The needle in the haystack: Application of breast fine-needle aspirate samples to quantitative protein microarray technology


  • This article is a US Government work and, as such, is in the public domain in the United States of America.



There is an unmet clinical need for economic, minimally invasive procedures that use a limited number of cells for the molecular profiling of tumors in individual patients. Reverse-phase protein microarray (RPPM) technology has been applied successfully to the quantitative analysis of breast, ovarian, prostate, and colorectal cancers using frozen surgical specimens.


For this report, the authors investigated the novel use of RPPM technology for the analysis of both archival cytology aspirate smears and frozen fine-needle aspiration (FNA) samples. RPPMs were printed with 63 breast FNA samples that were obtained before, during, and after treatment from 21 patients who were enrolled in a Phase II trial of neoadjuvant capecitabine and docetaxel therapy for breast cancer.


Based on an MCF7 cell line model of breast adenocarcinoma, the sensitivity of the RPPM detection method was in the femtomolar range with a coefficient of variance <13.5% for the most dilute sample. Assay linearity was noted from 1.0 μg/μL to 7.8 ng/μL total protein/array spot (R2 = 0.9887) for a membrane receptor protein (epidermal growth factor receptor; R2 = 0.9935).


The results from this study indicated that low-abundance analytes and phosphorylated and nonphosphorylated proteins in specimens that consist of a few thousand cells obtained through FNA can be quantified with RPPM technology. The ability to monitor the in vivo state of cell-signaling proteins before and after treatment potentially will augment the ability to design individualized therapy regimens through the mapping of aberrant cell-signaling phenotypes. The mapping of these protein pathways will further the development of rational drug targets. Cancer (Cancer Cytopathol) 2007. Published 2007 by the American Cancer Society.

The molecular analysis of breast tumors for the purpose of making individualized therapy decisions is expanding beyond immunohistochemistry and in situ hybridization. Until recently, molecular tumor characteristics have been based primarily on the immunohistochemical examination of tumor receptors, including estrogen, progesterone, and Her2/neu. However, evidence is emerging to support the concept that each patient's tumor has a unique complement of pathogenic molecular derangements that may impart prognostic significance.1–5 This concept was highlighted in genomic studies that were performed using gene expression profiling of breast tumors from young patients who were diagnosed with lymph node-negative disease.2 The authors of those studies reported that tumors not normally expected to produce poor outcomes based on clinical parameters, in fact, displayed poor prognostic genomic signatures. Conversely, women who would have been eligible for systemic adjuvant chemotherapy based on their clinical and histologic staging status, based on these same prognostic genomic signatures, were at low risk for the development of metastatic disease. Using this “genomic” risk fingerprint, patients in this subset had an excellent prognosis with a low risk of recurrence; therefore, most of these patients would not have benefited from adjuvant chemotherapy and would have been exposed to the toxicities and risks of such treatment. These studies have illustrated the power of translational molecular medicine related to its impact not only on treatment decisions but also on disease outcomes.2, 6

The advances in treatment and early screening programs have led to improvements in the prognosis associated with a diagnosis of breast cancer, although, clearly, additional improvements can be made. Once a diagnosis of breast cancer is rendered, the patient's prognosis conventionally has been determined by a host of clinical and histologic factors, including age, lymph node status, tumor size, pathologic grade, and histologic type.7 The process of diagnosing a palpable breast lesion may require a series of procedures: fine-needle aspiration (FNA) to characterize the mass and a core-needle biopsy or open biopsy for morphologic examination, including diagnosis confirmation, immunohistochemical analysis, grading, and surgical resection of the tumor with additional histopathologic analysis.8, 9 Realizing the new paradigm of molecular profiling with individualized treatment regimens hopefully will improve the treatment outcome for patients with breast cancer. The role of cytology in the future of molecular profiling will be significant only if the available technology is sensitive enough to work with small numbers of cells obtained in a cytology specimen. The amount of tissue necessary for the performance of these tests has decreased as ancillary studies and the molecular profiling of tumors have gained increasing roles in the individualized treatment of cancer. The role of a minimally invasive, cost-effective tissue-procurement and diagnostic procedure like FNA will play a huge role in the diagnosis and treatment paradigms for virgin, treated, and recurrent neoplasms such as breast cancer.

