Molecular characterization of scant lung tumor cells using iron-oxide nanoparticles and micro-nuclear magnetic resonance

Abstract

Advances in nanotechnology and microfluidics are enabling the analysis of small amounts of human cells. We tested whether recently developed micro-nuclear magnetic resonance (μNMR) technology could be leveraged for diagnosing pulmonary malignancy using fine needle aspirate (FNA) of primary lesions and/or peripheral blood samples. We enrolled a cohort of 35 patients referred for CT biopsy of primary pulmonary nodules, liver or adrenal masses and concurrently obtained FNA and peripheral blood samples. FNA sampling yielded sufficient material for μNMR analysis in 91% of cases and had a sensitivity and specificity of 91.6 and 100% respectively. Interestingly, among blood samples with positive circulating tumor cells (CTC), μNMR analysis of each patient's peripheral blood led to similar diagnosis (malignant vs benign) and differential diagnosis (lung malignancy subtype) in 100% and 90% (18/20) of samples, respectively. μNMR appears to be valuable, non-invasive adjunct in the diagnosis of lung cancer.

Introduction

Lung cancer is among the most common and fatal cancers worldwide. In 2013, over 225,000 new cases and159,000 deaths are estimated to occur in the US alone¹. While most patients are currently treated with a combination of surgery, radiation and chemotherapy, novel therapies are emerging for specific lung cancer subtypes^2-4. Both treatment initiation and confirmation of recurrence commonly rely on primary tissue sampling occurring through bronchoscopy or CT guided lung biopsy. Either approach yields small cores of tissue which are embedded, sectioned and then stained for immunohistochemistry. Percutaneous core biopsies with 17-19G coaxial needles, however, confer morbidity and throughput is generally low ^5,6.

Alternative sources of malignant cells for diagnostic and molecular testing include peripheral blood (circulating tumor cells or CTC) ^7-9, pleural fluid (thoracentesis) and fine needle aspirates (FNA). While these cells are attainable through less invasive measures, their challenges include their scant nature^10-12, limited viability, epithelial-to-mesenchymal transition (EMT) ¹³ and heterogeneous protein expression levels. CTC are rare (∼1-100 cells/mL of blood)^11,14, while a single FNA pass often yields < 200 cells depending on technique ¹². Irrespective of the source and method, malignant cells are outnumbered by co-existing host cells from blood and tissue samples (immune cells, mesothelial cells, fibroblast).

Advances in nanotechnology and microfluidics have expanded the feasibility of molecular analyses using limited clinical specimens. A spectrum of new methods have appeared ^15,16 varying in throughput, accuracy and ease of use. We have recently developed a μNMR approach that allows profiling of fine needle aspirates ^12,17, exosomes ¹⁸, and potentially individual cells ¹⁹ across the spectrum of solid tumors. The method is fast, accurate and applicable in a point of care setting. To inaugurate testing of our technology for pulmonary malignancies, we enrolled 35 patients referred for biopsies of primary lung lesions or their potential liver or adrenal metastases. We then compared our FNA and CTC analyses to conventional pathology interpretation of each patient's core biopsy (Figure. 1). We show that our method is accurate and when combined with CTC analysis, could potentially avoid the need for core biopsies.

Figure 1. Study outline.

Patients were consented for a peripheral blood sample prior to the procedure as well as an FNA sample during a routine CT guided lung biopsy. Aliquots of samples were incubated with either TCO-modified antibody cocktails (quad, cocktail) or single, specific antibodies. After washing, antibodies were revealed with magnetic nanoparticles detected by μNMR.

Methods

Patient population and analysis

The study was approved by the Institutional Review Board, and the procedures followed were in accordance with institutional guidelines. Informed consent was obtained from all subjects. Thirty-five subjects requiring biopsy were enrolled in this study. Thirty-two of them had pulmonary nodules that required tissue diagnosis, while 3 had known primary lung cancer with suspected adrenal (2) or liver (1) metastases. On the day of enrollment, both peripheral blood (7 mL) and fine needle aspirates from intraparenchymal masses were collected from each subject. Five clinicians (M.P., A.S., C.M.C, J.A.S. and R.W.), reviewed the documented clinical, imaging, and pathology data for each subject with cancer. Tables 1 and 2 summarize the different cohorts used for analyses.

