molecular profiling is a new discipline that uses a variety of approaches to generate a global view of mRNA, protein patterns, and DNA alterations in various cell types. Thus molecular profiles of disease processes may be seen as distinct from normal cells, and therapeutic approaches may be tailored based on molecular profiles.
- Microarray:
- A miniature array of regularly spaced DNA or oligonucleotide sequences printed on a solid support at high density that is used in a hybridization assay. The sequences may be cDNAs or oligonucleotide sequences that are synthesized in situ to make a DNA chip.
- RT-PCR (reverse-transcriptase polymerase chain reaction):
- PCR is a method that allows logarithmic amplification of short DNA sequences within a longer, double-stranded DNA molecule. Gene expression can be measured after extraction of total RNA and preparation of cDNA by a reverse-transcription step. Thus RT-PCR enables the detection of PCR products on a real-time basis, making it a sensitive technique for quantitating changes in gene expression.
Footnotes
-
Supported in part by grants from Veridex and the Minnie Pearl Cancer Foundation.
-
Authors’ disclosures of potential conflicts of interest and author contributions are found at the end of this article.
- Received September 14, 2007.
- Accepted May 16, 2008.