Molecular profiling with gene arrays provides insight into the levels of expression for genes involved in oncogenesis; however, it is known that gene transcript levels often do not correlate with the phosphorylated or functional state of the transcribed proteins.10 Posttranslational protein modifications, such as phosphorylation events at specific tyrosine (Tyr), threonine (Thr), and serine (ser) residues, play a significant role in orchestrating cellular protein-protein interactions. These finely tuned events are a significant part of cell-signaling cascades, driving cellular processes such as proliferation, apoptosis, and migration. The analysis of the cellular proteome with reverse-phase protein microarrays (RPPM) can provide information regarding the protein-protein interactions and the integrated state of cellular circuitry.11 Protein microarray technology is an emerging asset to the field of functional proteomics, because it provides the ability to investigate signaling pathways and to construct a “circuit map” of the normal and diseased state of the cellular proteome.11 In conjunction with laser-capture microdissection, protein microarrays can be used to interrogate signaling pathways in distinct cell populations. This approach is being used to study the in vivo state of the phosphoproteome in prostate12 and ovarian cancers13 and in cultured cell lines. The ability to multiplex and quantify protein-protein interactions in a parallel fashion with a single probe for multiple patient samples will permit the elucidation of many of the cell-signaling pathways that are believed to be involved in breast cancer (Fig. 1).

Figure 1.

These fine-needle aspirate (FNA) samples were analyzed with reverse-phase protein microarrays. (A) FNA samples were processed in 1 of 3 ways: stained with Papanicolaou, stained with Diff-Quik, or placed in RPMI and subsequently frozen. The archived Diff-Quik smear and the frozen samples were evaluated for their applicability to reverse-phase protein microarrays. (B) Multiple lysates were immobilized on a nitrocellulose-coated slide using a robotic printing device. The lysates were printed in a 2-fold dilution series. Quality control of the protein printing and immunostaining processes was achieved by printing cultured cell lysates and phosphopeptides on each array.

It is becoming clear that the molecular profile of a patient's tumor may dictate both important prognostic information and information for treatment selection. There is an unmet need in the medical community for economic diagnostic and prognostic assays that are able to provide molecular profiling information while minimizing invasive procedures. Currently, obtaining sufficient tissue from a patient's tumor can be a limiting factor in the analysis of protein biomarkers. Herein, we report the novel use of protein microarray technology for the analysis of archival cytology aspirate smears as well as frozen FNA material obtained from patients with breast carcinoma both before and after treatment.


FNA Samples

Specimens and relevant clinical data were obtained from patients who were enrolled in a National Institutes of Health (NIH), National Cancer Institute, Institutional Review Board-approved clinical trial with informed consent (NIH no. 00-C-0149). Details of this study and the clinical results have been published.14 Breast FNA samples from 30 women with newly diagnosed stage II or III breast cancer (median age, 50 years) were collected and prepared between 1999 and 2003. Breast FNA samples were obtained at various time points before, during, and after treatment. Twenty-one patients had pretherapy and posttherapy pairs of FNA samples available for our analysis. In addition, 21 of 30 patients in this analysis had both archival breast FNA smears and frozen FNA cell pellets from aspirates collected before and after they received neoadjuvant treatment with capecitabine and docetaxel in addition to samples that were collected prior to surgery. In total, there were 63 frozen cell aspirates. Fifty of these frozen aspirates were positive for adenocarcinoma. Five samples did not have adequate total protein and were not used for analysis. Specimens were processed within minutes of aspiration, and cell pellets were snap-frozen in liquid nitrogen.

FNA Cytologic Slides

At baseline, an FNA was performed on a palpable breast mass by a board-certified cytopathologist (A.F. or A.A.). Subsequent FNAs for the remaining time points were obtained from the same anatomic location. If the previous core-biopsy sampling of the lesion precluded direct palpation of the mass, then the FNAs were performed using random sampling of the general anatomic location where the mass had been localized prior to biopsy.

An initial FNA using a 23-gauge needle was performed. A portion of this first sample was placed on 1 slide, which was air dried and then stained with Diff-Quik for an immediate assessment of cellularity and the confirmation of malignancy. Another small portion of the sample was placed immediately into 95% ethanol for subsequent staining with the Papanicolaou stain (Fig. 1A).