Table 1. Overview of patient population.

Characteristic	Number	Percentage
Number of patients	35
Age
Median	70.5
Range	21-87
Gender
Male	21	60
Female	14	40
Primary Lung Cancer	24	68.6
Tumor Subtypes
Adenocarcinoma	17	48.6
Squamous Cell Carcinoma	4	11.4
Large Cell Carcinoma	1	2.9
Small Cell Carcinoma (SCLC)	2	5.7
Stage
IA	4	11.4
IB	3	8.6
IIA	1	2.9
IIB	4	11.4
IIIA	2	5.7
IIIB	0	0
IV	8	22.9
Limited Stage SCLC	2	5.7
Advanced Stage SCLC	0	0
Lung Metastases	7	20
Melanoma	4	11.4
Renal Cell Carcinoma	1	2.9
Ewing's Sarcoma	1	2.9
Unknown Primary	1	2.9
Benign Lesions	4	11.4

Table 2. Detailed patient information.

The table shows patients' clinical information and whether FNA and blood were obtained from each patient (last 2 columns). FNA samples from three patients (pts 4, 6 and 12) were acellular; with not enough material to proceed with DMR. All patient had pulmonary nodules or masses ranging from 0.7 -9.0 cm in diameter. There were 17 adenocarcinomas, 4 squamous cell carcinoma, 2 small cell carcinomas, 1 large cell carcinoma, 7 metastases and 4 benign nodules.

#	Pathology	Sex	Age	Diameter	Stage	FNA	Blood
1	Adeno	Female	69	2.6 cm	IIB	yes	yes
2	Adeno	Female	71	3.0 cm	IIB	yes	yes
3	Adeno	Female	71	1.2 cm	IA	yes	yes
4	Adeno	Female	75	0.7 cm	IV	acellular	yes
5	Adeno	Female	70	4.7 cm	IV	yes	yes
6	Adeno	Female	67	2.8 cm	IV	acellular	yes
7	Adeno	Male	56	2.3 cm	IA	yes	yes
8	Adeno	Male	71	8.2 cm	IIIA	yes	yes
9	Adeno	Male	74	1.9 cm	IV	yes	yes
10	Adeno	Male	73	6.4 cm	IV	yes	yes
11	Adeno	Female	77	1.0 cm	IA	yes	yes
12	Adeno	Male	78	3.7 cm	IIB	acellular	yes
13	Adeno	Male	87	2.4 cm	IIIA	yes	yes
14	Adeno	Male	58	2.4 cm	IA	yes	yes
15	Adeno	Female	47	4.2 cm	IIA	yes	yes
16	Adeno	Male	57	3.5 cm	IB	yes	yes
17	Adeno	Male	65	2.0 cm	IV	yes	yes
18	Squamous	Male	75	5.0 cm	IV	yes	yes
19	Squamous	Male	64	2.2 cm	IB	yes	yes
20	Squamous	Male	75	3.2 cm	IB	yes	yes
21	Squamous	Female	81	2.0 cm	IV	yes	yes
22	Small	Female	81	3.2 cm	Limited Stage	yes	yes
23	Small	Female	73	2.0 cm	Limited Stage	yes	yes
24	Large	Male	54	6.5 cm	IIB	yes	yes
25	Met	Male	73	1.7 cm	IV	yes	yes
26	Met	Male	78	3.1 cm	IV	yes	yes
27	Met	Female	43	2.0 cm	IV	yes	yes
28	Met	Male	65	1.9 cm	IV	no	yes
29	Met	Male	70	4.3 cm	IV	yes	yes
30	Met	Male	21	1.2 cm	Metastatic	yes	yes
31	Met	Female	78	2.3 cm	IV	yes	yes
32	Benign	Male	68	4.9 cm		yes	yes
33	Benign	Male	78	2.3 cm		no	yes
34	Benign	Female	59	1.4 cm		yes	yes
35	Benign	Male	66	2.2 cm		yes	yes

Selection of biomarkers

Detection of malignancy

We used a previously identified cocktail of four (quad) markers (EGFR, EpCAM, HER-2, MUC-1) for CTC detection, as the combined application of these markers allows more accurate CTC counting than a single marker (EpCAM) based detection ²⁰. Subclassification of lung cancer: we selected clinically relevant markers used in differentiating subytpes of lung cancer: adenocarcinoma (TTF1, Napsin A), squamous cell carcinoma (p40), small cell carcinoma (a cocktail of Synaptophysin + Chromogranin). Table 3, Figure S1 and Figure S2 show the markers, antibodies, and diagnostic algorithm used for the detection of cancer and differentiation of lung cancer subtypes.^21-24.

