Although results from several studies have indicated that molecular profiling of solid tumors improves treatment stratification (33
) and efficacy monitoring (34
), failure to detect molecular heterogeneity in cancer patients can lead to underpowered clinical trials (35
). However, the availability of serial tumor tissue to make such decisions during treatment is often limited because core biopsies carry procedural risks and are time-consuming and costly. Moreover, biopsies often yield small amounts of tissue (several cubic millimeters), which allow for only limited analyses. Conventional methods for molecular profiling (immunohistochemistry, flow cytometry, or proteomics analyses) often require considerable quantities of cells or tissues, both of which are not easily obtained with fine-needle aspirates. These caveats have fueled an intense interest in developing more sensitive technologies for enabling broader profiling of tissue in limited clinical specimens. Recent advances in nanotechnology and device miniaturization have made sophisticated molecular and cellular analyses of scant tumor samples possible, albeit often under well-controlled experimental conditions (15
). Yet, despite the development of various platforms through the National Cancer Institute's Cancer Nanotechnology Initiative (38
), only a few of these have advanced into clinical feasibility trials (3
The current clinical study revealed several unexpected findings. First, we observed considerable expression heterogeneity for all bio-markers across the sample populations. For example, EpCAM, the current marker of choice to define circulating tumor cells, was only highly expressed in ~60% of cancers and completely absent in ~20%. Second, we observed significant expression heterogeneity along identical and distal biopsy sites within a given tumor lesion. These findings have important implications for both molecular diagnostics and therapeutic drug targeting. Third, we obtained time courses on protein viability that demonstrated rapid decay, informing the need for prompt proteomic and other molecular measurements of human samples. Finally, we show that molecular profiling based on multimarker diagnostics in a point-of-care setting can have higher diagnostic accuracies when compared to state-of-the-art conventional pathology.
Of the individual markers investigated, MUC-1, HER2, EGFR, and EpCAM provided the highest diagnostic accuracy. Combining these four markers established correct diagnoses in 48 of the 50 patients in the initial cohort and in all 20 patients in the independent test set. This accuracy was superior to conventional clinical analysis. In the two misclassified cases, core biopsy showed significant inflammation and an absence of cancer cells. Although not attempted in this particular study, incorporating additional markers (for example, CD163, CD14, CD16, CD33, and 5B5), which define monocyte, macrophage, and fibroblast populations more accurately, would make characterizing the inflammatory and stromal components of fine-needle aspirate samples possible and could increase specificity (43
). The current study was specifically designed to include a range of intra-abdominal tumor types so as to simulate the typical clinical referral pattern seen at an interventional service. Although peripheral to the central study aims, it is noteworthy that the three- and four-marker combinations were found to offer similar predictive accuracies, both being superior to EpCAM alone and to conventional cytopathology. This study was not powered to be specific for particular cancer subsets, but it is likely that protein markers could be tailored to recognize specific epithelial cancers (for example, those of prostate and lung) or nonepithelial cancers (for example, melanoma, sarcoma, and lymphoma).
With respect to protein stability, we discovered early on that all cancer markers displayed relatively short half-lives once the tissue was harvested. To date, limited information has been available on the half-life of protein expression levels in aspirated cancer cells, and marker degradation may be one of the reasons for the lower detection sensitivities reported in some studies. Proteomic studies of freshly harvested nonmalignant cells have demonstrated that up to 40% of protein markers are differentially expressed when in vivo and in vitro conditions are compared (18
). Within the first hour after harvesting, we observed a mean decrease of ~100% in marker expression across the different markers studied (). The magnitude of this effect was unexpected and indicates that samples either require rapid analysis (for example, within minutes) or careful preservation using optimal methods to maintain the integrity of protein expression. For all of our clinical samples, we fixed the cells immediately after procurement using methods optimized with cell lines. This included treatment with either formaldehyde or a combination of formaldehyde and the detergent saponin, depending on whether the target was expressed on the surface or inside of the cell, respectively.
The μNMR device in its current form (DMR-3) is an advanced prototype specifically designed for clinical operation and has been significantly improved over previous prototypes such as DMR-1 (15
) and DMR-2 (17
). The DMR-3 system incorporates several new features including an array of solenoidal coils for multiplexed detection; a disposable, thin-walled sample container, which tightly slides into the coils; custom-designed and easy-to-use NMR hardware, which automatically tunes measurement settings (NMR frequency, pulse width, and power), to compensate for environmental factors such as temperature fluctuations; and a user-friendly interface for use with a smart phone (). Compared to other analytical techniques, the main advantage of μNMR is its capacity for rapid measurements with little interference from blood (that is, it allows for analyses of nonpurified samples). The platform is also versatile and scalable to easily accommodate additional bio-markers of interest, offers robust portable operation, and is relatively inexpensive compared to conventional histopathology, all key attributes for emerging nanotechnology-based diagnostics. Despite these advances and advantages, we believe that DMR-3 could be further enhanced to maximize clinical utility. We anticipate furnishing the device with more advanced multichannel measurement and microfluidic (for example, separation) capabilities to facilitate on-chip processing of whole-blood samples. Likewise, we are currently exploring additional technologies for combining sensitive μNMR measurements with higher-throughput purification chips (19
). We are also investigating technologies for the analysis of individual, magnetically tagged cells using miniature magnetometer sensors. Achieving single-cell resolution will enable the diagnostic study of rare cells, such as circulating tumor cells, for screening or monitoring cancer recurrence (47
). Finally, sensitive and detailed analyses of other cell types, including immune cells, stem cells, or non-epithelial neoplasms, could be performed with μNMR, which could thus facilitate the development of additional surrogate endpoints for clinical trials (7
Here, we show that the μNMR technology can yield highly sensitive and reproducible data, with implications for enhancing clinical decision-making. The method relies on a sophisticated technology rooted in the basic principles of NMR (15
) as well as on exploiting advanced nanoparticle-targeting strategies (24
). Extracting concurrent molecular information from fine-needle aspirates could minimize the incidence of nondiagnosis associated with existing standards of care and even improve diagnostic accuracies. Moreover, this minimally invasive procedure paves the way for repeated tumor samplings at various time points. Neoadjuvant treatment, for example, is a clinically accepted approach where chemotherapy precedes surgical resection. Serially interrogating tumor lesions during treatment would offer multiple windows into the biology of the tumor and its response to drug treatment. Moreover, the clinical utility of repeated biopsies for such purposes has longstanding precedence in clinical research (50
). More recently, it has been shown that patients with various treatment refractory malignancies benefit clinically from the use of conventional laboratory methods to measure therapy-specific proteins or genetic markers in core or surgical tumor biopsies (6
). Harnessing the rapid, multiplexed, and sensitive detection attributes of μNMR could enable future molecular profile–directed studies in sample-restricted trials.
We envision a number of specific clinical applications not tested here in which μNMR could be particularly useful, namely, forrapid detection and serial profiling of commonly attained specimens (fine-needle aspirates of thyroid tissue, paracentesis, thoracentesis, peripheral blood, and image-guided or surgical biopsies), for repeat treatment assessment (“pharmacodynamics”), and for robust and tumor-specific profiling of circulating microvesicles (exosomes) in blood that are shed by tumor cells. We anticipate that this versatile technology will find a wide range of applications in oncology because it enables molecular diagnostics at the bedside and has the potential for redefining the standard of care during diagnostic workup.