The remainder of the sample was placed into a 15-mL polypropylene tube that contained 5.0 mL of RPMI (American Type Culture Collection [ATCC], Manassas, Va). One or 2 additional FNA passes of the mass were performed with the sample placed in its entirety into the 5 mL of RPMI. A single cytospin was prepared from 150 mL of this pooled sample tumor suspension, which was reviewed again to confirm the malignant nature of the cells and to perform an approximate cell count.

The RPMI/tumor cell suspension was centrifuged at ×1000g for 5 minutes until a pellet formed. The supernatant was removed; then, the tumor pellet was snap-frozen in dry ice for 5 minutes and stored immediately in a −80°C freezer, remaining in the initial 15-mL polypropylene tube. Either A.A. or A.F. screened both the Papanicolaou- and Diff-Quik-stained slides. According to standard protocol, we assumed that, because the aspirate Diff-Quik smear was mostly tumor, the FNA that was obtained from the same area also would be comprised mostly of tumor cells. The Diff-Quik-stained slides were estimated to contain >95% tumor cells and determined by prior studies to yield superior protein signals.27 The Papanicolaou-stained slides were retained for cytologic evaluation.

Archival Smear Preparation for RPPM Analysis

The archived FNA slides (1 slide per patient per sample) were soaked in a Coplin jar that contained xylene (Mallinkrodt, St. Louis, Mo) for an average of 48 hours to remove the coverslips. The slides were laid flat and air dried briefly. The cellular material was solubilized directly on the slide with from 15 μL to 20 μL of protein-extraction buffer (2.5% solution of 2-mercaptoethanol [Sigma, St. Louis, Mo] in 2 × sodium dodecyl sulfate [SDS]/Tris-glycine loading buffer [Invitrogen, Carlsbad, Calif] and equal volumes of T-PER [Pierce Biotechnology, Rockford, Ill]). The solubilized contents of the slide were scraped into a tube, boiled for 5 minutes at 100°C, and spun at 14,000 revolutions per minute (rpm) for 3 minutes. The cell lysate supernatant was placed on ice until further dilution for microarray printing.

Control Lysates

Lysates were prepared from cultured Jurkat cells (clone E6-1; ATCC) that were treated with activating, anti-Fas monoclonal antibody and epidermal growth factor (EGF)-treated HeLa cells (ATCC).11 Fas ligand-treated Jurkat cells, EGF-treated HeLa cells, and MCF7 cell lysates (BD Pharmingen, San Diego, Calif) were printed on RPPMs as quality-control samples for assessing sensitivity, precision, and system quality control.

FNA Cellular Lysates

Frozen patient aspirate samples were allowed to thaw on ice. The samples were spun at 14,000 rpm for 1 minute, the supernatant was removed, and the cell pellets were lysed immediately in protein-extraction buffer as described previously.11 Briefly, the cell pellet was lysed in a 2.5% solution of 2-mercaptoethanol in 2 × SDS/Tris-glycine loading buffer (Invitrogen) and equal volumes of T-PER (Pierce Biotechnology) with the addition of a protease inhibitor cocktail (Sigma). Lysates were heated for 30 minutes at 70°C. Next, the lysates were spun at 14,000 rpm for 3 minutes to pellet any cellular debris. The whole cell lysate was stored at −80°C until microarray printing. Prior to microarray printing, the samples were boiled at 100°C for 5 minutes.

RPPM Construction

Each sample or control lysate was printed in duplicate on glass-backed, nitrocellulose array slides (FAST slides; Whatman, Florham Park, NJ) using a GMS 417 arrayer (Affymetrix, Santa Clara, Calif) equipped with 500-μm pins. Each lysate was printed in a serial, 2-fold dilution curve (neat and at dilutions of 1:2, 1:4, 1:8, and 1:16) and a diluent control using approximately 30.0 nL lysate per spot11 (Fig. 1B). Array slides either were stored with desiccant at −20°C or were processed immediately for immunostaining. Multiple samples were printed on each microarray, and each microarray slide comprised samples that represented lysates from archival smears, frozen aspirate material, controls, and reference peptides. The samples were printed randomly to eliminate slide and positional bias, thus allowing the comparison of samples across arrays.16