Table 3. Molecular markers used for diagnosis and sub-typing.

The quad marker test was used to determine if a malignancy was present. The quad marker contains four antibodies against common tumor markers (EGFR, EpCAM, MUC-1 and Her2). Adenocarcinoma was identified by positivity for TTF-1 and/or Napsin A. Squamous cell carcinoma was identified by positivity of p40. Small cell lung cancer was diagnosed by a positive signature for synaptophysin and/or chromogranin. Large cell carcinoma and metastasis were negative for all lung cancer specific panel of markers (TTF-1, Napsin A, p40, synaptophysin, chromogranin).

Category	Question	Quad EGFR + EpCAM + Muc + Her2	TTF1	Napsin A	p40	Cocktail Synaptophysin + Chromogranin
		Malignancy				Malignant	+
Benign	-
Primary	Lung	+	+
	Met	+	-
Type	Adenocarcinoma	+	+	+	+/-	-
	Squamous cell	+	-	-	+	-
	Small cell carcinoma	+	+/-	-	-	+
	Large cell carcinoma or metastasis	+	-	-	-	-

Sample processing and labeling

Each peripheral blood sample (7 ml) was lysed and the cell pellet resuspended in buffer (100 μl of 1X PBS/1% FBS). FNA samples were fixed as described previously ¹². For molecular profiling experiments of both blood and FNA specimens, primary antibodies were separately added to aliquots of parent samples. Cell pellets were incubated for 20 minutes. TCO-modified secondary IgG antibodies (10 μg/ml) were then added to the pellet and incubated for 20 minutes. Cell pellets were then washed twice with 1 × phosphate buffered saline (PBS) and incubated with magnetic nanoparticles (100 nM Tz-CLIO) for 10 minutes. Excess Tz-CLIO was removed by washing the pellet twice with 1XPB, before being resuspended in 20 μl of 1XPBS for μNMR measurements. Fluorescent images were taken for visualization of CTCs using EasySep™ enrichment method. CTCs were isolated using the FITC Positive Selection kit (STEMCELL Technologies Inc). Cells were labeled with antibodies against EPCAM, EFGR, Her-2 and Muc and targeted with FITC-conjugated antibodies and dextran-coated magnetic nanoparticles. Labeled CTCs are separated using an EasySep™ magnet. All labeling experiments were performed at room temperature.

Preparation of transcyclooctene (TCO)-modified antibodies

Monoclonal antibodies against targets were used for the primary labeling of antigens (Table S1). To enhance the labeling efficiency, TCO modification of antibodies were performed on secondary IgG antibodies instead of primary antibodies, (i.e. primary antibodies were incubated with cells per manufacturer's protocol). Secondary IgG antibodies were conjugated with (E)-cyclooct-4-enyl 2,5-dioxopyrrolidin-1-yl carbonate (TCO-NHS), as previously reported^{20,25,26,26,27}. TCO-conjugation was performed with 0.5 mg of mouse or rabbit IgG antibody in the presence of 1000 equivalents of TCO-NHS in phosphate-buffered saline (PBS) and 10% dimethylformamide at room temperature for 3 hours. Unreacted TCO-NHS was subsequently washed using 2 ml of Zeba desalting columns (Thermo Fisher, Rockford, IL), and antibody concentrations were determined by absorbance measurement.

Preparation of tetrazine (Tz)-modified nanoparticles

Cross-linked iron oxide (CLIO) nanoparticles were prepared as previously discussed ²⁶. Briefly, Tz-modified nanoparticles were synthesized by reacting NH2-MNPs with 500 equivalents of 2,5-dioxopyrrolidin-1-yl 5-(4-(1,2,4,5-tetrazin-3-yl)benzylamino)-5-oxopentanoate (Tz-NHS). The reaction was carried out in PBS/5% dimethylformamide at room temperature for 3 hours, as previously described ^20,26,27. Excess Tz-NHS was removed by gel filtration using Sephadex G-50 (GE Healthcare, Pittsburgh, PA).