The protein microarray slides were prepared for immunostaining by washing with 1 × Reblot mild solution (Chemicon, Temecula, Calif) for 15 minutes followed by 2 washes with phosphate-buffered saline (PBS) without calcium or magnesium. Slides were blocked with I-Block in PBS plus 0.5% Tween-20 (Applied Biosystems, Waltham, Mass) for a minimum of 1 hour. Analysis of protein expression on the arrays was assessed through a catalyzed signal-reporter system essentially as described previously.12, 13, 17 Briefly, the slides were placed on an automated slide stainer (Autostainer; Dako, Carpinteria, Calif) and immunostained according to the manufacturer's instructions (CSA kit; Dako). A set of specific, validated antibodies to phosphorylated or cleaved proteins in the c-erbB2 and c-erbB1 signal pathways was used to immunostain the RPPAs: EGF receptor (EGFR) Tyr1148 (Biosource International, Camarillo, Calif), EGFR Tyr1173, EGFR Tyr1068, protein kinase B (Akt) ser473, Akt Thr308, Bcl2 ser70, erbB2, erbB2 Tyr1248, eIF4E ser209, estrogen receptor α (ERα), ERα ser118, 4E-binding protein 1 (4EBP1) Thr37/46, 4EBP1 ser65, phosphatase and tensin homolog (PTEN) ser380, cleaved caspases 3 and 9, and signal transducer and activator of transcription 3 Tyr705 (Cell Signaling Technology, Beverly, Mass). The negative control slide was incubated with antibody diluent without primary antibody. The secondary antibody probe was goat antirabbit immunoglobulin-G H&L chain (1:5000 dilution; Vector Laboratories, Burlingame, Calif). Total protein per microarray spot was determined with Sypro Ruby protein blot stain (Molecular Probes, Eugene, Ore) or colloidal gold protein stain (AuroDye Forte; Amersham Biosciences, Piscataway, NJ) according to the manufacturer's directions and was imaged with a charged-coupled device camera (Alpha Innotech, San Leandro, Calif). The spot intensity was integrated over a fixed area as described previously.11 Each array was scanned on an Epson flatbed scanner at 1200 dots per inch, the spot intensity was analyzed (ImageQuant, version 5.2; Molecular Dynamics, Sunnyvale, Calif), data were normalized to total protein, and a standardized, single data value was generated for each sample on the array. This single data point was used for subsequent data analysis.

Statistical Analysis

Unsupervised hierarchical clustering (Ward method) and Spearman ρ nonparametric chi-square correlations were performed with JMP software (version 5.1; SAS Institute, Cary, NC).


Adequacy of Archival FNA Slide Samples

Previous studies of microdissected tissue samples with the RPPM platform indicated that from 1.0 μg/μL to 0.0625 μg/μL of total protein per spot is required for total protein determinations with Sypro Ruby protein blot stain or a colloidal gold stain.13, 17, 18 In an initial assessment of the applicability of FNA material for RPPMs, we compared total protein per spot for 8 cases, each of which consisted of an archival smear, as well as pretreatment and posttreatment frozen aspirate samples (Fig. 2). Semiquantitative cell counts performed at the time of sample collection for 7 of 9 of the baseline samples ranged from 1000 cells to 50,000 cells. Each sample was solubilized in an appropriate volume of extraction buffer based on cell count (100 cells/1.0 μL). Samples were printed in duplicate on 50 microarray slides. For a quality control of protein loading on the array, a Jurkat cell lysate of 2.0 μg/μL total protein was printed in duplicate on the same array.

Figure 2.

This chart illustrates the assessment of total protein yield from fine-needle aspirate (FNA) samples. Eight sample sets, consisting of coverslipped, archival, Diff-Quik-stained FNA slide smears and pretreatment and posttreatment frozen samples, were solubilized and printed on a reverse-phase protein microarray. Mean spot intensities were calculated from a Sypro Ruby protein blot-stained array. The undiluted pretreatment (squares) and posttreatment (triangles) frozen samples had total protein concentrations equivalent to a 1.0 μg/μL Jurkat cell lysate (crosses), whereas the archival smear sample (circles) had total protein concentrations 2- or 3-fold less. The large standard deviation noted for the frozen samples is because of the variation in cell numbers in the frozen aspirate samples.

The mean total protein (n = 8 samples) per spot was from 2- to 3-fold greater in the frozen tissue samples compared with the archival smears. The protein content of these samples was sufficient for protein quantitation with Sypro Ruby stain based on the relative spot intensities. The large standard deviation noted for the frozen samples was because of the variation in cell numbers in the frozen aspirate samples (range, 1000–50,000 cells). Thus, material obtained from archival cytology smears is adequate for performing RPPM; however, the frozen material is superior in terms of the relative intensity of the measured protein signal.

RPPM Sensitivity and Precision

Paweletz et al. thoroughly evaluated the sensitivity and precision of the RPPM using recombinant prostate-specific antigen molecules (sensitivity, 5 × 10−20 mol) and microdissected tissue samples (interslide coefficient of variation [CV], 1.26–39.4%; intraslide CV, 1.66–21.18% for 1/10,000 cell equivalents).19 Because variations in robotic printing devices will affect microarray spot size and protein loading, all FNA samples in this study were printed with a GMS 417 arrayer equipped with 500-μm pins. The sensitivity of any given antigen-antibody interaction is dependent on the detection system as well as the relative abundance of the antigen-antibody species and the relative binding affinities of the antibody.20 Although we were able to quantitate protein in the archival smears with Sypro Ruby staining, more sensitive determinations of total protein were needed for the analysis of low-abundance analytes within the linear dynamic range of the assay.

Previous work in our laboratory indicated that a sample with approximately 1.0 μg/μL to 2.0 μg/μL total protein contained adequate total protein, when diluted and arrayed on the RPPM, for detection with a Sypro Ruby stain (manufacturer's detection limits, 0.25–1.0 ng; Molecular Probes Sypro Ruby blot stain product information sheet revision March 3, 2005). Initial results comparing the relative mean spot intensities between archival FNA smears and pretreatment and posttreatment frozen FNA samples indicated the FNA samples were within the detectable range of the total protein assay (Fig. 2). Therefore, we determined the sensitivity and precision of the RPPM with respect to total protein and cellular equivalents from an MCF7 cell culture lysate as a representative breast adenocarcinoma sample.

MCF7 lysates were printed in 12-point, 2-fold dilution series (n = 5) (Fig. 3A) and stained with colloidal gold for total protein determination. Total protein sensitivity was determined to be 7.8 ng/μL (linear correlation [R2] = 0.988) (Fig. 3B). Protein concentration expressed as cellular equivalents on the microarray were 78 cell equivalents/μL lysate or 1/10,000 cell equivalents loaded per spot (Fig. 3C).

Figure 3.

The sensitivity of reverse-phase protein microarrays. (A) MCF7 cell lysates were printed in triplicate, 12-point, 2-fold dilution series and stained with a colloidal gold total protein for determination of total protein sensitivity. EGF indicates epidermal growth factor. (B) The MCF7 lysate had a total protein linear correlation over 8 dilution series (correlation coefficient [R2] = 0.9887) with a limit of detection of 7.8 ng/μL. (C) The sensitivity represented as cell equivalents corresponded to 78 cells per μL lysate or as 1/10,000 cellular equivalents loaded per spot.

To determine interslide precision, the MCF7 cell line, which is known to overexpress the EGFR, was used as a model. MCF7 lysates were printed in triplicate on 5 slides and immunostained with anti-EGFR (1:100 dilution). Excellent linear correlations were observed (R2 = 0.9821, n = 5) (Fig. 4A) across the data set. The CV for interslide reproducibility ranged from 5.1% for the undiluted spot, to 4.1% for the 1:2 dilution, and up to 13.5% for the 1:16 dilution (Fig. 4A). The greater variance noted in the undiluted spot may have been caused by saturation of the antigen binding sites at higher antigen concentrations for a given antibody concentration. Intraslide variation was <8.6% (n = 12) for all dilution points analyzed with good linearity of R2 = 0.9957 (Fig. 4B).

Figure 4.

The precision of reverse-phase protein microarrays. MCF7 cells were used as a model lysate for epidermal growth factor (EGF) receptor precision studies. (A) Arrays (n = 5) that were immunostained with anti-EGF receptor had an interslide coefficient of variation (CV) for EGF receptor concentrations ranging from 5.1% for the most concentrated spot to 13.5% for the most dilute spot (R2 = 0.9821). (B) The CV for intraslide reproducibility for 5 slides (total number of spots at 15 per dilution point) ranged from 8.6% for the most concentrated spot to 5% for the most dilute spot (R2 = 0.9957).