μNMR measurements

The μNMR technology and magnetic nanoparticle labeling have been extensively tested in cell lines ^17,20,26, mouse models of cancer ¹⁷ and in clinical samples ^12,20,28,29. Moreover, the performance μNMR for cancer detection was compared to gold standard method (CellSearch) ²⁰. In the present study, sample labeling and μNMR measurements (Figure 1) were performed as previously optimized ^20,26. The transverse relaxation rates (R2) were determined in an effective sample volume of 1μL. μNMR measurements and analysis of cellular expression were performed as previously descried ²⁰. Briefly, the following Carr-Purcell-Meiboom-Gill pulse sequences were used: echo time, 4 ms; repetition time, 6 s; number of 180° pulses per scan, 500; number of scans, 8. R2 values from samples were subtracted from those of PBS buffer to calculate ΔR2. The ΔR2 of samples were normalized to ΔR2 obtained from negative control to account for background signal. Negative controls for each sample were prepared similarly to the test samples without inclusion of the antibody incubation step. For each sample, measurement was done three times and the average value was obtained for analysis.

Results

FNA yield sufficient numbers of cells for analysis

Table 1 summarizes the enrolled patient cohort consisting of consecutive patients referred for CT guided biopsy of lung lesions or suspected lung cancer metastases in the adrenal glands or liver, and who consented to participate in the study. Table 2 summarizes the different cohorts and final pathologic diagnosis. Just before each procedure and on the same day peripheral blood was drawn for CTC analysis.

Following placement of the co-axial needle by CT guidance, 3-5 FNA passes were placed in saline from each lesion (Figure 1). Samples were incubated with TCO-modified antibodies for key marker proteins (Table 3) revealed by bioorthogonal reactions using magnetic nanoparticles and μNMR analysis. Figure 2 summarizes the yields of total and malignant cell number for each specimen. In two of the 35 patients, the biopsy procedure was aborted due to medical reasons, leaving 33 patients for analysis. Overall, the attained specimens were sufficient for μNMR analysis in 30 of the 33 patients (91 %) yielding an average of 1900 cancer cells per patient. In 3 patients, insufficient numbers of FNA cells were noted after the procedure.

Figure 2. Obtaining cells from fine needle aspirates (FNA).

Left: Representative example showing CT guided placement of a 20G coaxial needle within a nodule in the middle lobe of the right lung. Right, top: Total number of cells harvested via FNA. Right bottom: Number of cancer cells (quad positive) harvested by FNA sampling.

DMR analyses are accurate in establishing subtypes

To establish the diagnosis of malignancy, we used a cocktail of 4 antibodies (EGFR, MUC, Her2, EpCAM, collectively known as the “Quad marker”). Indeed, this signature is more sensitive than EpCAM alone for diagnosing epithelial malignancies ^20,25,12. Differentiating between different subtypes of primary lung cancer and lung metastases was based on the expression levels of key and validated proteins (TTF-1, napsin A, synaptophysin, chromogranin, p40; Table 3) ^21-24.

Figure 3 summarizes the FNA profiles across the 30 patients with sufficient FNA material available for analyses. The quad marker was positive (i.e. μNMR⁺ diagnosis = malignant) in all 27 cancer patients, including in lung cancer patients with stage 1 disease with a primary tumor as small as 0.7 cm. μNMR finding was negative in 3 patients with benign nodules and in 12 peripheral blood samples obtained from healthy individuals. TTF-1 and/or napsin A were positive in all adenocarcinomas, while p40 was positive in the four squamous call cancers. Altogether, FNA analysis by μNMR showed excellent concordance with conventional histopathology, both for establishing overall diagnoses and differential diagnoses (Figure 3). Table 4 summarizes the sensitivity, specificity and predictive values of μ-NMR analyses.

Figure 3. μNMR analysis of FNA samples.