Phosphoproteomic Kinase Analysis of Frozen FNA Samples

RPPMs offer the promise of being able to coordinately investigate multiple phosphorylated kinases, thus developing a proteomic molecular profile of cellular processes for individual patients. Therefore, we immunostained the RPPMs with 17 antibodies, which represented apoptotic and prosurvival cell-signaling pathways. In this feasibility study set, unsupervised Bayesian clustering analysis for 21 breast FNA samples revealed heterogeneity between samples and treatment time points (Fig. 5A). The baseline samples appeared to be heterogeneous for the protein endpoints measured and did not segregate into distinct clusters. In contrast, for the 7 samples for which there was a surgical sample available, 6 of 7 surgical samples were in the same cluster. These data compliment the results from previous studies of human epithelial tumors in which heterogeneous signaling also was noted,18 suggesting that FNA material indeed is an acceptable sample for phosphoproteomic analysis. Even though there was evident heterogeneity in the patient samples, there was significant clustering of prosurvival proteins in the PTEN-Akt and EGFR pathways to indicate common pathway activation for many of these samples. In the entire cohort of patients studied, as reported previously, only 31% of the patients had a complete response to therapy, suggesting that multiple prosurvival proteins are activated, possibly through a variety of pathways, for patients in whom therapy fails.14

Figure 5.

Phosphoproteomic analysis of fine-needle aspiration (FNA) breast samples. (A) Unsupervised Bayesian clustering (Ward method) for pretreatment and posttreatment frozen FNA samples. Each row represents an individual sample at a treatment time point, and each column represents a protein endpoint. Relative protein expression is denoted as elevated (red) or decreased (blue). (B) Nonparametric Spearman ρ analysis revealed significant correlations for protein nodes in the phosphatase and tensin homolog-protein kinase B (PTEN-Akt) and epidermal growth factor receptor (EGFR) prosurvival pathways.

Nonparametric Spearman ρ analysis for the entire study set revealed significant correlations between PTEN ser380 and Akt Thr308 (P < .001; R2 = 0.9096) (Figs. 5B, 6A). Additional nonparametric associations were noted for proteins that have been related in breast tumors, again confirming the ability to use FNA material for phosphoproteomic analysis (Fig. 6B-D). These data suggest that FNA material may be used to evaluate treatment effects and to redesign therapy regimens without the need for surgery.

Figure 6.

Spearman ρ nonparametric correlation of prosurvival pathway protein endpoints. (A) A strong association (R2 = 0.9096) is observed between the relative protein levels of phosphatase and tensin homolog (PTEN) serine 380 (ser380) and protein kinase B (AKT) threonine 308 (Thr308) for the pretreatment and posttreatment frozen fine-needle (FNA) samples (n = 63). (B-D) Additional nonparametric correlations were not identified as significant (PTEN ser380 and epidermal growth factor receptor tyrosine 1178 [EGFR Y1178], R2 = 0.5998; PTEN ser380 and Her2, R2 = 0.638; AKT Thr308 and 4E-binding protein 1 Thr37/46, R2 = 0.5469).


The application of transcriptional profiling to cytologic material obtained by FNA has been effective for messenger RNA (mRNA) expression of traditional, single-gene markers, such as ER or HER-2.21, 22 However, DNA and mRNA alterations may not necessarily correlate with posttranslational protein modifications, such as phosphorylation, acetylation, ubiquitination, or glycosylation, which confer a variety of biologic properties.23–25 Therefore, understanding the protein-protein and/or protein-DNA interactions that occur during tumorigenesis is necessary for the development of therapeutic intervention strategies targeted to an individual patient's molecular profile.12, 26

Proteomic analysis using surface-enhanced laser desorption/ionization (SELDI) of FNA material has been performed successfully on archival slide material and clinical samples.15, 27 Despite the ability of SELDI technology to generate a tumor-specific protein biosignature, the identification and quantification of proteins are limiting factors for the utility of SELDI in the diagnostic arena. In contrast, RPPM technology is a currently available proteomics tool for examining the activation status of a number of proteins that, it is believed, participate in oncogenesis and chemotherapy resistance; using this technology with FNA samples has important clinical implications.

The ability to use in vivo, liquid-based FNA material and archival slide material to profile the activation state of specific proteins in known signaling pathways has far-reaching clinical implications. We have demonstrated that the presence of low-abundance proteins in specimens that consist of a few thousand cells obtained by FNA can be assayed with RPPM technology. The individual heterogeneity and complexity of tumors has been established based on proteomic and genomic studies.1, 28 Microarrays used to profile tumors at the DNA, RNA, and protein levels have revealed disease-susceptibility genes that, when mutated, can alter protein signaling networks, which may vary between patients with similar tumors.1, 11, 28, 29 Similar to the results demonstrated with RPPM by Wulfkuhle et al. in ovarian cancer,13 there is considerable variation in the levels of activated proteins, such as EGFR and PTEN, both between patients and during various treatment time points for a single patient (Fig. 5A). The complex cell-signaling activation that occurs during oncogenesis and treatment may be patient-specific rather than disease-specific, emphasizing the need for individual patient profiling.

There are several advantages to using FNA specimens for proteomic analysis. The utility of an FNA performed by a skilled pathologist for the evaluation of a palpable breast lesion is well established. FNA, as a diagnostic procedure, when combined with the other components of the “triple test” (ie, physical examination, FNA, and mammography), has a reported sensitivity equivalent to an open biopsy, ranging from 82% to 99%.30–34 The advantages of an FNA over more invasive procedures, such as a core-needle biopsy, include not only cost effectiveness35 but more rapid diagnosis reporting and less trauma for the patient. Furthermore, the tumor sample obtained from an FNA contains a comparatively purer representative sample of tumor cells than a core-needle biopsy, which may contain larger numbers of tumor-infiltrating lymphocytes and stromal elements.21, 36 Symmans et al. demonstrated that FNA material obtained from breast carcinomas provided sufficient material to perform combinational DNA microarray gene-expression profiles with a total RNA yield similar to that of core-needle biopsy samples.36 The ability to obtain sufficient cells was substantiated by our results, which indicated that both archival and frozen FNA material contained ample amounts of protein for RPPM analysis using either Sypro Ruby or colloidal gold assays (Figs. 2, 3). Although less total protein was detected in archival smear preparations than in frozen material, we were able to quantify individual phosphoproteins with archival FNA samples (data not shown). Although serum, erythrocytes, and white blood cells may contribute to the overall total amount of protein per sample, this additional protein should not skew relative differences in the data. The total protein value is a scalar quantity, permitting comparison of relative differences in phosphoprotein levels in the background context of their total protein matrix. Measurement of total protein on the actual array spot, rather than measurement of a separate aliquot of lysate, conserves precious sample and allows each array spot to be normalized to total protein concentration. The smaller quantity of protein on the archival smears may pose a limitation for this methodology when attempting to detect very low abundance proteins. It is not unexpected that archival samples would exhibit a lower yield of protein extraction because of drying and the effects of storage temperature and time. Alternative detection strategies, such as Quantum Dots (Invitrogen), currently are being used in our laboratory as a means of increasing the sensitivity of the RPPM for archival material.37

Cellular kinase concentrations exhibit a large, dynamic range, necessitating the use of a detection system that is capable of detecting low-abundance kinases, such as transcription factors, and more abundant cell-membrane receptors. Polymerase chain reaction-style amplification chemistries do not exist for protein; thus, the amplification method must be linear, reproducible, and sufficiently sensitive to provide reliable, quantitative analysis and clinical utility.11 The total number of cells collected from an FNA may vary from a few thousand cells to 1 million cells. RPPMs incorporate sample dilution curves and amplification chemistries to address the challenges of linearity, sensitivity, and quantitation. Paweletz et al. demonstrated sensitivity in the femtomolar range with strong linearity and low CVs.19 Using an MCF7 breast adenocarcinoma cell line, we demonstrated good linearity both for a membrane receptor protein (EGFR; R2 = 0.9935) (Fig. 3B) and for total protein (R2 = 0.9887) (Fig. 4A).