Left: Raw μNMR data for the different biomarkers tested. Analysis of the different biomarkers into histologic types shows good agreement between μNMR and conventional histology (from surgical samples or core biopsies). All malignant and benign masses were correctly classified. Right: Histology images of lung cancer subtypes. Top: Lung adenocarcinoma exhibiting gland-forming malignant cells. Middle: Moderately differentiated squamous cell carcinoma demonstrating a nest of polygonal cells with eosinophilic cytoplasm and intercellular bridges. Bottom: Small cell carcinoma characterized by nests of malignant cells with a high nuclear cytoplasmic ratio, prominent apoptosis and mitotic activity. All images were obtained at 400X magnification.

Table 4. Diagnostic accuracy of μNMR.

Data are shown separately for FNA analysis of primary lung lesions and for CTC analysis. Note that calculations are based on all samples, including those with insufficient material for DMR processing (thus the overall reported sensitivity is reduced).

	FNA (n = 33)	CTC (n = 35)
Sensitivity	22/24 (91.6%)	20/24 (83.3%)
Specificity	8/8 (100%)	11/11 (100%)
PPV	22/22 (100%	20/20 (100%)
NPV	8/10 (80%)	11/15 (73.3%)
Total samples for analysis	33	35
Not enough cells	3
True neg	8	11
False neg	2	4
True pos (“lung” cancer cases)	22	20
False pos	0	0

CTC and FNA analyses are concordant

We next explored whether CTC analyses could be performed in the same patient cohort described above. Peripheral blood analyses from 35 patients demonstrated, on average, 38 CTC per patient (7ml blood) with proven malignancy (Figure 4). There were 11 true negative (i.e. diagnosis of negative for lung malignancy based on gold standard and μNMR- finding for CTC), four false negative (i.e. diagnosis of positive for lung malignancy based on gold standard but μNMR- finding for CTC) and zero false positive samples (Table 4). The overall sensitivity for detecting lung cancer in blood was 83.3% with the specificity of 100% (Table 4). We noted good correlation between lung cancer biomarkers across CTC and FNA (r²=0.83). Using patient by patient analyses, we then asked whether peripheral blood analysis could have obviated a lung biopsy. In 70% of the patient population, CTC were positive while 20% displayed FNA positive / CTC negative profiles, and 10% were negative for both CTC and FNA (Figure 4).

Figure 4. Analysis of circulating cancer cells.

Quad-NMR analysis of peripheral blood samples showed variable numbers of CTC. In 71% of all cases with positive FNA, there were also CTC present. CTC were negative in 19% of positive FNA samples. There was a correlation between circulating cancer cells and number of cancer cells harvested by FNA analysis (r² = 0.83).

Differential diagnosis of CTC samples

Since the same differential diagnostic markers can be applied to CTC, we next determined whether such analyses could establish the primary tumor type based on blood sample alone (Figure 5). Our data indicate that CTC analyses correctly identified histologic subtypes in 18 of 20 patients (90%) with available matched samples.

Figure 5. Differential diagnosis of CTC.

Left: raw μNMR data for the different biomarkers tested in CTC. Analysis of the different biomarkers into histologic types shows reasonable agreement in over 85% of cases. Right: fluorescent images of quad marker labeling of CTC (top) and CTC and DAPI staining (bottom). Fluorescent images were obtained at 40X magnification.

Discussion

Advances in nanotechnology, microfabrication and chemistries have catalyzed miniaturized diagnostic platforms to potentially analyze scant clinical specimens. Success here would potentially reduce the morbidity associated with more invasive procedures, improve cost effectiveness through efficient use of reagents and time, allow for serial testing to tailor treatments, and promote point of care testing. There are two generic approaches: I) methods to isolate and purify malignant cells for subsequent analysis ^7,30 and II) methods that allow direct measurements without the need for purification^12,19. Each analytical approach has its unique advantages and applications in the clinical workflow. For any method to succeed it must be accurate, reproducible, fast and cost-effective. In particular, methods that allow real-time feedback would be highly advantageous to guide decisions during interventional procedures.