Both apoptotic and prosurvival pathways with different subcellular pathways were probed in this feasibility study set. It has been demonstrated that apoptotic pathways involving Bcl-2, caspase-3, and caspase-9 are modulated with chemotherapy. HER2 is a validated target in breast cancer, and growth factor receptors like EGFR are emerging as promising molecular-targeted therapies.

In the current study, we demonstrated detectable differences in the EGFR mitogenesis pathway (EGFR Tyr1148, Akt Thr308, and PTEN ser308) after chemotherapy (Fig. 5B). Thus, FNA can be used successfully to investigate the subcellular circuitry of a patient's cancer by providing both diagnostic material and a representative tumor population. Direct, quantitative measurements can be made from RPPMs at any point on the dilution curve that is within the linear dynamic range of the antibody-analyte interaction28 (Fig. 1B).

The small sample pool precluded our ability to draw statistically significant correlations between treatment and response. However, a number of interesting correlations were observed. Loss of the tumor-suppressor gene PTEN has been implicated in variety of cancers, including endometrial, prostate, breast, and colon cancers as well as glioblastoma.38–42 By removing the phosphate group from phosphoinositide-3,4,5-triphosphate, PTEN inactivates Akt signaling. The down-regulation of Akt has a multitude of downstream effects, including the promotion of proapoptotic pathways and the inhibition of cell cycle progression.43 The level of phosphorylation and, subsequently, stability of PTEN may be influenced by hormones like estrogen and progesterone. It is hypothesized that both PTEN and ERα are phosphorylated by the protein kinase casein kinase II, which is under the influence of estrogen.44, 45 In the current study, we observed strong, nonparametric correlations (Spearman ρ) between PTEN ser380 and Akt Thr308 (P < 8.16 × 10−16), between PTEN ser380 and EGFR Y1148 (P < 1.69 × 10−13), between PTEN ser380 and ErbB2 (P < 7.55 × 10−9), and between Akt Thr308 and 4EBP1 (P < 4.77 × 10−7) (Fig. 6A-D). It is believed that loss of PTEN activity in patients with breast cancer confers resistance to trastuzumab treatment.46 Zhou et al. reported increasing levels of phosphorylation of Akt, mammalian target of rapamycin, and 4EBP1 as ductal epithelium progressed to breast cancer, and those authors also reported a positive correlation with ErbB2 overexpression.47 Although we were unable to demonstrate a significant correlation between ErbB2 and Akt or ERα and PTEN, we detected changes in these proteins for individual patients at various time points during therapy (Fig. 5A). There is a clear need for novel therapeutic strategies like combination therapy aimed at specific targets within key prosurvival signaling pathways, such as PTEN, Akt, and 4EBP1.

The global scientific community is moving rapidly toward the goal of prioritizing molecular research and development as an integral part of breast cancer evaluation. The Danish Center for Translational Breast Cancer Research has outlined the allocation of a mastectomy specimen to include functional genomics, proteomics, transcriptomics, immunohistochemistry, and 3-dimensional culturing48 with the intention of advancing biomarker discovery as well as understanding the molecular mechanisms of oncogenesis for therapeutic applications. Overall, our findings demonstrate the ability to obtain molecular profiles from FNA samples by RPPM. Further studies with larger numbers of patients are needed to determine in more detail how this technology can be used to provide valuable information to impact patient management. The allocation of patient samples, including FNA material, may include sending a portion of patient material for genomic and proteomic evaluation as translational molecular medicine becomes more widely established.

The search for cancer biomarkers and unique cell-signaling aberrations that may be targets for individualized pharmacologic therapies now has been set in motion. Commissions such as the National Cancer Institute/Food and Drug Administration Interagency Oncology Task Force are dedicated to furthering the discovery of biomarkers and molecular cancer targets. Minimally invasive FNA samples, as part of this discovery process, appear to have adequate sensitivity, precision, and ease of use as an acceptable medium for translational research and development. We have demonstrated that quantitative differences can be detected in a small number of cells from an FNA sample and can be used to investigate hundreds of phosphorylated kinase-specific endpoints. FNA samples potentially may be used to address the following needs of translational research: 1) identification of patients for molecular-targeted therapy, 2) understanding molecular interconnections in tumors that develop chemoresistance, 3) the potential to develop molecular profiles of recurrent tumors without surgery, and 4) the ability to redesign therapy based on functionally significant pathways.