Here we leveraged the benefits of μNMR to analyze clinically accessible samples. The approach has a number of advantages including fast turn around, high sensitivity and readouts without need for lengthy purification. The technology has previously been optimized for FNA analyses of abdominal malignancies¹², CTC²⁰ and exosome analyses¹⁸. The current study capitalized on some of the previously established malignancy markers (notably the quad signature of EpCAM, EGFR, Muc1 and Her2)¹² but also added clinically relevant differential diagnostic markers of lung cancers (napsin A, p40, TTF-1, synaptophysin and chromgranin; Table 3) ^21-24.

Our results show that point-of-care, operator-independent μNMR protein analysis of patient-derived FNA and peripheral blood samples is highly accurate in identifying the presence and establishing the type of lung malignancy. Achieving this in peripheral blood and/or limited biopsies can lead to a shift in the current standard of practice, with peripheral blood analysis preceding any invasive procedure, and FNA acquisition replacing the need of a more invasive core biopsy.

While the current results are very encouraging, we did not explore all potential aspects in this feasibility study. For example did not test for specific mutations to inform eligibility for molecular drug treatments (e.g. EGFR or Alk translocations) but such assays could be easily established through use of TCO labeled antibodies ²⁷ (or other affinity ligands) with sufficient specificity. Second, the μNMR method used here is based on read-outs of bulk samples (requiring ∼ 50 cells depending on target and antibodies) and is not a single cell profiling technology. For the latter, alternative approaches such as μHall sensors ¹⁹ or others could be used³¹. While not tested here, the μNMR method is capable of broader genetic analyses as recently demonstrated for bacterial profiling^32,33. It is conceivable to integrate such testing with antibody approaches on the same chip platform. Finally, we did not test the method with bronchoscopy specimens although we would expect similar results given the typical cell numbers that are harvested ³⁴. Similarly, we expect that the method could be used for analyses of other clinically relevant samples such as pleural fluid ³⁵.

The current data justifies future lung cancer focused studies to further develop the technology and inform preclinical and early phase drug testing. We anticipate the use of additional markers to accompany diagnostic panels by leveraging the advanced multiplexing capabilities of digital NMR technology. Attaining lung cancer subtype information along with any actionable molecular alterations rapidly (∼1 hour) and at the point-of-care can help streamline the clinical research process. Same visit readouts can help initiate treatment strategies (e.g. selecting targeted therapies if adenocarcinoma versus chemotherapy if squamous cell) and promptly identify resistance markers to alter treatment course. Coupled with CTC as surrogates to invasive core biopsies, as demonstrated here, there is also strong potential for cost effective and safety advantages. Point-of-care μNMR lung cancer cell analyses could thus function as a peripheral window into each individual's underlying biology and help meet the goals of precision medicine in this prevalent and often lethal disease.

Supplementary Material

01

Figure S1: Algorithm used in identification and sub classification of lung cancer samples using μNMR technology.

02

Figure S2: ROC analysis on Quad marker. Quad marker levels on CTC in the peripheral blood of cancer patients were compared with those of 15 healthy individuals, as assessed by quad-μNMR (A). Receiver operating characteristic (ROC) analysis was performed for quad μNMR. An Az (area under the ROC curve) of 0.5 was used to indicate no differences between two groups, whereas an Az of 0.84 was used to indicate a perfect separation between the groups. The optimal cut-off value for identifying malignant status was then defined as 1.66, the point on the ROC curve with the minimum distance between the 0% false-negative and the 100% true-positive.

03

Figure S3: ROC analysis on TTF1 marker. TTF1 levels in the peripheral blood of lung cancer patients were compared with a control group that includes benign nodules and non-lung cancer cases (i.e. expected to be negative for TTF1) (A). Receiver operating characteristic (ROC) analysis was performed for TTF1. An Az (area under the ROC curve) of 0.5 was used to indicate no differences between two groups, whereas an Az of 0.79 was used to indicate a perfect separation between the groups. The optimal cut-off value for identifying malignant status was then defined as 1.62, the point on the ROC curve with the minimum distance between the 0% false-negative and the 100% true-positive.

04

Table S1: A list of antibodies and relevant details utilized in the study.

We tested our recently developed point-of-care, operator-independent μNMR device to perform protein analysis of tumor cells for diagnosing pulmonary malignancy. In this clinical study, analysis of both fine needle aspirate (FNA) of primary tumors and blood circulating tumor cells (CTC) was shown to be highly accurate in identifying the presence and establishing the type of lung malignancy.